Investigating rain area delineation over land using data from TRMM

Significant changes in the hydrologic cycle both globally and regionally have been predicted in the future related to an enhanced greenhouse effect. Precipitation is a key parameter of the hydrologic cycle. Satellites can potentially provide measurements of rainfall with a high degree of accuracy and spatially and temporally consistent errors. Precipitation algorithms using passive microwave data have been developed using differing methodologies and for varying purposes. Many algorithms have been developed from cases with medium to high rainfall intensities, where rainfall gradients are large, and therefore precise delineation of rain extent has not been necessarily of a high priority. These same algorithms, however, are often less reliable at low rainfall rates and delineating the extent of the rainfall. Over land surfaces, in particular, algorithms have difficulty distinguishing between land surface effects and atmospheric effects, resulting in false rain signatures and rain underestimation. To a large part this is because the emission of radiation from the ground tends to be high in magnitude, variable in nature and controlled by a large number of factors unrelated to precipitation. These characteristics create problems to the simulation of upwelling radiation, and become particularly troublesome when retrievals are global or over large areas with a variety of surfaces likely to be encountered at any particular time.

In response to this ‘surface’ problem a number of methods have been developed for accounting for the presence of surface artefacts in rain signatures and false rain signatures where no rain is present. On the whole these algorithms comprise statistically derived and empirically calibrated relationships for individual surface types, including deserts, semi-arid land and snow cover. An alternative philosophy for solving this problem, in which every surface is treated as problematic, was pursued by Smith et al. (1998b) and Kniveton (1994). The methods in these studies relied on the assumption that changeable non -precipitating conditions could be characterised from a rolling set of brightness temperature measurements, for different time and space windows, by applying statistical techniques alone, to the satellite measurements. In this paper we explore the use of this statistical modelling approach combined with inverse and forward surface radiative transfer modelling to improve accurate rain area delineation.

The Tropical Rainfall Measuring Mission (TRMM) provides a unique opportunity to assess rainfall delineation approaches using a space borne precipitation radar with the broadly the same temporal and spatial ‘perspective’ as the passive microwave sensor onboard. Data from the TRMM Microwave instrument and precipitation radar are used in this paper to compare rain area delineation techniques.