Practical limitations of in situ observation networks for measuring spatially averaged precipitation over large and inaccessible areas have promoted the use of remote sensing to quantify precipitation. Both radar and satellites are important contributors to providing real-time precipitation measurements for large areas. Meteorological radars have a spatial resolution of 1-2 km and temporal revisit times of 15-30 minutes. Satellite remote sensing products have a spatial resolution of 10-20 km and temporal repeats of 1-2 times daily. However, the quantitative estimation of errors associated with radar and satellite precipitation data is severely hampered often by the lack of an independent, precise, and accurate rain gauge value with which to compare the remotely sensed data. Aggregating the data over longer periods and increasing the number of rain gauges used in the comparison reduces uncertainties (Yuter, 2003), and calibrated estimates from weather radars provide improved spatial representation of short-term precipitation patterns (Legates, 2000a).
The weather services of many industrialized nations have networks of land-based operational precipitation radars used to determine the location, size, and intensity and motion of precipitation events and to identify the type of precipitation. Ground-based radars Europe, Japan, Canada, and the United States are used for short-term weather and flood forecasting, to estimate the distribution and amount of cumulative precipitation for a specified region, to map the three-dimensional structure of storms, and to produce precipitation estimates for hydro-logic models (Yuter, 2003; Neary et al., 2004). Precipitation measurement by radar is an attractive alternative because it provides coverage of a large area with high spatial and temporal resolution from a single observing point. In addition, the area can be extended by compositing data from several radars, and in this way ground-based radars provide better spatial and temporal coverage than rain gauges.
Europe and the United States have the largest number ofweather radars. The Next Generation Weather Radar (NEXRAD) program in the United States supports 166 Weather Surveillance Radar-1988 Doppler (WSR-88D) systems and is managed by a joint agreement of three Federal agencies. The network of WSR-88D systems integrates advanced radar capabilities, real-time signal processing techniques, meteorological and hydrological algorithms, and automated product processing and analysis that are continually updated (Klazura and Imy, 1993; Crum and Alberty, 1993). An overview of the basic characteristics of radar-based rainfall observations is presented here, but detailed treatments are found in WMO (1996) and Yuter (2003).
The physical foundation for precipitation radars is the relationship between range-corrected, backscattered returned power and the size and number of reflecting targets in the radar beam (Fig. 5.10). Precipitation radars have a
Fig. 5.10. Simplified diagram of a pulse weather radar and selected characteristics that determine the radar's effectiveness in detecting precipitation.
Fig. 5.10. Simplified diagram of a pulse weather radar and selected characteristics that determine the radar's effectiveness in detecting precipitation.
I Target transmitter that switches on and off to transmit a pulse of electromagnetic energy via an antenna. The pulses are typically spaced on the order of a millisecond apart. Radar frequencies are divided into several bands, but practical considerations favor longer S-band and C-band wavelengths of 4 to 15 cm for stationary ground-based radars. These radars require larger diameter antennas that are more costly to operate than the antennas required for shorter wavelength radars deployed on satellites, aircraft, and ships (Yuter, 2003). Most precipitation radars use a circular parabolic antenna for both transmission and reception.
When the transmitted energy encounters a raindrop, some of the energy is absorbed and some is scattered in all directions. Although the raindrop serves as a reflector for the radar transmitted energy, only a small fraction of the incident radiation is scattered back toward the antenna where it is received and amplified. The backscattered power measured by the radar relates to the radar characteristics and the precipitation target characteristics. The total energy backscattered is the sum of the energy backscattered by each scattering particle. The return signal is quantitatively expressed by the radar equation, which has a radar term and a target term. The radar constant, cr, groups numerical constants and radar hardware parameters and is expressed as c ^PtG2l2MvCrpP3
where Pt is the peak power of the pulse transmitted by the radar in W, G is the antenna gain and is dimensionless, l is the wavelength of the transmitted wave in m, 6h is the horizontal bandwidth in radians, dv is the vertical bandwidth in radians, c is the speed of light (3 x 108ms~1), and rp is the pulse duration in seconds. The physical characteristics of the precipitation particles within the radar resolution volume are represented by the complex index of refraction, |K|2, which has a value of 0.93 for water and 0.197 for ice. The intensity of the return signal related to the density and size of the targets is expressed by the radar reflectivity factor, Z, which is determined by
where D is the drop diameter in mm and N(D) dD is the number of particles per unit volume in the diameter range D to D + dD. The radar equation for spherical drops is expressed in simplified form as
where Pr is the average backscattered power received by the radar over several pulses in W, rr is the range to the target relative to the radar, and the other terms are defined previously. The time delay between the original pulse transmission and receipt of the backscattered energy by the antenna is used to determine the distance to the raindrop. The relationship between the backscattered, returned energy and the size and number of the reflecting targets is the physical foundation for interpreting precipitation radar data (Yuter, 2003). Radar is capable of detecting precipitation and variations of the atmospheric refractive index generated by local variations of temperature or humidity. There are a number of basic assumptions inherent in these equations, but they serve as the basis for reasonable estimates of precipitation amounts from radar measurements (WMO, 1996).
The return signal from a radar transmitted pulse encountering a target is called an echo, and the most widely adopted approach for measuring rainfall using radar is based on the radar echo or reflectivity (Ward and Robinson, 2000). The radar echo has amplitude, a phase, and a polarization. Most operational radars are limited to analysis of the amplitude related to the size distribution and numbers of particles in the volume illuminated by the radar beam. The amplitude is used with empirical relations to determine the reflective factor to estimate the mass of precipitation per unit volume or the intensity of precipitation. Doppler radar has the capability of determining the phase difference between the transmitted and received pulse that is a measure of the mean Doppler velocity of hydrometeors (i.e. raindrops, snowflakes, or hail stones) or their motion (WMO, 1996). Radio waves reflected by objects moving away from the Doppler antenna change to a lower frequency, and waves reflected from an object moving toward the antenna change to a higher frequency. The frequency shift relative to the transmitted signal is expressed as fd = -2 V; (5-7)
where fd is the Doppler frequency or shift, Vd is the velocity, and lr is the radar wavelength. The change of the backscatter phase from pulse to pulse provides a measure of the change in range from the radar to the hydrometeor and the Doppler shift (Doviak and Doviak, 2003). The velocity component of a target relative to the radar beam is referred to as the radial velocity. The centimetric waves used by Doppler radar permit them to penetrate extensive fields of precipitation to identify the morphology of weather systems. This characteristic is a distinct advantage over optical and infrared waves which do not penetrate far into clouds and precipitation.
Indications of hydrometeor size serve as the basis for estimating precipitation intensity and amount by employing the reflectivity of targets or the power returned from a pulse volume. Rainfall rates are proportional to the volume of the raindrops, but the reflectivity is proportional to their surface area. The relation between radar reflectivity and rainfall rate is not constant, but extensive experimental results suggest it has the form
where Z is radar reflectivity in mm6 m~3 or dBZ, R is the rainfall rate in mm h~\ and a and b are coefficients. The most common value for a is 200, but it ranges from 70 to 500. The most common value for b is 1.6 with a range of 1.0 to 2.0. Variations in the Z-R relationship are related to physical differences in the form and size of the precipitation, radar clutter produced by ground echoes, the presence of an enhanced "bright band'' related to a melting snow layer, and radar signal attenuation due to heavy rainfall. An equivalent radar reflectivity factor Ze may be used in the Z-R relationship when precipitation aloft is measured by the radar and compared to R measured at the ground (WMO, 1996).
Reflectivity is commonly expressed in log scale decibel units (dBZ) with higher dBZ values indicating more power reflected and received by the radar. The radar computer system determines the rainfall rate and produces an estimate of the rainfall amount using a series of empirically derived equations. Light rainfall produces a reflectivity of 20-30 dBZ, moderate rainfall 30-45 dBZ, and intense rainfall 60-70 dBZ.
The NEXRAD WSR-88D radar systems deployed in the United States are active S-band Doppler radar systems operating at a wavelength of about 10 cm. WSR-88D antennas continually scan their environment in a sequence of preprogrammed 360° azimuthal sweeps at various elevations that comprise a volume scan. Two common volume scans represent basic operational modes that accommodate the sensitivity range of the radar. The clear air mode is the normal operational mode in which the radar rotates slower and is sensitive to the smallest echoes. The radar rotates faster in the precipitation mode and provides more rapid data updates, but it sacrifices sensitivity at lower reflective values. Processing the returned power spectral density provides the data necessary to estimate reflectivity, mean radial velocity, and velocity spectral width, which is a measure of the variability of the radial velocities in the sample volume (Crum et al., 1993; Klazura and Imy, 1993).
Validation of radar-rainfall products is a major challenge for broad utilization of these products in hydroclimatic applications, and understanding the error structure of radar-rainfall estimates is especially critical in utilizing this information to improve quantitative precipitation forecasts and to estimate extreme rainfall and flooding (Krajewski and Smith, 2002; Yuter 2003; Fritsch and Carbone, 2004). Experimental results indicate that improvements in quantitative precipitation forecasts are achieved by combining satellite real-time rainfall estimates with mesoscale model generated relative humidity and precipitable water and surface-radar derived instantaneous rainfall estimates (Vicente et al. 1998).
Satellites are used increasingly to provide precipitation estimates over oceans and continents and especially for data-sparse regions such as deserts, mountainous regions, and humid tropical regions. Unfortunately, direct measurement of rainfall from satellites is hindered by the presence of clouds that prevents observation of precipitation with visible, IR, or microwave sensors, and this requires reliance on passive methods for satellite remote sensing of precipitation. Passive methods determine precipitation indirectly using algorithms that transform satellite-sensed radiance from clouds or raindrops into precipitation. One widely used method is to employ the flux of outgoing longwave radiation estimated from satellite observations as a basis for estimating precipitation at a variety of temporal and spatial scales (Xie and Arkin, 1998).
Passive methods are based on the radiative intensities emitted or reflected by cloud and precipitation hydrometeors using visible, IR, and microwave portions of the electromagnetic spectrum. IR and visible methods are physically indirect because precipitation is derived from the radiative properties near the cloud top. Visible methods are less widely applied with the advent of IR and microwave measurements and the lack ofnighttime visible observations (Greene and Morrissey, 2000). Microwave techniques use more direct information on the vertical distribution of hydrometeors in a column of the atmosphere because the measured microwave radiation is directly related to the actual raindrops.
Active sensing of precipitation by satellite radar is accomplished by the TRMM precipitation radar, which was the first radar designed specifically for rainfall monitoring from space. The joint Japanese and United States TRMM is a low-orbit satellite stationed between 35° N and 35° S that has produced a wealth of detailed information on tropical rainfall (Kummerow et al., 2000). In the microwave region, precipitation is estimated by the MSU of the TOVS on TIROS polar-orbiting satellites, the SSM/I on DMS series polar-orbiting satellites, the AMSR-E on NASA's EOS satellites, and by the TRMM Microwave Imager (TMI). Satellite measurements of precipitation by TOVS, SSM/I, and AMSR-E provide spatially uniform global coverage over both land and water for support of hydroclimatological analysis (Susskind, et al., 1997; Ferraro, 1997; Wilheit et al., 2003).
Radiative intensity for IR and microwave wavelengths is expressed in terms of brightness temperature, which is the temperature required to match the measured intensity to the Planck blackbody function. IR brightness temperature commonly represents the physical temperature of the cloud top because most clouds are optically thick for IR radiation. Colder IR brightness temperatures often indicate higher cloud heights and higher rainfall rates at the surface. The relationship between IR brightness and precipitation is strongest for deep convection in the tropics and is less consistent in the mid-latitudes where most precipitation is produced by frontal stratiform clouds.
The radiative intensity for microwave radiation is the integrated contribution by all water drops and ice particles in the atmospheric column because microwave radiation can penetrate through cloud and precipitation layers. Microwave brightness temperatures may increase or decrease with increasing rainfall rate depending on the microwave frequency and the cloud microphysi-cal properties. At frequencies below 20 GHz, scattering by ice particles becomes negligible and rainfall cannot be detected over land because of high surface emissivity. Ocean surface temperature and emissivity do not vary dramatically and changes in brightness temperature can be attributed to the change in the optical depth of raindrops, which is approximately proportional to integrated total rainwater amount. However, the brightness temperature increase with rainfall rate reaches a maximum that indicates saturation of microwave radiation and further increases in rainfall beyond this point have decreased brightness temperatures.
Scattering by ice particles is the dominant signature of rain clouds at microwave radiation frequencies above 80 GHz. For high-frequency microwave radiation, the brightness temperature decreases with increasing optical depth of ice particles. The lower brightness temperature indicates more large ice particles aloft, which are commonly an indication of heavier rainfall at the surface.
Measured radiative intensity in the visible spectrum is due to sunlight reflection by clouds and surface features and is limited to daylight hours. Reflectivity in the visible spectrum increases with cloud optical depth, which is proportional to the vertically integrated liquid water path if the effective droplet size remains constant. Clouds with high optical depth are more reflective and more likely to be associated with precipitation. However, the sensitivity of visible reflectivity to the liquid water path decreases with increasing optical depth and becomes virtually insensitive to optical depth at values representing a cloud producing rainfall. Consequently, reflected visible radiation indicates the physical properties near the top portion of the cloud and the relation to the rainfall rate at the surface is rather indirect (Liu, 2003).
Visible and IR images from polar-orbiting and geostationary satellites provide information on cloud tops only, but the frequent observations by these satellites permit identification of characteristics that can be related to rainfall rates and cumulative rainfall. One widely used technique is the GOES precipitation index, which is based on the fraction of cloud colder than 235 K in the IR and a fixed rainfall rate. The main disadvantage of such approaches is that they infer surface rainfall from cloud-top characteristics. Rainfall estimates are improved by using passive microwave calibrated IR estimates of precipitation at a high spatial and temporal resolution (Kidd et al., 2003). An adaptation of the GOES precipitation index provides globally complete 3-hour precipitation estimates using a combination of microwave data from multisatellite observations (Huffman et al., 2001). Georgakakos et al. (2001) use visible, IR, and water vapor data from Meteosat to estimate small-scale daily rainfall over the Nile River Basin. Chen and Staelin (2003) describe an algorithm for the Aqua AIRS for estimating precipitation using an opaque-channel approach at 15 km resolution that has potential for global precipitation applications.
A comparison of 25 satellite-based monthly global precipitation estimates is described by Adler et al. (2001). The satellite-based products were compared with four model and two climatological products, and the evaluation concluded that merged data products provided the best overall results. Combined precipitation estimates from low-orbit satellite microwave data, geosynchronous-orbit satellite IR data, and rain gauge data are used by the Global Precipitation Climatology Project (GPCP) established by the World Climate Research Program (WCRP) to provide monthly mean data for 1979-2001 on a global 2.5° x 2.5° latitude-longitude grid (Huffman et al., 1997; Adler et al., 2003). These monthly data have served as the basis for 5-day global analysis (Xie et al. (2003), and a daily 1° x 1° latitude-longitude analysis for 1997-2001 (Huffman et al., 2001). New et al. (2001) address how satellite precipitation estimates can be incorporated in assessing trends as multidecadal merged precipitation datasets become available.
Was this article helpful?