The measurement of precipitation at sea remains one of the most challenging observational issues. Wind flow over ships influences the collection efficiency of gauges, and the short space and time scales of precipitation make point measurements difficult to interpret. Thus, remote sensing techniques must be used. Several satellite based systems, shipboard radar, and ocean acoustic noise measurements show potential for monitoring on the necessary space and time scales. Rasmusson and Arkin [1993] have reviewed the climate scale estimates of global precipitation derived from satellites. However, calibration of such techniques remains an area requiring new approaches and much work.
There are several indicators of rainfall which can be derived from satellite observations [ Arkin and Ardanuy, 1989]:
Highly Reflective Cloud (HRC) --- This technique uses visible light to identify those clouds which are most likely to be rain sources. The frequency of HRC can be related to rainfall measurements from atoll stations. It is somewhat subjective, suffers from poor diurnal sampling, and is restricted to the tropics. It does have the advantages of a long record, beginning in the early 1970s, and high spatial resolution. This approach is one of those used in the WCRP Global Precipitation Climatology Project (GPCP).
Outgoing Longwave Radiation (OLR) --- Values of OLR in the tropics respond strongly to variations in cloudiness, and is thus a useful index of convective activity. It has been used as both a qualitative and quantitative indicator of precipitation. Its sampling of the diurnal cycle, at twice per day, is better than HRC, though still suboptimum. It is restricted to low latitudes where precipitation is primarily from deep cumulonimbus convection (+-40 ). Janowiak and Arkin [1991] and Arkin and Xie [1994] describe some result of its use within the GPCP. OLR data are also being used as indirect forcing of the heating in NWP systems; these systems in turn are giving improved estimates of tropical precipitation.
Passive Microwave --- a. SSM/I (Special Sensor Microwave/Imager) This technique works over the oceans, where a uniform
background of microwave emission allows rainfall to be detected by
the absorption of microwave energy by raindrops or the scattering of
higher-frequency microwave energy by ice particles. There are polar
orbiting satellites carrying the appropriate sensors, so it shows
promise for the important problem of the high latitude, open ocean
detection of rainfall [ Wilheit et al., 1991]. However, the technique
cannot be used over land, snow or ice. The 30-km footprint is larger
than most rain cells; the partial occupation of the footprint and a
nonlinear response to rain degrade its utility in the tropics. Larger
scale rain systems at higher latitudes may be adequately resolved by
this technique. Because of spotty space and time coverage, accurate
climate scale estimates of accumulations (5
1 month)
may be problematic, though less so outside the tropics. In the
tropics, the passive microwave satellites may be used to ``calibrate''
the OLR technique, which has better space/time sampling. Chang, Chiu
and Wilheit [1993] provide a global precipitation map derived from
the SSM/I.
b. MSU (Microwave Sounding Unit). The microwave brightness temperature measured from satellites can be corrected for air mass temperature; when it exceeds a certain threshold precipitation is inferred because of cloud-water and rainwater induced warming. Spencer [1993] has developed a global ocean precipitation climatology for 1979-1991 based on the MSU. It compares favorably to the Legates and Wilmott [1990] compilation (based largely on the Dorman and Bourke application of the Tucker method), though it indicates higher precipitation in the eastern tropical Pacific (to 5.6 m/yr).
Tropical Rainfall Measuring Mission (TRMM) --- This is a rain radar
satellite to be launched in 1997 as a joint Japan/U. S. mission
( Simpson et al., 1998). It will
fly in a low-altitude, low inclination orbit to provide good resolution at
latitudes less than 
35 and will be non-sun-synchronous in order to
sample the diurnal cycle over monthly periods. Rain radars must use
additional information such as the reflectivity as a function of dropsize
distribution, or direct rain gauge measurements, in order to estimate a
rain rate. Comparisons with point measurements are good to only a factor
of 2; however, averaging over larger space and time scales improves the
accuracy.