Overcoming Parcel Delineation Constraints in Georgia Using Multi-Source Satellite Imagery

Accurate parcel delineation is a fundamental prerequisite for effective land administration, agricultural monitoring, and evidence-based policy making. In Georgia, the parcel delineation encounters multiple challenges. First, high-resolution aerial imagery is not available on a regular, annual basis, limiting consistent updating of parcel boundaries. Second, agricultural land in Georgia is highly fragmented, with small, irregularly shaped parcels that complicate boundary identification with medium-resolution data. Although the systematic registration of agricultural parcels was completed in 2025, major shortcomings persist concerning the current cultivation status of registered parcels and the spatial scope of abandoned agricultural land. This study explores the potential of multi-source satellite imagery to overcome these restrictions using automated parcel delineation. The research integrates Sentinel-2 multispectral time series, high-resolution PlanetScope imagery, aerial photographs, and official cadastral data to train and validate parcel delineation models within the Google Earth Engine (GEE) environment. Multi-temporal spectral metrics, texture features, and phenological information are used to enhance the separability of homogeneous agricultural units and improve boundary detection accuracy. The method is tested in two municipalities: Dedoplistskaro, dominated by large-scale annual crops, and Gori, characterized by orchards with complex canopy structures and fragmented parcel patterns. The findings show that combining medium-resolution temporal information from Sentinel-2 with the spatial detail of PlanetScope significantly improves automated parcel delineation compared to single-sensor approaches. The proposed workflow provides a framework for identifying cultivated and potentially abandoned parcels, offering a useful basis for agricultural monitoring and land management in data-limited environments such as Georgia.