Leaf area index (LAI) is an important parameter for evaluating the growth status and yield prediction of winter wheat. Spectral reflectance of leaves and concurrent LAI parameters of samples in Tongzhou and Shunyi Districts, Beijing City, China, were acquired during 2008/2009 winter wheat growth season. The correlation coefficient (|r|) Akaike’s information criterion (AIC), grey relational analysis (GRA)AIC, variable importance for projection (VIP)AIC, VIPpredictive residual error sum of square (PRESS) were integrated with partial least squares regression for estimating LAI, and the estimation models of optimal LAI were presented by using AIC and compared with traditional PRESS function. The results indicated that the R2 of |r|-PLS-AIC, GRA-PLS-AIC, VIP-PLS-AIC and VIP-PLS-PRESS models were 0.72, 0.67, 0.73 and 0.70, respectively. The VIP-PLS-AIC had higher predictive ability for winter wheat LAI than VIP-PLS-PRESS. Considering time domain characteristics of LAI, the relevant data acquired in Tongzhou and Shunyi Districts, Beijing City, China, during 2009/2010 winter wheat growth seasons were used to validate the models in different years. The results showed that VIP-PLS-AIC with RMSE of 081 had higher predictive ability than |r|-PLS-AIC with RMSE of 0.87, GRA-PLS-AIC with RMSE of 0.96 and VIP-PLS-PRESS with RMSE of 0.83. The results indicated that AIC could not only obtain the estimation model of optimal LAI at the same year, but also could be applied to the LAI detection in different years.