Scholarlink

<p>In the dynamic landscape of agricultural forecasting, the utilization of Machine Learning (ML) algorithms has proven superior in unveiling intricate nonlinear relationships compared to classical statistical methodologies. This research aims to predict wheat yield by harnessing the power of ML techniques, employing remote sensing and climatic data to safeguard the food security. Leveraging four remote sensing indices (GNDVI, NDVI, EVI, SAVI) alongside four climatic variables (T<sub>max</sub>, T<sub>min</sub>, PPT, WS) and one reconnaissance drought index (RDI), the study explores eight distinct model combinations within two scenarios—one encompassing RDI (Scenario-1) and the other excluding it (Scenario-2). Employing Random Forest (RF) as a nonlinear ML algorithm and LASSO as a linear model, the research seeks to identify the most effective combination and ML algorithm for wheat yield prediction. The results uncover compelling insights: in Scenario-1, RF regression excelled notably with the model combination (SAVI + RDI + WS), yielding an impressive R<sup>2</sup> of 0.82 and RMSE of 1.87. Likewise, in Scenario-2, RF surpassed LASSO performance, showcasing the (SAVI + T<sub>max</sub> + T<sub>min</sub> + PPT + PET + WS) model combination with the highest R<sup>2</sup> of 0.88 and the lowest RMSE of 1.58, closely followed by (GNDVI + T<sub>max</sub> + T<sub>min</sub> + PPT + PET + WS; R<sup>2</sup> = 0.83). Remarkably, linear LASSO demonstrated comparable performance with RF in both scenarios, displaying R<sup>2</sup> values ranging from 0.66 to 0.78 in Scenario-1 and 0.71 to 0.81 in Scenario-2. Based on these findings, we recommend using Scenario-2 for wheat yield prediction, integrating climate data and vegetation indices for more reliable forecasting. Policymakers and agricultural planners can leverage these insights to implement data-driven strategies for enhancing food security, optimizing resource allocation, and mitigating climate-induced risks in wheat production. This study highlights the critical role of ML in improving agricultural decision-making, offering a scalable and precise approach for yield prediction to support sustainable food production and climate resilience strategies.</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

Enhancing food security: machine learning-based wheat yield prediction using remote sensing and climate data in Pakistan

  • 期刊论文

  • 发表日期：

    2025年4月25日

查看全文
  • Nadeem Ahmed,
  • Syed Amer Mahmood,
  • Muhammad Haseeb,
  • Zainab Tahir,
  • Saira Batool

摘要

In the dynamic landscape of agricultural forecasting, the utilization of Machine Learning (ML) algorithms has proven superior in unveiling intricate nonlinear relationships compared to classical statistical methodologies. This research aims to predict wheat yield by harnessing the power of ML techniques, employing remote sensing and climatic data to safeguard the food security. Leveraging four remote sensing indices (GNDVI, NDVI, EVI, SAVI) alongside four climatic variables (Tmax, Tmin, PPT, WS) and one reconnaissance drought index (RDI), the study explores eight distinct model combinations within two scenarios—one encompassing RDI (Scenario-1) and the other excluding it (Scenario-2). Employing Random Forest (RF) as a nonlinear ML algorithm and LASSO as a linear model, the research seeks to identify the most effective combination and ML algorithm for wheat yield prediction. The results uncover compelling insights: in Scenario-1, RF regression excelled notably with the model combination (SAVI + RDI + WS), yielding an impressive R2 of 0.82 and RMSE of 1.87. Likewise, in Scenario-2, RF surpassed LASSO performance, showcasing the (SAVI + Tmax + Tmin + PPT + PET + WS) model combination with the highest R2 of 0.88 and the lowest RMSE of 1.58, closely followed by (GNDVI + Tmax + Tmin + PPT + PET + WS; R2 = 0.83). Remarkably, linear LASSO demonstrated comparable performance with RF in both scenarios, displaying R2 values ranging from 0.66 to 0.78 in Scenario-1 and 0.71 to 0.81 in Scenario-2. Based on these findings, we recommend using Scenario-2 for wheat yield prediction, integrating climate data and vegetation indices for more reliable forecasting. Policymakers and agricultural planners can leverage these insights to implement data-driven strategies for enhancing food security, optimizing resource allocation, and mitigating climate-induced risks in wheat production. This study highlights the critical role of ML in improving agricultural decision-making, offering a scalable and precise approach for yield prediction to support sustainable food production and climate resilience strategies.