Diameter frequency distributions provide essential information for estimating timber assortment, monitoring carbon stocks, and formulating forest management measures. In this study, we estimated diameter distribution utilizing unmanned aerial vehicle laser scanning (ULS) data by applying three Weibull distribution modeling methods: (1) parameter prediction method (PPM); (2) moment-based parameter recovery method (PRMM); and (3) percentile-based parameter recovery method (PRMP). The variables used in Weibull distribution modeling methods were combined with stand density as response groups to be modeled with ULS metrics. Considering the hierarchical structure of ULS data and the autocorrelation among sub-models, mixed-effects seemingly unrelated regression (SURM) were applied to take into account both spatial and cross-model correlations. The experiments were conducted for 11 sites of larch plantations using leave-one-out cross-validation (LOOCV). The diameter distribution was estimated and calibrated by the observed stand density considering the correlations of sub-models' random-effects. The results demonstrated that applying a relatively small number of plots (1 to 6) and estimated best linear predictor (EBLUP) for local calibration could improve the prediction performance. The optimal results were obtained from PRMM with six calibrated plots, and the average Reynolds error index was 45.30. Furthermore, simulation applications with different pulse densities were applied and suggested that calibration could also improve the estimation performance but brought little improvement on estimation stability, far lesser than the impact of point cloud density. This study provides an improved approach for diameter distribution estimation and benefits for operational forest applications using remote sensing data.