Assessment and comparison of five satellite precipitation products in Australia

Abstract

Quality satellite-derived precipitation products (SPPs) are needed for water resources inventories and management, particularly in poorly gauged regions around the world. The latest version of five SPPs were assessed against SILO (Scientific Information for Land Owners) gauge-based gridded precipitation dataset in Australia over a 5-year period from October 2014 to September 2019. The evaluation was carried out using a 0.50° grid at daily, seasonal, and annual temporal scales. The assessed SPPs were the Integrated Multi-satellitE Retrievals for GPM (Global Precipitation Measurement) (IMERG), TRMM (Tropical Rainfall Measuring Mission) Multi-satellite Precipitation Analysis (TMPA), Climate Prediction Centre (CPC) MORPHing technique (CMORPH), Precipitation Estimation from Remotely Sensed Information using Artificial Neural Networks (PERSIANN), and PERSIANN-CDR (Climate Data Record). PERSIANN does not include any ground-based observations for bias-correction, while the other four products are bias-corrected against gauge-based data. Bias ratio and correlation coefficient for the five SPPs showed that the overall performance of IMERG and TMPA was better than that of CMORPH, PERSIANN, and PERSIANN-CDR for Australia. Seasonal analysis showed that IMERG had the better skill in winter. Overall, IMERG appeared to be the best SPP for Australia. However, TMPA also performed reasonably well, considering the climatological calibration implemented recently in the precipitation processing algorithm. The Structural Similarity Index (SSI), a map comparison technique using a moving window-based approach, was used to compare similarities between a pair of gridded precipitation maps in terms of local mean, variance, and covariance. All the SPPs showed discrepancies in the spatial structure of the mean annual precipitation, predominantly over some high precipitation areas in Australia. These local scale differences were not detectable in conventional cell by cell comparison or simply by visual inspection. Therefore, SSI could be an effective method to evaluate satellite precipitation estimation algorithm.