Radar-derived estimates of rainfall intensity and accumulations offer unsurpassed spatial and temporal resolution, and are thus critical not only for issuance of flash-flood warnings but more broadly as input to agricultural and hydrological models for cropland and river/streamflow management. These measurements are thus a critically important product of the nationwide WSR-88D "NEXRAD" radar network. This research effort is focused on improved rainfall estimates through detection of departures from well-behaved "Rayleigh-type" radar signal behavior that may induce errors in deduced rainfall. The presumption that well-behaved Rayleigh-type statistics dominate observed storm properties is at the foundation of current radar-based precipitation estimation techniques. Though strong spatial gradients of rainfall intensity characteristic of thunderstorms are one potential source of non-Rayleigh signal behavior, research suggests that this complication may also arise in the context of more homogeneous precipitation for certain types of radars and radar scanning strategies (including measurements of differential radar reflectivity and derived hydrometeor type from polarized radars). In extreme cases--again generally associated with the heaviest areas of precipitation--induced errors may locally approach the magnitude of the derived rain rate itself. Within the context of this study, the existence of this statistical complication will be conveyed via computation of a "clustering index (CI)."
With the advent of high-resolution weather forecast models, radar observations will likely soon be assimilated into these in real-time to help better guide their predictions. The ability to reduce radar errors (or even simply better quantify the degree of uncertainty inherent in these measurements) would be of particular value in the context of modern data assimilation schemes. This proposal seeks to process high-resolution radar data collected by the NSF-supported CHILL radar in eastern Colorado to better relate the volumetric structure and evolution of CI anomalies to storm morphology and underlying cloud microphysical processes, as well as to extend CI measurements to snow, graupel and hail events. Other potential contributors to radar reflectivity bias (including "Bragg scattering," generally regarded as arising from the turbulent mixing of media with differing indices of refraction, as may occur at cloud/plume edges) will also be examined. The availability of high resolution measurements from CHILL (viz. radial data spacing as fine as 30 m) will further facilitate this work. Subsidiary efforts will address independent data sources such as surface-based disdrometer raindrop size distributions to shed additional light on non-Rayleigh precipitation behavior. The intellectual merit of this effort thus focuses on improved radar signal processing and associated understanding of the structure and dynamics of a wide variety of precipitating clouds. As suggested above, Broader Impacts will include the potential for appreciably improved measurements and short-term forecasts of precipitation intensity and amount.
As part of the development of radar for use in WWII, it was discovered that raindrops and ice particles sent measurable signals back to the radar. After the war, there was a great deal of interest to apply radars to peaceful purposes and the question of using radars in meteorology emerged. Simultaneously, the theory of how raindrops reflect radar signals was developed which allowed a quantitative interpretation in terms of rain intensity, for example. This development was important because it meant that there was hope for rapidly measuring precipitation over large areas, something never previously possible except in a few instances of expensive and immobile networks of rain gages. A key ingredient of the theory of radar signals from rain and snow that was developed is the assumption that each drop scatters energy back to the radar independently of what other drops are doing because they are spread uniformly in space. However, this assumption has never been verified, and there is reason to question it. Specifically, just by looking at streaks of precipitation falling from clouds and as they sweep across the pavement during thunderstorms, it is apparent that raindrops are not always scattered uniformly in space, but rather they can show what is referred to as ‘clustering’ or ‘bunching’ as streaks. That is, at times, drops (and snow flakes) can form clusters over many spatial scales. It turns out that this characteristic of precipitation is important because when spaced just right with respect to the wavelength of the radar, the signals from each drop can add together coherently to produce a much bigger signal than the sum of each separate drops spread randomly. This is analogous to the effect of soldiers marching in step across a suspension bridge causing ever increasing vibration in the bridge. The major advance under this grant has been the direct observation that this ‘coherent’ part of the signals scattered back to the radar does occur, and that it can, at times, become a significant portion of the total radar signal, often around 30% in rain and up to 70% in snow. When this happens it screws up the interpretation of radar measurements for the quantitative measurement of precipitation that have been developed over the last 60 years under the assumption that all the signals are ‘incoherent’. Unfortunately, there is no way to ‘correct’ for coherent scatter except to remove it from the observations. But most radar antennas are always moving. It turns out, however, that when a radar antenna is scanning it is not possible to measureand to remove the coherent signal contribution to the total radar signal even though it is still there. Thus, for all current operational radars, no correction is presently possible. Fortunately, there is a way out of this dilemma by using technology just now being added to the nation’s radar network system. This technology is referred to as ‘dual-polarization’. That is, radars can send out two kinds of electromagnetic waves. When the signals rain for the two different waves are combined in just the right way, any effect caused by coherent scatter will cancel out. It will not always work, but a lot of the time when radar measurements of precipitation are most important, the new technology will succeed in eliminating the effects of any coherent scatter particularly in rain. The applicability of this approach to snow is not so clear, however. Thus this research provided new insights into the physics behind the radar measurements of precipitation with obvious practical implications whenever there is a need for quantitative radar estimates of precipitation such as for flood warnings and for the optimization of city water management during storms to prevent flooding. This grant produced 6 refereed journal articles.
