WSEAS Transactions on Environment and Development
Print ISSN: 1790-5079, E-ISSN: 2224-3496
Volume 22, 2026
Detecting Temporal Land-Use Regimes in Mediterranean Agricultural Lowlands Using Landsat Time Series and Unsupervised Learning: Evidence from Albania
Authors: , , ,
Search Articles
Abstract: Mediterranean agricultural lowlands are undergoing intense land-use transformations driven by climatic variability, agricultural intensification, as well as urban pressure. Despite the availability of long satellite archives, long-term seasonal land-use dynamics and their functional heterogeneity remain insufficiently captured by conventional linear trend analyses. This study proposes a geo-information–driven analytical framework that integrates long-term satellite time series with unsupervised machine learning in order to identify functional land-use regimes beyond linear assumptions. Monthly Landsat time series spanning 25 years (2000–2025), derived from Landsat 5 TM, 7 ETM+, 8 OLI/TIRS, and 9 using the Google Earth Engine platform, were analysed within the R environment in order to ensure methodological transparency and reproducibility. Spectral indices revealed pronounced seasonality and increasing vegetation activity. K-means clustering, it identified three temporal spectral regimes—intensive agriculture, transitional urban/sparsely vegetated areas, and mixed agriculture–vegetation systems—delivering a transferable framework for land-use monitoring and spatial planning.The proposed framework improves land-use analysis and also supports spatial planning.
Keywords:
Geo-information science, Spatial time-series analysis, Landsat, Machine learning, Unsupervised clustering, Land-use dynamics, NDVI, Mediterranean agricultural landscapes
Pages: 1249-1269
DOI: 10.37394/232015.2026.22.103