The Relationship between TRMM Climate Rainfall Biases and the Synoptic Environment

The latest (version 6) Tropical Rainfall Measuring Mission (TRMM) retrieval algorithms show very good agreement in zonal mean rainfall between estimates from the precipitation radar (PR) and the TRMM microwave imager (TMI). Unfortunately, time-dependent regional differences still exist, which can have a significant impact for many climate applications. For example, estimates from TMI show an increase in tropical mean rainfall associated with the 1997/98 El Niño, while corresponding estimates from the PR do not. Due to the underconstrained nature of the rainfall retrieval problem all satellite retrieval algorithms must make assumptions regarding the cloud structure and microphysical properties. As a result, systematic changes in these assumed parameters between regions and/or over time result in regional and/or temporal biases in the rainfall estimates. These biases tend to be relatively small compared to random errors in the retrieval, however, when random errors are reduced through spatial and temporal averaging for climate applications, they become the dominant source of error.

Because such systematic biases are associated with distinct climate regimes, it follows that the associated algorithm errors must be related to changes in the synoptic environment. As shown in a previous study, the characteristics of rain systems in the central and east Pacific differ significantly from those over the west Pacific warm pool region. The environment between these regions differs significantly with warmer SSTs and more water vapor in the west and enhanced surface convergence in the central and east Pacific, although this changed dramatically during the 1997/98 El Niño. To investigate how the synoptic environment relates to systematic differences between the rainfall estimates, the PR (2A25) and TMI (2A12) rain estimates were matched in space and time at the TMI field-of-view for the periods from December through February of 1997, 98, and 99. The rain estimates were then clustered based on cloud structure and synoptic variables observed from TRMM to identify those most closely related to TMI/PR differences. Although comparing algorithm differences with satellite observed synoptic variables does not directly identify the assumptions leading to systematic differences in the retrievals, it provides a physical basis for understanding the behavior of the retrieval algorithms. It also provides a way to investigate algorithm differences using ground-based measurements, which is independent of location and time.