Relative errors in TRMM satellite Version 5 and Version 6 products: Steps forward and backward

Error estimates for satellite precipitation retrievals are needed globally for NASA/JAXA's current Tropical Rainfall Measuring Mission (TRMM) and the planned Global Precipitation Measurement (GPM) mission, but ground validation (GV) sites are sparse, particularly over the oceans. Information on uncertainties associated with satellite-derived precipitation products aids in the diagnosis and refinement of physical assumptions within algorithms, and adds value for applications such as forecasting, data assimilation, and climate diagnostics. Over regions without GV sites, relative errors between independent estimates derived from satellite passive microwave and precipitation radar are used to assess and diagnose errors in satellite precipitation retrievals. Local observations at GV sites can help guide the physical interpretation of the error characteristics.

TRMM Version 5 (V5) and Version 6 (V6) orbit product data sets are used to calculate statistics for probabilistic comparison among PR and TMI precipitation retrievals at sensor native scales and PR data rescaled to spatial scales similar to TMI 10 GHz and 19 GHz pixels. Comparisons are performed for 47 day increments, corresponding to the sampling of a full diurnal cycle by the TRMM satellite. Additionally, a composite analysis following Bony et al. (2004) is used to decompose the global oceanic frequency distribution of rainfall by meteorological regime.

For land areas, there is a marked improvement in the shape of the rainfall frequency distribution between TMI V5 and V6. Version 5 TMI Level 2 products had a physically unrealistic discontinuous rainfall distribution over land areas, i.e. some ranges of rain rates never occurred in the TMI products. These errors were not obvious in the area and time-integrated Level 3 products. Version 6 TMI has improved the rainfall frequency distribution over land areas so that it is continuous.

Over ocean areas, there is an apparent degradation between TMI V5 and V6. Composite analysis indicates a marked decrease in TMI rainfall associated with oceanic regions with the highest SST (> 28 deg C) and strongest upward motion in V6 as compared to V5. Although there is better agreement between monthly mean values of rainfall averaged over 40 deg N to 40 deg S latitude between TMI and PR in V6 as compared to V5, changes to the TMI algorithm appear to have had the unintended side effect of degrading the relationship between rainfall, SST, and omega.