A Study On Remote Sensing Inversion Of Nitrogen Content In The Canopy Of Phyllostachys Edulis Forest

Phyllostachys edulis has the characteristics of fast growth,early maturity,versatility and high efficiency,and is an excellent economic bamboo species in the low mountainous areas of southern China.As an indispensable nutrient element in the growth and development of moso bamboo,nitrogen is an important indicator to monitor the growth status of moso bamboo.The quantitative inversion of nitrogen content in moso bamboo forest canopy based on remote sensing data is of great significance for forestry production management.The study takes the moso bamboo forest in Tianbaoyan Reserve,Yongan City,Fujian Province,China as the research object,and uses a combination of UAV multispectral,satellite multispectral and star-machine fusion multispectral images to screen the sensitive characteristic variables of canopy nitrogen content and construct four statistical models(linear stepwise regression,random forest regression,K-nearest neighbor regression,Light GBM regression)and two physical models(PROSAIL-Dy N,INFORM-Dy N)to invert the nitrogen content of the canopy and determine the best inversion model.The study realized the spatial expansion of nitrogen content in the canopy of regional moso bamboo forest from"point"to"surface"and the fusion of multiple types of sensors.The main findings are as follows.(1)The PROSPECT-Dy N model was coupled with the 4SAIL model and the INFORM model to obtain the PROSAIL-Dy N model and the INFORM-Dy N model,respectively.Sensitivity analysis of the model parameters was carried out,and it was found that the leaf protein content(Ccp)sensitive area was mainly in the After 700 nm,the increase of its value causes a decrease in canopy reflectivity,but its overall sensitivity is low,below 20%.(2)Through the correlation analysis between the nitrogen content of the canopy and the various bands of the UAV multispectral image and the vegetation index constructed,the Red Edge,NIR,Optimized Soil Adjusted Vegetation Index(OSAVI),Triangular Vegetation Index(TVI),Soil Adjusted Vegetation Index(SAVI),Renormalized Difference Vegetation Index(RDVI),Difference Vegetation Index(DVI),and Nonlinear Vegetation Index(NLI)were screened out as the spectral characteristic parameters of the canopy nitrogen content,and then based on the filtered spectrum Characteristic parameters to construct a canopy nitrogen content inversion model based on UAV multispectral data.Comparing the accuracy validation results of each model,it was found that the K-nearest neighbor regression model performed the best(R~2=0.6280,RMSE=100.2897μg/cm~2,n RMSE=16.54%,RA=84.97%).(3)Through the correlation analysis between the nitrogen content of the canopy and the various bands of the Sentinel-2B multispectral image and the vegetation index constructed,B6,B7,B8,B8A,Normalized Difference Infrared Index(NDII),Difference Vegetation Index(DVI),Normalized Vegetation Moisture Index(NDWI),and Enhanced Vegetation Index(EVI)were screened out as the spectral characteristic parameters of the canopy nitrogen content,and then based on the filtered spectrum Characteristic parameters to construct a canopy nitrogen content inversion model based on Sentinel-2B multispectral data.Comparing the accuracy validation results of each model,it was found that the K-nearest neighbor regression model performed the best(R~2=0.6812,RMSE=105.6525μg/cm~2,n RMSE=17.43%,RA=83.55%).(4)Based on the GS(Gram-Schmidt)fusion algorithm,the fusion of the visible light single-band of the UAV and the Sentinel-2B satellite multispectral data is realized,and the"star machine"fusion multispectral image is obtained.By constructing and screening out Normalized Difference Infrared Index(NDII),Normalized Vegetation Moisture Index(NDWI)and band combinations((B8-B8A),(B7/B11),(B8/B11),(B6/B11)),(B8/B8A),(B8A/B11))8 characteristic variables,and then constructed an inversion model of bamboo canopy nitrogen content based on fusion multispectral data.Compared with the accuracy verification results of the model,the K-nearest neighbor model performed the best(R~2=0.7024,RMSE=95.8168μg/cm~2,n RMSE=15.80%,RA=86.90%).(5)Comparing the inversion models of canopy nitrogen content from multi-source remote sensing data,it was found that the nonlinear model performed better than the linear and physical models.Among the canopy nitrogen content inversion models constructed from multi-source remote sensing data,the best performing model was the K-nearest neighbor regression model based on star-machine fusion multispectral(R~2=0.7024,RA=86.90%);followed by the K-nearest neighbor regression model based on UAV multispectral(R~2=0.6280,RA=84.97%);followed by the Sentinel-2B multispectral K-nearest neighbor regression model(R~2=0.6812,RA=83.55%).The K-nearest neighbor regression model based on fused multispectral showed some advantages in the inversion study of nitrogen content in moso bamboo forest.In the absence of multispectral or hyperspectral equipment,the use of K-nearest neighbor regression model using UAV visible fused satellite multispectral can be an alternative for regional canopy nitrogen content inversion.Comparing the inversion results of canopy nitrogen content in the same area with multi-source remote sensing data,it was found that the inversion results of fused images were closer to those of UAV multispectral,with the largest proportion of600-700μg/cm~2 image elements,while the largest proportion of 500-600μg/cm~2 image elements were found in the inversion results based on Sentinel-2B images.The differences in the regional canopy nitrogen content inversion results may be related to the lower spatial resolution of satellite data and the greater influence of mixed pixels.