Correlations between Topography and Land Cover with Tornado Intensity Using Rapid-Scan Mobile Radar Observations in a Geographic Information System Framework

Despite a wide variety of tornado research that has been performed over the past 50+ years, very little work has been done to investigate the role of geographic surface features on tornadoes. Specifically, observational studies targeting this topic are nominal; and those that have been performed almost ubiquitously examine an isolated event and/or suffer from poor temporal and/or spatial resolution. As such, the impact that topography and surface friction have on tornado behavior is poorly understood. The goal of this work is to begin to fill the knowledge gap that exists around this topic by examining the statistical relationships between tornado intensity and i) topographic features, and/or ii) land cover type (e.g. forest, grassland, water, shrubs, etc.). Radial velocity data acquired from two rapid-scan mobile radars: the Mobile Weather Radar 2005 X-band Phased Array, (MWR-05XP) and the Rapid Scan X-band Polarimetric radar (RaXPol) are used as a proxy for tornado intensity for four case studies: the 2009 Goshen, WY tornado; the 2011 El Reno, OK tornado; the 2013 Carney, OK tornado, and the 2013 Shawnee, OK tornado together comprising a dataset of nearly 400 observations. Additionally, WSR-88D radar data are used to investigate 51 tornado events that occurred in the SE United States.

Radar observations of the tornadoes (latitude/longitude of tornado center, intensity, diameter of the 35 ms^-1 isodop) from the lowest elevation angle data available (ranging from ~50-550 m above radar level) are coupled with digital elevation model (DEM) GIS data and the 2011 national land cover database (NLCD) GIS data using ArcMap. The categorical NLCD data are reclassified with quantitative values of surface roughness lengths as specified by the EPA’s AERSURFACE user guide. Two separate datasets (one for the mobile radar observations and one for the WSR-88Ds) are constructed that couple intensity, elevation, and land cover classification for the location of the tornado center (for elevation) and within the specified 35 ms^-1 tornado radius (for land cover). The high spatio-temporal resolution of the mobile radar dataset (2 – 56 s updates, range gate spacing as low as 30 m) supersedes the resolution of any previous observational study relating tornadoes with topography or land cover in the peer-reviewed literature. The high resolution data allows for improved confidence in the geographic placement of tornado features in the context of the land cover/topography. Statistical correlations, multivariate linear regression models, and bootstrap non-parametric statistics are used to evaluate the quantitative relationships between tornado intensity and both topography and land cover type, and their statistical significance. Preliminary results indicate that there are statistically significant relationships, but these relationships are quite complex and resist generalization.