Hyperspectral Characteristics And Response On Nitrogen Nutrition Diagnosis For Different Position Leaves Of Winter Wheat

The winter wheat yield production is highly related to the food security.In order to obtain the high winter wheat production,excessive nitrogen fertilizers were always applied in current years,which have caused serious environmental pollution and other problems.Therefore,timely monitor of crop nitrogen nutrition in winter wheat will supply some guidance on scientific and reasonable fertilize application,and it is significant to food production and environmental protection.Currently,the hyperspectral technology has been realized the rapid acquisition of canopy information.However,the accuracy and stability of the prediction model can be directly influenced by the accuracy of spectral information.In order to achieve accurate spectral information and improve the predictive accuracy of nitrogen nutrition,the response changes and relationships of SPAD,nitrogen content,canopy spectrum and leaf spectrum were observed under different nitrogen rates in the study.The spectral response sensitive bands of SPAD value and nitrogen content were found,and the predictive models were also constructed and compared by using partial least squares(PLS)method based on the canopy spectra and leaf spectra.Moreover,the pretreatment methods and their multi-pretrements combinations were analyzed to improve the model performance.Therefore,the main conclusions of this thesis were as follows:(1)The SPAD value of winter wheat was a parabolic trend during the whole growth period,and there were the highest SPAD value during the flowering period.The SPAD value declined with the order:the 1st leaf>2nd leaf>3rd leaf at jointing stage,while the SPAD value of the 1st leaf was highest than that in the 2nd and 3rd leaf at other stage with large decreased in each leaf position.The leaf nitrogen content of the 2nd leaf at booting stage was the highest than that in the 1st and 3rd leaf,while declined with the order:the 1st leaf>2nd leaf>3rd leaf from the top at other growth stages.Leaf nitrogen content in the 2nd and 3rd leaf was sensitive to the increased nitrogen supply.Plant nitrogen content decreased with the growth processes and increased with increasing nitrogen application rates.The leaf nitrogen content was significantly correlated with the total nitrogen content of the plant.(2)The canopy spectral reflectance of winter wheat decreased with the increase of the nitrogen rates in the visible region,while increased in the near-infrared and mid-infrared region.The difference between leaf position reflectance was small with increasing nitrogen application rates in the visible and near infrared region.The correlation between the leaves nitrogen nutrition and the canopy spectra declined with the order:the 1st leaf>2nd leaf>3rd leaf from the top.But the correlation between nitrogen nutrition content and leaf spectra was stable in different leaf position.(3)Most of nitrogen nutrition sensitive wavelengths were mostly located in the green and near infrared area based on the canopy and leaf spectra.The SPAD value prediction model for the 1st,2nd and 3rd leaf based on leaf spectra had a good prediction with R2=0.722,0.691 and 0.681,respectively,and the predictive model of nitrogen content also performed well with R2=0.732,0.719 and 0.706,respectively.1st leaf nitrogen nutrition prediction model had the best effect based on the canopy spectra,but the model accuracy was lower than that of the model based on leaf spectra model.(4)The spectra baseline drift in the visible region was significantly inhibited after baseline correction(BC),multiple scattering correction(MSC),noise correction(NC)and their multi-pretrements combinations pretreatment.The spectral concentration processed with the BC and MSC method was improved in the near infrared region.The pre-treatment of NC method had the highest predictive model performance,and the R2 for the SPAD value model and nitrogen model were 0.686 and 0.683,respectively,and RMSE were 1.150 and 0.326,respectively.The SPAD value model and nitrogen model performance were the lowest by MSC pre-treatment.The estimating model of SPAD value(R2=0.768,RMSE=0.988)based on the combination MSC+NC was the best than other two pretreatment methods combinations,and the model of nitrogen content based on NC+MSC was the best(R2=0.762,RMSE=0.282).However,the best model of SPAD and the nitrogen content model were obtained based on the NC+BC+MSC.It indicated that the spectral pre-treatment obviously increased the predictive accuracy,however,improper pre-treatment would produce a negative effect on the prediction ability.The study indicated the single or the combinations of the spectral processed methods could not always increase the model performance,though under the different pre-treatment method order.