Performance of algorithms for rainfall retrieval from dual-polarization X-band radar measurements

Abstract

Quantitative precipitation measurement remains a key topic in radar meteorology. On the one hand, in urban hydrology at the small catchment scale, it is required the repetitive high sampling and space resolution that also required over a wide area that can only be achieved with weather radar. On the other hand, the study of the hydrological cycle at global scale requires large scale observations such as those provided by satellites. Satellites use onboard microwave active or passive sensors for measuring rainfall or Infrared radiometers for measuring cloud top temperature. However, there are open questions on the accuracy of precipitation retrievals from satellites. Ground validation on the basis of in situ measurements is impaired since in different climate regimes there are different cloud and precipitation microphysical processes that can affect satellite rain retrievals. Weather radars capability to monitor precipitation at high spatial and temporal scales has stimulated great interest and support within the hydrologic community. The US National Weather Service (NWS) is using an extensive network of weather surveillance Doppler radar (WSR-88D) systems (Heiss et al. 1990), which can bring dramatic advancements to the precipitation monitoring with direct implications on the improvement of real-time forecasting of river floods and flash floods. Precipitation, though, may originate from varying meteorological systems, ranging from cold frontal systems to thunderstorms and tropical systems, where rainfall estimates based on these classical single polarization radar observations have quantitative limitations (e.g., Smith et al. 1996; Fulton et al. 1998; Anagnostou et al. 1999). These limitations arise from uncertainties associated with the lack of uniqueness in reflectivity to rainfall intensity transformation, radar system calibration and contamination by ground returns problems, as well as precipitation profile and complex terrain effects. Recent considerations concern the upgrade of WSR-88D systems to include dual-polarization capability, expected to moderate the effect of Z-R variability and radar calibration (Ryzhkov et al. 2005; Bringi et al. 2004; Zrnic and Ryzhkov 1999), while deploying local radar units is an option to fill up critical gaps in the WSR-88D network. Use of small and cost effective X-band radar units for this purpose (e.g., CASA NSF Engineering Research Center) is particularly stressed in cases of regions prone to localized severe weather phenomena, like tornados and flash floods, and over mountainous basins not well covered (due to terrain blockage) by operational weather radar networks. Advances in weather radar technology have led to the development of polarimetric systems that are becoming more suitable to hydrological and hydrometeorological applications. First, Seliga and Bringi (1976, 1978) used the anisotropy information arising from the oblateness of raindrops to estimate rainfall. This information was exploited by producing new parameters such as the differential reflectivity (ZDR) and the differential propagation phase shift (ψDP). The ψDP is a powerful tool for the quantification of rain-path attenuation in short wavelength radar observations (Anagnostou et al. 2006a, b; Park et al. 2005; Matrosov et al. 2005) and for the estimation of precipitation parameters including hydrometeor size distributions (Brandes et al. 2004; Zhang et al. 2001; Vivekanandan et al. 2004; Gorgucci et al. 2000). Due to rain-path attenuation, such shorter wavelengths (X- and C-band) undergo makes longer wavelengths (S-band) more attractive in the quantification of rainfall. Note that even at C-band significant attenuation issues associated with convective storms can occur. Several polarimetric relations for rain rate estimation have been suggested during the last two decades, using ZH, ZDR and the specific propagation differential phase shift KDP (Ryzhkov et al. 2001; Brandes et al. 2001; May et al. 1999; Anagnostou et al. 2004; Matrosov et al. 2002). All these studies have shown that (a) there is an improvement in rainfall estimation if polarimetric radar is used, and (b) polarimetric rainfall estimation techniques are more robust with respect to Drop Size Distribution (DSD) variation than conventional Z-R relations. However, there is still no definitive compromise on the degree of improvement and the choice of the optimal polarimetric relationship (Ryzhkov et al. 2005). Furthermore, improving local flood and flash flood forecasting requires accurate quantitative rainfall measurements at small temporal (minutes) and spatial (hundred of meters to few kilometers) scales. The ability of short wavelength radar (X-band) to monitor precipitation at high spatiotemporal scales has stimulated great interest and support within the hydrologic community. As mentioned earlier in this Section, the use of small size X-band radar units is sought as an approach to fill up critical gaps in operational weather 'radar networks (consisting primarily of Sband, e.g., WSR-88D network in US and C-band radars, e.g., radar networks in Europe). This would be particularly significant for providing high resolution rainfall observations over small scale watersheds, urban areas and mountainous basins not well covered by operational radar networks. A primary disadvantage of X-band frequency is the enhanced rain path attenuation in ZH and ZDR measurements, compared to S-band (and moderately to C-band), including the potential for complete signal loss in cases of signal propagation through more than 10 km paths of high rainfall intensity, on the one hand. On the other hand, power independent parameters such as DP exhibit greater phase change per unit rainfall rate at shorter wavelengths. As a result, the sensitivity of DP to rainfall intensity at X-band can be about three times that of Sband observations. Consequently, X-band frequency offers an increased sensitivity on differential phase-based estimation of weak targets (such as stratiform rain rates) compared to S-band and C-band systems. Furthermore, a radar beam at X-band is associated with greater resolution than the lower frequencies (S-/C-band) for the same antenna size and is less susceptible to side lobe effects. As a result, X-band systems offer mobility and therefore cost efficiency, since they require low power units and small antenna sizes. This Chapter is based on the following three testable hypotheses: (a) The use of differential propagation phase shift can provide accurate estimates of rain-path specific and differential attenuation provided there is no total lose of the transmitted power; (b) attenuation-corrected X-band dual polarization radar measurements offer higher sensitivity compared to lower frequency radar in the estimation of low rain rates, and a comparable accuracy in the estimation of moderate-to-high rainfall rates and DSD parameters; and (c) X-band can provide high spatial and temporal resolution estimates but is limited by range to less than 50 km and up to 120 km in heavy and low-to-moderate rain rates, respectively. This Chapter explores the synergy of rainfall observations from multiple sensors needed to test the above hypotheses. The second part of the Chapter presents the state-of-the-art technology of todays research X-band polarimetric radar systems. Most of these systems are mobile and some are static systems. However, as we discussed earlier, the major drawback in systems is the atmospheric attenuation effect. Anagnostou et al. (2006a), Matrosov et al. (2005) and Park et al. (2005) have shown that ψDP can provide stable estimates of the path specific attenuation at horizontal polarization AH and specific differential attenuation ADP along a radar ray. The last two parts of the Chapter focuses on methods to estimate rainfall from X-band dual-polarization radar systems based on the microphysical properties of rain. Rainfall estimation techniques can be broadly quantified to physical based and empirical techniques. Physical based techniques are mainly power law algorithms where the coefficients have been calculated based on simulations. These algorithms are evaluated based on in-situ disdrometer spectra observations.