The influence of bioclimatic and topographic variables on grassland fire occurrence within an urbanized landscape

Abstract

Unplanned veldfires (or wildfires) characterize vegetation landscapes and offer a range of ecological benefits that promote, among others, the health of the grasslands and other fire-adapted ecosystems. However, in urbanized areas, uncontroled fires are often a threat to property, life, the environment, and economy. The eThekwini Municipal area, a global biodiversity hot-spot experiences frequent unplanned veld fires that threatens the valuable remnant grasslands. This necessitates an understanding of key drivers to fire occurrence as a first step towards the remnant grasslands sustainability. In this study, the probability of fire occurrence within the study area was determined using the Near Real-Time (NRT) MODIS Collection 6 Active Fire Data, topographic and bioclimatic variables within the Maximum Entropy (Maxent) environment. The predictor variables were assessed using jackknife analysis, percentage contribution, and Area Under Curve (AUC). The results showed that the mean temperature of the coldest quarter (33%), isothermality (12.3%), elevation (8.9%), and precipitation of the warmest month (8.8%) were the most influential predictor variables affecting fire occurrence within the study area. The Area Under Curve (AUC) values for training and test data-sets were 0.728 and 0.716 respectively, indicating good accuracy for the fire occurrence probability modeling. The study concludes that the Maxent modeling algorithm is suitable for determining fire occurrence and identifying key topographic and bioclimatic fire drivers within an urban landscape. These results are valuable in informing the protection and conservation of urban ecological systems, useful for the provision of urban ecosystem goods and services.