For NASA (2007)
A novel wildfire risk prediction algorithm, based on support vector machines, is presented. The algorithm depends on the observational data from previous weather conditions to predict the fire hazard level of a day. The model considered parameters to build the spread index are the slopes or position of the slope as well as weather data (precipitation, wind speed and direction, temperature) and fuel type, to predict the spreading of the fire, and the wind speed in addition to historical data. These parameters are integrated to make fire predictions using Support Vector Machines (SVM).
For NASA (2010)
A novel approach for modeling anthropogenically-initiated wildfire ignition was developed that significantly advances the theoretical knowledge of human-wildfire interactions. Gravity interaction models that are commonly used for economic analyses associated to increase business competition; which herewith combined with fluid dynamics models that mimic human movement patterns to predict the probability of anthropogenically-initiated wildfire.
The study identified population centers and transportation corridors, in particular: proximity to railroads and roads; traffic volume; and density of the corridors as the most influential factors for wildfire ignition. The population centers are identified as global influencing factors, and are modeled as the gravity term. The transportation corridors are identified as local influencing factors, and are modeled using fluid flow analogy as diffusion and convection terms. An analytic convection diffusion model (CDM) model is derived and the model coefficients calibrated using historic wildfire data. The outputs of the proposed method is based on a multivariate statistical analysis used in combination with artificial intelligence (AI) / machine learning algorithms are utilized as inputs to the CDM model.
The model helped fire managers to better plan wildfire mitigation (fuel reduction) strategies and effectively stage equipment and personnel geographically in the areas of drought that are coincident with high ignition probability. Land use and transportation managers will gain better understanding of the changes in wildfire risk pattern due to urban fringe development.