Leaf area index (LAI), an important biophysical parameter of canopy structure, is not only a variable of significance to ecological and climatic models, but also a indicator for monitoring ecosystem dynamic. LAI can be estimated using various indirect and direct methods. The leaf litter collection method is widely used as a direct method to measure LAI in deciduous forests. Indirect measurements using radiation transmittance and gap fraction theory are often compared and calibrated by direct litter fall measurement, which is considered as a reference method. However, few studies have addressed the question of sampling strategy and the precision of LAI estimation by leaf litter collection so far. This study examined the above methodology with disposable leaf litter collection in the deciduous forests in Changbai Mountain and Beijing region after defoliation. The main results were showed as follows: (1) The difference of moisture content between the upper and lower layer of the leaf litter was significant with the absolute variation of 10.0% and the diurnal variation of the leaf litter moisture content was also significant with the absolute variation of 20%. Therefore, litter samples must be directly collected as much as possible from top to bottom layer in order to reduce the error of LAI measurement caused by leaf litter moisture content variability. (2)The distribution of LAI measured either in plantation or natural forests was uneven, regardless of sampling size (1 m2 or 25 m2) and the size of sampling units at resolution of 1 hm2 plot or 30 m×30 m plot. The variation ranged from 0 to 15.5 (at 1m2 resolution) or from 2.6 to 9.1 (at 25m2 resolution). (3) The larger area of the litter sampling, the higher accuracy of LAI estimation in deciduous forests, and the sampling should be conducted in a flat land surface, For a plot of 30 m×30 m or 1 hm2, a 100 m2(10m×10m) subplot should be randomly sampled, and this led to the precision of LAI estimation up to 85% and 80%, respectively. (4) For the subplot (10m×10m), the four adjacent 5 m × 5m sub-plots could be further divided and sampled. The rapid measurement for each 5 m × 5m sub-plot was conducted in the following: Ⅰ Six sampling units of each 1m2 in size were randomly set up and the precision of LAI estimation in 100 m2 plot, 30 m×30 m plot and 1hm2 plot were 90%, 75% and 70% respectively (at the level of 99% probability). Ⅱ Eleven sample units of each 1m2 in size were randomly set up and the precision of LAI estimating in 100 m2 plot,30 m×30 m plot and 1hm2 plot were about 94%, 80%, and 75% respectively (at the level of 99% probability).