Estimation Of Chlorophyll Contents Based On Model Simulation And Remote Sensing Image

Green plants are the most important components of terrestrial ecosystem.They have many ecological functions such as maintaining water and soil,preventing wind and conserving water sources,and beautifying environment.Leaf structure parameter,equivalent water thickness,dry matter contentand chlorophyll content are called biochemical components of plant.Chlorophyll content is a kind of biophysical parameter that can quantitatively express the photosynthesis ability of plant.It is an important indicator to monitor the growth stage and nitrogen element of plant.Therefore,dynamic estimation of chlorophyll content can provide information to monitor crop growth and to prevent pests and diseases for agricultural decision makers in time,and to provide an important theoretical basis for sustainable development of modern agriculture.On may 16-18,2017,we made field measurements of 43 sampling points in Luancheng District of Shijiazhuang city.The latitude,longitude,LAI,mean leaf angles,winter wheat height were obtained.The laboratory method was used to measure chlorophyll contents and water contents of winter wheat leaves.The average values of field data were used as the parameter values of PROSAIL model to simulate wheat canopy spectra,to make research on the relationship between vegetation indices and chlorophyll contents;Normalized difference vegetation index and photochemical reflectance index with high correlations to chlorophyll content were used to simulate the sensitivity of vegetation indices to biochemical components by using PROSAIL model.Estimation model were established by using vegetation indices and chlorophyll content estimation.Using Landsat-8 OLI image on May 15,2017,chlorophyll contents of winter wheat were estimated.Based on neural network algorithm(BPNN),a BPNN-based chlorophyll content estimation model was established and accuracy was verified.The conclusions were as follows.By using PROSAIL model,the sensitivity of NDVI and PRI to various biochemical components was obtained.NDVI had high sensitivity to leaf structure parameters(N)and chlorophyll concentration(Cab);NDVI was insensitive to dry matter concentration(Cm)and equivalent water thickness(Cw).At canopy scale,NDVI had higher sensitivity to leaf area index(LAI),observation zenith angle(VZA),leaf inclination angle distribution(LAD)and solar elevation.PRI had higher sensitivity to N,Cm and Cab;But it had lower sensitivity to Cw.At canopy scale,PRI was less sensitive to LAI and VZA and more sensitive to LAD and solar elevation.Through the simulation of PROSAIL model,the relationship between different vegetation indices and chlorophyll contents were analyzed.When VZA was 0°,30°,60°,and90°,CVI,PRI,RVI,NDGI,OSAVI and DVI values increased with the increase of chlorophyll contents.MCARI,NDGRI and EVI values decreased when chlorophyll contents increased.CVI,PRI,RVI,NDVI,OSAVI and NDGRI values decreased with the increase of VZA.MCARI,EVI and DVI increased with the increase of VZA.Chlorophyll content estimation model was established to estimate chlorophyll content of winter wheat in Luancheng District,by integrating field measurement data with Landsat-8OLI image.The determination coefficient R~2 between CVI and chlorophyll content was 0.485,followed by RVI,that of NDVI was the lowest.The determination coefficient of CVI between measured and estimated values was 0.697,which was higher than those of RVI and NDVI.The root mean square error(RMSE)and relative error(RE)of RVI were the lowest,which were 2.729 and 0.095,respectively.Therefore,based on OLI images,vegetation indices RVI and CVI achieve higher accuracy and can be used to estimate chlorophyll content.Based on neural network model(BPNN),chlorophyll estimation models were established by using vegetation indices of CVI,NDVI and RVI,and the accuracy was verified.Using data of 20 points to build chlorophyll estimation models,R2between CVI,RVI,NDVI and estimated chlorophyll contents were 0.547,0.522 and 0.478,respectively.Using data of other20 points to verify the accuracy,R~2 of CVI between estimated and measured chlorophyll contents was 0.729,higher than those of RVI and NDVI.RMSE and RE of RVI between estimated and measured chlorophyll contents were the lowest,which were 2.558 and 0.046,respectively,Therefore,three chlorophyll estimation models based on BPNN,RVI had the highest estimation accuracy,and can be used to estimate chlorophyll content.