Abstract: MODIS sensors, carried onboard Terra and Aqua satellites, scan the same location daily at a fixed time. Because of the sequential multidirectional information contributed by satellite orbit drift along with multi-channel spectral responses, MODIS data greatly enriches the observations of land surface targets, which makes it possible to estimate the land surface parameters accurately and timely, such as leaf area index (LAI). Many researchers have focused on LAI estimation using MODIS data, among whom most used the multispectral data of a single satellite in one day or eight days, while few comprehensively utilized the multispectral and multidirectional information obtained by the both MODIS sensors in some sequence of days. MODIS LAI products have developed a series of generations, the fifth version (MODIS V005) has integrated data from both Terra and Aqua. It is proven that this version is improved with single satellite data, however, it only utilizes red and near-infrared band observations. It has been suggested that taking the shortwave infrared band observations into consideration can help improve the accuracy of LAI estimation. Moreover, some validation studies indicate that there are still some limitations in applying current MODIS LAI products, e.g., LAI is overestimated or underestimated to some extent in different regions. Therefore, this paper investigates the methods of winter-wheat LAI retrieval using multispectral and multidirectional observations of Terra/Aqua MODIS in consecutive days.In this study, data preprocessing, including cloud status and data quality checking, was used first to remove the observations with partial or complete cloud cover, cloud shadow, or low pixel quality in the study area. Then, LAI, average leaf angle (ALA), chlorophyll content (Cab), water content (Cw) and dry matter content (Cm) were selected as the inversion parameters through sensitivity analysis. Other parameters were fixed by drawing upon previous studies and a priori knowledge obtained from field measurements. Accordingly, a look-up table (LUT) of the PROSAIL model was generated. In order to determine the optimal bands and angles of observations, some tests were done with simulated data before the inversion. The RMSE (root mean square error) and R2 (determination coefficient ) between the estimated and the true LAI illustrates that the accuracy is improved when the data of 648, 858, 550, 1240, 1 640 and 2130 nm wavelengths and multidirectional observations are chosen. Finally, LAI was estimated by searching the LUT and the mean of the 50 best cases were taken as the final solution.Comparison between the LAI results and the NDVI derived from HJ-1 CCD and MODIS data shows that they are consistent in spatial pattern. Validation using field-measured LAI illustrates the results of our method are better than MODIS LAI products in temporal variation characteristics and closer to the field-measured LAI. Nevertheless, the retrieved LAI of winter wheat is usually lower than the field-measured values, regardless of whether the PROSAIL model or MODIS 3D radiative transfer model is used. This may be partly caused by the effect of mixed pixels, which needs to be verified by further studies with high spatial resolution data. Another reason is that a saturation phenomenon often occurs at high LAI levels resulting from the low sensitivity of canopy reflectance in this domain.