Retrieval of Raindrop Size Distribution Parameters from Dual-Polarization Radar Measurements

One of the main goals of the National Aeronautics and Space Administration (NASA) Global Precipitation Measurement (GPM) mission is to retrieve the parametric form of the raindrop size distribution (DSD) through its Dual-frequency Precipitation Radar (DPR) on board GPM core satellite. The GPM mission adopted a specific form of the three-parameter gamma model DSD but there is a lack of information to obtain the three unknown parameters of the gamma DSD with the dual frequency reflectivity measurements. At this stage, the shape parameter is set to 2 in combined radar-radiometer algorithm and set to 3 in DPR algorithm. The normalized intercept parameter, N_w, and the mean mass diameter, D_mass, are the other two parameters of the gamma model DSD and are the standard outputs from the DPR algorithm. The retrieval of D_mass within ±0.5 mm is one of the Level One (L1) requirements of the GPM mission. The accuracy of N_w and D_mass retrieval is determined through a comparative study of ground based and space borne products. While the ground based products are often considered as a reference, they inherit their have their own errors. Both N_w and D_mass rely on empirical relationships that are function of horizontal (Z_H) and differential (Z_DR) reflectivity. The D_mass-Z_DR and N_w-Z_H-D_mass relationships were derived using the sequential intensity filtering technique (SIFT, Lee and Zawadzki, 2005) utilizing the two-dimensional Video Disdrometer (2DVD) and Parsivel² disdrometer (P2) database collected during GPM Ground Validation (GV) field campaigns. These relationships were then applied to the polarimetric observables of selected operational and research radar across the United States and elsewhere including NASA’s S-band dual-polarization radar (NPOL).

This study investigates the accuracy of the D_mass-Z_DR and N_w-Z_H-D_mass relationships and their sensitivity to the climate regime and the choice of disdrometer, as well as the accuracy of the DSD parameters estimation from radar measurements. The accuracy of D_mass-Z_DR and N_W-Z_H-D_mass relationship was evaluated by comparing D_mass and logN_w that was either directly calculated from 2DVD or retrieved from 2DVD based-Z_DR and Z_H-D_mass measurements. Good agreement was evident for both parameters with absolute bias of 0.12-13 mm and 0.06-0.07 m^-3 mm^-1 for D_mass and logN_w, respectively. The D_mass and N_w estimated from NPOL-based ZDR and Z_H-D_mass measurements were in a reasonable agreement with those that were directly measured by 2DVDs. The D_mass estimation as compared to the 2DVD generally satisfied the L1 requirements with an absolute bias ranging between 0.30 and 0.40 mm for the most sites. A number of sites exhibited larger error due to ground clutter and time-height ambiguity. While not defined in terms of L1 requirement limits, LogN_w had a wider range of values with respect to the 2DVD measurements (2-4.5 m^-3 mm^-1), but was within acceptable range of 0.5-6 m^-3 mm^-1. This was partly due to the fact that NPOL measured Z_H-Z_DR values outside the observational Z_H-Z_DR envelope of the 2DVD. These outliers were found at high Z_DR-low Z_H and low Z_DR-high Z_H regimes and did not necessarily occur in a particular storm or segment of the storm. A sensitivity study was performed by both including and excluding these samples showed that the high Z_DR samples generally produce an overestimation D_mass, and underestimation of N_w, while vice versa was true for the low Z_DR samples.

Considering the sensitivity of D_mass-Z_DR relationship to the climate regime, three years of P2 data from a tropical site (Kwajalein - 8.8˚N, 163.7˚E) were analyzed and compared to two mid-latitude GV field campaigns, IFloodS and OLYMPEx representing mid-continent springtime rainfall and coastal orographic effect through pre- and post-frontal rainfall, respectively. Orographic lifting resulted in an abundance of small drops shifting the probability distribution of logNw to higher values with respect to both the Kwajalein and IFloodS datasets. The use of OLYMPEx D_mass-Z_DR relationship resulted in absolute bias of 0.17 mm in D_mass when it was applied to the Kwajalein dataset. This was a higher absolute bias with respect to the use of Kwajalein or IFloodS D_mass-Z_DR relationships for the other datasets.

Considering the sensitivity of D_mass estimation to the instrument type the co-located 2DVD and P2 measurements from IFloodS and OLYMPEx were employed to derive D_mass-Z_DR relationships from coincident datasets. The difference in Dmass based on 2DVD and P2 D_mass-Z_DR relationships remained within ± 0.2 mm until Z_DR of 1.6 dB and 2.6 dB for IFloodS and OLYMPEx, respectively. The absolute bias was 0.10 mm or lower in D_mass indicating a low sensitivity to the instrument type.

Reference:

Lee, G.W. and Zawadzki, I.: Variability of Drop Size Distributions: Noise and Noise Filtering in Disdrometric Data, 2005, J. Appl. Meteor., 44, 634-652.