Three Methods for Comparing Reflectivity Values Detected By Satellite and Ground-Based Radars during Landfalling Tropical Cyclones

Measurements from satellite-based radars provide detailed information about tropical cyclone rainbands while they are over ocean or during landfall. However, the limited swath over which the instrument can detect and time between overpasses means that the entire TC cannot be sampled at the same time. Ground-based radars have their own limitations with the distance over which they can provide meaningful data in the lower troposphere. Mosaicking neighboring radars can allow most of a TC to be detected once over land yet variations among the radars as well as factors such as beam spreading can cause a rainband to be detected differently by multiple radars. Being that data from the satellite- and ground-based radars are derived from different beam widths, scanning strategies, and scan times, how can we best compare the rainband structures of TCs using both sensors? This study uses three methods to compare these data. For method one, we compare values from GPM Ground Validation Data Archive for 4D matched Dual frequency Precipitation Radar (DPR), and Weather Surveillance Radar 1988-Doppler (WSR-88D), and 1 x 1 x 0.5 km gridded WSR at individual stations. Second, we re-grid DPR and mosaicked WSR to a common grid at 5 x 5 km and compare matching grid points at an altitude of 3 km over the entire DPR swath. Third, we take these commonly gridded values and identify objects using reflectivity thresholds between 20 and 40 dBZ and compare their spatial properties. Our case study examines data during two overpasses of Hurricane Laura (2020), the first at 0300 UTC 27 August prior to landfall and the second at 1244UTC on 27 August when two WSR sites that primarily detected Laura during the first overpass were not functional. Results from method one show a difference in stratiform and convective precipitation. WSR site KPOE sampled the edge of Laura where precipitation was mostly stratiform. Mean values of the DPR and WSR were similar according to two-way analysis of variance by ranks tests. Site KLCH was closer to the TC center and sampled the eyewall. Stratiform (convective) regions had dissimilar (similar) mean values. For KSHV which was only sampled during overpass two, mean values were significantly lower than for DPR. Comparisons from method two yield mean error of -0.7 (3.0) and mean absolute error of 2.6 (4.2) dBZ for the first (second) overpass, indicating that data from KLCH and KPOE were more closely matched to DPR values than for KSHV. For the object-based analysis, objects had similar dispersion and closure values during overpass one, with biases in closure being only 1 dBZ higher 20-25 dBZ, and 1 dBZ lower than WSR 35-40 dBZ. During the second overpass, DPR objects were larger and more compact at the same reflectivity value. The bias in closure is > 5 dBZ for DPR at 20-25 dBZ. This reduces to a 1-2 dBZ bias 30-35 dBZ and then rises to 3-4 dBZ for 35-40 dBZ. These results demonstrate the impact of missing WSR data on the south side of Laura’s circulation and suggest that the remaining radar, KSHV, was detecting values much lower compared to DPR. However, the bias is not consistent across different reflectivity levels which could be due to objects located on the edge of the storm or where sharp reflectivity gradients exist near convective regions. These results demonstrate that object-based analyses can provide different and insightful information about the spatial distribution of TC rainbands as compared to more traditional point-based analyses.