Development of a flood forecasting system on the upper Indus catchment using IFAS

Flow forecast in the Upper Indus catchment in Pakistan is based on average peak flow travel time between key dams and barrages. There was also no flow forecasting system for the Kabul river sub-basin where most of the 2010 floods victims were reported. A 5-km spatially distributed tank model using Integrated Flood Analysis System (IFAS) was developed from the Indus upstream reach until the Taunsa barrage. A preliminary model parameterisation, which relied on global data, was updated with newly surveyed soil hydraulic data. The model calibration was performed with Pakistan Meteorological Department (PMD) rain gauge data. Nash-Sutcliffe efficiencies E-NS were calculated for simulated discharges at key gauges. E-NS were very low or negative for past flood simulations. One of the reasons for these low efficiencies was the scarcity of local hydro-meteorological data, including rainfall. Therefore, corrected JAXA satellite-based rainfall estimates GSMaP-NRT were considered input data. GSMaP-NRT self-correction method coefficients were calibrated for Upper Indus, but the results only improved slightly. However, upstream discharges as boundary conditions resulted in E-NS reaching satisfactory averages over 0.7. Simulated hydrographs were then acceptable in terms of peak timing and/or height. Therefore, the conclusion was to recommend relying on upstream discharges as boundary conditions for operational use of the model.