Tools for predicting rainfall from lightning records: events identification and rain prediction using a Bayesian hierarchical model

TL;DR

This paper introduces a Bayesian hierarchical model to predict rainfall from lightning data, identifying storm events and estimating rainfall volume, with application to storms in Central Italy.

Contribution

The paper develops a novel statistical protocol combining event detection and Bayesian modeling for rainfall prediction using lightning data, incorporating spatial and temporal dependencies.

Findings

Effective identification of rainy events using scan statistics.

Accurate prediction of 15- and 30-minute rainfall at unobserved locations.

Model captures storm propagation and spatial dependence.

Abstract

We propose a new statistical protocol for the estimation of precipitation using lightning data. We first identify rainy events using a scan statistics, then we estimate Rainfall Lighting Ratio (RLR) to convert lightning number into rain volume given the storm intensity. Then we build a hierarchical Bayesian model aiming at the prediction of 15- and 30-minutes cumulated precipitation at unobserved locations and time using information on lightning in the same area. More specifically, we build a Bayesian hierarchical model in which precipitation is modeled as function of lightning count and space time variation is handled using specific structured (random) effects. The mean component of the model relates precipitation and lightning assuming that the number of lightning recorded on a regular grid depends on the number of lightning occurring in neighboring cells. We analyze several model formulations where storms propagation speed, spatial dependence and time variation incorporates different descriptions of the phenomena at hand. The space-time variation is assumed separable. The study area is located in Central Italy, where two storms, that differ for duration and intensity, are presented.