Abstract: (Summary adapted from the AMSR Rainfall ATBD)

Algorithm Theoretical Basis Documents (ATBDs) are being developed for every EOS
instrument product. These documents, consist of a detailed physical and
mathematical description of the algorithm, variance or uncertainty estimates,
and practical considerations, such as calibration and validation, exception
handling, quality assessments and diagnostics. ...

The aim of the precipitation algorithm is threefold: The first is to produce
instantaneous rainfall intensity on a pixel-by-pixel basis (Level 2). In
addition to rainfall intensity, over ocean, rainfall will be categorized as
convective or stratiform. The rainfall, as well as the latent heating which can
be derived from the convective/stratiform separation play an important role in
Global Circulation Model (GCM) initialization as well as data assimilation
efforts currently underway; The second objective is to produce estimates of
monthly totals on a 5' x 5' basis (Level 3). This product is aimed particularly
at climate monitoring as well GCM validation. The distinction between
instantaneous and monthly products, aside from the applications, is
necessitated by the poor temporal sampling of the AMSR instrument.
Fortunately, there are statistical properties of rainfall, which can be
exploited in order to gain greater confidence in monthly rainfall
accumulations. The third objective is to generate credible uncertainty
estimates. This requires that the algorithms, to the maximum extent possible,
be based on models with well-established physics. The detection of changes on a
wide variety of space and time scales is extremely important. By avoiding the
use of arbitrary tuning parameters, especially those that vary seasonally and
regionally, we have a high degree of confidence that any change we see in the
retrieved rainfall is a change in the actual rainfall rather than of some
tuning parameter. The physical basis further insures that rainfall estimates
made from AMSR can be easily adapted to previous as well as future sensors. Is
such a way, it will be possible to extend AMSR rainfall climatologies backward
to 1987 using SSM/I observations. It will also insure that knowledge gained
from the TRMM mission is easily incorporated into the AMSR retrieval. In the
ensuing discussion, the reader will see that we have been reasonably, if
imperfectly, successful in approaching this goal for the ocean algorithms,
where a large and reasonably constant reflectivity of the ocean background
enables approaches that are simply not possible if there is a significant
amount of land in the field of view. Unfortunately, for the land algorithms,
much less has proved possible.