Date Available


Year of Publication


Degree Name

Master of Arts (MA)

Document Type

Master's Thesis


Arts and Sciences



First Advisor

Dr. Tony Stallins

Second Advisor

Dr. Liang Liang


Severe thunderstorm warnings (SVTs) are released by meteorologists in the local forecast offices of the National Weather Service (NWS). These warnings are issued with the intent of alerting areas in the path of severe thunderstorms that human and property risk are elevated, and that appropriate precautionary measures should be taken. However, studies have shown that the spatial distribution of severe storm warnings demonstrates bias. Greater numbers of severe thunderstorm warnings sometimes are issued where population is denser. By contrast, less populated areas may be underwarned. To investigate the spatial patterns of these biases for the central and southeastern United States, geographically weighted regression was implemented on a set of demographic and land cover descriptors to ascertain their patterns of spatial association with counts of National Weather Service severe thunderstorm warnings. GWR was performed for each our independent variables (total population, median income, and percent impervious land cover) and for all three of these variables as a group. Global R2 values indicate that each individual variable as well as all three collectively explain approximately 60% of the geographical variation in severe thunderstorm warning counts. Local R2 increased in the vicinity of several urban regions, notably Atlanta, Washington, D.C., St. Louis, and Nashville. However, the independent variables did not exhibit the same spatial patterning of R2. Some cities had high local R2 for all variables. Other cities exhibited high local R2 for only one or two of these independent variables. Median income had the highest local R2 values overall. Standardized residuals confirmed significant differences among several NWS forecast offices in the number and pattern of severe thunderstorm warnings. Overall, approximately half of the influences on the distribution of severe thunderstorm warnings across the study area are related to underlying land cover and demographics. Future studies may find it productive to investigate the extent to which the spatial bias mapped in this study is an artifact of forecast culture, background thunderstorm regime, or a product of urban anthropogenic weather modification.