• Crown models are derived from terrestrial laser data for 3 NW USA conifer species.• Crown models require only crown length for implementation.• Beta and Weibull curves fit to 95th percentile widths describe crown extent.• Crown profile curves are species-specific and not interchangeable.• Crown shape is not strongly conditioned by tree size or site. AbstractRegional crown profile models were derived for three conifer species of the interior northwestern USA from terrestrial laser scans of eighty-six trees across a range of sizes and growing conditions. Equations were developed to predict crown shape from crown length for Pseudotsuga menziesii, Pinus ponderosa, and Abies lasiocarpa from parametric curves applied to crown-length normalized laser point clouds. The 95th width percentile adequately described each crown's outer limit; alternate width percentiles produced little profile shape variation. For P. menziesii and P. ponderosa, a scaling parameter-modified beta curve gave the most accurate fit (using cross-validated Mean Absolute Error) to aggregated 95th width percentile points. For A. lasiocarpa, beta and Weibull curves (equivalently modified) produced similar results. For all species, modified beta and Weibull curves fit crown points with less error than conic or cylindrical profiles. Crown profile curves were species-specific; interchanging among species increased error significantly. Laser-derived crown base metrics provided objectivity and consistency, but underestimated field-derived base heights through inclusion of dead branches. Profile curve parameters were not correlated with tree or stand characteristics suggesting that crown shape is not strongly conditioned by size and site factors. However, laser sampling necessarily favored more open growing conditions, potentially under-representing variations in crown shape associated with social position. Overall, Terrestrial Laser Scanning (TLS) lends itself to detailed measurements of external crown architecture with occlusion-imposed limits to characterization of internal features. Yet, the time and cost of collecting and processing individual tree data precludes use of TLS as a common field sampling tool.