Modeling stand fire probabilities with unobserved heterogeneity. Estimating stand age and climate change effects in Chilean radiata pine plantations

Evenly managed forest plantations are potentially vulnerable to fires because of their high fuel load build-up in each rotation. We use panel data analysis that considers the possible correlation between the observed covariates of interest with spatial unobserved heterogeneity. We compared two alternative approaches to estimate the stand burn probability function of the stand age: the average structural function (ASF) and the local average structural function (LASF). While our results show a significant positive effect of the stand age for the mature stage under both approaches, more differentiation in the stand burn probability for different ages is captured with the LASF. Also, under the LASF, the stand age functions are more sensitive to changes in site productivity. We predicted the burned area under the RCP 4.5 and RCP 8.5 climate scenarios considering adaptation in management regimes to site productivity changes. The largest impact is projected for the coastal areas where site productivity increases are combined with more suitable climate conditions for flammability. For the dryer hinterland, however, stand burn probabilities and the burned area are predicted to decline in the second period and the RCP 8.5 because of the dominant negative effect resulting from the site productivity reduction.