Abstract: Abstract: In order to solve the problem of low spatial heterogeneity of MODIS LAI products when used in regions of small and medium scale, in this paper, we presented a new method for the fast estimation of time-series LAI. LAI growth function is closely related with forest type, forest age and forest density. The normalized LAI growth function is associated with forest type and age of stand and is irrelevant with forest density. Due to the differences of LAI from different position caused by different forest density, LAI growth function of a large area can not represent the partial regions inside it, which means the LAI extracted from MODIS LAI product pixel can not represent the LAI of the internal parts. But leaves of the same forest type at the same age have a similar growth trend, namely, the normalized LAI growth function from the same forest type and age is basically the same. So the normalized LAI growth function obtained from a larger MODIS grid region with the same forest type and forest age can basically represent the normalized growth state of leaves from the sub-region within it. If the maximum LAI and normalized LAI function of the subregion are obtained, LAI growth function can be estimated based on the basis of formula. First of all, the MODIS LAI data sets were pre-processed. The MODIS time-series LAI curves were extracted from LAI MODIS raster data sets, and non normal seasonal fluctuations of LAI time-series curves were found. This study used Locally Adjusted Cubic-Spline Capping, namely LACC, to solve the problem of the non normal seasonal fluctuations. Based on the local weighted regression method, the LACC algorithm proposed by Chen et al was developed for the curve fitting and smoothing of abnormal data. In the framework of the method proposed by this paper, firstly, we extracted the MODIS time-series LAI cures of two broad-leaved forest samples and a mixed forest sample, and compared them with their own observed time-series LAI. A large deviation between the MODIS time-series LAI value and observed time-series LAI value was presented, but their normalized LAI values matched well, indicating that the normalized LAI growth curve obtained from the larger regions (MODIS pixels) could simulate the normalized LAI growth of the partial region of the large region (sample plots). Based on those conclusions, the normalized growth curves of MODIS LAI were extracted to simulate the annual variation of LAI in the studied area , and then the normalized LAI curves was fitted by using the three spline interpolation function to simulate the variation of LAI in study area at daily step. Secondly, according to the research results of LAI estimation model by Zhu Gaolong for Maoershan study area in 2011, RSR ( reduced simple ratio) had the best correlation with effective LAI, and it was most suitable for the LAI estimation. Based on the traditional remote sensing statistical model proposed by Zhu and TM remote sensing data, LAI was estimated after leaves completely expanded, namely the maximum LAI, which was the peak of the curve and by which the curve was controlled. Twenty LAI values observed in July 31, 2011 were used to verify the LAI maximum layer with R2 of 0.747. Finally, the normalized LAI fitting function and the maximum LAI were multiplied to obtain the time-series LAI data sets. The results showed that for directly estimated value of higher spatial resolution LAI on small and medium regional scale, the MODIS LAI products were not accurate and suitable, but the normalized MODIS LAI growth curve maintained high consistency with the normalized observed LAI and was used to simulate the annual change of real LAI. Moreover, the proposed method in this paper simply and efficiently provided time-series LAI data for other studies on small and medium regional scale. The fast estimation of time-series LAI proposed by this paper also had the following features: 1) the method can be applied to the region where the forest types are relatively simple and the age structure is not complicated. If the study area does not meet such situation, researchers should optimize this method by distinguishing age groups in each forest type; 2) the estimation accuracy of the LAI maximum layer estimated from TM data, which this method has certain dependence on, should be ensured above a certain precision.