Modeling forest fire risk using remote sensing and GIS in Virunga massif

Abstract

Forest fire is a serious concern in many ecosystems across the globe due to environmental and human loss they induce. In Virunga Massif, forest fires are not frequent and in some very few cases they are caused by illegal activities. However, forest fires in this area can be more devastating and accompanied by several impacts on the forest and biodiversity as it was the case during the fire outbreak in Virunga Massif in 2009. In this study forest fire risk model was developed using a combination of GIS and Remote Sensing tools with the aim of identifying vulnerability of different areas of the ecosystem to forest fire hazard. To achieve this, a forest fire model was designed by combining and weighting different variables into sub-models which later were integrated into the final forest fire model using the spatial multi-criteria evaluation approach. Regarding data processing, Landsat 8 satellite imagery was classified to obtained different vegetation classes used as fuel input to the model. In addition, topographic variables were derived from the Digital Elevation Model, and major anthropogenic activities which are considered as source of forest fire were identified and analyzed. Furthermore, the extent to which roads and human settlements contribute to fire detection and fire extinction response were analyzed. Finally, the thematic maps were assigned subjective weights and added up in Arc GIS to produce the fire risk model which was validated with both the burnt area points map and ground truth. Results showed that forest fire risk depend mostly on ignition factors which are illegal activities associated with fire and proximity to beehives sites (weight = 0.496); and topographic factors (weight = 0.243) which plays a crucial role in fire spreading. Other factors (fuel, detection and response) contributed less to the model. According to the final risk map, the high risk zone areas are located in steep slope areas around volcanoes cones while areas located in the interior of the forest were less vulnerable to forest fire. These results are in agreement with the field situation since the overall accuracy of the model was 75%; which suggests that the forest fire model is reliable and can be adopted. The results suggest therefore that in order to prevent forest fire hazard, it is important to control the ignition factors by preventing illegal activities in critical areas as indicated in the model map especially during the dry season, and lastly, relocating beehives sites at farther distance from the park edge. Key Words: Forest Fire Modeling, Biodiversity, Illegal Activities, Ignition factors, Topographic Factors, Spatial Multi-Criteria Evaluation