Canopy height models and airborne lasers to estimate forest biomass: two problems

Three ground datasets were used to simulate the canopy height characteristi cs of tropical forests in Costa Rica for the purposes of forest biomass est imation. The canopy height models (CHMs) were used in conjunction with airb orne laser data. Gross biomass estimation errors on the order of 50-90% aro se in two of the three analyses. The characteristics of the datasets and th e biomass estimation procedure are reviewed to identify sources of error. I n one dataset, the width of the fixed-area ground plots were small enough ( 5m) that significant portions of the overstory canopy above the plots were not accounted for in the ground samples. The use of mapped stand data from thin ground plots resulted in inaccurate CHMs, which in turn lead to gross overestimates of forest biomass (about 90% larger than the ground reference value). CHMs generated using mensuration data collected on thin, fixed-are a plots may significantly underestimate the average canopy height and crown closure actually found on that plot. This underestimation problem is direc tly related to plot width. Below a critical threshold, the thinner the plot , the greater the underestimation bias. In the second dataset, it is believ ed that lower-than-normal rainfalls at the beginning and end of the wet sea son may have produced a forest canopy with reduced leaf area. The airborne laser pulses penetrated further into the canopy, resulting in airborne lase r estimates of forest biomass which grossly underestimated reference values by about 50%. Changing canopy conditions (e.g. leaf loss due to drought, i nsect defoliation, storm damage) can affect the accuracy of a CHM. Sources of error are reported in order to forewarn those researchers who produce an d utilize canopy height models.