Resumo

The Amazon biome plays an important role in the climate system, with relevance at regional and global scales. Fire occurrences, related to both natural and anthropogenic activities, are relevant disturbances in the Legal Amazon, with significant effects. Changes in the patterns of fire occurrence in the Amazon region have been widely reported in the literature and are related with a variety of factors, including dry conditions, deforestation, agricultural expansion, climate changes, and climatic anomalies such as El Ninõ events.

The purpose of this paper is to analyze the existence of changes in the patterns of the fire occurrence in the Legal Amazon, within the spatio-temporal point process framework. To do this, we propose a novel methodology to extent the trend-cycle decomposition in spatio-temporal models to spatio-temporal point pattern data, by proposing to use a dynamic representation of a Log Gaussian Cox process (LGCP) where the intensity function is modeled through the decomposition of components into trend, seasonality, cycles, covariates and spatial effects.

The LGCP is a particular case of the Cox process, where the log-intensity function is a Gaussian field. Due to the stochastic property of the LGCP, fitting this model is often computationally expensive. In this sense, to perform the estimation in a computationally effective way, we use the stochastic partial differential equation approach to transform the initial Gaussian field to a Gaussian Markov Random Field, which is defined by sparse matrices. Furthermore, the resulting Bayesian hierarchical model fits within the integrated nested Laplace approximations framework (Rue et al., 2009).

To perform inference procedures, we proposed a structural decomposition to spatio-temporal point pattern data. In particular, we proposed to use a dynamic representation of a Log Gaussian Cox process where the intensity function was modeled through the decomposition of components into trend, seasonality, cycles, covariates and spatial effects. This useful formulation was able to capture permanent changes in the fire occurrence and also, to identify seasonal and cyclic effects (Laurini, 2019; Valente e Laurini, 2020).

The results show that long-term movements of fire activity dropped considerably between 2006 and 2012, which suggest that conservation regulations and/or market conditions in the mid-2000s were effective in reducing the fire events. Also, our model captured an increase in the trend component between 2013 and 2016, and after 2018, which may be explained by localized drivers associated with political measures that encourage the expansion of agriculture and livestock.

The estimated components suggested relevant changes in the patterns of the fire activity in the Legal Amazon. In particular, it is possible to observe how the long-term component is affected by conservation regulations and/or market conditions, i.e., the obtained evidences suggest that the changes in the fire ocorrences are mostly related to human-induced activities. Furthermore, the seasonal component provided evidence that fire events in the Legal Amazon has become more consistent throughout the year, suggesting the increase of fuel management practices occurring during the nonfire season.

Laurini, M., 2019. A spatio-temporal approach to estimate patterns of climate change. Environmetrics 30, e2542. Valente, F., Laurini, M., 2020. Tornado occurrences in the united states: A spatio-temporal point process approach. Econometrics8, 25. Rue, H., Martino, S., Chopin, N., 2009. Approximate bayesian inference for latent Gaussian models by using integrated nestedLaplace approximations. Journal of the Royal Statistical Society: Series B (Statistical Methodology) 71, 319–392.