A dynamic geospatial digital twin resolves the riparian management paradox

A dynamic geospatial digital twin resolves the riparian management paradox | 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 A dynamic geospatial digital twin resolves the riparian management paradox Jong-hwa Park, Jae hun Shin, Seung-Hwan Go This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-7421152/v1 This work is licensed under a CC BY 4.0 License Status: Posted Version 1 posted You are reading this latest preprint version Abstract The intensification of extreme rainfall under climate change amplifies fluvial geohazards, creating a critical management paradox for riparian vegetation, which is essential for bank stability yet can increase flood risks. Conventional management, reliant on static hydraulic models, fails to capture the dynamic nature of these systems. Here we show a Geospatial Digital Twin (GDT) framework that integrates high-resolution unmanned aerial vehicle imagery with a deep learning model to create a living, high-fidelity representation of a river corridor. Applied to the Yi-dong stream, South Korea, our GDT accurately segmented seasonal vegetation changes and translated them into dynamic hydraulic roughness maps. By simulating distinct management scenarios, we quantified the trade-offs between flood mitigation and geomorphic stability, revealing that a balanced strategy involving willow planting enhances stability without dramatically increasing flood risk. By providing a virtual laboratory to test management outcomes before implementation, the GDT framework offers a powerful, scalable tool for developing climate-adaptive strategies. Earth and environmental sciences/Environmental sciences/Environmental impact Earth and environmental sciences/Hydrology Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Introduction The global climate crisis is accelerating the hydrological cycle, increasing the frequency and magnitude of extreme rainfall events and associated fluvial geohazards such as flash floods and erosion¹⁻³. This "new climate normal" poses a direct threat to global security and economic stability, forcing countries to manage adaptation costs⁴⁻⁶. In Northeast Asia, a recognized climate change hotspot, these risks are particularly acute, with intensified monsoon rains causing catastrophic damage⁷. South Korea, for example, has committed over $ 85 billion USD to its national climate adaptation plans, yet a growing portion of this budget is allocated to reactive disaster recovery rather than proactive risk reduction, highlighting a critical gap in predictive management tools⁸. Headwater catchments, which comprise over 75% of global river networks, are the front line of this crisis⁹. These systems are highly sensitive to hydro-climatic shifts, and their management is complicated by a fundamental paradox centred on riparian vegetation¹⁰. This vegetation is a cornerstone of Nature-based Solutions (NbS), providing critical bank stabilization, filtering pollutants, and supporting biodiversity¹¹⁻¹³. However, the riparian vegetation increases hydraulic roughness, impeding flood conveyance, raising water levels, and exacerbating inundation risk for adjacent communities and infrastructure¹⁴⁻¹⁶. This conflict forces a difficult trade-off: clearing vegetation to mitigate flood risk can trigger erosion and ecosystem collapse, while promoting dense vegetation for ecological health can increase flood hazards¹⁷. This dilemma is compounded by deep-rooted methodological limitations in conventional hydraulic modeling. Standard practice relies on static, spatially-averaged roughness coefficients (e.g., Manning's ' n ') that fail to capture the complex, three-dimensional structure and significant seasonal phenology of vegetation¹⁸⁻²⁰. This oversimplification can lead to critical errors in flood forecasting, undermining the efficacy of multi-billion-dollar adaptation investments and hindering the design of nuanced management strategies that balance competing objectives²¹ , ²². The inability to dynamically represent vegetation-flow interactions remains a primary source of uncertainty in flood risk assessment. To address this challenge, we propose a paradigm shift enabled by the convergence of robotics, artificial intelligence (AI), and environmental science. We introduce a Geospatial Digital Twin (GDT)—a virtual replica of a physical environment, dynamically updated with near-real-time data—as a framework for proactive geohazard management²³⁻²⁵. Our GDT integrates ultra-high-resolution data from Unmanned Aerial Vehicles (UAVs) with a U-Net deep learning model that automatically translates sensor data into physically meaningful model parameters²⁶⁻²⁸. This creates a living, predictive laboratory for testing management interventions before they are implemented on the ground. This study develops and validates this GDT framework on the Yi-dong stream in South Korea, demonstrating an end-to-end pipeline from data acquisition to dynamic hydraulic parameterization, quantifying how seasonal vegetation changes alter hydraulic conditions, and simulating management scenarios to reveal the trade-offs between flood mitigation and geomorphic stability. Results Study site and GDT framework The study was conducted along a 300-m reach of the Yi-dong stream in Goesan, Republic of Korea (36°52'11.2"N 127°51'21.3"E), a representative gravel-bed headwater channel in the Geum River basin (Fig. 2). The site experiences a temperate climate with hot, humid summers and cold, dry winters, driving significant seasonal changes in vegetation. The GDT framework integrates four key components (Fig. 1): (a) multi-temporal data acquisition using UAVs; (b) AI-powered semantic segmentation to classify the landscape; (c) dynamic simulation using AI-derived parameters to model hydro-geomorphic conditions; and (d) decision support through quantitative assessment of management trade-offs. AI-driven mapping of dynamic riparian vegetation A U-Net deep learning model was developed to automatically classify multi-temporal UAV orthomosaics into water and vegetation classes. The model demonstrated high performance for the peak vegetation (August) and dormant (December) seasons, achieving a Mean Intersection over Union (Mean IoU) of 0.612 and 0.974, respectively, with overall accuracies exceeding 98% (Table 1 ). This ensured a reliable data foundation for the GDT. Performance was lower for the senescence period in October (Mean IoU = 0.596), reflecting the challenge of classifying transitional vegetation states using only RGB imagery. This highlights a practical limitation and underscores the need for more advanced sensors (e.g., multispectral) to improve classification during phenological transitions. Despite this, the AI-generated maps accurately delineated the water-land interface and captured significant seasonal shifts in vegetation cover (Fig. 3). Table 1 U-Net model performance metrics on independent test data. Season (2023) Overall Accuracy Mean IoU F1 Score (Water) F1 Score (Vegetation) Aug 18 (Peak) 0.985 0.612 0.992 0.919 Oct 27 (Senescence) 0.761 0.596 0.699 0.771 Dec 4 (Dormant) 0.987 0.974 0.976 0.972 Dynamic 3D hydraulic roughness parameterization The 2D segmentation maps were transformed into dynamic 3D surface models by integrating them with a high-resolution Digital Surface Model (DSM) and a bare-earth Digital Terrain Model (DTM). For each date, AI-classified vegetation pixels were assigned height values derived from the Canopy Height Model (CHM = DSM - DTM), creating a 3D representation of the vegetation's physical structure (Fig. 3). From this dynamic 3D mesh, we derived a spatially distributed Manning's ' n ' roughness coefficient for each date. The spatially averaged baseline Manning's ' n ' varied dramatically by season, from a low of 0.025 in December to a high of 0.0677 in August. This nearly threefold increase underscores the importance of capturing seasonal phenology to accurately model flood conveyance. Scenario analysis of hydro-geomorphic trade-offs We simulated four distinct management scenarios using the August (peak vegetation) dataset: S1 (Baseline), S2 (Flood Conveyance: 50% shrub removal), S3 (Bank Stabilization: 20% increase in willow plantings), and S4 (Full Restoration: 50% increase in vegetation area and height). For each scenario, we recalculated the GDT's hydraulic parameters and ran 2D Shallow Water Equation (SWE) simulations to quantify the resulting trade-offs (Table 2 , Table 3 ). The Flood Conveyance scenario (S2) reduced the area-weighted Manning's ' n ' by 10.49%, resulting in higher average flow velocity but at the cost of a 50% reduction in stabilizing shrub cover. In contrast, the Full Restoration scenario (S4) increased hydraulic roughness by 2.51%, impeding flow. The Bank Stabilization scenario (S3) emerged as a balanced approach, slightly increasing roughness (+ 1.48%) alongside a 20% enhancement of stabilizing vegetation. These results are visualized in Fig. 4 and Fig. 5, which clearly position the Bank Stabilization scenario (S3) as an optimized strategy that enhances stability without dramatically compromising conveyance. Table 2 Quantitative summary of management scenario outcomes for the August dataset. Values in parentheses indicate the percentage change from the S1: Baseline scenario. The velocity proxy is calculated as 1/ n . Scenario Management Action Manning's n Shrub Area (m 2 ) Velocity Proxy (m/s) S1: Baseline Current condition 0.0677 10.01 14.77 S2: Flood Conveyance 50% Shrub removal 0.0606 (-10.49%) 5.05 (-49.55%) 16.5 (+ 11.71%) S3: Bank Stabilization 20% Willow planting 0.0687 (+ 1.48%) 12.01 (+ 19.98%) 14.56 (-1.42%) S4: Full Restoration 50% buffer planting 0.0694 (+ 2.51%) 15.02 (+ 50.05%) 14.40 (-2.50%) Table 3 Summary of 2D SWE Model Hydraulic Outputs. Wave passage time used Kinematic wave approximation. Scenario Spatially-Averaged Manning's n Peak Water Depth (m) Average Flow Velocity (m/s) Wave Passage Time (s) Aug-S1 0.0677 0.294 0.097 29.32 Aug-S2 0.0606 0.294 0.106 26.24 Aug-S3 0.0687 0.294 0.096 29.75 Aug-S4 0.0694 0.294 0.11 30.05 Oct-S1 0.0374 0.164 0.132 16.2 Dec-S1 0.0250 0.153 0.194 10.83 Discussion This study develops and validates a Geospatial Digital Twin framework that represents a significant methodological advance for environmental science and a powerful tool for climate adaptation. The core innovation is the creation of a direct, automated pipeline from raw UAV sensor data to dynamic, physically meaningful parameters for hydraulic models. This approach directly addresses a long-standing limitation in hydrology—the reliance on static, generalized roughness coefficients—by providing a mechanism to continuously update models with high-fidelity, real-world data that captures both the 3D structure and seasonal phenology of vegetation²⁹⁻³¹. The GDT's ability to function as a virtual laboratory for scenario testing is its most powerful attribute. The trade-off analysis (Fig. 5, Table 2 ) translates abstract management goals into quantifiable hydro-geomorphic consequences, providing clear, evidence-based answers to critical questions faced by river managers globally³²⁻³⁴. For instance, the "Flood Conveyance" scenario (S2) confirms the hydraulic efficiency of vegetation clearing but also quantifies the ecological cost: a 50% loss of stabilizing vegetation and an implied increase in long-term erosion risk³⁵⁻³⁷. Conversely, the "Full Restoration" scenario (S4) aligns with the principles of NbS but reveals that maximizing ecological vegetation can increase hydraulic roughness, potentially elevating flood risk for adjacent land³⁸⁻⁴⁰. Our framework allows stakeholders to navigate this complex decision landscape, moving beyond single-objective optimization towards designing for multi-objective resilience⁴¹⁻⁴³. This capability is crucial for implementing effective and defensible climate adaptation strategies that align with international frameworks like the Sendai Framework for Disaster Risk Reduction and the UN's Sustainable Development Goals⁴⁴⁻⁴⁶. The methodology presented here is designed to be a scalable and transferable blueprint. The integration of globally available UAV technology and open-source AI models means the framework can be deployed in diverse geographic and climatic settings⁴⁷⁻⁴⁹. It offers a pathway for developing national-scale digital twin infrastructures for water resource management, enabling a paradigm shift from reactive, post-disaster response to proactive, predictive, and risk-informed governance⁵⁰⁻⁵². Our study has limitations that open clear avenues for future research. The reduced performance of the U-Net model during the senescent period highlights the limitations of using only RGB data for classifying transitional vegetation states. Future works should explore the integration of multispectral or hyperspectral imagery to improve classification accuracy across all phenological stages. The hydraulic analysis used a simplified channel geometry; future iterations should incorporate more complex, spatially variable topography to improve simulation fidelity⁵³⁻⁵⁵. Furthermore, the GDT is currently one-way coupled. The next frontier is to create a two-way coupled and prognostic twin, where simulated hydraulic forces and erosion patterns feedback to influence models of vegetation growth and mortality, enabling long-term forecasts of landscape evolution under different climate scenarios⁵⁶⁻⁵⁸. In conclusion, by transforming high-resolution remote sensing data into actionable intelligence, the GDT framework offers a robust methodology for navigating the complex challenges of river management in the Anthropocene. It empowers scientists, managers, and policymakers to design resilient river landscapes that balance the imperatives of public safety, economic activity, and ecological vitality in a rapidly changing world⁵⁹ , ⁶⁰. Methods GDT Framework Overview Our workflow consists of four stages: ( 1 ) multi-temporal data acquisition using a UAV; ( 2 ) AI-powered semantic segmentation of UAV imagery; ( 3 ) dynamic 3D surface reconstruction and hydraulic parameterization; and ( 4 ) scenario-based trade-off analysis using a 2D hydraulic model (Fig. 1). UAV Data Acquisition and Photogrammetric Processing Imagery was captured using an eBee + fixed-wing UAV (senseFly) equipped with a S.O.D.A. 3D camera on August 18, October 27, and December 4, 2023. Flights were conducted at an altitude of 100 m, yielding a 2.5 cm Ground Sampling Distance (GSD). The collected imagery was processed using Structure-from-Motion (SfM) photogrammetry in Pix4Dmapper software to generate high-resolution orthomosaics and Digital Surface Models (DSMs) for each date. A bare-earth Digital Terrain Model (DTM) was derived from the leaf-off (December) dataset. AI-based Semantic Segmentation A U-Net convolutional neural network²⁶ was implemented in Keras with a TensorFlow backend to segment orthomosaics into three classes: (0) Background, ( 1 ) Water, and ( 2 ) Vegetation. The model was trained on manually labelled image patches (32x32 pixels for August and October; 64x64 pixels for December) with a 20% validation split. To address class imbalance, a weighted sparse categorical cross-entropy loss function was used⁶⁷. Three U-Net models were developed with progressively enhanced architectures. The August model used a basic U-Net trained for 30 epochs. The October model incorporated residual connections and spatial attention, trained with focal loss. The December model added batch normalization and refined skip connections, trained for up to 50 epochs with early stopping⁶⁹. Preprocessing steps were tailored for each dataset, including greenness filters, morphological operations, and Conditional Random Field (CRF) refinement to optimize classification accuracy⁷⁰ , ⁷¹. Hydraulic Parameterization and Roughness Calculation A Canopy Height Model (CHM) was calculated for each date as CHM = DSM − DTM. A dynamic 3D surface mesh was constructed by assigning height values from the CHM to pixels classified as vegetation. A spatially distributed Manning's 'n' roughness coefficient was derived using a modified quadratic sum equation⁶¹: $$\:n={n}_{b}^{2}+\frac{{C}_{D}ah{R}^{4/3}}{2g}$$ where \(\:{n}_{b}\) is the bed roughness (0.025), \(\:{C}_{D}\) is the vegetation drag coefficient (0.7), \(\:a\) is the frontal area density (m − 1 ), \(\:h\) is the vegetation height from the CHM, and \(\:g\) is the acceleration due to gravity. \(\:R\) is hydraulic radius (m). The frontal density, \(\:a\) , was adjusted for management scenarios. To obtain a single representative roughness value, an area-weighted average of Manning's 'n' was calculated for the inundated area. 2D Hydraulic Modeling (Shallow Water Equations) 2D hydraulic modeling was performed using a finite difference scheme solving the shallow water equations⁶² , ⁶³. The initial condition used a Gaussian Hump to initiate localized height variation⁶⁴ , ⁶⁵: $$\:h\left(x,y\right)={h}_{0}+A\bullet\:exp\left(-\frac{{(x-{x}_{c})}^{2}+{(y-{y}_{c})}^{2}}{2{\sigma\:}^{2}}\right)\:$$ where \(\:{h}_{0}\) is the initial water depth, \(\:A\) is the hump amplitude, ( \(\:{x}_{c}\) , \(\:{y}_{c}\) ) is the hump center, and \(\:{\sigma\:}^{2}\) controls the spread. The model solves the continuity equation for mass conservation and momentum equations in the x and y directions. The friction slope terms ( \(\:{S}_{fx}\) , \(\:{S}_{fy}\) ) were calculated using Manning's equation: $$\:{S}_{fx}=\frac{{n}^{2}u\sqrt{{u}^{2}+{v}^{2}}}{{h}^{4/3}}g,\:{S}_{fy}=\frac{{n}^{2}v\sqrt{{u}^{2}+{v}^{2}}}{{h}^{4/3}}g,$$ The model was run on a trapezoidal channel geometry with fixed water levels, a time step of 0.005 s, and a simulation duration of 50 s. Management Scenario Formulation Four management scenarios were formulated for the August dataset: S1 (Baseline): current conditions; S2 (Flood Conveyance): 50% reduction in shrub area and height; S3 (Bank Stabilization): 20% increase in willow planting area; and S4 (Full Restoration): 50% increase in total vegetation area and height. For each scenario, area of frontal cover was adjusted. Statistical Analysis To compare the velocity distributions among the different seasonal and management scenarios (Fig. 4d), a one-way analysis of variance (ANOVA) was conducted using R software. This was followed by post hoc pairwise comparisons using Tukey's HSD test to identify which specific groups differed significantly. The alpha level for all tests was set at 0.05. Declarations Data Availability The minimal dataset necessary to interpret, replicate and build upon the findings reported in the article will be deposited in a publicly available repository and the accession codes will be provided upon acceptance. Code Availability The custom R and Python scripts used for the U-Net segmentation, hydraulic parameterization, and 2D SWE model analysis are available from the corresponding author upon reasonable request. Competing Interests The authors declare no competing interests. Author Contributions J.H.S. conceived the study, performed the analysis, and wrote the initial draft. S.H.G. assisted with data collection and processing. J.H.P. supervised the project and revised the manuscript. All authors contributed to the final version of the manuscript. Acknowledgements This research received no external funding. References IPCC. Climate Change 2021: The Physical Science Basis. Contribution of Working Group I to the Sixth Assessment Report of the Intergovernmental Panel on Climate Change (Cambridge Univ. Press, 2021). Tabari, H. Climate change impact on flood and extreme precipitation increases with water availability. Sci. Rep. 10, 13768 (2020). Myhre, G. et al. Multi-model simulations of aerosol and ozone forcing due to anthropogenic emissions. Geosci. Model Dev. 6, 1853–1876 (2013). Bauer, P., Thorpe, A. & Brunet, G. The quiet revolution of numerical weather prediction. Nature 525, 47–55 (2015). United Nations Office for Disaster Risk Reduction (UNDRR). Global Assessment Report on Disaster Risk Reduction 2022 (UNDRR, 2022). Hallegatte, S., Rentschler, J. & Rozenberg, J. Lifelines: The Resilient Infrastructure Opportunity (World Bank, 2019). Carmichael, G. R., Streets, D. G. & Wang, T. A new paradigm for air quality in East Asia. Environ. Sci. Technol. 54, 13781–13782 (2020). National Assembly Budget Office (NABO). Evaluation of Climate Crisis Adaptation Policy (NABO, 2024). Allen, D. C. Headwater streams and rivers. in River Ecology and Management 53–73 (Springer, 2019). Wohl, E. The functions of headwater streams. Nat. Rev. Earth Environ. 3, 499–512 (2022). Cohen-Shacham, E., Walters, G., Janzen, C. & Maginnis, S. Nature-based Solutions to address global societal challenges (IUCN, 2016). Seddon, N. et al. Getting the message right on nature-based solutions to climate change. Glob. Chang. Biol. 27, 1518–1546 (2021). Gurnell, A. M. Plants as river system engineers. Earth Surf. Process. Landforms 39, 4–25 (2014). Darby, S. E. Effect of riparian vegetation on flow resistance and flood potential. J. Hydraul. Eng. 125, 443–454 (1999). Järvelä, J. Determination of flow resistance of vegetated channels. Water Eng. Res. 4, 13–27 (2003). Nepf, H. M. Hydrodynamics of vegetated channels. J. Hydraul. Res. 50, 262–279 (2012). Dufour, S. & Piégay, H. From the myth of a lost paradise to targeted river restoration: forget natural references and focus on human benefits. River Res. Appl. 25, 568–581 (2009). Zhang, H., Wang, Z., Xu, W. & Wang, H. Determination of Emergent Vegetation Effects on Manning’s Coefficient of Gradually Varied Flow. IEEE Access 7, 12345–12356 (2019). Papaioannou, G., Vasiliades, L. & Loukas, A. A review of methods for calibrating Manning’s roughness in river hydraulic models. Hydrology 5, 27 (2018). Horritt, M. S. & Bates, P. D. Evaluation of 1D and 2D numerical models for predicting river flood inundation. J. Hydrol. 268, 87–99 (2002). Pappenberger, F. et al. Visualizing probabilistic flood forecast information: expert preferences and perceptions of best practice in uncertainty communication. Hydrol. Process. 27, 132–146 (2013). Kundzewicz, Z. W. et al. Flood risk and climate change: global and regional perspectives. Hydrol. Sci. J. 59, 1–28 (2014). Bauer, P., Stevens, B. & Hazeleger, W. A digital twin of Earth for the green transition. Nat. Clim. Chang. 11, 80–83 (2021). Nolet, C., Poerbandono, & van der Wegen, M. Developing a digital twin for river systems: A review of its potential. J. Hydrol. 601, 126693 (2021). Kraft, M., Vosse, J. & Noss, C. Digital twins in hydrology – A review of opportunities and limitations. WIREs Water 9, e1591 (2022). Ronneberger, O., Fischer, P. & Brox, T. U-Net: Convolutional networks for biomedical image segmentation. in Medical Image Computing and Computer-Assisted Intervention – MICCAI 2015 234–241 (Springer, 2015). LeCun, Y., Bengio, Y. & Hinton, G. Deep learning. Nature 521, 436–444 (2015). Lhoest, S. et al. River-vegetation mapping with deep learning and UAV imagery. Remote Sens. 14, 1374 (2022). Marcus, W. A. & Fonstad, M. A. Optical remote mapping of rivers. in Tools in Fluvial Geomorphology 196–221 (John Wiley & Sons, Ltd, 2010). Michez, A. et al. Classification of riverbank vegetation by UAV-acquired photogrammetry and hyperspectral imagery. Catena 168, 14–26 (2018). Petryk, S. & Bosmajian, G. Analysis of flow through vegetation. J. Hydraul. Div. 101, 871–884 (1975). Downs, P. W. & Gregory, K. J. River Channel Management: Towards Sustainable Catchment Hydrosystems (Arnold, 2004). Wohl, E. et al. The role of wood in river process. Geomorphology 279, 1–20 (2017). Palmer, M. A. et al. Standards for ecologically successful river restoration. J. Appl. Ecol. 42, 208–217 (2005). Simon, A. & Collison, A. J. C. Quantifying the mechanical and hydrologic effects of riparian vegetation on stream-bank stability. Earth Surf. Process. Landforms 27, 527–546 (2002). Abernethy, B. & Rutherfurd, I. D. The distribution and strength of riparian tree roots in relation to riverbank reinforcement. Hydrol. Process. 15, 63–79 (2001). Thorne, C. R. Effects of vegetation on riverbank erosion and stability. in Vegetation and Erosion 125–144 (Wiley, 1990). Keesstra, S. et al. The superior effect of nature-based solutions in land management for enhancing ecosystem services. Sci. Total Environ. 610–611, 997–1009 (2018). Dadson, S. J. et al. A restatement of the natural science evidence concerning catchment-based natural flood management in the UK. Proc. R. Soc. A. 473, 20160706 (2017). Burgess-Gamble, L. et al. Working with Natural Processes to Reduce Flood Risk (Environment Agency, 2017). Beck, M. W. et al. Rigorously valuing the role of U.S. coral reefs in coastal hazard risk reduction . (U.S. Geological Survey, 2018). Di Baldassarre, G. et al. Sociohydrology: A new science for a changing world. Hydrol. Process. 27, 1259–1263 (2013). Sivapalan, M., Savenije, H. H. G. & Blöschl, G. Socio-hydrology: A new science of people and water. Hydrol. Process. 26, 1270–1276 (2012). United Nations. Sendai Framework for Disaster Risk Reduction 2015–2030 (United Nations, 2015). United Nations. Transforming our world: the 2030 Agenda for Sustainable Development (United Nations, 2015). World Bank. Managing climate risks for a more resilient future (World Bank, 2021). Westoby, M. J. et al. ‘Structure-from-Motion’ photogrammetry: A low-cost, effective tool for geoscience applications. Geomorphology 179, 300–314 (2012). Colomina, I. & Molina, P. Unmanned aerial systems for photogrammetry and remote sensing: A review. ISPRS J. Photogramm. Remote Sens. 92, 79–97 (2014). Go, S. H., Jeong, K. S. & Park, J. H. Assessing Stream Vegetation Dynamics and Revetment Impact Using Time-Series RGB UAV Images and ResNeXt101 CNNs. Korean J. Remote Sens. 40, 9–18 (2024). El Saddik, A. Digital twins: The convergence of multimedia, IoT, and artificial intelligence. IEEE MultiMedia 25, 87–92 (2018). Wan, L. et al. A review of digital twin in the architecture, engineering and construction industry. J. Inf. Technol. Constr. 26, 539–560 (2021). VoPham, T., Hart, J. E., Laden, F. & Chiang, Y.-Y. Emerging trends in geospatial artificial intelligence (geoAI): potential for environmental health. Environ. Health 17, 40 (2018). Bates, P. D., Horritt, M. S. & Fewtrell, T. J. A simple inertial formulation of the shallow water equations for modelling flood inundation. J. Hydrol. 387, 33–45 (2010). Nicholas, A. P. & Smith, G. H. S. A numerical model of river meandering. Hydrol. Process. 13, 983–998 (1999). Schubert, J. E. & Sanders, B. F. Building a digital twin of a coastal floodplain using a coupled 1D-2D hydrodynamic model. J. Hydrol. 436–437, 48–63 (2012). Corenblit, D., Tabacchi, E., Steiger, J. & Gurnell, A. M. Reciprocal interactions and adjustments between fluvial landforms and vegetation dynamics in river corridors: A review of complementary approaches. Earth-Sci. Rev. 84, 56–86 (2007). Van De Wiel, M. J. & Darby, S. E. A new model to analyse the impact of woody riparian vegetation on the geotechnical stability of riverbanks. Earth Surf. Process. Landforms 32, 2185–2198 (2007). Murray, A. B. & Paola, C. A cellular model of braided rivers. Nature 371, 54–57 (1994). Grill, G. et al. Mapping the world’s free-flowing rivers. Nature 569, 215–221 (2019). Vörösmarty, C. J. et al. Global threats to human water security and river biodiversity. Nature 467, 555–561 (2010). Baptist, M. J. et al. On inducing equations for vegetation resistance. J. Hydraul. Res. 45, 435–450 (2007). Bates, P. D., Horritt, M. S. & Fewtrell, T. J. A simple inertial formulation of the shallow water equations for efficient two-dimensional flood inundation modelling. J. Hydrol. 387, 33–45 (2010). Hunter, N. M., Bates, P. D., Horritt, M. S. & Wilson, M. D. Simple spatially-distributed models for predicting flood inundation: A review. Geomorphology 90, 208–225 (2007). Ge, L., Cheung, K. F. & Kobayashi, M. H. Stochastic Solution for Uncertainty Propagation in Nonlinear Shallow-Water Equations. J. Hydraul. Eng. 134, 17–27 (2008). Yoon, S. B. Propagation of distant tsunamis over slowly varying topography. J. Geophys. Res. 107, 3140 (2002). Sohrabi, S., Afzalimehr, H. & Singh, V. P. Estimation of drag coefficient of emergent and submerged vegetation patches with various densities and arrangements in open channel flow. ISH J. Hydraul. Eng. 29, 297–307 (2022). Lin, T.-Y., Goyal, P., Girshick, R., He, K. & Dollár, P. Focal Loss for Dense Object Detection. IEEE Trans. Pattern Anal. Mach. Intell. 42, 318–327 (2020). Pu, Y., Lu, X., Soubry, I. & Guo, X. Early detection of woody plant encroachment in Canadian prairies using UAV imagery and transformer-based deep learning. Ecol. Inform. 90, 103354 (2025). Benameur, N., Mahmoudi, R., Deriche, M. et al. An Improved Approach for Cardiac MRI Segmentation based on a 3D UNet Combined with Papillary Muscle Exclusion. Arab J Sci Eng (2025). https://doi.org/10.1007/s13369-025-10310-x Serouart, M. et al. SegVeg: Segmenting RGB Images into Green and Senescent Vegetation by Combining Deep and Shallow Methods. Plant Phenomics 2022, 9803570 (2022). Huang, Q., Liang, G., Tan, C. & Dong, F. Res2-UNet++: a deep learning image post-processing method for electrical resistance tomography. Meas. Sci. Technol. 35, 105403 (2024). Akbari, R. & Hessami-Kermani, M.-R. A new method for dividing flood period in the variable-parameter Muskingum models. Hydrol. Res. 53, 241–257 (2022). Additional Declarations There is NO Competing Interest. Cite Share Download PDF Status: Posted Version 1 posted You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-7421152","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Article","associatedPublications":[],"authors":[{"id":510569113,"identity":"81017057-4a48-4230-971d-88b6ea9d0951","order_by":0,"name":"Jong-hwa Park","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAAyklEQVRIiWNgGAWjYBACxhlA3PjPhsEAKmCAVzlcSwNbGglaGCTAWg6ToIV5du/BjzN4zsubS/cYMPyoYTA2byDksDnnkiU3SNw23DnnjAFjzzEGM5kDhLTMyDGQfGBwO8HgRo4BA28Dg40EIYcBtRj/fJBwDqyF8S+RWswkNxw4ANbCDLTFjAgteWmWMxuSDXfOSCs4LHNMwpigFsMZuYdv9jbYyZtLJG98+KbGxnAGQS0NPAjOAXA0EQLyDDyEFY2CUTAKRsEIBwDm9z1yzikJkAAAAABJRU5ErkJggg==","orcid":"https://orcid.org/0000-0003-0592-6323","institution":"Chungbuk National University","correspondingAuthor":true,"prefix":"","firstName":"Jong-hwa","middleName":"","lastName":"Park","suffix":""},{"id":510569114,"identity":"a6fa61fb-d01c-404d-85b1-c4ff70d5392e","order_by":1,"name":"Jae hun Shin","email":"","orcid":"https://orcid.org/0000-0002-2606-0568","institution":"Chungbuk National University","correspondingAuthor":false,"prefix":"","firstName":"Jae","middleName":"hun","lastName":"Shin","suffix":""},{"id":510569115,"identity":"aed90617-d4ca-488e-9fad-22800830d4a6","order_by":2,"name":"Seung-Hwan Go","email":"","orcid":"https://orcid.org/0000-0002-3800-044X","institution":"Chungbuk National University","correspondingAuthor":false,"prefix":"","firstName":"Seung-Hwan","middleName":"","lastName":"Go","suffix":""}],"badges":[],"createdAt":"2025-08-21 01:10:09","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-7421152/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-7421152/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":91085580,"identity":"6f23fa07-bbb3-4d83-be58-406a2eb773df","added_by":"auto","created_at":"2025-09-11 12:22:00","extension":"jpg","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":1454951,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eConceptual framework of the Geospatial Digital Twin.\u003c/strong\u003e \u003cstrong\u003ea)\u003c/strong\u003e Data Acquisition: Multi-temporal, high-resolution data (RGB imagery, topography) are captured by UAVs. \u003cstrong\u003eb)\u003c/strong\u003e AI-Powered Analysis: A U-Net deep learning model performs semantic segmentation on the imagery to classify the landscape and extract dynamic vegetation parameters. \u003cstrong\u003ec)\u003c/strong\u003e Dynamic Simulation: The AI-derived parameters are fed into a hydraulic model to generate dynamic roughness maps and simulate hydro-geomorphic conditions. \u003cstrong\u003ed)\u003c/strong\u003e Decision Support: The simulation outputs provide a quantitative assessment of management trade-offs, enabling stakeholders to select optimized, climate-resilient strategies.\u003c/p\u003e","description":"","filename":"Figure10821.jpg","url":"https://assets-eu.researchsquare.com/files/rs-7421152/v1/60c66ecc32bce3e3e0946fd4.jpg"},{"id":91083484,"identity":"d1381f31-2b63-416e-8a35-3951d1794bd0","added_by":"auto","created_at":"2025-09-11 12:06:00","extension":"jpg","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":2513655,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eStudy area location.\u003c/strong\u003e \u003cstrong\u003ea)\u003c/strong\u003e Location of South Korea and the Geum River basin, with Goesan County highlighted. \u003cstrong\u003eb)\u003c/strong\u003e Location of the study site on the Yi-dong stream. \u003cstrong\u003ec)\u003c/strong\u003e A high-resolution UAV orthomosaic of the 23-m section where 2D hydraulic analysis is implemented.\u003c/p\u003e","description":"","filename":"Figure20821.jpg","url":"https://assets-eu.researchsquare.com/files/rs-7421152/v1/98bc46a63cdbee71b038a6bd.jpg"},{"id":91083486,"identity":"a3f46fd9-619e-4561-964b-957711404ebc","added_by":"auto","created_at":"2025-09-11 12:06:00","extension":"jpg","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":3647344,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eThe UAV-AI to 3D model pipeline.\u003c/strong\u003e \u003cstrong\u003ea)\u003c/strong\u003e The original high-resolution UAV orthomosaic. \u003cstrong\u003eb)\u003c/strong\u003e The AI-generated semantic segmentation map, classifying the landscape into water (blue) and vegetation (green). \u003cstrong\u003ec)\u003c/strong\u003e The resulting 3D surface mesh where vegetation height is explicitly represented, forming the basis for dynamic hydraulic roughness calculation.\u003c/p\u003e","description":"","filename":"Figure38021.jpg","url":"https://assets-eu.researchsquare.com/files/rs-7421152/v1/adc144be634eba308a8f627f.jpg"},{"id":91084523,"identity":"af0919cf-86fc-482e-a40b-e9cd3c8ef922","added_by":"auto","created_at":"2025-09-11 12:14:00","extension":"jpg","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":2916490,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eSeasonal variations in roughness and 2D SWE dynamics.\u003c/strong\u003e \u003cstrong\u003ea)\u003c/strong\u003e Seasonal dynamics of area-weighted Manning's '\u003cem\u003en\u003c/em\u003e' reveal clear seasonal and management-driven differences. \u003cstrong\u003eb,c)\u003c/strong\u003e Transient water depth dynamics at a downstream point for August and December, respectively, show how different roughness coefficients alter wave passage. \u003cstrong\u003ed)\u003c/strong\u003e Boxplots showing velocity distributions across the six series. Statistical significance was assessed using one-way ANOVA followed by post hoc pairwise comparisons. Asterisks denote significant differences: \u003cem\u003ep\u003c/em\u003e \u0026lt; 0.05, *\u003cem\u003ep\u003c/em\u003e \u0026lt; 0.01, **\u003cem\u003ep\u003c/em\u003e \u0026lt; 0.001, and ns: not significant.\u003c/p\u003e","description":"","filename":"Figure48021.jpg","url":"https://assets-eu.researchsquare.com/files/rs-7421152/v1/73f4ed0e790ebb1c8c5b3eb6.jpg"},{"id":91083489,"identity":"98810116-b117-4e1d-97ef-e18bbf643f05","added_by":"auto","created_at":"2025-09-11 12:06:00","extension":"jpg","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":1075101,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eGeohazard trade-off analysis of management scenarios.\u003c/strong\u003e The plot positions the three management scenarios (S2, S3, S4) relative to the baseline (S1) based on their impacts on two competing objectives: change in shrub area (a proxy for bank stability) and change in a flow velocity proxy (a proxy for flood conveyance). The quadrants represent different management outcomes. This visualization provides an intuitive tool for decision-making, positioning S3 as a balanced strategy.\u003c/p\u003e","description":"","filename":"Figure50821.jpg","url":"https://assets-eu.researchsquare.com/files/rs-7421152/v1/540dfd1b5a4d654f2ec7cdf1.jpg"},{"id":92619128,"identity":"73a161e5-6ff0-42be-a0dc-64591947fabc","added_by":"auto","created_at":"2025-10-01 18:34:05","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":12567064,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-7421152/v1/28052afa-6e8e-47a8-9c5e-dc1891362a1e.pdf"}],"financialInterests":"There is \u003cb\u003eNO\u003c/b\u003e Competing Interest.","formattedTitle":"A dynamic geospatial digital twin resolves the riparian management paradox","fulltext":[{"header":"Introduction","content":"\u003cp\u003eThe global climate crisis is accelerating the hydrological cycle, increasing the frequency and magnitude of extreme rainfall events and associated fluvial geohazards such as flash floods and erosion\u0026sup1;⁻\u0026sup3;. This \"new climate normal\" poses a direct threat to global security and economic stability, forcing countries to manage adaptation costs⁴⁻⁶. In Northeast Asia, a recognized climate change hotspot, these risks are particularly acute, with intensified monsoon rains causing catastrophic damage⁷. South Korea, for example, has committed over \u003cspan\u003e$\u003c/span\u003e85\u0026nbsp;billion USD to its national climate adaptation plans, yet a growing portion of this budget is allocated to reactive disaster recovery rather than proactive risk reduction, highlighting a critical gap in predictive management tools⁸.\u003c/p\u003e\u003cp\u003eHeadwater catchments, which comprise over 75% of global river networks, are the front line of this crisis⁹. These systems are highly sensitive to hydro-climatic shifts, and their management is complicated by a fundamental paradox centred on riparian vegetation\u0026sup1;⁰. This vegetation is a cornerstone of Nature-based Solutions (NbS), providing critical bank stabilization, filtering pollutants, and supporting biodiversity\u0026sup1;\u0026sup1;⁻\u0026sup1;\u0026sup3;. However, the riparian vegetation increases hydraulic roughness, impeding flood conveyance, raising water levels, and exacerbating inundation risk for adjacent communities and infrastructure\u0026sup1;⁴⁻\u0026sup1;⁶. This conflict forces a difficult trade-off: clearing vegetation to mitigate flood risk can trigger erosion and ecosystem collapse, while promoting dense vegetation for ecological health can increase flood hazards\u0026sup1;⁷.\u003c/p\u003e\u003cp\u003eThis dilemma is compounded by deep-rooted methodological limitations in conventional hydraulic modeling. Standard practice relies on static, spatially-averaged roughness coefficients (e.g., Manning's '\u003cem\u003en\u003c/em\u003e') that fail to capture the complex, three-dimensional structure and significant seasonal phenology of vegetation\u0026sup1;⁸⁻\u0026sup2;⁰. This oversimplification can lead to critical errors in flood forecasting, undermining the efficacy of multi-billion-dollar adaptation investments and hindering the design of nuanced management strategies that balance competing objectives\u0026sup2;\u0026sup1;\u003csup\u003e,\u003c/sup\u003e\u0026sup2;\u0026sup2;. The inability to dynamically represent vegetation-flow interactions remains a primary source of uncertainty in flood risk assessment.\u003c/p\u003e\u003cp\u003eTo address this challenge, we propose a paradigm shift enabled by the convergence of robotics, artificial intelligence (AI), and environmental science. We introduce a Geospatial Digital Twin (GDT)\u0026mdash;a virtual replica of a physical environment, dynamically updated with near-real-time data\u0026mdash;as a framework for proactive geohazard management\u0026sup2;\u0026sup3;⁻\u0026sup2;⁵. Our GDT integrates ultra-high-resolution data from Unmanned Aerial Vehicles (UAVs) with a U-Net deep learning model that automatically translates sensor data into physically meaningful model parameters\u0026sup2;⁶⁻\u0026sup2;⁸. This creates a living, predictive laboratory for testing management interventions before they are implemented on the ground. This study develops and validates this GDT framework on the Yi-dong stream in South Korea, demonstrating an end-to-end pipeline from data acquisition to dynamic hydraulic parameterization, quantifying how seasonal vegetation changes alter hydraulic conditions, and simulating management scenarios to reveal the trade-offs between flood mitigation and geomorphic stability.\u003c/p\u003e"},{"header":"Results","content":"\u003cdiv id=\"Sec3\" class=\"Section2\"\u003e\n\u003ch2\u003eStudy site and GDT framework\u003c/h2\u003e\n\u003cp\u003eThe study was conducted along a 300-m reach of the Yi-dong stream in Goesan, Republic of Korea (36\u0026deg;52'11.2\"N 127\u0026deg;51'21.3\"E), a representative gravel-bed headwater channel in the Geum River basin (Fig.\u0026nbsp;2). The site experiences a temperate climate with hot, humid summers and cold, dry winters, driving significant seasonal changes in vegetation. The GDT framework integrates four key components (Fig.\u0026nbsp;1): (a) multi-temporal data acquisition using UAVs; (b) AI-powered semantic segmentation to classify the landscape; (c) dynamic simulation using AI-derived parameters to model hydro-geomorphic conditions; and (d) decision support through quantitative assessment of management trade-offs.\u003c/p\u003e\n\u003c/div\u003e\n\u003ch3\u003eAI-driven mapping of dynamic riparian vegetation\u003c/h3\u003e\n\u003cp\u003eA U-Net deep learning model was developed to automatically classify multi-temporal UAV orthomosaics into water and vegetation classes. The model demonstrated high performance for the peak vegetation (August) and dormant (December) seasons, achieving a Mean Intersection over Union (Mean IoU) of 0.612 and 0.974, respectively, with overall accuracies exceeding 98% (Table\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e1\u003c/span\u003e). This ensured a reliable data foundation for the GDT. Performance was lower for the senescence period in October (Mean IoU\u0026thinsp;=\u0026thinsp;0.596), reflecting the challenge of classifying transitional vegetation states using only RGB imagery. This highlights a practical limitation and underscores the need for more advanced sensors (e.g., multispectral) to improve classification during phenological transitions. Despite this, the AI-generated maps accurately delineated the water-land interface and captured significant seasonal shifts in vegetation cover (Fig.\u0026nbsp;3).\u003c/p\u003e\n\u003cdiv class=\"gridtable\"\u003e\n\u003cdiv class=\"colspec\" align=\"left\"\u003e\u0026nbsp;\u003c/div\u003e\n\u003ctable id=\"Tab1\" border=\"1\"\u003e\u003ccaption\u003e\n\u003cdiv class=\"CaptionNumber\"\u003eTable 1\u003c/div\u003e\n\u003cdiv class=\"CaptionContent\"\u003e\n\u003cp\u003eU-Net model performance metrics on independent test data.\u003c/p\u003e\n\u003c/div\u003e\n\u003c/caption\u003e\n\u003cthead\u003e\n\u003ctr\u003e\n\u003cth align=\"left\"\u003e\n\u003cp\u003eSeason (2023)\u003c/p\u003e\n\u003c/th\u003e\n\u003cth align=\"left\"\u003e\n\u003cp\u003eOverall Accuracy\u003c/p\u003e\n\u003c/th\u003e\n\u003cth align=\"left\"\u003e\n\u003cp\u003eMean IoU\u003c/p\u003e\n\u003c/th\u003e\n\u003cth align=\"left\"\u003e\n\u003cp\u003eF1 Score (Water)\u003c/p\u003e\n\u003c/th\u003e\n\u003cth align=\"left\"\u003e\n\u003cp\u003eF1 Score (Vegetation)\u003c/p\u003e\n\u003c/th\u003e\n\u003c/tr\u003e\n\u003c/thead\u003e\n\u003ctbody\u003e\n\u003ctr\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003eAug 18 (Peak)\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e0.985\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e0.612\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e0.992\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e0.919\u003c/p\u003e\n\u003c/td\u003e\n\u003c/tr\u003e\n\u003ctr\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003eOct 27 (Senescence)\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e0.761\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e0.596\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e0.699\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e0.771\u003c/p\u003e\n\u003c/td\u003e\n\u003c/tr\u003e\n\u003ctr\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003eDec 4 (Dormant)\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e0.987\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e0.974\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e0.976\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e0.972\u003c/p\u003e\n\u003c/td\u003e\n\u003c/tr\u003e\n\u003c/tbody\u003e\n\u003c/table\u003e\n\u003c/div\u003e\n\u003ch3\u003eDynamic 3D hydraulic roughness parameterization\u003c/h3\u003e\n\u003cp\u003eThe 2D segmentation maps were transformed into dynamic 3D surface models by integrating them with a high-resolution Digital Surface Model (DSM) and a bare-earth Digital Terrain Model (DTM). For each date, AI-classified vegetation pixels were assigned height values derived from the Canopy Height Model (CHM\u0026thinsp;=\u0026thinsp;DSM - DTM), creating a 3D representation of the vegetation's physical structure (Fig.\u0026nbsp;3). From this dynamic 3D mesh, we derived a spatially distributed Manning's '\u003cem\u003en\u003c/em\u003e' roughness coefficient for each date. The spatially averaged baseline Manning's '\u003cem\u003en\u003c/em\u003e' varied dramatically by season, from a low of 0.025 in December to a high of 0.0677 in August. This nearly threefold increase underscores the importance of capturing seasonal phenology to accurately model flood conveyance.\u003c/p\u003e\n\u003cdiv id=\"Sec8\" class=\"Section2\"\u003e\n\u003ch2\u003eScenario analysis of hydro-geomorphic trade-offs\u003c/h2\u003e\n\u003c/div\u003e\n\u003cp\u003eWe simulated four distinct management scenarios using the August (peak vegetation) dataset: S1 (Baseline), S2 (Flood Conveyance: 50% shrub removal), S3 (Bank Stabilization: 20% increase in willow plantings), and S4 (Full Restoration: 50% increase in vegetation area and height). For each scenario, we recalculated the GDT's hydraulic parameters and ran 2D Shallow Water Equation (SWE) simulations to quantify the resulting trade-offs (Table\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e2\u003c/span\u003e, Table\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e3\u003c/span\u003e). The Flood Conveyance scenario (S2) reduced the area-weighted Manning's '\u003cem\u003en\u003c/em\u003e' by 10.49%, resulting in higher average flow velocity but at the cost of a 50% reduction in stabilizing shrub cover. In contrast, the Full Restoration scenario (S4) increased hydraulic roughness by 2.51%, impeding flow. The Bank Stabilization scenario (S3) emerged as a balanced approach, slightly increasing roughness (+\u0026thinsp;1.48%) alongside a 20% enhancement of stabilizing vegetation. These results are visualized in Fig.\u0026nbsp;4 and Fig.\u0026nbsp;5, which clearly position the Bank Stabilization scenario (S3) as an optimized strategy that enhances stability without dramatically compromising conveyance.\u003c/p\u003e\n\u003cdiv class=\"gridtable\"\u003e\n\u003cdiv class=\"colspec\" align=\"left\"\u003e\u0026nbsp;\u003c/div\u003e\n\u003ctable id=\"Tab2\" border=\"1\"\u003e\u003ccaption\u003e\n\u003cdiv class=\"CaptionNumber\"\u003eTable 2\u003c/div\u003e\n\u003cdiv class=\"CaptionContent\"\u003e\n\u003cp\u003e\u003cstrong\u003eQuantitative summary of management scenario outcomes for the August dataset.\u003c/strong\u003e Values in parentheses indicate the percentage change from the S1: Baseline scenario. The velocity proxy is calculated as 1/\u003cem\u003en\u003c/em\u003e.\u003c/p\u003e\n\u003c/div\u003e\n\u003c/caption\u003e\n\u003cthead\u003e\n\u003ctr\u003e\n\u003cth align=\"left\"\u003e\n\u003cp\u003eScenario\u003c/p\u003e\n\u003c/th\u003e\n\u003cth align=\"left\"\u003e\n\u003cp\u003eManagement Action\u003c/p\u003e\n\u003c/th\u003e\n\u003cth align=\"left\"\u003e\n\u003cp\u003eManning's n\u003c/p\u003e\n\u003c/th\u003e\n\u003cth align=\"left\"\u003e\n\u003cp\u003eShrub Area (m\u003csup\u003e2\u003c/sup\u003e)\u003c/p\u003e\n\u003c/th\u003e\n\u003cth align=\"left\"\u003e\n\u003cp\u003eVelocity Proxy (m/s)\u003c/p\u003e\n\u003c/th\u003e\n\u003c/tr\u003e\n\u003c/thead\u003e\n\u003ctbody\u003e\n\u003ctr\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003eS1: Baseline\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003eCurrent condition\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e0.0677\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e10.01\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e14.77\u003c/p\u003e\n\u003c/td\u003e\n\u003c/tr\u003e\n\u003ctr\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003eS2: Flood Conveyance\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003e50% Shrub removal\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e0.0606\u003c/p\u003e\n\u003cp\u003e(-10.49%)\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e5.05 (-49.55%)\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e16.5\u003c/p\u003e\n\u003cp\u003e(+\u0026thinsp;11.71%)\u003c/p\u003e\n\u003c/td\u003e\n\u003c/tr\u003e\n\u003ctr\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003eS3: Bank Stabilization\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003e20% Willow planting\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e0.0687\u003c/p\u003e\n\u003cp\u003e(+\u0026thinsp;1.48%)\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e12.01 (+\u0026thinsp;19.98%)\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e14.56\u003c/p\u003e\n\u003cp\u003e(-1.42%)\u003c/p\u003e\n\u003c/td\u003e\n\u003c/tr\u003e\n\u003ctr\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003eS4: Full Restoration\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003e50% buffer planting\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e0.0694\u003c/p\u003e\n\u003cp\u003e(+\u0026thinsp;2.51%)\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e15.02 (+\u0026thinsp;50.05%)\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e14.40\u003c/p\u003e\n\u003cp\u003e(-2.50%)\u003c/p\u003e\n\u003c/td\u003e\n\u003c/tr\u003e\n\u003c/tbody\u003e\n\u003c/table\u003e\n\u003c/div\u003e\n\u003cdiv class=\"gridtable\"\u003e\n\u003cdiv class=\"colspec\" align=\"left\"\u003e\u0026nbsp;\u003c/div\u003e\n\u003cdiv class=\"colspec\" align=\"char\"\u003e\u0026nbsp;\u003c/div\u003e\n\u003ctable id=\"Tab3\" border=\"1\"\u003e\u003ccaption\u003e\n\u003cdiv class=\"CaptionNumber\"\u003eTable 3\u003c/div\u003e\n\u003cdiv class=\"CaptionContent\"\u003e\n\u003cp\u003e\u003cstrong\u003eSummary of 2D SWE Model Hydraulic Outputs.\u003c/strong\u003e Wave passage time used Kinematic wave approximation.\u003c/p\u003e\n\u003c/div\u003e\n\u003c/caption\u003e\n\u003cthead\u003e\n\u003ctr\u003e\n\u003cth align=\"left\"\u003e\n\u003cp\u003eScenario\u003c/p\u003e\n\u003c/th\u003e\n\u003cth align=\"left\"\u003e\n\u003cp\u003eSpatially-Averaged Manning's n\u003c/p\u003e\n\u003c/th\u003e\n\u003cth align=\"left\"\u003e\n\u003cp\u003ePeak Water Depth (m)\u003c/p\u003e\n\u003c/th\u003e\n\u003cth align=\"left\"\u003e\n\u003cp\u003eAverage Flow Velocity (m/s)\u003c/p\u003e\n\u003c/th\u003e\n\u003cth align=\"left\"\u003e\n\u003cp\u003eWave Passage Time (s)\u003c/p\u003e\n\u003c/th\u003e\n\u003c/tr\u003e\n\u003c/thead\u003e\n\u003ctbody\u003e\n\u003ctr\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003eAug-S1\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e0.0677\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e0.294\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e0.097\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e29.32\u003c/p\u003e\n\u003c/td\u003e\n\u003c/tr\u003e\n\u003ctr\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003eAug-S2\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e0.0606\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e0.294\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e0.106\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e26.24\u003c/p\u003e\n\u003c/td\u003e\n\u003c/tr\u003e\n\u003ctr\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003eAug-S3\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e0.0687\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e0.294\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e0.096\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e29.75\u003c/p\u003e\n\u003c/td\u003e\n\u003c/tr\u003e\n\u003ctr\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003eAug-S4\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e0.0694\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e0.294\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e0.11\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e30.05\u003c/p\u003e\n\u003c/td\u003e\n\u003c/tr\u003e\n\u003ctr\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003eOct-S1\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e0.0374\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e0.164\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e0.132\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e16.2\u003c/p\u003e\n\u003c/td\u003e\n\u003c/tr\u003e\n\u003ctr\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003eDec-S1\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e0.0250\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e0.153\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e0.194\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e10.83\u003c/p\u003e\n\u003c/td\u003e\n\u003c/tr\u003e\n\u003c/tbody\u003e\n\u003c/table\u003e\n\u003c/div\u003e\n\u003ch3\u003e\u0026nbsp;\u003c/h3\u003e\n\u003cdiv id=\"Sec11\" class=\"Section2\"\u003e\u0026nbsp;\u003c/div\u003e"},{"header":"Discussion","content":"\u003cp\u003eThis study develops and validates a Geospatial Digital Twin framework that represents a significant methodological advance for environmental science and a powerful tool for climate adaptation. The core innovation is the creation of a direct, automated pipeline from raw UAV sensor data to dynamic, physically meaningful parameters for hydraulic models. This approach directly addresses a long-standing limitation in hydrology—the reliance on static, generalized roughness coefficients—by providing a mechanism to continuously update models with high-fidelity, real-world data that captures both the 3D structure and seasonal phenology of vegetation²⁹⁻³¹.\u003c/p\u003e\u003cp\u003eThe GDT's ability to function as a virtual laboratory for scenario testing is its most powerful attribute. The trade-off analysis (Fig.\u0026nbsp;5, Table\u0026nbsp;\u003cspan refid=\"Tab2\" class=\"InternalRef\"\u003e2\u003c/span\u003e) translates abstract management goals into quantifiable hydro-geomorphic consequences, providing clear, evidence-based answers to critical questions faced by river managers globally³²⁻³⁴. For instance, the \"Flood Conveyance\" scenario (S2) confirms the hydraulic efficiency of vegetation clearing but also quantifies the ecological cost: a 50% loss of stabilizing vegetation and an implied increase in long-term erosion risk³⁵⁻³⁷. Conversely, the \"Full Restoration\" scenario (S4) aligns with the principles of NbS but reveals that maximizing ecological vegetation can increase hydraulic roughness, potentially elevating flood risk for adjacent land³⁸⁻⁴⁰. Our framework allows stakeholders to navigate this complex decision landscape, moving beyond single-objective optimization towards designing for multi-objective resilience⁴¹⁻⁴³.\u003c/p\u003e\u003cp\u003eThis capability is crucial for implementing effective and defensible climate adaptation strategies that align with international frameworks like the Sendai Framework for Disaster Risk Reduction and the UN's Sustainable Development Goals⁴⁴⁻⁴⁶. The methodology presented here is designed to be a scalable and transferable blueprint. The integration of globally available UAV technology and open-source AI models means the framework can be deployed in diverse geographic and climatic settings⁴⁷⁻⁴⁹. It offers a pathway for developing national-scale digital twin infrastructures for water resource management, enabling a paradigm shift from reactive, post-disaster response to proactive, predictive, and risk-informed governance⁵⁰⁻⁵².\u003c/p\u003e\u003cp\u003eOur study has limitations that open clear avenues for future research. The reduced performance of the U-Net model during the senescent period highlights the limitations of using only RGB data for classifying transitional vegetation states. Future works should explore the integration of multispectral or hyperspectral imagery to improve classification accuracy across all phenological stages. The hydraulic analysis used a simplified channel geometry; future iterations should incorporate more complex, spatially variable topography to improve simulation fidelity⁵³⁻⁵⁵. Furthermore, the GDT is currently one-way coupled. The next frontier is to create a two-way coupled and prognostic twin, where simulated hydraulic forces and erosion patterns feedback to influence models of vegetation growth and mortality, enabling long-term forecasts of landscape evolution under different climate scenarios⁵⁶⁻⁵⁸.\u003c/p\u003e\u003cp\u003eIn conclusion, by transforming high-resolution remote sensing data into actionable intelligence, the GDT framework offers a robust methodology for navigating the complex challenges of river management in the Anthropocene. It empowers scientists, managers, and policymakers to design resilient river landscapes that balance the imperatives of public safety, economic activity, and ecological vitality in a rapidly changing world⁵⁹\u003csup\u003e,\u003c/sup\u003e⁶⁰.\u003c/p\u003e"},{"header":"Methods","content":"\u003ch2\u003eGDT Framework Overview\u003c/h2\u003e\u003cp\u003eOur workflow consists of four stages: (\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e) multi-temporal data acquisition using a UAV; (\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e) AI-powered semantic segmentation of UAV imagery; (\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e) dynamic 3D surface reconstruction and hydraulic parameterization; and (\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e) scenario-based trade-off analysis using a 2D hydraulic model (Fig.\u0026nbsp;1).\u003c/p\u003e\u003ch2\u003eUAV Data Acquisition and Photogrammetric Processing\u003c/h2\u003e\u003cp\u003eImagery was captured using an eBee + fixed-wing UAV (senseFly) equipped with a S.O.D.A. 3D camera on August 18, October 27, and December 4, 2023. Flights were conducted at an altitude of 100 m, yielding a 2.5 cm Ground Sampling Distance (GSD). The collected imagery was processed using Structure-from-Motion (SfM) photogrammetry in Pix4Dmapper software to generate high-resolution orthomosaics and Digital Surface Models (DSMs) for each date. A bare-earth Digital Terrain Model (DTM) was derived from the leaf-off (December) dataset.\u003c/p\u003e\u003ch2\u003eAI-based Semantic Segmentation\u003c/h2\u003e\u003cp\u003eA U-Net convolutional neural network²⁶ was implemented in Keras with a TensorFlow backend to segment orthomosaics into three classes: (0) Background, (\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e) Water, and (\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e) Vegetation. The model was trained on manually labelled image patches (32x32 pixels for August and October; 64x64 pixels for December) with a 20% validation split. To address class imbalance, a weighted sparse categorical cross-entropy loss function was used⁶⁷. Three U-Net models were developed with progressively enhanced architectures. The August model used a basic U-Net trained for 30 epochs. The October model incorporated residual connections and spatial attention, trained with focal loss. The December model added batch normalization and refined skip connections, trained for up to 50 epochs with early stopping⁶⁹. Preprocessing steps were tailored for each dataset, including greenness filters, morphological operations, and Conditional Random Field (CRF) refinement to optimize classification accuracy⁷⁰\u003csup\u003e,\u003c/sup\u003e⁷¹.\u003c/p\u003e\u003ch2\u003eHydraulic Parameterization and Roughness Calculation\u003c/h2\u003e\u003cp\u003eA Canopy Height Model (CHM) was calculated for each date as CHM = DSM − DTM. A dynamic 3D surface mesh was constructed by assigning height values from the CHM to pixels classified as vegetation. A spatially distributed Manning's 'n' roughness coefficient was derived using a modified quadratic sum equation⁶¹:\u003c/p\u003e\u003cdiv id=\"Equa\" class=\"Equation\"\u003e\u003cdiv format=\"TEX\" class=\"mathdisplay\" id=\"FileID_Equa\" name=\"EquationSource\"\u003e\n$$\\:n={n}_{b}^{2}+\\frac{{C}_{D}ah{R}^{4/3}}{2g}$$\u003c/div\u003e\u003c/div\u003e\u003cp\u003e\u003c/p\u003e\u003cp\u003ewhere \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{n}_{b}\\)\u003c/span\u003e\u003c/span\u003e is the bed roughness (0.025), \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{C}_{D}\\)\u003c/span\u003e\u003c/span\u003e is the vegetation drag coefficient (0.7), \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:a\\)\u003c/span\u003e\u003c/span\u003e is the frontal area density (m\u003csup\u003e− 1\u003c/sup\u003e), \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:h\\)\u003c/span\u003e\u003c/span\u003e is the vegetation height from the CHM, and \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:g\\)\u003c/span\u003e\u003c/span\u003e is the acceleration due to gravity. \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:R\\)\u003c/span\u003e\u003c/span\u003e is hydraulic radius (m). The frontal density, \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:a\\)\u003c/span\u003e\u003c/span\u003e, was adjusted for management scenarios. To obtain a single representative roughness value, an area-weighted average of Manning's 'n' was calculated for the inundated area.\u003c/p\u003e\u003cp\u003e\u003cb\u003e2D Hydraulic Modeling (Shallow Water Equations)\u003c/b\u003e\u003c/p\u003e\u003cp\u003e2D hydraulic modeling was performed using a finite difference scheme solving the shallow water equations⁶²\u003csup\u003e,\u003c/sup\u003e⁶³. The initial condition used a Gaussian Hump to initiate localized height variation⁶⁴\u003csup\u003e,\u003c/sup\u003e⁶⁵:\u003c/p\u003e\u003cdiv id=\"Equb\" class=\"Equation\"\u003e\u003cdiv format=\"TEX\" class=\"mathdisplay\" id=\"FileID_Equb\" name=\"EquationSource\"\u003e\n$$\\:h\\left(x,y\\right)={h}_{0}+A\\bullet\\:exp\\left(-\\frac{{(x-{x}_{c})}^{2}+{(y-{y}_{c})}^{2}}{2{\\sigma\\:}^{2}}\\right)\\:$$\u003c/div\u003e\u003c/div\u003e\u003cp\u003e\u003c/p\u003e\u003cp\u003ewhere \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{h}_{0}\\)\u003c/span\u003e\u003c/span\u003e is the initial water depth, \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:A\\)\u003c/span\u003e\u003c/span\u003e is the hump amplitude, (\u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{x}_{c}\\)\u003c/span\u003e\u003c/span\u003e, \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{y}_{c}\\)\u003c/span\u003e\u003c/span\u003e) is the hump center, and \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{\\sigma\\:}^{2}\\)\u003c/span\u003e\u003c/span\u003e controls the spread. The model solves the continuity equation for mass conservation and momentum equations in the x and y directions. The friction slope terms (\u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{S}_{fx}\\)\u003c/span\u003e\u003c/span\u003e, \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{S}_{fy}\\)\u003c/span\u003e\u003c/span\u003e) were calculated using Manning's equation:\u003c/p\u003e\u003cdiv id=\"Equc\" class=\"Equation\"\u003e\u003cdiv format=\"TEX\" class=\"mathdisplay\" id=\"FileID_Equc\" name=\"EquationSource\"\u003e\n$$\\:{S}_{fx}=\\frac{{n}^{2}u\\sqrt{{u}^{2}+{v}^{2}}}{{h}^{4/3}}g,\\:{S}_{fy}=\\frac{{n}^{2}v\\sqrt{{u}^{2}+{v}^{2}}}{{h}^{4/3}}g,$$\u003c/div\u003e\u003c/div\u003e\u003cp\u003e\u003c/p\u003e\u003cp\u003eThe model was run on a trapezoidal channel geometry with fixed water levels, a time step of 0.005 s, and a simulation duration of 50 s.\u003c/p\u003e\u003ch2\u003eManagement Scenario Formulation\u003c/h2\u003e\u003cp\u003eFour management scenarios were formulated for the August dataset: S1 (Baseline): current conditions; S2 (Flood Conveyance): 50% reduction in shrub area and height; S3 (Bank Stabilization): 20% increase in willow planting area; and S4 (Full Restoration): 50% increase in total vegetation area and height. For each scenario, area of frontal cover was adjusted.\u003c/p\u003e\u003ch2\u003eStatistical Analysis\u003c/h2\u003e\u003cp\u003eTo compare the velocity distributions among the different seasonal and management scenarios (Fig.\u0026nbsp;4d), a one-way analysis of variance (ANOVA) was conducted using R software. This was followed by post hoc pairwise comparisons using Tukey's HSD test to identify which specific groups differed significantly. The alpha level for all tests was set at 0.05.\u003c/p\u003e"},{"header":"Declarations","content":"\u003ch2\u003eData Availability\u003c/h2\u003e\n\u003cp\u003eThe minimal dataset necessary to interpret, replicate and build upon the findings reported in the article will be deposited in a publicly available repository and the accession codes will be provided upon acceptance.\u003c/p\u003e\n\u003ch2\u003eCode Availability\u003c/h2\u003e\n\u003cp\u003eThe custom R and Python scripts used for the U-Net segmentation, hydraulic parameterization, and 2D SWE model analysis are available from the corresponding author upon reasonable request.\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\u003eAuthor Contributions\u003c/h2\u003e\u003cp\u003eJ.H.S. conceived the study, performed the analysis, and wrote the initial draft. S.H.G. assisted with data collection and processing. J.H.P. supervised the project and revised the manuscript. All authors contributed to the final version of the manuscript.\u003c/p\u003e\u003ch2\u003eAcknowledgements\u003c/h2\u003e\u003cp\u003eThis research received no external funding.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\u003cli\u003e\u003cspan\u003eIPCC. \u003cem\u003eClimate Change 2021: The Physical Science Basis. Contribution of Working Group I to the Sixth Assessment Report of the Intergovernmental Panel on Climate Change\u003c/em\u003e (Cambridge Univ. Press, 2021).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eTabari, H. Climate change impact on flood and extreme precipitation increases with water availability. \u003cem\u003eSci. Rep.\u003c/em\u003e 10, 13768 (2020).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eMyhre, G. \u003cem\u003eet al.\u003c/em\u003e Multi-model simulations of aerosol and ozone forcing due to anthropogenic emissions. \u003cem\u003eGeosci. Model Dev.\u003c/em\u003e 6, 1853\u0026ndash;1876 (2013).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eBauer, P., Thorpe, A. \u0026amp; Brunet, G. The quiet revolution of numerical weather prediction. \u003cem\u003eNature\u003c/em\u003e 525, 47\u0026ndash;55 (2015).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eUnited Nations Office for Disaster Risk Reduction (UNDRR). \u003cem\u003eGlobal Assessment Report on Disaster Risk Reduction 2022\u003c/em\u003e (UNDRR, 2022).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eHallegatte, S., Rentschler, J. \u0026amp; Rozenberg, J. \u003cem\u003eLifelines: The Resilient Infrastructure Opportunity\u003c/em\u003e (World Bank, 2019).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eCarmichael, G. R., Streets, D. G. \u0026amp; Wang, T. A new paradigm for air quality in East Asia. \u003cem\u003eEnviron. Sci. Technol.\u003c/em\u003e 54, 13781\u0026ndash;13782 (2020).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eNational Assembly Budget Office (NABO). \u003cem\u003eEvaluation of Climate Crisis Adaptation Policy\u003c/em\u003e (NABO, 2024).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eAllen, D. C. Headwater streams and rivers. in \u003cem\u003eRiver Ecology and Management\u003c/em\u003e 53\u0026ndash;73 (Springer, 2019).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eWohl, E. The functions of headwater streams. \u003cem\u003eNat. Rev. Earth Environ.\u003c/em\u003e 3, 499\u0026ndash;512 (2022).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eCohen-Shacham, E., Walters, G., Janzen, C. \u0026amp; Maginnis, S. \u003cem\u003eNature-based Solutions to address global societal challenges\u003c/em\u003e (IUCN, 2016).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eSeddon, N. \u003cem\u003eet al.\u003c/em\u003e Getting the message right on nature-based solutions to climate change. \u003cem\u003eGlob. Chang. Biol.\u003c/em\u003e 27, 1518\u0026ndash;1546 (2021).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eGurnell, A. M. Plants as river system engineers. \u003cem\u003eEarth Surf. Process. Landforms\u003c/em\u003e 39, 4\u0026ndash;25 (2014).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eDarby, S. E. Effect of riparian vegetation on flow resistance and flood potential. \u003cem\u003eJ. Hydraul. Eng.\u003c/em\u003e 125, 443\u0026ndash;454 (1999).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eJ\u0026auml;rvel\u0026auml;, J. Determination of flow resistance of vegetated channels. \u003cem\u003eWater Eng. Res.\u003c/em\u003e 4, 13\u0026ndash;27 (2003).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eNepf, H. M. Hydrodynamics of vegetated channels. \u003cem\u003eJ. Hydraul. Res.\u003c/em\u003e 50, 262\u0026ndash;279 (2012).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eDufour, S. \u0026amp; Pi\u0026eacute;gay, H. From the myth of a lost paradise to targeted river restoration: forget natural references and focus on human benefits. \u003cem\u003eRiver Res. Appl.\u003c/em\u003e 25, 568\u0026ndash;581 (2009).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eZhang, H., Wang, Z., Xu, W. \u0026amp; Wang, H. Determination of Emergent Vegetation Effects on Manning\u0026rsquo;s Coefficient of Gradually Varied Flow. \u003cem\u003eIEEE Access\u003c/em\u003e 7, 12345\u0026ndash;12356 (2019).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003ePapaioannou, G., Vasiliades, L. \u0026amp; Loukas, A. A review of methods for calibrating Manning\u0026rsquo;s roughness in river hydraulic models. \u003cem\u003eHydrology\u003c/em\u003e 5, 27 (2018).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eHorritt, M. S. \u0026amp; Bates, P. D. Evaluation of 1D and 2D numerical models for predicting river flood inundation. \u003cem\u003eJ. Hydrol.\u003c/em\u003e 268, 87\u0026ndash;99 (2002).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003ePappenberger, F. \u003cem\u003eet al.\u003c/em\u003e Visualizing probabilistic flood forecast information: expert preferences and perceptions of best practice in uncertainty communication. \u003cem\u003eHydrol. Process.\u003c/em\u003e 27, 132\u0026ndash;146 (2013).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eKundzewicz, Z. W. \u003cem\u003eet al.\u003c/em\u003e Flood risk and climate change: global and regional perspectives. \u003cem\u003eHydrol. Sci. J.\u003c/em\u003e 59, 1\u0026ndash;28 (2014).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eBauer, P., Stevens, B. \u0026amp; Hazeleger, W. A digital twin of Earth for the green transition. \u003cem\u003eNat. Clim. Chang.\u003c/em\u003e 11, 80\u0026ndash;83 (2021).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eNolet, C., Poerbandono, \u0026amp; van der Wegen, M. Developing a digital twin for river systems: A review of its potential. \u003cem\u003eJ. Hydrol.\u003c/em\u003e 601, 126693 (2021).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eKraft, M., Vosse, J. \u0026amp; Noss, C. Digital twins in hydrology \u0026ndash; A review of opportunities and limitations. \u003cem\u003eWIREs Water\u003c/em\u003e 9, e1591 (2022).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eRonneberger, O., Fischer, P. \u0026amp; Brox, T. U-Net: Convolutional networks for biomedical image segmentation. in \u003cem\u003eMedical Image Computing and Computer-Assisted Intervention \u0026ndash; MICCAI 2015\u003c/em\u003e 234\u0026ndash;241 (Springer, 2015).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eLeCun, Y., Bengio, Y. \u0026amp; Hinton, G. Deep learning. \u003cem\u003eNature\u003c/em\u003e 521, 436\u0026ndash;444 (2015).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eLhoest, S. \u003cem\u003eet al.\u003c/em\u003e River-vegetation mapping with deep learning and UAV imagery. \u003cem\u003eRemote Sens.\u003c/em\u003e 14, 1374 (2022).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eMarcus, W. A. \u0026amp; Fonstad, M. A. Optical remote mapping of rivers. in \u003cem\u003eTools in Fluvial Geomorphology\u003c/em\u003e 196\u0026ndash;221 (John Wiley \u0026amp; Sons, Ltd, 2010).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eMichez, A. \u003cem\u003eet al.\u003c/em\u003e Classification of riverbank vegetation by UAV-acquired photogrammetry and hyperspectral imagery. \u003cem\u003eCatena\u003c/em\u003e 168, 14\u0026ndash;26 (2018).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003ePetryk, S. \u0026amp; Bosmajian, G. Analysis of flow through vegetation. \u003cem\u003eJ. Hydraul. Div.\u003c/em\u003e 101, 871\u0026ndash;884 (1975).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eDowns, P. W. \u0026amp; Gregory, K. J. \u003cem\u003eRiver Channel Management: Towards Sustainable Catchment Hydrosystems\u003c/em\u003e (Arnold, 2004).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eWohl, E. \u003cem\u003eet al.\u003c/em\u003e The role of wood in river process. \u003cem\u003eGeomorphology\u003c/em\u003e 279, 1\u0026ndash;20 (2017).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003ePalmer, M. A. \u003cem\u003eet al.\u003c/em\u003e Standards for ecologically successful river restoration. \u003cem\u003eJ. Appl. Ecol.\u003c/em\u003e 42, 208\u0026ndash;217 (2005).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eSimon, A. \u0026amp; Collison, A. J. C. Quantifying the mechanical and hydrologic effects of riparian vegetation on stream-bank stability. \u003cem\u003eEarth Surf. Process. Landforms\u003c/em\u003e 27, 527\u0026ndash;546 (2002).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eAbernethy, B. \u0026amp; Rutherfurd, I. D. The distribution and strength of riparian tree roots in relation to riverbank reinforcement. \u003cem\u003eHydrol. Process.\u003c/em\u003e 15, 63\u0026ndash;79 (2001).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eThorne, C. R. Effects of vegetation on riverbank erosion and stability. in \u003cem\u003eVegetation and Erosion\u003c/em\u003e 125\u0026ndash;144 (Wiley, 1990).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eKeesstra, S. \u003cem\u003eet al.\u003c/em\u003e The superior effect of nature-based solutions in land management for enhancing ecosystem services. \u003cem\u003eSci. Total Environ.\u003c/em\u003e 610\u0026ndash;611, 997\u0026ndash;1009 (2018).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eDadson, S. J. \u003cem\u003eet al.\u003c/em\u003e A restatement of the natural science evidence concerning catchment-based natural flood management in the UK. \u003cem\u003eProc. R. Soc. A.\u003c/em\u003e 473, 20160706 (2017).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eBurgess-Gamble, L. \u003cem\u003eet al. Working with Natural Processes to Reduce Flood Risk\u003c/em\u003e (Environment Agency, 2017).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eBeck, M. W. \u003cem\u003eet al. Rigorously valuing the role of U.S. coral reefs in coastal hazard risk reduction\u003c/em\u003e. (U.S. Geological Survey, 2018).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eDi Baldassarre, G. \u003cem\u003eet al.\u003c/em\u003e Sociohydrology: A new science for a changing world. \u003cem\u003eHydrol. Process.\u003c/em\u003e 27, 1259\u0026ndash;1263 (2013).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eSivapalan, M., Savenije, H. H. G. \u0026amp; Bl\u0026ouml;schl, G. Socio-hydrology: A new science of people and water. \u003cem\u003eHydrol. Process.\u003c/em\u003e 26, 1270\u0026ndash;1276 (2012).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eUnited Nations. \u003cem\u003eSendai Framework for Disaster Risk Reduction 2015\u0026ndash;2030\u003c/em\u003e (United Nations, 2015).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eUnited Nations. \u003cem\u003eTransforming our world: the 2030 Agenda for Sustainable Development\u003c/em\u003e (United Nations, 2015).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eWorld Bank. \u003cem\u003eManaging climate risks for a more resilient future\u003c/em\u003e (World Bank, 2021).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eWestoby, M. J. \u003cem\u003eet al.\u003c/em\u003e \u0026lsquo;Structure-from-Motion\u0026rsquo; photogrammetry: A low-cost, effective tool for geoscience applications. \u003cem\u003eGeomorphology\u003c/em\u003e 179, 300\u0026ndash;314 (2012).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eColomina, I. \u0026amp; Molina, P. Unmanned aerial systems for photogrammetry and remote sensing: A review. \u003cem\u003eISPRS J. Photogramm. Remote Sens.\u003c/em\u003e 92, 79\u0026ndash;97 (2014).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eGo, S. H., Jeong, K. S. \u0026amp; Park, J. H. Assessing Stream Vegetation Dynamics and Revetment Impact Using Time-Series RGB UAV Images and ResNeXt101 CNNs. \u003cem\u003eKorean J. Remote Sens.\u003c/em\u003e 40, 9\u0026ndash;18 (2024).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eEl Saddik, A. Digital twins: The convergence of multimedia, IoT, and artificial intelligence. \u003cem\u003eIEEE MultiMedia\u003c/em\u003e 25, 87\u0026ndash;92 (2018).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eWan, L. \u003cem\u003eet al.\u003c/em\u003e A review of digital twin in the architecture, engineering and construction industry. \u003cem\u003eJ. Inf. Technol. Constr.\u003c/em\u003e 26, 539\u0026ndash;560 (2021).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eVoPham, T., Hart, J. E., Laden, F. \u0026amp; Chiang, Y.-Y. Emerging trends in geospatial artificial intelligence (geoAI): potential for environmental health. \u003cem\u003eEnviron. Health\u003c/em\u003e 17, 40 (2018).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eBates, P. D., Horritt, M. S. \u0026amp; Fewtrell, T. J. A simple inertial formulation of the shallow water equations for modelling flood inundation. \u003cem\u003eJ. Hydrol.\u003c/em\u003e 387, 33\u0026ndash;45 (2010).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eNicholas, A. P. \u0026amp; Smith, G. H. S. A numerical model of river meandering. \u003cem\u003eHydrol. Process.\u003c/em\u003e 13, 983\u0026ndash;998 (1999).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eSchubert, J. E. \u0026amp; Sanders, B. F. Building a digital twin of a coastal floodplain using a coupled 1D-2D hydrodynamic model. \u003cem\u003eJ. Hydrol.\u003c/em\u003e 436\u0026ndash;437, 48\u0026ndash;63 (2012).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eCorenblit, D., Tabacchi, E., Steiger, J. \u0026amp; Gurnell, A. M. Reciprocal interactions and adjustments between fluvial landforms and vegetation dynamics in river corridors: A review of complementary approaches. \u003cem\u003eEarth-Sci. Rev.\u003c/em\u003e 84, 56\u0026ndash;86 (2007).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eVan De Wiel, M. J. \u0026amp; Darby, S. E. A new model to analyse the impact of woody riparian vegetation on the geotechnical stability of riverbanks. \u003cem\u003eEarth Surf. Process. Landforms\u003c/em\u003e 32, 2185\u0026ndash;2198 (2007).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eMurray, A. B. \u0026amp; Paola, C. A cellular model of braided rivers. \u003cem\u003eNature\u003c/em\u003e 371, 54\u0026ndash;57 (1994).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eGrill, G. \u003cem\u003eet al.\u003c/em\u003e Mapping the world\u0026rsquo;s free-flowing rivers. \u003cem\u003eNature\u003c/em\u003e 569, 215\u0026ndash;221 (2019).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eV\u0026ouml;r\u0026ouml;smarty, C. J. \u003cem\u003eet al.\u003c/em\u003e Global threats to human water security and river biodiversity. \u003cem\u003eNature\u003c/em\u003e 467, 555\u0026ndash;561 (2010).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eBaptist, M. J. \u003cem\u003eet al.\u003c/em\u003e On inducing equations for vegetation resistance. \u003cem\u003eJ. Hydraul. Res.\u003c/em\u003e 45, 435\u0026ndash;450 (2007).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eBates, P. D., Horritt, M. S. \u0026amp; Fewtrell, T. J. A simple inertial formulation of the shallow water equations for efficient two-dimensional flood inundation modelling. \u003cem\u003eJ. Hydrol.\u003c/em\u003e 387, 33\u0026ndash;45 (2010).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eHunter, N. M., Bates, P. D., Horritt, M. S. \u0026amp; Wilson, M. D. Simple spatially-distributed models for predicting flood inundation: A review. \u003cem\u003eGeomorphology\u003c/em\u003e 90, 208\u0026ndash;225 (2007).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eGe, L., Cheung, K. F. \u0026amp; Kobayashi, M. H. Stochastic Solution for Uncertainty Propagation in Nonlinear Shallow-Water Equations. \u003cem\u003eJ. Hydraul. Eng.\u003c/em\u003e 134, 17\u0026ndash;27 (2008).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eYoon, S. B. Propagation of distant tsunamis over slowly varying topography. \u003cem\u003eJ. Geophys. Res.\u003c/em\u003e 107, 3140 (2002).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eSohrabi, S., Afzalimehr, H. \u0026amp; Singh, V. P. Estimation of drag coefficient of emergent and submerged vegetation patches with various densities and arrangements in open channel flow. \u003cem\u003eISH J. Hydraul. Eng.\u003c/em\u003e 29, 297\u0026ndash;307 (2022).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eLin, T.-Y., Goyal, P., Girshick, R., He, K. \u0026amp; Doll\u0026aacute;r, P. Focal Loss for Dense Object Detection. \u003cem\u003eIEEE Trans. Pattern Anal. Mach. Intell.\u003c/em\u003e 42, 318\u0026ndash;327 (2020).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003ePu, Y., Lu, X., Soubry, I. \u0026amp; Guo, X. Early detection of woody plant encroachment in Canadian prairies using UAV imagery and transformer-based deep learning. \u003cem\u003eEcol. Inform.\u003c/em\u003e 90, 103354 (2025).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eBenameur, N., Mahmoudi, R., Deriche, M. \u003cem\u003eet al.\u003c/em\u003e An Improved Approach for Cardiac MRI Segmentation based on a 3D UNet Combined with Papillary Muscle Exclusion. \u003cem\u003eArab J Sci Eng\u003c/em\u003e (2025). \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1007/s13369-025-10310-x\u003c/span\u003e\u003cspan address=\"10.1007/s13369-025-10310-x\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eSerouart, M. \u003cem\u003eet al.\u003c/em\u003e SegVeg: Segmenting RGB Images into Green and Senescent Vegetation by Combining Deep and Shallow Methods. \u003cem\u003ePlant Phenomics\u003c/em\u003e 2022, 9803570 (2022).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eHuang, Q., Liang, G., Tan, C. \u0026amp; Dong, F. Res2-UNet++: a deep learning image post-processing method for electrical resistance tomography. \u003cem\u003eMeas. Sci. Technol.\u003c/em\u003e 35, 105403 (2024).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eAkbari, R. \u0026amp; Hessami-Kermani, M.-R. A new method for dividing flood period in the variable-parameter Muskingum models. \u003cem\u003eHydrol. Res.\u003c/em\u003e 53, 241\u0026ndash;257 (2022).\u003c/span\u003e\u003c/li\u003e\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":true,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true},"keywords":"","lastPublishedDoi":"10.21203/rs.3.rs-7421152/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-7421152/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eThe intensification of extreme rainfall under climate change amplifies fluvial geohazards, creating a critical management paradox for riparian vegetation, which is essential for bank stability yet can increase flood risks. Conventional management, reliant on static hydraulic models, fails to capture the dynamic nature of these systems. Here we show a Geospatial Digital Twin (GDT) framework that integrates high-resolution unmanned aerial vehicle imagery with a deep learning model to create a living, high-fidelity representation of a river corridor. Applied to the Yi-dong stream, South Korea, our GDT accurately segmented seasonal vegetation changes and translated them into dynamic hydraulic roughness maps. By simulating distinct management scenarios, we quantified the trade-offs between flood mitigation and geomorphic stability, revealing that a balanced strategy involving willow planting enhances stability without dramatically increasing flood risk. By providing a virtual laboratory to test management outcomes before implementation, the GDT framework offers a powerful, scalable tool for developing climate-adaptive strategies.\u003c/p\u003e","manuscriptTitle":"A dynamic geospatial digital twin resolves the riparian management paradox","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-09-11 12:05:55","doi":"10.21203/rs.3.rs-7421152/v1","editorialEvents":[{"type":"communityComments","content":0}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"5546d3f6-4c7a-44a9-8dc9-dfb5b4ce92ea","owner":[],"postedDate":"September 11th, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[{"id":54234225,"name":"Earth and environmental sciences/Environmental sciences/Environmental impact"},{"id":54234226,"name":"Earth and environmental sciences/Hydrology"}],"tags":[],"updatedAt":"2025-10-01T18:25:54+00:00","versionOfRecord":[],"versionCreatedAt":"2025-09-11 12:05:55","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-7421152","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-7421152","identity":"rs-7421152","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}