Retrieval uncertainties in passive microwave rainfall estimations as inferred from the TRMM observation and simulation data

Dong-Bin Shin and Long S. Chiu
Center for Earth Observing and Space Research
George Mason University
Fairfax, VA

Recent advances in remote sensing sensor technology and algorithms have reduced the uncertainties in satellite rainfall measurements from passive microwave radiometers. However, there are still a few unresolved problems. This study investigates two types of algorithm errors in the microwave estimation of rainfall: the inherent and simulation errors. The inherent error is associated with the variability in horizontal and vertical rainfall structures within a satellite's field of view (FOV) in conjunction with the non-linear relationship between brightness temperature and rain intensity. This error is also called as the beam-filling error. Analyses of data collected by the Tropical Rainfall Measuring Mission (TRMM) Microwave Imager (TMI) and Precipitation Radar (PR) show that the horizontal inhomogeneity of rainfall is related to its vertical structure and hence the combined effect contributes to increasing the uncertainty in the retrievals. For physically-based rainfall retrieval algorithms there is an error in a-priori knowledge, which is used to invert observations into physical parameters. The error arises from the uncertainties in cloud-radiative transfer models, including the assumptions about the microphysical properties of hydrometeors, surface emissivity model, parameterization of melting particles, three-dimensional effect of precipitating system, and so on. This study considers the error as the simulation error. The simulation error has been discussed in terms of the difference between simulated precipitation fields and matching observations. A simple Bayesian rainfall algorithm is also adopted to have an insight on the contribution of the two errors to the retrieval accuracy. This study discusses that the inherent error depends on the rainfall characteristics and sensor responses, but is independent on cloud-radiative transfer models. However, more precise cloud-radiative transfer models will reduce the simulation error. It is also demonstrated that lower resolution retrievals are less sensitive to the errors than the retrievals on higher resolutions.

[back]