556
A.D. Del Genio
observed attenuation A is given by A '" kRH, where R is the path-averaged rain rate and H
the path length through the raining layer. If we assume that H is equal to the slant range
between the ground and the freezing level (identified by the radar bright band), then R can
be determined from a measurement of H. In practice, though, supercooled liquid water exists
above the freezing level, limiting the accuracy of this simple estimate.
A variety of other algorithms have been developed for space rain estimation. The surface
reference method (Meneghini et al., 1983) compares the surface power return in raining and
non- raining conditions to isolate the attenuation due to rain. This procedure is effective unless
the nature of the surface return changes with time or is different underneath a rain system.
The utility of such techniques is of course limited by the sensitivity of the radar to different
rain rates and its ability to "see" the surface through the rain. For the PR, Okamoto and Kozu
(1987) estimate that rain rates < 1 mm/hr are detectable, with maximum observable rain rates
of 40 mm/hr and the surface return detectable through rain rates as high as 90 mm/hr. The
mirror image method (Meneghini and Atlas, 1986) uses the surface in a different way: Energy
reflected from the surface to the rain layer and then back again to the surface produces a virtual
image of the precipitation below the surface in the radar return. The weaker return from the
backscatter of the mirror image relative to that directly scattered back to the radar by the rain
contains information about the rain rate. Since this method requires data from an even further
attenuated portion of the signal, it fails at lower rain rates than does the surface reference
technique.
The Height Area Rainfall Threshold technique takes a completely different approach to rain
estimation (Atlas et al., 1990). Since rain distributions seem to be fairly well described by the
log normal pdf, the fraction of a given area that is observed to be raining above some threshold
detectable rain rate (i.e., above a threshold value of Z) is correlated with the total area of rain
(and thus total rain volume, the integral of the rain rate over area and time). The rain volume
in turn is related to rain area by the average rain rate in areas that are raining; this depends
on the height of the storm, which can be measured also by the radar (or by other TRMM
instruments). The strength of this approach is that it potentially provides a "snapshot" of the
area-wide rain, rather than requiring multiple TRMM visits to a region over a month to build
up suitable accuracy.
The weakness is its dependence on the existence of a universal rain pdf; a priori this is not
known to be the case, although TRMM data should constrain the degree of spatial and temporal
variability of the pdf.
20.4.2 TRMM Microwave Imager (TMI)
Figure 20.4 illustrates schematically the nature of the sources of radiation seen by a passive
microwave sensor. Note first that the Planck function is approximately linear in the microwave
region of the spectrum, so that microwave radiances can conveniently be discussed in terms of
the brightness temperature of the emitting object. Microwave rain detection depends heavily
on the fact that the ocean surface has low microwave emittance (typically e ::; 0.5). The
brightness temperature of the ocean surface (n.) is thus much lower than the actual sea surface
temperature (T.). Microwave emission from the top of a rain- producing system (e '" 1)
consequently appears as a warm anomaly above a cold surface (just the opposite of infrared
detection of cumulonimbus, which appear as cold cloud tops against a warm surface).
A number of complicating factors arise, though, in uniquely estimating rain rate from passive
microwave data. Land surfaces have much higher microwave emissivities than the ocean, and
the emittance varies with vegetation type, soil moisture, etc., so the contrast between rain
A.D. Del Genio
observed attenuation A is given by A '" kRH, where R is the path-averaged rain rate and H
the path length through the raining layer. If we assume that H is equal to the slant range
between the ground and the freezing level (identified by the radar bright band), then R can
be determined from a measurement of H. In practice, though, supercooled liquid water exists
above the freezing level, limiting the accuracy of this simple estimate.
A variety of other algorithms have been developed for space rain estimation. The surface
reference method (Meneghini et al., 1983) compares the surface power return in raining and
non- raining conditions to isolate the attenuation due to rain. This procedure is effective unless
the nature of the surface return changes with time or is different underneath a rain system.
The utility of such techniques is of course limited by the sensitivity of the radar to different
rain rates and its ability to "see" the surface through the rain. For the PR, Okamoto and Kozu
(1987) estimate that rain rates < 1 mm/hr are detectable, with maximum observable rain rates
of 40 mm/hr and the surface return detectable through rain rates as high as 90 mm/hr. The
mirror image method (Meneghini and Atlas, 1986) uses the surface in a different way: Energy
reflected from the surface to the rain layer and then back again to the surface produces a virtual
image of the precipitation below the surface in the radar return. The weaker return from the
backscatter of the mirror image relative to that directly scattered back to the radar by the rain
contains information about the rain rate. Since this method requires data from an even further
attenuated portion of the signal, it fails at lower rain rates than does the surface reference
technique.
The Height Area Rainfall Threshold technique takes a completely different approach to rain
estimation (Atlas et al., 1990). Since rain distributions seem to be fairly well described by the
log normal pdf, the fraction of a given area that is observed to be raining above some threshold
detectable rain rate (i.e., above a threshold value of Z) is correlated with the total area of rain
(and thus total rain volume, the integral of the rain rate over area and time). The rain volume
in turn is related to rain area by the average rain rate in areas that are raining; this depends
on the height of the storm, which can be measured also by the radar (or by other TRMM
instruments). The strength of this approach is that it potentially provides a "snapshot" of the
area-wide rain, rather than requiring multiple TRMM visits to a region over a month to build
up suitable accuracy.
The weakness is its dependence on the existence of a universal rain pdf; a priori this is not
known to be the case, although TRMM data should constrain the degree of spatial and temporal
variability of the pdf.
20.4.2 TRMM Microwave Imager (TMI)
Figure 20.4 illustrates schematically the nature of the sources of radiation seen by a passive
microwave sensor. Note first that the Planck function is approximately linear in the microwave
region of the spectrum, so that microwave radiances can conveniently be discussed in terms of
the brightness temperature of the emitting object. Microwave rain detection depends heavily
on the fact that the ocean surface has low microwave emittance (typically e ::; 0.5). The
brightness temperature of the ocean surface (n.) is thus much lower than the actual sea surface
temperature (T.). Microwave emission from the top of a rain- producing system (e '" 1)
consequently appears as a warm anomaly above a cold surface (just the opposite of infrared
detection of cumulonimbus, which appear as cold cloud tops against a warm surface).
A number of complicating factors arise, though, in uniquely estimating rain rate from passive
microwave data. Land surfaces have much higher microwave emissivities than the ocean, and
the emittance varies with vegetation type, soil moisture, etc., so the contrast between rain