Abstract: The objective of the paper is to establish a hyperspectral estimating model of rapid and nondestructive for the apple leaf water contents and to provide the theoretical basis of drought warning for the fruit tree. The apple leaves were collected as the research object on two different growthes and studied the different water contents of apple leaves of hyperspectral features. The relationship was analyzed between the different leaves water contents and spectral parameters. Estimating models of apple leaf water contents were set up by means of constructing six common spectral indexes which were WI, WBI, PWI, GVWI, MSI and NDWI. The results indicated that sensitive bands of the apple leaf hyperspectral were focused mainly on the near infrared and short wave infrared light areas. All single variable estimating models reached a significant level (P＜0.01) by six vegetation moisture indexes. But the moisture index established estimation model was y=29503x²-57746x+28317, its fitting determination coefficient R²=0.5401, was the largest. The model tested, its RMSE was 2.4 and RE was 5.8% . The test accuracy of the model reached 94.2% . Using principal component analysis method, the model was y=-556.819+347.838x₁-17.815x₂-27.864x₃+299.492x₄+ 25.647x₅+9.835x₆, its fitting determination coefficient R2=0.6371. The model tested, its RMSE was 1.26 and RE was 1.8%. The test accuracy of model reached 98.2%. It showed that the estimating model of apple leaves water contents had good sensitivity and stability using principal component regression analysis method.