This study presents an analysis of the influence of general landscape-level indicators on wildfire and its spatial susceptibility across a fire-prone landscape in the southeast of China using an integrated WOE-AHP model that consists of a statistical/probabilistic Weights-of-Evidence (WOE) model and a knowledge-based Analytical Hierarchy Process (AHP). Multi-class landscape indicators (i.e., slope, aspect, altitude, NDVI, annual rainfall, wind speed, land use, and proximity to rivers, roads, and human settlements) are weighted by the WOE model and act as the input data for pairwise analyses, which improved the commonly used AHP procedure that relies on a traditional nine-point pairwise rating scale and expert opinion. The model performance was evaluated using the ROC-AUC method that revealed that the integrated WOE-AHP model performed well both in terms of goodness-of-fit with the training dataset (AUC success rate = 0.94) and the capability to predict future ignitions (AUC prediction rate = 0.91). The efficiency of the proposed model was compared to a logistic regression and single WOE models and comparative analyses using the Wilcoxon signed-rank tests demonstrated a significant improvement of wildfire prediction using the integrated WOE-AHP model over these other models. Overall, given the proven capability of integrated modeling in identifying very influential landscape indicators, excluding the indicators with null predictive utility, and improving the prediction of wildfires, modelers can now use this alternative to the current modeling approach and tailor its use to any research related to disturbances to which landscapes are differentially susceptible.
All Science Journal Classification (ASJC) codes
- Decision Sciences(all)
- Ecology, Evolution, Behavior and Systematics