An Atmospheric–Hydrologic Forecasting Scheme for the Indus River Basin

Abstract

Abstract South Asia is vulnerable to extreme weather events, and the Indus River basin (IRB) is subject to flood risk. Flooding occurred in Pakistan during the monsoon seasons from 2010 to 2012, causing major economic damage and significant loss of life. The nature of slow-rise discharge on the Indus and overtopping of riverbanks in the IRB indicates the need for extended warning time (1–10 days). Daily mean streamflow at Tarbela Dam, a major control structure on the Indus River, is reproduced by using numerical weather prediction initialization data from the ECMWF to drive the Variable Infiltration Capacity macroscale hydrologic model. A one-dimensional routing model conducts base flow and surface runoff from each grid cell through a stream network. Comparisons of reconstructions with inflow data at Tarbela Dam over a 3-yr period yield an R2 correlation of 0.93 and RMSE of 692 m3 s−1. From 2010 to 2012, 276 daily ensemble hindcasts are generated. In comparing the ensemble mean of 10-day hindcasts to reconstructed inflow, the R2 correlation was >0.85 for 2010, >0.70 for 2011, and >0.70 for 2012. The RMSE was <24% of mean streamflow for 2010, <28% for 2011, and <34% for 2012. A method to translate this type of probabilistic streamflow forecast data into communicable information is described. A two-dimensional model is described to simulate movement of overflow water into the floodplain.