| 摘要: |
| 【 目的】探讨基于特征光谱优化组合的马铃薯叶片SPAD值高光谱估算方法，为实现马铃薯叶片SPAD值精准、高效、大面积监测提供理论依据。【 方法】于2020?2022年分别测定不同供水条件下马铃薯苗期、块茎 形成期和块茎膨大期叶片高光谱数据和SPAD值，提取各生育期敏感波段、一阶微分、光谱吸收特征参数和植被指数4种特征光谱，建立单一类型的特征光谱与SPAD值之间的线性模型；在此基础上，构建基于敏感波段、一阶微分特征波 段、光谱吸收特征参数和敏感植被指数优化组合的SPAD估算模型，并对模型精度进行验证。【 结果】在不同供水条 件下，马铃薯倒四叶片SPAD值随干旱胁迫程度的加剧而显著增加；苗期和块茎形成期的冠层高光谱反射率在充足灌 溉处理下最大，适度胁迫处理次之；而块茎膨大期则在充足灌溉处理下明显高于其他处理，其次是过量灌溉处理。利 用单一敏感波段、一阶微分、光谱吸收特征参数或植被指数的马铃薯叶片SPAD估算模型效果均较差，但利用上述4种特征光谱优化组合建立的马铃薯各生育时期叶片SPAD估算模型均优于单一特征光谱。基于优化组合构建的SPAD值预测模型的精度，在苗期以RF模型（R2=0.630）最优，其训练集和测试集R2分别为0.833 和0.490；在块茎形 成期和块茎膨大期均以Poly模型（R2分别为0.490和0.526）最优，前者的训练集和测试集R2分别为0.661和0.471，后 者的训练集和测试集R2分别为0.538和0.438。【 结论】基于4种特征光谱的优化组合模型对不同供水条件下马铃薯 叶片SPAD值具有较好的预测能力，是估算马铃薯SPAD的一种实时高效方法。 |
| 关键词: 马铃薯 SPAD值 高光谱 特征光谱组合 估测模型 |
| DOI：10. 13207/j. jnwafu. 2026. 07. 012 |
| 分类号: |
| 基金项目:国家自然科学基金项目“基于高光谱的马铃薯水分状况实时监测及节水灌溉推荐”（31960388）；内蒙古自然科学基金项目“基于高光谱特征参数的马铃薯长势关键参数实时监测及精准灌溉研究”（2023LHMS03046） |
|
| Estimation of SPAD value of potato leaves based on the optimized combination model of characteristic spectra |
|
SUYALA Qiqige1, ZOU Cunhui1, YANG Xuanman1, XING Ke1, FAN Mingshou2, TIAN Yanhua3, LIU Guangjing3
|
|
1.College of Grassland Science，Inner Mongolia Agricultural University，Hohhot，Inner Mongolia 010019，China;2.College of Agronomy，Inner Mongolia Agricultural University，Hohhot，Inner Mongolia 010011，China;3.Inner Mongolia Zhongjia Agricultural Biotechnology Co. ，LTD，Ulanqab，Inner Mongolia 011808，China
|
| Abstract: |
| 【 Objective】This research aims to explore the hyperspectral estimation method of SPAD value of potato leaves based on the optimized combination of characteristic spectra，to provide a theoretical basis for the accurate，efficient and large-scale monitoring of the SPAD value of potato leaves.【Method】In this study， the hyperspectral data and SPAD value of potato leaves were measured respectively during the seedling stage，tuber formation stage，and tuber expansion stage from 2020 to 2022. Four types of data，namely sensitive bands，first-order differential characteristic spectra，spectral absorption characteristic parameters，and spectral indices， were recorded in each growth period. The linear models between the single type characteristic spectra and the SPAD values were analyzed. On this basis，the SPAD estimation models were constructed based on the opti?mized combination of the characteristic bands of the original spectra，first-order differential characteristic bands，spectral absorption characteristic parameters，and sensitive spectral indices，and the accuracy of the model was verified.【Result】Under different water supply conditions for soil，the SPAD value of the fourth leaf from the
top of potato plants increased significantly with the intensification of drought stress. During the seedling stage and tuber formation stage，the canopy hyperspectral reflectance was the highest under the sufficient irrigation treatment，followed by that under the moderate stress treatment. During the tuber expansion stage，the canopy hyperspectral reflectance was the highest under the sufficient irrigation treatment，followed by that under the excessive irrigation treatment. The estimation models of SPAD value of potato leaves based on the combination of single sensitive band，first-order differential characteristic spectra，spectral absorption characteristic parameters or indices showed poor performance. However，the estimation models of SPAD values of potato leaves at various growth stages established by the optimized combination of the above four types of characteristic spectra all showed better performance than those based on single characteristic spectrum. For the SPAD value estimation models based on the optimized combinations，during the seedling stage，the RF model（R2=0.630） showed the highest accuracy，with the R2 value being 0. 833 for its training set and 0.490 for its test set. During the tuber formation stage and tuber expansion stage，the Poly model showed the highest accuracy（with R2 values being 0.490 and 0.526，respectively）. For the tuber formation stage，the R2 values of the training set and test set were 0.661 and 0.471，respectively. For the tuber expansion stage，the R2 values of the training set and test set were 0.538 and 0.438，respectively.【Conclusion】The optimized combination of the four characteristic spectra shows a good predictive ability for the SPAD values of potato leaves under different water supply conditions，and it is an efficient real-time method for estimating the SPAD value of potato. |
| Key words: potato SPAD value hyperspectral characteristic spectral combination estimation models |
