Parameterization of Lightning Flashes in the NASA GEOS Model Using GLM Data and Machine Learning

Lightning is the primary free-troposphere source of nitrogen oxide (NOx), and a key factor determining background ozone (O3) concentrations and the oxidation capacity of the atmosphere. To determine O3 production accurately, the NOx source from lightning must be consistent in time and space with the model’s deep convective mixing to ensure the proper balance of NOx from lightning and O3 precursors lofted from the boundary layer. Thus, NO_x production in the Goddard Earth Observing System (GEOS) suite of models is parameterized in terms of meteorological fields from GEOS constrained by satellite-based estimates of lightning climatology from the Optical Transient Detector and Lightning Imaging Sensor (OTD-LIS) . Since 2017, estimates of lightning products including flash count, group count, flash multiplicity, and flash optical energy have become available from the GOES-R series of Geostationary Lightning Mappers (GLMs). When positioned at ~75° W and ~137° W, the coverage region for two GLMs includes nearly all the Americas equatorward of 54°. The GEOS system includes a module containing multiple flash rate and LNO_x production schemes in support of chemical modeling. In this study, the GEOS system was constrained using re-analysis fields to reproduce observed meteorology during the July 2019 – June 2020 time period. In this presentation, we will examine the relationship between GEOS thermodynamic and cloud microphysical variables and GLM products including flash count, group count, flash multiplicity, and flash optical energy. We will also show preliminary results from a scheme that uses machine learning (ML) to obtain flash rates consistent with the GEOS fields and GLM products. Finally, we will discuss the additional issues that must be considered when using the schemes to predict mean NO_x production per flash. The ultimate goal is to create a robust parameterization that uses ML analysis (decision trees and neural networks) to obtain new parameterizations for flash rates and NOx production per flash over continental and marine locations in the GEOS atmospheric composition models.