Predicting postfire mortality of Pinus halepensis Mill. and Pinus pinea L.

The probability of mortality of Pinus halepensis and P pinea trees after 6 separate wildfires (which took place from 1989 to 1991 in South Eastern France) was modelled with logistic regression analysis using data on tree size and fire-damage descriptors from 998 trees in 13 stands. For both species, probability of mortality increased with increasing percentage of crown scorched and estimated depth of bark charring, and with decreasing tree DBH. The best fitted and field convenient model was selected for each species. Fire-caused mortality of P halepensis is a function of these three variables and for P. pinea is a function of the first two variables only. Analysis of receiver operating characteristic (ROC) curves indicated that the selected models perform well for both species with 85 and 95 percent concordance between predicted probabilities and observed outcomes for P halepensis and P pinea respectively. Models may be used for guiding salvage fire-damaged P haiepensis or P pinea stands. This paper reports on cases of different management scenarios by selecting different decision criteria when using the proposed models. Comparative analysis of fire resistance and survival of the two Mediterranean pine species shows that P. halepensis is more fire sensitive than P pinea.