Pinus elliottii  is commonly considered to be one of the most important pioneer tree species for forest restoration and reconstruction in subtropical degraded red soil area of China

due to its high productivity and leannessresistance. However

they also consume certain amount of water during the course of growing and maintaining ecological balance. Therefore

quantitative research on tree water consumption by transpiration has become a hotspot in the field of tree physiological ecology in recent years. In order to provide an effective way for conducting sustainable management of  Pinus elliottii  plantation and associated water resource of similar condition

in this study

we chose the  Pinus elliottii  plantation in degraded red soil area of Jiangxi province as the research object

and the log-sigmoid type function (i.e. tansig) of MATLAB software was selected as the action function of neurons. Based on the correlation analysis between sap flow velocity and meteorological factors

air temperature

relative air humidity

average net radiation and vapor pressure deficit were chosen as the input variables and sap flow velocity as the output variable. Then the optimum 3-layer BP artificial network model of individual sap flow velocity was established with the topological structure of 4-10-1. Nineteen hundreds groups of individual tree data were used to train the very neutral network both with Bayesian regularization algorithm and Levenberg-Marquardt algorithm

while the remaining nineteen hundreds groups were used to test the model. Good fitting results were obtained for linear regression between predictive and measured values under two algorithm model

with the R higher as 0.98. The results showed that fitting accuracy of training samples was 88.12% and 88.11%

respectively

and the simulating accuracy of testing samples was 88.11% and 87.98%

respectively. This model can well reflect the non-linear relationship between sap flow velocity and meteorological factors owing to its higher accuracy and fine generalization than linear regression model

which indicated the availability of BP neutral network for the analysis

simulation and prediction of sap flow velocity.