17:15 〜 19:15
[AHW22-P14] How to Model Mountain Rivers: A Multi-Objective Validation Approach in the Central Caucasus
キーワード：mountain hydrology, runoff modeling, glacier modelling, Baksan River, ECOMAG, A-Melt
This study investigates the possibilities and limitations of a multi-objective validation approach for the distributed hydrological model ECOMAG in the high-altitude Baksan River Basin, Central Caucasus. Runoff generation in the region is driven by snow and glacier melt, making accurate modelling crucial for water resource management. Traditional validation approaches that rely solely on observed runoff are often inadequate in mountainous regions due to limited monitoring stations and complex hydrological processes. To improve model reliability, we applied a multi-objective validation strategy using different datasets. The ECOMAG model was validated using observed discharge data, MODIS-derived snow cover maps, stable isotope hydrograph separation, glacier mass balance observations and glacier runoff simulations from the A-Melt energy-balance model.
The results showed high agreement between simulated and observed runoff, with the Nash-Sutcliffe efficiency (NSE) exceeding 0.8 for both the calibration (2000-2008) and validation (2009-2017) periods.
The model successfully captured seasonal snow cover variations, achieving an R² of 0.85 when compared with the MODIS data. However, the model tended to overestimate snow cover in the spring period, possibly due to the lack of meteorological stations in the basin.
Isotopic hydrograph separation further confirmed the model’s reliability in simulating the contributions of meltwater and rainfall to runoff. The proportion of snowmelt-derived runoff naturally decreased downstream, while rainfall contributions increased. The relative error of meltwater and rainfall separation was 19%, with R² values of 0.81 and 0.86, respectively.
Glacier ablation simulations showed mixed results. For the Garabashi Glacier, model performance was satisfactory, with a relative error of –12%. However, for the Djankuat Glacier, the model significantly underestimated ablation (–55%), likely due to the complex topography and the influence of avalanche-fed ice, which is not represented in ECOMAG.
A comparative analysis of glacial runoff estimates from the ECOMAG and A-Melt models revealed a good overall agreement (pBIAS = 18%). However, differences in the estimated snow and ice contributions highlighted challenges in parameterizing ice melt within hydrological models. ECOMAG simulations tended to overestimate the snowmelt component while underestimating ice melt compared to A-Melt, which can be attributed to differences in surface classification methods and melt coefficient calibration.
These findings confirm that a multi-objective validation approach significantly improves the accuracy of hydrological models in high-altitude basins by integrating diverse datasets. The approach enhances confidence in model predictions, making them more reliable for long-term water resource management, climate impact assessments, and adaptation strategies in glacier-fed river systems.
Future research should focus on refining glacial process representation by integrating high-resolution meteorological inputs, improving snow redistribution modeling, and coupling hydrological models with energy-balance glacial models. These advances will help improve predictions of future runoff changes in response to climate variability, ensuring more sustainable water management in mountain regions.
This study was supported by the grant of the The Government of the Russian Federation (Agreement №075-15-2024-614 date 13.06.2024) (multi-objective validation results analysis), was carried out under the Governmental Order to Water Problems Institute, Russian Academy of Sciences, subject no. FMWZ-2025-0003 (ECOMAG model adaptation and calibration), and under the Governmental Order to Institute of Geography, Russian Academy of Sciences, subject no. FMWZ-2022-0001 (A-Melt modelling).
The results showed high agreement between simulated and observed runoff, with the Nash-Sutcliffe efficiency (NSE) exceeding 0.8 for both the calibration (2000-2008) and validation (2009-2017) periods.
The model successfully captured seasonal snow cover variations, achieving an R² of 0.85 when compared with the MODIS data. However, the model tended to overestimate snow cover in the spring period, possibly due to the lack of meteorological stations in the basin.
Isotopic hydrograph separation further confirmed the model’s reliability in simulating the contributions of meltwater and rainfall to runoff. The proportion of snowmelt-derived runoff naturally decreased downstream, while rainfall contributions increased. The relative error of meltwater and rainfall separation was 19%, with R² values of 0.81 and 0.86, respectively.
Glacier ablation simulations showed mixed results. For the Garabashi Glacier, model performance was satisfactory, with a relative error of –12%. However, for the Djankuat Glacier, the model significantly underestimated ablation (–55%), likely due to the complex topography and the influence of avalanche-fed ice, which is not represented in ECOMAG.
A comparative analysis of glacial runoff estimates from the ECOMAG and A-Melt models revealed a good overall agreement (pBIAS = 18%). However, differences in the estimated snow and ice contributions highlighted challenges in parameterizing ice melt within hydrological models. ECOMAG simulations tended to overestimate the snowmelt component while underestimating ice melt compared to A-Melt, which can be attributed to differences in surface classification methods and melt coefficient calibration.
These findings confirm that a multi-objective validation approach significantly improves the accuracy of hydrological models in high-altitude basins by integrating diverse datasets. The approach enhances confidence in model predictions, making them more reliable for long-term water resource management, climate impact assessments, and adaptation strategies in glacier-fed river systems.
Future research should focus on refining glacial process representation by integrating high-resolution meteorological inputs, improving snow redistribution modeling, and coupling hydrological models with energy-balance glacial models. These advances will help improve predictions of future runoff changes in response to climate variability, ensuring more sustainable water management in mountain regions.
This study was supported by the grant of the The Government of the Russian Federation (Agreement №075-15-2024-614 date 13.06.2024) (multi-objective validation results analysis), was carried out under the Governmental Order to Water Problems Institute, Russian Academy of Sciences, subject no. FMWZ-2025-0003 (ECOMAG model adaptation and calibration), and under the Governmental Order to Institute of Geography, Russian Academy of Sciences, subject no. FMWZ-2022-0001 (A-Melt modelling).