Generating Probabilistic Next-Day Severe Weather Forecasts from Convection-Allowing Ensembles Using Random Forests

Abstract

Extracting explicit severe weather forecast guidance from convection-allowing ensembles (CAEs) is challenging since CAEs cannot directly simulate individual severe weather hazards. Currently, CAE-based severe weather probabilities must be inferred from one or more storm-related variables, which may require extensive calibration and/or contain limited information. Machine learning (ML) offers a way to obtain severe weather forecast probabilities from CAEs by relating CAE forecast variables to observed severe weather reports. This paper develops and verifies a random forest (RF)-based ML method for creating day 1 (1200–1200 UTC) severe weather hazard probabilities and categorical outlooks based on 0000 UTC Storm-Scale Ensemble of Opportunity (SSEO) forecast data and observed Storm Prediction Center (SPC) storm reports. RF forecast probabilities are compared against severe weather forecasts from calibrated SSEO 2–5-km updraft helicity (UH) forecasts and SPC convective outlooks issued at 0600 UTC. Continuous RF probabilities routinely have the highest Brier skill scores (BSSs), regardless of whether the forecasts are evaluated over the full domain or regional/seasonal subsets. Even when RF probabilities are truncated at the probability levels issued by the SPC, the RF forecasts often have BSSs better than or comparable to corresponding UH and SPC forecasts. Relative to the UH and SPC forecasts, the RF approach performs best for severe wind and hail prediction during the spring and summer (i.e., March–August). Overall, it is concluded that the RF method presented here provides skillful, reliable CAE-derived severe weather probabilities that may be useful to severe weather forecasters and decision-makers.

Using Machine Learning to Generate Storm-Scale Probabilistic Guidance of Severe Weather Hazards in the Warn-on-Forecast System

“…To accomplish this goal, we trained gradient-boosted classification trees (Friedman 2002;Chen and Guestrin 2016), random forests, and logistic regression models on WoFS forecasts from the 2017-19 Hazardous Weather Testbed Spring Forecasting Experiments (HWT-SFE; Gallo et al 2017) to determine which storms predicted by the WoFS will produce a tornado, severe hail, and/or severe wind report. These three ML algorithms are fairly common and have recently shown success in a variety of meteorological applications (e.g., Mecikalski et al 2015;Erickson et al 2016;Gagne et al 2017;Lagerquist et al 2017;Herman and Schumacher 2018a,b;Burke et al 2020;Loken et al 2019;McGovern et al 2019a,b;Hill et al 2020;Jergensen et al 2020;Steinkruger et al 2020).…”

“…Recent ML studies using real CAM ensemble output for severe weather prediction have been restricted to the next-day (24-36 h) paradigm and producing grid-based guidance (e.g., Gagne et al 2017;Burke et al 2020;Loken et al 2019;Hill et al 2020;Sobash et al 2020). Next-day forecasting methods, however, operate on a larger spatial scale because of the limited intrinsic predictability of storms at those lead times (Lorenz 1969).…”

Forecasting excessive rainfall with random forests and a deterministic convection-allowing model

“…Post-processing methods applied to model output, including machine learning (ML), have shown promise in producing calibrated, probabilistic forecasts of extreme precipitation (e.g., Gagne et al 2014;Scheuerer and Hamill 2015;Herman and Schumacher 2018b;Whan and Schmeits 2018;Loken et al 2019) and other convection-based hazards like tornadoes, hail, and severe wind (e.g., Gagne et al 2017;McGovern et al 2017;Burke et al 2020;Hill et al 2020;Loken et al 2020;Sobash et al 2020;Flora et al 2021). These postprocessing examples have used NWP model forecasts as inputs to random forests (RFs; Breiman 2001) and artificial neural networks to produce both quantitative and probabilistic event-based forecasts of weather hazards (e.g., Herman and Schumacher 2018b;Loken et al 2019). As an example, Herman and Schumacher (2018b) leveraged the National Oceanic and Atmospheric Administration (NOAA) Global Ensemble Forecast System Reforecast (GEFS/R; Hamill et al 2013) dataset to train RFs to probabilistically predict the occurrence of excessive rainfall events over 24-h periods analogous to day-2 and day-3 WPC EROs.…”

“…On the other hand, the application of CAMs with ML comes with two primary limitations: 1) CAMs have limited temporal range, often restricted to less than 60 forecast hours due to computational constraints; and 2) CAMs often undergo development upgrades routinely, which alter their biases. The latter limitation is important because ML models are effective at learning input biases-e.g., high temperature bias during the daytime-when the input biases remain static (Loken et al 2019). These limitations have likely hampered the development of CAM-based ML models for hazard forecasting.…”

“…These limitations have likely hampered the development of CAM-based ML models for hazard forecasting. A recently study conducted by Loken et al (2019) used high-resolution CAM ensemble forecasts as inputs to RF models and forecast accumulated precipitation over forecast hours 12-36, effectively calibrating the CAM ensemble QPFs. They showed that the RFs were able to correct systemic biases in QPF, and the RF forecasts outperformed the dynamic model QPF guidance.…”