Multisource remote sensing time series for mapping and long term monitoring of grapevine vineyards in the Northern Zagros

Abstract Forest conversion to agriculture is a critical threat to global ecosystems, with specialized viticulture driving significant land-use changes. In the Northern Zagros forests of Iran, high economic profitability has accelerated the expansion of grapevine ( Vitis vinifera L.) vineyards. This study mapped the spatio-temporal expansion of vineyards in Sardasht County (2016–2025) using multi-source remote sensing and machine-learning techniques. Time-series Sentinel-2 imagery was used as the primary data source for classification, while high-resolution PlanetScope imagery was used as supplementary reference information for visual interpretation and the collection of reference samples. A Random Forest classifier was implemented in Google Earth Engine using 63 input variables, including spectral bands, topographic factors, and phenological metrics. Classification achieved high accuracy, with Overall Accuracy (OA) ranging from 93.6% to 94.8% and Kappa values exceeding 0.91. Key predictors included topographic variables, SWIR and Red-Edge bands, and the NDMI index. Results show that vineyard area increased by 38.7%, growing from 5,820.5 ha in 2016 to 8,074.7 ha in 2025. Notably, over 54% of this expansion originated from converting sparse (32.04%) and dense (22.13%) forests, while 37.96% occurred on existing agricultural lands. The proposed remote sensing and machine-learning framework demonstrates high effectiveness for monitoring vineyards in heterogeneous mountain environments. The substantial expansion of vineyards highlights their growing economic importance but reveals critical pressure on Zagros forest ecosystems, emphasizing the urgent need for sustainable land-use planning to balance agricultural development with forest conservation.