Analyzing the Error Components in Blended Satellite Precipitation Analyses

In recent years the capability to quantify precipitation from space has been greatly enhanced with the addition of several new measurement capabilities, most notably from passive radiometric (PMW) sensors such as the Tropical Rainfall Measuring Mission (TRMM) Microwave Imager (TMI) and its companion Precipitation Radar (PR), and the EOS-Aqua Advanced Microwave Scanning Radiometer (AMSR-E). Additionally, the requirements in climate modeling, precipitation data assimilation, and hydrologic applications have necessitated the need for daily (and sub-daily) precipitation analyses and their associated accuracy. Owing to intermittent PMW orbit coverage, this has necessitated the blending of these data with the rapid-time capability of geostationary-based thermal infrared (IR) observations. The types of blending schemes can be subdivided into two main categories, those that utilize the IR observations to advect precipitation in between PMW revisits, and those that utilize the PMW to calibrate the IR observations in-between satellite revisits. While capturing more of the underlying cloud evolution and areas of extreme precipitation intensification, the blending operation does introduce a complicated error that has several components. Common to all schemes is the error that is introduced from the instantaneous PMW rainfall estimates. A related error results from the inter-sensor bias between each PMW sensor type, which is primarily a function of the sensor field-of-view, rainfall rate, latitude, season, and the background surface type. A more complex error is related to the revisit schedule of the underlying PMW satellite constellation. Other more scheme-specific errors result from the way that the IR data are used to maintain temporal persistence in-between satellite revisits. Since extreme precipitation events are often associated with orographic conditions, the extreme events (the tail of the rainfall distribution) may be under-represented unless terrain and environmental conditions are taken into account. The overall combined error characterization is complicated by the fact that it is constantly changing due to the PMW overpass schedule (and may involve periods of missing data), and where these observations lie with respect to the space and time scales of the blending process.

In this presentation we will demonstrate some of these factors using two recent blending schemes, the Climate Prediction Center Morphing (CMORPH) and the Naval Research Laboratory Blended-Satellite (NRL-Blend) techniques. Examples will be shown which depict the differences in the accumulated precipitation that results when the PMW satellite configuration is artificially modified, and when inter-sensor rainfall bias is taken into account.