In our latest Performance Report, we analyze BWI’s discharge forecast model’s performance for the Le Salat – Saint Lizier catchment.
As climate change amplifies hydrological extremes, such as droughts and floods, regular monitoring of water resources is becoming increasingly important for optimal water resource management.
BWI employs a semi-distributed hydrological model. With the integration of machine learning techniques, it monitors and forecasts river discharges. This blog post delves into a detailed performance evaluation of BWI’s discharge prediction model. We specifically focus on the ‘Le Salat’ river, with its outlet located at Saint-Lizier, in the Occitanie region of France.
Resting in the awe-inspiring landscapes of the Couserans region in South-Western region of France, the Le Salat River weaves flows through the departments of Ariège and Haute-Garonne, providing the Occitanie region with its pristine waters. As an essential tributary of the Garonne river, this river hides under the stunning beauty of the Pyrenees.
Originating from multiple water sources, Le Salat’s journey begins at the region of Salau where they call it affectionately the Ruisseau de Bégé. The river has a 74.1-kilometer length, winding its way through several valleys. It then passes through the quaint region of Saint-Girons, where it is marked by two notable tributaries, namely the Lez and the Baup. It then flows back to Saint-Lizier where the outlet for this report is located.
BWI utilizes a robust model performance evaluation framework that considers a range of metrics. Some of these include Nash-Sutcliffe model efficiency scores (NSE), normalized NSE (NNSE), Kling-Gupta Efficiency (KGE), and their respective modified and normalized versions. The preferred metric within this framework is NNSE, which ranges between 0 and 1. A higher NNSE score indicates superior model performance. This metric is particularly useful for tracking the model’s performance over time and effectively communicating the results.
At the Saint-Lizier outlet, the model achieves an NNSE of 0.82. This is a score that scientific publications classify as indicative of good model performance (NNSE > 0.65). The optimal parameter sets and the final state of the basin at the end of the training are recorded and utilized to initialize the subsequent training batch.
