Machine learning classification of significant tornadoes and hail in the U.S. using ERA5 proximity soundings

Abstract

Previous studies have identified environmental characteristics that skillfully discriminate between severe and significant-severe weather events, but they have largely been limited by sample size and/or population of predictor variables. Given the heightened societal impacts of significant-severe weather, this topic was revisited using over 150 000 ERA5 reanalysis-derived vertical profiles extracted at the grid-point nearest—and just prior to—tornado and hail reports during the period 1996–2019. Profiles were quality-controlled and used to calculate 84 variables. Several machine learning classification algorithms were trained, tested, and cross-validated on these data to assess skill in predicting severe or significant-severe reports for tornadoes and hail. Random forest classification outperformed all tested methods as measured by cross-validated critical success index scores and area under the receiver operating characteristic curve values. In addition, random forest classification was found to be more reliable than other methods and exhibited negligible frequency bias. The top three most important random forest classification variables for tornadoes were wind speed at 500 hPa, wind speed at 850 hPa, and 0–500-m storm-relative helicity. For hail, storm-relative helicity in the 3–6 km and -10 to -30 °C layers, along with 0–6-km bulk wind shear, were found to be most important. A game theoretic approach was used to help explain the output of the random forest classifiers and establish critical feature thresholds for operational nowcasting and forecasting. A use case of spatial applicability of the random forest model is also presented, demonstrating the potential utility for operational forecasting. Overall, this research supports a growing number of weather and climate studies finding admirable skill in random forest classification applications.

Reconstruction of Violent Tornado Environments in Europe: High‐Resolution Dynamical Downscaling of ERA5

“…To derive convective parameters from vertical profiles of pressure, altitude, temperature, humidity, U and V from ERA5 and WRF we used the thundeR R language package (Taszarek et al., 2021). We calculate parameters that are commonly used in the operational forecasting and climatological evaluations of significant tornadoes (Brooks et al., 2003; Grams et al., 2012; Gensini et al., 2021; Hua & Anderson‐Frey, 2021; Ingrosso et al., 2020; R. L. Thompson et al., 2003, 2012, 2013; Taszarek, Allen, Púčik, et al., 2020). These include: convective available potential energy (CAPE), convective inhibition (CIN), 0–500 m mean mixing ratio (MIXR), lifted condensation level (LCL), 0–1 km vertical wind shear (S01), 0–6 km vertical wind shear (S06), 0–500 m storm‐relative helicity (SRH), and the significant tornado parameter (STP; updated formula from Coffer et al.…”

“…To derive convective parameters from vertical profiles of pressure, altitude, temperature, humidity, U and V from ERA5 and WRF we used the thundeR R language package (Taszarek et al, 2021). We calculate parameters that are commonly used in the operational forecasting and climatological evaluations of significant tornadoes (Brooks et al, 2003;Grams et al, 2012;Gensini et al, 2021;Hua & Anderson-Frey, 2021;Ingrosso et al, 2020;R. L. Thompson et al, 2003R.…”

Why we need to focus on developing ethical, responsible, and trustworthy artificial intelligence approaches for environmental science

“…Artificial intelligence (AI) and machine learning (ML) have recently exploded in popularity for a wide variety of environmental science applications (e.g., McGovern et al, 2019; Reichstein et al, 2019; Gagne et al, 2020; Gensini et al, 2021; Hill and Schumacher, 2021; Lagerquist et al, 2021; Schumacher et al, 2021). Like other fields, environmental scientists are seeking to use AI/ML to build a linkage from raw data, such as satellite imagery and climate models, to actionable decisions.…”

Supercell convective environments in Spain based on ERA5: hail and non-hail differences

“…Following the methodology of Calvo-Sancho and Martín (2021) and Gensini et al (2021), supercell soundings for SP-HAIL and SP-NONHAIL events are built for t 0 . Each vertical profile is computed from ERA5 using the nearest grid point to the supercell location at t 0 .…”

“…Each vertical profile is computed from ERA5 using the nearest grid point to the supercell location at t 0 . A quality control is carried out to remove any sounding related to convective boundary propagation (Brooks et al, 2003(Brooks et al, , 2007Gensini et al, 2021). Accordingly, each vertical profile must record a non-zero most-unstable convective available potential energy (MUCAPE) and mixed-layer level of free convection (MLLFC) to be included in the study.…”