The Emergence of Inversion?Type Profile Algorithms for Estimation of Precipitation From Satellite Passive Microwave Measurements

Eric A. Smith

Dept. of Meteorology and Supercomputer Computations Research Institute, Florida State University, Tallahassee, FL 32306, USA

Christian Kummerow

Severe Storms Branch, Code 612, NASAIGoddard Space Flight Center, Greenbelt, MD 20771, USA

Alberto Mugnai

Istituto di Fisica dell'Atmosfera, Consiglio Nazionale delle Ricerche, Frascati 00044, Italy

ABSTRACT

The evolution of satellite precipitation retrieval from a single input-single output regression-type perspective to a multiple input-multiple output inversion-type perspective is reviewed. In essence, because the first attempts to retrieve rainfall from satellite measurements involved using a single input variable obtained from either a visible or infrared channel, estimates were limited to a single rainrate or single rain fallout parameter, and the prevailing view concerning the nature of the problem became fixated on estimating the surface rainfall. This view persisted even after multispectral passive microwave measurements capable of describing the vertical rain profile became available, and continues to persist today along side an emerging view which holds that the basic physics of the retrieval problem is more naturally connected to estimating the vertical profiles of liquidlice water contents, and then relating these retrieval vectors to surface rainrates. The theoretical basis for the inversion-type approach is examined, focusing on explaining why the natural problem in using multispectral passive microwave measurements for precipitation retrieval should involve the measurement of profiles of various hydrometeors whose varying vertical distributions dictate the magnitudes of the brightness temperatures at different frequencies. The physical and numerical designs of three current algorithms which utilize multispectral inversion principles are described (the American GSFC and FSU schemes, and the Italian IFASAP scheme). Because such inversion schemes are computationally intensive, this issue is given critical examination. Various computational strategies for overcoming the prohibitive CPU costs associated with global applications using a brute force pixel-by-pixel approach are discussed, including the current benchmark-grid approaches being used with the FSU and GSFC schemes. Finally, future trends are examined, recognizing that because this is a relatively new tropic-area, additional research must be done before inversion-type schemes can fully mature.