UAV-based Multispectral Multi-feature Parametric Inversion Of Biomass Of Cenchrus Fungigraminus Z.X.Lin & D.M.Lin & S.R.Lan Sp.Nov.under Nitrogen Fertilizer Treatment

In recent years,Giant Juncao（Cenchrus fungigraminus Z.X.Lin&D.M.Lin&S.R.Lan sp.nov.）has played an important role in ecological governance and other aspects.Because of its large biomass,rapid growth,wide adaptability,strong resistance,no biological invasion risk and other characteristics,Giant Juncao has been widely used in ecological governance applications,creating substantial ecological benefits.Biomass can not only be used for estimating the yield of Giant Juncao,but also for monitoring its growth.And SPAD（soil and plant analyzer development）can evaluate the health status of plants.However,measuring the aboveground indicators of Giant Juncao biomass is time-consuming and labor-intensive.Remote sensing has the characteristics of high efficiency,non-contact,large area,time-saving,and labor-saving.However,in the process of biomass inversion,the vegetation index has certain limitations.It is necessary to compensate for the shortcomings of the vegetation index through various feature parameters.Moreover,there are few remote sensing studies on the characteristics of Giant Juncao,which has the characteristics of high height,high coverage,regeneration,and tillering.In this study,Giant Juncao was used as a material and planted in the experimental field of Yanping District,Nanping City.The Giant Juncao was harvested during the seedling stage（2022-06-28）,tillering stage（2022-07-31）,tillering stage（2022-09-05）,and jointing stage（2022-10-01）for regeneration and no cutting（2022-12-14）.The experiment used a randomized block design and set three nitrogen application gradients:0,150,and 300 Kg·hm^-2,denoted as N1,N2,and N3,respectively.And use the DJI Elf Four multispectral camera to obtain multiple remote-sensing band images.Based on measured ground data,analyze the morphological dynamics of Giant Juncao,explore the changes in canopy reflectance of Giant Juncao,and the modeling properties between various feature parameters（86 vegetation indexes,40 image textures,unmanned aerial vehicle extraction of canopy height and coverage）and SPAD and biomass of Giant Juncao canopy.Use four modeling methods,including univariate linear regression（LM）,generalized additive model（GAM）,Support vector machine（SVM）,and random forest（RF）were used to selected the model with the highest accuracy.The model was applied to the growth dynamics of Giant Juncao,and it was clear that the model had good applicability.And establish a SPAD inversion model to monitor the growth of Giant Juncao.The research results are as follows:（1）The nitrogen application rate has a significant impact on the agronomic parameters of Giant Juncao.During the final growth period,the SPAD,N,natural plant height,and biomass of the community Giant Juncao under N3 growth conditions were 6.67,18.9 mg·g^-1,87.73 cm,and 70.47 kg higher than those under N1 conditions,respectively.The most relevant indicator for the total fresh weight of the community with different growth days is the natural plant height,with a correlation coefficient of 0.86（p<0.01）;it maintains a high correlation with natural plant height and SPAD in all growth days.Considering the experimental needs,using the total fresh weight of the community as the aboveground biomass（AGB）of the Giant Juncao,SPAD shows great potential in monitoring the growth of Giant Juncao.The drone extraction height has a good fitting effect with the measured natural plant height,but the overall distribution value is low,which is 22.69%lower than the measured natural height.（2）The higher the nitrogen application rate,the lower the reflectance of the red band absorption valley and the higher the near-infrared reflectance of the Giant Juncao leaves.The correlation analysis of various characteristic covariates with biomass and SPAD of Giant Juncao revealed that:the original bands most related to AGB and SPAD were all in the blue band;There is a high correlation between the extraction height of drones and the biomass of Giant Juncao（r=0.927）;The vegetation index established by spectral information reflects the horizontal changes of chlorophyll and is highly correlated with SPAD;The vegetation indices most closely related to SPAD and AGB are BNDVI（r=0.885）and CIRE（r=0.702）,respectively,and the correlation between texture features is not as good as that of vegetation indices.（3）By combining three feature parameters with vegetation index and estimating the biomass of Giant Juncao,only texture information has a poor effect on improving the accuracy of vegetation index modeling.Adding drone extraction of height information can greatly improve the modeling accuracy of Giant Juncao.The accuracy of the RF model and SVM model optimized using the cross-validation method has significantly improved.In the comparison of the application results of the four modeling methods,the RF model has the best accuracy and strong applicability,which can be applied to various parameters.The inversion accuracy is 0.97,4.41,and 0.17 for the modeling set R²,RMSE,and MRE,respectively.The validation set R²,RMSE,and MRE are 0.93,4.78,and 0.23,respectively.GAM and SVM models can also achieve high accuracy,but they require the addition of high correlation indicators and have poor universality.Using multiple feature parameters to invert the SPAD of Giant Juncao,it was found that neither texture features nor coverage information extracted by drones can improve the accuracy of the model.The RF model using only the original band and index has the best effect,with modeling R²,RMSE,and MRE are 0.95,1.28,and 0.03,respectively.The validation R²,RMSE,and MRE are 0.89,1.82,and 0.06,respectively.（4）Add data from other periods to test the built model,with AGB and SPAD testing R²,RMSE,and MRE values of 0.58,4.96,0.35,0.46,4.01,and 0.11,respectively.The best model for inverting AGB and SPAD predicts that the distribution areas of high and low values are generally consistent,which can achieve the effect of monitoring the growth and health of giant fungus grass SPAD by inverting it to achieve higher AGB.