Research On Monitoring Waterlogging Stress Level Of Winter Wheat Based On Hyperspectral Remote Sensing

Under the background of climate change,the incidence of extreme weather events such as waterlogging disasters increases.Frequent agricultural meteorological disasters are increasingly affecting agricultural production and affecting food security seriously.Waterlogging disaster is one of the important causes of crop yield reduction in the middle and lower reaches of the Yangtze River in China.It is of great significance for disaster prevention,reduction and food security to strengthen the monitoring of crop waterlogging stress in this area.However,traditional waterlogging stress monitoring method has shortcomings such as high labor intensity and strong subjectivity.Hyperspectral remote sensing provides a timely and reliable method for monitoring waterlogging stress,but at present,this method is rarely used in monitoring waterlogging stress level.In this paper,Yangzhou,Jiangsu Province,in the middle and lower reaches of the Yangtze River,was used as a research area.Based on pot and field control experiments,winter wheat was subjected to different gradients of waterlogging stress treatment at the jointing stage.Leaf spectrum,canopy hyperspectral imaging data and leaf water content(LWC)of winter wheat were collected(The data of pot experiment were collected every 7 days after waterlogging treatment until the winter wheat was mature;the data of field experiment were collected once at jointing,heading,flowering,filling and maturity stages respectively).The main purposes of the experiment were as follows:(1)Combined with vegetation index,harmonic analysis and spectral differential information entropy,leaf and canopy spectral characteristics of winter wheat under waterlogging stress were sorted out,and the spectral identification indexes were selected to distinguish the stress level.(2)Combined with methods such as vegetation index construction,deep learning and mathematical modeling,the characteristics of LWC in winter wheat under different waterlogging stress were analyzed,and a model suitable for inverting LWC of winter wheat under waterlogging stress was selected.(3)Finally,the above research results were applied and verified in the field.It aimed to monitor waterlogging stress level of winter wheat based on two aspects:spectral identification indexes and LWC.The results showed that the vegetation index SRPI was the best vegetation index for identifying winter wheat under normal and waterlogging stress.Red light absorption valley(RW:640-680nm),red edge(RE:670-737nm)and near infrared(NIR:750-900nm)band had strong ability to identify waterlogging stress.The RW band was the best band for monitoring waterlogging stress level.Within this band,the spectral differential information entropy and the first three harmonic amplitudes c1,c2,and c3during heading,filling,and flowering stages could all be used to identify waterlogging stress level of winter wheat.Within RE band,c3 could also be used to distinguish the stress level of JM31(more sensitive to water).BPNN model with the original spectral value at 648nm,the first-order differential value at 500nm,the red edge position,the new vegetation index RVI(437,466),NDVI(437,466)and NDVI'(747,1956)as independent variables was the best model for inverting LWC of winter wheat under waterlogging stress(modeling set R~2=0.889,RMSE=0.138,verification set R~2=0.891,RMSE=0.518,field verification set R~2=0.729,RMSE=1.425).The results showed that it was feasible to monitor waterlogging stress level of winter wheat from two aspects:spectral identification indexes(based on vegetation index,harmonic analysis and spectral differential information entropy)and inversion model of LWC(based on vegetation index construction,deep learning and mathematical modeling).It could identify and monitor waterlogging stress level accurately,and was of great significance to the control of waterlogging stress,the timely implementation of disaster recovery management technology.