Evaluation of TRMM rain-rate estimates using ground measurements at different time scales

The Tropical Rainfall Measuring Mission (TRMM) is a satellite-based program to estimate global tropical rainfall. Since the launch of the TRMM satellite in Nov 1997, a number of rainfall products have been produced from the TRMM-onboard Precipitation Radar (PR) and Microwave Imager (TMI) sensors. Rain products derived from the PR and TMI suffer substantial sampling, retrieval, and systematic errors. To improve TRMM rainfall algorithms and help data users be aware of the limitation of the TRMM products, it is necessary to evaluate TRMM products against other types of rainfall observations such as more conventional ground-based radar (GR) and tipping-bucket rain gauge (TG) measurements.

In this study, 12 years (1998-2009) of GR and TG products are used to evaluate the coincident TRMM products derived from TMI, PR and COM (a combined product using rain information from both TMI and PR) over the TRMM ground Validation (GV) site at Melbourne, Florida. These products are spatiotemporally matched and inter-compared at 0.25-deg grid resolution and various time scales ranging from 5 minutes to 30 days. On monthly and yearly scales, the TG shows excellent agreement with the GR since the GR rain rates are generated using the TG data as a constraint on a monthly basis. However, large disagreements exist between the GR and TG at shorter time scales due to their different spatial and temporal samplings. The yearly biases relative to the GR for the PR and TMI are mainly negative with a few exceptions. The COM bias fluctuates from year to year over the 12-yr period. The PR, TMI and COM are in good overall agreement with the GR in the lower range of rain rates, but the agreement is notably worse at higher rain rates. The diurnal cycle of rainfall over Melbourne is well captured by all products, but the peak satellite-derived rainfall (PR, TMI and COM) lags the peak from the ground measurements (GR and TG) by ~1 h.

It is important to note that the evaluation of TRMM rain estimates in this study is carried out over the land area of central Florida. The results drawn from this limited land area might not be applicable to oceanic areas and other land sites as the rain error statistics can be highly regime-dependent. In addition, the GV measurements themselves contain some biases although they are used as a reference in the TRMM satellite rain-rate evaluation in this study. Nevertheless, the GV measurements provide an independent ground-based reference for the comparison with TRMM satellite rain estimates, which can aid in the improvement of space-based rain retrieval algorithms for the TRMM and anticipated Global Precipitation Measurement (GPM) satellites.