Search for: All records

Raindrop size distributions (DSD) and rain rate have been estimated from polarimetric radar data using different approaches with the accuracy depending on the errors both in the radar measurements and the estimation methods. Herein, a deep neural network (DNN) technique was utilized to improve the estimation of the DSD and rain rate by mitigating these errors. The performance of this approach was evaluated using measurements from a two-dimensional video disdrometer (2DVD) at the Kessler Atmospheric and Ecological Field Station in Oklahoma as ground truth with the results compared against conventional estimation methods for the period 2006–17. Physical parameters (mass-/volume-weighted diameter and liquid water content), rain rate, and polarimetric radar variables (including radar reflectivity and differential reflectivity) were obtained from the DSD data. Three methods—physics-based inversion, empirical formula, and DNN—were applied to two different temporal domains (instantaneous and rain-event average) with three diverse error assumptions (fitting, measurement, and model errors). The DSD retrievals and rain estimates from 18 cases were evaluated by calculating the bias and root-mean-squared error (RMSE). DNN produced the best performance for most cases, with up to a 5% reduction in RMSE when model errors existed. DSD and rain estimated from a nearby polarimetric radar using the empirical and DNN methods were well correlated with the disdrometer observations; the rain-rate estimate bias of the DNN was significantly reduced (3.3% in DNN vs 50.1% in empirical). These results suggest that DNN has advantages over the physics-based and empirical methods in retrieving rain microphysics from radar observations.

 

Abstract The time preceding supercell tornadogenesis and tornadogenesis “failure” has been studied extensively to identify differing attributes related to tornado production or lack thereof. Studies from the Verification of the Origins of Rotation in Tornadoes Experiment (VORTEX) found that air in the rear-flank downdraft (RFD) regions of non- and weakly tornadic supercells had different near-surface thermodynamic characteristics than that in strongly tornadic supercells. Subsequently, it was proposed that microphysical processes are likely to have an impact on the resulting thermodynamics of the near-surface RFD region. One way to view proxies to microphysical features, namely drop size distributions (DSDs), is through use of polarimetric radar data. Studies from the second VORTEX used data from dual-polarization radars to provide evidence of different DSDs in the hook echoes of tornadic and non-tornadic supercells. However, radar-based studies during these projects were limited to a small number of cases preventing result generalizations. This study compiles 68 tornadic and 62 non-tornadic supercells using Weather Surveillance Radar–1988 Doppler (WSR-88D) data to analyze changes in polarimetric radar variables leading up to, and at, tornadogenesis and tornadogenesis failure. Case types generally did not show notable hook echo differences in variables between sets, but did show spatial hook echo quadrant DSD differences. Consistent with past studies, differential radar reflectivity factor (Z DR ) generally decreased leading up to tornadogenesis and tornadogenesis failure; in both sets, estimated total number concentration increased during the same times. Relationships between DSDs and the near-storm environment, and implications of results for nowcasting tornadogenesis, also are discussed. 

Abstract Tornadic supercells moved across parts of Oklahoma on the afternoon and evening of 9 May 2016. One such supercell, while producing a long-lived tornado, was observed by nearby WSR-88D radars to contain a strong anticyclonic velocity couplet on the lowest elevation angle. This couplet was located in a very atypical position relative to the ongoing cyclonic tornado and to the supercell’s updraft. A storm survey team identified damage near where this couplet occurred, and, in the absence of evidence refuting otherwise, the damage was thought to have been produced by an anticyclonic tornado. However, such a tornado was not seen in near-ground, high-resolution radar data from a much closer, rapid-scan, mobile radar. Rather, an elongated velocity couplet was observed only at higher elevation angles at altitudes similar to those at which the WSR-88D radars observed the strong couplet. This paper examines observations from two WSR-88D radars and a mobile radar from which it is argued that the anticyclonic couplet (and a similar one ~10 min later) were actually quasi-horizontal vortices centered ~1–1.5 km AGL. The benefits of having data from a radar much closer to the convective storm being sampled (e.g., better spatial resolution and near-ground data coverage) and providing more rapid volume updates are readily apparent. An analysis of these additional radar data provides strong, but not irrefutable, evidence that the anticyclonic tornado that may be inferred from WSR-88D data did not exist; consequently, upon discussions with the National Weather Service, it was not included in Storm Data. 

Abstract This study presents an investigation into relationships among topographic elevation, surface land cover, and tornado intensity using rapid scan, mobile Doppler radar observations of four tornadoes from the U.S. Central Plains. High spatiotemporal resolution observations of tornadic vortex signatures from the radar’s lowest elevation angle data (in most cases ranging from ~100 to 350 m above ground level) are coupled with digital elevation model (DEM) and 2011 National Land Cover Database (NLCD) data using a geographic information system (GIS). The relationships between 1) tornado intensity and topographic elevation or surface roughness and 2) changes in tornado intensity and changes in topographic elevation or surface roughness are investigated qualitatively, and statistical relationships are quantified and analyzed using a bootstrap permutation method for individual case studies and all cases collectively. Results suggest that there are statistically significant relationships for individual cases, but the relationships defy generalization and are different on a case-by-case basis, which may imply that they are coincidental, indicating a null correlation.