Analysis Of Ground-based Hyperspectral Characteristics And Satellite Remote Sensing Estimation Of Damage Degree Of Rice Leaf Roller In Field Rice

Rice leaf roller,Cnaphalocrocis medinalis,is one of the important pests of rice and poses a major threat to the food security in China.In this paper,some fixed-point and continuous observations on the damage of C.medinalis in the paddy field conducted which a ground non-imaging hyperspectrometer,SPAD-502 chlorophyll meter and a LP-80 canopy analyzer were used to obtain the rice canopy spectrum,leaf curling rate,relative chlorophyll content(Soil and Plant Analyzer Development,SPAD)and leaf area index(Leaf Area Index,LAI)data.Then,the original spectrum and first derivative spectrum characteristics of rice canopy of different damage levels under the damage of C.medinalis were analyzed.The correlation analysis of rice leaf rolling rate,SPAD,LAI with the original spectral reflectance of canopy,first derivative spectrum of canopy and vegetation index under different damage levels were carried out.Estimation models of leaf curling rate,SPAD and LAI were established by using highly correlated spectral characteristic bands and vegetation index.Based on the band range of Gaofen-2 remote sensing data and the ground hyperspectral data,7 broad-band vegetation indexes were selected and calculated and the relationship models between the vegetation indexes and the leaf rolling rate and SPAD were established at the jointing stage of rice.Finally,the Gaofen-2 remote sensing data was used to invert the leaf rolling rate and SPAD of the ground rice,and the ground measured point data was used to verify the accuracy of the inversion data.The main results were showed as follows:(1)On the near-infrared waveband,except for the filling period,the spectral reflectance of other growth periods was decreasing with the increasing of the damage degree,while the features of the visible light waveband were relatively insignificant.The first-order spectrum curve of rice canopy had a"main peak"in the 680?760nm band.The difference here was most obvious at booting stage and maturity stage.Within this range,the first-order spectrum value decreased with the increase in the degree of damage.(2)The correlation between SPAD,LAI,leaf curling rate and the first-order spectral reflectance of rice canopy was better than that of the original canopy spectral reflectance,and the spectral susceptible bands were mainly on the red and near-infrared bands.The correlation between SPAD and various vegetation indices was better than LAI and leaf rolling rate.Among the single factor estimation models,the cubic polynomial model had the best simulation effect,and the multiple stepwise regression estimation models had a better fitting effect than the single factor estimation models.The single factor estimation models and the multiple stepwise regression estimation models of leaf curling rate and SPAD had the best fitting effect in the jointing stage.The single factor estimation models of LAI had a poor fitting effect,but the multiple stepwise regression estimation model of LAI had a better fitting effect in the grouting period,with R~2 above 0.5.(3)The 7 wide-band vegetation indices were selected and calculated combined with the band ranges of Gaofen-2 remote sensing and they had the good correlations with the leaf rolling rate and SPAD.The fitting effect of the univariate polynomial models on the leaf curling rate was better than that of the multiple stepwise linear regression models,and the fitting effect of the multiple stepwise linear regression models of SPAD was better than that of the univariate models.The leaf curling rate and SPAD reversed by the Gaofen-2 remote sensing data had a significant linear correlation with the ground-measured leaf curling rate and SPAD,and the inversion effect of SPAD had a better effect than the leaf curling rate.These results indicated that it is feasible to use the Gaofen-2 data for the remote sensing inversion of leaf curling rate and SPAD on ground.