Metsa katvuse ja liituse hindamine lennukilt laserskanneriga

Many of the forest definitions include a criterion based on the share of ground covered by tree crowns. However, the lack of a clear definition for the share and mixing of the terms is common in laws and documentation causing the variables canopy cover (K) and crown cower (L) to be used loosely (Jennings et al., 1999). Different methods exist to estimate K and L (Korhonen et al., 2006). Airborne lidar data are now widely used for estimation of forest inventory variables via regression methods (N sset, 2004; N sset et al., 2004; Suvanto & Maltamo, 2010), leaf area index, LAI (Ria o et al., 2004; Morsdorf et al., 2006) and are an attractive source to estimate canopy cover and crown cover. Tests were carried out in mature Scots pine, Norway spruce and Silver birch stands (Table 1) which are used also for RAMI experiment (RAMI, 2010) in J rvselja, Estonia, to study the options for estimating canopy cover from airborne lidar data. Lidar data were collected with Leica ALS50-II on 30th July 2009 at 500 meter over ground. The scanner beam divergence at 1/e2 energy criterion was 0,22 mrad. Scan angle ranged up to 11 degrees and with two perpendicular flights the final point density on the ground was 20 p m-2 (Kuusk et al., 2009c). Lidar data were processed with FUSION/LDV (McGaughey, 2010) to extract data from sample plot area, create digital terrain model and to calcluate return height statistics. Canopy cover was estimated from lidar data by using all returns (Eq 1), using first returns only (Eq 2) or using single returns (Eq 3). For crown cover estimate the ratio of all returns to first returns D, was calculated (Eq 4). Reference height z was varied in the range 0.2 m ≤ z ≤ 10.0 m. Results were compared to the canopy cover estimates (Kc) calculated from the Cajanus tube (Rautiainen et al., 2005) readings from the ground corresponding to the z = 1.3 m (Table 2). Lidar return distributions by height (Figure 1) were different in studied stands. N sset (2004) recommended to use the height distribution information of return counts in regression models for predicting forest inventory variables. Cover estimates from lidar data depended significantly on the estimator (K1, Kk, Ky) and stand structure (Figure 2). The value of all lidar based estimators decreased with increasing reference height z (Figure 2). Compared to Kc K1(1.3) was positively biased (3-10%) in all stands. However, only in the birch stand the K1(1.3) estimate was outside the confidence intervals of Kc (Table 2, Figure 2). The single return (Ky) and all return (Kk) based canopy cover estimates depended more on the stand structure compared to K1. In the Scots pine stand K1(1.3) gave most similar canopy cover estimate to the ground estimate Kc whereas Ky(1.3) and Kk(1.3) underestimated Kc significantly (>15%). The pine stand structure was rather simple - only one layer of pine trees having minor overlaps between crowns. Therefore the Cajanus tube based estimates of canopy cover Kc, crown cover Lc