Global Fire Season Severity Analysis and Forecasting

11 Mar 2019  ·  Leonardo N. Ferreira, Didier A. Vega-Oliveros, Liang Zhao, Manoel F. Cardoso, Elbert E. N. Macau ·

Global fire activity has a huge impact on human lives. In recent years, many fire models have been developed to forecast fire activity. They present good results for some regions but require complex parametrizations and input variables that are not easily obtained or estimated. In this paper, we evaluate the possibility of using historical data from 2003 to 2017 of active fire detections (NASA's MODIS MCD14ML C6) and time series forecasting methods to estimate global fire season severity (FSS), here defined as the accumulated fire detections in a season. We used a hexagonal grid to divide the globe, and we extracted time series of daily fire counts from each cell. We propose a straightforward method to estimate the fire season lengths. Our results show that in 99% of the cells, the fire seasons have lengths shorter than seven months. Given this result, we extracted the fire seasons defined as time windows of seven months centered in the months with the highest fire occurrence. We define fire season severity (FSS) as the accumulated fire detections in a season. A trend analysis suggests a global decrease in length and severity. Since FSS time series are concise, we used the monthly-accumulated fire counts (MA-FC) to train and test the seven forecasting models. Results show low forecasting errors in some areas. Therefore we conclude that many regions present predictable variations in the FSS.

PDF Abstract

Datasets


  Add Datasets introduced or used in this paper