Unraveling the Role of Temperature and Rainfall on Active Fires in the Brazilian Amazon Using a Nonlinear Poisson Model

Abstract

Extreme droughts and high temperatures have become more frequent in the last two decades, increasing fire risk in the Amazon. The overarching goal of this study is to shed light on the influence of temperature and rainfall on fire occurrence for the 1998–2013 period. We use a Poisson regression to model satellite-based monthly fire counts across the Brazilian Amazon as a function of observed rainfall and temperature. Specifically, in the nonlinear regression framework we explore the fire count at month t as a function of the total monthly rainfall and monthly average of maximum daily temperatures at month t as well as lagged observations of these two predictors and of the fire counts. Our results indicate that the influence of temperature on fire counts can be larger than the effects of rainfall. Considering the temperature-fire relationship, the extra variance explained by rainfall is about 9%. Assuming average rainfall, we show that a 1°C increase in the monthly average of maximum daily temperatures results in a 30% increase in fire counts (19 fires per Mha more than the average), which translates into significant changes in the likelihood of fires within the Amazon. Our findings provide new insight about the role of temperature and rainfall in regulating fire occurrence in the Amazon, and the sensitivity of fire counts to relatively small changes in temperature highlights the vulnerability of the Amazon in a warming climate, where much higher temperatures are expected by the end of this century.