Scholarlink

<p>Cultivated land quality (CLQ) is crucial for sustainable agriculture. However, understanding its spatiotemporal dynamics and driving mechanisms remains challenging. This study introduces a novel hybrid framework in which geostatistics, machine learning, and the Geodetector model are integrated to comprehensively assess CLQ changes and their determinants. This framework was applied to 7253 samples from Zhejiang Province (2018–2022), and the results suggested that according to Yangtze River Delta nutrient standards, the soil nutrient status in Zhejiang, characterized by organic matter and available nutrients, was predominantly at medium-to-high levels. Moreover, overall improvement in CLQ was observed but with divergent regional trends: the high-quality lower Yangtze River region experienced a decrease, while the quality in other regions improved. The optimal model (extreme gradient boosting (XGBoost); <i>R</i>² = 0.670) identified natural factors (52.0%) as the primary driver of variation, followed by soil (30.4%) and socioeconomic factors (17.6%). Moreover, the risk detector revealed the nonlinear influences of the dominant drivers: CLQ variation was significantly positively correlated with increases in soil nitrogen, and available potassium and phosphorus, with <i>q</i> values of 0.092, 0.066 and 0.050, respectively. Additionally, the thresholds for meteorological factors were determined (annual ground temperature above 20.9&#xa0;°C and evaporation greater than 1000&#xa0;mm), beyond which the highest positive CLQ variation occurs. This study advances the field by providing a scalable, mechanism-aware framework for spatially targeted land management, moving beyond static mapping to inform dynamic, evidence-based strategies for CLQ protection and enhancement.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Analysis of the spatiotemporal evolution and determinants of cultivated land quality under a hybrid-driven framework integrating multiple machine learning and geostatistical approaches

  • 期刊论文

  • 发表日期：

    2026年4月1日

查看全文
  • Meiling Sheng,
  • Zhaohan Lou,
  • Mingtao Xiang,
  • Rui Xiao,
  • Zhouqiao Ren,
  • Xiaonan Lv,
  • Xufeng Fei

摘要

Cultivated land quality (CLQ) is crucial for sustainable agriculture. However, understanding its spatiotemporal dynamics and driving mechanisms remains challenging. This study introduces a novel hybrid framework in which geostatistics, machine learning, and the Geodetector model are integrated to comprehensively assess CLQ changes and their determinants. This framework was applied to 7253 samples from Zhejiang Province (2018–2022), and the results suggested that according to Yangtze River Delta nutrient standards, the soil nutrient status in Zhejiang, characterized by organic matter and available nutrients, was predominantly at medium-to-high levels. Moreover, overall improvement in CLQ was observed but with divergent regional trends: the high-quality lower Yangtze River region experienced a decrease, while the quality in other regions improved. The optimal model (extreme gradient boosting (XGBoost); R² = 0.670) identified natural factors (52.0%) as the primary driver of variation, followed by soil (30.4%) and socioeconomic factors (17.6%). Moreover, the risk detector revealed the nonlinear influences of the dominant drivers: CLQ variation was significantly positively correlated with increases in soil nitrogen, and available potassium and phosphorus, with q values of 0.092, 0.066 and 0.050, respectively. Additionally, the thresholds for meteorological factors were determined (annual ground temperature above 20.9 °C and evaporation greater than 1000 mm), beyond which the highest positive CLQ variation occurs. This study advances the field by providing a scalable, mechanism-aware framework for spatially targeted land management, moving beyond static mapping to inform dynamic, evidence-based strategies for CLQ protection and enhancement.