Benchmarking Real-Time Streamflow Forecast Skill in the Himalayan Region

Abstract

<jats:p>Improving decision-making in various areas of water policy and management (e.g., flood and drought preparedness, reservoir operation and hydropower generation) requires skillful streamflow forecasts. Despite the recent advances in hydrometeorological prediction, real-time streamflow forecasting over the Himalayas remains a critical issue and challenge, especially with complex basin physiography, shifting weather patterns and sparse and biased in-situ hydrometeorological monitoring data. In this study, we demonstrate the utility of low-complexity data-driven persistence-based approaches for skillful streamflow forecasting in the Himalayan country Nepal. The selected approaches are: (1) simple persistence, (2) streamflow climatology and (3) anomaly persistence. We generated the streamflow forecasts for 65 stream gauge stations across Nepal for short-to-medium range forecast lead times (1 to 12 days). The selected gauge stations were monitored by the Department of Hydrology and Meteorology (DHM) Nepal, and they represent a wide range of basin size, from ~17 to ~54,100 km2. We find that the performance of persistence-based forecasting approaches depends highly upon the lead time, flow threshold, basin size and flow regime. Overall, the persistence-based forecast results demonstrate higher forecast skill in snow-fed rivers over intermittent ones, moderate flows over extreme ones and larger basins over smaller ones. The streamflow forecast skill obtained in this study can serve as a benchmark (reference) for the evaluation of many operational forecasting systems over the Himalayas.</jats:p>

Understanding the impact of vegetation dynamics on the water cycle in the Noah-MP model

“…Compared to traditional fit metrics, such as RMSE or Nash-Sutcliffe Efficiency (NSE; Nash and Sutcliffe, 1970), the KGE provides more insight into the model skill and the ability to evaluate different components of overall error (Gupta et al, 2009;Fowler et al, 2018;Ghimire et al, 2020). A meaningful benchmark for the KGE is one in which the observed mean is used as a predictor and yields a KGE score of 1 -√ 2 ≈ −0.41 (Knoben et al, 2019).…”

Integrated modeling for assessing climate change impacts on water resources and hydropower potential in the Himalayas

“…We obtain these datasets from the Department of Hydrology and Meteorology (DHM), Government of Nepal. A 24-h precipitation accumulation is observed using a standard (8″ diameter) manual rain gauges every day at 03 UTC, i.e., 8:45 AM local time (Talchabhadel et al 2017), temperature observations are taken with the mercury (alcohol) containing thermometers for maximum and minimum temperatures at 1.25-2 m above the surface (Karki et al 2020), and daily streamflow data corresponds to the streamflow based on the stage-discharge relation and the mean values of three water level readings recorded at 8 AM, 12 noon and 4 PM, local time (Ghimire et al 2020).…”

Forecasting Convective Storms Trajectory and Intensity by Neural Networks

“…Besides testing the neural network on the performance metrics presented above, the FFNN is compared with "persistence" as a benchmark [52][53][54][55][56][57][58][59]. The persistent model assumes the phenomenon maintains its characteristics, which means the velocity vector, the average and maximum reflectivity, and the area at time t remain constant for the following steps until t + 60.…”