Abstract:
This research study offers a quantitative understanding of the environmental factors related to fire occurrence and its potential distribution. The Huai Kha Khaeng Wildlife Sanctuary, located in the northwestern region of Thailand, serves as an example to analyze and expand the knowledge base of forest fire ecology in tropical environments. Specific objectives focus on establishing the relationship between physiognomic variables that are related to forest fuel loading among different forest types. In addition, this study aims at modeling and quantifying the relative importance of the biophysical variables associated with the occurrence of tropical forest fires. The methodological framework links Geographic Information Systems (GIS) with the potential of statistical analysis. Thematic layers of several biophysical variables are combined in GIS, along with field measurements of fuel loading and stand physiognomy. Under the statistical analysis, the variability and interactions of spatial attributes related to fires are synthesized using the Decision Tree modeling. GIS is further employed to display the modeling results. Rainfall pattern, geological material, aspect and vegetation index variables significantly influence the ability of the Decision Tree model to predict the likely occurrence of fire. They explain most of the processes underlying a hierarchical set of rules that help to distinguish the varying levels of fire hazards. The vegetation index in particular was found to be a strong potential indicator of fire incidents and an underlying driving factor behind fuel moisture dynamics. At a certain vegetation index, two types of forest were distinguished as having wet and dry fuel conditions. The difference in the amount of fuel load between the physiognomically distinct evergreen and deciduous forests is proven to be insignificant, except in the variation of moisture content. Factors contributing to the varying levels of fuel moisture in evergreen forests are controlled by the micro-climate created by its intact crown cover. However, there is no distinct relationship concerning the stand structure of deciduous forest with regard to the dryness of fuel on its floor. The dominance of weather over the fuel variables suggests that forest fire situations in an open and dry stand of deciduous forest is driven by extreme weather conditions. A GlS-generated map of the sanctuary illustrates the spatial variation in fire hazard probabilities as predicted by the Decision Tree model. The prediction accuracy of fire hazard zones based on bio-physical factors is further enhanced by incorporating the proportion of neighbouring areas with high potential for ignition. The potential combustibility and danger rating are determined for the predicted hazard zones. In addition, the spatial association of the neighbouring human settlements is analyzed. This research expands the value of GIS from the usual selective retrieval and investigation of spatial patterns into the evaluation of the complex hierarchical combinations of spatial attributes. The combined effect of GIS and statistical modeling eliminates the problems of handling a mixture of environmental data and identifying both variables and attributes interactions. Likewise, the need to design site-specific fire managemet strategies, as guided by particular combinations of environmental attributes, takes into account the applicability of the data-driven Decision Tree modeling. Keywords: fire hazard modeling, fire management, GIS, Decision Tree analysis, Thailand.