Analysis of recent spatial–temporal evolution of human driving factors of wildfires in Spain

Abstract

Fire regimes are strongly dependent on human activities. Understanding the relative influence of human factors on wildfire is an important ongoing task especially in human-dominated landscapes such as the Mediterranean, where anthropogenic ignitions greatly surpass natural ignitions and human activities are modifying historical fire regimes. Most human drivers of wildfires have a temporal dimension, far beyond the appearance of change, and it is for this reason that we require an historical/temporal analytical perspective coupled to the spatial dimension. In this paper, we investigate and analyze spatial-temporal changes in the contribution of major human factors influencing forest fire occurrence, using Spanish historical statistical fire data from 1988 to 2012. We hypothesize that the influence of socioeconomic drivers on wildfires has changed over this period. Our method is based on fitting yearly explanatory regression models-testing several scenarios of wildfire data aggregation-using logit and Poisson Generalized Linear Models to determine the significance thresholds of the covariates. We then conduct a trend analysis using the Mann-Kendall test to calculate and analyze possible trends in the explanatory power of human driving factors of wildfires. Finally, Geographically Weighted Regression Models are explored to examine potential spatial-temporal patterns. Our results suggest that some of the explanatory factors of logistic models do vary over time and that new explanatory factors might be considered (such as arson-related variables or climate factors), since some of the traditional ones seem to be losing significance in presence-absence models, opposite to fire frequency models. In particular, the Wildland-Agricultural Interface and Wildland-Urban Interface appear to be losing explanatory power regarding ignition probability, and Protected Areas is becoming less significant in fire frequency models. GWR models revealed that this temporal behavior is not stationary neither over space or time.

Exploring spatial–temporal dynamics of fire regime features in mainland Spain

“…However, this behaviour is reversed during winter, although trend magnitude is less marked. This is probably due to the improvement in fire extinguishing or encouraging monitoring and prevention (MAPA, 1988;Rodrigues et al, 2016), particularly encouraged during summer. Fires ignited by lightning perhaps show the most contrast as there is a marked dichotomy between west and northeast.…”

A Comparison between Spatial Econometric Models and Random Forest for Modeling Fire Occurrence

“…However, as to urban infrastructure fires, the important predictors are much different and, therefore, the measures for preventing infrastructure fires from happening will be very different too. We should not deny that the occurrence of fire often shows spatial and temporal clustering and lagging [41][42][43]. Moreover, many natural phenomena such as chemical toxicants and PM 2.5 are spatially auto-correlative and the SE model is a useful tool for explaining the structure of natural hazards [44][45][46][47][48].…”

Sustainable Land Management, Wildfire Risk and the Role of Grazing in Mediterranean Urban-Rural Interfaces: A Regional Approach from Greece

“…Factors underlying an increased fire risk during the most recent decades are multifaceted and differentiated at the geographical scale, evolving rapidly over time [15,28,29]. Sustainable land management is fundamental in these circumstances [1,[30][31][32][33][34].…”