The Role of Calibration in Large-Scale Averages Computed from Multi-satellite Precipitation Datasets

One key step in computing (quasi-)global multi-satellite precipitation datasets is to decide how the various input precipitation datasets will be calibrated. This step is generally considered necessary to avoid unphysical changes in average values in the combined dataset as the various input datasets start, stop, and (potentially) change characteristics in response to changes in sensor availability and quality. What is not generally appreciated is that the resulting large-scale (space and time) averages display behavior that is a combination of the calibration source, the kind of calibration approach, and the sensor(s) being calibrated.

This issue is gaining some prominence as both climate-oriented precipitation datasets and High-Resolution Precipitation Products achieve sufficiently long records that comparisons start to reveal differing behaviors. In this talk the issues will be illustrated by comparing the Version 2 .2 Global Precipitation Climatology Project Satellite-Gauge (GPCP SG), the Version 7 Tropical Rainfall Measuring Mission (TRMM) Multi-satellite Precipitation Analysis (TMPA), and the Version 7 Real-Time TMPA (TMPA-RT), together with selected single-sensor datasets, such as the Version 7 TRMM Combined-Instrument (TCI) and the standard Version 7 TRMM Microwave Imager product as computed with the Goddard Profiling algorithm (TMI GPROF). One obvious point of comparison is the long-term average, which varies by data set. Over ocean, the TMPA is somewhat higher, while the GPCP SG and the TMPA-RT are close to each other and somewhat lower. In addition, there are inter-annual variations among the data sets. The GPCP SG and TMPA-RT, in common with the TMI GPROF, show larger interannual variations and a phase lag compared to the TMPA and its calibrator, the TCI. The key difference between these two sets of estimates is that the TCI (and so by calibration the TMPA) has both microwave and radar as input, while the other set is predominantly driven over ocean by passive microwave inputs. Over land both the GPCP SG and TMPA use the same Global Precipitation Climatology Centre (GPCC) precipitation gauge analysis, which tends to dominate the land average, although the details are sensitive to the definition of land and ocean.

The interannual differences raise the need to go back to the individual calibrating datasets and determine the basis for the differences as a way of improving our estimate of the long-term global precipitation record.