Naive Bayes Assessment of Mediterranean Basin Ecosystem Wildfires Associated With Vegetation Vigor in August 2021

Abstract

The Mediterranean basin, located in the Southern Mediterranean Sea in the North of Africa, is one of the ecosystems that are vulnerable to wildfire hazards. A massive wildfire occurred in this ecosystem in August 2021. Here, this study aims to detect the burnt areas and assess the ecosystem covariates related to the wildfire occurrences. Wildfire assessments were conducted using remote sensing (RS) analysis. Measured ecosystem covariates include vegetation cover, moisture, and the Normalized Difference Vegetation Index (NDVI). The RS analysis confirms at least 17090 ha of vegetation cover were damaged. The burnt vegetation cover was associated with the vegetation moisture and NDVI. The result shows 61.93% of burnt areas were observed within areas that had low vegetation moisture. While 51.75% of burnt areas occurred within areas that had low NDVI values. Naive Bayes classifier confirms that vegetation moisture and NDVI covariates can be used to estimate the presences of wildfires with values of Kappa, accuracy, precision, recall, and F1 score are 0.7, 0.85, 0.89, 0.8, and 0.84. The area under (AUC) the receiver operating characteristic (ROC) curve values for vegetation moisture and NDVI covariates were 0.945 and 0.399. This value indicates that low vegetation moisture contributes more to the wildfire presences than low NDVI.