Electronic Supplementary Materials (ESM)

1Kwak et al. 1 2

1 Electronic Supplementary Materials (ESM) 2 3 Estimating stem volume and biomass of Pinus koraiensis using LiDAR data 4 5 Doo-Ahn Kwak, Woo-Kyun Lee, Hyun-Kook Cho, Seung-Ho Lee, 6 Yowhan Son, Menas Kafatos, So-Ra Kim 7 8 9 Supplementary tables 10Table S1 Descriptive statistics of the field measurements (where SD is the standard deviation for 11 each parameter and N the number of measured trees according to class) 12Table S2 Shape values and formulae according to the crown shape of individual trees (Coder, 13 2000), where CD is the crown width and CH the crown height. 14Table S3 Definitions and coefficients of factors required for estimating stem and above ground 15 biomass 16Table S4 Categories for the accuracy analysis of tree crown segmentation 17 18 Supplementary figures 19Fig. S1 An alternate type of laser altimeter, known as surface LiDAR, utilizes the complete time- 20 varying distribution of returning pulse energy or waveform resulting from the reflection of 21 a single pulse with a large footprint. 22Fig. S2 A small footprint LiDAR system can acquire precise information on a crown or canopy 23 using several LiDAR signals (returns) reflected onto- and into the canopy. In Figure 2b, 24 each signal includes a geographical coordinate and elevation. 25Fig. S3 Figure S3a represents the direction along- and across-track according to the flight direction 26 of the aircraft. False strips on the DCM produced without the filtering process are shown in 27 the direction from the upper left to the lower right in the left DCM of figure S3b. The 28 yellow points are LiDAR returns reflected from the surfaces of objects. As the distribution 29 of LiDAR returns was more uniform due to the filtering process, unnecessary false strips 30 were eliminated. 31Fig. S4 These figures depict how the watershed concept can be applied to individual tree 32 segmentation. The canopy surface (DCM in this study) has to be reversed because if not the 33 boundary lines are generated along with tree tops on the DCM. 34Fig. S5 Example two-dimensional side view of idealized crown shapes, with the crown shape factor 35 number and generic name (Coder, 2000) 1Kwak et al. 2 2

1Table S1 Descriptive statistics of the field measurements (where Std. is the standard deviation for 2each parameter and N the number of measured trees according to class) 3 Classes Measured growth factors N Max. Min. Mean Std. High Tree height (m) 67 22.7 15.2 19.1 1.5 density plot (Size: 0.05ha) DBH (cm) 34.0 11.0 19.9 5.7 Crown base height (m) 18.3 13.2 15.6 1.0 Crown width (m) 4.7 3.3 4.1 0.4 Medium Tree height (m) 37 28.2 20.7 24.1 2.1 density plot (Size: 0.1ha) DBH (cm) 45.0 28.0 35.9 4.1 Crown base height (m) 20.2 13.2 17.2 1.7 Crown width (m) 6.9 4.5 5.6 0.7 Low Tree height (m) 24 29.5 18.1 23.6 3.2 density plot (Size: 0.1ha) DBH (cm) 70.0 26.0 40.6 9.0 Crown base height (m) 22.1 11.6 15.8 3.1 Crown width (m) 8.5 5.3 6.8 0.9 4 5 6Table S2 Shape values and formulae according to the crown shape of individual trees (Coder, 72000), where CD is the crown width and CH the crown height. 8 Shape number Shape value Shape formula Shape name S1 8/8 (1.000) 0.7854CD2 CH Cylinder S2 7/8 (0.875) 0.6872CD2 CH Rounded-edge cylinder S3 3/4 (0.750) 0.5891CD2 CH Elongated spheroid S4 2/3 (0.667) 0.5236CD2 CH Spheroid S5 5/8 (0.625) 0.4909CD2 CH Expanded parabolic S6 1/2 (0.500) 0.3927 CD2 CH Parabolic S7 3/8 (0.375) 0.2945CD2 CH Fat cone S8 1/3 (0.333) 0.2619CD2 CH Cone S9 1/4 (0.250) 0.1964CD2 CH Neiloid S10 1/8 (0.125) 0.0982CD2 CH Thin neiloid 1Kwak et al. 3 2

1Table S3 Definitions and coefficients of factors required for estimating stem and above ground 2biomass 3 Factor Definition Coefficient Wood basic density Oven-dried weight (g) Coniferous: 0.48 /Green volume (cm3) Deciduous: 0.65

Biomass conversion Above ground biomass (kg) Coniferous: 1.29 and expansion factors /Stem biomass (kg) Deciduous: 1.22 4 5 6Table S4 Categories for the accuracy analysis of tree crown segmentation 7 Categories Definitions Correct delineation 1:1 correspondence between reference tree and detected tree exactly

Satisfactory One automatically detected tree corresponds to one reference tree, but the delineation boundaries are larger or smaller than reference tree crown

Merged tree More than one reference tree lies within the automatically delineated tree Split tree More than one automatically-delineated tree lies within one reference tree

Not found There exists a reference tree, but no corresponding automatically delineated tree 8 1Kwak et al. 4 2

1 (a) Scanning forest stand with a large footprint (b) Interpretation of a forest stand by a large LiDAR system footprint LiDAR signal 2 3Fig. S1 An alternate type of laser altimeter, known as surface LiDAR, utilizes the complete time- 4varying distribution of returning pulse energy or waveform resulting from the reflection of a single 5pulse with a large footprint. 6 7 8

9 (a) Scanning a forest stand with a small (b) Visualizing the vertical structure of a forest footprint LiDAR system stand using a small footprint LiDAR signal 10 11Fig. S2 A small footprint LiDAR system can acquire precise information on a crown or canopy 12using several LiDAR signals (returns) reflected onto- and into the canopy. In Fig. S2b, each signal 13includes a geographical coordinate and elevation. 1Kwak et al. 5 2

2 3(a) Directions by tracks (b) Filtering effect (left: before filtering, right: after filtering) 4 5Fig. S3 Figure S3a represents the direction along- and across-track according to the flight direction 6of the aircraft. False strips on the DCM produced without the filtering process are shown in the 7direction from the upper left to the lower right in the left DCM of figure S3b. The yellow points are 8LiDAR returns reflected from the surfaces of objects. As the distribution of LiDAR returns was 9more uniform due to the filtering process, unnecessary false strips were eliminated. 10 11

12 13 (a) Real tree in a forest stand (b) Real tree crowns’ surface 1Kwak et al. 6 2

1 2 (c) Reverse of real tree crowns’ surface (d) Formation of watershed by flooding 3 4Fig. S4 These figures depict how the watershed concept can be applied to individual tree 5segmentation. The canopy surface (DCM in this study) has to be reversed because if not the 6boundary lines are generated along with tree tops on the DCM.

7 8 9Fig. S5 Example two-dimensional side view of idealized crown shapes, with the crown shape factor 10number and generic name (Coder, 2000) 11