UAV-based Multispectral Remote Sensing Research On Monitoring And Regulation Of Nitrogen Nutrition Indicators In Spring Maize

Rapid,non-destructive and accurate monitoring of nitrogen nutrition during the crop reproductive period is of great significance to achieve precise fertilization management of maize,reduce environmental pollution and improve resource utilization.In this study,the nitrogen fertilizer gradient test and field nitrogen regulation test were set up based on water and fertilizer integration conditions,with northeastern spring maize as the research object and the former Xi’aili village test field in Tongliao City as the research area,and the maize canopy spectral data under different treatments in two years were obtained by using UAV with multi-spectrum,and the above-ground biomass,leaf area index,plant nitrogen uptake and specific leaf nitrogen were monitored by fusing 19 vegetation indices through random forest algorithm.This study provides a theoretical basis for improving the accuracy of monitoring maize N indicators.The main findings of the study are as follows:(1)Nitrogen application greatly influences spring corn yield,biomass,leaf area index,plant nitrogen uptake,and specific leaf nitrogen,and appropriate nitrogen application can lead to higher corn yields.Fertilizer application was more efective in promoting corn growth and development than a single application.The results of this study showed that within the range of 270-315 kg/ha of total N application in this area,six drip fertilization follow-up applications during the whole reproductive period could achieve yields in the range of 16.11-18.69 t/ha.(2)The inverse model of nitrogen nutrient indexes during the nutrient growth period constructed separately using the random forest algorithm incorporating 19 vegetation indices had high model prediction accuracy for each index,and the accuracy of aboveground biomass inversion(nitrogen gradient:R2=0.83,regulation:R2=0.88),leaf area index inversion(nitrogen gradient:R2=0.9,nitrogen gradient:R2=0.95),plant nitrogen uptake inversion accuracy(nitrogen gradient:R2=0.70,modulation:R2=0.78)and specific leaf nitrogen inversion accuracy(nitrogen gradient:R2=0.61,nitrogen gradient:R2=0.68).Based on the random forest algorithm and multiple vegetation indices can accurately monitor the nitrogen nutrition index of maize during the nutrient growth period,the random forest algorithm constructed the highest accuracy of leaf area index model in different experiments,and the results of this study show that the random forest algorithm has good potential for development in leaf area index prediction.(3)Using the models constructed under the nitrogen gradient test for inversion prediction under the regulation test,it was found that the prediction accuracy of the four models under the gradient test was poor,and the inversion accuracy of above-ground biomass(R2=0.08),leaf area index(R2=0.29)and plant nitrogen uptake(R2=0.43)were extremely poorer than the inversion accuracy of the leaf nitrogen model.The results of this study showed that the inverse model constructed under the traditional nitrogen fertilizer gradient test was less effective in the field regulation test,and the monitoring model based on the perspective of the nitrogen fertilizer gradient test could not accurately monitor the nitrogen nutrient index in the field.(4)By analyzing the correlation between vegetation index and each N nutrient index and its importance in the model,the results showed that structural insensitive pigment index(SIPI)had the highest correlation with aboveground biomass,leaf area index,plant nitrogen uptake and specific leaf nitrogen in the nitrogen fertilization gradient test,and this index had the highest importance in the four N nutrient index models;in the regulation test,structural insensitive pigment index had a higher correlation with the four indexes,and The structural insensitive pigment index had the highest correlation with the four indices in the regulation experiment,and the highest importance in the leaf area index and specific leaf nitrogen inversion model,and the highest importance in the aboveground biomass inversion model.The results showed that the structural insensitive pigment index could be used to monitor nitrogen nutrient.(5)According to the importance of vegetation indices in the model,the optimal set of vegetation index variables for constructing the nitrogen nutrient index model in the nitrogen gradient experiment was found to be SIPI,MSR,RVI1,RVI2,NDVI,MTCI,GNDVI,PPR,NDRE,NDVI by gradually reducing the vegetation index variables using the random forest algorithm;the optimal set of vegetation index variables for constructing the plant nitrogen uptake model in the regulation experiment was found to be:PPR,GNDVI,CIred-edge,RVI2,OSAVI,SIPI.The optimal vegetation index variable groups in the control experiment were PPR,GNDVI,CIred-edge,RVI2,OSAVI,and SIPI.Reducing the number of vegetation indices in the model in the control experiment had no effect on the accuracy of aboveground biomass,leaf area index,and specific leaf nitrogen model.The results of this study showed that reducing the number of variables for constructing models not only had less effect on the model accuracy,but also could reduce the model training test time.