Dynamic maps of plastic-mulched farmlands in Northeast China from 1985 to 2025

Abstract We provide PMF-NEChina-40D, the first plastic-mulched farmlands (PMFs) dynamic monitoring dataset in Northeast (NE) China for the past four decades from 1985 to 2025. Multi-sensor Landsat series and Sentinel-2 imagery were processed on the Google Earth Engine platform, and a dual-temporal PMFs mapping framework integrating film-on and film-off stages was proposed. Specifically, the Random Forest classifier, combined with grid-based partitioned modeling, was used to produce PMFs maps with spatial resolutions of 30 m (1985–2015) and 10 m (2020–2025) for nine time steps and every five years, including 1985, 1990, 1995, 2000, 2005, 2010, 2015, 2020 and 2025. Training and validation samples were generated through visual interpretation supported by multi-source satellite imageries and field surveys. Classification accuracy was evaluated using multiple metrics, with average user’s accuracy, producer’s accuracy, F1 score and overall accuracy exceeding 90%. The resulting dataset yields the first long time-series PMFs distribution maps of Northeast China with high spatial resolution across the past four decades. The dataset offers a reusable resource for studies of agricultural land management, environmental assessment and sustainable development in Northeast China black soil region, and can support applications in other regions with similar agricultural practices.