This performance report assesses the effectiveness of BWI’s discharge forecast model for the Tarn River, specifically at the Countal outlet with a catchment area of 15969 km2, located in southern France. The Tarn River basin features dramatic elevation changes, descending from mountainous regions to the plains, and it passes through varied terrain, including limestone plateaus, deep gorges, and agricultural lowlands. Mediterranean and oceanic influences characterize the climate in the area. The summer droughts and Cévennes rains – with monsoon-like intense rainfall, result in very high peak flows. BWI’s model demonstrates strong performance on the validation dataset (NNSE = 0.72), with very well-captured baseflow and peak-flow occurrences.
The Mediterranean region is susceptible to anthropogenic and climatic changes, mostly affecting its water resources and related practices. This sensitivity results from its complex cyclonic system sweeping over a large evaporative basin (Allam et al., 2020).
A major influence on this sensitivity is the phenomenon of “Cévennes episodes” (French: épisodes cévenols) —intense, heavy rainfall events that primarily affect parts of southern France, especially the Cévennes region. These episodes, often referred to as Mediterranean episodes, are notorious for triggering severe floods. The interplay between concentrated precipitation and high evaporation rates contributes to unique and variable hydrological conditions. Additionally, steep slopes and impermeable soils lead to rapid runoff, exacerbating flood risks.
The hydrological patterns in the Tarn River reflect a combination of both Mediterranean and Atlantic (oceanic) influences. This results in variable flow regimes, with the potential for sudden increases in water levels during heavy rainfall events. This performance report examines the effectiveness of BWI’s semi-distributed hydrological model, which leverages machine learning techniques to deliver near real-time estimates and forecasts of river discharge. Specifically, the report focuses on the Gard River at La Countal, providing insights into the model performance and forecasting capacity in managing river flow data.
In hydrological modeling, the components of performance metrics like the Nash-Sutcliffe Efficiency (NSE) and the Kling-Gupta Efficiency (KGE) provide crucial insights into various aspects of discharge prediction. BWI considers both NSE and KGE for model performance evaluation and deploys the Normalized NSE (NNSE) to communicate the model performance evaluation with stakeholders.
NSE primarily assesses the overall agreement between observed and simulated discharge. It does not have specific components like KGE, but it indirectly provides information on the peak flow accuracy, overall variability, and bias (indirectly) of the discharges produced by the hydrological model at the catchment outlet. It ranges from −∞ to 1 and evaluates how well the simulated data matches the observed data. NNSE adjusts the NSE value to make it more comparable across different catchments or periods by standardizing to provide indices expressed between 0 and 1.
NSE is particularly sensitive to peak flows. It emphasizes how well the model captures high discharge events, such as floods. A high NSE value indicates that the model is accurately predicting peak discharges, reflecting the catchment’s response to extreme rainfall.
By comparing the squared differences between observed and simulated discharges to the variability in observed discharges, NSE indicates how well the model captures the overall variability of flows in the catchment.
Although NSE does not explicitly separate bias, a low NSE score could indicate that the model is consistently underestimating or overestimating the flow. It suggests that the simulated and observed discharges differ significantly, often due to bias in the model.
In a hydrological model, NNSE values below 0.5 indicate poor performance. Values above 0.67 suggest that baseflow correlates well with observations and that the timing of flood events is generally well-identified. Models with an NNSE of 0.67 or higher are deployed for 10-day discharge forecasts. Recognizing that NNSE is not an absolute performance metric, if other performance metrics (e.g., KGE) show good agreement, we accept NNSE starting from 0.65. When NNSE exceeds 0.85, the model demonstrates good agreement in the timing and magnitude of peak flows during extreme flood events. An NNSE value of 1 represents a perfect model.
The figure below illustrates the performance of “Le Tarn – La Countal” model. On the left, the map displays the catchment area for the outlet, while on the right, a graphical comparison shows observed discharges (in gold) alongside simulated discharges (in green).
The model demonstrates strong performance at the selected La Countal outlet, achieving an NNSE of 0.72 during the testing period. Base flows up to 600 m³/s are generally well estimated. However, an extreme flood event occurred in March 2024, reaching flows of up to 1200 m³/s. Although the model successfully forecasted the event’s occurrence, it underestimated the peak flow, as shown in the performance report. This discrepancy may stem from both limited weather data resolution and a relatively short model training period. Extending the training period would increase the model’s exposure to extreme events, providing additional learning opportunities to enhance its ability to forecast such events’ magnitude and timing.
The estimated hydrograph also shows a faster recession of peak flows than observed in reality—another area for improvement in future model iterations. As more data and data sources become available, both accuracy in extreme flow forecasting and hydrograph recession rates should improve.
Overall, while there are areas for refinement, the model’s NNSE of 0.72 and accurate base flow capture offer valuable contributions to water resource management, early warning systems, and the planning of critical flood prevention infrastructures.
Allam, A., Moussa, R., Najem, W., and Bocquillon, C. (2020) ”Specific climate classification for Mediterranean hydrology and future evolution under Med-CORDEX regional climate model scenarios”, Hydrol. Earth Syst. Sci., 24 (9), pp. 4503–4521. Available at: https://hess.copernicus.org/articles/24/4503/2020/ (Accessed 12 November 2024).
Bassin Versant Tarn Aval (2024) ”Géographie”. Available at: https://www.tarn-aval.com/pages/le-territoire/bassin-versant-tarn-aval.html (Accessed 12 November 2024).
