Airborne laser profiling is used to estimate tropical forest basal area, volume, and biomass. A procedure is developed and tested to divorce the laser and ground data collection efforts. The procedure involves 1) the collection of airborne laser data over the area of interest; 2) the collection of ground data in areas of similar forest cover type (in most cases in the same study area, but not necessarily coincident with the laser transects); 3) computer simulation of the forest canopy based on spatial and forest mensuration data available from the ground survey; 4) development of regression relationships between laser and ground measurements based on the simulated forest canopy; and 5) the use of these regressions with the airborne laser data using simple random sampling or line intercept sampling techniques. Variants of this simulation procedure were tested using three distinct data sets acquired in and over the tropical forests of Costa Rica. On two of the three study sites, airborne laser estimates of basal area, volume, and biomass grossly misrepresented ground estimates of same. On the third study site, where the widest ground transect samples were acquired, airborne and ground estimates agree within 10%. Basal area, volume, and biomass prediction inaccuracies in the first two study areas are tied directly to disagreement between simulated laser estimates and their airborne measurements of average canopy height, height variability, and canopy density.

A number of sampling issues were investigated in order to define a large area inventory procedure which utilizes airborne laser data in conjunction with a limited ground sampling effort divorced spatially and temporally from the airborne laser sampling phase. The following results were noted in the analyses of the three study areas. 1) Four ground transect segment lengths were considered: 25, 50, 75, and 100m. The 25m segment length introduces a level of variability which may severely degrade prediction accuracy in these Costa Rican primary tropical forests. This effect is more pronounced as transect width decreases. A minimum transect length which mitigates significant mean square errors by capturing a representative spatial sample of the primary tropical forest is on the order of 50m. 2) Gaps between airborne laser segments on the order of 15m mitigate the effects of spatial autocorrelation. 3) The decision to transform or not to transform the dependent variable (eg., biomass) is by far the most important factor of those considered in this experiment. The natural log transformation of the dependent variable increases prediction error, and error increases dramatically at the shorter segment lengths. The most accurate models are simple linear models with forced zero intercept and an untransformed dependent variable. 4) Results do not suggest any apparent, consistent advantage to the use of parametric or nonparametric regression techniques; either is appropriate. 5) General linear models are developed to predict basal area, volume, and biomass using airborne laser height metrics. Laser metrics include average canopy height, all pulses (h̅_a,), average canopy height, canopy hits (h̅_c), and the coefficients of variation of these terms (c_a and c_c). Coefficients of determination, where calculated for comparable models with intercepts, range from 0.4 to 0.6. 6) Line Intercept Sampling techniques may be used to estimate h̅_a, h̅_c, canopy profile area (p), canopy density (g), canopy area, and canopy volume directly using airborne laser data without the need to identify individual tree crowns in the airborne laser data. Efforts to extend these metrics to estimates of basal area, volume, or biomass by attempting to relate canopy area to basal area and canopy volume to merchantable woody volume or above-ground dry biomass were unsuccessful. 7) The assumption of canopy shape needed to develop the simulated airborne laser measurements has a significant impact on the accuracy of the basal area, volume, and biomass estimates. Estimates of volume or biomass developed assuming an elliptical crown, parabolic crown, and conical crown are respectively 5%, 12%, and 23% larger than estimates developed based on spherical cross-sectional assumptions. Based on this body of research, airborne laser and ground sampling procedures are proposed for use for reconnaissance level surveys of inaccessible, forested regions.