H200-0004
Retrieving DSD moments from GPM-DPR: A simulation study based on the full DSD spectra characterized by the Generalized Gamma model
Retrieving DSD moments from GPM-DPR: A simulation study based on the full DSD spectra characterized by the Generalized Gamma model
Wednesday, 16 December 2020
Poster
Abstract:
Recently, a method to retrieve rain drop size distribution (DSD) moments from X-band dual polarization radar measurements has been developed using copolar reflectivity, differential reflectivity and specific attenuation. Two reference moments are retrieved first, followed by reconstructing the DSDs and calculating other moments using a generalized gamma model to represent the underlying shape, h(x), corresponding to the pair of chosen reference moments. Here we explore a similar approach but the retrieval method in this study uses dual-frequency radar measurements. In the case of GPM-DPR, the two frequencies are 13.8 and 35 GHz, and their products include attenuation corrected reflecivities at both frequencies (ZKu and ZKa) and the specific attenuation (kKu, kKa, if available). Our approach is to use these products to determine two reference moments, namely M3 and M6 representing the third and the sixth moments respectively. Then, as with the polarimetric radar retrievals, we use the most probable h(x) to reconstruct the full DSD spectra, from which other moments are calculated. The best two DPR products for estimating M3 and M6 appear to be AKa and ZKu. Figure 1 shows the retrievals versus the ‘true’ moments. 2930 three-minute DSDs were used for the (T-matrix) scattering calculations at Ku and Ka bands but only cases with AKa > 0.5 dB/km were chosen. For the retrieved moments, the integration was performed only up to 6 mm drop diameter.
The [1:1] line is included in Fig. 1. Even with just two DPR products as input the results seem promising, although the lower order moments show somewhat more scatter, especially the zeroth moment, M0. We will quantify the retrieval errors, and additionally examine the stability of h(x). For the latter, data from (i) Greeley, Colorado, (ii) Huntsville, Alabama, and (iii) Wallops, Virginia, will be used. Finally, a GPM overpass case over Huntsville, during a widespread rain event on 11 April 2016, will be considered as an initial test case.
The [1:1] line is included in Fig. 1. Even with just two DPR products as input the results seem promising, although the lower order moments show somewhat more scatter, especially the zeroth moment, M0. We will quantify the retrieval errors, and additionally examine the stability of h(x). For the latter, data from (i) Greeley, Colorado, (ii) Huntsville, Alabama, and (iii) Wallops, Virginia, will be used. Finally, a GPM overpass case over Huntsville, during a widespread rain event on 11 April 2016, will be considered as an initial test case.