耕地是人类赖以生存与发展的基础和保障。尽管过去研究已经表明中国近几十年来的快速城市化进程侵占了广泛的耕地，但是目前有关两者间精细尺度作用关系的探索依旧有限。制图层面，现有工作受制于算法精度低、采样效率差等不确定性，缺乏统一、准确的耕地提取方法和数据产品，使得难以正确认识大范围、高频率、细粒度的耕地时空动态及其伴随城市发展的变化情况。实践层面，当前有关城市扩张与耕地流失的研究大多局限于转化面积等指标的简单量化，尤其缺乏其在不同规模城市内演变规律与影响因素的深入讨论。针对上述挑战与不足，本文围绕“较高时空分辨率下中国耕地动态变化与城市扩张发展关系的影响机制是什么？”的科学问题，结合遥感云计算、机器学习、变化检测、地理空间分析、面板数据回归等研究方法，首先提出了一种高分辨率长时序年度耕地制图框架，并基于此生成了中国30米年度耕地数据集（CACD），其次量化了耕地动态的时空变化与转移情况，再者分析了耕地流失与城市扩张的作用联系，最后讨论了耕地-城市土地转换的影响因素。研究结果表明：（1）CACD的总体精度平均为0.93±0.01，变化图层的准确率为0.84。进一步通过跨产品间的比较分析，发现CACD在制图准确性、与统计数据的相关性、以及时空细节方面优于其他同类数据产品。（2）1986-2021年期间，中国耕地总面积基本保持稳定，呈现出时间上 “先增后减再缓增”的趋势和空间上“北增南减、西多东少”的格局。超过三分之一的耕地发生了用地类型转变，另外每年约有1.62万平方千米的耕地发生了撂荒。（3）研究期间全国大约有11万平方千米的耕地被用于城市建设，占到全部新建城市用地的74%。特别地，城市的等级越高，因城市扩张而流失的耕地面积越多，并且两者间的依赖度与贡献度更大。未来至2050年，城市扩张将再导致3.9-4.9万平方千米的耕地流失，其中尤以二、三、四线城市所面临的压力最大。（4）耕地-城市土地转换与经济增长间的倒U型关系符合环境库兹涅茨曲线（EKC）的理论假设。这种关系受到规模效应、构成效应、技术效应、政策效应的多重因素影响，而城市人口数量在其中起到决定性作用。综上所述，本文的研究结果填补了有关精细尺度中国耕地时空动态变化及其与城市扩张发展关系的知识空白，为未来中国耕地资源的统筹利用以及可持续农业发展提供了重要数据支撑、科学依据、以及决策指导。
Cropland is the foundation and guarantee for human survival and development. While previous studies have indicated that China‘s rapid urbanization over recent decades has encroached upon extensive cropland, exploration of the fine-scale interaction between the two remains limited. On the mapping front, existing efforts are constrained by low algorithm precision, poor sampling efficiency, and lack of unified and accurate methods for cropland extraction and data products, which impedes the accurate understanding of large-scale, high-frequency, fine-grained spatiotemporal dynamics of cropland and its associated changes along with urban development. On the practical side, current research on urban expansion and cropland loss mostly confines itself to simple quantification of conversion areas, particularly lacking in-depth discussions on the evolutionary patterns and influencing factors within cities of different scales. In response to the aforementioned challenges and shortcomings, this paper focused on the scientific question of "What are the mechanisms underlying the relationship between cropland dynamic changes and urban expansion development in China at relatively high spatiotemporal resolutions?". Through the integration of remote sensing cloud computing, machine learning, change detection, geospatial analysis, and panel data regression methods, this research firstly proposed a high-resolution, long-term annual cropland dynamic mapping framework and generated the 30-meter annual cropland dataset of China (CACD) based on it. Secondly, this research quantified the spatiotemporal changes and transfers of cropland dynamics. Subsequently, it analyzed the relationship between cropland loss and urban expansion. Finally, this study discussed the influencing factors of cropland-urban land conversion.The research results indicated that: (1) The overall accuracy of CACD averaged 0.93±0.01, with an accuracy of 0.84 for the change layer. Further comparative analysis across products demonstrated that CACD outperformed other existing datasets in terms of mapping accuracy, correlation with statistical data, and spatiotemporal details.(2) From 1986 to 2021, the total cropland area in China remained relatively stable, showing a temporal trend of "initial increase, subsequent decrease, followed by gradual increase" and a spatial pattern of "increase in the north and west, decrease in the south and east". Over one-third of the cropland underwent land use type conversion, with an additional 1.62×104 square kilometers of cropland being abandoned annually.(3) During the study period, approximately 11×104 square kilometers of croplands nationwide were used for urban construction, accounting for 74% of all newly developed urban land. In particular, the higher the city hierarchy, the greater the amount of cropland lost due to urban expansion, with a greater dependency and contribution between the two. It is projected that urban expansion will lead to a further loss of 3.9×104 to 4.9×104 square kilometers of cropland by 2050, with particularly high pressures faced by second-, third-, and fourth-tier cities.(4) The inverted U-shaped relationship between cropland-urban land conversion and economic growth aligned with the theoretical assumptions of the Environmental Kuznets Curve (EKC). This relationship was affected by multifaceted factors including scale effects, composition effects, technological effects, and policy effects, with urban population size playing a decisive role.In conclusion, the research findings of this paper fill the knowledge gap concerning the fine-scale spatiotemporal dynamics of cropland in China and its relationship with urban expansion, which provide crucial data support, scientific foundation, and decision guidance for China‘s future coordinated utilization of cropland resources and sustainable agricultural development.
