AbstractLand degradation due to mismanagement is widespread globally and may threaten the achievement of several UN Sustainable Development Goals. Yet the differences in land productivity degradation under various land management patterns (land sparing vs. land sharing) are poorly known. In this research, we used remote sensing data to develop a machine learning model for assessing the risk of land productivity degradation and interpreted the model using state‐of‐the‐art interpretable artificial intelligence techniques. In 2018, the risk level of land productivity degradation in the agricultural production space of the Yangtze River Delta urban agglomeration (YRD) was 0.230. More than half of the area was at low risk (68.19% of the area), mainly in mountainous and hilly areas. The degradation risk of the land sharing management pattern is lower than that of the land sparing pattern, but there are significant differences among provinces/municipalities. The four most influential factors for land productivity degradation in YRD were Normalized Vegetation Difference Index, nighttime light, elevation, and nitrogen deposition, which together explained 72.75% of the degradation risk. This study provides a methodological framework for land degradation governance in emerging urban agglomerations. It strongly recommends that policymakers explore locally appropriate land management patterns based on regional contexts.
Read full abstract