Uncertainty and Bias in Satellite-Based Precipitation Estimates over Indian Subcontinental Basins: Implications for Real-Time Streamflow Simulation and Flood Prediction*

摘要

Real-time streamflow monitoring is essential over the Indian subcontinental river basins, as a large population is affected by floods. Moreover, streamflow monitoring helps in managing water resources in the agriculture-dominated region. In this study, the authors systematically investigated the bias and uncertainty in satellite-based precipitation products [Climate Prediction Center morphing technique (CMORPH); Precipitation Estimation from Remotely Sensed Information Using Artificial Neural Networks (PERSIANN); PERSIANN Climate Data Record (PERSIANN-CDR); and Tropical Rainfall Measuring Mission (TRMM), version 7, real-time (3B42RTV7) and gauge-adjusted (3B42V7) products] over the Indian subcontinental river basins for the period of 2000-13. Moreover, the authors evaluated the influence of bias in the satellite precipitation on real-time streamflow monitoring and flood assessment over the Mahanadi river basin. Results showed that CMORPH and PERSIANN underestimated daily mean precipitation over the majority of the subcontinental river basins. On the other hand, TRMM-3B42RTV7 overestimated daily mean precipitation over most of the river basins in the subcontinent. While gauge-adjusted products of PERSIANN (PERSIANN-CDR) and TRMM (TRMM-3B42V7) performed better than their real-time products, large biases remain in their performance to capture extreme precipitation (both frequency and magnitudes) over the subcontinental basins. Among the real-time precipitation products, TRMM-3B42RTV7 performed better than CMORPH and PERSIANN over the majority of the Indian subcontinental basins. Daily streamflow simulations using the Variable Infiltration Capacity model (VIC) for the Mahanadi river basin showed a better performance by the TRMM-3B42RTV7 product than the other real-time datasets. Moreover, daily streamflow simulations over the Mahanadi river basin showed that bias in real-time precipitation products affects the initial condition and precipitation forcing, which in turn affects flood peak timing and magnitudes.