2025 , Volume 30, № 4, p.77-89

Satellite imagery is successfully employed for urban studies, flood forecasting, fire detection, and vegetation classification. Machine learning models analyze and classify these images based on their spatial, spectral, and temporal characteristics. To streamline Earth observation data processing, platform-based solutions have been developed, which integrate a comprehensive set of tools from image acquisition to result dissemination within a unified environment. The platform of the ISDCT SB RAS has introduced a repository of multispectral satellite imagery with an intuitive timeline based navigation system. This functionality allows researchers to efficiently track environmental and anthropogenic changes in a given study area over time. Additionally, the platform includes a semi automated training sample generation tool, significantly accelerating the preparation of datasets for machine learning model training. The integration of JupyterHub enables users to develop custom classification models or apply pre-trained models for image analysis. Classification of satellite images allows to identify different types of the land cover on the imagery and track their changes over time. Furthermore, the platform supports the seamless publication of derived geospatial maps on an interactive geoportal, ensuring that results are presented in a decision-ready format. This holistic approach not only optimizes remote sensing data analysis workflows but also bridges the gap between advanced computational methods and practical applications in environmental monitoring