Models or users - who is guilty? A large-scale case study of sediment transport modelling - validation datasets

• DOI: 10.5194/egusphere-egu25-21288
• Publication Date: 2025-03-15T06:29:37Z
• Authors: Miroslav Bauer; Tomáš Dostál; Josef Krása; Barbora Jáchymová; Jan Devátý; Martina Mazancová

The environment of the Czech Republic has long been significantly affected by anthropogenic influences that negatively impact the condition of landscape ecosystems. One such influence is accelerated soil erosion and excessive sediment loading of watercourses, primarily from agricultural landscapes. This process reduces soil fertility, degrades water quality, and increases the transfer of contaminants into aquatic systems.This study analyzes sediment transport in the Elbe basin using monitoring data from the Labe (Elbe) and Vltava (Moldau) river basin authorities. The investigated area covers approximately 49,000 km² and includes 600 watercourse sampling profiles with monthly suspended solids measurements. The focus is on comparing measured and modeled sediment transport. WaTEM/SEDEM (based on USLE/RUSLE methods and sediment transport capacity assessment) was chosen as the modeling tool. The main objectives were to (i) assess long-term and episodic sediment loading, (ii) identify factors affecting the agreement between measured and modeled values, and (iii) evaluate the potential for model calibration and validation.Calculations indicate that 1.5 million tons of erosive sediment enter the streams of the study area annually. Of this amount, 59% is captured in reservoirs, corresponding to 627,000 tons deposited each year. The analysis showed a strong correlation (R² = 0.94) between modeled and measured data for the entire dataset. However, after excluding 21 high-transport profiles (above 20,000 t/year), the coefficient of determination dropped to 0.50, revealing that outliers significantly affect the model’s match. This study provides a detailed comparison of modeled and measured sediment, including the influence of catchment characteristics and major rainfall episodes on model suitability.The paper further discusses the limitations of the data sources used for both calibration and validation, with a strong emphasis on constraints arising from user-level mistakes—both minor and major—in data preparation and model computation. It is crucial for all users (and scientists) to acknowledge these limitations, address them openly, and strive to reduce errors in future modeling, validation, and publications.Research has been supported by project TUDI (European Union's Horizon 2020 research and innovation program under grant agreement No. 101000224), QL24020309 (The Ministry of Agriculture of the Czech Republic) and TAČR SS03010332 (Technology Agency of the Czech Republic).