Scholarlink

<p>Monitoring farmland mulching areas using remote sensing data is essential for assessing agricultural sustainability and plastic pollution. However, large-scale estimation is often hindered by the sparsity and non-probability nature of available satellite samples. To address this, we propose a Bayesian hierarchical model that incorporates spatially structured priors and auxiliary geospatial covariates to infer total mulching coverage at multiple administrative levels. We validate the proposed approach by estimating farmland mulching areas using mulched blocks identified from satellite imagery in Qin’an County of Gansu Province, China. The model generates village-level estimates which are aggregated to town and county-level totals, accompanied by uncertainty quantification derived from Bayesian posterior distributions. The estimated mulching areas show good agreement with ground truth data, demonstrating the accuracy of the model. Additionally, the robustness of the method is confirmed through comparative experiments under varying sampling rates and sampling uniformity. These results highlight the potential of the proposed approach as a novel and effective framework for estimating farmland mulching areas from sparse remote sensing data. Moreover, the ability to provide credible uncertainty assessments supports more informed and data-driven decision-making for policymakers.</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

Bayesian estimation of farmland mulching area from sparse remote sensing observations: a hierarchical spatial modeling framework

  • 期刊论文

  • 发表日期：

    2026年7月7日

查看全文
  • Xinlai Kang,
  • Hongyu Chen,
  • Yanbing Bai,
  • Bohai Zhang

摘要

Monitoring farmland mulching areas using remote sensing data is essential for assessing agricultural sustainability and plastic pollution. However, large-scale estimation is often hindered by the sparsity and non-probability nature of available satellite samples. To address this, we propose a Bayesian hierarchical model that incorporates spatially structured priors and auxiliary geospatial covariates to infer total mulching coverage at multiple administrative levels. We validate the proposed approach by estimating farmland mulching areas using mulched blocks identified from satellite imagery in Qin’an County of Gansu Province, China. The model generates village-level estimates which are aggregated to town and county-level totals, accompanied by uncertainty quantification derived from Bayesian posterior distributions. The estimated mulching areas show good agreement with ground truth data, demonstrating the accuracy of the model. Additionally, the robustness of the method is confirmed through comparative experiments under varying sampling rates and sampling uniformity. These results highlight the potential of the proposed approach as a novel and effective framework for estimating farmland mulching areas from sparse remote sensing data. Moreover, the ability to provide credible uncertainty assessments supports more informed and data-driven decision-making for policymakers.