Soil Moisture and Tornadogenesis in the United States

This dataset was created to explore how soil moisture may influence where tornadoes form. It includes nearly 6,000 tornadoes that occurred across the contiguous U.S. from 2015 to 2019. The location of tornadogenesis is obtained from the Storm Prediction Center ONETOR database, which uses official NCEI Storm Data Reports to stitch together reports of tornadoes that cross county lines into a single tornado event. Using this dataset eliminates the possibility of falsely identifying a new tornadogenesis point when an ongoing tornado simply moves into a new county. We utilized all tornadoes that occurred in CONUS from 2015-2019 to generate this dataset, without respect to parent storm mode (e.g., supercell, quasi-linear convective systems, tropical cyclones). This resulted in 5854 tornadoes between January 2015 and December 2019 that will be evaluated in this study. More details on the tornado data are available from: https://www.spc.noaa.gov/wcm/data/SPC_severe_database_description.pdf Observed soil moisture data were acquired from the National Soil Moisture network (nationalsoilmoisture.com; NSM). NSM blends in-situ soil moisture measurements from >1000 stations across the CONUS with precipitation and soil texture data to objectively analyze soil water content at regular, spatially gridded intervals. Values are obtained daily and based upon the local conditions present at 7:00 AM local time. NSM provides data at three soil depths (5 cm, 20 cm, and 50 cm) at a 4 km spatial resolution. More details on the quality control and standardization of the soil moisture measurements and the Regression Kriging method used to spatially interpolate soil moisture are described in Zhao et al. (2020). The location of tornadogenesis, given by start latitude and longitude in the ONETOR dataset, was used to find the closest 4-km soil moisture grid cell. Values for each soil moisture depth on the day the tornado occurred were extracted from the NSM dataset so that each tornado record from 2015 through 2019 had a soil moisture percentile associated with the initiation point. Because the soil moisture observations are acquired at 7:00 local time, the values generally represent the pre-storm land surface conditions, which is desirable. However, if rain occurred between 7:00 AM local time and the time of tornadogenesis, this additional moisture will not be accounted for. While this may be a source for some error, particularly in the upper-most soil moisture level of 5 cm, the effect on our results are likely minimal given the large sample size. Because the volumetric water content (VWC) varies as a function of soil properties, local hydroclimate and seasonal variations, soil moisture data are converted from VWC – which is an absolute quantity – to climatological percentiles. This conversion enables standardization of the observations across large regions that have different soil types and climates (Ford et al., 2016). Soil moisture percentiles describe how wet or dry the soil is compared to its usual range for a specific location and time of year. They are calculated by comparing each measurement to the period of record at that location and at that time of year following the approach of Ford et al. (2016). It is necessary to standardize the soil moisture data to compare relationships over space (e.g., Arizona to Florida) and over time (e.g., January versus July), and to account for variations in physical characteristics that will affect the soil moisture content values. The soil moisture percentiles associated with each tornado formation location are recorded in the "SM_percentile" column.