Evaluation of PERSIANN-CCS Satellite Derived Rainfall Product with Raingauge Data over Kelani River Basin, Sri Lanka

Abstract

Satellite rainfall estimates (SREs) are high in spatial and temporal resolution and particularly important for regions with sparse raingauges. However, SREs are required to evaluate with gauged rainfall data before applying for hydrological studies. In this research, the accuracy of Precipitation Estimation from Remotely Sensed Information using Artificial Neural Networks Cloud Classification System (PERSIANN-CCS) product was evaluated at daily, monthly, yearly, and seasonal scale upon the raingauge data of the Kelani River basin of Sri Lanka for the period 2004 to 2010. The performance of the SREs was evaluated using both continuous and categorical verification statistics. PERSIANN-CCS rainfall estimates follow the bi-modal rainfall pattern and showed greater underestimation in South West Monsoon (SWM) season (May-Sep.) and overestimation in InterMonsoon 1 (IM1) period (March-April). PERSIANN-CCS is more capable of recognizing conventional and depressional rains than monsoonal rains. On the other hand, it produces low false alarms in the high rainy season than in the low rainy season. The daily categorical statistics show above average scores (Accuracy>0.69; POD>0.65; FAR<0.34; 0.76>FBias<1.11), however, estimations were with low CC (<0.53) and high bias (<24 & >-64%). Bias corrected PERSIANN-CCS may be a high-resolution rainfall source for flood forecasting applications in the Kelani River basin.