Effects on Estimation Accuracy of Forest Variables Using Different Pulse Density of Laser Data
Collection of airborne laser scanner data for forest inventory is becoming a common practice today. To reduce cost when laser data are collected over large areas, the flight altitude or flight speed may be increased, resulting in low pulse density laser data. The effect of using a different pulse density of laser data on estimation accuracy of forest variables was investigated at stand level. Laser data were acquired by the airborne TopEye system at a 1,200-ha forest area located in southern Sweden (58°30′N, 13°40′E). The 70 selected stands were dominated by Norway spruce [Picea abies (L.) Karst.] and Scots pine (Pinus sylvestris L.) with tree height in the range of 6–28 m (mean 19 m) and stem volume in the range of 30–620 m3 ha−1 (average 286 m3 ha−1). Regression analysis was used to establish empirical functions at stand level. The pulse density of laser data was reduced from 25,000 to 40 returns ha−1. By reducing the pulse density the root mean square error (RMSE) for the tree height and stem volume estimation increased from 0.7 to 1.8 m and from 13% to 29%, respectively. A substantial decrease in estimation accuracy of tree height and stem volume could be observed at pulse densities <80 returns ha−1 (corresponding to about 10 m between adjacent laser returns). The rapid increase in RMSE at these pulse densities could be explained by the less accurate classification of ground returns. With access to a high resolution digital elevation model the RMSE for the tree height and stem volume estimation increased from 0.7 to 1.1 m and from 13% to 23%, respectively, as an effect of the reduced pulse density. Even though the pulse density was reduced to several meters between adjacent laser returns, the estimation accuracies were equal to or better than those commonly obtained by using conventional forest inventory methods, e.g., aerial photo interpretation. This finding implies that low pulse density airborne laser scanner data could be cost efficient to use in inventory for estimation of forest variables at stand level.