The applications of geomatics to predicate forest fire based on earlier pre-fire conditions. A case study: Belezma national park -Algeria

Abstract Remote sensing has become a primary data source for fire hazard prediction, including other techniques, but not limited to geographic information systems (GIS), artificial intelligence (AI), and machine learning. This research paper aims to measure the vulnerability to forest fires and define pre-forest fire criteria in semi-dry climates. The study area is "Belezma" national park; which represents one of the most vital national parks in Algeria since it contains numerous rarest plants in the world; and over 20% of reserved animals; as a result, the park represents an example of fire prevention in areas of high environmental values. The vulnerability is assessed by using various pre-fire factors: Normalized Difference Vegetation Index (NDVI), Normal Difference Water Index (NDWI), and land surface temperature (LST). Once the variables are generated based on Landsat 8 satellite imagery collected between May and August of the same year, the proposed fire model is implemented into a GIS environment using the analytical hierarchy process (AHP). The results demonstrated that elevation had a significant impact compared to other factors; as a result, most fires are triggered during the dry period where the vulnerability is low in the highest elevation area; vice versa, the vulnerability values are higher in the louder elevation zones. Finally, the validation was performed by using two approaches: firstly, we checked the results on the ground, where we detected 150–300 years old non-burned trees in the low vulnerability area using the dendrochronology method; secondly, we geolocated every local wildfire between 2018 and 2020; then measuring the spatial correlation between recorded wildfires and the vulnerability using Moran index method.