Study On Influencing Factors Analysis And Rate Prediction Of Sap Flow

Sap flow is an important process in soil-plant-atmosphere continuous water cycle system(SPAC).It controls the transpiration of the tree crown.It is often used as an important index to study the characteristics of tree water consumption and the mechanism of tree water transport.At present,the sap flow monitoring method and prediction model are not perfect.Therefore,a single probe sap flow monitor is designed and developed in this thesis.Then,the influencing factors of sap flow are explored,and a sap flow prediction model combining CNN and GRU is proposed.The research work of this thesis is as follows:1)A single-probe heat pulse velocity sap flow monitor is designed to realize the monitoring of sap flow and the local storage and remote transmission of relevant data.Compared with the instrument based on the heat ratio method,the R~2 of the monitoring results are 0.984 and 0.947,and the RMSE are 2.85 and6.29,respectively.The instrument designed in this thesis only needs a single needle.Compared with the existing monitoring equipment,it has the advantages of simple operation,stable performance,low power consumption and convenient carrying.2)The diurnal variation of sap flow is analyzed in this thesis,then,the influencing factors of sap flow were analyzed by image method and Pearson correlation coefficient method.The diurnal variation of sap flow shows a certain regularity,and the diurnal peak of sap flow rate is significantly higher in sunny days than in rainy days.Sap flow of Ginkgo biloba and Sapindus nipponica was strongly positively correlated with Ta,VPD and ST,strongly negatively correlated with RH,and weakly correlated with WS and SWC.3)This thesis proposes a CNN-GRU(CGRU)hybrid model.In the experiment,through the correlation analysis of sap flow and environmental factors,the environmental factors with weak correlation are eliminated,and the CGRU network is used to predict the sap flow.The performance of the model is verified on the public data set SAPFLUXNET and self-test data set.When the number of historical environmental factors(NHEF)is 16 and 18 respectively,the MSE,MAPE and R~2 of CGRU model on the two data sets reach 0.00231,22.319%and 0.948,0.00222,59.221%and 0.952 respectively.