Lightning modeling needs for short-range numerical weather prediction

Speaker: Amanda Back (NOAA)

Date: March 28, 2024 @ 3pm

Workman 101

Abstract:

The Rapid Refresh Analysis and Forecast System was the first hourly-updating operational numerical weather prediction capability in the world. Since 2014, the High-Resolution Rapid Refresh (HRRR) has provided hourly near-real-time analyses, and forecasts with up to two days’ lead time, at a 3-km horizontal resolution over the conterminous United States. The “convection-allowing” scale of the model enables greater fidelity for processes including cloud physics, convection, and storm rotation; however, the fine spatial and temporal scales required for explicit thunderstorm electrification remain out-of-reach given operational timeliness requirements. To meet the need for lightning prediction, models lacking explicit electrification schemes diagnose the threat using a combination of kinematic (e.g., updrafts) and microphysical (e.g., ice hydrometeor) model fields relevant to charge buildup. These proxies, however, lack accuracy in placement and intensity compared to forecasts of radar reflectivity for the same storms.

Beyond constituting a hazard that HRRR can predict, lightning often occurs alongside tornadoes, flash floods, and hail, and unlike these is straightforward to remotely sense. Both long-range ground-based RF networks and the geostationary lightning mapper optical sensors on the GOES-R satellites detect lightning in observation-sparse regions. Thus, given a robust, theoretically sound, model-simulated lightning field, lightning observations can be used as “truth” datasets against which to evaluate HRRR’s simulation skill for lightning, as well as to infer skill for other hazards with which lightning correlates.

HRRR owes its predictive skill not just to better-resolved processes, but also to the use of novel observational data. In comparison to systems that initialize two or four forecast runs daily, the hourly updates of the Rapid Refresh systems add value through the assimilation of frequently-updating observational data including measurements from aircraft and weather radars. These sources are rich over much of the US but lack coverage over large bodies of water, over complex terrain (for weather radars), and in many places outside this country. Lightning detections offer unique storm-related information for these regions, but to initialize weather models with lightning data again requires a relationship to be defined between lightning detections and model variables. The currently operational Rapid Refresh systems ingest lightning detections from ground-based networks through a purely empirical relationship with radar reflectivity, while a method, pioneered by the University of Oklahoma, relating satellite-detected lightning flash extent density to graupel mass is included in the prototype next-generation Rapid Refresh Forecast System, which will run at 3 km horizontal resolution over continent-scale domain.. Though beneficial on the whole, these methods have certain drawbacks and limitations; the assimilation of lightning data into convection-allowing models is an active area of research for which many open questions remain.