Estimation of spatial correlation of rain drop size distribution parameters and rain rates using NASA's S-band polarimetric radar and 2D video disdrometer network: Two case studies from MC3E

The spatial correlation of DSD parameters and rain rate at distances varying from <0.5 km to ~5-10 km is important in understanding their spatial variability for example, as related to down-scaling methodologies/modeling, to estimate the “point-to-area” variance when comparing gage/disdrometer data to radar retrievals, and application to non-uniform beam filling (NUBF) “corrections” for satellite-borne radar data which necessarily represent large pixel sizes (~4 km: TRMM and future GPM). It even applies to coarse-scale radar estimates, e.g., at long ranges where the radar beam becomes broad, or even grid-averaged products. While the spatial correlation of R has been studied extensively with dense gage networks, recently it has been shown that polarimetric radar data obtained at high spatial (close ranges<30 km) and high time resolution (PPI/RHI scan cycles <40s) offers a distinct advantage in estimating the spatial correlation function over fixed network of gages/disdrometers. On several occasions during the MC3E campaign in northern Oklahoma, NASA's S-band polarimetric radar, NPOL, was made to perform repeated PPI scans over six 2D video disdrometer (2DVD) sites, located 20 to 30 km from the radar. The scans were repeated approximately every 40 seconds. We consider here two cases, one a rapidly evolving multi-cell rain event (with large drops) on 24 April 2011 and the second a somewhat more uniform rain event on 11 May 2011.

The repeated PPI scans were used to determine the spatial correlations of two of the main DSD parameters, namely, D0 and log(Nw) as well as rain rate (R). The correlations were determined along each radial over the whole azimuthal range of the PPI sector scans. The spatial correlations show azimuthal dependence, as expected, especially for the highly convective 24 April 2011 event. The time series at every polar pixel were constructed which in turn were used to compute the Pearson correlation coefficient for various separations (radially-outward from an a priori fixed range location set at 20 km). Further, we construct the CDF along each range circle starting at the reference range of 20 km, and compute the 10th, 50th (median) and 90th percentile values. This is repeated at range increments of 150 m (radar gate spacing). Such an approach gives a “pseudo”-1D spatial correlation at the same time giving an estimate of its cross-beam (azimuthal) variability. For the long duration 11th May 2011 event (over 4 hours), the spatial correlation of the DSD parameters and rain rate showed good agreement with 2DVD-based spatial correlations. The 24 April event was too short (<80 mins) and highly convective to yield “stable” spatial correlations from the 2DVD network but the radar-based correlations were quite stable due to large number of high resolution pixels available, further showing the advantage of using radar data. Note that normal PPI updates of 5 min (eg, WSR-88D) is too large to estimate the spatial correlations at high spatial resolution.

As part of the aforementioned scan sequence, the NPOL also made repeated RHI scans (updates <40 s) along one azimuth centered over the 2DVD network. These scans were used to determine the vertical correlations of the DSD parameters as well as the liquid water content, for stratiform versus convective rain.