Identifying Spatial Patterns of LULC Change with K-Means: A Case Study Using Satellite Data and Google Earth Engine

Rapid land use and land cover (LULC) changes in Klungkung Regency—marked by declining vegetation and expanding urban areas—present significant challenges for sustainable land management. This study employs the K-Means clustering algorithm on multi-temporal Sentinel-2 imagery from 2019, 2021, and 2023, using 1,000 spatial sample points distributed across Klungkung Regency, Bali. Six spatial features were extracted and standardized for analysis, comprising vegetation indices (NDVI, EVI, FVC), NDBI, elevation, and population density. All data collection and preprocessing steps utilized the Google Earth Engine (GEE) platform, enabling efficient handling of large-scale geospatial datasets. Clustering was conducted in an unsupervised manner, with the optimal configuration identified at k = 2, validated by consistent Silhouette Scores above 0.76 throughout the study years. The results reveal a notable temporal shift from built-up (Cluster 1) to vegetated (Cluster 0) land, with Cluster 0 increasing from 41% of the area in 2019 to 61% in 2023. This trend may reflect a combination of reforestation initiatives, land use regulations, or reduced urban expansion post-pandemic. The study demonstrates the utility of integrating GEE with unsupervised clustering to detect subtle spatial transitions, providing robust outputs for evidence-based policymaking, long-term environmental monitoring, and adaptive spatial planning in regional land resource management.