Modeling And Prediction Of Plant Responses To Climate Changes Based On Sunlight-Induced Chlorophyll Fluorescence

In recent years,a series of climate problems such as temperature warming and frequent extreme weather have brought great negative impacts on human production and life,and also affected the growth of terrestrial vegetation and CO2 absorption.The uncertainty of growth can not only effectively evaluate and predict its growth and development,but also guide vegetation regional management and promote sustainable development.At present,most of the research on plant response to climate conditions is based on specific experimental areas,and the analysis results have certain contingencies,which are difficult to be used in the assessment of largescale vegetation areas.At present,the climate variables considered in large-area vegetationclimate response research based on remote sensing data are usually limited to three or less,and the model results are not good,making it difficult to evaluate the impact of multiple climate coupling effects on vegetation growth.In addition,in terms of vegetation growth and development monitoring,Sun-Induced Chlorophyll Fluorescence(SIF)is a vegetation remote sensing technology developed rapidly in recent years,which can directly reflect the dynamic changes of plant photosynthesis However,the acquisition of chlorophyll fluorescence data is currently limited by satellite capture and calculation capabilities,and lacks long-term and timeliness,making the amount of data insufficient.Aiming at the current lack of vegetation growth assessment under the coupling effect of multiple climate conditions and the limited acquisition of chlorophyll fluorescence signals,this paper established a response model of vegetation SIF to climate change through multiple regression methods and structural equation modeling,and studied the multivariate coupling.The direct and indirect impact of each variable on SIF under the action;at the same time,by constructing the SIF prediction model,the accurate prediction of vegetation SIF is realized,and the problem of SIF data limitation is effectively solved.The details of this research work are as follows:1.Research on the construction of vegetation response model to climate under multiclimate conditions based on regression method.Through time trend analysis,the time series changes of global temperature,precipitation,radiation,drought index,wind speed,and sunlight-induced chlorophyll fluorescence from 2001 to 2018 were studied,and they were further divided into three categories: Areas without significant changes(Area Without Significant Change,AWSC),significantly increased area(Area With Significant Increase,AWSI),significantly decreased area(Area With Significant Decrease,AWSD);using partial correlation analysis,it is confirmed that most vegetation zone SIF will increase with temperature and precipitation(radiation,wind speed,humidity)increase(decrease),and map the results to the global longitude and latitude points for visualization;use the adjusted LASSO regression and ridge regression to solve the problems of multicollinearity and overfitting,The irrelevant and key factors affecting vegetation SIF were identified,among which temperature,precipitation,radiation,wind speed,and humidity as irrelevant variables accounted for 21.07%,38.15%,77.54%,55.77%,and 91.39% of the study area,respectively,while The proportions of key positive(negative)impacts are 61.85%,28.37%,7.15%,0.97%,and 1.66%(4.49%,10.60%,23.47%,45.37%,and 16.07%),and the impact of each Climatic factors of vegetation zone change.2.Research on model building of vegetation response to climate based on structural equations.Based on the multiple regression results of vegetation on multiple climate conditions,and further considering the characteristics of climate variable coupling,the structural equation model was used to establish the response model of vegetation SIF to various environmental variables under the condition of multi-factor coupling;six typical vegetation in the world were selected Coupling regression analysis was carried out on the data of regional fluorescence and environmental factors;the study analyzed the direct impact,indirect impact and total impact of each environmental variable on SIF.According to the size of the impact,the environmental variables accounting for more than 25% are extracted and mapped to the map.The research results intuitively provide a basis for the protection of global vegetation under the current climate change.3.Research on the prediction method of sunlight-induced chlorophyll fluorescence based on deep learning.In order to expand the SIF data set and establish a SIF change early warning method,the study selected the station data of 24 major cities in China,and proposed a SIF prediction model based on CNN-LSTM.Firstly,outliers and missing values are processed by the box plot method and time similarity method,and the traditional long-short-term memory network(Long-Short Term Memory,LSTM)model is improved,and the convolutional neural network(Convolutional Neural Networks,CNN)is added.),and compared it with other machine learning methods,the results show that whether from a single site or a global perspective,the CNN-LSTM proposed in this paper is more accurate in predicting SIF,and the average R2 of 24 sites is 0.932,RMSE is 0.06,and the CV-RMSE is 28.67.