Nondestructive Diagnosis Of Tree Nitrogen Status In Zanthoxylum Armatum V. Novemfolius Based On Integrated Imaging Of UAV-Borne Hyperspectral And Lidar Technology

Nitrogen is the basis for the growth and development,yield formation and quality improvement of Zanthoxylum armatum.It is of great significance to clarify the nitrogen nutrition status of Zanthoxylum armatum in the key growth period,which is of great significance to accurately guide the nitrogen nutrition management of it.In recent years,with the development of remote sensing technology,UAV equipped with multi-sensor has been widely used in the extraction of vegetation parameters.Among them,optical remote sensing can obtain the physiological and biochemical spectral characteristics of vegetation canopy,while the laser pulse emitted by Li DAR can penetrate the vegetation canopy and obtain its accurate three-dimensional structure information.There are certain complementary advantages between them.In this study,Zanthoxylum armatum（8-year-old）in Fruiting period was taken as the research object.Firstly,the nutrient deficiency and Nitrogen level test was set up to clarify the law of nitrogen demand during the growth and development of Zanthoxylum bungeanum,and determine the Vigorous growing and Fruiting period as the key diagnostic period.Secondly,the inversion models of CNC,CNA and AGNA based on the fusion of single hyperspectral data,single LIDAR point cloud data and hyperspectral Li DAR data are established respectively.At the same time,the influence of flight altitude on the inversion accuracy was analyzed.The non-destructive diagnosis model of optimal nitrogen nutrition of Zanthoxylum bungeanum at Vigorous growing and Fruiting stages were established.The main results were as follows:（1）The main limiting nutrient elements of adult Zanthoxylum bungeanum were N>P>K.The combined application of N,P and K can achieve high and excellent yield of Zanthoxylum bungeanum.The effect of balanced fertilization was the best,while the yield of N,P and K fertilizer was reduced by 17.93%,6.55%and 5.86%respectively.At the same time,the contents of volatile oil,hemp flavor,alcohol soluble extract and ether extract of Zanthoxylum bungeanum were reduced by 14.23%～26.74%,12.41%～22.78%,8.81%～20.68%and 17.47%～31.97%respectively.It showed that N was the first limiting factor to limit the high and excellent yield of Zanthoxylum bungeanum.In addition,the nitrogen concentration in each part of Zanthoxylum bungeanum tree decreased gradually with time,and reached the lowest level in the Mature period.The CNC was relatively stable in the Vigorous growing-Fruit period,with an amplitude of 32.82 and 34.13 g·kg^-1,respectively.The demand for nitrogen was large in the Shooting-Vigorous growing period and Fruiting-Mature period.The CNA accounts for 40.64%and 32.48%of its annual net accumulation,and the AGNA accounts for 46.64%and 33.57%of its annual net accumulation,respectively.In the Vigorous growing and Fruiting periods,CNC,CNA and AGNA had a very significant correlation with yield and quality（P<0.01）.The correlation coefficient was higher than that in other periods.The Vigorous growing and Fruiting periods of Zanthoxylum bungeanum were the most important periods for nitrogen nutrition diagnosis and recommendation of nitrogen application.（2）Taking the CNC,CNA and AGNA of Zanthoxylum bungeanum in the Vigorous growing and Fruiting periods as diagnostic indexes,the spectral reflectance of Zanthoxylum bungeanum canopy at flight altitude of 60,80 and 100 m was obtained by UAV Hyperspectral.Through correlation analysis and reconstruction of vegetation index（NDSI,RSI and DSI）,it was clear that the sensitive bands of Zanthoxylum bungeanum CNC were R₅₆₀,R₆₉₀,R₇₃₂ and R₈₇₉,The best reconstructed vegetation index in Vigorous growing and Fruiting periods were NDSI_{（879,732）}and NDSI_{（560,690）};The sensitive bands of CNA were R₅₂₇,R₇₁₁,R₇₂₃ and R₉₈₆.The best reconstructed vegetation index in Vigorous growing and Fruiting periods were NDSI_{（986,711）}and NDSI_{（723,527）};The sensitive bands of AGNA were R₅₁₅,R₇₁₁,R₇₃₆ and R₉₈₆.The best reconstructed vegetation index in Vigorous growing and Fruiting periods were NDSI_{（986,711）}and NDSI_{（736,515）}.Moreover,the modeling effect of the reconstructed vegetation index was better than that of empirical vegetation index under most conditions.Among them,the NDSI_{（986,711）}-CNA and NDSI_{（986,711）}-AGNA estimation models have the best accuracy at 60 m flight altitude,R² was 0.85 and 0.83 respectively,RMSE was3.46 and 4.46 g·plant^-1 respectively,and MRE was 13.4%and 12.0%respectively.The accuracy of NDSI_{（560,690）}-CNC estimation model at 100 m flight altitude in Fruiting period was the best,R² was 0.88,RMSE was 1.96 g·kg^-1 and MRE was 5.1%.The model R² of NDSI_{（732,879）}-CNC at 100 m flight altitude in Vigorous growing period and CNA and AGNA at 100 m flight altitude in Fruiting period were 0.68,0.60 and0.62 respectively,RMSE were 2.17 g·kg^-1,4.60 and 6.95 g·plant^-1 respectively,and MRE were 5.0%,15.2%and 14.6%respectively.The modeling accuracy was poor.（3）Taking the plant height（PH）,crown width（CD）and canopy volume（CV）as diagnostic indexes at the Vigorous growing and Fruiting periods,the point cloud data of Zanthoxylum bungeanum at flight altitudes of 60,80 and 100 m were obtained by UAV Li DAR,and the model was constructed by single variable linear regression and multivariable multiple linear regression.The results showed that the multiple linear regression model based on PH,CD and CV could significantly improve the estimation accuracy of nitrogen nutrition diagnostic indexes.The characteristic parameters of point cloud were not sensitive to Zanthoxylum bungeanum CNC,and have good fitting accuracy to CNA and AGNA.Among them,MLR（PH+CD+CV）-CNA and MLR（PH+CD+CV）-AGNA models have the best accuracy at 60 m flight altitude,R² was0.72,RMSE was 4.77 and 5.79 g·plant^-1 respectively,and MRE was 18.4%and 15.1%respectively.The accuracy of MLR（PH+CD+CV）-CNA and MLR（PH+CD+CV）-AGNA models was the best at 60 m flight altitude in Fruiting period,R² were 0.74and 0.85,RMSE were 3.71 and 4.36 g·plant^-1,and MRE were 14.9%and 9.8%respectively.（4）Using the Hyperspectral and Li DAR feature information fusion modeling method,the optimal vegetation index（NDSI）and point cloud feature parameters（PH,CD and CV）at different flight altitudes were used to model the nitrogen nutrition diagnosis indexes（CNC,CNA and AGNA）.The results showed that the multiple linear regression model based on hyperspectral and Li DAR fusion derived index had significantly improved the estimation accuracy of nitrogen nutrition diagnosis index than single hyperspectral Vegetation Index and point cloud characteristic parameter model.The accuracy of MLR（NDSI+Li DAR）-CNA and MLR（NDSI+Li DAR）-AGNA models were the best at 60 m,R² were 0.92 and 0.91,RMSE were 2.54 and 3.36g·plant^-1,and MRE were 11.2%and 10.0%,respectively.When the Fruiting period was100 m,the accuracy of MLR（NDSI+lidar）-CNC model was the best,R² was 0.98,RMSE was 0.78 g·kg^-1 and MRE was 1.9%.To sum up,the vegetation information in the key nitrogen nutrition diagnosis period of Zanthoxylum bungeanum was obtained through near earth remote sensing technology of UAV Hyperspectral and Li DAR.The nitrogen nutrition diagnosis model based on hyperspectral and Li DAR information fusion and collaborative inversion was constructed,which had higher accuracy than the nitrogen nutrition inversion model based on single sensor information.It solved the problem of low accuracy of using hyperspectral technology to diagnose the nitrogen nutrition status of trees in the early stage of vegetative growth,and provided a new idea and method for obtaining more accurate nitrogen nutrition inversion in the future.