A machine-learning approach to thunderstorm forecasting through post-processing of simulation data

Kianusch Vahid Yousefnia, Tobias Bölle, Isabella Zöbisch, Thomas Gerz

arXiv preprint · 15 March 2023

In one sentence. SALAMA, a feedforward network post-processing convection-resolving NWP fields, gives calibrated thunderstorm probabilities with skill beyond NWP reflectivity up to eleven hours lead.

Abstract

Thunderstorms pose a major hazard to society and economy, which calls for reliable thunderstorm forecasts. In this work, we introduce a Signature-based Approach of identifying Lightning Activity using MAchine learning (SALAMA), a feedforward neural network model for identifying thunderstorm occurrence in numerical weather prediction (NWP) data. The model is trained on convection-resolving ensemble forecasts over Central Europe and lightning observations. Given only a set of pixel-wise input parameters that are extracted from NWP data and related to thunderstorm development, SALAMA infers the probability of thunderstorm occurrence in a reliably calibrated manner. For lead times up to eleven hours, we find a forecast skill superior to classification based only on NWP reflectivity. Varying the spatiotemporal criteria by which we associate lightning observations with NWP data, we show that the time scale for skillful thunderstorm predictions increases linearly with the spatial scale of the forecast.

Labels

Labels and summary are generated by Claude and reviewed weekly. Wrong? Edit this entry.