We evaluated the accuracies of Digital Terrain Models (DTMs) generated from LiDAR data in an area of montane Atlantic Forest in the municipality of Rio de Janeiro, Brazil, and explored the use of those data to estimate tree heights. We employed the interpolation of data on a regular grid and a Triangular Irregular Network (TIN) using two softwares (ArcGis 10.3 and Fusion), with 0.4 and 1 m pixels, at scales of 1:2,000 and 1:5,000. The DTMs were evaluated using statistical inferences, considering 165 points surveyed using a geodetic Total Station (TS) in a 20x50 m plot. The models were classified according to the standard regulatory instructions of National Cartography technical norms. Accuracy varied from 0.59 m to 0.63 m among the specifications use for scales of 1:2,000 and 1:5,000. The maximum tree crown heights encountered with the eight Canopy Height Models (CHMs) employed were very similar to field measurements. When our data were analyzed by height classes, however, differences were encountered between the different softwares and interpolation methods used, indicating the influence of DTMs on tree height estimates based on LiDAR data.