Mechanism analysis and risk assessment of cultivated land loss based on Bayesian network in the urban periphery

Spatial and temporal features of the cultivated land loss in the urban periphery are of major significance for food security. The influencing factors can also dominate the resource allocation of the cultivated land. This study aims to analyze the spatiotemporal features of the cultivated land loss using the land use change survey data and GIS spatial analysis. A case study was taken of the Yinzhou District in Ningbo City, Zhejiang Province, China. Then, a Bayesian network model was constructed to identify the key influencing factors on the cultivated land loss using the node importance. The risk of the cultivated loss was also predicted using diagnostic analysis and probabilistic inference. The results indicate that: 1) A total of 11 153 original cultivated land patches were lost from 2009 to 2021. And 50 345 new patches emerged to cover a cumulative area of 8,599.54 hectares after transformation and fragmentation, which accounted for 40.6% of the total cultivated land area in 2009. Only 59.4% of the cultivated land remained under stable utilization. The primary pathways of the cultivated land loss were converted to construct the cultivated and forest land. Specifically, approximately 38.7% of the lost cultivated land was converted into construction land. While about 26.2% was transformed into forest land. Such a transformation pattern represented the urban expansion, ecological-economic reconstruction, and the characteristic industries of the cultivated land use. 2) The driving force of the agricultural structure had a significant effect on the probability of the loss of cultivated land, followed by the driving force of the construction occupation. There was a relatively low importance of the cultivated land functions on the cultivated land loss. While decision-making factors had a significant impact on the probability of the cultivated land. Especially, the subsidies for the cultivated land also played a relatively strong role in the loss of cultivated land. There were the polarized patterns of the agricultural protection subsidy factor. The agricultural structure was significantly impacted by farmland protection subsidies, either very large or extremely low. “Threshold Effect” in behavioral economics was attributed to the polarization of the agricultural protection subsidies. Behavioral change only occurred when the subsidy level surpassed the “significance threshold” in farmers’ minds. Besides, there was the relatively small influence of the neighborhood and natural factors on the probability of cultivated land loss. 3) The risk of the cultivated land exhibited spatial differentiation. Urbanization frontier regions (like Xiaying, Panhuo, and Meixu) were geographically associated with the high-risk locations. The sub-high-risk locations were observed in the intersection of Shounan, Yunlong, and Dongqian Lake, as well as the shoreline of Zhanqi and Xianxiang. The majority of the medium-risk locations were found in the transitional spatial units, including Wuxiang, Tangxi, and Hengxi. The primary locations of the low-risk areas were found in the towns and villages with flat terrain and an endowment of cultivated land. Overall, the high risks were observed in both favorable ecological and advantageous geographic locations, due to the multiple overlapping pressures. In conclusion, some strategies must be employed according to the varying risk levels of the cultivated land. While the various measures were concurrently implemented to enhance the interactions among the management, the market, and agricultural entities in the cultivated land use. Thereby, the findings can also provide a strong reference for the decision-making on the cultivated land resources.