Monitoring Growth Characters With Multi-angular Hyperspectral Remote Sensing In Wheat

Remote sensing can rapidly,non-destructive and accurately determine the growth status of crop in the field,which offers important technical support for implementation of precision farming.Multi-angular remote sensors may improve detection of whole canopy physiological and biochemical parameters.The objective of this study was to explore the optimum spectral analysis method and sensitive view zenith angle（VZA）for estimating leaf nitrogen content（LNC）,leaf area index（LAI）and pigment density,on the basis of different year,sites,cultivars,growth stages,N rates,and planting densities.The anticipated results would provide new vegetation index（VI）choice for multi-angular satellite remote sensing,and thus assist in real-time estimation and precise diagnosis of plant growth status in wheat.The relationship of LNC to traditional VIs and Normalized Difference Spectral Indices（ND）and Simple Ratio Indices（SR）was compared under different VZA.The results showed that the coefficient of determination（R²）of spectral reflectance and traditional VIs with LNC decreased with increasing VZA in both the forward and backward scattering directions and reached maximum values at a viewing angle of-20°.Ratio index（RI-1dB）exhibited the best linear relationship to LNC at the-20°viewing angle,but Enhanced vegetation index（EVI-1）showed the highest R² with LNC at the nadir direction.The relationships between the LNC and two-band combinations indicate that there are three sensitive regions with high R²,which vary with VZA,usually comprising combinations of blue-red wavelengths,green-red edge wavelengths,and between-red edge wavelengths.To further analyze the relationship between the combination of the three sensitive regions and the sensitive VZAs with LNC,the MAVI_SR index was calculated and found to be highly correlated with LNC.When independent data were fit to the derived equations,the novel MAVI_SRR is more effective for estimating LNC than previously reported VIs,independent of years,sites,and growth periods.Relation of LNC to VI and factor analysis-back propagation neural network（FA-BPNN）under different VZAs was revealed in this study.Results showed that the back-scatter direction gave improved index performance,relative to the nadir and forward-scattering direction.Red-edge VIs（e.g.,mND705,GND[750,550],NDRE,RI-1dB）were highly correlated with LNC.However,the relationships strongly depended on experimental conditions,and these VIs tended to saturate at the highest LNC（4.5%）.To further overcome the influence of different experimental conditions and VZAs on VIs,we developed a novel index,Angular Insensitivity Vegetation Index（AIVI）,revealed that performance was the highest at–20°and was relatively homogenous between–10°and–40°.This provided a united,predictive model across this wide-angle range,which enhances the possibility of N monitoring by using portable monitors.FA-BPNN gave better performance than vegetation index under different view zenith angle.These results suggest that the novel AIVI and FA-BPNN are more effective for multi-angular remote sensing monitoring LNC under different experimental conditions.The relationships of LAI to ground-based canopy hyper-spectral reflectance,spectral parameters and FA-BPNN under different VZAs was compared,and to derive regression equations for monitoring LAI in winter wheat with canopy multi-angular remote sensing data.The results showed that small view zenith angles nearby the vertical direction are more suitable for monitoring LAI than extreme larger view zenith angles.The spectral reflectance,traditional indices,ND and SR give better performance for LAI under backward scattering direction than forward direction.ND and SR did not show any advantage under all view zenith angles,but SR gave better performance than ND for monitoring LAI.The factor loading of green wavelength decreased with increasing view zenith angle in the first factor,while the value increased in the second factor.Testing of the monitoring models with independent dataset indicated that the spectral index of VIopt gave accurate growth estimation.The saturation,species-specific and angle sensitivity of common VIs for estimating LAI were compared in this study.The relationship between VIs and LAI are species-specific for erect and horizontal cultivars.The off-nadir VZAs did not give better index performance than the nadir direction.NDVI,SAVI,OSAVI,MSAVI,WDRVI,MTVI and mND705 tended to saturate when the value of LAI exceeded 4,except for the EVI and TVI.To further overcome the saturation and angular sensitive,we developed a floating angle coefficient Kf,based on NIR and green bands.All the VIs×Kf sharply improved the association with LAI across all VZAs.Mostly VIs×Kf could construct universal algorithms across VZAs for accurate estimation of LAI,except for the WDRVI×Kf,EVI×Kf,and TVI×Kf.Testing of predicting models with independent data indicated that the spectral indices mND705 and OSAVI gave more reliable estimation of LAI in wheat.Comparison the relationships of canopy leaf pigment density to ground-based hyper-spectral reflectance,spectral parameters and FA-BPNN under different VZAs.The results showed that the sensitive bands focused on 720-900 nm.VOG1,RI-1dB,NDRE,SDr/SDb and DD were highly correlated with leaf pigment density.ND and SR concentrated on red-edge and NIR bands under backward scattering direction and blue and red bands under forward direction.The first factor of FA-BPNN was blue and red spectral bands,and the second factor was NIR bands.The backward view zenith angles nearby the vertical direction are more suitable for monitoring leaf pigment density based on comparison of various monitoring models under different VZAs.The models with SDr/SDb,DD,ND（720,760）and ND（732,738）could be used to reliably estimate pigment density in wheat.