| The maize-soybean intercropping system plays an important role in ensuring the food security in China.This system is improving soybean acreage and yield and ensuring there is no significant reduction in maize yield.Precision agriculture requires the managers to get the information of crop in timely,accurately and non-destructive.Therefore,the hyperspectral remote that could obtain the information about the target in rapidly and accurately has huge advantages.Other researchers have use the hyperspectral reflectance accurately estimate the agronomic parameters,but there are few results about the crop in the intercropping system.Therefore,it is important to find the effects of different nitrogen application levels on the agricultural parameters,yield and quality of maize under intercropping system and to construct the mathematical model between agronomic parameters and grain yield and quality.What`s more,hyperspectral remote in scientific and timely and accurate plays a very positive role in guiding the rational application of nitrogen fertilizer and realizing modern agriculture with high yield,high quality and ecology.This study is conducted based on the different nitrogen application levels for maize under maize-soybean intercropping system to find the mathematical relationship between agronomic parameters and grain yield and quality and to construct the model for assessing the agronomic parameters.The results showed that with the increase of nitrogen application,leaf area index,content and density of photosynthetic pigment,content and density of carbon and nitrogen,and content and density of soluble sugar of maize increased.Finally,the grain yield of maize also increased with the increase of nitrogen application.Simultaneously,the grain protein content and starch content also increased with increased nitrogen,but the grain fat content decreased firstly and then increased with the increase of nitrogen application.Through the stepwise regression analysis between agronomic parameters and grain yield and quality,it is indicated that the grain yield model constructed by three parameters of carotenoid content,nitrogen content and soluble sugar content in milky period.The coefficient of determination?R2?of the model reaches 0.993,and the root mean square error?RMSE?is only 14.194 g.The model of percentage and content of grain protein is composed by three indices contain nitrogen density?filling-bell mouthed?,soluble sugar content?filling?and carotenoid density?bell mouthed?,and two indices contain chlorophyll a content?milky?and nitrogen content?milky?,respectively.The coefficients of determination and root mean square errors are 0.997,0.92 and 0.168,1.25 g,respectively.The model of percentage and content of grain fat is determined by three indices contain nitrogen density?milky?,carotenoid density?milky-trumpet?and carbon percentage?milky-trumpet?,and three indices contain carbon density?milky?,carotenoid density?milky-trumpet?and chlorophyll a/b?milky–filling?,respectively.The coefficients of determination and root mean square errors are 0.952,0.919 and 0.233,0.486 g.The model of percentage and content of grain starch is determined by two indices contain carotenoid content?milky-trumpet?and soluble sugar content?filling?and three indices contain carotenoid content?milky?,nitrogen content?milky?and chlorophyll a content?milky-trumpet?.The determination coefficients and root mean square errors were 0.766,0.982 and 1.265,6.630 g,respectively.The hyperspectral reflectance of maize canopy decreased with the increase of nitrogen application in the vicinity of“green peak”,and increased with the increase of nitrogen application in the near-infrared platform.Compared with other treatments,the red edge position of no nitrogen treatment?T1?occurred“blue shift”,and the red edge amplitude increased with the increase of nitrogen application at any time.The red edge position of each treatment in the bell-mouthed occurred“red shift”compared with the seedling stage.By analyzing the correlation between agronomic parameters and the hyperspectral primary reflectance and the first-order differential of a single band,it is found that after the first-order differential processing,the number of sensitive bands of each parameter is significantly decreased and more concentrated.In addition,the vegetation index,the new vegetation index based on the free combination of bands,and the wavelet coefficients obtained by continuous wavelet transform were analyzed and compared for assessing of agronomic parameters:chlorophyll a content and carotenoid content were computed with db2?D744,8?and gaus2?D786,6?,respectively.The carotenoid density,carbon percentage,carbon density,nitrogen content,nitrogen density and soluble sugar content are freely combined with the first-order differential passband based on hyperspectral reflectance to obtain the vegetation index DVI?D753,D981?,RVI?D370,D661?,RVI?D737,D730?,RVI?D539,D743?,DVI?D756,D871?and NDVI?D570,D532?have the highest correlation.The linear and nonlinear regression analysis is used to construct each parameter estimation model through the optimal vegetation index or wavelet coefficient is obtained through correlation analysis.In addition,the continuous wavelet transform processing is combined with UVE to select the bands,then using perform partial least squares regression to construct the models.The comparative analysis of the constructed hyperspectral estimation model and verification found that the partial least squares regression model constructed by combining continuous wavelet transform processing with UVE has higher performance than the model constructed using vegetation index or wavelet coefficient.The coefficient of determination?R2?of chlorophyll a content,carotenoid content,carotenoid density,chlorophyll a/b,carbon density,nitrogen content,nitrogen density and soluble sugar content model were0.968,0.925,0.981,0.968,0.99,0.963,0.431,0.99,respectively.The corresponding root mean square error?RMSE?were 14.605 mg,2.19 mg,2.948 mg/m2,0.032,0.047g/m2,0.076 g,0.161 g/m2 and 0.002 g,respectively.In summary,the using of maize canopy hyperspectral reflectance can achieve quantitative estimation of agronomic parameters.In addition,we realized hyperspectral estimation of maize grain yield and quality by combining the mathematical relationship between maize grain yield and quality and agronomic parameters.The remote sensing could provide technology for achieving high yield,quality and ecology of maize production.In addition,it will contribute to the development of modern precision agriculture in China. |
PDF Full Text Request