Scholarlink

<p>Accurate estimation of maize photosynthetic pigments using hyperspectral remote sensing is critical for evaluating crop physiological status. This study investigates the photosynthetic pigment estimations for maize at both leaf and canopy scales using spectral data, focusing on the role of Leaf Area Index (LAI) in improving canopy-scale predictions. Field experiments under varying nitrogen levels are conducted. Estimations for carotenoids (Car), chlorophyll (Chl) and the Car/Chl ratio are established at both the canopy and leaf scales, using Multiple Stepwise Regression (MSR), Partial Least Squares Regression (PLSR), Random Forest Regression (RFR) and eXtreme Gradient Boosting (XGBoost). PLSR achieves the highest accuracy for Car and the Car/Chl ratio at the canopy scale (R² = 0.767, NRMSE = 12.624%; R² = 0.798, NRMSE = 16.552%), while RFR excels in estimating Chl at the canopy (R² = 0.887, NRMSE = 9.595%) and all pigments at the leaf scale (R²: 0.57–0.9; NRMSE: 6.59–17.89%). Uncertainties in estimating photosynthetic pigments arise from LAI synergistic spectral parameters, which depends on the degree to which the pigments are affected by the mixed background. The adjustment and optimization of machine learning algorithms can weaken this uncertainty and improve the accuracy of the estimation. SHAP (SHapley Additive exPlanations) analysis reveals that variable importance is not solely determined by correlation strength, highlighting the complexity of model interpretation. These findings lay a foundation for scalable maize pigment monitoring using UAVs or satellite platforms; however, when transitioning from ground-based spectral probes to UAV observations, critical factors affecting reflectance should be accounted for to maximize the utility of the proposed models.</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

Synergistic estimation of photosynthetic pigments in maize based on leaf area index: from leaf spectrum to canopy spectrum

  • 期刊论文

  • 发表日期：

    2025年12月5日

查看全文
  • Zhaohong Lu,
  • Chenyao Yang,
  • Zhonglin Wang,
  • Xianming Tan,
  • Jiawei Zhang,
  • Junxu Chen,
  • Jing Gao,
  • Qi Wang,
  • Jie Zhang,
  • Xintong Wei,
  • Jiaqi Zou,
  • Feng Yang,
  • Wenyu Yang

摘要

Accurate estimation of maize photosynthetic pigments using hyperspectral remote sensing is critical for evaluating crop physiological status. This study investigates the photosynthetic pigment estimations for maize at both leaf and canopy scales using spectral data, focusing on the role of Leaf Area Index (LAI) in improving canopy-scale predictions. Field experiments under varying nitrogen levels are conducted. Estimations for carotenoids (Car), chlorophyll (Chl) and the Car/Chl ratio are established at both the canopy and leaf scales, using Multiple Stepwise Regression (MSR), Partial Least Squares Regression (PLSR), Random Forest Regression (RFR) and eXtreme Gradient Boosting (XGBoost). PLSR achieves the highest accuracy for Car and the Car/Chl ratio at the canopy scale (R² = 0.767, NRMSE = 12.624%; R² = 0.798, NRMSE = 16.552%), while RFR excels in estimating Chl at the canopy (R² = 0.887, NRMSE = 9.595%) and all pigments at the leaf scale (R²: 0.57–0.9; NRMSE: 6.59–17.89%). Uncertainties in estimating photosynthetic pigments arise from LAI synergistic spectral parameters, which depends on the degree to which the pigments are affected by the mixed background. The adjustment and optimization of machine learning algorithms can weaken this uncertainty and improve the accuracy of the estimation. SHAP (SHapley Additive exPlanations) analysis reveals that variable importance is not solely determined by correlation strength, highlighting the complexity of model interpretation. These findings lay a foundation for scalable maize pigment monitoring using UAVs or satellite platforms; however, when transitioning from ground-based spectral probes to UAV observations, critical factors affecting reflectance should be accounted for to maximize the utility of the proposed models.