Monitoring Interannual Variability in Spatial Distribution of Plastic-Mulched Farmland in Black Soil Areas Using Google Earth Engine

Plastic mulch is widely used in agriculture, providing significant benefits. However, improper use can harm farmland ecosystems and threaten food security, especially in black soil regions, which are primary grain-producing areas. Timely and accurate monitoring of spatial and temporal changes in plastic-mulched farmland (PMF) is crucial for controlling mulch pollution and promoting sustainable development. However, dynamic changes of PMF, diversity of background environment, and cloud cover seriously hinder the long-term and large-scale monitoring of PMF. Therefore, this article proposed a new process for monitoring PMF based on Google Earth Engine and multitemporal random forest probability synthesis (MRFPS) algorithms. Taking three typical black soil mulched areas, the random forest performance was compared in five feature scenarios by combining spectral bands, spectral indices, and texture information. The key features and combinations were optimized, and then the MRFPS algorithm was used to minimize the impact of cloud contamination and map the distribution of PMF. The results showed that the sowing period and vigorous growth period were the key periods for PMF identification; the red edge bands, BSI, retrogressive plastic greenhouse index, plastic-mulched citrus index, NDBBI, and EVI were the important features for PMF identification while the texture information had less influence. The classification results had high accuracy with an OA of more than 90%, outperforming other methods. Analysis of the spatial distribution from 2017 to 2023 revealed a continued shrinkage in PMF area, with regional differences in the frequency of PMF, which may be closely related to farming practices and government policies. This study provides essential support for exploring PMF distribution change patterns.