Global River Models (GRMs), which simulate river flow and flood processes, have rapidly developed in recent decades. However, these advancements necessitate meaningful and standardized quality assessments and comparisons against a suitable set of observational variables using appropriate metrics, a requirement currently lacking within GRM communities. This study proposes the implementation of a benchmark system designed to facilitate the assessment of river models and enables comparisons against established benchmarks. The benchmark system incorporates satellite remote sensing data, including water surface elevation and inundation extent information, with necessary preprocessing. Consequently, this evaluation system encompasses a larger geographical area compared to traditional methods relying solely on in-situ river discharge measurements for GRMs. A set of evaluation and comparison metrics has been developed, including a quantile-based comparison metric that allows for a comprehensive analysis of multiple simulation outputs. The test application of this benchmark system to a global river model (CaMa-Flood), utilizing diverse runoff inputs, illustrates that the incorporation of bias-corrected runoff data leads to improved model performance across various observational variables and performance metrics. The current iteration of the benchmark system is suitable for global-scale assessments and can effectively evaluate the impact of model development as well as facilitate intercomparisons among different models. The source codes are accessiable from https://doi.org/10.5281/zenodo.10903211.

Understanding spatial and temporal variations in terrestrial waters is key to assessing the global hydrological cycle. The future Surface Water and Ocean Topography (SWOT) satellite mission will observe the elevation and slope of surface waters at <100 m resolution. Methods for incorporating SWOT measurements into river hydrodynamic models have been developed to generate spatially and temporally continuous discharge estimates. However, most of SWOT data assimilation studies have been performed on a local scale. We developed a novel framework for estimating river discharge on a global scale by incorporating SWOT observations into the CaMa-Flood hydrodynamic model. The local ensemble transform Kalman filter with adaptive local patches was used to assimilate SWOT observations. We tested the framework using multi-model runoff forcing and/or inaccurate model parameters represented by corrupted Manning’s coefficient. Assimilation of virtual SWOT observations considerably improved river discharge estimates for continental-scale rivers at high latitudes (>50°) and also downstream river reaches at low latitudes. High assimilation efficiency in downstream river reaches was due to both local state correction and the propagation of corrected hydrodynamic states from upstream river reaches. Accurate global river discharge estimates were obtained (Kling–Gupta efficiency [KGE] > 0.90) in river reaches with > 270 accumulated overpasses per SWOT cycle when no model error was assumed. Introducing model errors decreased this accuracy (KGE ≈ 0.85). Therefore, improved hydrodynamic models are essential for maximizing SWOT information. These synthetic experiments showed where discharge estimates can be improved using SWOT observations. Further advances are needed for data assimilation on global-scale.

Continental-scale river hydrodynamic modeling is useful for understanding the global hydrological cycle, and model evaluation is essential for robust calibration and assessing model performance. Although many models have been robustly evaluated using several variables separately, methods for the integrated multivariable evaluation of models have yet to be established. Here, we propose an evaluation method using the overall basin skill score (OSK), based on considering the spatial distribution of different variables via a sub-basin approach. The OSK approach integrates multiple variables to overcome observation-related limitations, such as the distinct temporal and spatial dimensions and unit of measurement unique to each variable, thus judging model performance objectively at the sub-basin and basin scales. As a case study, the global river model, CaMa-Flood, was evaluated using three variables¾discharge, water surface elevation, and flooded area¾for the Amazon Basin, focusing on the impact of using different types of baseline topography data (SRTM and MERIT digital elevation models [DEMs]). CaMa-Flood with the MERIT DEM performed robustly well over a wide range of river depth parameters with a maximum OSK of 0.51 against 0.46 for the SRTM DEM. Single-variable evaluation for all three variables proved inadequate due to low sensitivity for river bathymetry, with good performance outcomes potentially arising for the wrong reasons. This study confirmed that model evaluation using this method enables a balanced evaluation of different variables and a robust estimation of the best parameter set. The proposed method proved useful for flexible, integrated multivariable model evaluation, with modifications allowed per the user’s requirements.

Large-scale river hydrodynamic model act as fundamental tool for many scientific applications related to water cycle, biogeochemistry, and carbon cycle. Even though process representation in the physically based hydrodynamic models has improved significantly in recent times, due to many error sources the uncertainty reduction and evaluation remains a key issue. Previously most of the research focused on the evaluation of hydrodynamic model considering single variable only i.e., discharge due limitations related to models and data availability. The recent advances in hydrodynamic modelling and remote sensing helped to overcome limitation. Some recent studies performed calibration and validation considering multiple variables but were unable to integrate them into a single evaluation score due to different spatial and temporal dimension of variables and thus make it hard to judge the overall performance. Here, we have evaluated the performance of Catchment-based Macroscale Floodplain (CaMa-Flood) hydrodynamic model over Amazon basin considering multiple variables i.e., discharge (Q), water surface elevation (WSE) and flooded area (FA) for a topography data multi-error removed improved terrain (MERIT) DEM. We proposed an evaluation method and introduced a metric “overall basin skill score” (OSK) to integrate the performances due to multiple (three here) variables considering their spatial distribution via a sub-basin approach and provide the evaluation on a scale of 0 to 1. The integrated method showed the robustness in the method and able to detect the best river channel depth parameter set with maximum OSK of 0.57, whereas the evaluation using single variable proved inadequate due to different sensitivity of variables and maximum metric score were obtained for many parameters sets. The proposed method enables a balanced evaluation of different variables and proved useful to integrated multivariable model evaluation with reducing the chances of getting the right results due to wrong reasons. Preprint related to this work: https://doi.org/10.1002/essoar.10506596.1

River bathymetry is an important parameter for hydrodynamic modeling; however, it is associated with large bias because direct large-scale measurements are impractical. Recent studies adjusted river bathymetry data based on assessment of the difference between modeled and observed water surface elevation (WSE); however, river discharge model uncertainties can lead to unintended errors in the correction. In this study, we propose a simple but robust correction method using the WSE bias, based on stage–discharge rating curves rather than time series data. The rating curve represents only the characteristics of the river section, and is not sensitive to instantaneous errors. Our results showed that the “corrected river bathymetry” converged among discharge experiments driven by noise-corrupted runoff forcing. Given the difficulty of eliminating discharge errors, a method for correcting river bathymetry that is free from discharge errors is important for improving hydrodynamic modeling.

The Amazon River basin harbors some of the world’s largest wetland complexes, which are of major importance for biodiversity, the water cycle and climate, and human activities. Accurate estimates of inundation extent and its variations across spatial and temporal scales are therefore fundamental to understand and manage the basin’s resources. More than fifty inundation estimates have been generated for this region, yet major differences exist among the datasets, and a comprehensive assessment of them is lacking. Here we present an intercomparison of 29 inundation datasets for the Amazon basin derived from remote sensing-based products, hydrological models and multi-source products. Spatial resolutions range from 12.5 m to 25 km, and temporal resolution from static to monthly intervals, covering up to a few decades. Overall, 26% of the lowland Amazon basin is estimated as subject to inundation by at least one product. The long-term maximum inundated area across the entire basin (lowland areas with elevation < 500 m) is estimated at 599,700 ± 81,800 km² if considering only higher quality SAR-based products and 490,300 ± 204,800 km² if considering 18 basin-scale datasets. However, even the highest resolution SAR-based product underestimates the local maximum values, as estimated by subregional products, suggesting a basin-wide underestimation of ~10%. The minimum inundation extent shows greater disagreements among products than the maximum extent: 139,300 ± 127,800 km² for SAR-based products and 112,392 ± 79,300 km² for the overall average. Discrepancies arise from differences among sensors, time periods, dates of acquisition, spatial resolution, and data processing algorithms. The median total area subject to inundation in medium to large river floodplains (drainage area > 1,000 km²) is 323,700 km². The highest spatial agreement is observed for floodplains dominated by open water such as along the lower mainstem rivers, whereas intermediate agreement is found along major vegetated floodplains fringing larger rivers (e.g., Amazon mainstem floodplain). Especially large disagreements exist among estimates for interfluvial wetlands (Llanos de Moxos, Pacaya-Samiria, Negro, Roraima), where inundation tends to be shallower and more variable in time. Our data inter-comparison helps identify the current major knowledge gaps regarding inundation mapping in the Amazon and their implications for multiple applications. In the context of forthcoming hydrology-oriented satellite missions, we make recommendations for future developments of inundation estimates in the Amazon and present a WebGIS application (https://amazon-inundation.herokuapp.com/) we developed to provide user-friendly visualization and data acquisition of current Amazon inundation datasets.