Fine-Scale Mapping of Natural Ecological Communities Using Machine Learning Approaches

Remote sensing technology has been used widely in mapping forest and wetland communities, primarily with moderate spatial resolution imagery traditional classification techniques. The success of these efforts varies widely. natural communities the Laurentian Mixed Forest are an important component Upper Great Lakes ecosystems. Mapping monitoring using high benefits resource management, conservation restoration efforts. This study developed a robust approach to delineate habitat utilizing multispectral high-resolution (60 cm) National Agriculture Imagery Program (NAIP) data. For accurate training set delineation, NAIP imagery, soils data spectral enhancement techniques such as principal analysis (PCA) independent (ICA) were integrated. evaluated importance biogeophysical parameters topography, soil characteristics gray level co-occurrence matrix (GLCM) textures, together normalized difference vegetation index (NDVI) water (WINAIP) indices, joint mutual information maximization (JMIM) feature selection method various machine learning algorithms (MLAs) accurately map communities. Individual community user’s accuracies (UA) ranged from 60 100%. An overall accuracy (OA) 79.45% (kappa coefficient (k): 0.75) random (RF) OA 75.85% (k: 0.70) support vector (SVM) achieved. showed that use ancillary layers was critical improve interclass separation accuracy. Utilizing available free coupled integrated MLAs, fine-scale successfully delineated spatially spectrally complex environment.