Wildfire Susceptibility Modelling in Queen Elizabeth National Park - Uganda.
Irumba, Derrick Robert
MetadataShow full item record
Existence of vast Savannah grasslands that provide flammable fuel as well as rhythmic wet and dry seasons contribute to the occurrence of wildfires in protected areas in the Sub-Saharan region of Africa. Wildfires cause ecological disturbances in these landscapes affecting both flora and fauna. To mitigate against the negative impacts of wildfires, prevention practices such as creation of fire breaks and suppression practices such as fire extinguishing are applied for management of wildfires in Queen Elizabeth National Park (QENP) - Uganda. Effective use of fire management practices requires precise information to guide their application. However, due to lack of spatial information about the areas susceptible to wildfires over time, the prevention and suppression practices are often inefficient. To bridge this information gap, this research was therefore carried out to determine the variability of wildfire susceptibility of QENP using MODIS and VIIRS satellite data. This was achieved by characterization of wildfire conditioning factors including altitude, aspect, NDVI, Precipitation, Proximity to lakes, rivers and settlements in relation to previous wildfires and the modelling of areas susceptible to fire within the park using Weights of Evidence (WOE) statistical method. Results of the study revealed that fire occurrences were seasonally dependent with most fires observed in dry seasons because of accumulation of dry flammable vegetation. NDVI, altitude and proximity to lakes were the conditioning factors that exhibited the highest correlation categories with the occurrence of fires. This was attributed to water stress in vegetation, physio-graphic influence and socioeconomic activities of the fishing villages around the lakes. WOE achieved 70.3% and 69.9% success rates as validated using Receiver Operating Characteristics - Area Under the Curve method for VIIRS and MODIS derived models respectively which implied that the modeling results were reliable. Wildfire susceptibility maps were developed with 5 susceptibility levels obtained as very high, high, moderate, low, very low. 19% of the study area was classified with susceptibility level as very high and 20% as high with the remaining 61% as moderate, low and very low. The 19% of the study area with very high susceptibility is thereby the most prone to wildfires and these wildfires peak in the periods of January-March and June-August. This effectively narrowed down the areas and periods in which to focus application of fire prevention and the limited suppression resources. As a recommendation, the study could be up-scaled for development of early warning systems that can detect conditions potentially leading to causation of wildfires so that advance preparations could be made to prevent them.