GIS Characterization For Peatland Forest Fire Index Using GeoAI Technique Nadia Fitriana Rahimi, Fareesha Irdeena Azman, Aduwati Sali, Sheriza Mohd Razali
Universiti Putra Malaysia
Abstract
Peatland forests are highly fire sensitive ecosystems, particularly during prolonged drought conditions. Incorporating Groundwater Level, GWL into Fire Weather Index, FWI system is crucial as GWL significantly influences fire risk. when GWL decreases, peat soil becomes increasingly exposed to the oxygen which raising its flammability and the likelihood of fire outbreaks. To enhance the fire risk prediction, the study proposes an improved method over the traditional Inverse Distance Weighting, IDW approach. Specifically, it integrates Geospatial Artificial Intelligence, GeoAI with Geographic Information System, GIS technologies for fire risk assessment and geospatial mapping. The paper explores the application of GeoAI in predicting GWL using both machine learning and deep learning techniques. Meteorological and spatial parameters such as temperature, rainfall, soil moisture, latitude, and longitude were used to increase the reliability of GWL predictions. The methodology includes data preparation, feature identification, model development and spatial mapping of actual versus predicted GWL values. Model performance was evaluated using RMSE, MAE, R2 score. Among the tested models, Long Short Term Memory, LSTM demonstrated the highest accuracy and lowest error and effectively capturing both temporal and spatial environmental patterns. predicted GWL values were taken then spatially mapped to visualize groundwater conditions and assess fire prone zones. The findings highlight the potential of advanced predictive models to enhance environmental monitoring, thereby strengthening efforts towards ecological sustainability and resilience to form a strong basis for future research on real time sensor integration and geospatial fire risk analysis.
