Driven geospatial artificial intelligence modeling for climate change impact assessment: a global perspective

Driven geospatial artificial intelligence modeling for climate change impact assessment: a global perspective | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Systematic Review Driven geospatial artificial intelligence modeling for climate change impact assessment: a global perspective Francis Quayson This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-8913236/v1 This work is licensed under a CC BY 4.0 License Status: Posted Version 1 posted You are reading this latest preprint version Abstract The challenge of anthropogenic climate change requires a paradigm shift in analysis, moving beyond traditional modeling constraints toward the application of computational intelligence. Geospatial Artificial Intelligence (GeoAI), an emergent interdisciplinary domain synthesizing spatial science, artificial intelligence (AI), and statistical learning, can reshape the epistemology of climate change impact assessment. Although previous reviews have examined discrete elements of geospatial technologies or AI in environmental science, a comprehensive, critical synthesis that captures the methodological and application-based convergence of GeoAI and climate change is rare. This study analyzes GeoAI applications in climate change science, employing a dual scientometric and systematic review methodology. A scientometric analysis of 152 core publications, retrieved from Scopus, Web of Science, IEEE Xplore, and Google Scholar, was conducted to quantitatively elucidate publication trends, thematic networks, and the global collaborative landscape. This mapping reveals a rapidly accelerating, interdisciplinary field characterized by robust clusters around deep learning architectures, multi-modal remote sensing, and specific climate impact pathways. The review deconstructs the GeoAI technological stack from convolutional and recurrent neural networks to transformative transformer-based models and hybrid physics-AI frameworks. It also evaluates their roles in climate variable prediction, extreme event attribution, ecosystem vulnerability diagnostics, and socio-economic impact modeling. The review identifies profound epistemological and practical challenges, including data-centric limitations (heterogeneity, paucity), computational intractability, the “black box” conundrum, model transferability failures, and critical ethical imperatives. This review outlines a strategic prospective research agenda, suggesting that the evolution of scalable, interpretable, and context-aware GeoAI systems is crucial for constructing a resilient and equitable planetary future. Planetary Science Environmental Engineering GeoAI Deep Learning Remote Sensing Scientometric Analysis Systematic Review Explainable AI Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Figure 6 Figure 7 Figure 8 1. Introduction Anthropogenic climate change constitutes the defining socio-ecological challenge of the Anthropocene, a complex, polycausal phenomenon with cascading, non-linear implications for Earth system stability, biodiversity integrity, human security, and global economic equity (Ripple et al., 2020; Steffen et al., 2018). The Intergovernmental Panel on Climate Change (IPCC) has pointed out an accelerating trajectory of global warming and duration of compound extreme weather events, thereby pushing natural and human systems toward adaptation limits (IPCC, 2023; Shukla et al., 2022). Understanding, projecting, and mitigating the multifaceted impacts of these perturbations requires processing the vast, heterogeneous, multi-scale, and multi-modal datasets that characterize the modern Earth observation paradigm (Irrgang et al., 2021a; Reichstein et al., 2019a). While foundational to historical understanding, traditional mechanistic climate and impact models typically face computational, parametric, and structural limitations in assimilating the petabyte-scale data deluge from satellite remote sensing, unmanned aerial vehicles (UAVs), and pervasive in-situ sensor networks (Karpatne et al., 2017; Schneider et al., 2017). Therefore, Geospatial Artificial Intelligence (GeoAI) has emerged as a transformative scientific frontier. GeoAI is the interdisciplinary synthesis of geospatial theories, spatiotemporal data models, and computational methods with advanced AI paradigms, particularly from the statistical machine learning (ML) and deep learning (DL) domains (Janowicz et al., 2020; W. Li, 2020a). The proliferation of high-resolution satellite imagery (e.g., Sentinel, Landsat, and Planet), hyperspectral and Synthetic Aperture Radar (SAR) data, and the Internet of Things (IoT) has created a “geospatial big data” environment of unprecedented volume, velocity, and variety, suited for data-driven AI methodologies (Ma et al., 2019a; W. Yang et al., 2017). GeoAI techniques effectively extract non-obvious patterns, making high-dimensional predictions in ill-posed problems and revealing latent insights from these complex spatiotemporal datasets. Thus, they can radically enhance the granularity, accuracy, and temporal frequency of climate change impact assessments. (H. Liu et al., 2024; Y. Wang et al., 2022) GeoAI applications in climate science are diverse and expanding rapidly, representing a shift from correlation-based analysis to a more mechanistic, although data-driven understanding. For instance, convolutional neural networks (CNNs), particularly encoder–decoder architectures such as U-Net and DeepLab, are being deployed to delineate glacier boundaries, monitor ice sheet dynamics, and classify sea ice from multi-spectral and SAR satellite imagery with sub-pixel accuracy (Gavahi et al., 2023; R. Wang et al., 2024; L. Zhang et al., 2016). Similarly, recurrent neural networks (RNNs), Long Short-Term Memory (LSTM), and Gated Recurrent Unit (GRU) networks have demonstrated superiority in forecasting essential climate variables by learning long-term dependencies from historical time-series data from ground stations and gridded reanalysis products (Fang et al., 2017; Le et al., 2019; Shi et al., 2015). Additionally, ensemble ML methods, such as Random Forests and Gradient Boosting Machines (XGBoost and LightGBM), are used to model species distribution shifts in response to climate change. They are also used to assess the multi-hazard vulnerability of coastal zones and predict wildfire risk in a warming world (Elith et al., 2008; Jain et al., 2020; Krizhevsky et al., 2017; Mei et al., 2022). Despite these advancements, the field lacks a comprehensive, critical synthesis that maps its intellectual structure, evaluates its methodological trajectories and epistemological assumptions, and consolidates its persistent challenges into a coherent, prospective research agenda. Existing reviews have either focused on AI in environmental science without a geospatial focus (Rolnick et al., 2023) or on specific geospatial technologies (e.g., remote sensing) without a concentrated AI focus (Wulder et al., 2018) or have been limited in their temporal scope or methodological depth (Table 1). A thorough, state-of-the-art analysis that targets the epistemological and practical relationship between GeoAI and climate change impact assessment is thus essential to consolidate knowledge, identify methodological synergies and limitations, and guide future research investments and policies. This study fills this critical scholarly gap by analyzing GeoAI applications for climate change impact assessment from a global perspective. The objectives are as follows: 1. To identify the relevant corpus of literature in the GeoAI-climate change domain and conduct a scientometric analysis to quantitatively elucidate annual publication trends, thematic network structures, and global research contributions. 2. To systematically review and synthesize the predominant GeoAI technologies, methodological approaches, and application domains in climate change impact assessment, discussing their associated capabilities, limitations, and implementation contexts. 3. To identify persistent theoretical and applied research gaps and challenges within the literature and provide a strategic, prospective agenda for future scientific inquiry and technological development. Table 1 Overview of previous review articles on the AI–Geospatial–Climate nexus Authors (Year) Review Focus & Title Methodological Approach Timespan Key Findings / Scope Identified Gaps / Limitations Rolnick et al. (2022) Broad applications of machine learning (ML) for climate change. Topical synthesis and taxonomy development. 2015–2021 Cataloged ML opportunities across climate domains: mitigation, adaptation, modeling, and societal impacts. Lacks a dedicated geospatial theory and data perspective; scope is overly broad, missing depth in GeoAI methodologies. Zhu et al. (2017) Deep learning (DL) in remote sensing. Technical review and resource listing. 2012–2017 Comprehensive tutorial on DL architectures (CNNs, RNNs) for remote sensing tasks such as classification and detection. Focuses on computer vision techniques; does not systematically address climate change applications or impact assessment. Wang et al. (2022a) GeoAI for hydroinformatics. Systematic literature review. 2010–2021 Reviewed AI applications in water resources, including flood forecasting and water quality monitoring. Domain-specific (hydrology); does not cover the full spectrum of climate impacts (e.g., ecosystems, urban areas, cryosphere). Li (2020) Foundations and concepts of GeoAI. Conceptual analysis and vision paper. 2015–2020 Defined the GeoAI paradigm, discussing spatial statistics, ML integration, and future trajectories. Theoretical and foundational; lacks a systematic review of empirical applications, especially in climate science. Reichstein et al. (2019) DL for process understanding in Earth system science. Perspective and selective review. 2016–2019 Advocated for combining DL with physical models for better causality and generalization in Earth science. Selective, non-systematic review; focuses on the philosophy of integration rather than a comprehensive application survey. Wulder et al. (2018) Land cover mapping and change detection using remote sensing. Critical review of technological trends. 2000–2018 Described the evolution from manual digitization to automated methods for land cover monitoring. Focused on a specific application (land cover); does not review the broader AI toolkit or other climate impact domains. 2. Research methodology This study adopts a convergent parallel mixed-methods approach (Creswell and Poth, 2018), synergizing quantitative scientometric analysis with qualitative systematic review to holistically examine the GeoAI-climate change literature. The quantitative component involves a systematic, multi-database retrieval and scientometric mapping of relevant studies, while the qualitative component entails a detailed, critical, and thematic assessment of the conceptual and methodological content of the retrieved literature. This mixed-methods approach mitigates the inherent limitations and biases of either approach used in isolation and ensures a comprehensive, evidence-based understanding of the field’s current state, dynamics, and intellectual structure. The literature retrieval procedure is shown in Fig. 1, and the overall analytical framework of the study is shown in Fig. 2 . 2.1. Retrieval of relevant studies A robust literature retrieval protocol was employed, as shown in Fig. 1, following the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines (Page et al., 2021). To ensure comprehensive coverage and minimize database-specific bias, the primary search was conducted across four major scholarly databases: Scopus, Web of Science Core Collection, IEEE Xplore Digital Library, and Google Scholar. These platforms were selected for their extensive coverage of high-quality, peer-reviewed literature in engineering, environmental science, computer science, and geoinformatics (Martín-Martín et al., 2021; Zhu et al., 2017). Through an iterative process, the search query was designed to capture the core conceptual domains of GeoAI and climate change, using the TITLE-ABS-KEY fields in Scopus and Web of Science and analogous advanced search functions in IEEE Xplore and Google Scholar. The initial multi-database search, conducted for the period up to June 1, 2025, returned 892 records (Scopus: 231, Web of Science: 198, IEEE Xplore: 145, Google Scholar: 318). All records were exported and consolidated using the Zotero reference management software, and duplicates were removed using both automated deduplication and manual inspection, resulting in 485 unique publications. To ensure the review’s conceptual focus, methodological rigor, and contemporaneity, the following inclusion and exclusion criteria were applied in a two-stage screening process: 2.1.1. Inclusion criteria 1. Peer-reviewed journal articles or full-length conference proceedings. 2. Primary focus on the application, development, or critical evaluation of AI/ML/DL methods to geospatial data for climate change-related analysis, impact assessment, prediction, or adaptation/mitigation planning. 3. Publications dated between January 2015 and June 2025 were included. 4. Written in the English language. 2.1.2. Exclusion criteria 1. Articles where AI/ML was mentioned peripherally without substantive methodological application or empirical validation. 2. Studies using only traditional statistical methods (e.g., linear regression and GIS overlay) on geospatial data without an AI/ML component. 3. Editorials, opinion pieces, book reviews, books, book chapters, and dissertations. 4. Studies where the geospatial component was incidental or not integral to the AI methodology. After screening the title and abstract of the 485 unique publications against these criteria, 333 records were excluded. The remaining 192 articles underwent a full-text assessment of eligibility. This intensive screening resulted in the exclusion of 40 articles, primarily owing to a lack of substantive AI-geospatial integration or a weak climate change focus. This process yielded a final corpus of 152 articles for quantitative scientometric analysis. The Python programming language (v3.11), with libraries such as Pandas, NumPy, and “pybliometrics,” was used for managing, cleaning, and performing initial analyses on the bibliographic data. Subsequently, to ensure a thorough and comprehensive coverage for the qualitative systematic review, a forward and backward snowballing technique (Wohlin, 2014) was applied to the reference lists and citing articles of these 152 publications. This iterative process identified an additional 4 highly relevant articles that were missed by the initial database search, resulting in a final expanded corpus of 77 articles for the in-depth systematic review. This expanded set ensures the research field is fully covered while maintaining a manageable scope for qualitative content analysis. 2.2. Scientometric analysis Scientometric analysis is a quantitative methodology for mapping the structural and dynamic evolution of scientific fields by measuring and analyzing patterns in scholarly publications (Ellegaard & Wallin, 2015; Zupic & Čater, 2015). In this review, the bibliographic data of the 152 publications were exported from the databases in standardized RIS and BibTeX formats and harmonized. The primary scientometric analysis was conducted using the Bibliometrix R package (Aria & Cuccurullo, 2017), with custom scripts in Python (v3.11) using the Matplotlib, Seaborn, NetworkX, and “scikit-learn” libraries for customized visualizations, network analysis, and centrality calculations. The scientometric investigation focused on four key analytical dimensions: 1. Annual Publication Trend: To visualize the growth trajectory, identify key inflection points, and assess the maturation and velocity of the research field from 2015 to June 2025. 2. Keyword Co-occurrence Analysis: To identify the main research themes, conceptual underpinnings, and their intellectual interrelationships. A minimum keyword frequency threshold of 5 was set to filter out noise and focus on core concepts. Natural Language Processing (NLP) techniques in Python were used for keyword lemmatization and synonym merging (e.g., “cnn” and “convolutional neural network”). 3. Analysis of Publication Outlets: To identify the core journals and conferences that serve as the primary dissemination channels for knowledge in this domain and to assess their relative impact and interconnectedness. 4. Mapping of Contributing Countries and Institutions: To reveal global and institutional research leadership, collaboration networks, and the geographical distribution of scientific production using author affiliation data. 2.3. Systematic analysis The systematic analysis involved a qualitative, in-depth, and critical examination of the full corpus of 77 articles. The objective was to synthesize empirical findings, evaluate methodological rigor and innovation, and identify convergent themes, challenges, and knowledge gaps. The analysis followed a structured, iterative coding framework developed through inductive (data-driven) and deductive (theory-driven) approaches, consistent with established qualitative content analysis methodologies (Elo & Kyngäs, 2008; Hsieh & Shannon, 2005). Using Python for data management and the qualitative data analysis software NVivo for coding, the articles were categorized and analyzed across five interconnected analytical domains: 1. An Overview of GeoAI Systems and Architectures: This domain involves classifying and critiquing the dominant AI/ML paradigms (e.g., supervised, unsupervised, and self-supervised) and specific architectural families (e.g., CNNs, RNNs, GANs, and Transformers) used in geospatial climate studies. 2. The Integration of GeoAI in Climate Change Impact Assessment: This domain deals with examining the specific mechanisms, data fusion strategies, and methodological pipelines through which GeoAI models are applied to distinct climate-related problems, from detection to attribution and projection. 3. Technological and Methodological Attributes: This domain delves into analyzing data sources (e.g., MODIS, Sentinel, and CMIP6), feature engineering practices, model validation and evaluation metrics (e.g., RMSE, IoU, F1-score), hyperparameter tuning strategies, and the prevalence of specific software tools and libraries (e.g., TensorFlow, PyTorch, Scikit-learn, and GDAL). 4. GeoAI-Climate Application Domains: This domain thematically categorizes studies into major application areas such as cryosphere and hydrosphere monitoring, terrestrial and marine ecosystem dynamics, agricultural security and food systems, urban climate resilience, and human health impacts. 5. Challenges, Limitations, and Ethical Considerations: This domain synthesizes reported and inferred impediments related to data quality and availability, computational complexity, model interpretability and explainability, generalizability and transferability, reproducibility, and profound ethical concerns including algorithmic bias and socio-environmental justice. This structured, multi-faceted content analysis helped identify convergent themes, methodological trends, technological bottlenecks, and critical research gaps, which form the evidentiary basis for the recommendations presented in Section 5. 3. Result of the scientometric analysis 3.1. Annual publication trend The annual publication trend (Fig. 3) indicates the rapid emergence and subsequent advancement of the GeoAI-climate change field as a distinct and vibrant research domain. The period from 2015 to 2017 can be considered a nascent phase, with a modest output of 2 to 4 publications per year. This phase coincided with the initial exploration of basic ML models (e.g., Random Forests and Support Vector Machines) on geospatial data for climate applications, and the early adoption of DL in computer vision began to influence remote sensing analysis (Ball et al., 2017; Längkvist et al., 2016). A notable inflection point occurred around 2018 to 2019, with annual publications jumping to 8 and then 12. This surge aligns with the widespread maturation and accessibility of DL frameworks (e.g., TensorFlow and PyTorch). Additionally, it shows increased availability of curated geospatial datasets (e.g., Earth Engine catalog) and a growing recognition within the climate science community of the value of data-driven approaches for addressing problems intractable to purely physical models (Reichstein et al., 2019b). A steady upward trajectory is observed from 2020 onwards, peaking at 22 publications in 2024. Projections based on the first half of 2025 suggest this trend will continue, with an estimated 25+ publications for the entire year. This exponential growth pattern (R² > 0.95 for a polynomial fit) indicates a rapidly consolidating research area attracting increasing scholarly attention. The growth is driven by a confluence of pressing global climate concerns, continuous advancements in AI algorithms and computing hardware (e.g., GPUs and TPUs), and the unprecedented flow of Earth observation data from public and private satellite constellations. 3.2. Keyword co-occurrence The keyword co-occurrence network, generated using a minimum frequency of 5, yielded 72 keywords forming 4 distinct and semantically coherent clusters (Fig. 4). The network’s overall density and the strength of linkages indicate a well-connected, interdisciplinary field. Cluster 1 (Red - AI/ML Methodological Core): This cluster forms the methodological backbone of the field, centered around core AI concepts such as “deep learning,” “machine learning,” “artificial intelligence,” “convolutional neural networks (CNN),” and “long short-term memory (LSTM).” It also includes related terms such as “feature extraction” and “predictive modelling,” highlighting the focus on automated pattern recognition and forecasting. Cluster 2 (Green - Remote Sensing & Geospatial Data): This cluster encompasses the critical data sources and platforms that enable GeoAI, including “remote sensing,” “satellite imagery,” “Landsat,” “Sentinel-2,” “Synthetic Aperture Radar (SAR),” “Earth observation,” and “big data.” The strong linkages between this cluster and the red AI/ML cluster demonstrate the tight, symbiotic coupling of methodological innovation with data availability and characteristics. Cluster 3 (Blue - Specific Climate Impacts & Vulnerabilities): This cluster focuses on the application domains and specific climate change impacts. Key terms include “sea level rise,” “land use and land cover change (LULC),” “drought monitoring,” “wildfire risk,” “biodiversity,” “agriculture,” and “climate adaptation.” This cluster connects the methods (Red) and data (Green) to tangible environmental and societal challenges. Cluster 4 (Yellow - Spatial Analysis & Modeling Context): This cluster contains foundational geospatial concepts and broader modeling contexts, such as “geographic information system (GIS),” “spatial analysis,” “climate models,” “risk assessment,” “vulnerability,” and “sustainability.” It situates the GeoAI work within the larger frameworks of geospatial science and climate risk management. The network’s core is primarily occupied by “geospatial artificial intelligence,” “climate change,” and “machine learning,” confirming their central role as the field’s defining concepts. The high betweenness centrality of “climate change” indicates it connects the methodological and application-oriented clusters. Table 2, generated using Python’s NetworkX library, ranks the top 20 keywords by degree centrality, a measure of a node’s connectedness within the network. “Machine Learning,” “Remote Sensing,” and “Deep Learning” exhibit the highest degree centrality, indicating that they are the most influential and interconnected concepts that facilitate knowledge flow across the entire domain. Table 2 Top 20 keywords by degree centrality in the GeoAI-Climate Change domain Rank Keyword Occurrence Degree Centrality Betweenness Centrality 1 Machine Learning 58 52 0.118 2 Remote Sensing 55 50 0.132 3 Climate Change 53 49 0.211 4 Deep Learning 49 46 0.095 5 Convolutional Neural Network (CNN) 31 35 0.054 6 Geographic Information System (GIS) 28 33 0.048 7 Land Use and Land Cover Change (LULC) 25 30 0.041 8 Satellite Imagery 24 29 0.037 9 Long Short-Term Memory (LSTM) 22 27 0.033 10 Drought Monitoring 19 25 0.029 11 Sea Level Rise 18 24 0.031 12 Spatial Analysis 17 23 0.025 13 Big Data 16 22 0.022 14 Risk Assessment 15 21 0.028 15 Sentinel-2 14 20 0.019 16 Biodiversity 13 19 0.020 17 Wildfire 13 18 0.017 18 Climate Adaptation 12 17 0.023 19 Synthetic Aperture Radar (SAR) 11 16 0.015 20 Explainable AI (XAI) 10 15 0.026 3.3. Publication outlets The analysis of publication outlets identified the primary channels for scholarly communication and knowledge dissemination in this field. The threshold was set to sources with at least 2 publications from the core corpus, resulting in 21 key outlets. A network map (Fig. 5) shows the co-citation relationships between these journals, revealing a dense web of intellectual exchange. Regarding publication volume, the most prominent venues are “Remote Sensing of Environment,” “ISPRS Journal of Photogrammetry and Remote Sensing,” “International Journal of Applied Earth Observation and Geoinformation,” and “IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing.” This concentration in premier remote sensing and geoinformation journals indicates that the field is currently anchored within the geospatial science and engineering communities. However, a significant and growing presence is noted in high-impact, broad-scope journals such as “Nature Climate Change,” “Science Advances,” and “Proceedings of the National Academy of Sciences,” suggesting the increasing recognition of GeoAI’s transformative potential by the wider scientific community. Furthermore, computer science conferences such as “NeurIPS,” “ICML,” and “CVPR” are emerging as crucial venues for presenting foundational algorithmic advances with geospatial applications. Table 3 ranks the top 10 outlets based on the average normalized citation count, a metric that corrects for the age and size of the journal, providing a measure of impact per paper. “Nature Climate Change” and “Remote Sensing of Environment” lead in this regard, highlighting the high impact of the research published therein. Table 3 Top 10 publication outlets in the GeoAI-Climate Change domain by average normalized citations Publication Outlet Total Link Strength (TLS) Documents Total Citations Normalized Citations Avg. Normalized Citations Nature Climate Change 4 3 185 28.45 9.48 Remote Sensing of Environment 22 9 412 25.18 2.80 Science Advances 3 2 97 15.11 7.56 ISPRS Journal of Photogrammetry and Remote Sensing 18 7 201 12.33 1.76 International Journal of Applied Earth Observation and Geoinformation 15 6 134 8.91 1.49 IEEE Transactions on Geoscience and Remote Sensing 12 5 118 7.25 1.45 Global Change Biology 5 2 68 6.89 3.45 Environmental Research Letters 8 3 75 5.12 1.71 IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing 10 4 81 4.95 1.24 Climatic Change 6 2 52 4.87 2.44 3.4. Top contributing countries Mapping the geographical distribution of research output (Fig. 6) reveals a clear but evolving global leadership structure. The minimum thresholds were set at 2 documents and 10 citations per country. Among the 45 countries represented in the corpus, 18 met these criteria. Based on corresponding author and co-author affiliations, the analysis shows that the United States and China are the dominant contributors, both in terms of raw publication volume and Total Link Strength (TLS), a measure of international co-authorship. This indicates high domestic productivity and their extensive and central roles in global collaborative networks. The United Kingdom, Germany, Switzerland, and the Netherlands also demonstrate strong and influential contributions, often acting as bridges in the collaboration network. Canada and Australia are other significant nodes of the Anglosphere. Notably, emerging economies such as India and Brazil appear as meaningful contributors, reflecting a growing but uneven global engagement with GeoAI for addressing regional and global climate challenges. Table 4 presents a detailed ranking, highlighting the USA’s central role in the collaborative network and the high impact of Swiss research, as indicated by its high average normalized citation count. Table 4 Top contributing countries in GeoAI-Climate change research Country TLS Documents Total Citations Normalized Citations Avg. Normalized Citations United States 105 41 1456 58.91 1.44 China 88 38 987 42.15 1.11 United Kingdom 45 18 445 19.87 1.10 Germany 38 15 388 16.54 1.10 Canada 32 12 321 14.12 1.18 Australia 28 11 278 11.95 1.09 Switzerland 25 8 305 13.88 1.74 Netherlands 22 9 234 10.01 1.11 France 20 7 198 8.45 1.21 India 18 6 123 5.28 0.88 Brazil 15 5 115 4.91 0.98 Italy 14 5 104 4.44 0.89 Spain 12 4 97 4.15 1.04 Japan 10 3 85 3.63 1.21 South Korea 9 3 78 3.33 1.11 Sweden 8 3 71 3.03 1.01 Norway 7 2 65 2.78 1.39 Austria 6 2 58 2.48 1.24 4. Result of the systematic analysis 4.1. GeoAI systems and architectures in climate science GeoAI systems involve diverse and evolving algorithms and computational architectures designed to process, learn from, and reason about spatiotemporal data. Based on the systematic review of 77 articles, these can be categorized into several dominant but often overlapping paradigms with distinct strengths and application contexts in climate science. 4.1.1. CNNs and spatial feature learning CNNs have become the foundational models for analyzing spatially explicit data, particularly raster-based satellite and aerial imagery (Lecun et al., 2015; Zhu et al., 2017). Through the use of convolutional filters and pooling operations, their intrinsic inductive bias allows them to automatically learn hierarchical spatial features from low-level edges and textures to high-level object representations without manual feature engineering (Krizhevsky et al., 2017). This makes them powerful for tasks such as land use and land cover change (LULC) classification (Y. Liu et al., 2018; E. Zhang et al., 2021; Z. Zhang et al., 2018), glacier and sea ice mapping (Gavahi et al., 2023; R. Wang et al., 2024), urban heat island analysis (Aviles Toledo et al., 2024), and building footprint extraction for exposure modeling (Fayaz et al., 2024; Schneider et al., 2017b). Fully Convolutional Network (FCN) architectures, particularly the U-Net family (Ronneberger et al., 2015), are widely used for semantic segmentation of environmental features, achieving state-of-the-art accuracy in delineating complex boundaries, such as coastlines, burn scars, and agricultural fields (Iglovikov et al., 2018; Längkvist et al., 2016). The application of CNNs has considerably transcended standard RGB imagery to multi-spectral, hyper-spectral, and SAR data. This has enabled robust monitoring of vegetation health (e.g., through NDVI), soil moisture, flood extents, and ocean surface winds under varying atmospheric and climatic conditions (Lacoste et al., 2021; Ma et al., 2019b). 4.1.2. RNNs and temporal dynamics modeling To analyze time-series climate data, RNNs and their more advanced variants, LSTM (Hochreiter & Schmidhuber, 1997) and GRU networks (Cho et al., 2014), are essential. These architectures are designed to handle sequential data by maintaining an internal state or “memory” of previous inputs (Yu et al., 2019). LSTMs particularly address the vanishing/exploding gradient problem of vanilla RNNs, allowing them to capture long-term dependencies in temporal sequences, which is crucial for climate processes (Shi et al., 2015). They are used for forecasting essential climate variables such as precipitation, temperature, evapotranspiration, and streamflow (Fang et al., 2017; Lees et al., 2021). LSTM networks learn from historical time-series data from ground stations, gridded reanalysis products (e.g., ERA5), or satellite-derived data cubes. Consequently, LSTM networks can model non-linear and chaotic dynamics that elude traditional statistical methods (ARIMA) or simplified physical models, providing early warning systems for droughts, floods, and heatwaves (Frame et al., 2022; Xiang et al., 2020). 4.1.3. Transformer-Based models and attention mechanisms Inspired by their revolutionary success in NLP (Vaswani et al., 2023), transformer architectures are now being rapidly adapted for geospatial tasks (Irvin et al., 2020; Reed et al., 2023). Their core innovation is the self-attention mechanism, which allows the model to dynamically weigh the importance of all other elements in a sequence (or image patches) when processing a particular element. Therefore, they capture global context more effectively than CNNs with local receptive fields (Dosovitskiy et al., 2021). Vision Transformers (ViTs) are being explored for image-based tasks such as scene classification and change detection, potentially outperforming CNNs on extremely large datasets (Tuia et al., 2022; R. Wang et al., 2024). Furthermore, transformers are being developed for spatiotemporal forecasting, showing promise in modeling complex, large-scale climate phenomena such as El Niño-Southern Oscillation (ENSO) by capturing long-range teleconnections across the Pacific Ocean (Ham et al., 2019; Kurth et al., 2023). The flexibility of transformers to handle multi-modal data (e.g., imagery, point data, and text) also positions them as a key architecture for future integrated assessment models. 4.1.4. Ensemble, Hybrid, and Physics-Informed Models Ensemble methods, such as Random Forests (Breiman, 2001) and Gradient Boosting Machines (e.g., XGBoost) (Chen and Guestrin, 2016), LightGBM (Ke et al., 2017), are popular and effective for tabular geospatial data that integrates satellite-derived indices, climate reanalysis data, topographic variables, and socio-economic factors (Lacoste et al., 2021; Maxwell et al., 2018). They are robust to overfitting, provide native feature importance scores, and provide strong baseline performance for tasks such as species distribution modeling (Valavi et al., 2022), climate risk vulnerability mapping (Alnajjar et al., 2025), and social vulnerability assessment (Tate et al., 2021). Hybrid models that combine the strengths of different architectures are common. For instance, CNN–LSTM models use CNNs to extract spatial features from each timestep of a satellite image time series and LSTMs to model the temporal evolution of these features. This is highly effective for forecasting crop yield (R. Wang et al., 2024; You et al., 2017), soil erosion (C. Yang et al., 2017), and chlorophyll-a concentration in water bodies (Pahlevan et al., 2020). A major and promising frontier is Physics-Informed Neural Networks (PINNs) (Raissi et al., 2019), which embed physical laws (e.g., conservation of energy and Navier-Stokes equations) directly into the loss function or architecture of the neural network. This constrains the solution space to be physically plausible, which is particularly valuable for forecasting in data-sparse regions (Kashinath et al., 2021; Read et al., 2019). 4.2. GeoAI in climate change impact assessment The application of GeoAI is revolutionizing the methodological approaches and empirical precision with which climate change impacts across the Earth’s systems are assessed. 4.2.1. Cryosphere and hydrosphere monitoring GeoAI is proving critical for tracking and projecting changes in the Earth’s frozen regions (cryosphere) and water cycles (hydrosphere). CNNs, specifically U-Net and DeepLab, automate the mapping of glacier retreat, calving fronts, and supraglacial lakes from multi-spectral (e.g., Landsat 8 and Sentinel-2) and SAR (e.g., Sentinel-1) imagery with sub-pixel precision. This provides essential, high-frequency data for constraining sea-level rise projections (Heidler et al., 2022; L. Zhang et al., 2016). Similarly, AI models are used to classify sea ice types, concentration, and motion from SAR data, improving maritime navigation safety and the parameterization of sea ice in climate models (Khaki et al., 2020; Y. Liu et al., 2018; R. Wang et al., 2024). In hydrology, LSTMs and CNN–LSTMs are deployed for predicting river discharge, forecasting flash floods, and modeling groundwater levels, effectively integrating complex rainfall-runoff processes with real-time satellite precipitation data (e.g., GPM) and soil moisture information (Fang et al., 2017; Frame et al., 2022; Kratzert et al., 2023). 4.2.2. Terrestrial and marine ecosystems Climate-induced shifts in biodiversity, phenology, and ecosystem functioning are being monitored and projected with unprecedented detail using GeoAI. Species Distribution Models (SDMs) have evolved from using traditional logistic regression and MaxEnt to employing ensemble ML methods and CNNs that can incorporate raw remote sensing imagery directly as input. They learn imperceptible habitat features, leading to more accurate and mechanistically informed predictions of habitat suitability under future climate scenarios (Botella et al., 2018; Harris et al., 2022; Valavi et al., 2022). In marine ecology, AI models, including CNNs for image analysis and RNNs for time-series prediction, analyze satellite-derived ocean color data (e.g., from MODIS and VIIRS) to monitor harmful algal blooms, detect coral bleaching events, and track changes in phytoplankton biomass and community structure. These are critical indicators of ocean health and are highly sensitive to ocean warming, acidification, and deoxygenation (Free et al., 2021; Khaki et al., 2020; F. Li et al., 2024). 4.2.3. Agriculture and food security Ensuring global food security under a changing climate is a prime application area for GeoAI. ML models, particularly XGBoost and Random Forests, integrate multi-source data including climate projections (CMIP6), soil maps, and high-resolution satellite imagery (e.g., Sentinel-2). This is to predict crop yields at regional to global scales, identify croplands susceptible to drought and heat stress, and recommend optimal planting dates and irrigation schedules (Jin et al., 2019; R. Wang et al., 2024). For instance, CNNs can detect early signs of crop disease and pest infestation exacerbated by changing weather patterns from UAV and satellite imagery, enabling targeted and timely interventions that reduce yield loss and pesticide overuse (Su et al., 2022; L. Zhang et al., 2016). 4.2.4. Urban environments, infrastructure, and human health Cities are hotspots of climate risk and vulnerability. GeoAI helps assess urban exposure and vulnerability by combining very-high-resolution imagery, LiDAR data, land surface temperature (LST) retrievals, and demographic census information. This is to map heat stress risk, identify neighborhoods with poor green space access, and model flood inundation under different precipitation scenarios (F. Li et al., 2024; Schneider et al., 2017b; Stark, 2018). Furthermore, AI models are used to analyze the cooling effect of urban green infrastructure and the impact of building morphology on energy consumption, informing urban planning and climate adaptation strategies (Bi et al., 2023; W. Li, 2020b; Tehrani et al., 2024). In public health, GeoAI models the climate-driven spatial and temporal dynamics of vector-borne diseases (e.g., malaria and dengue), correlating environmental variables such as temperature, precipitation, and vegetation with disease incidence data to create early warning systems (Lowe et al., 2021; Messina et al., 2019). 4.3. GeoAI-Climate application domains The systematic review allowed for a fine-grained, thematic classification of studies into two overarching domains: monitoring and detection of ongoing impacts and the forecasting and projection of future risks. 4.3.1. Climate impact monitoring and detection This domain focuses on observing, quantifying, and attributing ongoing changes, constituting the majority of the reviewed studies. It leverages GeoAI for automated, large-scale, and high-frequency analysis. Deforestation and Forest Degradation Detection: CNNs and change detection algorithms (e.g., those using Siamese networks) are used to identify illegal logging, forest degradation, and fire scars in near-real-time from dense satellite image time series (e.g., with PlanetScope data). This enables rapid response and enforcement (Bullock et al., 2020; Längkvist et al., 2016; Schwartz et al., 2020). Wildfire Burn Scar Mapping and Severity Assessment: AI models, particularly semantic segmentation networks, rapidly and accurately delineate the spatial extent and severity of wildfires from post-fire satellite imagery, providing critical data for carbon emission estimates, ecosystem recovery monitoring, and disaster response (Andela et al., 2017; Ban et al., 2020). Coastal Erosion, Inundation, and Wetland Loss: ML models predict shoreline changes and map saltwater intrusion and coastal flood risk from sea-level rise and storm surges, often using data fusion techniques that combine optical satellite imagery, SAR, LiDAR, and topographic data (Mentaschi et al., 2018; Vos et al., 2022). 4.3.2. Climate impact forecasting and projection This domain uses GeoAI to predict future states and probabilities, which is more challenging but crucial for proactive adaptation planning and policy formulation. Extreme Weather Prediction and Attribution: DL models, including transformers and graph neural networks (GNNs), are being developed to improve the sub-seasonal to seasonal forecasting of extreme events such as tropical cyclones, atmospheric rivers, and heatwaves. They sometimes outperform traditional numerical weather prediction models in specific contexts (Bi et al., 2023; Kurth et al., 2023; Rasp et al., 2018). Long-term Climate Projection Downscaling: AI methods, particularly deep neural networks and generative adversarial networks (GANs), are used to “downscale” coarse-resolution outputs from Global Climate Models (GCMs) to higher, more decision-relevant resolutions (e.g., 1 km). They provide localized projections of temperature, precipitation, and other variables for regional impact assessments (Harris et al., 2022; Nowack et al., 2020; Vandal et al., 2017). Ecosystem Service Projection and Tipping Points: Ensemble models and process-guided AI project future changes in carbon sequestration, water purification, and other ecosystem services under different climate and land-use scenarios (Y. Liu et al., 2018; Runting et al., 2020; Tehrani et al., 2024). Furthermore, AI is being explored to identify early-warning signals of potential climate tipping points in systems such as the Amazon rainforest and coral reefs (Boers & Rypdal, 2021; Boulton et al., 2022). 4.4. Challenges associated with GeoAI applications in climate science Despite its transformative promise, the deployment and operationalization of GeoAI face several interconnected challenges (Table 5). Table 5 Challenges facing GeoAI applications in climate science Challenge Category Specific Description and Manifestations Data Quality, Heterogeneity & Paucity Issues include sensor noise, cloud contamination in optical imagery, spatial and temporal inconsistencies across different satellite platforms, a critical lack of high-quality, labeled training data for supervised learning (especially in remote regions and for rare events), and inherent biases in datasets that can perpetuate environmental injustices (Ball et al., 2017; Karpatne, Atluri, et al., 2017; Tuia et al., 2022). Computational Intractability & Resource Demands Training complex DL models (e.g., Vision Transformers) on high-resolution, global, multi-temporal datasets requires immense computational resources (e.g., clusters of GPUs/TPUs), creating a significant barrier to entry for researchers in low-resource institutions and countries, and raising the carbon footprint of AI research (Schwartz et al., 2020; Strubell et al., 2019). Model Generalizability, Transferability & Robustness Models trained on data from one geographic region, time period, or sensor often perform poorly when applied to another (domain shift), limiting their global utility and operational robustness. This is exacerbated by non-stationarity in climate systems, where past relationships may not hold in the future (Reichstein et al., 2019b; Rolnick et al., 2023; R. Wang et al., 2024). Explainability, Interpretability and Trust (The “Black Box” Problem) The complex, high-dimensional nature of many DL models hinders understanding of how and why they arrive at a particular prediction. This lack of transparency is a major barrier to adoption by domain scientists, policymakers, and stakeholders who require interpretable and physically plausible evidence to inform decisions (Roscher et al., 2020; Samek & Müller, 2019; Toms et al., 2020) Ethical Considerations, Algorithmic Fairness & Justice Biases in training data (e.g., uneven global distribution of weather stations) can lead to biased model outcomes that systematically under-predict risks for marginalized communities, potentially exacerbating existing climate vulnerabilities and violating principles of climate justice (Messina et al., 2019; Rolnick et al., 2023). Data sovereignty, privacy (e.g., from high-resolution imagery), and the potential for misuse (e.g., in geoengineering) are also profound concerns. Interoperability, Reproducibility & Workflow Integration Integrating diverse data formats (netCDF, GeoTIFF, and point clouds), software ecosystems (Python, R, and GIS), and reconciling the outputs from data-driven AI models with well-established physical models remain technically challenging. A lack of standardized benchmarking datasets and model reporting hinders reproducibility and fair comparison (Frame et al., 2022; Lundberg et al., 2017) 4.5. Synthesis and prioritization of GeoAI challenges While the preceding sections have detailed individual challenges, a synthesis is required to understand their relative severity, prevalence, and interrelationships to effectively prioritize mitigation efforts. Fig. 7 shows these impediments, combining a radar chart to assess perceived severity with a bar chart to indicate their prevalence in the reviewed literature. The radar chart (Fig. 7) shows expert-based severity scores (1–10 scale) for six major challenge categories. A higher score indicates a greater perceived impediment to field advancement. It reveals that the most severe challenges, as inferred from author discussions and critical analysis, are those related to Explainability & Trust and Data Quality & Heterogeneity. The “black box” nature of complex models (Explainability) is not simply a technical issue but a fundamental barrier to adoption by policymakers and domain scientists who require transparent, physically plausible evidence. Concurrently, the foundational issue of data (Quality & Heterogeneity) continues to undermine model reliability and generalization, a problem exacerbated by the unique noise characteristics of remote sensing data and the high cost of creating accurate labels. The prevalence analysis (Fig. 7) tells a slightly different story. Data Quality & Heterogeneity and Model Generalizability are the most frequently discussed challenges, indicating that they are the most immediate and commonly encountered setbacks in daily research. While Ethical Considerations are discussed with moderate frequency, they are perceived as highly severe, signaling a growing recognition of their profound implications for climate justice and the potential for algorithmic bias to exacerbate existing vulnerabilities. This synthesized view shows that the community faces three problems: 1) Foundational Data Issues, 2) Operational Model Limitations (generalizability and computational demands), and 3) Profound Epistemic and Ethical Concerns (explainability and ethics). Addressing these challenges requires a coordinated effort across computer science, geospatial theory, and social science. 5. Existing gaps and future directions The synthesis of the literature reveals several theoretical, methodological, and ethical gaps that present opportunities for future research (Fig. 8). 5.1. The explainability gap: from black box to trusted, transparent tool A predominant and widely acknowledged gap is the limited application of Explainable AI (XAI) and interpretability techniques in GeoAI-climate studies (Toms et al., 2020). While complex models often achieve high predictive accuracy, the physical reasoning and driving factors behind their predictions remain unclear, limiting their utility for scientific discovery and policy formulation. Future research should prioritize the systematic integration of post-hoc (e.g., SHAP (Lundberg et al., 2017) and LIME (Ribeiro et al., 2016) and intrinsic (e.g., attention mechanisms in transformers) XAI methods. The goal is to move beyond mere prediction to understanding, uncovering the spatiotemporal variables and interactions that the model considers essential and ensuring these align with domain knowledge and physical plausibility (Roscher et al., 2020; Toms et al., 2020). This is crucial for building trust, diagnosing model failures, and generating new, testable hypotheses about the climate system. 5.2. The causality gap: moving beyond correlation to causal understanding Most current GeoAI applications effectively identify complex, non-linear correlations but fail to establish causal relationships (Runge et al., 2019). Inferring causality from observational data helps elucidate the fundamental drivers of climate impacts, assess the efficacy of interventions, and make robust predictions under novel conditions (e.g., new policy scenarios). Future work should explore the integration of causal inference frameworks (Neuberg, 2003), such as causal directed acyclic graphs (DAGs), propensity score matching, and causal discovery algorithms, with GeoAI models. This represents a paradigm shift from purely predictive modeling to a causal learning paradigm that can answer “what if” questions, which is the cornerstone of effective climate adaptation and mitigation planning (Beucler et al., 2021; Nowack et al., 2020). 5.3. The generalizability gap: toward foundation models for earth observation The lack of model transferability and robustness across space, time, and sensors hinders the operational deployment of GeoAI (Tuia et al., 2022). A promising and transformative direction is the development of “foundation models” or “pre-trained models” for Earth observation (Irvin et al., 2020; Reed et al., 2023) and large-scale models pre-trained on diverse and unlabeled satellite data archives using self-supervised learning objectives. These models would learn general-purpose representations of the Earth’s surface, which could then be fine-tuned for specific climate tasks (e.g., flood mapping and crop classification) with minimal additional labeled data. This would democratize access to powerful GeoAI tools, enhance global consistency in monitoring, and significantly improve performance in data-sparse regions (Irrgang et al., 2021b; Sarmadi et al., 2023; R. Wang et al., 2024). 5.4. The integration gap: physics-informed and knowledge-guided GeoAI A significant frontier lies in the principled coupling of AI with physical knowledge and mechanistic understanding (Karpatne, Ebert-Uphoff, et al., 2017b). Pure data-driven models can produce physically inconsistent results, especially when extrapolating. PINNs (Raissi et al., 2019) and other approaches that embed physical laws (e.g., conservation equations and radiative transfer models) as soft constraints into the learning process can produce more robust, generalizable, and physically consistent models. This hybrid approach, called “theory-guided data science,” is particularly valuable for forecasting and simulation in regions with sparse observational data, ensuring that AI solutions respect the fundamental laws governing the Earth system(Kashinath et al., 2021; Read et al., 2019). 5.5. The equity and implementation gap: socially conscious, participatory, and equitable GeoAI. The social and ethical dimensions of GeoAI requires urgent attention (Y. Liu et al., 2018; Runting et al., 2020). Future research should proactively develop methods for auditing and mitigating algorithmic biases that could disadvantage vulnerable populations in climate risk assessments and resource allocation. Furthermore, studies should develop participatory and co-designed GeoAI frameworks that incorporate local knowledge, indigenous wisdom, and community priorities into the modeling process (Goldman et al., 2009). This ensures that GeoAI solutions are technically sophisticated, contextually appropriate, socially robust, and equitable, thereby aligning technological advancement with the principles of climate and environmental justice(Gavahi et al., 2023). 6. Conclusion This study comprehensively analyzed GeoAI applications in climate change impact assessment through scientometrics and a systematic review. The scientometric analysis of 152 core publications, augmented by a multi-database search strategy, quantitatively elucidated the field’s rapid, exponential growth from 2015 to 2025. It identified key thematic clusters around ML methodologies, remote sensing data, and specific climate impacts and mapped the influential roles of leading countries and premier journals in remote sensing and interdisciplinary science. The systematic review of 77 articles provided an in-depth examination of the field’s intellectual and technical aspects. It elucidated the dominant technological paradigms, with CNNs for spatial feature learning, RNNs/LSTMs for temporal dynamics, and emerging Transformer-based models for global context. Hybrid Physics-Informed AI frameworks and their transformative applications across critical domains, including cryosphere and hydrosphere monitoring, terrestrial and marine ecosystem analysis, agricultural security forecasting, and urban climate resilience planning, were also elucidated. The study critically examined the multi-faceted challenges of data quality and heterogeneity, computational intractability, the “black box” problem of model interpretability, limited generalizability, and profound ethical considerations that currently constrain the field’s full potential and equitable application. Finally, the synthesis identified pivotal research gaps, most notably the pressing need for explainable (XAI), causal, generalizable (e.g., via foundation models), physics-informed, and socially equitable GeoAI frameworks. These recommendations call for an interdisciplinary effort to develop these next-generation systems, which must be guided by the principles of transparency, physical consistency, and climate justice. By addressing these challenges and rigorously pursuing these future directions, the GeoAI research community can significantly augment our global capacity to understand, project, mitigate, and adapt to the profound and escalating impacts of climate change. Thus, we can provide the reliable, actionable, and trustworthy intelligence essential for navigating toward a resilient planetary future. 6.1. Limitations and research outlook This review is constrained by its scope and methodology. While multiple databases were used, some relevant studies in non-English languages or in less indexed venues may have been omitted. The focus on a specific set of keywords, although carefully selected, may not capture all nascent applications at the periphery of this rapidly evolving field. The systematic review’s qualitative synthesis, while rigorous, is inherently subject to researcher interpretation. Future work should include broader literature searches and more formal meta-analysis techniques where sufficient quantitative data exists. The field’s trajectory suggests that the integration of AI with Earth system science will only advance, prompting ongoing critical reviews and methodological innovations to ensure its responsible and effective deployment. Declarations Conflict of interest author declares no conflict of interest. References Alnajjar, S., García-Martínez, A., López-Cabeza, V. P., & Al-Azhari, W. (2025). A Multidimensional Approach to Mapping Urban Heat Vulnerability: Integrating Remote Sensing and Spatial Configuration. Smart Cities , 8 (4). https://doi.org/10.3390/smartcities8040137 Andela, N., Morton, D. C., Giglio, L., Chen, Y., Van Der Werf, G. R., Kasibhatla, P. S., DeFries, R. S., Collatz, G. J., Hantson, S., Kloster, S., Bachelet, D., Forrest, M., Lasslop, G., Li, F., Mangeon, S., Melton, J. R., Yue, C., & Randerson, J. T. (2017). A human-driven decline in global burned area. Science , 356 (6345), 1356–1362. https://doi.org/10.1126/science.aal4108 Aria, M., & Cuccurullo, C. (2017). bibliometrix: An R-tool for comprehensive science mapping analysis. Journal of Informetrics , 11 (4), 959–975. https://doi.org/10.1016/j.joi.2017.08.007 Aviles Toledo, C., Crawford, M. M., & Tuinstra, M. R. (2024). Integrating multi-modal remote sensing, deep learning, and attention mechanisms for yield prediction in plant breeding experiments. Frontiers in Plant Science , 15 . https://doi.org/10.3389/fpls.2024.1408047 Ball, J. E., Anderson, D. T., & Chan, C. S. (2017). Comprehensive survey of deep learning in remote sensing: theories, tools, and challenges for the community. Journal of Applied Remote Sensing , 11 (04), 1. https://doi.org/10.1117/1.jrs.11.042609 Ban, Y., Zhang, P., Nascetti, A., Bevington, A. R., & Wulder, M. A. (2020). Near Real-Time Wildfire Progression Monitoring with Sentinel-1 SAR Time Series and Deep Learning. Scientific Reports , 10 (1). https://doi.org/10.1038/s41598-019-56967-x Beucler, T., Pritchard, M., Rasp, S., Ott, J., Baldi, P., & Gentine, P. (2021). Enforcing Analytic Constraints in Neural Networks Emulating Physical Systems. Physical Review Letters , 126 (9). https://doi.org/10.1103/PhysRevLett.126.098302 Bi, K., Xie, L., Zhang, H., Chen, X., Gu, X., & Tian, Q. (2023). Accurate medium-range global weather forecasting with 3D neural networks. Nature , 619 (7970), 533–538. https://doi.org/10.1038/s41586-023-06185-3 Boers, N., & Rypdal, M. (2021). Critical slowing down suggests that the western Greenland Ice Sheet is close to a tipping point . https://doi.org/10.1073/pnas.2024192118/-/DCSupplemental.y Botella, C., Joly, A., Bonnet, P., Monestiez, P., & Munoz, F. (2018). A deep learning approach to Species Distribution Modelling . Boulton, C. A., Lenton, T. M., & Boers, N. (2022). Pronounced loss of Amazon rainforest resilience since the early 2000s. Nature Climate Change , 12 (3), 271–278. https://doi.org/10.1038/s41558-022-01287-8 Breiman, L. (2001). Random Forests (Vol. 45). Bullock, E. L., Woodcock, C. E., & Holden, C. E. (2020). Improved change monitoring using an ensemble of time series algorithms. Remote Sensing of Environment , 238 . https://doi.org/10.1016/j.rse.2019.04.018 Chen, T., & Guestrin, C. (2016). XGBoost: A scalable tree boosting system. Proceedings of the ACM SIGKDD International Conference on Knowledge Discovery and Data Mining , 13-17-August-2016 , 785–794. https://doi.org/10.1145/2939672.2939785 Cho, K., van Merrienboer, B., Gulcehre, C., Bahdanau, D., Bougares, F., Schwenk, H., & Bengio, Y. (2014). Learning Phrase Representations using RNN Encoder-Decoder for Statistical Machine Translation . http://arxiv.org/abs/1406.1078 Creswell, J. W. ., & Poth, C. N. . (2018). Qualitative inquiry and research design . SAGE. Dosovitskiy, A., Beyer, L., Kolesnikov, A., Weissenborn, D., Zhai, X., Unterthiner, T., Dehghani, M., Minderer, M., Heigold, G., Gelly, S., Uszkoreit, J., & Houlsby, N. (2021). An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale . http://arxiv.org/abs/2010.11929 Elith, J., Leathwick, J. R., & Hastie, T. (2008). A working guide to boosted regression trees. In Journal of Animal Ecology (Vol. 77, Issue 4, pp. 802–813). https://doi.org/10.1111/j.1365-2656.2008.01390.x Ellegaard, O., & Wallin, J. A. (2015). The bibliometric analysis of scholarly production: How great is the impact? Scientometrics , 105 (3), 1809–1831. https://doi.org/10.1007/s11192-015-1645-z Elo, S., & Kyngäs, H. (2008). The qualitative content analysis process. Journal of Advanced Nursing , 62 (1), 107–115. https://doi.org/10.1111/j.1365-2648.2007.04569.x Fang, K., Shen, C., Kifer, D., & Yang, X. (2017). Prolongation of SMAP to Spatiotemporally Seamless Coverage of Continental U.S. Using a Deep Learning Neural Network. Geophysical Research Letters , 44 (21), 11,030-11,039. https://doi.org/10.1002/2017GL075619 Fayaz, M., Nam, J., Dang, L. M., Song, H. K., & Moon, H. (2024). Land-Cover Classification Using Deep Learning with High-Resolution Remote-Sensing Imagery. Applied Sciences (Switzerland) , 14 (5). https://doi.org/10.3390/app14051844 Frame, J. M., Kratzert, F., Klotz, D., Gauch, M., Shelev, G., Gilon, O., Qualls, L. M., Gupta, H. V., & Nearing, G. S. (2022). Deep learning rainfall-runoff predictions of extreme events. Hydrology and Earth System Sciences , 26 (13), 3377–3392. https://doi.org/10.5194/hess-26-3377-2022 Free, G., Bresciani, M., Pinardi, M., Ghirardi, N., Luciani, G., Caroni, R., & Giardino, C. (2021). Detecting climate driven changes in chlorophyll-a in deep subalpine lakes using long term satellite data. Water (Switzerland) , 13 (6). https://doi.org/10.3390/w13060866 Gavahi, K., Forumandi, E., & Moradkhani, H. (2023). A deep learning-based framework for multi-source 2 precipitation fusion . https://www.sciencedirect.com/science/article/pii/S0034425723002742 Goldman, J., Shilton, K., Burke, J., Estrin, D., Hansen, M., Ramanathan, N., Reddy, S., Samanta, V., Srivastava, M., & West, R. (2009). Participatory Sensing . http://www.wilsoncenter.org/foresight/ Ham, Y. G., Kim, J. H., & Luo, J. J. (2019). Deep learning for multi-year ENSO forecasts. Nature , 573 (7775), 568–572. https://doi.org/10.1038/s41586-019-1559-7 Harris, L., McRae, A. T. T., Chantry, M., Dueben, P. D., & Palmer, T. N. (2022). A Generative Deep Learning Approach to Stochastic Downscaling of Precipitation Forecasts. Journal of Advances in Modeling Earth Systems , 14 (10). https://doi.org/10.1029/2022MS003120 Heidler, K., Mou, L., Baumhoer, C., Dietz, A., & Zhu, X. X. (2022). HED-UNet: Combined Segmentation and Edge Detection for Monitoring the Antarctic Coastline. IEEE Transactions on Geoscience and Remote Sensing , 60 . https://doi.org/10.1109/TGRS.2021.3064606 Hochreiter, S., & Schmidhuber, J. (1997). Long Short-Term Memory. Neural Computation , 9 (8), 1735–1780. https://doi.org/10.1162/neco.1997.9.8.1735 Hsieh, H. F., & Shannon, S. E. (2005). Three approaches to qualitative content analysis. Qualitative Health Research , 15 (9), 1277–1288. https://doi.org/10.1177/1049732305276687 Iglovikov, V., Seferbekov, S., Buslaev, A., & Shvets, A. (2018). TernausNetV2: Fully convolutional network for instance segmentation. IEEE Computer Society Conference on Computer Vision and Pattern Recognition Workshops , 2018-June , 228–232. https://doi.org/10.1109/CVPRW.2018.00042 Intergovernmental Panel on Climate Change (IPCC). (2023). Weather and Climate Extreme Events in a Changing Climate. In Climate Change 2021 – The Physical Science Basis (pp. 1513–1766). Cambridge University Press. https://doi.org/10.1017/9781009157896.013 Irrgang, C., Boers, N., Sonnewald, M., Barnes, E. A., Kadow, C., Staneva, J., & Saynisch-Wagner, J. (2021a). Towards neural Earth system modelling by integrating artificial intelligence in Earth system science. In Nature Machine Intelligence (Vol. 3, Issue 8, pp. 667–674). Nature Research. https://doi.org/10.1038/s42256-021-00374-3 Irrgang, C., Boers, N., Sonnewald, M., Barnes, E. A., Kadow, C., Staneva, J., & Saynisch-Wagner, J. (2021b). Towards neural Earth system modelling by integrating artificial intelligence in Earth system science. In Nature Machine Intelligence (Vol. 3, Issue 8, pp. 667–674). Nature Research. https://doi.org/10.1038/s42256-021-00374-3 Irvin, J., Sheng, H., Ramachandran, N., Johnson-Yu, S., Zhou, S., Story, K., Rustowicz, R., Elsworth, C., Austin, K., & Ng, A. Y. (2020). ForestNet: Classifying Drivers of Deforestation in Indonesia using Deep Learning on Satellite Imagery . http://arxiv.org/abs/2011.05479 Jain, P., Coogan, S. C. P., Subramanian, S. G., Crowley, M., Taylor, S., & Flannigan, M. D. (2020). A review of machine learning applications in wildfire science and management. In Environmental Reviews (Vol. 28, Issue 4, pp. 478–505). Canadian Science Publishing. https://doi.org/10.1139/er-2020-0019 Janowicz, K., Gao, S., McKenzie, G., Hu, Y., & Bhaduri, B. (2020). GeoAI: spatially explicit artificial intelligence techniques for geographic knowledge discovery and beyond. In International Journal of Geographical Information Science (Vol. 34, Issue 4, pp. 625–636). Taylor and Francis Ltd. https://doi.org/10.1080/13658816.2019.1684500 Jin, Z., Azzari, G., You, C., Di Tommaso, S., Aston, S., Burke, M., & Lobell, D. B. (2019). Smallholder maize area and yield mapping at national scales with Google Earth Engine. Remote Sensing of Environment , 228 , 115–128. https://doi.org/10.1016/j.rse.2019.04.016 Karpatne, A., Atluri, G., Faghmous, J. H., Steinbach, M., Banerjee, A., Ganguly, A., Shekhar, S., Samatova, N., & Kumar, V. (2017). Theory-guided data science: A new paradigm for scientific discovery from data. IEEE Transactions on Knowledge and Data Engineering , 29 (10), 2318–2331. https://doi.org/10.1109/TKDE.2017.2720168 Karpatne, A., Ebert-Uphoff, I., Ravela, S., Babaie, H. A., & Kumar, V. (2017a). Machine Learning for the Geosciences: Challenges and Opportunities . http://arxiv.org/abs/1711.04708 Karpatne, A., Ebert-Uphoff, I., Ravela, S., Babaie, H. A., & Kumar, V. (2017b). Machine Learning for the Geosciences: Challenges and Opportunities . http://arxiv.org/abs/1711.04708 Kashinath, K., Mustafa, M., Albert, A., Wu, J. L., Jiang, C., Esmaeilzadeh, S., Azizzadenesheli, K., Wang, R., Chattopadhyay, A., Singh, A., Manepalli, A., Chirila, D., Yu, R., Walters, R., White, B., Xiao, H., Tchelepi, H. A., Marcus, P., Anandkumar, A., … Prabhat. (2021). Physics-informed machine learning: Case studies for weather and climate modelling. In Philosophical Transactions of the Royal Society A: Mathematical, Physical and Engineering Sciences (Vol. 379, Issue 2194). Royal Society Publishing. https://doi.org/10.1098/rsta.2020.0093 Ke, G., Meng, Q., Finley, T., Wang, T., Chen, W., Ma, W., Ye, Q., & Liu, T.-Y. (2017). LightGBM: A Highly Efficient Gradient Boosting Decision Tree . https://www.researchgate.net/publication/378480234 Khaki, S., Wang, L., & Archontoulis, S. V. (2020). A CNN-RNN Framework for Crop Yield Prediction. Frontiers in Plant Science , 10 . https://doi.org/10.3389/fpls.2019.01750 Kratzert, F., Nearing, G., Addor, N., Erickson, T., Gauch, M., Gilon, O., Gudmundsson, L., Hassidim, A., Klotz, D., Nevo, S., Shalev, G., & Matias, Y. (2023). Caravan - A global community dataset for large-sample hydrology. Scientific Data , 10 (1). https://doi.org/10.1038/s41597-023-01975-w Krizhevsky, A., Sutskever, I., & Hinton, G. E. (2017). ImageNet classification with deep convolutional neural networks. Communications of the ACM , 60 (6), 84–90. https://doi.org/10.1145/3065386 Kurth, T., Subramanian, S., Harrington, P., Pathak, J., Mardani, M., Hall, D., Miele, A., Kashinath, K., & Anandkumar, A. (2023, June 26). FourCastNet: Accelerating Global High-Resolution Weather Forecasting Using Adaptive Fourier Neural Operators. Proceedings of the Platform for Advanced Scientific Computing Conference, PASC 2023 . https://doi.org/10.1145/3592979.3593412 Lacoste, A., Sherwin, E. D., Kerner, H., Alemohammad, H., Lütjens, B., Irvin, J., Dao, D., Chang, A., Gunturkun, M., Drouin, A., Rodriguez, P., & Vazquez, D. (2021). Toward Foundation Models for Earth Monitoring: Proposal for a Climate Change Benchmark . http://arxiv.org/abs/2112.00570 Längkvist, M., Kiselev, A., Alirezaie, M., & Loutfi, A. (2016). Classification and segmentation of satellite orthoimagery using convolutional neural networks. Remote Sensing , 8 (4). https://doi.org/10.3390/rs8040329 Le, X. H., Ho, H. V., Lee, G., & Jung, S. (2019). Application of Long Short-Term Memory (LSTM) neural network for flood forecasting. Water (Switzerland) , 11 (7). https://doi.org/10.3390/w11071387 Lecun, Y., Bengio, Y., & Hinton, G. (2015). Deep learning. In Nature (Vol. 521, Issue 7553, pp. 436–444). Nature Publishing Group. https://doi.org/10.1038/nature14539 Lees, T., Buechel, M., Anderson, B., Slater, L., Reece, S., Coxon, G., & Dadson, S. J. (2021). Benchmarking data-driven rainfall-runoff models in Great Britain: A comparison of long short-term memory (LSTM)-based models with four lumped conceptual models. Hydrology and Earth System Sciences , 25 (10), 5517–5534. https://doi.org/10.5194/hess-25-5517-2021 Li, F., Yigitcanlar, T., Nepal, M., Thanh, K. N., & Dur, F. (2024). A Novel Urban Heat Vulnerability Analysis: Integrating Machine Learning and Remote Sensing for Enhanced Insights. Remote Sensing , 16 (16). https://doi.org/10.3390/rs16163032 Li, W. (2020a). GeoAI: Where machine learning and big data converge in GIScience. Journal of Spatial Information Science , 20 , 71–77. https://doi.org/10.5311/JOSIS.2020.20.658 Li, W. (2020b). GeoAI: Where machine learning and big data converge in GIScience. Journal of Spatial Information Science , 20 , 71–77. https://doi.org/10.5311/JOSIS.2020.20.658 Liu, H., Zhong, H., Lin, W., & Wu, J. (2024). Tree species classification based on PointNet++ deep learning and true- colour point cloud. International Journal of Remote Sensing , 45 (16), 5577–5604. https://doi.org/10.1080/01431161.2024.2377837 Liu, Y., Bi, J., & Lv, J. (2018). Future impacts of climate change and land use on multiple ecosystem services in a rapidly urbanizing agricultural Basin, China. Sustainability (Switzerland) , 10 (12). https://doi.org/10.3390/su10124575 Lowe, R., Lee, S. A., Brady, O. J., Bastos, L., Colón-González, F. J., on Climate Change, C., Health Lowe, P. R., Gasparrini, A., Carvalho, M. S., Lowe, R., Lee, S. A., O, K. M., Brady, O. J., Bastos, L., Carrasco-Escobar, G., de Castro Catão, R., Colón-González, F. J., Barcellos, C., Sá Carvalho, M., … Gasparrini, A. (2021). Centre for Mathematical Modelling of Infectious Diseases ( Department of Public Health Environments and Society (Prof A Gasparrini), and Centre for Statistical Modelling (Prof A Gasparrini) Combined effects of hydrometeorological hazards and urbanisation on dengue risk in Brazil: a spatiotemporal modelling study. In Articles Lancet Planet Health (Vol. 5). www.thelancet.com/ Lundberg, S. M., Allen, P. G., & Lee, S.-I. (2017). A Unified Approach to Interpreting Model Predictions. In ACM Digital Library . https://github.com/slundberg/shap Ma, L., Liu, Y., Zhang, X., Ye, Y., Yin, G., & Johnson, B. A. (2019a). Deep learning in remote sensing applications: A meta-analysis and review. In ISPRS Journal of Photogrammetry and Remote Sensing (Vol. 152, pp. 166–177). Elsevier B.V. https://doi.org/10.1016/j.isprsjprs.2019.04.015 Ma, L., Liu, Y., Zhang, X., Ye, Y., Yin, G., & Johnson, B. A. (2019b). Deep learning in remote sensing applications: A meta-analysis and review. In ISPRS Journal of Photogrammetry and Remote Sensing (Vol. 152, pp. 166–177). Elsevier B.V. https://doi.org/10.1016/j.isprsjprs.2019.04.015 Martín-Martín, A., Thelwall, M., Orduna-Malea, E., & Delgado López-Cózar, E. (2021). Google Scholar, Microsoft Academic, Scopus, Dimensions, Web of Science, and OpenCitations’ COCI: a multidisciplinary comparison of coverage via citations. Scientometrics , 126 (1), 871–906. https://doi.org/10.1007/s11192-020-03690-4 Maxwell, A. E., Warner, T. A., & Fang, F. (2018). Implementation of machine-learning classification in remote sensing: An applied review. In International Journal of Remote Sensing (Vol. 39, Issue 9, pp. 2784–2817). Taylor and Francis Ltd. https://doi.org/10.1080/01431161.2018.1433343 Mei, W., Wang, H., Fouhey, D., Zhou, W., Hinks, I., Gray, J. M., Van Berkel, D., & Jain, M. (2022). Using Deep Learning and Very-High-Resolution Imagery to Map Smallholder Field Boundaries. Remote Sensing , 14 (13). https://doi.org/10.3390/rs14133046 Mentaschi, L., Vousdoukas, M. I., Pekel, J. F., Voukouvalas, E., & Feyen, L. (2018). Global long-term observations of coastal erosion and accretion. Scientific Reports , 8 (1). https://doi.org/10.1038/s41598-018-30904-w Messina, J. P., Brady, O. J., Golding, N., Kraemer, M. U. G., Wint, G. R. W., Ray, S. E., Pigott, D. M., Shearer, F. M., Johnson, K., Earl, L., Marczak, L. B., Shirude, S., Davis Weaver, N., Gilbert, M., Velayudhan, R., Jones, P., Jaenisch, T., Scott, T. W., Reiner, R. C., & Hay, S. I. (2019). The current and future global distribution and population at risk of dengue. Nature Microbiology , 4 (9), 1508–1515. https://doi.org/10.1038/s41564-019-0476-8 Neuberg, L. (2003). Causality: models, reasoning, and inference, by judea pearl, cambridge university press, 2000. Econometric Theory , 19 , 675–685. https://doi.org/10.1017/S0266466603004109 Nowack, P., Runge, J., Eyring, V., & Haigh, J. D. (2020). Causal networks for climate model evaluation and constrained projections. Nature Communications , 11 (1). https://doi.org/10.1038/s41467-020-15195-y Page, M. J., McKenzie, J. E., Bossuyt, P. M., Boutron, I., Hoffmann, T. C., Mulrow, C. D., Shamseer, L., Tetzlaff, J. M., Akl, E. A., Brennan, S. E., Chou, R., Glanville, J., Grimshaw, J. M., Hróbjartsson, A., Lalu, M. M., Li, T., Loder, E. W., Mayo-Wilson, E., McDonald, S., … Moher, D. (2021). The PRISMA 2020 statement: An updated guideline for reporting systematic reviews. Journal of Clinical Epidemiology , 134 , 178–189. https://doi.org/10.1016/j.jclinepi.2021.03.001 Pahlevan, N., Smith, B., Schalles, J., Binding, C., Cao, Z., Ma, R., Alikas, K., Kangro, K., Gurlin, D., Hà, N., Matsushita, B., Moses, W., Greb, S., Lehmann, M. K., Ondrusek, M., Oppelt, N., & Stumpf, R. (2020). Seamless retrievals of chlorophyll-a from Sentinel-2 (MSI) and Sentinel-3 (OLCI) in inland and coastal waters: A machine-learning approach. Remote Sensing of Environment , 240 . https://doi.org/10.1016/j.rse.2019.111604 Raissi, M., Perdikaris, P., & Karniadakis, G. E. (2019). Physics-informed neural networks: A deep learning framework for solving forward and inverse problems involving nonlinear partial differential equations. Journal of Computational Physics , 378 , 686–707. https://doi.org/10.1016/j.jcp.2018.10.045 Rasp, S., Pritchard, M. S., & Gentine, P. (2018). Deep learning to represent subgrid processes in climate models . https://doi.org/10.5281/zenodo.1402384 Read, J. S., Jia, X., Willard, J., Appling, A. P., Zwart, J. A., Oliver, S. K., Karpatne, A., Hansen, G. J. A., Hanson, P. C., Watkins, W., Steinbach, M., & Kumar, V. (2019). Process-Guided Deep Learning Predictions of Lake Water Temperature. Water Resources Research , 55 (11), 9173–9190. https://doi.org/10.1029/2019WR024922 Reed, C. J., Gupta, R., Li, S., Brockman, S., Funk, C., Clipp, B., Keutzer, K., Candido, S., Uyttendaele, M., & Darrell, T. (2023). Scale-MAE: A Scale-Aware Masked Autoencoder for Multiscale Geospatial Representation Learning . http://arxiv.org/abs/2212.14532 Reichstein, M., Camps-Valls, G., Stevens, B., Jung, M., Denzler, J., Carvalhais, N., & Prabhat. (2019a). Deep learning and process understanding for data-driven Earth system science. Nature , 566 (7743), 195–204. https://doi.org/10.1038/s41586-019-0912-1 Reichstein, M., Camps-Valls, G., Stevens, B., Jung, M., Denzler, J., Carvalhais, N., & Prabhat. (2019b). Deep learning and process understanding for data-driven Earth system science. Nature , 566 (7743), 195–204. https://doi.org/10.1038/s41586-019-0912-1 Ribeiro, M. T., Singh, S., & Guestrin, C. (2016). “Why should i trust you?” Explaining the predictions of any classifier. Proceedings of the ACM SIGKDD International Conference on Knowledge Discovery and Data Mining , 13-17-August-2016 , 1135–1144. https://doi.org/10.1145/2939672.2939778 Ripple, W. J., Wolf, C., Newsome, T. M., Barnard, P., & Moomaw, W. R. (2020). World Scientists’ Warning of a Climate Emergency SCIENTIST SIGNATORIES FROM 153 COUNTRIES (LIST IN SUPPLEMENTAL FILE S1) (Vol. 70, Issue 1). https://academic.oup.com/bioscience Rolnick, D., Donti, P. L., Kaack, L. H., Kochanski, K., Lacoste, A., Sankaran, K., Ross, A. S., Milojevic-Dupont, N., Jaques, N., Waldman-Brown, A., Luccioni, A. S., Maharaj, T., Sherwin, E. D., Mukkavilli, S. K., Kording, K. P., Gomes, C. P., Ng, A. Y., Hassabis, D., Platt, J. C., … Bengio, Y. (2023). Tackling Climate Change with Machine Learning. In ACM Computing Surveys (Vol. 55, Issue 2). Association for Computing Machinery. https://doi.org/10.1145/3485128 Ronneberger, O., Fischer, P., & Brox, T. (2015). U-Net: Convolutional Networks for Biomedical Image Segmentation . http://arxiv.org/abs/1505.04597 Roscher, R., Jülich, F., Bohn, B., Duarte, M., Garcke, J., & Duarte, M. F. (2020). Explainable Machine Learning for Scientific Insights and Discoveries. IEEE . https://doi.org/10.48550/arXiv.1905.08883 Runge, J., Nowack, P., Kretschmer, M., Flaxman, S., & Sejdinovic, D. (2019). Detecting and quantifying causal associations in large nonlinear time series datasets. In Sci. Adv (Vol. 5). https://www.science.org Runting, R. K., Phinn, S., Xie, Z., Venter, O., & Watson, J. E. M. (2020). Opportunities for big data in conservation and sustainability. In Nature Communications (Vol. 11, Issue 1). Nature Research. https://doi.org/10.1038/s41467-020-15870-0 Samek, W., & Müller, K. R. (2019). Towards Explainable Artificial Intelligence. In Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics): Vol. 11700 LNCS (pp. 5–22). Springer Verlag. https://doi.org/10.1007/978-3-030-28954-6_1 Sarmadi, H., Rögnvaldsson, T., Carlsson, N. R., Ohlsson, M., Wahab, I., & Hall, O. (2023). Towards Explaining Satellite Based Poverty Predictions with Convolutional Neural Networks . https://doi.org/10.1109/DSAA60987.2023.10302541 Schneider, T., Lan, S., Stuart, A., & Teixeira, J. (2017a). Earth System Modeling 2.0: A Blueprint for Models That Learn From Observations and Targeted High-Resolution Simulations. Geophysical Research Letters , 44 (24), 12,396-12,417. https://doi.org/10.1002/2017GL076101 Schneider, T., Lan, S., Stuart, A., & Teixeira, J. (2017b). Earth System Modeling 2.0: A Blueprint for Models That Learn From Observations and Targeted High-Resolution Simulations. Geophysical Research Letters , 44 (24), 12,396-12,417. https://doi.org/10.1002/2017GL076101 Schwartz, R., Dodge, J., Smith, N. A., & Etzioni, O. (2020). Green AI. Communications of the ACM , 63 (12), 54–63. https://doi.org/10.1145/3381831 Shi, X., Chen, Z., Wang, H., Yeung, D.-Y., Wong, W.-K., Woo, W.-C., & Kong Observatory, H. (2015). Convolutional LSTM Network: A Machine Learning Approach for Precipitation Nowcasting. ACM Digital Library . Shukla, P. R., Skea, J., Reisinger, A., Slade, R., Fradera, R., Pathak, M., Al, A., Malek, K., Renée Van Diemen, B., Hasija, A., Lisboa, G., Luz, S., Malley, J., Mccollum, D., & Some, S. (2022). Climate Change 2022 Mitigation of Climate Change Working Group III Contribution to the Sixth Assessment Report of the Intergovernmental Panel on Climate Change Summary for Policymakers Edited by . www.ipcc.ch Stark, T. (2018). Using Deep Convolutional Neural Networks for the Identification of Informal Settlements to Improve a Sustainable Development in Urban Environments . Steffen, W., Rockström, J., Richardson, K., Lenton, T. M., Folke, C., Liverman, D., Summerhayes, C. P., Barnosky, A. D., Cornell, S. E., Crucifix, M., Donges, J. F., Fetzer, I., Lade, S. J., Scheffer, M., Winkelmann, R., & Schellnhuber, H. J. (2018). Trajectories of the Earth System in the Anthropocene. PNAS . https://doi.org/10.1073/pnas.1810141115/-/DCSupplemental Strubell, E., Ganesh, A., & Mccallum, A. (2019). Energy and Policy Considerations for Deep Learning in NLP. In ACL Anthology . https://doi.org/doi.org/10.18653/v1/P19-1355 Su, J., Yi, D., Coombes, M., Liu, C., Zhai, X., McDonald-Maier, K., & Chen, W. H. (2022). Spectral analysis and mapping of blackgrass weed by leveraging machine learning and UAV multispectral imagery. Computers and Electronics in Agriculture , 192 . https://doi.org/10.1016/j.compag.2021.106621 Tate, E., Rahman, M. A., Emrich, C. T., & Sampson, C. C. (2021). Flood exposure and social vulnerability in the United States. Natural Hazards , 106 (1), 435–457. https://doi.org/10.1007/s11069-020-04470-2 Tehrani, A. A., Veisi, O., Fakhr, B. V., & Du, D. (2024). Predicting solar radiation in the urban area: A data-driven analysis for sustainable city planning using artificial neural networking. Sustainable Cities and Society , 100 , 105042. https://doi.org/10.1016/J.SCS.2023.105042 Toms, B. A., Barnes, E. A., & Ebert-Uphoff, I. (2020). Physically Interpretable Neural Networks for the Geosciences: Applications to Earth System Variability. Journal of Advances in Modeling Earth Systems , 12 (9). https://doi.org/10.1029/2019MS002002 Tuia, D., Kellenberger, B., Beery, S., Costelloe, B. R., Zuffi, S., Risse, B., Mathis, A., Mathis, M. W., van Langevelde, F., Burghardt, T., Kays, R., Klinck, H., Wikelski, M., Couzin, I. D., van Horn, G., Crofoot, M. C., Stewart, C. V., & Berger-Wolf, T. (2022). Perspectives in machine learning for wildlife conservation. In Nature Communications (Vol. 13, Issue 1). Nature Research. https://doi.org/10.1038/s41467-022-27980-y Valavi, R., Guillera-Arroita, G., Lahoz-Monfort, J. J., & Elith, J. (2022). Predictive performance of presence-only species distribution models: a benchmark study with reproducible code. Ecological Monographs , 92 (1). https://doi.org/10.1002/ecm.1486 Vandal, T., Kodra, E., Ganguly, S., Michaelis, A., Nemani, R., & Ganguly, A. R. (2017). DeepSD: Generating High Resolution Climate Change Projections through Single Image Super-Resolution . http://arxiv.org/abs/1703.03126 Vaswani, A., Brain, G., Shazeer, N., Parmar, N., Uszkoreit, J., Jones, L., Gomez, A. N., Kaiser, Ł., & Polosukhin, I. (2023). Attention Is All You Need . Vos, K., Deng, W., Harley, M. D., Turner, I. L., & Splinter, K. D. M. (2022). Beach-face slope dataset for Australia. Earth System Science Data , 14 (3), 1345–1357. https://doi.org/10.5194/essd-14-1345-2022 Wang, R., Ma, L., He, G., Johnson, B. A., Yan, Z., Chang, M., & Liang, Y. (2024). Transformers for Remote Sensing: A Systematic Review and Analysis. In Sensors (Vol. 24, Issue 11). Multidisciplinary Digital Publishing Institute (MDPI). https://doi.org/10.3390/s24113495 Wang, Y., Dong, P., Hu, W., Chen, G., Zhang, D., Chen, B., & Lei, G. (2022). Modeling the Climate Suitability of Northernmost Mangroves in China under Climate Change Scenarios. Forests , 13 (1). https://doi.org/10.3390/f13010064 Wohlin, C. (2014). Guidelines for snowballing in systematic literature studies and a replication in software engineering. ACM International Conference Proceeding Series . https://doi.org/10.1145/2601248.2601268 Wulder, M. A., Coops, N. C., Roy, D. P., White, J. C., & Hermosilla, T. (2018). Land cover 2.0. In International Journal of Remote Sensing (Vol. 39, Issue 12, pp. 4254–4284). Taylor and Francis Ltd. https://doi.org/10.1080/01431161.2018.1452075 Xiang, Z., Yan, J., & Demir, I. (2020). A Rainfall-Runoff Model With LSTM-Based Sequence-to-Sequence Learning. Water Resources Research , 56 (1). https://doi.org/10.1029/2019WR025326 Yang, C., Yu, M., Hu, F., Jiang, Y., & Li, Y. (2017). Utilizing Cloud Computing to address big geospatial data challenges. Computers, Environment and Urban Systems , 61 , 120–128. https://doi.org/10.1016/j.compenvurbsys.2016.10.010 Yang, W., Yuan, C. J., Tong, C., Yang, P., Yang, L., & Huang, B. (2017). Diurnal variation of CO2, CH4, and N2O emission fluxes continuously monitored in-situ in three environmental habitats in a subtropical estuarine wetland. Marine Pollution Bulletin , 119 (1), 289–298. https://doi.org/10.1016/j.marpolbul.2017.04.005 You, J., Li, X., Low, M., Lobell, D., & Ermon, S. (2017, February). Deep Gaussian Process for Crop Yield Prediction Based on Remote Sensing Data. ACM Digital Library . www.aaai.org Yu, Y., Si, X., Hu, C., & Zhang, J. (2019). A Review of Recurrent Neural Networks: LSTM Cells and Network Architectures. Neural Computation , 31 , 1235–1270. https://doi.org/10.1162/neco_a_01199 Zhang, E., Liu, L., Huang, L., & Ng, K. S. (2021). An automated, generalized, deep-learning-based method for delineating the calving fronts of Greenland glaciers from multi-sensor remote sensing imagery. Remote Sensing of Environment , 254 , 112265. https://doi.org/10.1016/J.RSE.2020.112265 Zhang, L., Zhang, L., & Du, B. (2016). Deep learning for remote sensing data: A technical tutorial on the state of the art. IEEE Geoscience and Remote Sensing Magazine , 4 (2), 22–40. https://doi.org/10.1109/MGRS.2016.2540798 Zhang, Z., Liu, Q., & Wang, Y. (2018). Road Extraction by Deep Residual U-Net. IEEE Geoscience and Remote Sensing Letters , 15 (5), 749–753. https://doi.org/10.1109/LGRS.2018.2802944 Zhu, X. X., Tuia, D., Mou, L., Xia, G.-S., Zhang, L., Xu, F., & Fraundorfer, F. (2017). Deep learning in remote sensing: a review . https://doi.org/10.1109/MGRS.2017.2762307 Zupic, I., & Čater, T. (2015). Bibliometric Methods in Management and Organization. Organizational Research Methods , 18 (3), 429–472. https://doi.org/10.1177/1094428114562629 Additional Declarations The authors declare no competing interests. Cite Share Download PDF Status: Posted Version 1 posted You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-8913236","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Systematic Review","associatedPublications":[],"authors":[{"id":593611090,"identity":"0e2c940e-81ad-431e-9143-b8f867a4d0d1","order_by":0,"name":"Francis Quayson","email":"data:image/png;base64,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","orcid":"https://orcid.org/0009-0004-8769-1461","institution":"Hong Kong Polytechnic University","correspondingAuthor":true,"prefix":"","firstName":"Francis","middleName":"","lastName":"Quayson","suffix":""}],"badges":[],"createdAt":"2026-02-19 02:14:56","currentVersionCode":1,"declarations":{"humanSubjects":false,"vertebrateSubjects":false,"conflictsOfInterestStatement":false,"humanSubjectEthicalGuidelines":false,"humanSubjectConsent":false,"humanSubjectClinicalTrial":false,"humanSubjectCaseReport":false,"vertebrateSubjectEthicalGuidelines":false},"doi":"10.21203/rs.3.rs-8913236/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-8913236/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":103312537,"identity":"ca281f3a-4abe-456d-901c-45fce6f5d4e9","added_by":"auto","created_at":"2026-02-24 10:20:41","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":204874,"visible":true,"origin":"","legend":"\u003cp\u003ePRISMA flow diagram for systematic review.\u003c/p\u003e","description":"","filename":"floatimage1.png","url":"https://assets-eu.researchsquare.com/files/rs-8913236/v1/3d3951607f2a2f868bc03f04.png"},{"id":103312536,"identity":"22f5890c-1d58-4811-8baa-d622c022904d","added_by":"auto","created_at":"2026-02-24 10:20:41","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":87253,"visible":true,"origin":"","legend":"\u003cp\u003eAnalytical framework showing methodology.\u003c/p\u003e","description":"","filename":"floatimage2.png","url":"https://assets-eu.researchsquare.com/files/rs-8913236/v1/3ffb0fa45bfa09846d3052f6.png"},{"id":103506428,"identity":"eb5606b4-a054-4f52-b382-2c998a14959b","added_by":"auto","created_at":"2026-02-26 13:36:21","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":117762,"visible":true,"origin":"","legend":"\u003cp\u003eAnnual publication trends from 2015 to June 2025.\u003c/p\u003e","description":"","filename":"floatimage3.png","url":"https://assets-eu.researchsquare.com/files/rs-8913236/v1/469ba344f755672f71dd78e5.png"},{"id":103312538,"identity":"64a7f81b-ca77-4609-83d6-d4596c4951f7","added_by":"auto","created_at":"2026-02-24 10:20:41","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":212136,"visible":true,"origin":"","legend":"\u003cp\u003eScience mapping of keyword co-occurrence network.\u003c/p\u003e","description":"","filename":"floatimage4.png","url":"https://assets-eu.researchsquare.com/files/rs-8913236/v1/6b2c620f175e37c6b0d061b3.png"},{"id":103312539,"identity":"83b3ac5e-c4d1-4a3e-8948-ec043fcc3e22","added_by":"auto","created_at":"2026-02-24 10:20:41","extension":"png","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":154946,"visible":true,"origin":"","legend":"\u003cp\u003eNetwork mapping of top publication outlets and their co-citation links.\u003c/p\u003e","description":"","filename":"floatimage5.png","url":"https://assets-eu.researchsquare.com/files/rs-8913236/v1/fb1c772d542e07e93ac10bdb.png"},{"id":103312543,"identity":"1c825a22-7eb0-43d0-8b63-cc4f1d8d7d6b","added_by":"auto","created_at":"2026-02-24 10:20:42","extension":"png","order_by":6,"title":"Figure 6","display":"","copyAsset":false,"role":"figure","size":103263,"visible":true,"origin":"","legend":"\u003cp\u003eNetwork mapping of active countries and international collaborations.\u003c/p\u003e","description":"","filename":"floatimage6.png","url":"https://assets-eu.researchsquare.com/files/rs-8913236/v1/6c416b142a884fd1b823ac5b.png"},{"id":103312541,"identity":"a6dab4c0-4119-45c8-a29c-23787402f638","added_by":"auto","created_at":"2026-02-24 10:20:41","extension":"png","order_by":7,"title":"Figure 7","display":"","copyAsset":false,"role":"figure","size":200802,"visible":true,"origin":"","legend":"\u003cp\u003eSynthesis of challenges with severity/prevalence facing GeoAI applications in climate science.\u003c/p\u003e","description":"","filename":"floatimage7.png","url":"https://assets-eu.researchsquare.com/files/rs-8913236/v1/54876ea528164788c1a04757.png"},{"id":103312542,"identity":"cb7dd780-3a99-49f4-a697-b146c44710ff","added_by":"auto","created_at":"2026-02-24 10:20:42","extension":"png","order_by":8,"title":"Figure 8","display":"","copyAsset":false,"role":"figure","size":113205,"visible":true,"origin":"","legend":"\u003cp\u003eResearch gaps and future direction analysis.\u003c/p\u003e","description":"","filename":"floatimage8.png","url":"https://assets-eu.researchsquare.com/files/rs-8913236/v1/5223ed33181817ea296020d9.png"},{"id":103509801,"identity":"06bd331b-50f6-4307-8564-8d31e177c8b8","added_by":"auto","created_at":"2026-02-26 14:01:21","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":1847301,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-8913236/v1/2d2b318b-778f-4e42-a243-09e23ce6107b.pdf"}],"financialInterests":"The authors declare no competing interests.","formattedTitle":"\u003cp\u003eDriven geospatial artificial intelligence modeling for climate change impact assessment: a global perspective\u003c/p\u003e","fulltext":[{"header":"1. Introduction","content":"\u003cp\u003eAnthropogenic climate change constitutes the defining socio-ecological challenge of the Anthropocene, a complex, polycausal phenomenon with cascading, non-linear implications for Earth system stability, biodiversity integrity, human security, and global economic equity (Ripple et al., 2020; Steffen et al., 2018). The Intergovernmental Panel on Climate Change (IPCC) has pointed out an accelerating trajectory of global warming and duration of compound extreme weather events, thereby pushing natural and human systems toward adaptation limits (IPCC, 2023; Shukla et al., 2022). Understanding, projecting, and mitigating the multifaceted impacts of these perturbations requires processing the vast, heterogeneous, multi-scale, and multi-modal datasets that characterize the modern Earth observation paradigm (Irrgang et al., 2021a; Reichstein et al., 2019a). While foundational to historical understanding, traditional mechanistic climate and impact models typically face computational, parametric, and structural limitations in assimilating the petabyte-scale data deluge from satellite remote sensing, unmanned aerial vehicles (UAVs), and pervasive in-situ sensor networks (Karpatne et al., 2017; Schneider et al., 2017). Therefore, Geospatial Artificial Intelligence (GeoAI) has emerged as a transformative scientific frontier. GeoAI is the interdisciplinary synthesis of geospatial theories, spatiotemporal data models, and computational methods with advanced AI paradigms, particularly from the statistical machine learning (ML) and deep learning (DL) domains (Janowicz et al., 2020; W. Li, 2020a). The proliferation of high-resolution satellite imagery (e.g., Sentinel, Landsat, and Planet), hyperspectral and Synthetic Aperture Radar (SAR) data, and the Internet of Things (IoT) has created a \u0026ldquo;geospatial big data\u0026rdquo; environment of unprecedented volume, velocity, and variety, suited for data-driven AI methodologies (Ma et al., 2019a; W. Yang et al., 2017). GeoAI techniques effectively extract non-obvious patterns, making high-dimensional predictions in ill-posed problems and revealing latent insights from these complex spatiotemporal datasets. Thus, they can radically enhance the granularity, accuracy, and temporal frequency of climate change impact assessments. (H. Liu et al., 2024; Y. Wang et al., 2022)\u003c/p\u003e\n\u003cp\u003eGeoAI applications in climate science are diverse and expanding rapidly, representing a shift from correlation-based analysis to a more mechanistic, although data-driven understanding. For instance, convolutional neural networks (CNNs), particularly encoder\u0026ndash;decoder architectures such as U-Net and DeepLab, are being deployed to delineate glacier boundaries, monitor ice sheet dynamics, and classify sea ice from multi-spectral and SAR satellite imagery with sub-pixel accuracy (Gavahi et al., 2023; R. Wang et al., 2024; L. Zhang et al., 2016). Similarly, recurrent neural networks (RNNs), Long Short-Term Memory (LSTM), and Gated Recurrent Unit (GRU) networks have demonstrated superiority in forecasting essential climate variables by learning long-term dependencies from historical time-series data from ground stations and gridded reanalysis products (Fang et al., 2017; Le et al., 2019; Shi et al., 2015). Additionally, ensemble ML methods, such as Random Forests and Gradient Boosting Machines (XGBoost and LightGBM), are used to model species distribution shifts in response to climate change. They are also used to assess the multi-hazard vulnerability of coastal zones and predict wildfire risk in a warming world (Elith et al., 2008; Jain et al., 2020; Krizhevsky et al., 2017; Mei et al., 2022). Despite these advancements, the field lacks a comprehensive, critical synthesis that maps its intellectual structure, evaluates its methodological trajectories and epistemological assumptions, and consolidates its persistent challenges into a coherent, prospective research agenda. Existing reviews have either focused on AI in environmental science without a geospatial focus (Rolnick et al., 2023) or on specific geospatial technologies (e.g., remote sensing) without a concentrated AI focus (Wulder et al., 2018) or have been limited in their temporal scope or methodological depth (Table 1). A thorough, state-of-the-art analysis that targets the epistemological and practical relationship between GeoAI and climate change impact assessment is thus essential to consolidate knowledge, identify methodological synergies and limitations, and guide future research investments and policies. This study fills this critical scholarly gap by analyzing GeoAI applications for climate change impact assessment from a global perspective. The objectives are as follows:\u003c/p\u003e\n\u003cp\u003e1. To identify the relevant corpus of literature in the GeoAI-climate change domain and conduct a scientometric analysis to quantitatively elucidate annual publication trends, thematic network structures, and global research contributions.\u003c/p\u003e\n\u003cp\u003e2. To systematically review and synthesize the predominant GeoAI technologies, methodological approaches, and application domains in climate change impact assessment, discussing their associated capabilities, limitations, and implementation contexts.\u003c/p\u003e\n\u003cp\u003e3. To identify persistent theoretical and applied research gaps and challenges within the literature and provide a strategic, prospective agenda for future scientific inquiry and technological development.\u003c/p\u003e\n\u003cp\u003eTable 1 \u0026nbsp;Overview of previous review articles on the AI\u0026ndash;Geospatial\u0026ndash;Climate nexus\u003c/p\u003e\n\u003ctable border=\"1\" cellspacing=\"0\" cellpadding=\"0\" width=\"914\"\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 83px;\"\u003e\n \u003cp\u003eAuthors (Year)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 159px;\"\u003e\n \u003cp\u003eReview Focus \u0026amp; Title\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 121px;\"\u003e\n \u003cp\u003eMethodological Approach\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 91px;\"\u003e\n \u003cp\u003eTimespan\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 212px;\"\u003e\n \u003cp\u003eKey Findings / Scope\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 249px;\"\u003e\n \u003cp\u003eIdentified Gaps / Limitations\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 83px;\"\u003e\n \u003cp\u003eRolnick et al. (2022)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 159px;\"\u003e\n \u003cp\u003eBroad applications of machine learning (ML) for climate change.\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 121px;\"\u003e\n \u003cp\u003eTopical synthesis and taxonomy development.\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 91px;\"\u003e\n \u003cp\u003e2015\u0026ndash;2021\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 212px;\"\u003e\n \u003cp\u003eCataloged ML opportunities across climate domains: mitigation, adaptation, modeling, and societal impacts.\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 249px;\"\u003e\n \u003cp\u003eLacks a dedicated geospatial theory and data perspective; scope is overly broad, missing depth in GeoAI methodologies.\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 83px;\"\u003e\n \u003cp\u003eZhu et al. (2017)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 159px;\"\u003e\n \u003cp\u003eDeep learning (DL) in remote sensing.\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 121px;\"\u003e\n \u003cp\u003eTechnical review and resource listing.\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 91px;\"\u003e\n \u003cp\u003e2012\u0026ndash;2017\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 212px;\"\u003e\n \u003cp\u003eComprehensive tutorial on DL architectures (CNNs, RNNs) for remote sensing tasks such as classification and detection.\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 249px;\"\u003e\n \u003cp\u003eFocuses on computer vision techniques; does not systematically address climate change applications or impact assessment.\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 83px;\"\u003e\n \u003cp\u003eWang et al. (2022a)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 159px;\"\u003e\n \u003cp\u003eGeoAI for hydroinformatics.\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 121px;\"\u003e\n \u003cp\u003eSystematic literature review.\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 91px;\"\u003e\n \u003cp\u003e2010\u0026ndash;2021\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 212px;\"\u003e\n \u003cp\u003eReviewed AI applications in water resources, including flood forecasting and water quality monitoring.\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 249px;\"\u003e\n \u003cp\u003eDomain-specific (hydrology); does not cover the full spectrum of climate impacts (e.g., ecosystems, urban areas, cryosphere).\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 83px;\"\u003e\n \u003cp\u003eLi (2020)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 159px;\"\u003e\n \u003cp\u003eFoundations and concepts of GeoAI.\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 121px;\"\u003e\n \u003cp\u003eConceptual analysis and vision paper.\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 91px;\"\u003e\n \u003cp\u003e2015\u0026ndash;2020\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 212px;\"\u003e\n \u003cp\u003eDefined the GeoAI paradigm, discussing spatial statistics, ML integration, and future trajectories.\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 249px;\"\u003e\n \u003cp\u003eTheoretical and foundational; lacks a systematic review of empirical applications, especially in climate science.\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 83px;\"\u003e\n \u003cp\u003eReichstein et al. (2019)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 159px;\"\u003e\n \u003cp\u003eDL for process understanding in Earth system science.\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 121px;\"\u003e\n \u003cp\u003ePerspective and selective review.\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 91px;\"\u003e\n \u003cp\u003e2016\u0026ndash;2019\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 212px;\"\u003e\n \u003cp\u003eAdvocated for combining DL with physical models for better causality and generalization in Earth science.\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 249px;\"\u003e\n \u003cp\u003eSelective, non-systematic review; focuses on the philosophy of integration rather than a comprehensive application survey.\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 83px;\"\u003e\n \u003cp\u003eWulder et al. (2018)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 159px;\"\u003e\n \u003cp\u003eLand cover mapping and change detection using remote sensing.\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 121px;\"\u003e\n \u003cp\u003eCritical review of technological trends.\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 91px;\"\u003e\n \u003cp\u003e2000\u0026ndash;2018\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 212px;\"\u003e\n \u003cp\u003eDescribed the evolution from manual digitization to automated methods for land cover monitoring.\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 249px;\"\u003e\n \u003cp\u003eFocused on a specific application (land cover); does not review the broader AI toolkit or other climate impact domains.\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n\u003c/table\u003e"},{"header":"2. Research methodology","content":"\u003cp\u003eThis study adopts a convergent parallel mixed-methods approach (Creswell and Poth, 2018), synergizing quantitative scientometric analysis with qualitative systematic review to holistically examine the GeoAI-climate change literature. The quantitative component involves a systematic, multi-database retrieval and scientometric mapping of relevant studies, while the qualitative component entails a detailed, critical, and thematic assessment of the conceptual and methodological content of the retrieved literature. This mixed-methods approach mitigates the inherent limitations and biases of either approach used in isolation and ensures a comprehensive, evidence-based understanding of the field\u0026rsquo;s current state, dynamics, and intellectual structure. The literature retrieval procedure is shown in \u003cstrong\u003eFig.\u003c/strong\u003e 1, and the overall analytical framework of the study is shown in \u003cstrong\u003eFig. 2\u003c/strong\u003e.\u003c/p\u003e\n\u003cp\u003e2.1. Retrieval of relevant studies\u003c/p\u003e\n\u003cp\u003eA robust literature retrieval protocol was employed, as shown in Fig. 1, following the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines (Page et al., 2021). To ensure comprehensive coverage and minimize database-specific bias, the primary search was conducted across four major scholarly databases: Scopus, Web of Science Core Collection, IEEE Xplore Digital Library, and Google Scholar. These platforms were selected for their extensive coverage of high-quality, peer-reviewed literature in engineering, environmental science, computer science, and geoinformatics (Mart\u0026iacute;n-Mart\u0026iacute;n et al., 2021; Zhu et al., 2017). Through an iterative process, the search query was designed to capture the core conceptual domains of GeoAI and climate change, using the TITLE-ABS-KEY fields in Scopus and Web of Science and analogous advanced search functions in IEEE Xplore and Google Scholar.\u003c/p\u003e\n\u003cp\u003eThe initial multi-database search, conducted for the period up to June 1, 2025, returned 892 records (Scopus: 231, Web of Science: 198, IEEE Xplore: 145, Google Scholar: 318). All records were exported and consolidated using the Zotero reference management software, and duplicates were removed using both automated deduplication and manual inspection, resulting in 485 unique publications. To ensure the review\u0026rsquo;s conceptual focus, methodological rigor, and contemporaneity, the following inclusion and exclusion criteria were applied in a two-stage screening process:\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e2.1.1. Inclusion criteria\u003c/p\u003e\n\u003cp\u003e1. Peer-reviewed journal articles or full-length conference proceedings.\u003c/p\u003e\n\u003cp\u003e2. Primary focus on the application, development, or critical evaluation of AI/ML/DL methods to geospatial data for climate change-related analysis, impact assessment, prediction, or adaptation/mitigation planning.\u003c/p\u003e\n\u003cp\u003e3. Publications dated between January 2015 and June 2025 were included.\u003c/p\u003e\n\u003cp\u003e4. \u0026nbsp;Written in the English language.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e2.1.2. Exclusion criteria\u003c/p\u003e\n\u003cp\u003e1. Articles where AI/ML was mentioned peripherally without substantive methodological application or empirical validation.\u003c/p\u003e\n\u003cp\u003e2. Studies using only traditional statistical methods (e.g., linear regression and GIS overlay) on geospatial data without an AI/ML component.\u003c/p\u003e\n\u003cp\u003e3. Editorials, opinion pieces, book reviews, books, book chapters, and dissertations.\u003c/p\u003e\n\u003cp\u003e4. Studies where the geospatial component was incidental or not integral to the AI methodology.\u003c/p\u003e\n\u003cp\u003eAfter screening the title and abstract of the 485 unique publications against these criteria, 333 records were excluded. The remaining 192 articles underwent a full-text assessment of eligibility. This intensive screening resulted in the exclusion of 40 articles, primarily owing to a lack of substantive AI-geospatial integration or a weak climate change focus. This process yielded a final corpus of 152 articles for quantitative scientometric analysis. The Python programming language (v3.11), with libraries such as Pandas, NumPy, and \u0026ldquo;pybliometrics,\u0026rdquo; was used for managing, cleaning, and performing initial analyses on the bibliographic data. Subsequently, to ensure a thorough and comprehensive coverage for the qualitative systematic review, a forward and backward snowballing technique (Wohlin, 2014) was applied to the reference lists and citing articles of these 152 publications. This iterative process identified an additional 4 highly relevant articles that were missed by the initial database search, resulting in a final expanded corpus of 77 articles for the in-depth systematic review. This expanded set ensures the research field is fully covered while maintaining a manageable scope for qualitative content analysis.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e2.2. Scientometric analysis\u003c/p\u003e\n\u003cp\u003eScientometric analysis is a quantitative methodology for mapping the structural and dynamic evolution of scientific fields by measuring and analyzing patterns in scholarly publications (Ellegaard \u0026amp; Wallin, 2015; Zupic \u0026amp; Čater, 2015). In this review, the bibliographic data of the 152 publications were exported from the databases in standardized RIS and BibTeX formats and harmonized. The primary scientometric analysis was conducted using the Bibliometrix R package (Aria \u0026amp; Cuccurullo, 2017), with custom scripts in Python (v3.11) using the Matplotlib, Seaborn, NetworkX, and \u0026ldquo;scikit-learn\u0026rdquo; libraries for customized visualizations, network analysis, and centrality calculations. The scientometric investigation focused on four key analytical dimensions:\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e1. Annual Publication Trend: To visualize the growth trajectory, identify key inflection points, and assess the maturation and velocity of the research field from 2015 to June 2025.\u003c/p\u003e\n\u003cp\u003e2. Keyword Co-occurrence Analysis: To identify the main research themes, conceptual underpinnings, and their intellectual interrelationships. A minimum keyword frequency threshold of 5 was set to filter out noise and focus on core concepts. Natural Language Processing (NLP) techniques in Python were used for keyword lemmatization and synonym merging (e.g., \u0026ldquo;cnn\u0026rdquo; and \u0026ldquo;convolutional neural network\u0026rdquo;).\u003c/p\u003e\n\u003cp\u003e3. Analysis of Publication Outlets: To identify the core journals and conferences that serve as the primary dissemination channels for knowledge in this domain and to assess their relative impact and interconnectedness.\u003c/p\u003e\n\u003cp\u003e4. Mapping of Contributing Countries and Institutions: To reveal global and institutional research leadership, collaboration networks, and the geographical distribution of scientific production using author affiliation data.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e2.3. Systematic analysis\u003c/p\u003e\n\u003cp\u003eThe systematic analysis involved a qualitative, in-depth, and critical examination of the full corpus of 77 articles. The objective was to synthesize empirical findings, evaluate methodological rigor and innovation, and identify convergent themes, challenges, and knowledge gaps. The analysis followed a structured, iterative coding framework developed through inductive (data-driven) and deductive (theory-driven) approaches, consistent with established qualitative content analysis methodologies (Elo \u0026amp; Kyng\u0026auml;s, 2008; Hsieh \u0026amp; Shannon, 2005). Using Python for data management and the qualitative data analysis software NVivo for coding, the articles were categorized and analyzed across five interconnected analytical domains:\u003c/p\u003e\n\u003cp\u003e1. An Overview of GeoAI Systems and Architectures: This domain involves classifying and critiquing the dominant AI/ML paradigms (e.g., supervised, unsupervised, and self-supervised) and specific architectural families (e.g., CNNs, RNNs, GANs, and Transformers) used in geospatial climate studies.\u003c/p\u003e\n\u003cp\u003e2. The Integration of GeoAI in Climate Change Impact Assessment: This domain deals with examining the specific mechanisms, data fusion strategies, and methodological pipelines through which GeoAI models are applied to distinct climate-related problems, from detection to attribution and projection.\u003c/p\u003e\n\u003cp\u003e3. Technological and Methodological Attributes: This domain delves into analyzing data sources (e.g., MODIS, Sentinel, and CMIP6), feature engineering practices, model validation and evaluation metrics (e.g., RMSE, IoU, F1-score), hyperparameter tuning strategies, and the prevalence of specific software tools and libraries (e.g., TensorFlow, PyTorch, Scikit-learn, and GDAL).\u003c/p\u003e\n\u003cp\u003e4. GeoAI-Climate Application Domains: This domain thematically categorizes studies into major application areas such as cryosphere and hydrosphere monitoring, terrestrial and marine ecosystem dynamics, agricultural security and food systems, urban climate resilience, and human health impacts.\u003c/p\u003e\n\u003cp\u003e5. Challenges, Limitations, and Ethical Considerations: This domain synthesizes reported and inferred impediments related to data quality and availability, computational complexity, model interpretability and explainability, generalizability and transferability, reproducibility, and profound ethical concerns including algorithmic bias and socio-environmental justice.\u003c/p\u003e\n\u003cp\u003eThis structured, multi-faceted content analysis helped identify convergent themes, methodological trends, technological bottlenecks, and critical research gaps, which form the evidentiary basis for the recommendations presented in Section 5.\u003c/p\u003e"},{"header":"3. Result of the scientometric analysis","content":"\u003cp\u003e3.1. Annual publication trend\u003c/p\u003e\n\u003cp\u003eThe annual publication trend (Fig. 3) indicates the rapid emergence and subsequent advancement of the GeoAI-climate change field as a distinct and vibrant research domain. The period from 2015 to 2017 can be considered a nascent phase, with a modest output of 2 to 4 publications per year. This phase coincided with the initial exploration of basic ML models (e.g., Random Forests and Support Vector Machines) on geospatial data for climate applications, and the early adoption of DL in computer vision began to influence remote sensing analysis (Ball et al., 2017; L\u0026auml;ngkvist et al., 2016). A notable inflection point occurred around 2018 to 2019, with annual publications jumping to 8 and then 12. This surge aligns with the widespread maturation and accessibility of DL frameworks (e.g., TensorFlow and PyTorch). Additionally, it shows increased availability of curated geospatial datasets (e.g., Earth Engine catalog) and a growing recognition within the climate science community of the value of data-driven approaches for addressing problems intractable to purely physical models (Reichstein et al., 2019b).\u003c/p\u003e\n\u003cp\u003eA steady upward trajectory is observed from 2020 onwards, peaking at 22 publications in 2024. Projections based on the first half of 2025 suggest this trend will continue, with an estimated 25+ publications for the entire year. This exponential growth pattern (R\u0026sup2; \u0026gt; 0.95 for a polynomial fit) indicates a rapidly consolidating research area attracting increasing scholarly attention. The growth is driven by a confluence of pressing global climate concerns, continuous advancements in AI algorithms and computing hardware (e.g., GPUs and TPUs), and the unprecedented flow of Earth observation data from public and private satellite constellations.\u003c/p\u003e\n\u003cp\u003e3.2. Keyword co-occurrence\u003c/p\u003e\n\u003cp\u003eThe keyword co-occurrence network, generated using a minimum frequency of 5, yielded 72 keywords forming 4 distinct and semantically coherent clusters (Fig. 4). The network\u0026rsquo;s overall density and the strength of linkages indicate a well-connected, interdisciplinary field.\u003c/p\u003e\n\u003cul\u003e\n \u003cli\u003eCluster 1 (Red - AI/ML Methodological Core): This cluster forms the methodological backbone of the field, centered around core AI concepts such as \u0026ldquo;deep learning,\u0026rdquo; \u0026ldquo;machine learning,\u0026rdquo; \u0026ldquo;artificial intelligence,\u0026rdquo; \u0026ldquo;convolutional neural networks (CNN),\u0026rdquo; and \u0026ldquo;long short-term memory (LSTM).\u0026rdquo; It also includes related terms such as \u0026ldquo;feature extraction\u0026rdquo; and \u0026ldquo;predictive modelling,\u0026rdquo; highlighting the focus on automated pattern recognition and forecasting.\u003c/li\u003e\n \u003cli\u003eCluster 2 (Green - Remote Sensing \u0026amp; Geospatial Data): This cluster encompasses the critical data sources and platforms that enable GeoAI, including \u0026ldquo;remote sensing,\u0026rdquo; \u0026ldquo;satellite imagery,\u0026rdquo; \u0026ldquo;Landsat,\u0026rdquo; \u0026ldquo;Sentinel-2,\u0026rdquo; \u0026ldquo;Synthetic Aperture Radar (SAR),\u0026rdquo; \u0026ldquo;Earth observation,\u0026rdquo; and \u0026ldquo;big data.\u0026rdquo; The strong linkages between this cluster and the red AI/ML cluster demonstrate the tight, symbiotic coupling of methodological innovation with data availability and characteristics.\u003c/li\u003e\n \u003cli\u003eCluster 3 (Blue - Specific Climate Impacts \u0026amp; Vulnerabilities): This cluster focuses on the application domains and specific climate change impacts. Key terms include \u0026ldquo;sea level rise,\u0026rdquo; \u0026ldquo;land use and land cover change (LULC),\u0026rdquo; \u0026ldquo;drought monitoring,\u0026rdquo; \u0026ldquo;wildfire risk,\u0026rdquo; \u0026ldquo;biodiversity,\u0026rdquo; \u0026ldquo;agriculture,\u0026rdquo; and \u0026ldquo;climate adaptation.\u0026rdquo; This cluster connects the methods (Red) and data (Green) to tangible environmental and societal challenges.\u003c/li\u003e\n \u003cli\u003eCluster 4 (Yellow - Spatial Analysis \u0026amp; Modeling Context): This cluster contains foundational geospatial concepts and broader modeling contexts, such as \u0026ldquo;geographic information system (GIS),\u0026rdquo; \u0026ldquo;spatial analysis,\u0026rdquo; \u0026ldquo;climate models,\u0026rdquo; \u0026ldquo;risk assessment,\u0026rdquo; \u0026ldquo;vulnerability,\u0026rdquo; and \u0026ldquo;sustainability.\u0026rdquo; It situates the GeoAI work within the larger frameworks of geospatial science and climate risk management.\u003c/li\u003e\n\u003c/ul\u003e\n\u003cp\u003eThe network\u0026rsquo;s core is primarily occupied by \u0026ldquo;geospatial artificial intelligence,\u0026rdquo; \u0026ldquo;climate change,\u0026rdquo; and \u0026ldquo;machine learning,\u0026rdquo; confirming their central role as the field\u0026rsquo;s defining concepts. The high betweenness centrality of \u0026ldquo;climate change\u0026rdquo; indicates it connects the methodological and application-oriented clusters. Table 2, generated using Python\u0026rsquo;s NetworkX library, ranks the top 20 keywords by degree centrality, a measure of a node\u0026rsquo;s connectedness within the network. \u0026ldquo;Machine Learning,\u0026rdquo; \u0026ldquo;Remote Sensing,\u0026rdquo; and \u0026ldquo;Deep Learning\u0026rdquo; exhibit the highest degree centrality, indicating that they are the most influential and interconnected concepts that facilitate knowledge flow across the entire domain.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eTable 2 \u0026nbsp;Top 20 keywords by degree centrality in the GeoAI-Climate Change domain\u003c/p\u003e\n\u003ctable border=\"1\" cellspacing=\"0\" cellpadding=\"0\" width=\"720\"\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 60px;\"\u003e\n \u003cp\u003eRank\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 263px;\"\u003e\n \u003cp\u003eKeyword\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 91px;\"\u003e\n \u003cp\u003eOccurrence\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 132px;\"\u003e\n \u003cp\u003eDegree Centrality\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 174px;\"\u003e\n \u003cp\u003eBetweenness Centrality\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 60px;\"\u003e\n \u003cp\u003e1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 263px;\"\u003e\n \u003cp\u003eMachine Learning\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 91px;\"\u003e\n \u003cp\u003e58\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 132px;\"\u003e\n \u003cp\u003e52\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 174px;\"\u003e\n \u003cp\u003e0.118\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 60px;\"\u003e\n \u003cp\u003e2\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 263px;\"\u003e\n \u003cp\u003eRemote Sensing\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 91px;\"\u003e\n \u003cp\u003e55\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 132px;\"\u003e\n \u003cp\u003e50\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 174px;\"\u003e\n \u003cp\u003e0.132\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 60px;\"\u003e\n \u003cp\u003e3\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 263px;\"\u003e\n \u003cp\u003eClimate Change\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 91px;\"\u003e\n \u003cp\u003e53\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 132px;\"\u003e\n \u003cp\u003e49\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 174px;\"\u003e\n \u003cp\u003e0.211\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 60px;\"\u003e\n \u003cp\u003e4\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 263px;\"\u003e\n \u003cp\u003eDeep Learning\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 91px;\"\u003e\n \u003cp\u003e49\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 132px;\"\u003e\n \u003cp\u003e46\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 174px;\"\u003e\n \u003cp\u003e0.095\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 60px;\"\u003e\n \u003cp\u003e5\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 263px;\"\u003e\n \u003cp\u003eConvolutional Neural Network (CNN)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 91px;\"\u003e\n \u003cp\u003e31\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 132px;\"\u003e\n \u003cp\u003e35\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 174px;\"\u003e\n \u003cp\u003e0.054\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 60px;\"\u003e\n \u003cp\u003e6\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 263px;\"\u003e\n \u003cp\u003eGeographic Information System (GIS)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 91px;\"\u003e\n \u003cp\u003e28\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 132px;\"\u003e\n \u003cp\u003e33\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 174px;\"\u003e\n \u003cp\u003e0.048\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 60px;\"\u003e\n \u003cp\u003e7\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 263px;\"\u003e\n \u003cp\u003eLand Use and Land Cover Change (LULC)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 91px;\"\u003e\n \u003cp\u003e25\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 132px;\"\u003e\n \u003cp\u003e30\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 174px;\"\u003e\n \u003cp\u003e0.041\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 60px;\"\u003e\n \u003cp\u003e8\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 263px;\"\u003e\n \u003cp\u003eSatellite Imagery\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 91px;\"\u003e\n \u003cp\u003e24\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 132px;\"\u003e\n \u003cp\u003e29\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 174px;\"\u003e\n \u003cp\u003e0.037\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 60px;\"\u003e\n \u003cp\u003e9\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 263px;\"\u003e\n \u003cp\u003eLong Short-Term Memory (LSTM)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 91px;\"\u003e\n \u003cp\u003e22\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 132px;\"\u003e\n \u003cp\u003e27\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 174px;\"\u003e\n \u003cp\u003e0.033\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 60px;\"\u003e\n \u003cp\u003e10\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 263px;\"\u003e\n \u003cp\u003eDrought Monitoring\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 91px;\"\u003e\n \u003cp\u003e19\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 132px;\"\u003e\n \u003cp\u003e25\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 174px;\"\u003e\n \u003cp\u003e0.029\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 60px;\"\u003e\n \u003cp\u003e11\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 263px;\"\u003e\n \u003cp\u003eSea Level Rise\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 91px;\"\u003e\n \u003cp\u003e18\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 132px;\"\u003e\n \u003cp\u003e24\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 174px;\"\u003e\n \u003cp\u003e0.031\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 60px;\"\u003e\n \u003cp\u003e12\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 263px;\"\u003e\n \u003cp\u003eSpatial Analysis\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 91px;\"\u003e\n \u003cp\u003e17\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 132px;\"\u003e\n \u003cp\u003e23\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 174px;\"\u003e\n \u003cp\u003e0.025\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 60px;\"\u003e\n \u003cp\u003e13\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 263px;\"\u003e\n \u003cp\u003eBig Data\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 91px;\"\u003e\n \u003cp\u003e16\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 132px;\"\u003e\n \u003cp\u003e22\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 174px;\"\u003e\n \u003cp\u003e0.022\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 60px;\"\u003e\n \u003cp\u003e14\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 263px;\"\u003e\n \u003cp\u003eRisk Assessment\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 91px;\"\u003e\n \u003cp\u003e15\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 132px;\"\u003e\n \u003cp\u003e21\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 174px;\"\u003e\n \u003cp\u003e0.028\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 60px;\"\u003e\n \u003cp\u003e15\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 263px;\"\u003e\n \u003cp\u003eSentinel-2\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 91px;\"\u003e\n \u003cp\u003e14\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 132px;\"\u003e\n \u003cp\u003e20\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 174px;\"\u003e\n \u003cp\u003e0.019\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 60px;\"\u003e\n \u003cp\u003e16\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 263px;\"\u003e\n \u003cp\u003eBiodiversity\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 91px;\"\u003e\n \u003cp\u003e13\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 132px;\"\u003e\n \u003cp\u003e19\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 174px;\"\u003e\n \u003cp\u003e0.020\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 60px;\"\u003e\n \u003cp\u003e17\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 263px;\"\u003e\n \u003cp\u003eWildfire\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 91px;\"\u003e\n \u003cp\u003e13\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 132px;\"\u003e\n \u003cp\u003e18\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 174px;\"\u003e\n \u003cp\u003e0.017\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 60px;\"\u003e\n \u003cp\u003e18\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 263px;\"\u003e\n \u003cp\u003eClimate Adaptation\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 91px;\"\u003e\n \u003cp\u003e12\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 132px;\"\u003e\n \u003cp\u003e17\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 174px;\"\u003e\n \u003cp\u003e0.023\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 60px;\"\u003e\n \u003cp\u003e19\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 263px;\"\u003e\n \u003cp\u003eSynthetic Aperture Radar (SAR)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 91px;\"\u003e\n \u003cp\u003e11\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 132px;\"\u003e\n \u003cp\u003e16\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 174px;\"\u003e\n \u003cp\u003e0.015\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 60px;\"\u003e\n \u003cp\u003e20\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 263px;\"\u003e\n \u003cp\u003eExplainable AI (XAI)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 91px;\"\u003e\n \u003cp\u003e10\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 132px;\"\u003e\n \u003cp\u003e15\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 174px;\"\u003e\n \u003cp\u003e0.026\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n\u003c/table\u003e\n\u003cp\u003e3.3. Publication outlets\u003c/p\u003e\n\u003cp\u003eThe analysis of publication outlets identified the primary channels for scholarly communication and knowledge dissemination in this field. The threshold was set to sources with at least 2 publications from the core corpus, resulting in 21 key outlets. A network map (Fig. 5) shows the co-citation relationships between these journals, revealing a dense web of intellectual exchange. Regarding publication volume, the most prominent venues are \u0026ldquo;Remote Sensing of Environment,\u0026rdquo; \u0026ldquo;ISPRS Journal of Photogrammetry and Remote Sensing,\u0026rdquo; \u0026ldquo;International Journal of Applied Earth Observation and Geoinformation,\u0026rdquo; and \u0026ldquo;IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing.\u0026rdquo; This concentration in premier remote sensing and geoinformation journals indicates that the field is currently anchored within the geospatial science and engineering communities. However, a significant and growing presence is noted in high-impact, broad-scope journals such as \u0026ldquo;Nature Climate Change,\u0026rdquo; \u0026ldquo;Science Advances,\u0026rdquo; and \u0026ldquo;Proceedings of the National Academy of Sciences,\u0026rdquo; suggesting the increasing recognition of GeoAI\u0026rsquo;s transformative potential by the wider scientific community. Furthermore, computer science conferences such as \u0026ldquo;NeurIPS,\u0026rdquo; \u0026ldquo;ICML,\u0026rdquo; and \u0026ldquo;CVPR\u0026rdquo; are emerging as crucial venues for presenting foundational algorithmic advances with geospatial applications. Table 3 ranks the top 10 outlets based on the average normalized citation count, a metric that corrects for the age and size of the journal, providing a measure of impact per paper. \u0026ldquo;Nature Climate Change\u0026rdquo; and \u0026ldquo;Remote Sensing of Environment\u0026rdquo; lead in this regard, highlighting the high impact of the research published therein.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eTable 3 \u0026nbsp;Top 10 publication outlets in the GeoAI-Climate Change domain by average normalized citations\u003c/p\u003e\n\u003ctable border=\"1\" cellspacing=\"0\" cellpadding=\"0\" width=\"686\"\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 226px;\"\u003e\n \u003cp\u003ePublication Outlet\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 84px;\"\u003e\n \u003cp\u003eTotal Link Strength (TLS)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 101px;\"\u003e\n \u003cp\u003eDocuments\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 78px;\"\u003e\n \u003cp\u003eTotal Citations\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 95px;\"\u003e\n \u003cp\u003eNormalized Citations\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 101px;\"\u003e\n \u003cp\u003eAvg. Normalized Citations\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 226px;\"\u003e\n \u003cp\u003eNature Climate Change\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 84px;\"\u003e\n \u003cp\u003e4\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 101px;\"\u003e\n \u003cp\u003e3\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 78px;\"\u003e\n \u003cp\u003e185\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 95px;\"\u003e\n \u003cp\u003e28.45\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 101px;\"\u003e\n \u003cp\u003e9.48\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 226px;\"\u003e\n \u003cp\u003eRemote Sensing of Environment\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 84px;\"\u003e\n \u003cp\u003e22\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 101px;\"\u003e\n \u003cp\u003e9\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 78px;\"\u003e\n \u003cp\u003e412\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 95px;\"\u003e\n \u003cp\u003e25.18\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 101px;\"\u003e\n \u003cp\u003e2.80\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 226px;\"\u003e\n \u003cp\u003eScience Advances\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 84px;\"\u003e\n \u003cp\u003e3\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 101px;\"\u003e\n \u003cp\u003e2\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 78px;\"\u003e\n \u003cp\u003e97\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 95px;\"\u003e\n \u003cp\u003e15.11\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 101px;\"\u003e\n \u003cp\u003e7.56\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 226px;\"\u003e\n \u003cp\u003eISPRS Journal of Photogrammetry and Remote Sensing\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 84px;\"\u003e\n \u003cp\u003e18\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 101px;\"\u003e\n \u003cp\u003e7\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 78px;\"\u003e\n \u003cp\u003e201\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 95px;\"\u003e\n \u003cp\u003e12.33\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 101px;\"\u003e\n \u003cp\u003e1.76\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 226px;\"\u003e\n \u003cp\u003eInternational Journal of Applied Earth Observation and Geoinformation\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 84px;\"\u003e\n \u003cp\u003e15\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 101px;\"\u003e\n \u003cp\u003e6\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 78px;\"\u003e\n \u003cp\u003e134\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 95px;\"\u003e\n \u003cp\u003e8.91\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 101px;\"\u003e\n \u003cp\u003e1.49\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 226px;\"\u003e\n \u003cp\u003eIEEE Transactions on Geoscience and Remote Sensing\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 84px;\"\u003e\n \u003cp\u003e12\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 101px;\"\u003e\n \u003cp\u003e5\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 78px;\"\u003e\n \u003cp\u003e118\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 95px;\"\u003e\n \u003cp\u003e7.25\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 101px;\"\u003e\n \u003cp\u003e1.45\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 226px;\"\u003e\n \u003cp\u003eGlobal Change Biology\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 84px;\"\u003e\n \u003cp\u003e5\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 101px;\"\u003e\n \u003cp\u003e2\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 78px;\"\u003e\n \u003cp\u003e68\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 95px;\"\u003e\n \u003cp\u003e6.89\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 101px;\"\u003e\n \u003cp\u003e3.45\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 226px;\"\u003e\n \u003cp\u003eEnvironmental Research Letters\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 84px;\"\u003e\n \u003cp\u003e8\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 101px;\"\u003e\n \u003cp\u003e3\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 78px;\"\u003e\n \u003cp\u003e75\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 95px;\"\u003e\n \u003cp\u003e5.12\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 101px;\"\u003e\n \u003cp\u003e1.71\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 226px;\"\u003e\n \u003cp\u003eIEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 84px;\"\u003e\n \u003cp\u003e10\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 101px;\"\u003e\n \u003cp\u003e4\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 78px;\"\u003e\n \u003cp\u003e81\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 95px;\"\u003e\n \u003cp\u003e4.95\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 101px;\"\u003e\n \u003cp\u003e1.24\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 226px;\"\u003e\n \u003cp\u003eClimatic Change\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 84px;\"\u003e\n \u003cp\u003e6\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 101px;\"\u003e\n \u003cp\u003e2\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 78px;\"\u003e\n \u003cp\u003e52\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 95px;\"\u003e\n \u003cp\u003e4.87\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 101px;\"\u003e\n \u003cp\u003e2.44\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n\u003c/table\u003e\n\u003cp\u003e3.4. Top contributing countries\u003c/p\u003e\n\u003cp\u003eMapping the geographical distribution of research output (Fig. 6) reveals a clear but evolving global leadership structure. The minimum thresholds were set at 2 documents and 10 citations per country. Among the 45 countries represented in the corpus, 18 met these criteria. Based on corresponding author and co-author affiliations, the analysis shows that the United States and China are the dominant contributors, both in terms of raw publication volume and Total Link Strength (TLS), a measure of international co-authorship. This indicates high domestic productivity and their extensive and central roles in global collaborative networks. The United Kingdom, Germany, Switzerland, and the Netherlands also demonstrate strong and influential contributions, often acting as bridges in the collaboration network. Canada and Australia are other significant nodes of the Anglosphere. Notably, emerging economies such as India and Brazil appear as meaningful contributors, reflecting a growing but uneven global engagement with GeoAI for addressing regional and global climate challenges. Table 4 presents a detailed ranking, highlighting the USA\u0026rsquo;s central role in the collaborative network and the high impact of Swiss research, as indicated by its high average normalized citation count.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eTable 4 \u0026nbsp;Top contributing countries in GeoAI-Climate change research\u003c/p\u003e\n\u003ctable border=\"1\" cellspacing=\"0\" cellpadding=\"0\" width=\"696\" class=\"fr-table-selection-hover\"\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 113px;\"\u003e\n \u003cp\u003eCountry\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 132px;\"\u003e\n \u003cp\u003eTLS\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 94px;\"\u003e\n \u003cp\u003eDocuments\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 113px;\"\u003e\n \u003cp\u003eTotal Citations\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 113px;\"\u003e\n \u003cp\u003eNormalized Citations\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 130px;\"\u003e\n \u003cp\u003eAvg. Normalized Citations\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 113px;\"\u003e\n \u003cp\u003eUnited States\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 132px;\"\u003e\n \u003cp\u003e105\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 94px;\"\u003e\n \u003cp\u003e41\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 113px;\"\u003e\n \u003cp\u003e1456\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 113px;\"\u003e\n \u003cp\u003e58.91\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 130px;\"\u003e\n \u003cp\u003e1.44\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 113px;\"\u003e\n \u003cp\u003eChina\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 132px;\"\u003e\n \u003cp\u003e88\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 94px;\"\u003e\n \u003cp\u003e38\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 113px;\"\u003e\n \u003cp\u003e987\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 113px;\"\u003e\n \u003cp\u003e42.15\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 130px;\"\u003e\n \u003cp\u003e1.11\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 113px;\"\u003e\n \u003cp\u003eUnited Kingdom\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 132px;\"\u003e\n \u003cp\u003e45\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 94px;\"\u003e\n \u003cp\u003e18\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 113px;\"\u003e\n \u003cp\u003e445\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 113px;\"\u003e\n \u003cp\u003e19.87\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 130px;\"\u003e\n \u003cp\u003e1.10\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 113px;\"\u003e\n \u003cp\u003eGermany\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 132px;\"\u003e\n \u003cp\u003e38\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 94px;\"\u003e\n \u003cp\u003e15\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 113px;\"\u003e\n \u003cp\u003e388\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 113px;\"\u003e\n \u003cp\u003e16.54\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 130px;\"\u003e\n \u003cp\u003e1.10\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 113px;\"\u003e\n \u003cp\u003eCanada\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 132px;\"\u003e\n \u003cp\u003e32\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 94px;\"\u003e\n \u003cp\u003e12\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 113px;\"\u003e\n \u003cp\u003e321\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 113px;\"\u003e\n \u003cp\u003e14.12\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 130px;\"\u003e\n \u003cp\u003e1.18\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 113px;\"\u003e\n \u003cp\u003eAustralia\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 132px;\"\u003e\n \u003cp\u003e28\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 94px;\"\u003e\n \u003cp\u003e11\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 113px;\"\u003e\n \u003cp\u003e278\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 113px;\"\u003e\n \u003cp\u003e11.95\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 130px;\"\u003e\n \u003cp\u003e1.09\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 113px;\"\u003e\n \u003cp\u003eSwitzerland\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 132px;\"\u003e\n \u003cp\u003e25\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 94px;\"\u003e\n \u003cp\u003e8\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 113px;\"\u003e\n \u003cp\u003e305\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 113px;\"\u003e\n \u003cp\u003e13.88\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 130px;\"\u003e\n \u003cp\u003e1.74\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 113px;\"\u003e\n \u003cp\u003eNetherlands\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 132px;\"\u003e\n \u003cp\u003e22\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 94px;\"\u003e\n \u003cp\u003e9\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 113px;\"\u003e\n \u003cp\u003e234\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 113px;\"\u003e\n \u003cp\u003e10.01\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 130px;\"\u003e\n \u003cp\u003e1.11\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 113px;\"\u003e\n \u003cp\u003eFrance\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 132px;\"\u003e\n \u003cp\u003e20\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 94px;\"\u003e\n \u003cp\u003e7\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 113px;\"\u003e\n \u003cp\u003e198\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 113px;\"\u003e\n \u003cp\u003e8.45\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 130px;\"\u003e\n \u003cp\u003e1.21\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 113px;\"\u003e\n \u003cp\u003eIndia\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 132px;\"\u003e\n \u003cp\u003e18\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 94px;\"\u003e\n \u003cp\u003e6\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 113px;\"\u003e\n \u003cp\u003e123\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 113px;\"\u003e\n \u003cp\u003e5.28\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 130px;\"\u003e\n \u003cp\u003e0.88\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 113px;\"\u003e\n \u003cp\u003eBrazil\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 132px;\"\u003e\n \u003cp\u003e15\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 94px;\"\u003e\n \u003cp\u003e5\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 113px;\"\u003e\n \u003cp\u003e115\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 113px;\"\u003e\n \u003cp\u003e4.91\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 130px;\"\u003e\n \u003cp\u003e0.98\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 113px;\"\u003e\n \u003cp\u003eItaly\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 132px;\"\u003e\n \u003cp\u003e14\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 94px;\"\u003e\n \u003cp\u003e5\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 113px;\"\u003e\n \u003cp\u003e104\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 113px;\"\u003e\n \u003cp\u003e4.44\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 130px;\"\u003e\n \u003cp\u003e0.89\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 113px;\"\u003e\n \u003cp\u003eSpain\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 132px;\"\u003e\n \u003cp\u003e12\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 94px;\"\u003e\n \u003cp\u003e4\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 113px;\"\u003e\n \u003cp\u003e97\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 113px;\"\u003e\n \u003cp\u003e4.15\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 130px;\"\u003e\n \u003cp\u003e1.04\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 113px;\"\u003e\n \u003cp\u003eJapan\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 132px;\"\u003e\n \u003cp\u003e10\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 94px;\"\u003e\n \u003cp\u003e3\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 113px;\"\u003e\n \u003cp\u003e85\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 113px;\"\u003e\n \u003cp\u003e3.63\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 130px;\"\u003e\n \u003cp\u003e1.21\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 113px;\"\u003e\n \u003cp\u003eSouth Korea\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 132px;\"\u003e\n \u003cp\u003e9\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 94px;\"\u003e\n \u003cp\u003e3\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 113px;\"\u003e\n \u003cp\u003e78\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 113px;\"\u003e\n \u003cp\u003e3.33\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 130px;\"\u003e\n \u003cp\u003e1.11\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 113px;\"\u003e\n \u003cp\u003eSweden\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 132px;\"\u003e\n \u003cp\u003e8\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 94px;\"\u003e\n \u003cp\u003e3\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 113px;\"\u003e\n \u003cp\u003e71\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 113px;\"\u003e\n \u003cp\u003e3.03\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 130px;\"\u003e\n \u003cp\u003e1.01\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 113px;\"\u003e\n \u003cp\u003eNorway\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 132px;\"\u003e\n \u003cp\u003e7\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 94px;\"\u003e\n \u003cp\u003e2\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 113px;\"\u003e\n \u003cp\u003e65\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 113px;\"\u003e\n \u003cp\u003e2.78\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 130px;\"\u003e\n \u003cp\u003e1.39\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 113px;\"\u003e\n \u003cp\u003eAustria\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 132px;\"\u003e\n \u003cp\u003e6\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 94px;\"\u003e\n \u003cp\u003e2\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 113px;\"\u003e\n \u003cp\u003e58\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 113px;\"\u003e\n \u003cp\u003e2.48\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 130px;\"\u003e\n \u003cp\u003e1.24\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n\u003c/table\u003e\n\u003cp\u003e\u0026nbsp;\u003c/p\u003e"},{"header":"4. Result of the systematic analysis","content":"\u003cp\u003e4.1. GeoAI systems and architectures in climate science\u003c/p\u003e\n\u003cp\u003eGeoAI systems involve diverse and evolving algorithms and computational architectures designed to process, learn from, and reason about spatiotemporal data. Based on the systematic review of 77 articles, these can be categorized into several dominant but often overlapping paradigms with distinct strengths and application contexts in climate science.\u003c/p\u003e\n\u003cp\u003e4.1.1. CNNs and spatial feature learning\u003c/p\u003e\n\u003cp\u003eCNNs have become the foundational models for analyzing spatially explicit data, particularly raster-based satellite and aerial imagery (Lecun et al., 2015; Zhu et al., 2017). Through the use of convolutional filters and pooling operations, their intrinsic inductive bias allows them to automatically learn hierarchical spatial features from low-level edges and textures to high-level object representations without manual feature engineering (Krizhevsky et al., 2017). This makes them powerful for tasks such as land use and land cover change (LULC) classification (Y. Liu et al., 2018; E. Zhang et al., 2021; Z. Zhang et al., 2018), glacier and sea ice mapping (Gavahi et al., 2023; R. Wang et al., 2024), urban heat island analysis \u0026nbsp;(Aviles Toledo et al., 2024), and building footprint extraction for exposure modeling (Fayaz et al., 2024; Schneider et al., 2017b). Fully Convolutional Network (FCN) architectures, particularly the U-Net family (Ronneberger et al., 2015), are widely used for semantic segmentation of environmental features, achieving state-of-the-art accuracy in delineating complex boundaries, such as coastlines, burn scars, and agricultural fields (Iglovikov et al., 2018; L\u0026auml;ngkvist et al., 2016). The application of CNNs has considerably transcended standard RGB imagery to multi-spectral, hyper-spectral, and SAR data. This has enabled robust monitoring of vegetation health (e.g., through NDVI), soil moisture, flood extents, and ocean surface winds under varying atmospheric and climatic conditions (Lacoste et al., 2021; Ma et al., 2019b).\u003c/p\u003e\n\u003cp\u003e4.1.2. RNNs and temporal dynamics modeling\u003c/p\u003e\n\u003cp\u003eTo analyze time-series climate data, RNNs and their more advanced variants, LSTM (Hochreiter \u0026amp; Schmidhuber, 1997)\u0026nbsp;and GRU networks (Cho et al., 2014), are essential. These architectures are designed to handle sequential data by maintaining an internal state or \u0026ldquo;memory\u0026rdquo; of previous inputs (Yu et al., 2019). LSTMs particularly address the vanishing/exploding gradient problem of vanilla RNNs, allowing them to capture long-term dependencies in temporal sequences, which is crucial for climate processes (Shi et al., 2015). They are used for forecasting essential climate variables such as precipitation, temperature, evapotranspiration, and streamflow (Fang et al., 2017; Lees et al., 2021). LSTM networks learn from historical time-series data from ground stations, gridded reanalysis products (e.g., ERA5), or satellite-derived data cubes. Consequently, LSTM networks can model non-linear and chaotic dynamics that elude traditional statistical methods (ARIMA) or simplified physical models, providing early warning systems for droughts, floods, and heatwaves (Frame et al., 2022; Xiang et al., 2020).\u003c/p\u003e\n\u003cp\u003e4.1.3. Transformer-Based models and attention mechanisms\u003c/p\u003e\n\u003cp\u003eInspired by their revolutionary success in NLP (Vaswani et al., 2023), transformer architectures are now being rapidly adapted for geospatial tasks (Irvin et al., 2020; Reed et al., 2023). Their core innovation is the self-attention mechanism, which allows the model to dynamically weigh the importance of all other elements in a sequence (or image patches) when processing a particular element. Therefore, they capture global context more effectively than CNNs with local receptive fields (Dosovitskiy et al., 2021). Vision Transformers (ViTs) are being explored for image-based tasks such as scene classification and change detection, potentially outperforming CNNs on extremely large datasets (Tuia et al., 2022; R. Wang et al., 2024). Furthermore, transformers are being developed for spatiotemporal forecasting, showing promise in modeling complex, large-scale climate phenomena such as El Ni\u0026ntilde;o-Southern Oscillation (ENSO) by capturing long-range teleconnections across the Pacific Ocean (Ham et al., 2019; Kurth et al., 2023). The flexibility of transformers to handle multi-modal data (e.g., imagery, point data, and text) also positions them as a key architecture for future integrated assessment models.\u003c/p\u003e\n\u003cp\u003e4.1.4. Ensemble, Hybrid, and Physics-Informed Models\u003c/p\u003e\n\u003cp\u003eEnsemble methods, such as Random Forests (Breiman, 2001) and Gradient Boosting Machines (e.g., XGBoost) (Chen and Guestrin, 2016), LightGBM (Ke et al., 2017), are popular and effective for tabular geospatial data that integrates satellite-derived indices, climate reanalysis data, topographic variables, and socio-economic factors (Lacoste et al., 2021; Maxwell et al., 2018). They are robust to overfitting, provide native feature importance scores, and provide strong baseline performance for tasks such as species distribution modeling (Valavi et al., 2022), climate risk vulnerability mapping (Alnajjar et al., 2025), and social vulnerability assessment (Tate et al., 2021). Hybrid models that combine the strengths of different architectures are common. For instance, CNN\u0026ndash;LSTM models use CNNs to extract spatial features from each timestep of a satellite image time series and LSTMs to model the temporal evolution of these features. This is highly effective for forecasting crop yield (R. Wang et al., 2024; You et al., 2017), soil erosion (C. Yang et al., 2017), and chlorophyll-a concentration in water bodies (Pahlevan et al., 2020). A major and promising frontier is Physics-Informed Neural Networks (PINNs) (Raissi et al., 2019), which embed physical laws (e.g., conservation of energy and Navier-Stokes equations) directly into the loss function or architecture of the neural network. This constrains the solution space to be physically plausible, which is particularly valuable for forecasting in data-sparse regions (Kashinath et al., 2021; Read et al., 2019).\u003c/p\u003e\n\u003cp\u003e4.2. GeoAI in climate change impact assessment\u003c/p\u003e\n\u003cp\u003eThe application of GeoAI is revolutionizing the methodological approaches and empirical precision with which climate change impacts across the Earth\u0026rsquo;s systems are assessed.\u003c/p\u003e\n\u003cp\u003e4.2.1. Cryosphere and hydrosphere monitoring\u003c/p\u003e\n\u003cp\u003eGeoAI is proving critical for tracking and projecting changes in the Earth\u0026rsquo;s frozen regions (cryosphere) and water cycles (hydrosphere). CNNs, specifically U-Net and DeepLab, automate the mapping of glacier retreat, calving fronts, and supraglacial lakes from multi-spectral (e.g., Landsat 8 and Sentinel-2) and SAR (e.g., Sentinel-1) imagery with sub-pixel precision. This provides essential, high-frequency data for constraining sea-level rise projections (Heidler et al., 2022; L. Zhang et al., 2016). Similarly, AI models are used to classify sea ice types, concentration, and motion from SAR data, improving maritime navigation safety and the parameterization of sea ice in climate models (Khaki et al., 2020; Y. Liu et al., 2018; R. Wang et al., 2024). In hydrology, LSTMs and CNN\u0026ndash;LSTMs are deployed for predicting river discharge, forecasting flash floods, and modeling groundwater levels, effectively integrating complex rainfall-runoff processes with real-time satellite precipitation data (e.g., GPM) and soil moisture information (Fang et al., 2017; Frame et al., 2022; Kratzert et al., 2023).\u003c/p\u003e\n\u003cp\u003e4.2.2. Terrestrial and marine ecosystems\u003c/p\u003e\n\u003cp\u003eClimate-induced shifts in biodiversity, phenology, and ecosystem functioning are being monitored and projected with unprecedented detail using GeoAI. Species Distribution Models (SDMs) have evolved from using traditional logistic regression and MaxEnt to employing ensemble ML methods and CNNs that can incorporate raw remote sensing imagery directly as input. They learn imperceptible habitat features, leading to more accurate and mechanistically informed predictions of habitat suitability under future climate scenarios (Botella et al., 2018; Harris et al., 2022; Valavi et al., 2022). In marine ecology, AI models, including CNNs for image analysis and RNNs for time-series prediction, analyze satellite-derived ocean color data (e.g., from MODIS and VIIRS) to monitor harmful algal blooms, detect coral bleaching events, and track changes in phytoplankton biomass and community structure. These are critical indicators of ocean health and are highly sensitive to ocean warming, acidification, and deoxygenation (Free et al., 2021; Khaki et al., 2020; F. Li et al., 2024).\u003c/p\u003e\n\u003cp\u003e4.2.3. Agriculture and food security\u003c/p\u003e\n\u003cp\u003eEnsuring global food security under a changing climate is a prime application area for GeoAI. ML models, particularly XGBoost and Random Forests, integrate multi-source data including climate projections (CMIP6), soil maps, and high-resolution satellite imagery (e.g., Sentinel-2). This is to predict crop yields at regional to global scales, identify croplands susceptible to drought and heat stress, and recommend optimal planting dates and irrigation schedules (Jin et al., 2019; R. Wang et al., 2024). For instance, CNNs can detect early signs of crop disease and pest infestation exacerbated by changing weather patterns from UAV and satellite imagery, enabling targeted and timely interventions that reduce yield loss and pesticide overuse (Su et al., 2022; L. Zhang et al., 2016).\u003c/p\u003e\n\u003cp\u003e4.2.4. Urban environments, infrastructure, and human health\u003c/p\u003e\n\u003cp\u003eCities are hotspots of climate risk and vulnerability. GeoAI helps assess urban exposure and vulnerability by combining very-high-resolution imagery, LiDAR data, land surface temperature (LST) retrievals, and demographic census information. This is to map heat stress risk, identify neighborhoods with poor green space access, and model flood inundation under different precipitation scenarios (F. Li et al., 2024; Schneider et al., 2017b; Stark, 2018). Furthermore, AI models are used to analyze the cooling effect of urban green infrastructure and the impact of building morphology on energy consumption, informing urban planning and climate adaptation strategies (Bi et al., 2023; W. Li, 2020b; Tehrani et al., 2024). In public health, GeoAI models the climate-driven spatial and temporal dynamics of vector-borne diseases (e.g., malaria and dengue), correlating environmental variables such as temperature, precipitation, and vegetation with disease incidence data to create early warning systems (Lowe et al., 2021; Messina et al., 2019).\u003c/p\u003e\n\u003cp\u003e4.3. GeoAI-Climate application domains\u003c/p\u003e\n\u003cp\u003eThe systematic review allowed for a fine-grained, thematic classification of studies into two overarching domains: monitoring and detection of ongoing impacts and the forecasting and projection of future risks.\u003c/p\u003e\n\u003cp\u003e4.3.1. Climate impact monitoring and detection\u003c/p\u003e\n\u003cp\u003eThis domain focuses on observing, quantifying, and attributing ongoing changes, constituting the majority of the reviewed studies. It leverages GeoAI for automated, large-scale, and high-frequency analysis.\u003c/p\u003e\n\u003cul class=\"decimal_type\"\u003e\n \u003cli\u003eDeforestation and Forest Degradation Detection: CNNs and change detection algorithms (e.g., those using Siamese networks) are used to identify illegal logging, forest degradation, and fire scars in near-real-time from dense satellite image time series (e.g., with PlanetScope data). This enables rapid response and enforcement (Bullock et al., 2020; L\u0026auml;ngkvist et al., 2016; Schwartz et al., 2020).\u003c/li\u003e\n \u003cli\u003eWildfire Burn Scar Mapping and Severity Assessment: AI models, particularly semantic segmentation networks, rapidly and accurately delineate the spatial extent and severity of wildfires from post-fire satellite imagery, providing critical data for carbon emission estimates, ecosystem recovery monitoring, and disaster response (Andela et al., 2017; Ban et al., 2020).\u003c/li\u003e\n \u003cli\u003eCoastal Erosion, Inundation, and Wetland Loss: ML models predict shoreline changes and map saltwater intrusion and coastal flood risk from sea-level rise and storm surges, often using data fusion techniques that combine optical satellite imagery, SAR, LiDAR, and topographic data (Mentaschi et al., 2018; Vos et al., 2022).\u003c/li\u003e\n\u003c/ul\u003e\n\u003cp\u003e4.3.2. Climate impact forecasting and projection\u003c/p\u003e\n\u003cp\u003eThis domain uses GeoAI to predict future states and probabilities, which is more challenging but crucial for proactive adaptation planning and policy formulation.\u003c/p\u003e\n\u003cul class=\"decimal_type\"\u003e\n \u003cli\u003eExtreme Weather Prediction and Attribution: DL models, including transformers and graph neural networks (GNNs), are being developed to improve the sub-seasonal to seasonal forecasting of extreme events such as tropical cyclones, atmospheric rivers, and heatwaves. They sometimes outperform traditional numerical weather prediction models in specific contexts (Bi et al., 2023; Kurth et al., 2023; Rasp et al., 2018).\u003c/li\u003e\n \u003cli\u003eLong-term Climate Projection Downscaling: AI methods, particularly deep neural networks and generative adversarial networks (GANs), are used to \u0026ldquo;downscale\u0026rdquo; coarse-resolution outputs from Global Climate Models (GCMs) to higher, more decision-relevant resolutions (e.g., 1 km). They provide localized projections of temperature, precipitation, and other variables for regional impact assessments (Harris et al., 2022; Nowack et al., 2020; Vandal et al., 2017).\u003c/li\u003e\n \u003cli\u003eEcosystem Service Projection and Tipping Points: Ensemble models and process-guided AI project future changes in carbon sequestration, water purification, and other ecosystem services under different climate and land-use scenarios (Y. Liu et al., 2018; Runting et al., 2020; Tehrani et al., 2024). Furthermore, AI is being explored to identify early-warning signals of potential climate tipping points in systems such as the Amazon rainforest and coral reefs (Boers \u0026amp; Rypdal, 2021; Boulton et al., 2022).\u003c/li\u003e\n\u003c/ul\u003e\n\u003cp\u003e4.4. Challenges associated with GeoAI applications in climate science\u003c/p\u003e\n\u003cp\u003eDespite its transformative promise, the deployment and operationalization of GeoAI face several interconnected challenges (Table 5).\u003c/p\u003e\n\u003cp\u003eTable 5 Challenges facing GeoAI applications in climate science\u003c/p\u003e\n\u003ctable border=\"1\" cellspacing=\"0\" cellpadding=\"0\"\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 194px;\"\u003e\n \u003cp\u003eChallenge Category\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 625px;\"\u003e\n \u003cp\u003eSpecific Description and Manifestations\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 194px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eData Quality, Heterogeneity \u0026amp; Paucity\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 625px;\"\u003e\n \u003cp\u003eIssues include sensor noise, cloud contamination in optical imagery, spatial and temporal inconsistencies across different satellite platforms, a critical lack of high-quality, labeled training data for supervised learning (especially in remote regions and for rare events), and inherent biases in datasets that can perpetuate environmental injustices\u0026nbsp;(Ball et al., 2017; Karpatne, Atluri, et al., 2017; Tuia et al., 2022).\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 194px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eComputational Intractability \u0026amp; Resource Demands\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 625px;\"\u003e\n \u003cp\u003eTraining complex DL models (e.g., Vision Transformers) on high-resolution, global, multi-temporal datasets requires immense computational resources (e.g., clusters of GPUs/TPUs), creating a significant barrier to entry for researchers in low-resource institutions and countries, and raising the carbon footprint of AI research\u0026nbsp;(Schwartz et al., 2020; Strubell et al., 2019).\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 194px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eModel Generalizability, Transferability \u0026amp; Robustness\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 625px;\"\u003e\n \u003cp\u003eModels trained on data from one geographic region, time period, or sensor often perform poorly when applied to another (domain shift), limiting their global utility and operational robustness. This is exacerbated by non-stationarity in climate systems, where past relationships may not hold in the future\u0026nbsp;(Reichstein et al., 2019b; Rolnick et al., 2023; R. Wang et al., 2024).\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 194px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eExplainability, Interpretability and Trust (The \u0026ldquo;Black Box\u0026rdquo; Problem)\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 625px;\"\u003e\n \u003cp\u003eThe complex, high-dimensional nature of many DL models hinders understanding of \u003cem\u003ehow\u003c/em\u003e and \u003cem\u003ewhy\u003c/em\u003e they arrive at a particular prediction. This lack of transparency is a major barrier to adoption by domain scientists, policymakers, and stakeholders who require interpretable and physically plausible evidence to inform decisions (Roscher et al., 2020; Samek \u0026amp; M\u0026uuml;ller, 2019; Toms et al., 2020)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 194px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eEthical Considerations, Algorithmic Fairness \u0026amp; Justice\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 625px;\"\u003e\n \u003cp\u003eBiases in training data (e.g., uneven global distribution of weather stations) can lead to biased model outcomes that systematically under-predict risks for marginalized communities, potentially exacerbating existing climate vulnerabilities and violating principles of climate justice\u0026nbsp;(Messina et al., 2019; Rolnick et al., 2023). Data sovereignty, privacy (e.g., from high-resolution imagery), and the potential for misuse (e.g., in geoengineering) are also profound concerns.\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 194px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eInteroperability, Reproducibility \u0026amp; Workflow Integration\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 625px;\"\u003e\n \u003cp\u003eIntegrating diverse data formats (netCDF, GeoTIFF, and point clouds), software ecosystems (Python, R, and GIS), and reconciling the outputs from data-driven AI models with well-established physical models remain technically challenging. A lack of standardized benchmarking datasets and model reporting hinders reproducibility and fair comparison\u0026nbsp;(Frame et al., 2022; Lundberg et al., 2017)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n\u003c/table\u003e\n\u003cp\u003e4.5. Synthesis and prioritization of GeoAI challenges\u003c/p\u003e\n\u003cp\u003eWhile the preceding sections have detailed individual challenges, a synthesis is required to understand their relative severity, prevalence, and interrelationships to effectively prioritize mitigation efforts. Fig. 7 shows these impediments, combining a radar chart to assess perceived severity with a bar chart to indicate their prevalence in the reviewed literature.\u003c/p\u003e\n\u003cp\u003eThe radar chart (Fig. 7) shows expert-based severity scores (1\u0026ndash;10 scale) for six major challenge categories. A higher score indicates a greater perceived impediment to field advancement. It reveals that the most severe challenges, as inferred from author discussions and critical analysis, are those related to Explainability \u0026amp; Trust and Data Quality \u0026amp; Heterogeneity. The \u0026ldquo;black box\u0026rdquo; nature of complex models (Explainability) is not simply a technical issue but a fundamental barrier to adoption by policymakers and domain scientists who require transparent, physically plausible evidence. Concurrently, the foundational issue of data (Quality \u0026amp; Heterogeneity) continues to undermine model reliability and generalization, a problem exacerbated by the unique noise characteristics of remote sensing data and the high cost of creating accurate labels.\u003c/p\u003e\n\u003cp\u003eThe prevalence analysis (Fig. 7) tells a slightly different story. Data Quality \u0026amp; Heterogeneity and Model Generalizability are the most frequently discussed challenges, indicating that they are the most immediate and commonly encountered setbacks in daily research. While Ethical Considerations are discussed with moderate frequency, they are perceived as highly severe, signaling a growing recognition of their profound implications for climate justice and the potential for algorithmic bias to exacerbate existing vulnerabilities.\u003c/p\u003e\n\u003cp\u003eThis synthesized view shows that the community faces three problems: 1) Foundational Data Issues, 2) Operational Model Limitations (generalizability and computational demands), and 3) Profound Epistemic and Ethical Concerns (explainability and ethics). Addressing these challenges requires a coordinated effort across computer science, geospatial theory, and social science.\u003c/p\u003e"},{"header":"5. Existing gaps and future directions","content":"\u003cp\u003eThe synthesis of the literature reveals several theoretical, methodological, and ethical gaps that present opportunities for future research (Fig. 8).\u003c/p\u003e\n\u003cp\u003e5.1. The explainability gap: from black box to trusted, transparent tool\u003c/p\u003e\n\u003cp\u003eA predominant and widely acknowledged gap is the limited application of Explainable AI (XAI) and interpretability techniques in GeoAI-climate studies (Toms et al., 2020). While complex models often achieve high predictive accuracy, the physical reasoning and driving factors behind their predictions remain unclear, limiting their utility for scientific discovery and policy formulation. Future research should prioritize the systematic integration of post-hoc (e.g., SHAP (Lundberg et al., 2017) and LIME (Ribeiro et al., 2016) and intrinsic (e.g., attention mechanisms in transformers) XAI methods. The goal is to move beyond mere prediction to understanding, uncovering the spatiotemporal variables and interactions that the model considers essential and ensuring these align with domain knowledge and physical plausibility (Roscher et al., 2020; Toms et al., 2020). This is crucial for building trust, diagnosing model failures, and generating new, testable hypotheses about the climate system. \u0026nbsp; \u0026nbsp;\u003c/p\u003e\n\u003cp\u003e5.2. The causality gap: moving beyond correlation to causal understanding\u003c/p\u003e\n\u003cp\u003eMost current GeoAI applications effectively identify complex, non-linear correlations but fail to establish causal relationships (Runge et al., 2019). Inferring causality from observational data helps elucidate the fundamental drivers of climate impacts, assess the efficacy of interventions, and make robust predictions under novel conditions (e.g., new policy scenarios). Future work should explore the integration of causal inference frameworks (Neuberg, 2003), such as causal directed acyclic graphs (DAGs), propensity score matching, and causal discovery algorithms, with GeoAI models. This represents a paradigm shift from purely predictive modeling to a causal learning paradigm that can answer \u0026ldquo;what if\u0026rdquo; questions, which is the cornerstone of effective climate adaptation and mitigation planning (Beucler et al., 2021; Nowack et al., 2020).\u003c/p\u003e\n\u003cp\u003e5.3. The generalizability gap: toward foundation models for earth observation\u003c/p\u003e\n\u003cp\u003eThe lack of model transferability and robustness across space, time, and sensors hinders the operational deployment of GeoAI (Tuia et al., 2022). A promising and transformative direction is the development of \u0026ldquo;foundation models\u0026rdquo; or \u0026ldquo;pre-trained models\u0026rdquo; for Earth observation (Irvin et al., 2020; Reed et al., 2023) and large-scale models pre-trained on diverse and unlabeled satellite data archives using self-supervised learning objectives. These models would learn general-purpose representations of the Earth\u0026rsquo;s surface, which could then be fine-tuned for specific climate tasks (e.g., flood mapping and crop classification) with minimal additional labeled data. This would democratize access to powerful GeoAI tools, enhance global consistency in monitoring, and significantly improve performance in data-sparse regions (Irrgang et al., 2021b; Sarmadi et al., 2023; R. Wang et al., 2024).\u003c/p\u003e\n\u003cp\u003e5.4. The integration gap: physics-informed and knowledge-guided GeoAI\u003c/p\u003e\n\u003cp\u003eA significant frontier lies in the principled coupling of AI with physical knowledge and mechanistic understanding (Karpatne, Ebert-Uphoff, et al., 2017b). Pure data-driven models can produce physically inconsistent results, especially when extrapolating. PINNs (Raissi et al., 2019) and other approaches that embed physical laws (e.g., conservation equations and radiative transfer models) as soft constraints into the learning process can produce more robust, generalizable, and physically consistent models. This hybrid approach, called \u0026ldquo;theory-guided data science,\u0026rdquo; is particularly valuable for forecasting and simulation in regions with sparse observational data, ensuring that AI solutions respect the fundamental laws governing the Earth system(Kashinath et al., 2021; Read et al., 2019).\u003c/p\u003e\n\u003cp\u003e5.5. The equity and implementation gap: socially conscious, participatory, and equitable GeoAI.\u003c/p\u003e\n\u003cp\u003eThe social and ethical dimensions of GeoAI requires urgent attention (Y. Liu et al., 2018; Runting et al., 2020). Future research should proactively develop methods for auditing and mitigating algorithmic biases that could disadvantage vulnerable populations in climate risk assessments and resource allocation. Furthermore, studies should develop participatory and co-designed GeoAI frameworks that incorporate local knowledge, indigenous wisdom, and community priorities into the modeling process (Goldman et al., 2009). This ensures that GeoAI solutions are technically sophisticated, contextually appropriate, socially robust, and equitable, thereby aligning technological advancement with the principles of climate and environmental justice(Gavahi et al., 2023).\u003c/p\u003e"},{"header":"6. Conclusion","content":"\u003cp\u003eThis study comprehensively analyzed GeoAI applications in climate change impact assessment through scientometrics and a systematic review. The scientometric analysis of 152 core publications, augmented by a multi-database search strategy, quantitatively elucidated the field\u0026rsquo;s rapid, exponential growth from 2015 to 2025. It identified key thematic clusters around ML methodologies, remote sensing data, and specific climate impacts and mapped the influential roles of leading countries and premier journals in remote sensing and interdisciplinary science.\u003c/p\u003e \u003cp\u003eThe systematic review of 77 articles provided an in-depth examination of the field\u0026rsquo;s intellectual and technical aspects. It elucidated the dominant technological paradigms, with CNNs for spatial feature learning, RNNs/LSTMs for temporal dynamics, and emerging Transformer-based models for global context. Hybrid Physics-Informed AI frameworks and their transformative applications across critical domains, including cryosphere and hydrosphere monitoring, terrestrial and marine ecosystem analysis, agricultural security forecasting, and urban climate resilience planning, were also elucidated. The study critically examined the multi-faceted challenges of data quality and heterogeneity, computational intractability, the \u0026ldquo;black box\u0026rdquo; problem of model interpretability, limited generalizability, and profound ethical considerations that currently constrain the field\u0026rsquo;s full potential and equitable application.\u003c/p\u003e \u003cp\u003eFinally, the synthesis identified pivotal research gaps, most notably the pressing need for explainable (XAI), causal, generalizable (e.g., via foundation models), physics-informed, and socially equitable GeoAI frameworks. These recommendations call for an interdisciplinary effort to develop these next-generation systems, which must be guided by the principles of transparency, physical consistency, and climate justice. By addressing these challenges and rigorously pursuing these future directions, the GeoAI research community can significantly augment our global capacity to understand, project, mitigate, and adapt to the profound and escalating impacts of climate change. Thus, we can provide the reliable, actionable, and trustworthy intelligence essential for navigating toward a resilient planetary future.\u003c/p\u003e \u003cdiv id=\"Sec35\" class=\"Section2\"\u003e \u003ch2\u003e6.1. Limitations and research outlook\u003c/h2\u003e \u003cp\u003eThis review is constrained by its scope and methodology. While multiple databases were used, some relevant studies in non-English languages or in less indexed venues may have been omitted. The focus on a specific set of keywords, although carefully selected, may not capture all nascent applications at the periphery of this rapidly evolving field. The systematic review\u0026rsquo;s qualitative synthesis, while rigorous, is inherently subject to researcher interpretation. Future work should include broader literature searches and more formal meta-analysis techniques where sufficient quantitative data exists. The field\u0026rsquo;s trajectory suggests that the integration of AI with Earth system science will only advance, prompting ongoing critical reviews and methodological innovations to ensure its responsible and effective deployment.\u003c/p\u003e \u003c/div\u003e"},{"header":"Declarations","content":"\u003cp\u003e \u003cstrong\u003eConflict of interest\u003c/strong\u003e \u003cp\u003eauthor declares no conflict of interest.\u003c/p\u003e \u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\n \u003cli\u003eAlnajjar, S., Garc\u0026iacute;a-Mart\u0026iacute;nez, A., L\u0026oacute;pez-Cabeza, V. P., \u0026amp; Al-Azhari, W. (2025). A Multidimensional Approach to Mapping Urban Heat Vulnerability: Integrating Remote Sensing and Spatial Configuration. \u003cem\u003eSmart Cities\u003c/em\u003e, \u003cem\u003e8\u003c/em\u003e(4). https://doi.org/10.3390/smartcities8040137\u003c/li\u003e\n \u003cli\u003eAndela, N., Morton, D. C., Giglio, L., Chen, Y., Van Der Werf, G. R., Kasibhatla, P. S., DeFries, R. S., Collatz, G. J., Hantson, S., Kloster, S., Bachelet, D., Forrest, M., Lasslop, G., Li, F., Mangeon, S., Melton, J. R., Yue, C., \u0026amp; Randerson, J. T. (2017). A human-driven decline in global burned area. \u003cem\u003eScience\u003c/em\u003e, \u003cem\u003e356\u003c/em\u003e(6345), 1356\u0026ndash;1362. https://doi.org/10.1126/science.aal4108\u003c/li\u003e\n \u003cli\u003eAria, M., \u0026amp; Cuccurullo, C. (2017). bibliometrix: An R-tool for comprehensive science mapping analysis. \u003cem\u003eJournal of Informetrics\u003c/em\u003e, \u003cem\u003e11\u003c/em\u003e(4), 959\u0026ndash;975. https://doi.org/10.1016/j.joi.2017.08.007\u003c/li\u003e\n \u003cli\u003eAviles Toledo, C., Crawford, M. M., \u0026amp; Tuinstra, M. R. (2024). Integrating multi-modal remote sensing, deep learning, and attention mechanisms for yield prediction in plant breeding experiments. \u003cem\u003eFrontiers in Plant Science\u003c/em\u003e, \u003cem\u003e15\u003c/em\u003e. https://doi.org/10.3389/fpls.2024.1408047\u003c/li\u003e\n \u003cli\u003eBall, J. E., Anderson, D. T., \u0026amp; Chan, C. S. (2017). Comprehensive survey of deep learning in remote sensing: theories, tools, and challenges for the community. \u003cem\u003eJournal of Applied Remote Sensing\u003c/em\u003e, \u003cem\u003e11\u003c/em\u003e(04), 1. https://doi.org/10.1117/1.jrs.11.042609\u003c/li\u003e\n \u003cli\u003eBan, Y., Zhang, P., Nascetti, A., Bevington, A. R., \u0026amp; Wulder, M. A. (2020). Near Real-Time Wildfire Progression Monitoring with Sentinel-1 SAR Time Series and Deep Learning. \u003cem\u003eScientific Reports\u003c/em\u003e, \u003cem\u003e10\u003c/em\u003e(1). https://doi.org/10.1038/s41598-019-56967-x\u003c/li\u003e\n \u003cli\u003eBeucler, T., Pritchard, M., Rasp, S., Ott, J., Baldi, P., \u0026amp; Gentine, P. (2021). Enforcing Analytic Constraints in Neural Networks Emulating Physical Systems. \u003cem\u003ePhysical Review Letters\u003c/em\u003e, \u003cem\u003e126\u003c/em\u003e(9). https://doi.org/10.1103/PhysRevLett.126.098302\u003c/li\u003e\n \u003cli\u003eBi, K., Xie, L., Zhang, H., Chen, X., Gu, X., \u0026amp; Tian, Q. (2023). Accurate medium-range global weather forecasting with 3D neural networks. \u003cem\u003eNature\u003c/em\u003e, \u003cem\u003e619\u003c/em\u003e(7970), 533\u0026ndash;538. https://doi.org/10.1038/s41586-023-06185-3\u003c/li\u003e\n \u003cli\u003eBoers, N., \u0026amp; Rypdal, M. (2021). \u003cem\u003eCritical slowing down suggests that the western Greenland Ice Sheet is close to a tipping point\u003c/em\u003e. https://doi.org/10.1073/pnas.2024192118/-/DCSupplemental.y\u003c/li\u003e\n \u003cli\u003eBotella, C., Joly, A., Bonnet, P., Monestiez, P., \u0026amp; Munoz, F. (2018). \u003cem\u003eA deep learning approach to Species Distribution Modelling\u003c/em\u003e.\u003c/li\u003e\n \u003cli\u003eBoulton, C. A., Lenton, T. M., \u0026amp; Boers, N. (2022). Pronounced loss of Amazon rainforest resilience since the early 2000s. \u003cem\u003eNature Climate Change\u003c/em\u003e, \u003cem\u003e12\u003c/em\u003e(3), 271\u0026ndash;278. https://doi.org/10.1038/s41558-022-01287-8\u003c/li\u003e\n \u003cli\u003eBreiman, L. (2001). \u003cem\u003eRandom Forests\u003c/em\u003e (Vol. 45).\u003c/li\u003e\n \u003cli\u003eBullock, E. L., Woodcock, C. E., \u0026amp; Holden, C. E. (2020). Improved change monitoring using an ensemble of time series algorithms. \u003cem\u003eRemote Sensing of Environment\u003c/em\u003e, \u003cem\u003e238\u003c/em\u003e. https://doi.org/10.1016/j.rse.2019.04.018\u003c/li\u003e\n \u003cli\u003eChen, T., \u0026amp; Guestrin, C. (2016). XGBoost: A scalable tree boosting system. \u003cem\u003eProceedings of the ACM SIGKDD International Conference on Knowledge Discovery and Data Mining\u003c/em\u003e, \u003cem\u003e13-17-August-2016\u003c/em\u003e, 785\u0026ndash;794. https://doi.org/10.1145/2939672.2939785\u003c/li\u003e\n \u003cli\u003eCho, K., van Merrienboer, B., Gulcehre, C., Bahdanau, D., Bougares, F., Schwenk, H., \u0026amp; Bengio, Y. (2014). \u003cem\u003eLearning Phrase Representations using RNN Encoder-Decoder for Statistical Machine Translation\u003c/em\u003e. http://arxiv.org/abs/1406.1078\u003c/li\u003e\n \u003cli\u003eCreswell, J. W. ., \u0026amp; Poth, C. N. . (2018). \u003cem\u003eQualitative inquiry and research design\u003c/em\u003e. SAGE.\u003c/li\u003e\n \u003cli\u003eDosovitskiy, A., Beyer, L., Kolesnikov, A., Weissenborn, D., Zhai, X., Unterthiner, T., Dehghani, M., Minderer, M., Heigold, G., Gelly, S., Uszkoreit, J., \u0026amp; Houlsby, N. (2021). \u003cem\u003eAn Image is Worth 16x16 Words: Transformers for Image Recognition at Scale\u003c/em\u003e. http://arxiv.org/abs/2010.11929\u003c/li\u003e\n \u003cli\u003eElith, J., Leathwick, J. R., \u0026amp; Hastie, T. (2008). A working guide to boosted regression trees. In \u003cem\u003eJournal of Animal Ecology\u003c/em\u003e (Vol. 77, Issue 4, pp. 802\u0026ndash;813). https://doi.org/10.1111/j.1365-2656.2008.01390.x\u003c/li\u003e\n \u003cli\u003eEllegaard, O., \u0026amp; Wallin, J. A. (2015). The bibliometric analysis of scholarly production: How great is the impact? \u003cem\u003eScientometrics\u003c/em\u003e, \u003cem\u003e105\u003c/em\u003e(3), 1809\u0026ndash;1831. https://doi.org/10.1007/s11192-015-1645-z\u003c/li\u003e\n \u003cli\u003eElo, S., \u0026amp; Kyng\u0026auml;s, H. (2008). The qualitative content analysis process. \u003cem\u003eJournal of Advanced Nursing\u003c/em\u003e, \u003cem\u003e62\u003c/em\u003e(1), 107\u0026ndash;115. https://doi.org/10.1111/j.1365-2648.2007.04569.x\u003c/li\u003e\n \u003cli\u003eFang, K., Shen, C., Kifer, D., \u0026amp; Yang, X. (2017). Prolongation of SMAP to Spatiotemporally Seamless Coverage of Continental U.S. Using a Deep Learning Neural Network. \u003cem\u003eGeophysical Research Letters\u003c/em\u003e, \u003cem\u003e44\u003c/em\u003e(21), 11,030-11,039. https://doi.org/10.1002/2017GL075619\u003c/li\u003e\n \u003cli\u003eFayaz, M., Nam, J., Dang, L. M., Song, H. K., \u0026amp; Moon, H. (2024). Land-Cover Classification Using Deep Learning with High-Resolution Remote-Sensing Imagery. \u003cem\u003eApplied Sciences (Switzerland)\u003c/em\u003e, \u003cem\u003e14\u003c/em\u003e(5). https://doi.org/10.3390/app14051844\u003c/li\u003e\n \u003cli\u003eFrame, J. M., Kratzert, F., Klotz, D., Gauch, M., Shelev, G., Gilon, O., Qualls, L. M., Gupta, H. V., \u0026amp; Nearing, G. S. (2022). Deep learning rainfall-runoff predictions of extreme events. \u003cem\u003eHydrology and Earth System Sciences\u003c/em\u003e, \u003cem\u003e26\u003c/em\u003e(13), 3377\u0026ndash;3392. https://doi.org/10.5194/hess-26-3377-2022\u003c/li\u003e\n \u003cli\u003eFree, G., Bresciani, M., Pinardi, M., Ghirardi, N., Luciani, G., Caroni, R., \u0026amp; Giardino, C. (2021). Detecting climate driven changes in chlorophyll-a in deep subalpine lakes using long term satellite data. \u003cem\u003eWater (Switzerland)\u003c/em\u003e, \u003cem\u003e13\u003c/em\u003e(6). https://doi.org/10.3390/w13060866\u003c/li\u003e\n \u003cli\u003eGavahi, K., Forumandi, E., \u0026amp; Moradkhani, H. (2023). \u003cem\u003eA deep learning-based framework for multi-source 2 precipitation fusion\u003c/em\u003e. https://www.sciencedirect.com/science/article/pii/S0034425723002742\u003c/li\u003e\n \u003cli\u003eGoldman, J., Shilton, K., Burke, J., Estrin, D., Hansen, M., Ramanathan, N., Reddy, S., Samanta, V., Srivastava, M., \u0026amp; West, R. (2009). \u003cem\u003eParticipatory Sensing\u003c/em\u003e. http://www.wilsoncenter.org/foresight/\u003c/li\u003e\n \u003cli\u003eHam, Y. G., Kim, J. H., \u0026amp; Luo, J. J. (2019). Deep learning for multi-year ENSO forecasts. \u003cem\u003eNature\u003c/em\u003e, \u003cem\u003e573\u003c/em\u003e(7775), 568\u0026ndash;572. https://doi.org/10.1038/s41586-019-1559-7\u003c/li\u003e\n \u003cli\u003eHarris, L., McRae, A. T. T., Chantry, M., Dueben, P. D., \u0026amp; Palmer, T. N. (2022). A Generative Deep Learning Approach to Stochastic Downscaling of Precipitation Forecasts. \u003cem\u003eJournal of Advances in Modeling Earth Systems\u003c/em\u003e, \u003cem\u003e14\u003c/em\u003e(10). https://doi.org/10.1029/2022MS003120\u003c/li\u003e\n \u003cli\u003eHeidler, K., Mou, L., Baumhoer, C., Dietz, A., \u0026amp; Zhu, X. X. (2022). HED-UNet: Combined Segmentation and Edge Detection for Monitoring the Antarctic Coastline. \u003cem\u003eIEEE Transactions on Geoscience and Remote Sensing\u003c/em\u003e, \u003cem\u003e60\u003c/em\u003e. https://doi.org/10.1109/TGRS.2021.3064606\u003c/li\u003e\n \u003cli\u003eHochreiter, S., \u0026amp; Schmidhuber, J. (1997). Long Short-Term Memory. \u003cem\u003eNeural Computation\u003c/em\u003e, \u003cem\u003e9\u003c/em\u003e(8), 1735\u0026ndash;1780. https://doi.org/10.1162/neco.1997.9.8.1735\u003c/li\u003e\n \u003cli\u003eHsieh, H. F., \u0026amp; Shannon, S. E. (2005). Three approaches to qualitative content analysis. \u003cem\u003eQualitative Health Research\u003c/em\u003e, \u003cem\u003e15\u003c/em\u003e(9), 1277\u0026ndash;1288. https://doi.org/10.1177/1049732305276687\u003c/li\u003e\n \u003cli\u003eIglovikov, V., Seferbekov, S., Buslaev, A., \u0026amp; Shvets, A. (2018). TernausNetV2: Fully convolutional network for instance segmentation. \u003cem\u003eIEEE Computer Society Conference on Computer Vision and Pattern Recognition Workshops\u003c/em\u003e, \u003cem\u003e2018-June\u003c/em\u003e, 228\u0026ndash;232. https://doi.org/10.1109/CVPRW.2018.00042\u003c/li\u003e\n \u003cli\u003eIntergovernmental Panel on Climate Change (IPCC). (2023). Weather and Climate Extreme Events in a Changing Climate. In \u003cem\u003eClimate Change 2021 \u0026ndash; The Physical Science Basis\u003c/em\u003e (pp. 1513\u0026ndash;1766). Cambridge University Press. https://doi.org/10.1017/9781009157896.013\u003c/li\u003e\n \u003cli\u003eIrrgang, C., Boers, N., Sonnewald, M., Barnes, E. A., Kadow, C., Staneva, J., \u0026amp; Saynisch-Wagner, J. (2021a). Towards neural Earth system modelling by integrating artificial intelligence in Earth system science. In \u003cem\u003eNature Machine Intelligence\u003c/em\u003e (Vol. 3, Issue 8, pp. 667\u0026ndash;674). Nature Research. https://doi.org/10.1038/s42256-021-00374-3\u003c/li\u003e\n \u003cli\u003eIrrgang, C., Boers, N., Sonnewald, M., Barnes, E. A., Kadow, C., Staneva, J., \u0026amp; Saynisch-Wagner, J. (2021b). Towards neural Earth system modelling by integrating artificial intelligence in Earth system science. In \u003cem\u003eNature Machine Intelligence\u003c/em\u003e (Vol. 3, Issue 8, pp. 667\u0026ndash;674). Nature Research. https://doi.org/10.1038/s42256-021-00374-3\u003c/li\u003e\n \u003cli\u003eIrvin, J., Sheng, H., Ramachandran, N., Johnson-Yu, S., Zhou, S., Story, K., Rustowicz, R., Elsworth, C., Austin, K., \u0026amp; Ng, A. Y. (2020). \u003cem\u003eForestNet: Classifying Drivers of Deforestation in Indonesia using Deep Learning on Satellite Imagery\u003c/em\u003e. http://arxiv.org/abs/2011.05479\u003c/li\u003e\n \u003cli\u003eJain, P., Coogan, S. C. P., Subramanian, S. G., Crowley, M., Taylor, S., \u0026amp; Flannigan, M. D. (2020). A review of machine learning applications in wildfire science and management. In \u003cem\u003eEnvironmental Reviews\u003c/em\u003e (Vol. 28, Issue 4, pp. 478\u0026ndash;505). Canadian Science Publishing. https://doi.org/10.1139/er-2020-0019\u003c/li\u003e\n \u003cli\u003eJanowicz, K., Gao, S., McKenzie, G., Hu, Y., \u0026amp; Bhaduri, B. (2020). GeoAI: spatially explicit artificial intelligence techniques for geographic knowledge discovery and beyond. In \u003cem\u003eInternational Journal of Geographical Information Science\u003c/em\u003e (Vol. 34, Issue 4, pp. 625\u0026ndash;636). Taylor and Francis Ltd. https://doi.org/10.1080/13658816.2019.1684500\u003c/li\u003e\n \u003cli\u003eJin, Z., Azzari, G., You, C., Di Tommaso, S., Aston, S., Burke, M., \u0026amp; Lobell, D. B. (2019). Smallholder maize area and yield mapping at national scales with Google Earth Engine. \u003cem\u003eRemote Sensing of Environment\u003c/em\u003e, \u003cem\u003e228\u003c/em\u003e, 115\u0026ndash;128. https://doi.org/10.1016/j.rse.2019.04.016\u003c/li\u003e\n \u003cli\u003eKarpatne, A., Atluri, G., Faghmous, J. H., Steinbach, M., Banerjee, A., Ganguly, A., Shekhar, S., Samatova, N., \u0026amp; Kumar, V. (2017). Theory-guided data science: A new paradigm for scientific discovery from data. \u003cem\u003eIEEE Transactions on Knowledge and Data Engineering\u003c/em\u003e, \u003cem\u003e29\u003c/em\u003e(10), 2318\u0026ndash;2331. https://doi.org/10.1109/TKDE.2017.2720168\u003c/li\u003e\n \u003cli\u003eKarpatne, A., Ebert-Uphoff, I., Ravela, S., Babaie, H. A., \u0026amp; Kumar, V. (2017a). \u003cem\u003eMachine Learning for the Geosciences: Challenges and Opportunities\u003c/em\u003e. http://arxiv.org/abs/1711.04708\u003c/li\u003e\n \u003cli\u003eKarpatne, A., Ebert-Uphoff, I., Ravela, S., Babaie, H. A., \u0026amp; Kumar, V. (2017b). \u003cem\u003eMachine Learning for the Geosciences: Challenges and Opportunities\u003c/em\u003e. http://arxiv.org/abs/1711.04708\u003c/li\u003e\n \u003cli\u003eKashinath, K., Mustafa, M., Albert, A., Wu, J. L., Jiang, C., Esmaeilzadeh, S., Azizzadenesheli, K., Wang, R., Chattopadhyay, A., Singh, A., Manepalli, A., Chirila, D., Yu, R., Walters, R., White, B., Xiao, H., Tchelepi, H. A., Marcus, P., Anandkumar, A., \u0026hellip; Prabhat. (2021). Physics-informed machine learning: Case studies for weather and climate modelling. In \u003cem\u003ePhilosophical Transactions of the Royal Society A: Mathematical, Physical and Engineering Sciences\u003c/em\u003e (Vol. 379, Issue 2194). Royal Society Publishing. https://doi.org/10.1098/rsta.2020.0093\u003c/li\u003e\n \u003cli\u003eKe, G., Meng, Q., Finley, T., Wang, T., Chen, W., Ma, W., Ye, Q., \u0026amp; Liu, T.-Y. (2017). \u003cem\u003eLightGBM: A Highly Efficient Gradient Boosting Decision Tree\u003c/em\u003e. https://www.researchgate.net/publication/378480234\u003c/li\u003e\n \u003cli\u003eKhaki, S., Wang, L., \u0026amp; Archontoulis, S. V. (2020). A CNN-RNN Framework for Crop Yield Prediction. \u003cem\u003eFrontiers in Plant Science\u003c/em\u003e, \u003cem\u003e10\u003c/em\u003e. https://doi.org/10.3389/fpls.2019.01750\u003c/li\u003e\n \u003cli\u003eKratzert, F., Nearing, G., Addor, N., Erickson, T., Gauch, M., Gilon, O., Gudmundsson, L., Hassidim, A., Klotz, D., Nevo, S., Shalev, G., \u0026amp; Matias, Y. (2023). Caravan - A global community dataset for large-sample hydrology. \u003cem\u003eScientific Data\u003c/em\u003e, \u003cem\u003e10\u003c/em\u003e(1). https://doi.org/10.1038/s41597-023-01975-w\u003c/li\u003e\n \u003cli\u003eKrizhevsky, A., Sutskever, I., \u0026amp; Hinton, G. E. (2017). ImageNet classification with deep convolutional neural networks. \u003cem\u003eCommunications of the ACM\u003c/em\u003e, \u003cem\u003e60\u003c/em\u003e(6), 84\u0026ndash;90. https://doi.org/10.1145/3065386\u003c/li\u003e\n \u003cli\u003eKurth, T., Subramanian, S., Harrington, P., Pathak, J., Mardani, M., Hall, D., Miele, A., Kashinath, K., \u0026amp; Anandkumar, A. (2023, June 26). FourCastNet: Accelerating Global High-Resolution Weather Forecasting Using Adaptive Fourier Neural Operators. \u003cem\u003eProceedings of the Platform for Advanced Scientific Computing Conference, PASC 2023\u003c/em\u003e. https://doi.org/10.1145/3592979.3593412\u003c/li\u003e\n \u003cli\u003eLacoste, A., Sherwin, E. D., Kerner, H., Alemohammad, H., L\u0026uuml;tjens, B., Irvin, J., Dao, D., Chang, A., Gunturkun, M., Drouin, A., Rodriguez, P., \u0026amp; Vazquez, D. (2021). \u003cem\u003eToward Foundation Models for Earth Monitoring: Proposal for a Climate Change Benchmark\u003c/em\u003e. http://arxiv.org/abs/2112.00570\u003c/li\u003e\n \u003cli\u003eL\u0026auml;ngkvist, M., Kiselev, A., Alirezaie, M., \u0026amp; Loutfi, A. (2016). Classification and segmentation of satellite orthoimagery using convolutional neural networks. \u003cem\u003eRemote Sensing\u003c/em\u003e, \u003cem\u003e8\u003c/em\u003e(4). https://doi.org/10.3390/rs8040329\u003c/li\u003e\n \u003cli\u003eLe, X. H., Ho, H. V., Lee, G., \u0026amp; Jung, S. (2019). Application of Long Short-Term Memory (LSTM) neural network for flood forecasting. \u003cem\u003eWater (Switzerland)\u003c/em\u003e, \u003cem\u003e11\u003c/em\u003e(7). https://doi.org/10.3390/w11071387\u003c/li\u003e\n \u003cli\u003eLecun, Y., Bengio, Y., \u0026amp; Hinton, G. (2015). Deep learning. In \u003cem\u003eNature\u003c/em\u003e (Vol. 521, Issue 7553, pp. 436\u0026ndash;444). Nature Publishing Group. https://doi.org/10.1038/nature14539\u003c/li\u003e\n \u003cli\u003eLees, T., Buechel, M., Anderson, B., Slater, L., Reece, S., Coxon, G., \u0026amp; Dadson, S. J. (2021). Benchmarking data-driven rainfall-runoff models in Great Britain: A comparison of long short-term memory (LSTM)-based models with four lumped conceptual models. \u003cem\u003eHydrology and Earth System Sciences\u003c/em\u003e, \u003cem\u003e25\u003c/em\u003e(10), 5517\u0026ndash;5534. https://doi.org/10.5194/hess-25-5517-2021\u003c/li\u003e\n \u003cli\u003eLi, F., Yigitcanlar, T., Nepal, M., Thanh, K. N., \u0026amp; Dur, F. (2024). A Novel Urban Heat Vulnerability Analysis: Integrating Machine Learning and Remote Sensing for Enhanced Insights. \u003cem\u003eRemote Sensing\u003c/em\u003e, \u003cem\u003e16\u003c/em\u003e(16). https://doi.org/10.3390/rs16163032\u003c/li\u003e\n \u003cli\u003eLi, W. (2020a). GeoAI: Where machine learning and big data converge in GIScience. \u003cem\u003eJournal of Spatial Information Science\u003c/em\u003e, \u003cem\u003e20\u003c/em\u003e, 71\u0026ndash;77. https://doi.org/10.5311/JOSIS.2020.20.658\u003c/li\u003e\n \u003cli\u003eLi, W. (2020b). GeoAI: Where machine learning and big data converge in GIScience. \u003cem\u003eJournal of Spatial Information Science\u003c/em\u003e, \u003cem\u003e20\u003c/em\u003e, 71\u0026ndash;77. https://doi.org/10.5311/JOSIS.2020.20.658\u003c/li\u003e\n \u003cli\u003eLiu, H., Zhong, H., Lin, W., \u0026amp; Wu, J. (2024). Tree species classification based on PointNet++ deep learning and true- colour point cloud. \u003cem\u003eInternational Journal of Remote Sensing\u003c/em\u003e, \u003cem\u003e45\u003c/em\u003e(16), 5577\u0026ndash;5604. https://doi.org/10.1080/01431161.2024.2377837\u003c/li\u003e\n \u003cli\u003eLiu, Y., Bi, J., \u0026amp; Lv, J. (2018). Future impacts of climate change and land use on multiple ecosystem services in a rapidly urbanizing agricultural Basin, China. \u003cem\u003eSustainability (Switzerland)\u003c/em\u003e, \u003cem\u003e10\u003c/em\u003e(12). https://doi.org/10.3390/su10124575\u003c/li\u003e\n \u003cli\u003eLowe, R., Lee, S. A., Brady, O. J., Bastos, L., Col\u0026oacute;n-Gonz\u0026aacute;lez, F. J., on Climate Change, C., Health Lowe, P. R., Gasparrini, A., Carvalho, M. S., Lowe, R., Lee, S. A., O, K. M., Brady, O. J., Bastos, L., Carrasco-Escobar, G., de Castro Cat\u0026atilde;o, R., Col\u0026oacute;n-Gonz\u0026aacute;lez, F. J., Barcellos, C., S\u0026aacute; Carvalho, M., \u0026hellip; Gasparrini, A. (2021). Centre for Mathematical Modelling of Infectious Diseases ( Department of Public Health Environments and Society (Prof A Gasparrini), and Centre for Statistical Modelling (Prof A Gasparrini) Combined effects of hydrometeorological hazards and urbanisation on dengue risk in Brazil: a spatiotemporal modelling study. In \u003cem\u003eArticles Lancet Planet Health\u003c/em\u003e (Vol. 5). www.thelancet.com/\u003c/li\u003e\n \u003cli\u003eLundberg, S. M., Allen, P. G., \u0026amp; Lee, S.-I. (2017). A Unified Approach to Interpreting Model Predictions. In \u003cem\u003eACM Digital Library\u003c/em\u003e. https://github.com/slundberg/shap\u003c/li\u003e\n \u003cli\u003eMa, L., Liu, Y., Zhang, X., Ye, Y., Yin, G., \u0026amp; Johnson, B. A. (2019a). Deep learning in remote sensing applications: A meta-analysis and review. In \u003cem\u003eISPRS Journal of Photogrammetry and Remote Sensing\u003c/em\u003e (Vol. 152, pp. 166\u0026ndash;177). Elsevier B.V. https://doi.org/10.1016/j.isprsjprs.2019.04.015\u003c/li\u003e\n \u003cli\u003eMa, L., Liu, Y., Zhang, X., Ye, Y., Yin, G., \u0026amp; Johnson, B. A. (2019b). Deep learning in remote sensing applications: A meta-analysis and review. In \u003cem\u003eISPRS Journal of Photogrammetry and Remote Sensing\u003c/em\u003e (Vol. 152, pp. 166\u0026ndash;177). Elsevier B.V. https://doi.org/10.1016/j.isprsjprs.2019.04.015\u003c/li\u003e\n \u003cli\u003eMart\u0026iacute;n-Mart\u0026iacute;n, A., Thelwall, M., Orduna-Malea, E., \u0026amp; Delgado L\u0026oacute;pez-C\u0026oacute;zar, E. (2021). Google Scholar, Microsoft Academic, Scopus, Dimensions, Web of Science, and OpenCitations\u0026rsquo; COCI: a multidisciplinary comparison of coverage via citations. \u003cem\u003eScientometrics\u003c/em\u003e, \u003cem\u003e126\u003c/em\u003e(1), 871\u0026ndash;906. https://doi.org/10.1007/s11192-020-03690-4\u003c/li\u003e\n \u003cli\u003eMaxwell, A. E., Warner, T. A., \u0026amp; Fang, F. (2018). Implementation of machine-learning classification in remote sensing: An applied review. In \u003cem\u003eInternational Journal of Remote Sensing\u003c/em\u003e (Vol. 39, Issue 9, pp. 2784\u0026ndash;2817). Taylor and Francis Ltd. https://doi.org/10.1080/01431161.2018.1433343\u003c/li\u003e\n \u003cli\u003eMei, W., Wang, H., Fouhey, D., Zhou, W., Hinks, I., Gray, J. M., Van Berkel, D., \u0026amp; Jain, M. (2022). Using Deep Learning and Very-High-Resolution Imagery to Map Smallholder Field Boundaries. \u003cem\u003eRemote Sensing\u003c/em\u003e, \u003cem\u003e14\u003c/em\u003e(13). https://doi.org/10.3390/rs14133046\u003c/li\u003e\n \u003cli\u003eMentaschi, L., Vousdoukas, M. I., Pekel, J. F., Voukouvalas, E., \u0026amp; Feyen, L. (2018). Global long-term observations of coastal erosion and accretion. \u003cem\u003eScientific Reports\u003c/em\u003e, \u003cem\u003e8\u003c/em\u003e(1). https://doi.org/10.1038/s41598-018-30904-w\u003c/li\u003e\n \u003cli\u003eMessina, J. P., Brady, O. J., Golding, N., Kraemer, M. U. G., Wint, G. R. W., Ray, S. E., Pigott, D. M., Shearer, F. M., Johnson, K., Earl, L., Marczak, L. B., Shirude, S., Davis Weaver, N., Gilbert, M., Velayudhan, R., Jones, P., Jaenisch, T., Scott, T. W., Reiner, R. C., \u0026amp; Hay, S. I. (2019). The current and future global distribution and population at risk of dengue. \u003cem\u003eNature Microbiology\u003c/em\u003e, \u003cem\u003e4\u003c/em\u003e(9), 1508\u0026ndash;1515. https://doi.org/10.1038/s41564-019-0476-8\u003c/li\u003e\n \u003cli\u003eNeuberg, L. (2003). Causality: models, reasoning, and inference, by judea pearl, cambridge university press, 2000. \u003cem\u003eEconometric Theory\u003c/em\u003e, \u003cem\u003e19\u003c/em\u003e, 675\u0026ndash;685. https://doi.org/10.1017/S0266466603004109\u003c/li\u003e\n \u003cli\u003eNowack, P., Runge, J., Eyring, V., \u0026amp; Haigh, J. D. (2020). Causal networks for climate model evaluation and constrained projections. \u003cem\u003eNature Communications\u003c/em\u003e, \u003cem\u003e11\u003c/em\u003e(1). https://doi.org/10.1038/s41467-020-15195-y\u003c/li\u003e\n \u003cli\u003ePage, M. J., McKenzie, J. E., Bossuyt, P. M., Boutron, I., Hoffmann, T. C., Mulrow, C. D., Shamseer, L., Tetzlaff, J. M., Akl, E. A., Brennan, S. E., Chou, R., Glanville, J., Grimshaw, J. M., Hr\u0026oacute;bjartsson, A., Lalu, M. M., Li, T., Loder, E. W., Mayo-Wilson, E., McDonald, S., \u0026hellip; Moher, D. (2021). The PRISMA 2020 statement: An updated guideline for reporting systematic reviews. \u003cem\u003eJournal of Clinical Epidemiology\u003c/em\u003e, \u003cem\u003e134\u003c/em\u003e, 178\u0026ndash;189. https://doi.org/10.1016/j.jclinepi.2021.03.001\u003c/li\u003e\n \u003cli\u003ePahlevan, N., Smith, B., Schalles, J., Binding, C., Cao, Z., Ma, R., Alikas, K., Kangro, K., Gurlin, D., H\u0026agrave;, N., Matsushita, B., Moses, W., Greb, S., Lehmann, M. K., Ondrusek, M., Oppelt, N., \u0026amp; Stumpf, R. (2020). Seamless retrievals of chlorophyll-a from Sentinel-2 (MSI) and Sentinel-3 (OLCI) in inland and coastal waters: A machine-learning approach. \u003cem\u003eRemote Sensing of Environment\u003c/em\u003e, \u003cem\u003e240\u003c/em\u003e. https://doi.org/10.1016/j.rse.2019.111604\u003c/li\u003e\n \u003cli\u003eRaissi, M., Perdikaris, P., \u0026amp; Karniadakis, G. E. (2019). Physics-informed neural networks: A deep learning framework for solving forward and inverse problems involving nonlinear partial differential equations. \u003cem\u003eJournal of Computational Physics\u003c/em\u003e, \u003cem\u003e378\u003c/em\u003e, 686\u0026ndash;707. https://doi.org/10.1016/j.jcp.2018.10.045\u003c/li\u003e\n \u003cli\u003eRasp, S., Pritchard, M. S., \u0026amp; Gentine, P. (2018). \u003cem\u003eDeep learning to represent subgrid processes in climate models\u003c/em\u003e. https://doi.org/10.5281/zenodo.1402384\u003c/li\u003e\n \u003cli\u003eRead, J. S., Jia, X., Willard, J., Appling, A. P., Zwart, J. A., Oliver, S. K., Karpatne, A., Hansen, G. J. A., Hanson, P. C., Watkins, W., Steinbach, M., \u0026amp; Kumar, V. (2019). Process-Guided Deep Learning Predictions of Lake Water Temperature. \u003cem\u003eWater Resources Research\u003c/em\u003e, \u003cem\u003e55\u003c/em\u003e(11), 9173\u0026ndash;9190. https://doi.org/10.1029/2019WR024922\u003c/li\u003e\n \u003cli\u003eReed, C. J., Gupta, R., Li, S., Brockman, S., Funk, C., Clipp, B., Keutzer, K., Candido, S., Uyttendaele, M., \u0026amp; Darrell, T. (2023). \u003cem\u003eScale-MAE: A Scale-Aware Masked Autoencoder for Multiscale Geospatial Representation Learning\u003c/em\u003e. http://arxiv.org/abs/2212.14532\u003c/li\u003e\n \u003cli\u003eReichstein, M., Camps-Valls, G., Stevens, B., Jung, M., Denzler, J., Carvalhais, N., \u0026amp; Prabhat. (2019a). Deep learning and process understanding for data-driven Earth system science. \u003cem\u003eNature\u003c/em\u003e, \u003cem\u003e566\u003c/em\u003e(7743), 195\u0026ndash;204. https://doi.org/10.1038/s41586-019-0912-1\u003c/li\u003e\n \u003cli\u003eReichstein, M., Camps-Valls, G., Stevens, B., Jung, M., Denzler, J., Carvalhais, N., \u0026amp; Prabhat. (2019b). Deep learning and process understanding for data-driven Earth system science. \u003cem\u003eNature\u003c/em\u003e, \u003cem\u003e566\u003c/em\u003e(7743), 195\u0026ndash;204. https://doi.org/10.1038/s41586-019-0912-1\u003c/li\u003e\n \u003cli\u003eRibeiro, M. T., Singh, S., \u0026amp; Guestrin, C. (2016). \u0026ldquo;Why should i trust you?\u0026rdquo; Explaining the predictions of any classifier. \u003cem\u003eProceedings of the ACM SIGKDD International Conference on Knowledge Discovery and Data Mining\u003c/em\u003e, \u003cem\u003e13-17-August-2016\u003c/em\u003e, 1135\u0026ndash;1144. https://doi.org/10.1145/2939672.2939778\u003c/li\u003e\n \u003cli\u003eRipple, W. J., Wolf, C., Newsome, T. M., Barnard, P., \u0026amp; Moomaw, W. R. (2020). \u003cem\u003eWorld Scientists\u0026rsquo; Warning of a Climate Emergency SCIENTIST SIGNATORIES FROM 153 COUNTRIES (LIST IN SUPPLEMENTAL FILE S1)\u003c/em\u003e (Vol. 70, Issue 1). https://academic.oup.com/bioscience\u003c/li\u003e\n \u003cli\u003eRolnick, D., Donti, P. L., Kaack, L. H., Kochanski, K., Lacoste, A., Sankaran, K., Ross, A. S., Milojevic-Dupont, N., Jaques, N., Waldman-Brown, A., Luccioni, A. S., Maharaj, T., Sherwin, E. D., Mukkavilli, S. K., Kording, K. P., Gomes, C. P., Ng, A. Y., Hassabis, D., Platt, J. C., \u0026hellip; Bengio, Y. (2023). Tackling Climate Change with Machine Learning. In \u003cem\u003eACM Computing Surveys\u003c/em\u003e (Vol. 55, Issue 2). Association for Computing Machinery. https://doi.org/10.1145/3485128\u003c/li\u003e\n \u003cli\u003eRonneberger, O., Fischer, P., \u0026amp; Brox, T. (2015). \u003cem\u003eU-Net: Convolutional Networks for Biomedical Image Segmentation\u003c/em\u003e. http://arxiv.org/abs/1505.04597\u003c/li\u003e\n \u003cli\u003eRoscher, R., J\u0026uuml;lich, F., Bohn, B., Duarte, M., Garcke, J., \u0026amp; Duarte, M. F. (2020). Explainable Machine Learning for Scientific Insights and Discoveries. \u003cem\u003eIEEE\u003c/em\u003e. https://doi.org/10.48550/arXiv.1905.08883\u003c/li\u003e\n \u003cli\u003eRunge, J., Nowack, P., Kretschmer, M., Flaxman, S., \u0026amp; Sejdinovic, D. (2019). Detecting and quantifying causal associations in large nonlinear time series datasets. In \u003cem\u003eSci. Adv\u003c/em\u003e (Vol. 5). https://www.science.org\u003c/li\u003e\n \u003cli\u003eRunting, R. K., Phinn, S., Xie, Z., Venter, O., \u0026amp; Watson, J. E. M. (2020). Opportunities for big data in conservation and sustainability. In \u003cem\u003eNature Communications\u003c/em\u003e (Vol. 11, Issue 1). Nature Research. https://doi.org/10.1038/s41467-020-15870-0\u003c/li\u003e\n \u003cli\u003eSamek, W., \u0026amp; M\u0026uuml;ller, K. R. (2019). Towards Explainable Artificial Intelligence. In \u003cem\u003eLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics): Vol. 11700 LNCS\u003c/em\u003e (pp. 5\u0026ndash;22). Springer Verlag. https://doi.org/10.1007/978-3-030-28954-6_1\u003c/li\u003e\n \u003cli\u003eSarmadi, H., R\u0026ouml;gnvaldsson, T., Carlsson, N. R., Ohlsson, M., Wahab, I., \u0026amp; Hall, O. (2023). \u003cem\u003eTowards Explaining Satellite Based Poverty Predictions with Convolutional Neural Networks\u003c/em\u003e. https://doi.org/10.1109/DSAA60987.2023.10302541\u003c/li\u003e\n \u003cli\u003eSchneider, T., Lan, S., Stuart, A., \u0026amp; Teixeira, J. (2017a). Earth System Modeling 2.0: A Blueprint for Models That Learn From Observations and Targeted High-Resolution Simulations. \u003cem\u003eGeophysical Research Letters\u003c/em\u003e, \u003cem\u003e44\u003c/em\u003e(24), 12,396-12,417. https://doi.org/10.1002/2017GL076101\u003c/li\u003e\n \u003cli\u003eSchneider, T., Lan, S., Stuart, A., \u0026amp; Teixeira, J. (2017b). Earth System Modeling 2.0: A Blueprint for Models That Learn From Observations and Targeted High-Resolution Simulations. \u003cem\u003eGeophysical Research Letters\u003c/em\u003e, \u003cem\u003e44\u003c/em\u003e(24), 12,396-12,417. https://doi.org/10.1002/2017GL076101\u003c/li\u003e\n \u003cli\u003eSchwartz, R., Dodge, J., Smith, N. A., \u0026amp; Etzioni, O. (2020). Green AI. \u003cem\u003eCommunications of the ACM\u003c/em\u003e, \u003cem\u003e63\u003c/em\u003e(12), 54\u0026ndash;63. https://doi.org/10.1145/3381831\u003c/li\u003e\n \u003cli\u003eShi, X., Chen, Z., Wang, H., Yeung, D.-Y., Wong, W.-K., Woo, W.-C., \u0026amp; Kong Observatory, H. (2015). Convolutional LSTM Network: A Machine Learning Approach for Precipitation Nowcasting. \u003cem\u003eACM Digital Library\u003c/em\u003e.\u003c/li\u003e\n \u003cli\u003eShukla, P. R., Skea, J., Reisinger, A., Slade, R., Fradera, R., Pathak, M., Al, A., Malek, K., Ren\u0026eacute;e Van Diemen, B., Hasija, A., Lisboa, G., Luz, S., Malley, J., Mccollum, D., \u0026amp; Some, S. (2022). \u003cem\u003eClimate Change 2022 Mitigation of Climate Change Working Group III Contribution to the Sixth Assessment Report of the Intergovernmental Panel on Climate Change Summary for Policymakers Edited by\u003c/em\u003e. www.ipcc.ch\u003c/li\u003e\n \u003cli\u003eStark, T. (2018). \u003cem\u003eUsing Deep Convolutional Neural Networks for the Identification of Informal Settlements to Improve a Sustainable Development in Urban Environments\u003c/em\u003e.\u003c/li\u003e\n \u003cli\u003eSteffen, W., Rockstr\u0026ouml;m, J., Richardson, K., Lenton, T. M., Folke, C., Liverman, D., Summerhayes, C. P., Barnosky, A. D., Cornell, S. E., Crucifix, M., Donges, J. F., Fetzer, I., Lade, S. J., Scheffer, M., Winkelmann, R., \u0026amp; Schellnhuber, H. J. (2018). Trajectories of the Earth System in the Anthropocene. \u003cem\u003ePNAS\u003c/em\u003e. https://doi.org/10.1073/pnas.1810141115/-/DCSupplemental\u003c/li\u003e\n \u003cli\u003eStrubell, E., Ganesh, A., \u0026amp; Mccallum, A. (2019). Energy and Policy Considerations for Deep Learning in NLP. In \u003cem\u003eACL Anthology\u003c/em\u003e. https://doi.org/doi.org/10.18653/v1/P19-1355\u003c/li\u003e\n \u003cli\u003eSu, J., Yi, D., Coombes, M., Liu, C., Zhai, X., McDonald-Maier, K., \u0026amp; Chen, W. H. (2022). Spectral analysis and mapping of blackgrass weed by leveraging machine learning and UAV multispectral imagery. \u003cem\u003eComputers and Electronics in Agriculture\u003c/em\u003e, \u003cem\u003e192\u003c/em\u003e. https://doi.org/10.1016/j.compag.2021.106621\u003c/li\u003e\n \u003cli\u003eTate, E., Rahman, M. A., Emrich, C. T., \u0026amp; Sampson, C. C. (2021). Flood exposure and social vulnerability in the United States. \u003cem\u003eNatural Hazards\u003c/em\u003e, \u003cem\u003e106\u003c/em\u003e(1), 435\u0026ndash;457. https://doi.org/10.1007/s11069-020-04470-2\u003c/li\u003e\n \u003cli\u003eTehrani, A. A., Veisi, O., Fakhr, B. V., \u0026amp; Du, D. (2024). Predicting solar radiation in the urban area: A data-driven analysis for sustainable city planning using artificial neural networking. \u003cem\u003eSustainable Cities and Society\u003c/em\u003e, \u003cem\u003e100\u003c/em\u003e, 105042. https://doi.org/10.1016/J.SCS.2023.105042\u003c/li\u003e\n \u003cli\u003eToms, B. A., Barnes, E. A., \u0026amp; Ebert-Uphoff, I. (2020). Physically Interpretable Neural Networks for the Geosciences: Applications to Earth System Variability. \u003cem\u003eJournal of Advances in Modeling Earth Systems\u003c/em\u003e, \u003cem\u003e12\u003c/em\u003e(9). https://doi.org/10.1029/2019MS002002\u003c/li\u003e\n \u003cli\u003eTuia, D., Kellenberger, B., Beery, S., Costelloe, B. R., Zuffi, S., Risse, B., Mathis, A., Mathis, M. W., van Langevelde, F., Burghardt, T., Kays, R., Klinck, H., Wikelski, M., Couzin, I. D., van Horn, G., Crofoot, M. C., Stewart, C. V., \u0026amp; Berger-Wolf, T. (2022). Perspectives in machine learning for wildlife conservation. In \u003cem\u003eNature Communications\u003c/em\u003e (Vol. 13, Issue 1). Nature Research. https://doi.org/10.1038/s41467-022-27980-y\u003c/li\u003e\n \u003cli\u003eValavi, R., Guillera-Arroita, G., Lahoz-Monfort, J. J., \u0026amp; Elith, J. (2022). Predictive performance of presence-only species distribution models: a benchmark study with reproducible code. \u003cem\u003eEcological Monographs\u003c/em\u003e, \u003cem\u003e92\u003c/em\u003e(1). https://doi.org/10.1002/ecm.1486\u003c/li\u003e\n \u003cli\u003eVandal, T., Kodra, E., Ganguly, S., Michaelis, A., Nemani, R., \u0026amp; Ganguly, A. R. (2017). \u003cem\u003eDeepSD: Generating High Resolution Climate Change Projections through Single Image Super-Resolution\u003c/em\u003e. http://arxiv.org/abs/1703.03126\u003c/li\u003e\n \u003cli\u003eVaswani, A., Brain, G., Shazeer, N., Parmar, N., Uszkoreit, J., Jones, L., Gomez, A. N., Kaiser, Ł., \u0026amp; Polosukhin, I. (2023). \u003cem\u003eAttention Is All You Need\u003c/em\u003e.\u003c/li\u003e\n \u003cli\u003eVos, K., Deng, W., Harley, M. D., Turner, I. L., \u0026amp; Splinter, K. D. M. (2022). Beach-face slope dataset for Australia. \u003cem\u003eEarth System Science Data\u003c/em\u003e, \u003cem\u003e14\u003c/em\u003e(3), 1345\u0026ndash;1357. https://doi.org/10.5194/essd-14-1345-2022\u003c/li\u003e\n \u003cli\u003eWang, R., Ma, L., He, G., Johnson, B. A., Yan, Z., Chang, M., \u0026amp; Liang, Y. (2024). Transformers for Remote Sensing: A Systematic Review and Analysis. In \u003cem\u003eSensors\u003c/em\u003e (Vol. 24, Issue 11). Multidisciplinary Digital Publishing Institute (MDPI). https://doi.org/10.3390/s24113495\u003c/li\u003e\n \u003cli\u003eWang, Y., Dong, P., Hu, W., Chen, G., Zhang, D., Chen, B., \u0026amp; Lei, G. (2022). Modeling the Climate Suitability of Northernmost Mangroves in China under Climate Change Scenarios. \u003cem\u003eForests\u003c/em\u003e, \u003cem\u003e13\u003c/em\u003e(1). https://doi.org/10.3390/f13010064\u003c/li\u003e\n \u003cli\u003eWohlin, C. (2014). Guidelines for snowballing in systematic literature studies and a replication in software engineering. \u003cem\u003eACM International Conference Proceeding Series\u003c/em\u003e. https://doi.org/10.1145/2601248.2601268\u003c/li\u003e\n \u003cli\u003eWulder, M. A., Coops, N. C., Roy, D. P., White, J. C., \u0026amp; Hermosilla, T. (2018). Land cover 2.0. In \u003cem\u003eInternational Journal of Remote Sensing\u003c/em\u003e (Vol. 39, Issue 12, pp. 4254\u0026ndash;4284). Taylor and Francis Ltd. https://doi.org/10.1080/01431161.2018.1452075\u003c/li\u003e\n \u003cli\u003eXiang, Z., Yan, J., \u0026amp; Demir, I. (2020). A Rainfall-Runoff Model With LSTM-Based Sequence-to-Sequence Learning. \u003cem\u003eWater Resources Research\u003c/em\u003e, \u003cem\u003e56\u003c/em\u003e(1). https://doi.org/10.1029/2019WR025326\u003c/li\u003e\n \u003cli\u003eYang, C., Yu, M., Hu, F., Jiang, Y., \u0026amp; Li, Y. (2017). Utilizing Cloud Computing to address big geospatial data challenges. \u003cem\u003eComputers, Environment and Urban Systems\u003c/em\u003e, \u003cem\u003e61\u003c/em\u003e, 120\u0026ndash;128. https://doi.org/10.1016/j.compenvurbsys.2016.10.010\u003c/li\u003e\n \u003cli\u003eYang, W., Yuan, C. J., Tong, C., Yang, P., Yang, L., \u0026amp; Huang, B. (2017). Diurnal variation of CO2, CH4, and N2O emission fluxes continuously monitored in-situ in three environmental habitats in a subtropical estuarine wetland. \u003cem\u003eMarine Pollution Bulletin\u003c/em\u003e, \u003cem\u003e119\u003c/em\u003e(1), 289\u0026ndash;298. https://doi.org/10.1016/j.marpolbul.2017.04.005\u003c/li\u003e\n \u003cli\u003eYou, J., Li, X., Low, M., Lobell, D., \u0026amp; Ermon, S. (2017, February). Deep Gaussian Process for Crop Yield Prediction Based on Remote Sensing Data. \u003cem\u003eACM Digital Library\u003c/em\u003e. www.aaai.org\u003c/li\u003e\n \u003cli\u003eYu, Y., Si, X., Hu, C., \u0026amp; Zhang, J. (2019). A Review of Recurrent Neural Networks: LSTM Cells and Network Architectures. \u003cem\u003eNeural Computation\u003c/em\u003e, \u003cem\u003e31\u003c/em\u003e, 1235\u0026ndash;1270. https://doi.org/10.1162/neco_a_01199\u003c/li\u003e\n \u003cli\u003eZhang, E., Liu, L., Huang, L., \u0026amp; Ng, K. S. (2021). An automated, generalized, deep-learning-based method for delineating the calving fronts of Greenland glaciers from multi-sensor remote sensing imagery. \u003cem\u003eRemote Sensing of Environment\u003c/em\u003e, \u003cem\u003e254\u003c/em\u003e, 112265. https://doi.org/10.1016/J.RSE.2020.112265\u003c/li\u003e\n \u003cli\u003eZhang, L., Zhang, L., \u0026amp; Du, B. (2016). Deep learning for remote sensing data: A technical tutorial on the state of the art. \u003cem\u003eIEEE Geoscience and Remote Sensing Magazine\u003c/em\u003e, \u003cem\u003e4\u003c/em\u003e(2), 22\u0026ndash;40. https://doi.org/10.1109/MGRS.2016.2540798\u003c/li\u003e\n \u003cli\u003eZhang, Z., Liu, Q., \u0026amp; Wang, Y. (2018). Road Extraction by Deep Residual U-Net. \u003cem\u003eIEEE Geoscience and Remote Sensing Letters\u003c/em\u003e, \u003cem\u003e15\u003c/em\u003e(5), 749\u0026ndash;753. https://doi.org/10.1109/LGRS.2018.2802944\u003c/li\u003e\n \u003cli\u003eZhu, X. X., Tuia, D., Mou, L., Xia, G.-S., Zhang, L., Xu, F., \u0026amp; Fraundorfer, F. (2017). \u003cem\u003eDeep learning in remote sensing: a review\u003c/em\u003e. https://doi.org/10.1109/MGRS.2017.2762307\u003c/li\u003e\n \u003cli\u003eZupic, I., \u0026amp; Čater, T. (2015). Bibliometric Methods in Management and Organization. \u003cem\u003eOrganizational Research Methods\u003c/em\u003e, \u003cem\u003e18\u003c/em\u003e(3), 429\u0026ndash;472. https://doi.org/10.1177/1094428114562629\u003c/li\u003e\n\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":true,"hideJournal":true,"highlight":"","institution":"Hong Kong Polytechnic University","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true},"keywords":"GeoAI, Deep Learning, Remote Sensing, Scientometric Analysis, Systematic Review, Explainable AI","lastPublishedDoi":"10.21203/rs.3.rs-8913236/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-8913236/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eThe challenge of anthropogenic climate change requires a paradigm shift in analysis, moving beyond traditional modeling constraints toward the application of computational intelligence. Geospatial Artificial Intelligence (GeoAI), an emergent interdisciplinary domain synthesizing spatial science, artificial intelligence (AI), and statistical learning, can reshape the epistemology of climate change impact assessment. Although previous reviews have examined discrete elements of geospatial technologies or AI in environmental science, a comprehensive, critical synthesis that captures the methodological and application-based convergence of GeoAI and climate change is rare. This study analyzes GeoAI applications in climate change science, employing a dual scientometric and systematic review methodology. A scientometric analysis of 152 core publications, retrieved from Scopus, Web of Science, IEEE Xplore, and Google Scholar, was conducted to quantitatively elucidate publication trends, thematic networks, and the global collaborative landscape. This mapping reveals a rapidly accelerating, interdisciplinary field characterized by robust clusters around deep learning architectures, multi-modal remote sensing, and specific climate impact pathways. The review deconstructs the GeoAI technological stack from convolutional and recurrent neural networks to transformative transformer-based models and hybrid physics-AI frameworks. It also evaluates their roles in climate variable prediction, extreme event attribution, ecosystem vulnerability diagnostics, and socio-economic impact modeling. The review identifies profound epistemological and practical challenges, including data-centric limitations (heterogeneity, paucity), computational intractability, the \u0026ldquo;black box\u0026rdquo; conundrum, model transferability failures, and critical ethical imperatives. This review outlines a strategic prospective research agenda, suggesting that the evolution of scalable, interpretable, and context-aware GeoAI systems is crucial for constructing a resilient and equitable planetary future.\u003c/p\u003e","manuscriptTitle":"Driven geospatial artificial intelligence modeling for climate change impact assessment: a global perspective","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2026-02-24 10:20:31","doi":"10.21203/rs.3.rs-8913236/v1","editorialEvents":[{"type":"communityComments","content":0}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"a10a8e56-408c-4cd1-b561-54001c74054c","owner":[],"postedDate":"February 24th, 2026","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[{"id":63166194,"name":"Planetary Science"},{"id":63166195,"name":"Environmental Engineering"}],"tags":[],"updatedAt":"2026-02-24T10:20:31+00:00","versionOfRecord":[],"versionCreatedAt":"2026-02-24 10:20:31","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-8913236","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-8913236","identity":"rs-8913236","version":["v1"]},"buildId":"XKTyCvWXoU3ODBz1xrDgd","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}