Abstract
Thirteen satellite precipitation products (SPPs), re-gridded to 1 km resolution, were evaluated in terms of the structural similarity index (SSI) over the Pra catchment in Ghana. Three SPP scenarios were considered: Scenario one (S1) was the original SPPs; Scenario two (S2) was bias-corrected SPPs; and Scenario three (S3) was the better of S1 and S2 for each wet day. For each scenario, the best SPP was selected to constitute the 14th SPP referred to as the BEST SPP. Each SPP was evaluated in terms of SSI against the rain gauge rainfield for each wet day. For S1, the top three SPPs were TMPA, GSMAP and CMORPH; for S2, CMORPH, PERCCS and MSWEP were the top three; and for S3, CMORPH, PERCCS and TMPA came out on top in order of decreasing performance. Bias correction led to improvement in the overall SSI measure (SSIM) for 73% of wet days. The BEST SPP increased the SSIM of the best individual SPP by over 50% for S1, and over 30% for both S2 and S3. Comparing the BEST SPP of the three scenarios, S2 increased the SSIM statistic by 20% over that for S1, and SSIM was further improved by 4% for S3. It is highly recommended to use BEST SPP (S3) to generate the required 1 km × 1 km rainfields for the Pra, or other catchments around the world with a sparse rain gauge network, through conditional merging with rain gauge data as demonstrated.