Remote Sensing Identification and Spatiotemporal Evolution of Cultivated Land Non-Agriculturalization in Qixian County Based on Google Earth Engine

Abstract

Cultivated land non-agriculturalization is an important manifestation of county-level land use transition. Timely identification of its spatial distribution and phased evolution is essential for farmland protection and rational land resource allocation. Taking Qixian County in Henan Province, China, as the study area, this paper used Sentinel-2 imagery from 2020, 2023, and 2025 on the Google Earth Engine platform. Percentile compositing was employed to generate multi-temporal feature images, and a random forest classifier was used to extract land use information. A unified cultivated land mask for 2020 was then used as a constraint, and post-classification comparison was applied to identify the conversion of cultivated land to non-agricultural uses. The spatiotemporal evolution of non-agriculturalization was analyzed in terms of area, transition type, and spatial distribution. The results show that the overall accuracy of land use classification in the three periods was higher than 83%, and the Kappa coefficient was above 0.77. The areas of cultivated land non-agriculturalization in 2020–2023, 2023–2025, and 2020–2025 were 54.5241 km², 10.3185 km², and 40.6306 km², respectively. Conversion to water dominated in 2020–2023, whereas conversion to orchard/woodland became dominant later. Spatially, non-agriculturalization was concentrated around the county seat, township centers, and transport corridors.