The article introduces novel ensemble data assimilation (DA) techniques for streamflow forecasting using WRF-Hydro and DART. Model-related biases are tackled through spatially and temporally varying adaptive prior and posterior inflation. Spurious and physically incorrect correlations, on the other hand, are mitigated using a topologically based along-the-stream localization. Hurricane Florence (2018) in the Carolinas, USA, is used as a test case to investigate the performance of DA techniques.

Several rivers basins in the Ganga plains suffer from very high sediment production in their catchment and there are no good estimates of sediment yield from these basins due to a lack of gauge data. The RUSLE model offers an alternative approach and the same has been applied in a small basin in the Ganga plains. The study demonstrated the usefulness of the proposed methodology for quantifying uncertainty in soil erosion and sediment yield estimates at ungauged basins.

Long-term catchment sediment and nutrient budgets are important for managing soil and nutrient resources for more sustainability. Here we construct a 14-year budget of water, sediment and nutrients across two subtropical reservoirs. A major flood in January 2011 dominated flow and loads in and out of both reservoirs. Sediment and nutrient budgets are inherently dynamic, and our results demonstrate that meaningful reservoir budgets require reliable estimates of uncertainty and variability.

This paper further develops the method of quantile regression (QR) to generate probabilistic river stage forecasts. Besides the forecast itself, this study uses the rate of rise of the river stage in the last 24 and 48h and the forecast error 24 and 48h before as predictors in QR configurations. When compared to just using the forecast as an independent variable, adding the latter four predictors significantly improved the forecasts, as measured by the Brier skill score and the CRPS.