Trend-based scaling for high-resolution lightning in climate projections

New geostationary satellites, together with ground networks, now provide high-resolution, continuous lightning observations, offering unprecedented insights into lightning activity across vast areas of the globe. In contrast, global climate models (GCMs) lack the spatial resolution and physical processes required to simulate lightning directly, leading to the need for parameterizations and scaling methods. In this study, we present a novel trend-based scaling approach that bridges the gap between coarse-resolution GCM output and high-resolution lightning flash rates to improve projections of lightning activity by the end of the century. The scaling method employs machine learning techniques to identify the atmospheric parameters that best reproduce observed current lightning activity, which are then combined with coarser GCM trends (individually for each quantile of the distribution) to project future lightning changes.

Italy was selected as a case study, using the past 15 years of lightning observations from the LINET network to identify lightning predictors among atmospheric parameters from the ERA5 reanalysis. The best predictors identified include a combination of convective available potential energy, relative humidity, temperature gradients, wind velocity and shear, geopotential height, and freezing level. This model accurately reproduces the spatial distribution and temporal variability of current lightning activity. Trend scaling from multiple future climate scenarios was then applied using CMIP6 projections to evaluate changes in lightning activity across different regions and time periods. 

Our results show that trend-based scaling significantly improves the spatial distribution and intensity of projected lightning flash rates compared to traditional parameterizations. This work provides a practical framework for integrating lightning projections into climate impact studies, enhancing the reliability of lightning future changes under various climate scenarios. The main advantage of the proposed method is that it can be applied to reanalysis datasets of any resolution, offering a flexible tool for assessing lightning-related risks in a warming world.