# Global Fire Season Severity Analysis and Forecasting

## TL;DR

This study analyzes global fire season severity using a hexagonal grid and evaluates the accuracy of short- and medium-term forecasting methods based on historical fire detection data, revealing high predictability in most regions.

## Contribution

It introduces a simple, fast, and globally applicable approach to estimate and forecast fire season severity using only historical fire detection data.

## Key findings

- Most fire seasons are shorter than seven months.
- Over half of the regions show decreasing fire season lengths.
- Forecasting errors are lower than average fire season severity in 95% of regions.

## Abstract

In this paper, we divide the globe into a hexagonal grid and we extracted time series of daily fire counts from each cell to estimate and analyze worldwide fire season severity (FSS), here defined as the accumulated fire detections in a season. The central question here is evaluating the accuracy of time series forecasting methods to estimate short-term (months) and medium-term (seasons) using only historical data of active fire detections. This approach is simple, fast, and use globally available data, making it easier for large scale prediction. Our results comprehend descriptive and predictive analyses of the worldwide seasonal fire activity. We verified that in 99% of the cells, the fire seasons have lengths shorter than seven months and that 57% have their lengths decrease. We also observed a declining tendency in the number of active fire counts during the seasons in 61% cells. However, some regions like the Northeast Brazil and the West Coast of the USA present an increasing trend. We verified that the forecasting error is lower than the mean FSS in 95% of the cells, indicating clear predictability in the FSS.