In order to investigate effects of freeze-thawing treatment on models of soil total phosphorus and available phosphorus by near infrared spectroscopy,the soil samples were treated by liquid nitrogen after ground.The results showed that the soil samples after freeze-thawing had certain effects on the soil near infrared spectroscopy and that the predicted accuracy of models for soil total phosphorus and available phosphorus content also improved greatly.The models with partial least squares regression(PLS1)or artificial neural networks(ANN)became better.After freeze-thawing the parameters of model for soil total phosphorus content using PLS1 was as follows:Rc2=0.980,R(p)2=0.955,RPD=4.551,and was Rc2=0.964,Rp2=0.933,RPD=3.926 for soil available phosphorus content;and was Rc2=0.972,Rp2=0.955,RPD=4.488 and Rc2=0.988,Rp2=0.893,RPD=3.212,for soil total and available phosphorus content using ANN,respectively.It showed that models built by PLS1 and ANN could predict content of soil total and available phosphorus quantitatively.