Search for: All records

Abstract We introduce the National Science Foundation (NSF) AI Institute for Research on Trustworthy AI in Weather, Climate, and Coastal Oceanography (AI2ES). This AI institute was funded in 2020 as part of a new initiative from the NSF to advance foundational AI research across a wide variety of domains. To date AI2ES is the only NSF AI institute focusing on environmental science applications. Our institute focuses on developing trustworthy AI methods for weather, climate, and coastal hazards. The AI methods will revolutionize our understanding and prediction of high-impact atmospheric and ocean science phenomena and will be utilized by diverse, professional user groups to reduce risks to society. In addition, we are creating novel educational paths, including a new degree program at a community college serving underrepresented minorities, to improve workforce diversity for both AI and environmental science. 

We present an overview of recent work on using artificial intelligence (AI)/machine learning (ML) techniques for forecasting convective weather and its associated hazards, including tornadoes, hail, wind, and lightning. These high-impact phenomena globally cause both massive property damage and loss of life, yet they are very challenging to forecast. Given the recent explosion in developing ML techniques across the weather spectrum and the fact that the skillful prediction of convective weather has immediate societal benefits, we present a thorough review of the current state of the art in AI and ML techniques for convective hazards. Our review includes both traditional approaches, including support vector machines and decision trees, as well as deep learning approaches. We highlight the challenges in developing ML approaches to forecast these phenomena across a variety of spatial and temporal scales. We end with a discussion of promising areas of future work for ML for convective weather, including a discussion of the need to create trustworthy AI forecasts that can be used for forecasters in real time and the need for active cross-sector collaboration on testbeds to validate ML methods in operational situations.

We provide an overview of recent machine learning research in predicting hazards from thunderstorms, specifically looking at lightning, wind, hail, and tornadoes. These hazards kill people worldwide and also destroy property and livestock. Improving the prediction of these events in both the local space as well as globally can save lives and property. By providing this review, we aim to spur additional research into developing machine learning approaches for convective hazard prediction.

 

Ensemble Kalman filter (EnKF) analyses of the storms associated with the 8 May 2017 Colorado severe hail event using either the Milbrandt and Yau (MY) or the NSSL double-moment bulk microphysics scheme in the forecast model are evaluated. With each scheme, two experiments are conducted in which the reflectivity ( Z) observations update in addition to dynamic and thermodynamic variables: 1) only the hydrometeor mixing ratios or 2) all microphysical variables. With fewer microphysical variables directly constrained by the Z observations, only updating hydrometeor mixing ratios causes the forecast error covariance structure to become unreliable, and results in larger errors in the analysis. Experiments that update all microphysical variables produce analyses with the lowest Z root-mean-square innovations; however, comparing the estimated hail size against hydrometeor classification algorithm output suggests that further constraint from observations is needed to more accurately estimate surface hail size. Ensemble correlation analyses are performed to determine the impact of hail growth assumptions in the MY and NSSL schemes on the forecast error covariance between microphysical and thermodynamic variables. In the MY scheme, Z is negatively correlated with updraft intensity because the strong updrafts produce abundant small hail aloft. The NSSL scheme predicts the growth of large hail aloft; consequently, Z is positively correlated with storm updraft intensity and hail state variables. Hail production processes are also shown to alter the background error covariance for liquid and frozen hydrometeor species. Results in this study suggest that EnKF analyses are sensitive to the choice of MP scheme (e.g., the treatment of hail growth processes).

 

An ensemble of 10 forecasts is produced for the 20 May 2013 Newcastle–Moore EF5 tornado and its parent supercell using a horizontal grid spacing of 50 m, nested within ensemble forecasts with 500-m horizontal grid spacing initialized via ensemble Kalman filter data assimilation of surface and radar observations. Tornadic circulations are predicted in all members, though the intensity, track, and longevity of the predicted tornado vary substantially among members. Overall, tornadoes in the ensemble forecasts persisted longer and moved to the northeast faster than the observed tornado. In total, 8 of the 10 ensemble members produce tornadoes with winds corresponding to EF2 intensity or greater, with maximum instantaneous near-surface horizontal wind speeds of up to 130 m s⁻¹and pressure drops of up to 120 hPa; values similar to those reported in observational studies of intense tornadoes. The predicted intense tornadoes all acquire well-defined two-cell vortex structure, and exhibit features common in observed tornadic storms, including a weak-echo notch and low reflectivity within the mesocyclone. Ensemble-based probabilistic tornado forecasts based upon near-surface wind and/or vorticity fields at 10 m above the surface produce skillful forecasts of the tornado in terms of area under the relative operating characteristic curve, with probability swaths extending along and to the northeast of the observed tornado path. When probabilistic swaths of 0–3- and 2–5-km updraft helicity are compared to the swath of wind at 10 m above the surface exceeding 29 m s⁻¹, a slight northwestward bias is present, although the pathlength, orientation, and the placement of minima and maxima show very strong agreement.

 

Day-ahead (20–22 h) 3-km grid spacing convection-allowing model forecasts are performed for a severe hail event that occurred in Denver, Colorado, on 8 May 2017 using six different multimoment microphysics (MP) schemes including: the Milbrandt–Yau double-moment (MY2), Thompson (THO), NSSL double-moment (NSSL), Morrison double-moment graupel (MOR-G) and hail (MOR-H), and Predicted Particle Properties (P3) schemes. Hail size forecasts diagnosed using the Thompson hail algorithm and storm surrogates predict hail coverage. For this case hail forecasts predict the coverage of hail with a high level of skill but underpredict hail size. The storm surrogate updraft helicity predicts the coverage of severe hail with the most skill for this case. Model data are analyzed to assess the effects of microphysical treatments related to rimed ice. THO uses diagnostic equations to increase the size of graupel within the hail core. MOR-G and MOR-H predict small rimed ice aloft; excessive size sorting and increased fall speeds cause MOR-H to predict more and larger surface hail than MOR-G. The MY2 and NSSL schemes predict large, dense rimed ice particles because both schemes predict separate hail and graupel categories. The NSSL scheme predicts relatively little hail for this case; however, the hail size forecast qualitatively improves when the maximum size of both hail and graupel is considered. The single ice category P3 scheme only predicts dense hail near the surface while above the melting layer large concentrations of low-density ice dominate.

 

Hail forecast evaluations provide important insight into microphysical treatment of rimed ice. In this study we evaluate explicit 0–90-min EnKF-based storm-scale (500-m horizontal grid spacing) hail forecasts for a severe weather event that occurred in Oklahoma on 19 May 2013. Forecast ensembles are run using three different bulk microphysics (MP) schemes: the Milbrandt–Yau double-moment scheme (MY2), the Milbrandt–Yau triple-moment scheme (MY3), and the NSSL variable density-rimed ice double-moment scheme (NSSL). Output from a hydrometeor classification algorithm is used to verify surface hail size forecasts. All three schemes produce forecasts that predict the coverage of severe surface hail with moderate to high skill, but exhibit less skill at predicting significant severe hail coverage. A microphysical budget analysis is conducted to better understand hail growth processes in all three schemes. The NSSL scheme uses two-variable density-rimed ice categories to create large hailstones from dense, wet growth graupel particles; however, it is noted the scheme underestimates the coverage of significant severe hail. Both the MY2 and MY3 schemes produce many small hailstones aloft from unrimed, frozen raindrops; in the melting layer, hailstones become much larger than observations because of the excessive accretion of water. The results of this work highlight the importance of using a MP scheme that realistically models microphysical processes.