利用间隙率模型反演LAI（Leaf Area Index），需要同时获取冠层间隙率和消光系数，后者与冠层叶倾角分布有关。基于点云数量构建激光雷达穿透指数LPI （LiDAR Penetration Index），用以代替冠层间隙率GF （Gap Fraction），并利用间隙率模型反演冠层LAI是利用LiDAR PCD（LiDAR Point Cloud Data）数据反演冠层LAI主要思路。冠层和背景的光谱差异是影响PCD数据中冠层和背景点云数量的重要因素，因此从LPI到GF的校正需要获取背景和冠层的后向散射系数比（μ=ρg/ρv）。本文基于PCD数据中点云强度进行μ值获取，用以实现LPI到GF的校正；在假设区域内叶倾角满足椭球形叶倾角分布的基础上，利用样地尺度下的多角度GF，采用有约束的非线性最优化方法获取椭球形叶倾角分布参数χ，实现冠层消光系数的获取；最后利用间隙率模型实现基于PCD数据的LAI反演。本文探讨了基于PCD数据进行冠层LAI反演时，样地尺度Rxy_Tile、样方尺度Rxy_Plot以及进行背景和冠层分割的高度阈值Ht对模型的影响。结果显示，由于区域内地衣植被广泛覆盖，基于点云强度的μ 值接近1，符合区域特点；经过μ值校正后的GF对冠层间隙率具有较好的反映能力（R2=0.78,RMSE=0.09）；对于优势种明显的区域，基于样地尺度内多角度GF的χ值反演受样地内冠间大间隙的影响，选择合适的样地尺度能够减小LAI反演过程中的系统性误差；结合地面参考数据，确定的最优Rxy_Tile、Rxy_Plot和Ht分别为950 m、10 m和2.6 m，在此基础上反演的LAI与地面测量数据具有高度的一致性（R2=0.84,RMSE=0.51）；与Rxy_Plot相比，基于间隙率模型的LAI反演对Ht的选择更为敏感。

Canopy gap fraction and extinction coefficient are two primary variables to retrieve Leaf Area Index (LAI) from light transmittance-based model. Currently, for the difficulty of calculating gap fraction from discrete LiDAR Point Cloud Data (PCD), LiDAR Penetration Index (LPI) is used as the alternative of gap fraction to estimate LAI. However, LPI ignores the target spectral difference which is an important factor affecting the number of canopy and background echoes. Therefore, the backscattering coefficient of the background and canopy, μ=ρg/ρv, is required to correct the LPI to GF. We extracted μ from intensity of the PCD data, which achieved by using a linear regression between the intensity of background and that of canopy in each pulse intensity groups, then the mean μ of all valid groups was used to transform LPI to gap. Given there was a dominant species of vegetation in study area, the light extinction coefficient (k) was extracted using constrained optimization method to obtain the ellipsoidal model parameter χ from multi-angle gap fraction at the large spatial scale (tile scale) under the assumption that the leaf angle distribution can be modeled by a ellipsoidal model and the leaf mean tilt angle is constant through study area. Finally, LiDAR LAI was estimated using retrieved gap fraction and extinction coefficient. Meanwhile, the impact of tile scale (Rxy_Tile), sample scale Rxy_Plot and height threshold (Ht) were also investigated. The results showed that the μ value was close to unit, and it is contributed by the extensive coverage of lichen vegetation in the area, which is consistent with the actual field characteristics. The gap fraction corrected by μ has a good ability to reflect the field measured data (R2=0.78, RMSE=0.09), and the leaf angle distribution parameter χ, is affected mainly by the large gap between the crowns for areas with dominant species. In terms of size of tile, the retrieval χ, the parameter of ellipsoidal model, was sensitive to the spatial size of tile, which means that attention should be paid to select tile size. An ill-suited tile size would result in a systematic underestimation of LAI. For the target parameter of LAI, the result revealed that it was highly consistent with the ground measurement (R2=0.84, RMSE=0.51) under the condition of Rxy_Tile, Rxy_Plot and Ht of 950 m, 10 m and 2.6 m respectively. It was concluded that the retrieved LAI was more sensitive to the choice of Ht, and it was noted that more attention would be paid to select appropriate Ht to ensuring the consistent result of LiDAR LAI and field measurements in the further work direction. We conclude that it is feasible to retrieve μ and further to produce LAI using ALS PCD data only. The significance of the proposed method is that it can produce reliable remotely sensed LAI from ALS PCD even with no ancillary spectral data.