A Statistical Framework for Detecting Environmental Variability in Satellite Imagery Using Covariance Determinants

This work proposes a statistical framework for detecting environmental variability in satellite imagery using multivariate analysis of spectral data. The method is based on covariance determinants computed within local pixel windows, providing a quantitative measure of spatial heterogeneity. The framework is particularly relevant for complex environments affected by atmospheric noise, such as coastal fog systems (e.g., the Camanchaca in northern Chile). This study establishes a methodological foundation for future applications in environmental monitoring, time series analysis, and machine learning-based predictive modeling. This work is part of an open research initiative led by Meqatec SpA. Collaboration is welcome.