Monitoring Of Cotton Drought Stress Based On UAV Spectral Remotesensing

【Object】This experiment takes drip irrigation cotton under the film as the research object,and monitors drought-treated cotton leaves and soil moisture through drones equipped with multi-spectral,thermal infrared and visible light cameras,and constructs cotton leaves and soil moisture inversion based on drone spectral remote sensing model.In order to quickly obtain the field scale cotton drought information,timely grasp the cotton water status,provide technical methods,and provide theoretical basis for agricultural remote sensing drought monitoring【Methods】The experiment is located in the farmland water precision control test field of Wulanwusu Agrometeorological Experimental Station in Shihezi,Xinjiang.Drought stress trails were set up at the boll stage of cotton with five water gradient treatments based on 90%(T90),70%(T70),50%(T50),30%(T30)and control 100%(CK)of the local average irrigation at boll stage.Through continuous sampling of dry matter accumulation,plant height,photosynthetic gas exchange,SPAD,leaf area index,changes of leaf and soil moisture of cotton under drought stress,the UAV spectral remote sensing images were acquired simultaneously to investigate the effects of drought stress on the growth and development of cotton,and the most moisture-sensitive remote sensing parameters were selected by correlating the UAV spectral data with the leaf and soil moisture of cotton fields,constructing a model for non-destructive estimation of cotton leaf and root domain soil moisture content based on UAV spectral remote sensing,and verifying the accuracy of each model.【Results】1)Insufficient water supply at flowering and boll stage would slow down the growth of stem,leaf and boll,inhibit the growth of plant height,decrease the photosynthetic capacity of leaves and SPAD,and make leaves wilt,resulting in the decrease of leaf area index(LAI),and a pattern of CK>T90>T70>T50>T30 was observed among the treatments.From 72d to 105d after emergence,T30,T50,T70,T90 and CK treatment of Lei Ling dry matter increase was 10.04g,15.41g,21.33g,24.64g and 28.81g,respectively,and the net photosynthetic rate,Pn,decreased by 80%,59%,45%,25%and 7%respectively;From 77d to 113d after emergence,the change range of LAI in T30,T50,T70,T90 and CK treatments was-1.60、-1.22、-0.85、-0.56 and 0.61.Drought stress significantly reduced leaf water content(LWC)and soil relative water content(RSWC),and with the extension of drought stress time,leaf water content of cotton was more consistent with soil water status.2)Drought-stressed cotton canopy spectral features were obtained using a UAV and correlated with leaf and soil moisture.The multi-spectral spectral reflectance at 450nm,550nm and 685nm,Canopy Temperature(Tc)were negatively correlated with LWC and RSWC of each soil layer in the 0-50cm root domain,while the spectral reflectance at 725nm,850nm and 905nm were positively correlated with LWC and RSWC of different soil layers.Canopy Temperature is extremely strongly correlated with LWC,with a Pearson correlation coefficient of-0.752.And Canopy Temperature was significantly or highly significantly negatively correlated with the RSWC of each soil layer,the correlation coefficient was between-0.535 and-0.743.The correlation coefficients of 0-50cm root RSWC for each band and Canopy Temperature increased first and then decreased from shallow to deep,and the correlation coefficients with 20-30cm and30-40cm RSWC were the highest,reaching-0.743 and-0.725,respectively.The results showed that Canopy Temperature had the best prediction ability for RSWC in 20-30cm and 30-40cm soil layers.3)The correlation analysis of the selected Vegetation Index(VIs)and Vegetation Water Supply Index(VSWI)with LWC and RSWC showed that Vegetation Index and VSWI were positively correlated with LWC and RSWC,The Vegetation Index with correlation coefficients with LWC above 0.79 are DVI1,DVI2,NDVI1,NDVI2 and MTCI1,among which DVI1 has the highest correlation coefficient of 0.863with LWC.The Vegetation Water Supply Index with correlation coefficients with LWC above 0.8 are VSWI＿DVI1,VSWI＿DVI2,VSWI＿NDVI1,VSWI＿NDVI2 and VSWI＿MTCI1,among which VSWI＿DVI1 has the highest correlation coefficient of 0.887 with LWC.The results showed that the correlation between Vegetation Water Supply Index and LWC was significantly stronger than that of Vegetation Index,and the correlation coefficient was 1%-16%higher.4)The Vegetation Index with significant or extremely significant correlation with RSWC were MTCI1,MTCI2,DVI1 and DVI2,among which MTCI2 had the highest correlation coefficients of 0.837 and 0.890,respectively.With the moisture content of RSWC with each soil layer are VSWI＿MTCI1,VSWI＿MTCI2,VSWI＿DVI1,VSWI＿DVI2,VSWI＿NDVI1 and VSWI＿NDVI2,among which VSWI＿MTCI2 having the highest correlation coefficients of 0.886 and 0.894 with the moisture content of 20-30cm and 30-40cm soils,respectively.With the increase of soil depth,the correlation coefficients of Vegetation Index and the Vegetation Water Supply Index with RSWC in root region increased first and then decreased,and the correlation coefficient with 20-30cm and 30-40cm soil layer was the highest.Except that VSWI＿DVI1 is slightly lower than DVI1,the Pearson correlation coefficients of other Vegetation Water Supply Index and RSWC are all improved compared with Vegetation Index.The results show that the Vegetation Water Supply Index is better than the Vegetation Index in predicting cotton field RSWC.5)The Vegetation Water Supply Index model constructed based on multi-spectral and thermal infrared spectroscopy data can better estimate cotton leaves and soil moisture.The estimation accuracy and stability of the Vegetation Water Supply Index model are significantly better than the multi-spectral single band,Vegetation Index and canopy temperature models,and it has higher inversion capabilities for cotton LWC,20-30cm and 30-40cm root zone RSWC,among the cotton leaf and soil moisture inversion models based on UAV spectral moisture sensitivity parameters,the best fitting effect is the polynomial model of vegetation water supply index,and the R~2is greater than 0.794.