Wildfires have intensified, causing substantial ecological and hydrological impacts. Post-wildfire areas experience vegetation changes, which leads to increasing erosion, causing flash floods, and degrading downstream water quality. Due to its effect on the environment, It is essential to understand these post-wildfire vegetation changes and map them for damage assessment. This study focused on understanding the performance of various Machine Learning (ML) algorithms for pre-fire and post-fire land cover classification. It also involved reviewing the hydrological effects of the fire on the ecosystem under consideration through a literature review. The US Forest Service dataset was used to identify five land cover classes in the research area: Trees, Grass/Forb/Herb-Tree-Tree Mix, Barren-Shrubs Mix, Grass/Forb/Herb-Shrubs Mix, and Shrubs. The regions of interest (ROIs) for the study area were created on the five identified vegetation classes. The study developed ML classification algorithms of Support Vector Machines, and Neural Networks by optimizing hyperparameters during the model training phase. By systematically selecting the best hyperparameters, we created classification models and deployed them for producing pre-fire and post-fire land cover classification maps. The model performance was measured through mean absolute error (MAE) and root mean square error (RMSE). The comparison between the pre-fire and post-fire land cover maps reveals significant changes to the vegetation cover potentially leading to flash floods in Harris Spring Canyon in 2014. The study's significance lies in developing ML models for creating land cover classification maps.
