Apr 30, 2026
Mapping and analyzing the spatial pattern of farmland abandonment of Zhejiang Province using Google Earth Engine based on multi-source data
Land use, as a key form of human-environment interaction, has sharply and continuously transformed the global land surface. Farmland abandonment (FA) is an extreme manifestation of land use marginalization, exerting both positive and negative impacts on the ecological environment and human well-being. To map the extent of FA in Zhejiang Province and clarify its spatial pattern, driving mechanisms, and ecological-socioeconomic implications, this study employed the Google Earth Engine (GEE) and Geographic Information System (GIS) platforms, integrating multi-source land use/land cover data and Landsat time-series images. We used the random forest (RF) algorithm for land use classification, the land use trajectory method for FA identification, landscape pattern analysis (Fragstats 4) to quantify landscape characteristics, and spatial autocorrelation analysis (Moran’s I and LISA indices) to explore spatial aggregation patterns. The results showed that farmland accounted for 16.32% of the total land use area in Zhejiang Province, equivalent to 1.89 × 10⁴ km². The total area of FA was 1.72 × 10⁸ m², with a farmland abandonment rate of 1.65%. The area of active farmland (AF) was approximately 1.95 × 10⁹ m², with a continuous cultivation rate of 18.69%. The landscape fragmentation, aggregation, and diversity of FA, AF, and fallow land (FL) differed significantly: FA exhibited the most severe fragmentation, while AF had the highest aggregation degree. Spatial autocorrelation analysis revealed that FA and AF both showed dominant high aggregation and low dispersion characteristics; the global Moran’s I index of FA was 0.93 (Z-score = 439.45, p < 0.01), indicating a strong positive spatial autocorrelation. Local spatial autocorrelation (LISA) analysis showed that high-high (HH) and low-low (LL) clusters were the main types, accounting for 38.97% and 37.95% of FA patches (passing p < 0.01 significance test), respectively. Comparative analysis with existing studies confirmed the scientific validity of our results (overall classification accuracy > 95%). Finally, we discussed the ecological and socio-economic driving mechanisms of FA, analyzed its potential threats to food security, proposed targeted policy implications, and identified research limitations and future research directions. Our findings provide a scientific basis for rational land use management, farmland protection, and food security assurance in Zhejiang Province.