Recognizing the practical limitations of rain gauges for measuring spatially averaged rainfall over large areas and inaccessible areas, hydrologists have increasingly turned to remote sensing as a possible means for quantifying the precipitation input, especially in areas where there are few surface gauges. Because the fundamental approach to measuring rainfall and snow are different with respect to remote sensing, snow is discussed separately. Although ground-based radar, which is a remote sensing technique, has advanced to an operational stage for locating regions of heavy rain, and for estimating rainfall rates, it will not be discussed in this paper.
Direct measurement of rainfall from satellites for operational purposes has not been generally feasible because the presence of clouds prevents observation of the precipitation directly with visible, near infrared and thermal infrared sensors. However, improved analysis of rainfall can be achieved by combining satellite and conventional gauge data. Useful data can be derived from satellites used primarily for meteorological purposes, including polar orbiters such as The National Oceanographic and Atmospheric Administration (NOAA) series and the Defense Meteorological Satellite Program (DMSP), and from geostationary satellites such as Global Operational Environmental Satellite (GOES), Geosynchronous Meteorological Satellite (GMS) and Meteosat, but their visible and infrared images can provide information only about the cloud tops rather than cloud bases or interiors. However, since these satellites provide frequent observations, (even at night with thermal sensors) the characteristics of potentially precipitating clouds and the rates of changes in cloud area and shape can be observed. From these observations, estimates of rainfall can be made which relate cloud characteristics to instantaneous rainfall rates and cumulative rainfall over time. For the practicing hydrologist, satellite rainfall methods are most valuable when there are no or very few surface gauges for measuring rainfall.
The availability of meteorological and Landsat satellite data has produced a number of techniques for inferring precipitation from the visible and/or infrared (VIS/IR) imagery of clouds. The GOES Precipitation Index [ Arkin, 1979], derived from thresholding the infrared brightness temperature of cloud tops has been used to study the distribution of tropical rainfall. The university of Bristol [See Barrett and Martin, 1981 and D'Souza and Barrett, 1988] has led the development of a cloud indexing [ Moses and Barrett, 1986] approach, a thresholding [ Barrett et al., 1988] approach. A life-history approach, developed by Scofield and Oliver, [1977] considers the rates of change in individual convective clouds or clusters of convective clouds. This approach is the basis of a flash flood system that assimilates GOES data with ground based and atmospheric data to forecast precipitation amounts for use in a flood forecast model [ Clark and Morris, 1986].
Whereas the visible/infrared techniques provide only indirect estimates of rain, microwave techniques have great potential for measuring precipitation because the measured microwave radiation is directly related to the rain drops themselves. Microwave techniques react to rain in two ways: by emission/absorption and by scattering. With the emission/absorption approach, rainfall is observed by the emission of thermal energy by the raindrops against a cold, uniform background. A number of algorithms have been developed to estimate the precipitation over the ocean [ Wilheit et al, 1977, Kummerow et al, 1989]
With the scattering approach the rain is observed through enhanced scattering primarily caused by frozen hydrometers and not directly by the rain. Thus rain rates must be established empirically or with cloud models but this method is not restricted to ocean backgrounds and may be the only feasible approach for estimating rain over land with microwave radiometry. Spencer et al [1989] have shown that the DMSP Special Sensor Microwave/Imager (SSM/I) data can identify rain areas and Adler et al, [1992] has used a cloud based model with 85 and 37 GHz SSM/I data to estimate rain rates. It should be noted that these approaches provide instantaneous rain rates that are then aggregated to yield monthly values which are valuable for climate studies but not for process or operational hydrology.