Estimation For The Main Biophysical And Biochemical Parameters Of Plants Based On Gene Expression Programming Algorithm And Hyperspectral Data

The occurrence of extreme weather events caused by global climate change has also led to the occurrence of disaster factors in local areas,and the changes in meteorological factors have different effects on the variety,quality and yield of agroforestry crops.In order to study the sensitivity and adaptability of crop varieties and quality to climate change and to assess the impact of crop yields on future climate change,monitoring and estimating the main biophysical parameters(such as leaf area index)and biochemical parameters(such as leaf nitrogen content,leaf dry matter content)is very necessary.In this paper,the LOPEX'93 database was used as the research data.The leaf optical model was optimized by genetic expression programming(GEP)algorithm,and the dry matter content of the leaves was estimated.Using the original and derivative spectra of the leaf,the relationship between the leaf nitrogen content and the leaf spectral data was fitted to study the hyperspectral band combination which was most suitable for estimating the leaf nitrogen content and construct the estimation model.In addition,a new spectral index was constructed based on the GEP spectral model to estimate the leaf area index by taking the rice canopy data as an example.The paper has obtained the following results and main conclusions:(1)The first derivative model established using the partial band(400-1195 nm)is the best for most plant species in the test data set,and it is less affected by the leaf moisture content.The GEP model for estimating the leaf nitrogen content was constructed by using the original spectrum and the first derivative spectral data respectively.The results of leaf nitrogen content estimation by using all bands or partial bands(such as visible and near infrared)were obtained.The first-order derivative spectra enhance the weak absorption spectral characteristic information which reflect the change of leaf nitrogen content in the visible and partial near-infrared bands,so the model was improved.(2)In the models of first-order derivative spectrum,the better bands are in green band(509nm-532nm),red edge region(704nm-741nm)and short wave infrared region(815nm-957nm,1011nm-1170nm).In the models for leaf nitrogen content estimation using original spectrum,the better bands are in the blue band at 440-485nm,green band at 505-570nm and short wave infrared band at 1325-2500nm.So,the first-order derivative spectrum in green and red edge and partial near infrared regions are more sensitive to leaf nitrogen content changes.Although the short wave infrared band contains the important spectral characteristic information which reflects the change of leaf nitrogen content,the effect of water should be taken into account.(3)The factors that affect the accuracy of the inversion results of the dry matter content mainly include the error of the leaf simulated spectra by PROSPECT model and the influence of moisture.Using the PROSPECT model and the leaf measured parameters,the simulated spectrum by measured parameters can be obtained.Compared with the measured spectrum and simulated spectrum,result showed that the PROSPECT model has a certain degree of error in the simulation of the leaf spectrum of different plants.Moreover,the absorption characteristics of the leaf moisture content in the near infrared and short wave infrared bands affect the inversion of the dry matter content of the leaves,which is the main influencing factor.(4)In the case of relatively low leaf moisture content,the estimation of leaf dry matter content is ideal.The GEP algorithm and PROSPECT model were used to estimate the leaf dry matter content of the whole data set.The accuracy of the whole estimation was not very high.However,when leaf equivalent water content is less than 0.0093g/cm2,the result is better.(5)The new spectral index ?-R1 can well reflect the change of rice LAI,and to a certain extent reduce the saturation phenomenon.In this paper,the GEP algorithm was used to establish the model of canopy spectrum and LAI based on the measured hyperspectral data.Then,the new spectral index ?-R1 of LAI was proposed by GEP model.In order to verify the new index,the PROSAIL model was used to simulate the changes of the spectral index values of ?-R1 at different chlorophyll and LAI levels.The ?-R1 index is affected by chlorophyll concentration at the higher LAI level,but on the whole,its value is more sensitive to the variety of rice LAI.