基于高光谱遥感的油菜叶片氮磷养分含量诊断  

Diagnosis of nitrogen and phosphorus nutrient content in rapeseed leaves based on hyperspectral remote sensing

作　　者：王清华 朱格格 方雯 刘诗诗[1] 鲁剑巍[1] WANG Qing-Hua;ZHU Ge-Ge;FANG Wen;LIU Shi-Shi;LU Jian-Wei(College of Resources and Environment,Huazhong Agricultural University/Key Laboratory of Arable Land Conservation(Middle and Lower Reaches of Yangtze River),Ministry of Agriculture and Rural Affairs,Wuhan 430070,Hubei,China)

机构地区：[1]华中农业大学资源与环境学院/农业农村部长江中下游耕地保育重点实验室,湖北武汉430070

出　　处：《作物学报》2025年第5期1326-1337,共12页Acta Agronomica Sinica

基　　金：国家自然科学基金项目(42171350);国家重点研发计划项目(2021YFD1600503)资助。

摘　　要：利用高光谱遥感技术准确、无损地诊断油菜氮磷养分亏缺,能够为精准施肥提供依据。本研究以多点、多年田间试验测定的越冬期冬油菜叶片氮含量(leaf nitrogen concentration,LNC)、叶片磷含量(leaf phosphorus concentration, LPC)、产量和冠层反射光谱为基础,利用竞争性自适应重加权平均算法、无信息变量消除法、连续投影算法筛选对LNC、LPC敏感的特征波段,基于筛选的波段利用偏最小二乘回归构建基于原初光谱和一阶微分光谱的LNC、LPC估测模型。根据养分含量估测结果结合研究区的氮营养指数(nitrogen nutrition index,NNI)和磷营养指数(phosphorous nutrition index, PNI)进行油菜养分亏缺诊断。结果表明,筛选出的油菜越冬期LNC、LPC特征波段主要集中在400~460 nm、650~730 nm、1140~1210 nm、2240~2370 nm和650~730 nm、2100~2310 nm。基于一阶微分光谱和无信息变量消除法的模型其估测精度要优于其他模型,在测试集上该模型也能准确估测油菜LNC(R^(2)=0.773,RMSE=0.528%)和LPC (R^(2)=0.785, RMSE=0.09%)。同时,本研究利用田间试验产量数据确定了油菜越冬期NNI和PNI的阈值,分别为1.20和0.75。基于高光谱遥感估测的LNC和LPC,进一步计算NNI和PNC,能够对油菜越冬期的养分亏缺进行诊断,为油菜生产可持续发展提供新的技术。Hyperspectral remote sensing technology provides an accurate and non-destructive method for diagnosing nitrogen(N)and phosphorus(P)deficiencies in rapeseed,laying the groundwork for precision fertilization.This study utilized multi-site,multi-year field trials to collect data on leaf nitrogen concentration(LNC),leaf phosphorus concentration(LPC),yield,and the canopy reflectance spectrum of winter rapeseed during the overwintering period.Feature bands sensitive to LNC and LPC were identified using competitive adaptive reweighted sampling(CARS),successive projections algorithm(SPA),and the elimination of non-informative variables(UVE).Partial least squares regression(PLSR)models were constructed to estimate LNC and LPC based on both the original spectrum and the first-order derivative spectrum.Nutrient deficiency diagnosis was achieved by integrating the nitrogen nutrition index(NNI)and phosphorus nutrition index(PNI)derived from the estimated nutrient concentrations.The results revealed that the characteristic bands for LNC and LPC were primarily concentrated in the ranges of 400-460 nm,650-730 nm,1140-1210 nm,and 2240-2370 nm for LNC,and 650-730 nm,2100-2310 nm for LPC.The model based on the first-order derivative spectrum and the UVE method demonstrated superior accuracy compared to other models.In the test set,the model achieved high estimation accuracy for LNC(R^(2)=0.773,RMSE=0.528%)and LPC(R^(2)=0.785,RMSE=0.09%).Threshold values for NNI and PNI during the overwintering period were established using yield data from field trials,which were 1.20 and 0.75,respectively.By employing hyperspectral remote sensing to estimate LNC and LPC,subsequent calculations of NNI and PNI can effectively diagnose nutrient deficiencies in rapeseed during the overwintering period.This approach provides a novel technological solution for the sustainable development of rapeseed production.

关 键 词：冬油菜 高光谱遥感 偏最小二乘 波段选择 叶片氮磷含量 

分 类 号：S127[农业科学—农业基础科学]