Consensus land-cover mapping improves grassland classification in European mountain landscapes

Consensus land-cover mapping improves grassland classification in European mountain landscapes | 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 Article Consensus land-cover mapping improves grassland classification in European mountain landscapes Šimon Opravil, Matthias Baumann, Tomáš Goga, Hamid Afzali, Tobias Kuemmerle, and 1 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-8096101/v1 This work is licensed under a CC BY 4.0 License Status: Published Journal Publication published 10 Feb, 2026 Read the published version in Scientific Reports → Version 1 posted 10 You are reading this latest preprint version Abstract Accurate land-cover information is essential for biodiversity monitoring, yet existing 10-m global and continental land-cover datasets vary in accuracy and thematic consistency, particularly for grasslands in complex mountain environments. We assessed six state-of-the-art land-cover products (Dynamic World, ESA WorldCover, Esri Land Cover, Corine Land Cover + Backbone, ELC10, S2GLC) across the Alps and Carpathians and developed three consensus maps using weighted voting, accuracy-confusion weighting, and an accuracy-weighted Random Forest ensemble. All datasets were validated against an independent set of expert-interpreted reference samples. Individual products showed large discrepancies in grassland extent, elevation distribution, and landscape structure. Global datasets (Dynamic World, Esri Land Cover) underestimated grassland extent, whereas ESA WorldCover and Corine Land Cover + reported higher proportions. Consensus approaches substantially reduced these inconsistencies. The Random Forest ensemble achieved the highest accuracy (overall 90–92%), outperforming individual datasets and improving both user’s and producer’s accuracies for grassland (> 84%). Consensus datasets also better captured expected elevation and slope gradients, producing more spatially coherent and ecologically realistic grassland patterns. By integrating multiple land-cover sources, consensus approaches effectively mitigated dataset-specific biases and increased the reliability of grassland mapping in heterogeneous mountain systems. Consequently, consensus land-cover products provide a robust and ecologically meaningful alternative to single-source datasets for environmental assessments in complex mountain regions. Biological sciences/Ecology Earth and environmental sciences/Ecology Earth and environmental sciences/Environmental sciences Land-cover mapping Consensus approach Grasslands Alps and Carpathians Earth observation Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Figure 6 1. Introduction Land cover mapping is key to biodiversity monitoring and environmental reporting because it provides spatially explicit indicators of ecosystem state and change. Land-cover datasets are used to describe broad patterns of land use, but also as critical predictors in species distribution models and ecosystem service assessments 1–3 . By linking mapped habitats to landscape metrics, land-cover information can be translated into indicators of fragmentation, habitat quality, and ecological connectivity 4–7 . However, the validity of land-cover derived indicators depends directly on the accuracy and thematic consistency of the maps 8,9 . This challenge is particularly evident in mountain ecosystems, which are among the most biodiverse regions, such as Alpine and Carpathian grasslands. These areas are considered biodiversity hotspots that have been shaped by centuries of traditional management, and are also linked to steep environmental gradients, complex climatic mosaics, and topographic heterogeneity 10–13 . To accurately capture them, land-cover maps are needed that reliably reflect the extent of grasslands, their distribution along steep environmental gradients, and landscape structure. Over the past decades, numerous land-cover datasets have been developed at regional to global scales. These products vary considerably in terms of thematic detail, spatial resolution, and temporal coverage. Improvements in spatial and temporal resolution have not always been matched by corresponding increases in thematic accuracy. Comparative studies have shown that new 10-m global land-cover products differ sharply in how they represent major classes. For example, recent assessments indicate that global datasets tend to underestimate the extent of grasslands. Dynamic World and Esri Land Cover report only 9–12% of the Alps and Carpathians as grassland, while continental products such as Corine Land Cover + and ESA WorldCover map 21–24%, and ELC10 and S2GLC yield intermediate values (11–21%) 14,15 . These inconsistencies leave users without clear guidance on which dataset is most reliable for ecological applications. On the other hand, this growing availability of concurrent land-cover datasets has opened opportunities to create consensus maps that combine multiple sources to exploit their strengths and offset individual weaknesses. Early approaches relied on majority voting or fuzzy logic, assigning degrees of class membership 16–18 . More sophisticated methods incorporate uncertainty and class-specific accuracy through the Dempster–Shafer theory or multi-criteria integration 19,20 , while conditional-probability models weight class assignments by reported accuracies 21 . Beyond these rule-based techniques, ensemble machine learning offers a flexible alternative, combining diverse classifiers to enhance predictive stability and reduce misclassification 22,23 . Despite its success in other domains, ensemble learning remains under-explored in land-cover harmonization, leaving open whether it can deliver more coherent and ecologically realistic consensus maps. In this study, we focus on the Alps and the Carpathians to examine how grasslands are across a suite of widely used land cover products and whether consensus approaches can improve their reliability. Specifically, we integrate multiple datasets containing grassland classes to generate consensus maps and evaluate their performance relative to individual products. In addition, we assess how these datasets capture grassland patterns along environmental gradients of elevation, slope, and aspect, and examine the implications of their differences for landscape metrics. Our research is guided by three questions: Does a consensus approach improve the accuracy of grassland mapping compared to individual land cover datasets? How consistently do different datasets map grasslands across variable terrain gradients? What's the influence of the different land cover datasets on the overall landscape characteristics as defined by the landscape metrics? 2. Data & Methods 2.1. Study areas The Alps and Carpathians are the two largest European mountain ranges, differing in environmental and socio-economic settings that shape grassland ecosystems. The Alps span seven countries with temperate to alpine climates and steep precipitation gradients, while the Carpathians stretch across eight countries with more continental conditions. Both regions contain mosaics of semi-natural grasslands managed by grazing and mowing but differ in management intensity and land-use history, from intensively used Alpine valleys to more extensive Carpathian systems shaped by collectivisation and post-socialist transitions. 2.2. Data 2.2.1. Base land cover datasets We integrated six state-of-the-art land-cover datasets to build a consensus land-cover map for the Alps and Carpathians: Dynamic World 24 (DW), ESA WorldCover 25 (ESA WC), Esri Land Cover 26 (ESRI LC), Corine Land Cover + Backbone (CLC+), ELC10 27 , and S2GLC 28 . All datasets provide 10 m spatial resolution and are derived from Sentinel-1 and Sentinel-2 observations, ensuring a consistent remote-sensing foundation. The selected products vary in scope and purpose, ranging from global (DW, ESA WC, ESRI LC) to continental (CLC+, ELC10, S2GLC). All employ extensive training data, advanced modelling approaches, and comparable thematic classification schemes (Table 1 – 2 ). While CLC + and S2GLC do not fully cover parts of Ukraine, each dataset contributes unique methodological strengths and weaknesses that together enable robust regional harmonisation (Supplementary S1). 2.2.2 Base datasets preprocessing We harmonised the land-cover datasets by clipping them to the region of interest and standardising land-cover classes to a common classification scheme (Table 2 .). We used the top-level land-cover classification nomenclature from the Land Use/Cover Area frame statistical Survey (LUCAS) 29 , which includes artificial land, arable land, forest, shrubs, grassland, bare land, water, and wetland. Additionally, we included a class for glaciers and permanent snow based on LUCAS subclass G50. The ELC10 and DW datasets already followed this classification scheme, and we standardized the classification schemes of the other datasets based on Table 2 . Additionally, we processed the DW datasets by applying the mode function to the multi-year dataset (2018–2019). This multi-year aggregation approach accounts for the seasonal variability inherent in DW’s temporal coverage, allowing for a more stable representation of persistent land-cover patterns. Consequently, it enhances the identification of long-term trends useful for distinguishing between cropland (both active and fallow) and grassland 14 . Table 1 Overview of base land-cover maps containing grassland classes used in our study. MMU refers to the minimal mapping unit. Dataset Name Abbreviation Generated by MMU (m2) Years Available Coverage Corine Land Cover + Backbone CLC+ Copernicus Land Monitoring Service 100 2018 Europe Dynamic Word DW Google 250 every Sentinel-2 image with cloud cover less than 35% global pan-European land cover map ELC10 Venter & Sydenham (2021) 100 2018 Europe Word Cover ESA ESA WC ESA 100 2020, 2021 global Esri Land Cover ESRI LC ESRI 250 2020 global Sentinel-2 Global Land Cover S2GLC Malinowski et al. 2020 100 2017 Europe 2.2.3. Sampling validation data We followed best practices for accuracy assessment 30 . Harmonised land-cover classes from the base maps defined sampling strata for validation pixels in each mountain region. To avoid bias, we used a single validation dataset stratified by an external land-cover map 31 , independent of the evaluated products but using identical class definitions. We targeted an overall accuracy standard error below 0.01, expecting higher user accuracy for artificial land, cropland, woodland, and water, and lower user accuracy for shrubland, grassland, bare land, wetland, and snow (Supplementary S2). Each land-cover class included at least 70 validation pixels, yielding 1,460 points in the Alps and 1,410 in the Carpathians (Supplementary S2). Three experts visually interpreted each point using 2018 high-resolution Google Earth and Sentinel-2 imagery. The final class was assigned by majority agreement or by a single expert when confident in their interpretation. Table 2 Legend harmonisation of base LC datasets Abbr. built crop forest shrub grass bare water wet snow Class Num. 1 2 3 4 5 6 7 8 9 LUCAS Artificial land Cropland Woodland Shrubland Grassland Bareland Water Wetlands Snow CLC+ Sealed Periodically herbaceous Woody Low-growing woody plants Permanent herbaceous Non and sparsely-vegetated, Lichens and mosses Water Snow and ice DW Built Area Crops Trees Scrub/Shrub Grass Bare Ground Water Flooded Vegetation Snow/Ice ELC10 Artificial land Cropland Woodland Shrubland Grassland Bareland Water Wetlands ESA WC Built-up Cropland Tree cover Shrubland Grassland Bare / sparse vegetation, Moss and lichen Permanent water bodies Herbaceous wetland Snow and Ice Esri Built Area Crops Trees Scrub/Shrub Grass Bare Ground Water Flooded Vegetation Snow/Ice S2GLC Artificial surfaces Cultivated areas Broadleaf, Coniferous tree cover Moors and heathland Herbaceous vegetation Natural material surfaces Water bodies Marshes, Peatbogs Permanent snow cover 2.3. Evaluation and Integration of Land-Cover Datasets 2.3.1. Accuracy assessment We validated each of the used land-cover datasets and the consensus dataset using the independent validation sample described in Section 2.2.3 . For each class in each map, we computed accuracy metrics including user accuracy, producer accuracy, overall accuracy, and the F1-score. Standard errors for these metrics were estimated using the equations provided by Stehman & Foody (2019, Eqs. 12–18) 32 . 2.3.2. Consensus approaches We tested three consensus mapping approaches to integrate multiple land-cover datasets. First, we implemented a weighted voting approach (Con_WV), where each pixel was assigned the class with the highest sum of weights, with the weights being the F1 scores for each class across all input datasets. To enhance this weighted voting approach, we introduced an accuracy-confusion weighing method (Con_AccCo), adapted from 21 . This method calculates a consensus probability for each class by summing correct classification probabilities (the diagonal elements of the error matrices) while subtracting misclassification probabilities (the off-diagonal elements), thereby emphasising classes that are more reliably detected across datasets. Finally, we developed an accuracy-weighted Random Forest model (Con_RF). This method integrates multiple land-cover datasets using F1-scores as explanatory variables, allowing each dataset to contribute proportionally based on its observed accuracy. Training data were derived from LUCAS Harmonised 29 and LUCAS Copernicus 33 and preprocessed to remove ambiguous or low-accuracy points. The Random Forest classifier (ee.Classifier.smileRandomForest) was trained in Google Earth Engine with 500 trees and default parameters. For a more detailed description of these approaches, refer to Supplementary Material S1. 2.4. Comparison of grasslands across the Alps and Carpathians We first quantified the total extent of grasslands, defined as the area of all grassland pixels, in the Alps, in the Carpathians, and separately for each country within these mountain ranges. We then examined how this extent varied along environmental gradients. For every pixel classified as grassland, we extracted elevation, slope, and aspect from the Copernicus 30-m global DEM, resampled to 10 m using nearest-neighbour resampling. To assess how grassland landscape structure differs among datasets, we compared the composition and configuration of landscapes in the Alps and Carpathians using a set of landscape metrics. Landscapes were selected by generating a regular point grid at 5-km spacing and drawing a circular buffer of 2500 m around each point. Within each buffer, we calculated metrics that describe grassland patch patterns in relation to the area and shape of grassland patches. The metrics include mean grassland patch area, total grassland area, patch fractal dimension (PAFRAC), and edge density. 3. Results 3.1 Comparison of individual land-cover maps and consensus map quality The accuracy of individual base land-cover datasets varied considerably, reflecting differences in classification methods and input data (Figure. 2). In the Alps, overall accuracy ranged from 70% to 87%, with ESA WC (86%) and CLC+ (87%) performing best, while S2GLC (70%) and ESRI LC (82%) had the lowest accuracy among the evaluated datasets (Fig. 2 A). Accuracy followed a similar pattern in the Carpathians, with ESA WC reaching 90% and CLC + and ELC10 (exceeding 87%). Both Con_WV and Con_AccCo improved overall classification accuracy relative to individual datasets, but Con_RF consistently achieved the highest accuracy, with an overall accuracy of 90% in the Alps and 92% in the Carpathians (Fig. 2 A). For grassland areas, the greatest improvement in the accuracy for this class was observed in Con_RF, where both user accuracy and producer accuracy exceeded 84% in the Alps and Carpathians (Fig. 2 B, C). Con_RF and CLC + emerged as the top-performing maps for grassland classification, with both datasets achieving user and producer accuracies above 80%. In contrast, Con_AccCo and Con_Votes demonstrated slightly higher omission errors by underestimating the grassland areas, which led to reduced user accuracy. Among the original datasets, only CLC + outperformed Con_RF in terms of user accuracy in the Alps. However, when considering producer accuracy, both datasets showed very similar performance (0.872 and 0.874 F1-scores for CLC + and Con_RF, respectively). 3.2. Inter-Dataset Differences in Grassland Extent and Distribution The comparison of land-cover datasets revealed clear discrepancies in mapped proportions of grasslands across countries in the Alps and Carpathians (Fig. 4 ). DW and ESRI LC consistently reported the lowest grassland coverage, whereas ESA WC and CLC + showed the highest proportions. ELC10 produced intermediate values overall, but substantially higher grassland shares in the northwestern Alps (Germany) compared to all other products. Consensus products generally reduced low and high proportions but still differed in magnitude. Con_RF indicated higher grassland proportions than Con_AccCon and Con_WV, aligning more closely with ESA WC. The distribution of grasslands according to elevation, slope, and aspect gradients differs between individual map sources, which were more pronounced in the Alps than in the Carpathians (Fig. 5 A). CLC+, ESA WC, and S2GLC covered the entire spectrum from secondary grasslands in lowlands to the alpine grasslands above 2,500 m a.s.l., while DW, ESRI LC and ELC10 mapped primarily lower elevation grasslands, with considerably lower representation in high mountain areas. Consensus products (Con_RF, Con_AccCo, Con_WV) produced more consistent spatial patterns but still varied in their depiction of high-elevation grasslands. Median values differed in the Alps, with a spread of 900 m between the lowest median (ELC10–810 m a.s.l.) and the highest ESA WC (1710 m a.s.l.). In the Carpathians, the differences between datasets were less pronounced in this respect, with all sources showing a distribution concentrated at lower elevations (with a median of 420 m a.s.l. for CLC + to 540 m a.s.l. for ESRI LC) and alpine locations almost absent. Slope distribution and orientation (Fig. 5 B) revealed less pronounced differences between the individual datasets as compared to elevation. In the Alps, CLC+, ELC10, and ESA WorldCover consistently mapped grasslands on gentler terrain, reflecting their tendency to classify grasslands on low slopes. In contrast, DW, ESRI LC, and S2GLC mapped grasslands to steeper slopes, especially in the upper mountain and subalpine zones. In the Carpathians, most grasslands occurred on gentle terrain below 20°, corresponding to the extensive valley and plateau systems typical of this region. The narrower range of slopes compared to the Alps partly explains the reduced variability among datasets, as steep alpine meadows are almost absent in the Carpathian environment. Differences in aspect distribution were the least pronounced yet still noticeable (Fig. 5 C). In the Alps, CLC+, ELC10, and ESA WorldCover showed a preference for east to south-facing slopes, which corresponds to the climatic conditions of these exposures for productive grasslands. DW, ESRI LC and S2GLC captured similar orientations but placed greater emphasis on south and southeast orientations. In contrast, the Carpathian datasets exhibited a broader and more even distribution of aspects, reflecting the region’s gentler relief and lower topographic variability. 3.3. Influence of Land-Cover Dataset Characteristics on Grassland Spatial Structure Landscape metrics reveal consistent structural patterns across datasets. ESRI LC and DW, with large mean patch areas, show low edge density and low shape complexity (PAFRAC), indicating spatially aggregated, contiguous, and geometrically simple grassland patches with limited but consolidated extent. Conversely, ELC10 and S2GLC, characterized by small mean patch areas, exhibit high edge density and elevated PAFRAC, reflecting fragmented, irregular, and heterogeneous landscapes. Their larger total class areas, concentrated mainly in the Carpathians, result from spatial dispersion rather than continuous coverage. Intermediate configurations (e.g., ESA WC, CLC+, and consensus datasets) occupy a transitional position between these extremes. Moderate mean patch areas correspond with intermediate edge density and PAFRAC, reflecting balanced spatial aggregation and geometric complexity. Within the consensus datasets, Con_RF exhibits slightly larger mean patch and class areas, together with moderate edge density, suggesting that ensemble averaging mitigates fragmentation and produces a more cohesive yet regionally nuanced depiction of grassland structure. 4. Discussion Accurate and detailed land-cover mapping is essential for synthesizing knowledge of ecosystems across large areas. This need is especially acute in mountainous regions, where complex topography and diverse management histories generate mosaics of heterogeneous landscapes 10 . Within this study, we demonstrated that the land cover datasets provided contrasting mapping of grasslands in the Alps and Carpathians, with possible implications for ecological analyses and biodiversity monitoring. Furthermore, we showed that integrating multiple land-cover products reduces dataset-specific biases and yields a more reliable composite map 14,15,31 . Our results confirm a central pattern reported for recent Sentinel-based land-cover datasets at 10-meter resolution: while overall accuracy can be high, performance across individual classes remains uneven, and substantial differences persist between datasets 15 . In line with previous cross-comparisons, we observe that global datasets differ most in their ability to map grasslands, shrublands, and wetlands, while showing stronger agreement for water, tree cover, built-up areas, and crops 14 . Our assessment also confirms these class-specific discrepancies where DW and ESRI LC systematically underestimate the area of grasslands compared to ESA WC and CLC + in both mountain systems. Similar biases have been documented globally, with ESA WC tending to overestimate grasslands, Esri overestimating shrublands, and DW exaggerating snow/ice 14 . These systematic tendencies likely explain the under and overestimations of grassland areas that we found in the Alps and Carpathians In our comparison, the regional datasets (ELC10, CLC+, S2GLC) performed relatively well, echoing their reported pan-European accuracies 27,28 . However, upon comparing the results, it is clear that the Alps and Carpathians pose particular challenges. The steep terrain, fine-scale mosaics, and spectral similarity between crops, grasslands, and shrubs exacerbate classification errors (Supplementary S4). Consequently, user and producer accuracies in our study deviated from continental assessments that documented persistent weaknesses in mapping herbaceous vegetation 34 . Consensus approaches substantially improve the reliability of grassland mapping in complex mountain environments. The observed increase in their mapping accuracy highlights the ability of ensemble integration to produce a more dependable map of grassland cover. While regional datasets such as CLC + and ELC10 already performed relatively well, our results show that combining multiple sources reduces class-specific biases and produces depictions of grasslands that are both more robust and ecologically consistent. This supports previous calls for ensemble or weighted-evidence strategies to increase the reliability of land cover information 19,35,36 . Nonetheless, consensus maps remain constrained by the quality and availability of the underlying datasets. For example, the lack of full coverage of the Carpathian Mountains in CLC+ (e.g., Ukraine) and the partial availability of S2GLC limited the potential use within this mountain range. Moreover, harmonising the consensus map with the LUCAS nomenclature thematic detail leads to inconsistencies in wetland or snow classes across products. These compromises are in line with other harmonisation efforts, highlighting that consensus approaches may reduce uncertainty, but cannot fully eliminate structural limitations in land-cover data 37,38 . The selection of land-cover datasets strongly influences the evaluation of environmental gradients and landscape metrics (Fig. 5 , 6 ). We documented that on grasslands, where their elevation distribution revealed strong divergences. Some datasets reproduced the expected bimodal pattern of secondary lowland and alpine grasslands in the Alps, while others mapped larger extents of grasslands in low elevations. Such bias in the datasets may underestimate the high-elevation grasslands, although this cannot be directly validated in this study, given that our reference data were not stratified by elevation. Even so, we found that in the Carpathians, differences between datasets were less pronounced, which is likely due to the location of grasslands at lower and mid elevations. Slope distributions showed narrower variation than elevation but still revealed potential biases. ELC10 and CLC + emphasised gentler slopes, while DW and ESRI LC extended into steeper terrain. Aspect patterns were more subtle, with ESA WC and CLC + depicting grasslands more evenly across orientations, while DW and Esri concentrated them on south- and east-facing slopes. Selected landscape metrics reinforced these distinctions, as DW and ESRI LC mapped grasslands as aggregated patches due to their minimum mapping unit area, whereas other datasets depicted finer, more fragmented mosaics. Consensus maps reconciled these extremes, producing more ecologically plausible patch structures and retaining regional contrasts between the Alps and Carpathians. These improvements have important implications for downstream applications. Venter et al. (2023) 39 , for example, showed that Sentinel-based land-cover maps improved species distribution models for solitary bees in Norway compared with manually digitised maps, largely because products such as ELC10 and ESA WC, with smaller minimum mapping units, better resolved fine-scale habitat patches. Natsukawa et al. (2024) 40 demonstrated that integrating fine-scale land cover data with topographic information substantially enhances habitat models, yielding significant insights for conservation. Our findings suggest that relying on a single dataset for similar broad-scale analyses may yield markedly different results, depending on the specific product chosen. By integrating multiple land-cover datasets, our consensus map mitigated these distortions, preserving ecologically relevant gradients while reducing outliers from individual datasets. This suggests that consensus approaches can provide land-cover maps that are ecologically plausible, are methodologically robust, and are more suitable for downstream applications. These include biodiversity modelling 21,39 , habitat quality assessments, connectivity analyses, and quantification of land use intensity in grassland systems. In this context, ensemble products also improve the comparability of studies that rely on land cover as a foundational variable. Relying on a single land-cover dataset risks systematic bias, whereas consensus products enhance the comparability and credibility of evidence. In this sense, ensemble approaches improve the practical utility of land cover information in heterogeneous mountain landscapes. 5. Conclusion Accurate land cover maps are fundamental for broad-scale monitoring of the earth-systems particularly in heterogeneous mountain systems. Thanks to the technical availabilities, there are multiple datasets that are candidates to reach sufficient accuracy. Our analysis revealed that state-of-the-art land-cover datasets, although derived from similar satellite sources, yield substantially different representations of grasslands in the Alps and Carpathians, which in turn affects their extent and spatial configuration. These discrepancies varied across gradients of elevation, slope, and aspect. By integrating multiple datasets through ensemble and weighted consensus approaches, we demonstrated a reduction of such discrepancies. The Random Forest consensus map achieved the highest overall and class-specific accuracies, outperforming individual datasets and producing an ecologically coherent map of grasslands in both mountain regions. This improvement results from combining the complementary strengths of mapping products, thereby mitigating dataset-specific biases and increasing robustness in complex terrain. Beyond accuracy gains, the consensus approach preserved key ecological gradients and generated more realistic spatial patterns of grassland distribution and fragmentation. Such qualitative improvements may be particularly valuable for downstream applications that rely on land-cover inputs. While it may directly impact the results of the spatial-based studies, such as habitat modelling, landscape connectivity analyses, and biodiversity monitoring, it may also indirectly influence the policy documents that shape the future biodiversity strategies. Declarations Ethics declarations The expert annotations used in this study were produced by the co‑authors Šimon Opravil, Tomáš Goga, and Hamid Afzali. Each co‑author provided informed consent for the use of their annotations in this manuscript and in any related supplementary materials. No personal or sensitive information was collected; the annotations concern land‑cover categories only. Competing interests The authors declare no competing interests. Funding Declaration This research was funded by the project ‘G4B: Grasslands for biodiversity: supporting the protection of the biodiversity-rich grasslands and related management practices in the Alps and Carpathians’ funded by Biodiversa+, the European Biodiversity Partnership under the 2021–2022 BiodivProtect joint call for research proposals, co-funded by the European Commission (GA N°101052342) and with the funding organisations SNSF, DFG, NCN, PROV BZ, SAS and UEFISCDI; by the project of Slovak Research and Development Agency-21-0226: “Species-rich Carpathian grasslands: mapping, history, drivers of change and conservation” pursued at the Institute of Geography of the Slovak Academy of Sciences; by the Slovak Scientific Grant Agency VEGA under Grant 2/0043/23 “Detection of landscape diversity and its changes in Slovakia based on remote sensing data in the context of the European Green Deal.” Author Contribution Conceptualization, S.O., M.B., T.K., R.P; methodology, S.O., M.B., T.B., R.P.; software, S.O.; validation, S.O., T.G., H.A.; formal analysis, S.O. and R.P.; investigation, S.O., T.G., H.A., R.P., resources, S.O., T.G.; data curation, S.O.; writing—original draft preparation, S.O. and R.P.; writing—review and editing, M.B., T.G., H.A., T.K.; visualization, S.O.; supervision, R.P. and T.K.; project administration, R.P, and T.K.; funding acquisition, R.P. and T.K. All authors have read and agreed to the published version of the manuscript. Acknowledgements This study was made possible using multiple existing land cover datasets, including Google© Dynamic World, ESA WorldCover, ESRI© Land Cover, CLMS CLC+, ELC10, and S2GLC. We acknowledge Google, the European Space Agency, ESRI, the Copernicus Land Monitoring Service, and other data providers for making these datasets publicly available. All maps used in this study are accessible under their respective open data licenses. Data Availability Consensus datasets are openly available in the Zenodo repository at [https://doi.org/10.5281/zenodo.13823832](https:/doi.org/10.5281/zenodo.13823832) . The code used for data processing, analysis, and all other datasets stored in Google Earth Engine is publicly accessible in the GitHub repository at [https://github.com/simonopravil/MAC-Land](https:/github.com/simonopravil/MAC-Land) . References Cochran, F., Daniel, J., Jackson, L. & Neale, A. Earth observation-based ecosystem services indicators for national and subnational reporting of the sustainable development goals. Remote Sens. Environ. 244 , 111796 (2020). Pereira, H. M. et al. Essential Biodiversity Variables. Science 339 , 277–278 (2013). Timmermans, J. & Daniel Kissling, W. Advancing terrestrial biodiversity monitoring with satellite remote sensing in the context of the Kunming-Montreal global biodiversity framework. Ecol. Indic. 154 , 110773 (2023). Gaujour, E., Amiaud, B., Mignolet, C. & Plantureux, S. Factors and processes affecting plant biodiversity in permanent grasslands. A review. Agron. Sustain. Dev. 32 , 133–160 (2012). Scherreiks, P. et al. Present and historical landscape structure shapes current species richness in Central European grasslands. Landsc. Ecol. 37 , 745–762 (2022). Schindler, S., von Wehrden, H., Poirazidis, K., Wrbka, T. & Kati, V. Multiscale performance of landscape metrics as indicators of species richness of plants, insects and vertebrates. Ecol. Indic. 31 , 41–48 (2013). Turner, M. G. & Gardner, R. H. Landscape Metrics. in Landscape Ecology in Theory and Practice: Pattern and Process (eds Turner, M. G. & Gardner, R. H.) 97–142 (Springer, New York, NY, 2015). doi:10.1007/978-1-4939-2794-4_4. Huang, C., Geiger, E. L. & Kupfer, J. A. Sensitivity of landscape metrics to classification scheme. Int. J. Remote Sens. 27 , 2927–2948 (2006). Shao, G. & Wu, J. On the accuracy of landscape pattern analysis using remote sensing data. Landsc. Ecol. 23 , 505–511 (2008). Bezák, P. & Mitchley, J. Drivers of change in mountain farming in Slovakia: from socialist collectivisation to the Common Agricultural Policy. Reg. Environ. Change 14 , 1343–1356 (2014). Coelho, M. T. P. et al. The geography of climate and the global patterns of species diversity. Nature 622 , 537–544 (2023). Rahbek, C. et al. Humboldt’s enigma: What causes global patterns of mountain biodiversity? Science 365 , 1108–1113 (2019). Grasslands of the World: Diversity, Management and Conservation . (CRC Press, Boca Raton London New York, 2018). Venter, Z. S., Barton, D. N., Chakraborty, T., Simensen, T. & Singh, G. Global 10 m Land Use Land Cover Datasets: A Comparison of Dynamic World, World Cover and Esri Land Cover. Remote Sens. 14 , 4101 (2022). Wang, Y. et al. A review of regional and Global scale Land Use/Land Cover (LULC) mapping products generated from satellite remote sensing. ISPRS J. Photogramm. Remote Sens. 206 , 311–334 (2023). Iwao, K. et al. Creation of New Global Land Cover Map with Map Integration. J. Geogr. Inf. Syst. 03 , 160–165 (2011). Tsendbazar, N.-E., De Bruin, S., Fritz, S. & Herold, M. Spatial Accuracy Assessment and Integration of Global Land Cover Datasets. Remote Sens. 7 , 15804–15821 (2015). Pérez-Hoyos, A., García-Haro, F. J. & San-Miguel-Ayanz, J. A methodology to generate a synergetic land-cover map by fusion of different land-cover products. Int. J. Appl. Earth Obs. Geoinformation 19 , 72–87 (2012). Huang, A. et al. A Methodology to Generate Integrated Land Cover Data for Land Surface Model by Improving Dempster-Shafer Theory. Remote Sens. 14 , 972 (2022). Pérez-Hoyos, A., Udías, A. & Rembold, F. Integrating multiple land cover maps through a multi-criteria analysis to improve agricultural monitoring in Africa. Int. J. Appl. Earth Obs. Geoinformation 88 , 102064 (2020). Tuanmu, M. & Jetz, W. A global 1-km consensus land‐cover product for biodiversity and ecosystem modelling. Glob. Ecol. Biogeogr. 23 , 1031–1045 (2014). Aguilar, R., Zurita-Milla, R., Izquierdo-Verdiguier, E. & A. De By, R. A Cloud-Based Multi-Temporal Ensemble Classifier to Map Smallholder Farming Systems. Remote Sens. 10 , 729 (2018). Witjes, M. et al. A spatiotemporal ensemble machine learning framework for generating land use/land cover time-series maps for Europe (2000–2019) based on LUCAS, CORINE and GLAD Landsat. PeerJ 10 , e13573 (2022). Brown, C. F. et al. Dynamic World, Near real-time global 10 m land use land cover mapping. Sci. Data 9 , 251 (2022). Zanaga, D. et al. ESA WorldCover 10 m 2020 v100. Zenodo https://doi.org/10.5281/ZENODO.5571936 (2021). Karra, K. et al. Global land use / land cover with Sentinel 2 and deep learning. in 2021 IEEE International Geoscience and Remote Sensing Symposium IGARSS 4704–4707 (2021). doi:10.1109/IGARSS47720.2021.9553499. Venter, Z. S. & Sydenham, M. A. K. Continental-Scale Land Cover Mapping at 10 m Resolution Over Europe (ELC10). Remote Sens. 13 , 2301 (2021). Malinowski, R. et al. Automated Production of a Land Cover/Use Map of Europe Based on Sentinel-2 Imagery. Remote Sens. 12 , 3523 (2020). d’Andrimont, R. et al. Harmonised LUCAS in-situ land cover and use database for field surveys from 2006 to 2018 in the European Union. Sci. Data 7 , 352 (2020). Olofsson, P. et al. Good practices for estimating area and assessing accuracy of land change. Remote Sens. Environ. 148 , 42–57 (2014). Pflugmacher, D., Rabe, A., Peters, M. & Hostert, P. Mapping pan-European land cover using Landsat spectral-temporal metrics and the European LUCAS survey. Remote Sens. Environ. 221 , 583–595 (2019). Stehman, S. V. & Foody, G. M. Key issues in rigorous accuracy assessment of land cover products. Remote Sens. Environ. 231 , 111199 (2019). d’Andrimont, R. et al. LUCAS Copernicus 2018: Earth-observation-relevant in situ data on land cover and use throughout the European Union. Earth Syst. Sci. Data 13 , 1119–1133 (2021). Van Cleemput, E., Vanierschot, L., Fernández-Castilla, B., Honnay, O. & Somers, B. The functional characterization of grass- and shrubland ecosystems using hyperspectral remote sensing: trends, accuracy and moderating variables. Remote Sens. Environ. 209 , 747–763 (2018). Herold, M., See, L., Tsendbazar, N.-E. & Fritz, S. Towards an Integrated Global Land Cover Monitoring and Mapping System. Remote Sens. 8 , 1036 (2016). Kerner, H. et al. How accurate are existing land cover maps for agriculture in Sub-Saharan Africa? Sci. Data 11 , 486 (2024). Jansen, L. J. M., Groom, G. & Carrai, G. Land-cover harmonisation and semantic similarity: some methodological issues. J. Land Use Sci. 3 , 131–160 (2008). Xu, Q., Yordanov, V., Bruzzone, L. & Brovelli, M. A. High-Resolution Global Land Cover Maps and Their Assessment Strategies. ISPRS Int. J. Geo-Inf. 14 , 235 (2025). Venter, Z. S. et al. Comparing Global Sentinel-2 Land Cover Maps for Regional Species Distribution Modeling. Remote Sens. 15 , 1749 (2023). Natsukawa, H. et al. Importance of the interplay between land cover and topography in modeling habitat selection. Ecol. Indic. 169 , 112896 (2024 Additional Declarations No competing interests reported. Supplementary Files Supplementary.docx Cite Share Download PDF Status: Published Journal Publication published 10 Feb, 2026 Read the published version in Scientific Reports → Version 1 posted Editorial decision: Revision requested 05 Jan, 2026 Reviews received at journal 02 Jan, 2026 Reviews received at journal 02 Jan, 2026 Reviewers agreed at journal 06 Dec, 2025 Reviewers agreed at journal 05 Dec, 2025 Reviewers invited by journal 04 Dec, 2025 Editor assigned by journal 04 Dec, 2025 Editor invited by journal 28 Nov, 2025 Submission checks completed at journal 19 Nov, 2025 First submitted to journal 19 Nov, 2025 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-8096101","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Article","associatedPublications":[],"authors":[{"id":556206219,"identity":"d0257481-a185-49a9-9325-3f4342b43dd3","order_by":0,"name":"Šimon Opravil","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA7ElEQVRIiWNgGAWjYDACZiT2gQcVQFICzCRWS8IZYrSggMQ2IrTIt3MnPi5gsMkzbz9jeCBxXp28we3mB4w/zuDWYnCYd7PxDIa0YpkzOQYHErcdNtxw55gBM88NPFqYebdJ8zAcTgRqSwBqOcA4c0aCATPDBzwOa4Zp4X8G1DKnzn7mjPQPjD/waGE4DNMikXzgQGIDc2K/RI4BA16HgfzCY5BWLCHx+MCBhGOHk/tlzhQc5sHjffn+sxsf81TY5EnwJzZ/+FBTZ9sm3b7x4Y9jeBwGsYshAYV/gJAGEEggqGIUjIJRMApGLgAAXNxVCFzWBxsAAAAASUVORK5CYII=","orcid":"","institution":"Slovak Academy of Sciences","correspondingAuthor":true,"prefix":"","firstName":"Šimon","middleName":"","lastName":"Opravil","suffix":""},{"id":556206220,"identity":"d085b521-fb51-4bea-94d1-8842d900661b","order_by":1,"name":"Matthias Baumann","email":"","orcid":"","institution":"Humboldt-Universität zu Berlin","correspondingAuthor":false,"prefix":"","firstName":"Matthias","middleName":"","lastName":"Baumann","suffix":""},{"id":556206221,"identity":"2c5aed2d-0f77-48fe-8b79-10881510d86e","order_by":2,"name":"Tomáš Goga","email":"","orcid":"","institution":"Slovak Academy of Sciences","correspondingAuthor":false,"prefix":"","firstName":"Tomáš","middleName":"","lastName":"Goga","suffix":""},{"id":556206222,"identity":"c945b885-3d5f-4425-aaa8-65f57f4400bf","order_by":3,"name":"Hamid Afzali","email":"","orcid":"","institution":"Slovak Academy of Sciences","correspondingAuthor":false,"prefix":"","firstName":"Hamid","middleName":"","lastName":"Afzali","suffix":""},{"id":556206223,"identity":"ff94479c-edec-4fda-8e97-191d3fa55c2f","order_by":4,"name":"Tobias Kuemmerle","email":"","orcid":"","institution":"Humboldt-Universität zu Berlin","correspondingAuthor":false,"prefix":"","firstName":"Tobias","middleName":"","lastName":"Kuemmerle","suffix":""},{"id":556206225,"identity":"0cd46e11-47d5-4d7a-a7b7-45aaf6782159","order_by":5,"name":"Róbert Pazúr","email":"","orcid":"","institution":"Slovak Academy of Sciences","correspondingAuthor":false,"prefix":"","firstName":"Róbert","middleName":"","lastName":"Pazúr","suffix":""}],"badges":[],"createdAt":"2025-11-12 12:08:20","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-8096101/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-8096101/v1","draftVersion":[],"editorialEvents":[{"content":"https://doi.org/10.1038/s41598-026-39197-w","type":"published","date":"2026-02-10T15:57:10+00:00"}],"editorialNote":"","failedWorkflow":false,"files":[{"id":97708411,"identity":"27fcea54-c0a0-4b84-aac2-5716587891dc","added_by":"auto","created_at":"2025-12-08 13:16:37","extension":"png","order_by":0,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":3952820,"visible":true,"origin":"","legend":"","description":"","filename":"Figure1.png","url":"https://assets-eu.researchsquare.com/files/rs-8096101/v1/2d0fddb26d15cde9bcc9df0d.png"},{"id":97894059,"identity":"789ed16f-6cee-42df-b477-ec729f143da5","added_by":"auto","created_at":"2025-12-10 15:31:53","extension":"docx","order_by":1,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":8108657,"visible":true,"origin":"","legend":"","description":"","filename":"MACLand.docx","url":"https://assets-eu.researchsquare.com/files/rs-8096101/v1/8014d70c0bc663966c7d2eae.docx"},{"id":97893372,"identity":"045b2ec6-fb12-40ef-8bf8-ad3017778083","added_by":"auto","created_at":"2025-12-10 15:30:13","extension":"png","order_by":2,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":163065,"visible":true,"origin":"","legend":"","description":"","filename":"Figure2.png","url":"https://assets-eu.researchsquare.com/files/rs-8096101/v1/23670d5303aac5137682ab2d.png"},{"id":97708412,"identity":"d98fe2dc-4ca9-4304-82e1-0f6cd513df80","added_by":"auto","created_at":"2025-12-08 13:16:37","extension":"png","order_by":3,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":2255487,"visible":true,"origin":"","legend":"","description":"","filename":"Figure3.png","url":"https://assets-eu.researchsquare.com/files/rs-8096101/v1/383c57d01671bdceb16e7f56.png"},{"id":97708416,"identity":"ba70d4f6-6ac7-49a2-825a-e2798b4ab78e","added_by":"auto","created_at":"2025-12-08 13:16:37","extension":"png","order_by":4,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":325548,"visible":true,"origin":"","legend":"","description":"","filename":"Figure4.png","url":"https://assets-eu.researchsquare.com/files/rs-8096101/v1/ee86f5b608438f3e2c86f8e3.png"},{"id":97895114,"identity":"12ae8a37-ed25-4824-a353-85f1584e5f31","added_by":"auto","created_at":"2025-12-10 15:33:36","extension":"png","order_by":5,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":617090,"visible":true,"origin":"","legend":"","description":"","filename":"Figure5.png","url":"https://assets-eu.researchsquare.com/files/rs-8096101/v1/f5273e1b4f7df2eaf537dafa.png"},{"id":97893395,"identity":"7cb7d1a7-1661-4d97-b5cb-371df07d9a3e","added_by":"auto","created_at":"2025-12-10 15:30:21","extension":"png","order_by":6,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":188139,"visible":true,"origin":"","legend":"","description":"","filename":"Figure6.png","url":"https://assets-eu.researchsquare.com/files/rs-8096101/v1/d951c2c86bf0a4b33723a25d.png"},{"id":97708419,"identity":"02780468-5bf8-4b7d-8574-897ae219fa54","added_by":"auto","created_at":"2025-12-08 13:16:37","extension":"json","order_by":7,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":8088,"visible":true,"origin":"","legend":"","description":"","filename":"3e428c5f187b4fefbd9fff0858d8cb47.json","url":"https://assets-eu.researchsquare.com/files/rs-8096101/v1/bbb5dbdd7e53c5cb70c42a69.json"},{"id":97895260,"identity":"e8f574a7-d98b-4881-a7dd-529608ea0cc7","added_by":"auto","created_at":"2025-12-10 15:33:54","extension":"docx","order_by":8,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":5727462,"visible":true,"origin":"","legend":"","description":"","filename":"Supplementary.docx","url":"https://assets-eu.researchsquare.com/files/rs-8096101/v1/9a3d0361de939f81785aa81b.docx"},{"id":97894200,"identity":"008846d8-8d3a-4594-b940-a4b799d0d311","added_by":"auto","created_at":"2025-12-10 15:32:02","extension":"xml","order_by":9,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":87522,"visible":true,"origin":"","legend":"","description":"","filename":"3e428c5f187b4fefbd9fff0858d8cb471enriched.xml","url":"https://assets-eu.researchsquare.com/files/rs-8096101/v1/8a295821d06e356704318d1e.xml"},{"id":97708422,"identity":"03cefe88-38a0-43b6-ac05-e1660158a8c5","added_by":"auto","created_at":"2025-12-08 13:16:37","extension":"png","order_by":10,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":3952820,"visible":true,"origin":"","legend":"","description":"","filename":"Figure1.png","url":"https://assets-eu.researchsquare.com/files/rs-8096101/v1/e16215f518a3d5c313bb99ec.png"},{"id":97895241,"identity":"9dbb1357-d95a-4a31-aa28-d67d7a218e56","added_by":"auto","created_at":"2025-12-10 15:33:52","extension":"png","order_by":11,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":163065,"visible":true,"origin":"","legend":"","description":"","filename":"Figure2.png","url":"https://assets-eu.researchsquare.com/files/rs-8096101/v1/134bd03c12e420a54865bc26.png"},{"id":97708428,"identity":"eceaf536-0dd4-4138-9f3f-508c80255913","added_by":"auto","created_at":"2025-12-08 13:16:37","extension":"png","order_by":12,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":2255487,"visible":true,"origin":"","legend":"","description":"","filename":"Figure3.png","url":"https://assets-eu.researchsquare.com/files/rs-8096101/v1/2c8c0dfcc7e1fca20fd27ca5.png"},{"id":97708424,"identity":"4ece63fd-d951-4e75-9d32-19c18b398890","added_by":"auto","created_at":"2025-12-08 13:16:37","extension":"png","order_by":13,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":325548,"visible":true,"origin":"","legend":"","description":"","filename":"Figure4.png","url":"https://assets-eu.researchsquare.com/files/rs-8096101/v1/427dc100d53dae327b00f67c.png"},{"id":97708423,"identity":"bf43fa29-9123-4b3d-b5d1-0e4ca98b1ce2","added_by":"auto","created_at":"2025-12-08 13:16:37","extension":"png","order_by":14,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":617090,"visible":true,"origin":"","legend":"","description":"","filename":"Figure5.png","url":"https://assets-eu.researchsquare.com/files/rs-8096101/v1/6ba0aa399ef48407493362a4.png"},{"id":97708426,"identity":"3f098178-2ece-4779-ae60-2257e4420075","added_by":"auto","created_at":"2025-12-08 13:16:37","extension":"png","order_by":15,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":188139,"visible":true,"origin":"","legend":"","description":"","filename":"Figure6.png","url":"https://assets-eu.researchsquare.com/files/rs-8096101/v1/9c8d498a24218bdcbd447146.png"},{"id":97895305,"identity":"91c0dd0c-14f8-4ff7-93e5-deb3dda1b1fd","added_by":"auto","created_at":"2025-12-10 15:33:57","extension":"png","order_by":16,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":3060743,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage1.png","url":"https://assets-eu.researchsquare.com/files/rs-8096101/v1/331496418ff40475bb538b07.png"},{"id":97895307,"identity":"e5c9e9bb-76d0-4ec1-b615-f5c7dd6f4d77","added_by":"auto","created_at":"2025-12-10 15:33:57","extension":"png","order_by":17,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":159504,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage2.png","url":"https://assets-eu.researchsquare.com/files/rs-8096101/v1/3c60f5ca94f2a15f8dfd2518.png"},{"id":97894214,"identity":"cb2a497d-3c81-4ef3-a496-8a65a77de072","added_by":"auto","created_at":"2025-12-10 15:32:03","extension":"png","order_by":18,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":2255487,"visible":true,"origin":"","legend":"","description":"","filename":"Figure3.png","url":"https://assets-eu.researchsquare.com/files/rs-8096101/v1/99bb0d84f3113292b4f3a721.png"},{"id":97893303,"identity":"5893b81e-319e-4495-929c-ab7222bb843a","added_by":"auto","created_at":"2025-12-10 15:29:48","extension":"png","order_by":19,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":276667,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage4.png","url":"https://assets-eu.researchsquare.com/files/rs-8096101/v1/947f96f5dcc940927ff01358.png"},{"id":97708445,"identity":"aea0c65e-e27e-4a3c-b68c-4ca422a91a90","added_by":"auto","created_at":"2025-12-08 13:16:37","extension":"png","order_by":20,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":478427,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage5.png","url":"https://assets-eu.researchsquare.com/files/rs-8096101/v1/a61c04f9f28e5e630760ae32.png"},{"id":97708434,"identity":"0d591a8e-015f-44e9-bd9d-6de8efa17899","added_by":"auto","created_at":"2025-12-08 13:16:37","extension":"png","order_by":21,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":159328,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage6.png","url":"https://assets-eu.researchsquare.com/files/rs-8096101/v1/687dc890c7ac475b1a78f7f2.png"},{"id":97893928,"identity":"07bc542b-323a-40cc-a67c-693f5ae961aa","added_by":"auto","created_at":"2025-12-10 15:31:33","extension":"png","order_by":22,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":462898,"visible":true,"origin":"","legend":"","description":"","filename":"OnlineFigure1.png","url":"https://assets-eu.researchsquare.com/files/rs-8096101/v1/0141bbdca8b906f631508add.png"},{"id":97895258,"identity":"281c339f-fc14-40d9-b0b6-76b3da08d857","added_by":"auto","created_at":"2025-12-10 15:33:54","extension":"png","order_by":23,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":37279,"visible":true,"origin":"","legend":"","description":"","filename":"OnlineFigure2.png","url":"https://assets-eu.researchsquare.com/files/rs-8096101/v1/59caae31844784f7f2b0a8a5.png"},{"id":97708441,"identity":"3d6b6293-7f9e-4a4d-8011-1de193581b0c","added_by":"auto","created_at":"2025-12-08 13:16:37","extension":"png","order_by":24,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":367165,"visible":true,"origin":"","legend":"","description":"","filename":"OnlineFigure3.png","url":"https://assets-eu.researchsquare.com/files/rs-8096101/v1/46808d90a3f61f4d97f770d5.png"},{"id":97708439,"identity":"46fb9a7a-ef2d-4ead-829b-75ae1888ce93","added_by":"auto","created_at":"2025-12-08 13:16:37","extension":"png","order_by":25,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":80338,"visible":true,"origin":"","legend":"","description":"","filename":"OnlineFigure4.png","url":"https://assets-eu.researchsquare.com/files/rs-8096101/v1/1630b87c9c5d8094c701640e.png"},{"id":97708430,"identity":"111aaea8-2ad4-4fc8-a211-5f98578525b2","added_by":"auto","created_at":"2025-12-08 13:16:37","extension":"png","order_by":26,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":128515,"visible":true,"origin":"","legend":"","description":"","filename":"OnlineFigure5.png","url":"https://assets-eu.researchsquare.com/files/rs-8096101/v1/75db39d717067949f5534298.png"},{"id":97708436,"identity":"ab757f9f-3b6a-43c1-8d46-06054ab09353","added_by":"auto","created_at":"2025-12-08 13:16:37","extension":"png","order_by":27,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":46587,"visible":true,"origin":"","legend":"","description":"","filename":"OnlineFigure6.png","url":"https://assets-eu.researchsquare.com/files/rs-8096101/v1/85ea4f2462548d43ac63687f.png"},{"id":97708447,"identity":"f9277636-387a-408d-af7b-9935d4edcfd7","added_by":"auto","created_at":"2025-12-08 13:16:37","extension":"png","order_by":28,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":462700,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage1.png","url":"https://assets-eu.researchsquare.com/files/rs-8096101/v1/e3e082de3256d9bb4ffa9fab.png"},{"id":97708442,"identity":"7a518a60-169e-42b0-9d93-b8c7abf7a066","added_by":"auto","created_at":"2025-12-08 13:16:37","extension":"png","order_by":29,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":39799,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage2.png","url":"https://assets-eu.researchsquare.com/files/rs-8096101/v1/4e4684880a01073cd7755a2f.png"},{"id":97893961,"identity":"bc529815-988f-46f1-95d7-8259e0c21b1a","added_by":"auto","created_at":"2025-12-10 15:31:41","extension":"png","order_by":30,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":367165,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage3.png","url":"https://assets-eu.researchsquare.com/files/rs-8096101/v1/614cd03bb901eafaa435fb5e.png"},{"id":97708437,"identity":"5eedba86-cc86-47f2-b659-5f24d3493f09","added_by":"auto","created_at":"2025-12-08 13:16:37","extension":"png","order_by":31,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":83494,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage4.png","url":"https://assets-eu.researchsquare.com/files/rs-8096101/v1/bcff2a5829d08e56d028b5a0.png"},{"id":97895341,"identity":"6f82a021-6414-40e9-a694-22b9352f1cfb","added_by":"auto","created_at":"2025-12-10 15:34:01","extension":"png","order_by":32,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":127025,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage5.png","url":"https://assets-eu.researchsquare.com/files/rs-8096101/v1/9188ef35ecc285fff69698a8.png"},{"id":97708443,"identity":"a740f7b6-1b86-4306-8dc0-30012e692857","added_by":"auto","created_at":"2025-12-08 13:16:37","extension":"png","order_by":33,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":49424,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage6.png","url":"https://assets-eu.researchsquare.com/files/rs-8096101/v1/505a4db04a5dce413ea0ced2.png"},{"id":97708444,"identity":"392833f8-f68f-4794-ac43-434648cbab70","added_by":"auto","created_at":"2025-12-08 13:16:37","extension":"xml","order_by":34,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":86173,"visible":true,"origin":"","legend":"","description":"","filename":"3e428c5f187b4fefbd9fff0858d8cb471structuring.xml","url":"https://assets-eu.researchsquare.com/files/rs-8096101/v1/bc9e4402d1880b68c5bdc321.xml"},{"id":97894860,"identity":"1314ad05-b294-4531-99cc-6f2faee597a3","added_by":"auto","created_at":"2025-12-10 15:33:10","extension":"html","order_by":35,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":95283,"visible":true,"origin":"","legend":"","description":"","filename":"earlyproof.html","url":"https://assets-eu.researchsquare.com/files/rs-8096101/v1/193e8224ad5a69906cf0040e.html"},{"id":97708408,"identity":"cde23a7c-57a0-4f7c-b474-dadd8f2be385","added_by":"auto","created_at":"2025-12-08 13:16:37","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":3952820,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cem\u003eStudy areas within Europe: the Alps and the Carpathians.\u003c/em\u003e\u003c/p\u003e","description":"","filename":"Figure1.png","url":"https://assets-eu.researchsquare.com/files/rs-8096101/v1/53ba4a9eafe0204d50820496.png"},{"id":97708407,"identity":"48f12d12-c3f9-49aa-a3ea-95fae4cd922f","added_by":"auto","created_at":"2025-12-08 13:16:36","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":163065,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cem\u003eStudy areas within Europe: the Alps and the Carpathians.\u003c/em\u003e\u003c/p\u003e","description":"","filename":"Figure2.png","url":"https://assets-eu.researchsquare.com/files/rs-8096101/v1/a2b220444d955fe476ab956b.png"},{"id":97894073,"identity":"d0cc18a5-5771-47e1-9483-594a86fda9ae","added_by":"auto","created_at":"2025-12-10 15:31:54","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":2255487,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cem\u003eComparison of the Con_RF consensus dataset with other land-cover datasets across the Alpine and Carpathian regions. (A) Con_RF land-cover map covering the Alps and Carpathians (B) Detailed example from the Alps (France) and (C) from the Carpathians (Romania)\u003c/em\u003e\u003c/p\u003e","description":"","filename":"Figure3.png","url":"https://assets-eu.researchsquare.com/files/rs-8096101/v1/7f9f0d5173011aec85e3da81.png"},{"id":97893370,"identity":"c00dd1b9-d4cf-448f-afbd-b6d4caee745f","added_by":"auto","created_at":"2025-12-10 15:30:13","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":325548,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cem\u003eCountry-level proportions of grasslands within the Alps and the Carpathians derived from different land-cover datasets.\u003c/em\u003e\u003c/p\u003e","description":"","filename":"Figure4.png","url":"https://assets-eu.researchsquare.com/files/rs-8096101/v1/3597ecf1f57a73eaaf0df670.png"},{"id":97895248,"identity":"44313da9-d8ad-41c4-8813-3d0fe4ee9cf5","added_by":"auto","created_at":"2025-12-10 15:33:52","extension":"png","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":617090,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cem\u003eTerrain distribution of grasslands across compared land-cover datasets in the Alps and Carpathians. (A) elevation, (B) slope, and (C) aspect\u003c/em\u003e\u003c/p\u003e","description":"","filename":"Figure5.png","url":"https://assets-eu.researchsquare.com/files/rs-8096101/v1/d9a4858e322fbf879c568516.png"},{"id":97895296,"identity":"fcc715e1-356c-421f-8ad4-de2a52e79e4d","added_by":"auto","created_at":"2025-12-10 15:33:57","extension":"png","order_by":6,"title":"Figure 6","display":"","copyAsset":false,"role":"figure","size":188139,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cem\u003eComparison of landscape metrics for grasslands across datasets and regions. Boxplots show the distribution of mean grassland patch area (log scale), total grassland area, edge density, and perimeter–area fractal dimension\u003c/em\u003e\u003c/p\u003e","description":"","filename":"Figure6.png","url":"https://assets-eu.researchsquare.com/files/rs-8096101/v1/4ca1bc5e3a1cc3b541648a8e.png"},{"id":102785491,"identity":"4bfaf78d-15dd-4dd8-a5fc-2d087aeac4dd","added_by":"auto","created_at":"2026-02-16 16:07:31","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":7634756,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-8096101/v1/f3f994e5-86d1-4b67-9e20-c349b99ecff9.pdf"},{"id":97894774,"identity":"873b064b-54f0-4d73-a87c-73c3b3efdac2","added_by":"auto","created_at":"2025-12-10 15:33:00","extension":"docx","order_by":0,"title":"","display":"","copyAsset":false,"role":"supplement","size":5727462,"visible":true,"origin":"","legend":"","description":"","filename":"Supplementary.docx","url":"https://assets-eu.researchsquare.com/files/rs-8096101/v1/eb263e152b8b1c718189977c.docx"}],"financialInterests":"No competing interests reported.","formattedTitle":"Consensus land-cover mapping improves grassland classification in European mountain landscapes","fulltext":[{"header":"1. Introduction","content":"\u003cp\u003eLand cover mapping is key to biodiversity monitoring and environmental reporting because it provides spatially explicit indicators of ecosystem state and change. Land-cover datasets are used to describe broad patterns of land use, but also as critical predictors in species distribution models and ecosystem service assessments\u003csup\u003e1\u0026ndash;3\u003c/sup\u003e. By linking mapped habitats to landscape metrics, land-cover information can be translated into indicators of fragmentation, habitat quality, and ecological connectivity\u003csup\u003e4\u0026ndash;7\u003c/sup\u003e. However, the validity of land-cover derived indicators depends directly on the accuracy and thematic consistency of the maps\u003csup\u003e8,9\u003c/sup\u003e.\u003c/p\u003e\u003cp\u003eThis challenge is particularly evident in mountain ecosystems, which are among the most biodiverse regions, such as Alpine and Carpathian grasslands. These areas are considered biodiversity hotspots that have been shaped by centuries of traditional management, and are also linked to steep environmental gradients, complex climatic mosaics, and topographic heterogeneity\u003csup\u003e10\u0026ndash;13\u003c/sup\u003e. To accurately capture them, land-cover maps are needed that reliably reflect the extent of grasslands, their distribution along steep environmental gradients, and landscape structure.\u003c/p\u003e\u003cp\u003eOver the past decades, numerous land-cover datasets have been developed at regional to global scales. These products vary considerably in terms of thematic detail, spatial resolution, and temporal coverage. Improvements in spatial and temporal resolution have not always been matched by corresponding increases in thematic accuracy. Comparative studies have shown that new 10-m global land-cover products differ sharply in how they represent major classes. For example, recent assessments indicate that global datasets tend to underestimate the extent of grasslands. Dynamic World and Esri Land Cover report only 9\u0026ndash;12% of the Alps and Carpathians as grassland, while continental products such as Corine Land Cover\u0026thinsp;+\u0026thinsp;and ESA WorldCover map 21\u0026ndash;24%, and ELC10 and S2GLC yield intermediate values (11\u0026ndash;21%)\u003csup\u003e14,15\u003c/sup\u003e. These inconsistencies leave users without clear guidance on which dataset is most reliable for ecological applications.\u003c/p\u003e\u003cp\u003eOn the other hand, this growing availability of concurrent land-cover datasets has opened opportunities to create consensus maps that combine multiple sources to exploit their strengths and offset individual weaknesses. Early approaches relied on majority voting or fuzzy logic, assigning degrees of class membership\u003csup\u003e16\u0026ndash;18\u003c/sup\u003e. More sophisticated methods incorporate uncertainty and class-specific accuracy through the Dempster\u0026ndash;Shafer theory or multi-criteria integration\u003csup\u003e19,20\u003c/sup\u003e, while conditional-probability models weight class assignments by reported accuracies\u003csup\u003e21\u003c/sup\u003e. Beyond these rule-based techniques, ensemble machine learning offers a flexible alternative, combining diverse classifiers to enhance predictive stability and reduce misclassification\u003csup\u003e22,23\u003c/sup\u003e. Despite its success in other domains, ensemble learning remains under-explored in land-cover harmonization, leaving open whether it can deliver more coherent and ecologically realistic consensus maps.\u003c/p\u003e\u003cp\u003eIn this study, we focus on the Alps and the Carpathians to examine how grasslands are across a suite of widely used land cover products and whether consensus approaches can improve their reliability. Specifically, we integrate multiple datasets containing grassland classes to generate consensus maps and evaluate their performance relative to individual products. In addition, we assess how these datasets capture grassland patterns along environmental gradients of elevation, slope, and aspect, and examine the implications of their differences for landscape metrics. Our research is guided by three questions:\u003c/p\u003e\u003cp\u003e\u003col\u003e\u003cspan\u003e\u003cli\u003e\u003cp\u003eDoes a consensus approach improve the accuracy of grassland mapping compared to individual land cover datasets?\u003c/p\u003e\u003c/li\u003e\u003c/span\u003e\u003cspan\u003e\u003cli\u003e\u003cp\u003eHow consistently do different datasets map grasslands across variable terrain gradients?\u003c/p\u003e\u003c/li\u003e\u003c/span\u003e\u003cspan\u003e\u003cli\u003e\u003cp\u003eWhat's the influence of the different land cover datasets on the overall landscape characteristics as defined by the landscape metrics?\u003c/p\u003e\u003c/li\u003e\u003c/span\u003e\u003c/ol\u003e\u003c/p\u003e"},{"header":"2. Data \u0026 Methods","content":"\u003cdiv id=\"Sec2\" class=\"Section2\"\u003e\u003ch2\u003e2.1. Study areas\u003c/h2\u003e\u003cp\u003eThe Alps and Carpathians are the two largest European mountain ranges, differing in environmental and socio-economic settings that shape grassland ecosystems. The Alps span seven countries with temperate to alpine climates and steep precipitation gradients, while the Carpathians stretch across eight countries with more continental conditions. Both regions contain mosaics of semi-natural grasslands managed by grazing and mowing but differ in management intensity and land-use history, from intensively used Alpine valleys to more extensive Carpathian systems shaped by collectivisation and post-socialist transitions.\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec3\" class=\"Section2\"\u003e\u003ch2\u003e2.2. Data\u003c/h2\u003e\u003cdiv id=\"Sec4\" class=\"Section3\"\u003e\u003ch2\u003e2.2.1. Base land cover datasets\u003c/h2\u003e\u003cp\u003eWe integrated six state-of-the-art land-cover datasets to build a consensus land-cover map for the Alps and Carpathians: Dynamic World\u003csup\u003e24\u003c/sup\u003e (DW), ESA WorldCover\u003csup\u003e25\u003c/sup\u003e (ESA WC), Esri Land Cover\u003csup\u003e26\u003c/sup\u003e (ESRI LC), Corine Land Cover\u0026thinsp;+\u0026thinsp;Backbone (CLC+), ELC10\u003csup\u003e27\u003c/sup\u003e, and S2GLC\u003csup\u003e28\u003c/sup\u003e. All datasets provide 10 m spatial resolution and are derived from Sentinel-1 and Sentinel-2 observations, ensuring a consistent remote-sensing foundation.\u003c/p\u003e\u003cp\u003eThe selected products vary in scope and purpose, ranging from global (DW, ESA WC, ESRI LC) to continental (CLC+, ELC10, S2GLC). All employ extensive training data, advanced modelling approaches, and comparable thematic classification schemes (Table\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e\u0026ndash;\u003cspan refid=\"Tab2\" class=\"InternalRef\"\u003e2\u003c/span\u003e). While CLC\u0026thinsp;+\u0026thinsp;and S2GLC do not fully cover parts of Ukraine, each dataset contributes unique methodological strengths and weaknesses that together enable robust regional harmonisation (Supplementary S1).\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec5\" class=\"Section3\"\u003e\u003ch2\u003e2.2.2 Base datasets preprocessing\u003c/h2\u003e\u003cp\u003eWe harmonised the land-cover datasets by clipping them to the region of interest and standardising land-cover classes to a common classification scheme (Table\u0026nbsp;\u003cspan refid=\"Tab2\" class=\"InternalRef\"\u003e2\u003c/span\u003e.). We used the top-level land-cover classification nomenclature from the Land Use/Cover Area frame statistical Survey (LUCAS)\u003csup\u003e29\u003c/sup\u003e, which includes artificial land, arable land, forest, shrubs, grassland, bare land, water, and wetland. Additionally, we included a class for glaciers and permanent snow based on LUCAS subclass G50. The ELC10 and DW datasets already followed this classification scheme, and we standardized the classification schemes of the other datasets based on Table\u0026nbsp;\u003cspan refid=\"Tab2\" class=\"InternalRef\"\u003e2\u003c/span\u003e. Additionally, we processed the DW datasets by applying the mode function to the multi-year dataset (2018\u0026ndash;2019). This multi-year aggregation approach accounts for the seasonal variability inherent in DW\u0026rsquo;s temporal coverage, allowing for a more stable representation of persistent land-cover patterns. Consequently, it enhances the identification of long-term trends useful for distinguishing between cropland (both active and fallow) and grassland\u003csup\u003e14\u003c/sup\u003e.\u003c/p\u003e\u003cp\u003e\u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab1\" border=\"1\"\u003e\u003ccaption language=\"En\"\u003e\u003cdiv class=\"CaptionNumber\"\u003eTable 1\u003c/div\u003e\u003cdiv class=\"CaptionContent\"\u003e\u003cp\u003eOverview of base land-cover maps containing grassland classes used in our study. MMU refers to the minimal mapping unit.\u003c/p\u003e\u003c/div\u003e\u003c/caption\u003e\u003ccolgroup cols=\"6\"\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c6\" colnum=\"6\"\u003e\u003c/div\u003e\u003cthead\u003e\u003ctr\u003e\u003cth align=\"left\" colname=\"c1\"\u003e\u003cp\u003eDataset Name\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c2\"\u003e\u003cp\u003eAbbreviation\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c3\"\u003e\u003cp\u003eGenerated by\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c4\"\u003e\u003cp\u003eMMU (m2)\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c5\"\u003e\u003cp\u003eYears Available\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c6\"\u003e\u003cp\u003eCoverage\u003c/p\u003e\u003c/th\u003e\u003c/tr\u003e\u003c/thead\u003e\u003ctbody\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eCorine Land Cover\u0026thinsp;+\u0026thinsp;Backbone\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eCLC+\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eCopernicus Land Monitoring Service\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e100\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e2018\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003eEurope\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eDynamic Word\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eDW\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eGoogle\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e250\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003eevery Sentinel-2 image with cloud cover less than 35%\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003eglobal\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003epan-European land cover map\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eELC10\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eVenter \u0026amp; Sydenham (2021)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e100\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e2018\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003eEurope\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eWord Cover ESA\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eESA WC\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eESA\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e100\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e2020, 2021\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003eglobal\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eEsri Land Cover\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eESRI LC\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eESRI\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e250\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e2020\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003eglobal\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eSentinel-2 Global Land Cover\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eS2GLC\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eMalinowski et al. 2020\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e100\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e2017\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003eEurope\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003c/tbody\u003e\u003c/colgroup\u003e\u003c/table\u003e\u003c/div\u003e\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec6\" class=\"Section3\"\u003e\u003ch2\u003e2.2.3. Sampling validation data\u003c/h2\u003e\u003cp\u003eWe followed best practices for accuracy assessment\u003csup\u003e30\u003c/sup\u003e. Harmonised land-cover classes from the base maps defined sampling strata for validation pixels in each mountain region. To avoid bias, we used a single validation dataset stratified by an external land-cover map\u003csup\u003e31\u003c/sup\u003e, independent of the evaluated products but using identical class definitions. We targeted an overall accuracy standard error below 0.01, expecting higher user accuracy for artificial land, cropland, woodland, and water, and lower user accuracy for shrubland, grassland, bare land, wetland, and snow (Supplementary S2). Each land-cover class included at least 70 validation pixels, yielding 1,460 points in the Alps and 1,410 in the Carpathians (Supplementary S2). Three experts visually interpreted each point using 2018 high-resolution Google Earth and Sentinel-2 imagery. The final class was assigned by majority agreement or by a single expert when confident in their interpretation.\u003c/p\u003e\u003cp\u003e\u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab2\" border=\"1\"\u003e\u003ccaption language=\"En\"\u003e\u003cdiv class=\"CaptionNumber\"\u003eTable 2\u003c/div\u003e\u003cdiv class=\"CaptionContent\"\u003e\u003cp\u003eLegend harmonisation of base LC datasets\u003c/p\u003e\u003c/div\u003e\u003c/caption\u003e\u003ccolgroup cols=\"10\"\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c6\" colnum=\"6\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c7\" colnum=\"7\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c8\" colnum=\"8\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c9\" colnum=\"9\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c10\" colnum=\"10\"\u003e\u003c/div\u003e\u003cthead\u003e\u003ctr\u003e\u003cth align=\"left\" colname=\"c1\"\u003e\u003cp\u003eAbbr.\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c2\"\u003e\u003cp\u003ebuilt\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c3\"\u003e\u003cp\u003ecrop\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c4\"\u003e\u003cp\u003eforest\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c5\"\u003e\u003cp\u003eshrub\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c6\"\u003e\u003cp\u003egrass\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c7\"\u003e\u003cp\u003ebare\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c8\"\u003e\u003cp\u003ewater\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c9\"\u003e\u003cp\u003ewet\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c10\"\u003e\u003cp\u003esnow\u003c/p\u003e\u003c/th\u003e\u003c/tr\u003e\u003c/thead\u003e\u003ctbody\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e\u003cb\u003eClass Num.\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e1\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e2\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e3\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e4\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e5\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c7\"\u003e\u003cp\u003e6\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c8\"\u003e\u003cp\u003e7\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c9\"\u003e\u003cp\u003e8\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c10\"\u003e\u003cp\u003e9\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e\u003cb\u003eLUCAS\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eArtificial land\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eCropland\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003eWoodland\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003eShrubland\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003eGrassland\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c7\"\u003e\u003cp\u003eBareland\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c8\"\u003e\u003cp\u003eWater\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c9\"\u003e\u003cp\u003eWetlands\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c10\"\u003e\u003cp\u003eSnow\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e\u003cb\u003eCLC+\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eSealed\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003ePeriodically herbaceous\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003eWoody\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003eLow-growing woody plants\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003ePermanent herbaceous\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c7\"\u003e\u003cp\u003eNon and sparsely-vegetated, Lichens and mosses\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c8\"\u003e\u003cp\u003eWater\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c9\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c10\"\u003e\u003cp\u003eSnow and ice\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e\u003cb\u003eDW\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eBuilt Area\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eCrops\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003eTrees\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003eScrub/Shrub\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003eGrass\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c7\"\u003e\u003cp\u003eBare Ground\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c8\"\u003e\u003cp\u003eWater\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c9\"\u003e\u003cp\u003eFlooded Vegetation\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c10\"\u003e\u003cp\u003eSnow/Ice\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e\u003cb\u003eELC10\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eArtificial land\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eCropland\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003eWoodland\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003eShrubland\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003eGrassland\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c7\"\u003e\u003cp\u003eBareland\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c8\"\u003e\u003cp\u003eWater\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c9\"\u003e\u003cp\u003eWetlands\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c10\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e\u003cb\u003eESA WC\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eBuilt-up\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eCropland\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003eTree cover\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003eShrubland\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003eGrassland\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c7\"\u003e\u003cp\u003eBare / sparse vegetation, Moss and\u003c/p\u003e\u003cp\u003elichen\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c8\"\u003e\u003cp\u003ePermanent water bodies\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c9\"\u003e\u003cp\u003eHerbaceous wetland\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c10\"\u003e\u003cp\u003eSnow and Ice\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e\u003cb\u003eEsri\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eBuilt Area\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eCrops\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003eTrees\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003eScrub/Shrub\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003eGrass\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c7\"\u003e\u003cp\u003eBare Ground\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c8\"\u003e\u003cp\u003eWater\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c9\"\u003e\u003cp\u003eFlooded Vegetation\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c10\"\u003e\u003cp\u003eSnow/Ice\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e\u003cb\u003eS2GLC\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eArtificial surfaces\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eCultivated areas\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003eBroadleaf, Coniferous tree cover\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003eMoors and heathland\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003eHerbaceous vegetation\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c7\"\u003e\u003cp\u003eNatural material surfaces\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c8\"\u003e\u003cp\u003eWater bodies\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c9\"\u003e\u003cp\u003eMarshes, Peatbogs\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c10\"\u003e\u003cp\u003ePermanent snow cover\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003c/tbody\u003e\u003c/colgroup\u003e\u003c/table\u003e\u003c/div\u003e\u003c/p\u003e\u003c/div\u003e\u003c/div\u003e\u003cdiv id=\"Sec7\" class=\"Section2\"\u003e\u003ch2\u003e2.3. Evaluation and Integration of Land-Cover Datasets\u003c/h2\u003e\u003cdiv id=\"Sec8\" class=\"Section3\"\u003e\u003ch2\u003e2.3.1. Accuracy assessment\u003c/h2\u003e\u003cp\u003eWe validated each of the used land-cover datasets and the consensus dataset using the independent validation sample described in Section \u003cspan refid=\"Sec6\" class=\"InternalRef\"\u003e2.2.3\u003c/span\u003e. For each class in each map, we computed accuracy metrics including user accuracy, producer accuracy, overall accuracy, and the F1-score. Standard errors for these metrics were estimated using the equations provided by Stehman \u0026amp; Foody (2019, Eqs.\u0026nbsp;12\u0026ndash;18)\u003csup\u003e32\u003c/sup\u003e.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec9\" class=\"Section3\"\u003e\u003ch2\u003e2.3.2. Consensus approaches\u003c/h2\u003e\u003cp\u003eWe tested three consensus mapping approaches to integrate multiple land-cover datasets. First, we implemented a weighted voting approach (Con_WV), where each pixel was assigned the class with the highest sum of weights, with the weights being the F1 scores for each class across all input datasets. To enhance this weighted voting approach, we introduced an accuracy-confusion weighing method (Con_AccCo), adapted from\u003csup\u003e21\u003c/sup\u003e. This method calculates a consensus probability for each class by summing correct classification probabilities (the diagonal elements of the error matrices) while subtracting misclassification probabilities (the off-diagonal elements), thereby emphasising classes that are more reliably detected across datasets. Finally, we developed an accuracy-weighted Random Forest model (Con_RF). This method integrates multiple land-cover datasets using F1-scores as explanatory variables, allowing each dataset to contribute proportionally based on its observed accuracy. Training data were derived from LUCAS Harmonised\u003csup\u003e29\u003c/sup\u003e and LUCAS Copernicus\u003csup\u003e33\u003c/sup\u003e and preprocessed to remove ambiguous or low-accuracy points. The Random Forest classifier (ee.Classifier.smileRandomForest) was trained in Google Earth Engine with 500 trees and default parameters. For a more detailed description of these approaches, refer to Supplementary Material S1.\u003c/p\u003e\u003c/div\u003e\u003c/div\u003e\u003cdiv id=\"Sec10\" class=\"Section2\"\u003e\u003ch2\u003e2.4. Comparison of grasslands across the Alps and Carpathians\u003c/h2\u003e\u003cp\u003eWe first quantified the total extent of grasslands, defined as the area of all grassland pixels, in the Alps, in the Carpathians, and separately for each country within these mountain ranges. We then examined how this extent varied along environmental gradients. For every pixel classified as grassland, we extracted elevation, slope, and aspect from the Copernicus 30-m global DEM, resampled to 10 m using nearest-neighbour resampling.\u003c/p\u003e\u003cp\u003eTo assess how grassland landscape structure differs among datasets, we compared the composition and configuration of landscapes in the Alps and Carpathians using a set of landscape metrics. Landscapes were selected by generating a regular point grid at 5-km spacing and drawing a circular buffer of 2500 m around each point. Within each buffer, we calculated metrics that describe grassland patch patterns in relation to the area and shape of grassland patches. The metrics include mean grassland patch area, total grassland area, patch fractal dimension (PAFRAC), and edge density.\u003c/p\u003e\u003c/div\u003e"},{"header":"3. Results","content":"\u003cdiv id=\"Sec12\" class=\"Section2\"\u003e\u003ch2\u003e3.1 Comparison of individual land-cover maps and consensus map quality\u003c/h2\u003e\u003cp\u003eThe accuracy of individual base land-cover datasets varied considerably, reflecting differences in classification methods and input data (Figure. 2). In the Alps, overall accuracy ranged from 70% to 87%, with ESA WC (86%) and CLC+ (87%) performing best, while S2GLC (70%) and ESRI LC (82%) had the lowest accuracy among the evaluated datasets (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003eA). Accuracy followed a similar pattern in the Carpathians, with ESA WC reaching 90% and CLC\u0026thinsp;+\u0026thinsp;and ELC10 (exceeding 87%). Both Con_WV and Con_AccCo improved overall classification accuracy relative to individual datasets, but Con_RF consistently achieved the highest accuracy, with an overall accuracy of 90% in the Alps and 92% in the Carpathians (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003eA).\u003c/p\u003e\u003cp\u003eFor grassland areas, the greatest improvement in the accuracy for this class was observed in Con_RF, where both user accuracy and producer accuracy exceeded 84% in the Alps and Carpathians (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003eB, C). Con_RF and CLC\u0026thinsp;+\u0026thinsp;emerged as the top-performing maps for grassland classification, with both datasets achieving user and producer accuracies above 80%. In contrast, Con_AccCo and Con_Votes demonstrated slightly higher omission errors by underestimating the grassland areas, which led to reduced user accuracy. Among the original datasets, only CLC\u0026thinsp;+\u0026thinsp;outperformed Con_RF in terms of user accuracy in the Alps. However, when considering producer accuracy, both datasets showed very similar performance (0.872 and 0.874 F1-scores for CLC\u0026thinsp;+\u0026thinsp;and Con_RF, respectively).\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec13\" class=\"Section2\"\u003e\u003ch2\u003e3.2. Inter-Dataset Differences in Grassland Extent and Distribution\u003c/h2\u003e\u003cp\u003eThe comparison of land-cover datasets revealed clear discrepancies in mapped proportions of grasslands across countries in the Alps and Carpathians (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003e). DW and ESRI LC consistently reported the lowest grassland coverage, whereas ESA WC and CLC\u0026thinsp;+\u0026thinsp;showed the highest proportions. ELC10 produced intermediate values overall, but substantially higher grassland shares in the northwestern Alps (Germany) compared to all other products. Consensus products generally reduced low and high proportions but still differed in magnitude. Con_RF indicated higher grassland proportions than Con_AccCon and Con_WV, aligning more closely with ESA WC.\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003cp\u003eThe distribution of grasslands according to elevation, slope, and aspect gradients differs between individual map sources, which were more pronounced in the Alps than in the Carpathians (Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003eA). CLC+, ESA WC, and S2GLC covered the entire spectrum from secondary grasslands in lowlands to the alpine grasslands above 2,500 m a.s.l., while DW, ESRI LC and ELC10 mapped primarily lower elevation grasslands, with considerably lower representation in high mountain areas. Consensus products (Con_RF, Con_AccCo, Con_WV) produced more consistent spatial patterns but still varied in their depiction of high-elevation grasslands. Median values differed in the Alps, with a spread of 900 m between the lowest median (ELC10\u0026ndash;810 m a.s.l.) and the highest ESA WC (1710 m a.s.l.). In the Carpathians, the differences between datasets were less pronounced in this respect, with all sources showing a distribution concentrated at lower elevations (with a median of 420 m a.s.l. for CLC\u0026thinsp;+\u0026thinsp;to 540 m a.s.l. for ESRI LC) and alpine locations almost absent.\u003c/p\u003e\u003cp\u003eSlope distribution and orientation (Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003eB) revealed less pronounced differences between the individual datasets as compared to elevation. In the Alps, CLC+, ELC10, and ESA WorldCover consistently mapped grasslands on gentler terrain, reflecting their tendency to classify grasslands on low slopes. In contrast, DW, ESRI LC, and S2GLC mapped grasslands to steeper slopes, especially in the upper mountain and subalpine zones.\u003c/p\u003e\u003cp\u003eIn the Carpathians, most grasslands occurred on gentle terrain below 20\u0026deg;, corresponding to the extensive valley and plateau systems typical of this region. The narrower range of slopes compared to the Alps partly explains the reduced variability among datasets, as steep alpine meadows are almost absent in the Carpathian environment.\u003c/p\u003e\u003cp\u003eDifferences in aspect distribution were the least pronounced yet still noticeable (Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003eC). In the Alps, CLC+, ELC10, and ESA WorldCover showed a preference for east to south-facing slopes, which corresponds to the climatic conditions of these exposures for productive grasslands. DW, ESRI LC and S2GLC captured similar orientations but placed greater emphasis on south and southeast orientations. In contrast, the Carpathian datasets exhibited a broader and more even distribution of aspects, reflecting the region\u0026rsquo;s gentler relief and lower topographic variability.\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec14\" class=\"Section2\"\u003e\u003ch2\u003e3.3. Influence of Land-Cover Dataset Characteristics on Grassland Spatial Structure\u003c/h2\u003e\u003cp\u003eLandscape metrics reveal consistent structural patterns across datasets. ESRI LC and DW, with large mean patch areas, show low edge density and low shape complexity (PAFRAC), indicating spatially aggregated, contiguous, and geometrically simple grassland patches with limited but consolidated extent.\u003c/p\u003e\u003cp\u003eConversely, ELC10 and S2GLC, characterized by small mean patch areas, exhibit high edge density and elevated PAFRAC, reflecting fragmented, irregular, and heterogeneous landscapes. Their larger total class areas, concentrated mainly in the Carpathians, result from spatial dispersion rather than continuous coverage.\u003c/p\u003e\u003cp\u003eIntermediate configurations (e.g., ESA WC, CLC+, and consensus datasets) occupy a transitional position between these extremes. Moderate mean patch areas correspond with intermediate edge density and PAFRAC, reflecting balanced spatial aggregation and geometric complexity. Within the consensus datasets, Con_RF exhibits slightly larger mean patch and class areas, together with moderate edge density, suggesting that ensemble averaging mitigates fragmentation and produces a more cohesive yet regionally nuanced depiction of grassland structure.\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003c/div\u003e"},{"header":"4. Discussion","content":"\u003cp\u003eAccurate and detailed land-cover mapping is essential for synthesizing knowledge of ecosystems across large areas. This need is especially acute in mountainous regions, where complex topography and diverse management histories generate mosaics of heterogeneous landscapes\u003csup\u003e10\u003c/sup\u003e. Within this study, we demonstrated that the land cover datasets provided contrasting mapping of grasslands in the Alps and Carpathians, with possible implications for ecological analyses and biodiversity monitoring. Furthermore, we showed that integrating multiple land-cover products reduces dataset-specific biases and yields a more reliable composite map\u003csup\u003e14,15,31\u003c/sup\u003e.\u003c/p\u003e\u003cp\u003eOur results confirm a central pattern reported for recent Sentinel-based land-cover datasets at 10-meter resolution: while overall accuracy can be high, performance across individual classes remains uneven, and substantial differences persist between datasets\u003csup\u003e15\u003c/sup\u003e. In line with previous cross-comparisons, we observe that global datasets differ most in their ability to map grasslands, shrublands, and wetlands, while showing stronger agreement for water, tree cover, built-up areas, and crops\u003csup\u003e14\u003c/sup\u003e. Our assessment also confirms these class-specific discrepancies where DW and ESRI LC systematically underestimate the area of grasslands compared to ESA WC and CLC\u0026thinsp;+\u0026thinsp;in both mountain systems. Similar biases have been documented globally, with ESA WC tending to overestimate grasslands, Esri overestimating shrublands, and DW exaggerating snow/ice\u003csup\u003e14\u003c/sup\u003e. These systematic tendencies likely explain the under and overestimations of grassland areas that we found in the Alps and Carpathians\u003c/p\u003e\u003cp\u003eIn our comparison, the regional datasets (ELC10, CLC+, S2GLC) performed relatively well, echoing their reported pan-European accuracies\u003csup\u003e27,28\u003c/sup\u003e. However, upon comparing the results, it is clear that the Alps and Carpathians pose particular challenges. The steep terrain, fine-scale mosaics, and spectral similarity between crops, grasslands, and shrubs exacerbate classification errors (Supplementary S4). Consequently, user and producer accuracies in our study deviated from continental assessments that documented persistent weaknesses in mapping herbaceous vegetation\u003csup\u003e34\u003c/sup\u003e.\u003c/p\u003e\u003cp\u003eConsensus approaches substantially improve the reliability of grassland mapping in complex mountain environments. The observed increase in their mapping accuracy highlights the ability of ensemble integration to produce a more dependable map of grassland cover. While regional datasets such as CLC\u0026thinsp;+\u0026thinsp;and ELC10 already performed relatively well, our results show that combining multiple sources reduces class-specific biases and produces depictions of grasslands that are both more robust and ecologically consistent. This supports previous calls for ensemble or weighted-evidence strategies to increase the reliability of land cover information\u003csup\u003e19,35,36\u003c/sup\u003e.\u003c/p\u003e\u003cp\u003eNonetheless, consensus maps remain constrained by the quality and availability of the underlying datasets. For example, the lack of full coverage of the Carpathian Mountains in CLC+ (e.g., Ukraine) and the partial availability of S2GLC limited the potential use within this mountain range. Moreover, harmonising the consensus map with the LUCAS nomenclature thematic detail leads to inconsistencies in wetland or snow classes across products. These compromises are in line with other harmonisation efforts, highlighting that consensus approaches may reduce uncertainty, but cannot fully eliminate structural limitations in land-cover data \u003csup\u003e37,38\u003c/sup\u003e.\u003c/p\u003e\u003cp\u003eThe selection of land-cover datasets strongly influences the evaluation of environmental gradients and landscape metrics (Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003e, \u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003e). We documented that on grasslands, where their elevation distribution revealed strong divergences. Some datasets reproduced the expected bimodal pattern of secondary lowland and alpine grasslands in the Alps, while others mapped larger extents of grasslands in low elevations. Such bias in the datasets may underestimate the high-elevation grasslands, although this cannot be directly validated in this study, given that our reference data were not stratified by elevation. Even so, we found that in the Carpathians, differences between datasets were less pronounced, which is likely due to the location of grasslands at lower and mid elevations. Slope distributions showed narrower variation than elevation but still revealed potential biases. ELC10 and CLC\u0026thinsp;+\u0026thinsp;emphasised gentler slopes, while DW and ESRI LC extended into steeper terrain. Aspect patterns were more subtle, with ESA WC and CLC\u0026thinsp;+\u0026thinsp;depicting grasslands more evenly across orientations, while DW and Esri concentrated them on south- and east-facing slopes. Selected landscape metrics reinforced these distinctions, as DW and ESRI LC mapped grasslands as aggregated patches due to their minimum mapping unit area, whereas other datasets depicted finer, more fragmented mosaics. Consensus maps reconciled these extremes, producing more ecologically plausible patch structures and retaining regional contrasts between the Alps and Carpathians.\u003c/p\u003e\u003cp\u003eThese improvements have important implications for downstream applications. Venter et al. (2023)\u003csup\u003e39\u003c/sup\u003e, for example, showed that Sentinel-based land-cover maps improved species distribution models for solitary bees in Norway compared with manually digitised maps, largely because products such as ELC10 and ESA WC, with smaller minimum mapping units, better resolved fine-scale habitat patches. Natsukawa et al. (2024)\u003csup\u003e40\u003c/sup\u003e demonstrated that integrating fine-scale land cover data with topographic information substantially enhances habitat models, yielding significant insights for conservation. Our findings suggest that relying on a single dataset for similar broad-scale analyses may yield markedly different results, depending on the specific product chosen.\u003c/p\u003e\u003cp\u003eBy integrating multiple land-cover datasets, our consensus map mitigated these distortions, preserving ecologically relevant gradients while reducing outliers from individual datasets. This suggests that consensus approaches can provide land-cover maps that are ecologically plausible, are methodologically robust, and are more suitable for downstream applications. These include biodiversity modelling \u003csup\u003e21,39\u003c/sup\u003e, habitat quality assessments, connectivity analyses, and quantification of land use intensity in grassland systems. In this context, ensemble products also improve the comparability of studies that rely on land cover as a foundational variable. Relying on a single land-cover dataset risks systematic bias, whereas consensus products enhance the comparability and credibility of evidence. In this sense, ensemble approaches improve the practical utility of land cover information in heterogeneous mountain landscapes.\u003c/p\u003e"},{"header":"5. Conclusion","content":"\u003cp\u003eAccurate land cover maps are fundamental for broad-scale monitoring of the earth-systems particularly in heterogeneous mountain systems. Thanks to the technical availabilities, there are multiple datasets that are candidates to reach sufficient accuracy. Our analysis revealed that state-of-the-art land-cover datasets, although derived from similar satellite sources, yield substantially different representations of grasslands in the Alps and Carpathians, which in turn affects their extent and spatial configuration. These discrepancies varied across gradients of elevation, slope, and aspect. By integrating multiple datasets through ensemble and weighted consensus approaches, we demonstrated a reduction of such discrepancies. The Random Forest consensus map achieved the highest overall and class-specific accuracies, outperforming individual datasets and producing an ecologically coherent map of grasslands in both mountain regions. This improvement results from combining the complementary strengths of mapping products, thereby mitigating dataset-specific biases and increasing robustness in complex terrain.\u003c/p\u003e\u003cp\u003eBeyond accuracy gains, the consensus approach preserved key ecological gradients and generated more realistic spatial patterns of grassland distribution and fragmentation. Such qualitative improvements may be particularly valuable for downstream applications that rely on land-cover inputs. While it may directly impact the results of the spatial-based studies, such as habitat modelling, landscape connectivity analyses, and biodiversity monitoring, it may also indirectly influence the policy documents that shape the future biodiversity strategies.\u003c/p\u003e"},{"header":"Declarations","content":"\u003cp\u003e\u003ch2\u003eEthics declarations\u003c/h2\u003e\u003cp\u003eThe expert annotations used in this study were produced by the co‑authors Šimon Opravil, Tom\u0026aacute;š Goga, and Hamid Afzali. Each co‑author provided informed consent for the use of their annotations in this manuscript and in any related supplementary materials. No personal or sensitive information was collected; the annotations concern land‑cover categories only.\u003c/p\u003e\u003c/p\u003e\u003cp\u003e\u003ch2\u003eCompeting interests\u003c/h2\u003e\u003cp\u003eThe authors declare no competing interests.\u003c/p\u003e\u003c/p\u003e\u003ch2\u003eFunding Declaration\u003c/h2\u003e\u003cp\u003e\u003c/p\u003e\u003cp\u003eThis research was funded by the project \u0026lsquo;G4B: Grasslands for biodiversity: supporting the protection of the biodiversity-rich grasslands and related management practices in the Alps and Carpathians\u0026rsquo; funded by Biodiversa+, the European Biodiversity Partnership under the 2021\u0026ndash;2022 BiodivProtect joint call for research proposals, co-funded by the European Commission (GA N\u0026deg;101052342) and with the funding organisations SNSF, DFG, NCN, PROV BZ, SAS and UEFISCDI; by the project of Slovak Research and Development Agency-21-0226: \u0026ldquo;Species-rich Carpathian grasslands: mapping, history, drivers of change and conservation\u0026rdquo; pursued at the Institute of Geography of the Slovak Academy of Sciences; by the Slovak Scientific Grant Agency VEGA under Grant 2/0043/23 \u0026ldquo;Detection of landscape diversity and its changes in Slovakia based on remote sensing data in the context of the European Green Deal.\u0026rdquo;\u003c/p\u003e\u003ch2\u003eAuthor Contribution\u003c/h2\u003e\u003cp\u003eConceptualization, S.O., M.B., T.K., R.P; methodology, S.O., M.B., T.B., R.P.; software, S.O.; validation, S.O., T.G., H.A.; formal analysis, S.O. and R.P.; investigation, S.O., T.G., H.A., R.P., resources, S.O., T.G.; data curation, S.O.; writing\u0026mdash;original draft preparation, S.O. and R.P.; writing\u0026mdash;review and editing, M.B., T.G., H.A., T.K.; visualization, S.O.; supervision, R.P. and T.K.; project administration, R.P, and T.K.; funding acquisition, R.P. and T.K. All authors have read and agreed to the published version of the manuscript.\u003c/p\u003e\u003ch2\u003eAcknowledgements\u003c/h2\u003e\u003cp\u003eThis study was made possible using multiple existing land cover datasets, including Google\u0026copy; Dynamic World, ESA WorldCover, ESRI\u0026copy; Land Cover, CLMS CLC+, ELC10, and S2GLC. We acknowledge Google, the European Space Agency, ESRI, the Copernicus Land Monitoring Service, and other data providers for making these datasets publicly available. All maps used in this study are accessible under their respective open data licenses.\u003c/p\u003e\u003ch2\u003eData Availability\u003c/h2\u003e\u003cp\u003eConsensus datasets are openly available in the Zenodo repository at [https://doi.org/10.5281/zenodo.13823832](https:/doi.org/10.5281/zenodo.13823832) . The code used for data processing, analysis, and all other datasets stored in Google Earth Engine is publicly accessible in the GitHub repository at [https://github.com/simonopravil/MAC-Land](https:/github.com/simonopravil/MAC-Land) .\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\n\u003cli\u003eCochran, F., Daniel, J., Jackson, L. \u0026amp; Neale, A. Earth observation-based ecosystem services indicators for national and subnational reporting of the sustainable development goals. \u003cem\u003eRemote Sens. Environ.\u003c/em\u003e \u003cb\u003e244\u003c/b\u003e, 111796 (2020).\u003c/li\u003e\n\u003cli\u003ePereira, H. M. \u003cem\u003eet al.\u003c/em\u003e Essential Biodiversity Variables. \u003cem\u003eScience\u003c/em\u003e \u003cb\u003e339\u003c/b\u003e, 277–278 (2013).\u003c/li\u003e\n\u003cli\u003eTimmermans, J. \u0026amp; Daniel Kissling, W. Advancing terrestrial biodiversity monitoring with satellite remote sensing in the context of the Kunming-Montreal global biodiversity framework. \u003cem\u003eEcol. Indic.\u003c/em\u003e \u003cb\u003e154\u003c/b\u003e, 110773 (2023).\u003c/li\u003e\n\u003cli\u003eGaujour, E., Amiaud, B., Mignolet, C. \u0026amp; Plantureux, S. Factors and processes affecting plant biodiversity in permanent grasslands. A review. \u003cem\u003eAgron. Sustain. Dev.\u003c/em\u003e \u003cb\u003e32\u003c/b\u003e, 133–160 (2012).\u003c/li\u003e\n\u003cli\u003eScherreiks, P. \u003cem\u003eet al.\u003c/em\u003e Present and historical landscape structure shapes current species richness in Central European grasslands. \u003cem\u003eLandsc. Ecol.\u003c/em\u003e \u003cb\u003e37\u003c/b\u003e, 745–762 (2022).\u003c/li\u003e\n\u003cli\u003eSchindler, S., von Wehrden, H., Poirazidis, K., Wrbka, T. \u0026amp; Kati, V. Multiscale performance of landscape metrics as indicators of species richness of plants, insects and vertebrates. \u003cem\u003eEcol. Indic.\u003c/em\u003e \u003cb\u003e31\u003c/b\u003e, 41–48 (2013).\u003c/li\u003e\n\u003cli\u003eTurner, M. G. \u0026amp; Gardner, R. H. Landscape Metrics. in \u003cem\u003eLandscape Ecology in Theory and Practice: Pattern and Process\u003c/em\u003e (eds Turner, M. G. \u0026amp; Gardner, R. H.) 97–142 (Springer, New York, NY, 2015). doi:10.1007/978-1-4939-2794-4_4.\u003c/li\u003e\n\u003cli\u003eHuang, C., Geiger, E. L. \u0026amp; Kupfer, J. A. Sensitivity of landscape metrics to classification scheme. \u003cem\u003eInt. J. Remote Sens.\u003c/em\u003e \u003cb\u003e27\u003c/b\u003e, 2927–2948 (2006).\u003c/li\u003e\n\u003cli\u003eShao, G. \u0026amp; Wu, J. On the accuracy of landscape pattern analysis using remote sensing data. \u003cem\u003eLandsc. Ecol.\u003c/em\u003e \u003cb\u003e23\u003c/b\u003e, 505–511 (2008).\u003c/li\u003e\n\u003cli\u003eBezák, P. \u0026amp; Mitchley, J. Drivers of change in mountain farming in Slovakia: from socialist collectivisation to the Common Agricultural Policy. \u003cem\u003eReg. Environ. Change\u003c/em\u003e \u003cb\u003e14\u003c/b\u003e, 1343–1356 (2014).\u003c/li\u003e\n\u003cli\u003eCoelho, M. T. P. \u003cem\u003eet al.\u003c/em\u003e The geography of climate and the global patterns of species diversity. \u003cem\u003eNature\u003c/em\u003e \u003cb\u003e622\u003c/b\u003e, 537–544 (2023).\u003c/li\u003e\n\u003cli\u003eRahbek, C. \u003cem\u003eet al.\u003c/em\u003e Humboldt’s enigma: What causes global patterns of mountain biodiversity? \u003cem\u003eScience\u003c/em\u003e \u003cb\u003e365\u003c/b\u003e, 1108–1113 (2019).\u003c/li\u003e\n\u003cli\u003e\u003cem\u003eGrasslands of the World: Diversity, Management and Conservation\u003c/em\u003e. (CRC Press, Boca Raton London New York, 2018).\u003c/li\u003e\n\u003cli\u003eVenter, Z. S., Barton, D. N., Chakraborty, T., Simensen, T. \u0026amp; Singh, G. Global 10 m Land Use Land Cover Datasets: A Comparison of Dynamic World, World Cover and Esri Land Cover. \u003cem\u003eRemote Sens.\u003c/em\u003e \u003cb\u003e14\u003c/b\u003e, 4101 (2022).\u003c/li\u003e\n\u003cli\u003eWang, Y. \u003cem\u003eet al.\u003c/em\u003e A review of regional and Global scale Land Use/Land Cover (LULC) mapping products generated from satellite remote sensing. \u003cem\u003eISPRS J. Photogramm. Remote Sens.\u003c/em\u003e \u003cb\u003e206\u003c/b\u003e, 311–334 (2023).\u003c/li\u003e\n\u003cli\u003eIwao, K. \u003cem\u003eet al.\u003c/em\u003e Creation of New Global Land Cover Map with Map Integration. \u003cem\u003eJ. Geogr. Inf. Syst.\u003c/em\u003e \u003cb\u003e03\u003c/b\u003e, 160–165 (2011).\u003c/li\u003e\n\u003cli\u003eTsendbazar, N.-E., De Bruin, S., Fritz, S. \u0026amp; Herold, M. Spatial Accuracy Assessment and Integration of Global Land Cover Datasets. \u003cem\u003eRemote Sens.\u003c/em\u003e \u003cb\u003e7\u003c/b\u003e, 15804–15821 (2015).\u003c/li\u003e\n\u003cli\u003ePérez-Hoyos, A., García-Haro, F. J. \u0026amp; San-Miguel-Ayanz, J. A methodology to generate a synergetic land-cover map by fusion of different land-cover products. \u003cem\u003eInt. J. Appl. Earth Obs. Geoinformation\u003c/em\u003e \u003cb\u003e19\u003c/b\u003e, 72–87 (2012).\u003c/li\u003e\n\u003cli\u003eHuang, A. \u003cem\u003eet al.\u003c/em\u003e A Methodology to Generate Integrated Land Cover Data for Land Surface Model by Improving Dempster-Shafer Theory. \u003cem\u003eRemote Sens.\u003c/em\u003e \u003cb\u003e14\u003c/b\u003e, 972 (2022).\u003c/li\u003e\n\u003cli\u003ePérez-Hoyos, A., Udías, A. \u0026amp; Rembold, F. Integrating multiple land cover maps through a multi-criteria analysis to improve agricultural monitoring in Africa. \u003cem\u003eInt. J. Appl. Earth Obs. Geoinformation\u003c/em\u003e \u003cb\u003e88\u003c/b\u003e, 102064 (2020).\u003c/li\u003e\n\u003cli\u003eTuanmu, M. \u0026amp; Jetz, W. A global 1-km consensus land‐cover product for biodiversity and ecosystem modelling. \u003cem\u003eGlob. Ecol. Biogeogr.\u003c/em\u003e \u003cb\u003e23\u003c/b\u003e, 1031–1045 (2014).\u003c/li\u003e\n\u003cli\u003eAguilar, R., Zurita-Milla, R., Izquierdo-Verdiguier, E. \u0026amp; A. De By, R. A Cloud-Based Multi-Temporal Ensemble Classifier to Map Smallholder Farming Systems. \u003cem\u003eRemote Sens.\u003c/em\u003e \u003cb\u003e10\u003c/b\u003e, 729 (2018).\u003c/li\u003e\n\u003cli\u003eWitjes, M. \u003cem\u003eet al.\u003c/em\u003e A spatiotemporal ensemble machine learning framework for generating land use/land cover time-series maps for Europe (2000–2019) based on LUCAS, CORINE and GLAD Landsat. \u003cem\u003ePeerJ\u003c/em\u003e \u003cb\u003e10\u003c/b\u003e, e13573 (2022).\u003c/li\u003e\n\u003cli\u003eBrown, C. F. \u003cem\u003eet al.\u003c/em\u003e Dynamic World, Near real-time global 10 m land use land cover mapping. \u003cem\u003eSci. Data\u003c/em\u003e \u003cb\u003e9\u003c/b\u003e, 251 (2022).\u003c/li\u003e\n\u003cli\u003eZanaga, D. \u003cem\u003eet al.\u003c/em\u003e ESA WorldCover 10 m 2020 v100. Zenodo https://doi.org/10.5281/ZENODO.5571936 (2021).\u003c/li\u003e\n\u003cli\u003eKarra, K. \u003cem\u003eet al.\u003c/em\u003e Global land use / land cover with Sentinel 2 and deep learning. in \u003cem\u003e2021 IEEE International Geoscience and Remote Sensing Symposium IGARSS\u003c/em\u003e 4704–4707 (2021). doi:10.1109/IGARSS47720.2021.9553499.\u003c/li\u003e\n\u003cli\u003eVenter, Z. S. \u0026amp; Sydenham, M. A. K. Continental-Scale Land Cover Mapping at 10 m Resolution Over Europe (ELC10). \u003cem\u003eRemote Sens.\u003c/em\u003e \u003cb\u003e13\u003c/b\u003e, 2301 (2021).\u003c/li\u003e\n\u003cli\u003eMalinowski, R. \u003cem\u003eet al.\u003c/em\u003e Automated Production of a Land Cover/Use Map of Europe Based on Sentinel-2 Imagery. \u003cem\u003eRemote Sens.\u003c/em\u003e \u003cb\u003e12\u003c/b\u003e, 3523 (2020).\u003c/li\u003e\n\u003cli\u003ed’Andrimont, R. \u003cem\u003eet al.\u003c/em\u003e Harmonised LUCAS in-situ land cover and use database for field surveys from 2006 to 2018 in the European Union. \u003cem\u003eSci. Data\u003c/em\u003e \u003cb\u003e7\u003c/b\u003e, 352 (2020).\u003c/li\u003e\n\u003cli\u003eOlofsson, P. \u003cem\u003eet al.\u003c/em\u003e Good practices for estimating area and assessing accuracy of land change. \u003cem\u003eRemote Sens. Environ.\u003c/em\u003e \u003cb\u003e148\u003c/b\u003e, 42–57 (2014).\u003c/li\u003e\n\u003cli\u003ePflugmacher, D., Rabe, A., Peters, M. \u0026amp; Hostert, P. Mapping pan-European land cover using Landsat spectral-temporal metrics and the European LUCAS survey. \u003cem\u003eRemote Sens. Environ.\u003c/em\u003e \u003cb\u003e221\u003c/b\u003e, 583–595 (2019).\u003c/li\u003e\n\u003cli\u003eStehman, S. V. \u0026amp; Foody, G. M. Key issues in rigorous accuracy assessment of land cover products. \u003cem\u003eRemote Sens. Environ.\u003c/em\u003e \u003cb\u003e231\u003c/b\u003e, 111199 (2019).\u003c/li\u003e\n\u003cli\u003ed’Andrimont, R. \u003cem\u003eet al.\u003c/em\u003e LUCAS Copernicus 2018: Earth-observation-relevant in situ data on land cover and use throughout the European Union. \u003cem\u003eEarth Syst. Sci. Data\u003c/em\u003e \u003cb\u003e13\u003c/b\u003e, 1119–1133 (2021).\u003c/li\u003e\n\u003cli\u003eVan Cleemput, E., Vanierschot, L., Fernández-Castilla, B., Honnay, O. \u0026amp; Somers, B. The functional characterization of grass- and shrubland ecosystems using hyperspectral remote sensing: trends, accuracy and moderating variables. \u003cem\u003eRemote Sens. Environ.\u003c/em\u003e \u003cb\u003e209\u003c/b\u003e, 747–763 (2018).\u003c/li\u003e\n\u003cli\u003eHerold, M., See, L., Tsendbazar, N.-E. \u0026amp; Fritz, S. Towards an Integrated Global Land Cover Monitoring and Mapping System. \u003cem\u003eRemote Sens.\u003c/em\u003e \u003cb\u003e8\u003c/b\u003e, 1036 (2016).\u003c/li\u003e\n\u003cli\u003eKerner, H. \u003cem\u003eet al.\u003c/em\u003e How accurate are existing land cover maps for agriculture in Sub-Saharan Africa? \u003cem\u003eSci. Data\u003c/em\u003e \u003cb\u003e11\u003c/b\u003e, 486 (2024).\u003c/li\u003e\n\u003cli\u003eJansen, L. J. M., Groom, G. \u0026amp; Carrai, G. Land-cover harmonisation and semantic similarity: some methodological issues. \u003cem\u003eJ. Land Use Sci.\u003c/em\u003e \u003cb\u003e3\u003c/b\u003e, 131–160 (2008).\u003c/li\u003e\n\u003cli\u003eXu, Q., Yordanov, V., Bruzzone, L. \u0026amp; Brovelli, M. A. High-Resolution Global Land Cover Maps and Their Assessment Strategies. \u003cem\u003eISPRS Int. J. Geo-Inf.\u003c/em\u003e \u003cb\u003e14\u003c/b\u003e, 235 (2025).\u003c/li\u003e\n\u003cli\u003eVenter, Z. S. \u003cem\u003eet al.\u003c/em\u003e Comparing Global Sentinel-2 Land Cover Maps for Regional Species Distribution Modeling. \u003cem\u003eRemote Sens.\u003c/em\u003e \u003cb\u003e15\u003c/b\u003e, 1749 (2023).\u003c/li\u003e\n\u003cli\u003eNatsukawa, H. \u003cem\u003eet al.\u003c/em\u003e Importance of the interplay between land cover and topography in modeling habitat selection. \u003cem\u003eEcol. Indic.\u003c/em\u003e \u003cb\u003e169\u003c/b\u003e, 112896 (2024\u003c/li\u003e\n\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":false,"highlight":"","institution":"","isAcceptedByJournal":true,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"scientific-reports","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"scirep","sideBox":"Learn more about [Scientific Reports](http://www.nature.com/srep/)","snPcode":"","submissionUrl":"","title":"Scientific Reports","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"stoa","reportingPortfolio":"Scientific Reports","inReviewEnabled":true,"inReviewRevisionsEnabled":true},"keywords":"Land-cover mapping, Consensus approach, Grasslands, Alps and Carpathians, Earth observation","lastPublishedDoi":"10.21203/rs.3.rs-8096101/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-8096101/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eAccurate land-cover information is essential for biodiversity monitoring, yet existing 10-m global and continental land-cover datasets vary in accuracy and thematic consistency, particularly for grasslands in complex mountain environments. We assessed six state-of-the-art land-cover products (Dynamic World, ESA WorldCover, Esri Land Cover, Corine Land Cover\u0026thinsp;+\u0026thinsp;Backbone, ELC10, S2GLC) across the Alps and Carpathians and developed three consensus maps using weighted voting, accuracy-confusion weighting, and an accuracy-weighted Random Forest ensemble. All datasets were validated against an independent set of expert-interpreted reference samples. Individual products showed large discrepancies in grassland extent, elevation distribution, and landscape structure. Global datasets (Dynamic World, Esri Land Cover) underestimated grassland extent, whereas ESA WorldCover and Corine Land Cover\u0026thinsp;+\u0026thinsp;reported higher proportions. Consensus approaches substantially reduced these inconsistencies. The Random Forest ensemble achieved the highest accuracy (overall 90\u0026ndash;92%), outperforming individual datasets and improving both user\u0026rsquo;s and producer\u0026rsquo;s accuracies for grassland (\u0026gt;\u0026thinsp;84%). Consensus datasets also better captured expected elevation and slope gradients, producing more spatially coherent and ecologically realistic grassland patterns. By integrating multiple land-cover sources, consensus approaches effectively mitigated dataset-specific biases and increased the reliability of grassland mapping in heterogeneous mountain systems. Consequently, consensus land-cover products provide a robust and ecologically meaningful alternative to single-source datasets for environmental assessments in complex mountain regions.\u003c/p\u003e","manuscriptTitle":"Consensus land-cover mapping improves grassland classification in European mountain landscapes","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-12-08 13:16:32","doi":"10.21203/rs.3.rs-8096101/v1","editorialEvents":[{"type":"communityComments","content":0},{"type":"decision","content":"Revision requested","date":"2026-01-05T19:46:14+00:00","index":"","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2026-01-02T15:41:12+00:00","index":"hide","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2026-01-02T14:46:55+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"306774115774347016538369524076331167040","date":"2025-12-06T11:29:41+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"140758174692562395642214112978663897761","date":"2025-12-05T08:59:03+00:00","index":"hide","fulltext":""},{"type":"reviewersInvited","content":"","date":"2025-12-04T11:13:36+00:00","index":"","fulltext":""},{"type":"editorAssigned","content":"","date":"2025-12-04T11:08:43+00:00","index":"","fulltext":""},{"type":"editorInvited","content":"","date":"2025-11-28T10:03:01+00:00","index":"","fulltext":""},{"type":"checksComplete","content":"","date":"2025-11-19T18:31:18+00:00","index":"","fulltext":""},{"type":"submitted","content":"Scientific Reports","date":"2025-11-19T18:28:36+00:00","index":"","fulltext":""}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"scientific-reports","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"scirep","sideBox":"Learn more about [Scientific Reports](http://www.nature.com/srep/)","snPcode":"","submissionUrl":"","title":"Scientific Reports","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"stoa","reportingPortfolio":"Scientific Reports","inReviewEnabled":true,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"857214a4-fa7c-4438-a45e-dfbb3a50d0fc","owner":[],"postedDate":"December 8th, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"published-in-journal","subjectAreas":[{"id":59201764,"name":"Biological sciences/Ecology"},{"id":59201765,"name":"Earth and environmental sciences/Ecology"},{"id":59201766,"name":"Earth and environmental sciences/Environmental sciences"}],"tags":[],"updatedAt":"2026-02-16T16:04:33+00:00","versionOfRecord":{"articleIdentity":"rs-8096101","link":"https://doi.org/10.1038/s41598-026-39197-w","journal":{"identity":"scientific-reports","isVorOnly":false,"title":"Scientific Reports"},"publishedOn":"2026-02-10 15:57:10","publishedOnDateReadable":"February 10th, 2026"},"versionCreatedAt":"2025-12-08 13:16:32","video":"","vorDoi":"10.1038/s41598-026-39197-w","vorDoiUrl":"https://doi.org/10.1038/s41598-026-39197-w","workflowStages":[]},"version":"v1","identity":"rs-8096101","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-8096101","identity":"rs-8096101","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}