Sampling Simulation of TRMM Rainfall Estimation Using Radar AMeDAS Composites

Abstract

Sampling errors of rainfall retrieved by a low-inclination, low-altitude satellite, such as the Tropical Rainfall Measuring Mission (TRMM) satellite, are estimated by using radar?AMeDAS (Automatic Meteorological Data Acquisition System) rainfall data around Japan. The data are composite radar-derived rainfall maps calibrated with a dense surface rain gauge network (AMeDAS). In this study, sampling errors of monthly rainfall have been estimated over 5° ? 5° and 2.5° ? 2.5° (latitude ? longitude) domains using a 43-month time series of radar?AMeDAS data. For a 5° ? 5° area the sampling error around Japan would be approximately 16% with the swath width of the TRMM microwave imager (TMI) and approximately 20% with the swath width of the TRMM precipitation radar. Using the swath width of TMI, the sampling error for a 2.5° ? 2.5° area was 24%. These errors are relatively larger than found in previous studies using Global Atmospheric Research Program (GARP) Tropical Atlantic Experiment (GATE) radar data, because this study is based on the long-term actual data around Japan, where 1) rain related to midlatitude cyclones and tropical convective rainfall are included, and 2) rainfall over the land as well as the ocean is included. A probable source of additional error for monthly averages is due to the fact that Japan is at the extreme northern edge of the TRMM orbit where it takes six weeks to cover the diurnal cycle. A strong seasonal dependency of sampling errors was found. For the 5° ? 5° area, errors in the summer rainy season (June?September) were 13% and 19% in the dry season (November?March) with the swath width of TMI. In terms of the autocorrelation coefficient of area-averaged rainfall, this difference of sampling errors between summer and winter is reasonable because the summer e-folding time is longer than that for winter. The sampling errors could be reduced considerably, especially at higher latitudes, by two TRMM satellites flying one-half orbit apart, providing a greater number of independent samples and a complete sampling of the diurnal cycle within one month.