Lorey's height, representative of mean height in uneven-aged forest stands, is a valuable parameter for forest ecosystem management. While in situ measures provide the most precise information, remote-sensing techniques may provide less expensive but denser and more operational alternative of Lorey's height estimation over highly mountainous areas. This research aims first to evaluate the performances of two nonparametric data mining methods, random forest (RF) and artificial neural network (ANN), for estimation of Lorey's height using ice, cloud and land elevation satellite/geoscience laser altimeter system (ICESat/GLAS) in Hyrcanian forests of Iran and then to provide Lorey's height map using a synergy of ICESat/GLAS and optical images (TM and SPOT). RF and ANN GLAS height models were developed using waveform deterministic metrics, principal components (PCs) from principal component analysis (PCA) and terrain index (TI) extracted from a digital elevation model (DEM). The best result was obtained using an ANN combining first three PCs of PCA and waveform extent "Wext" (RMSE = 3.4 m, RMSE% = 12.4). In order to map Lorey's height, GLAS-estimated heights were regressed against indices derived from optical images and also topographic information. The best model (RF regression with RMSE = 5.5 m and = 0.59) was applied on the entire study area, and a wall-to-wall height map was generated. This map showed relatively good compatibility with in situ measurements collected in part of the study area.