case study #3 identifying tree species from€¦ · case study #3 identifying tree species from...

10
Case Study #3 Identifying Tree Species From Geiger-Mode LiDAR Data: Compatibility and Accuracy Assessment CAPITOL STATE FOREST OLYMPIA, WASHINGTON ILMF 2017: IDENTIFYING TIMBER SPECIES FROM LIDAR DATA – CASE STUDY FEBRUARY 14, 2017 27

Upload: vuongdang

Post on 25-Aug-2018

220 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: Case Study #3 Identifying Tree Species From€¦ · Case Study #3 Identifying Tree Species From Geiger-Mode LiDAR Data: Compatibility and Accuracy Assessment CAPITOL STATE FOREST

Case Study #3 Identifying Tree Species From

Geiger-Mode LiDAR Data: Compatibility and Accuracy Assessment

CAPITOL STATE FOREST

OLYMPIA, WASHINGTON

ILM

F 20

17: I

DEN

TIFY

ING

TIM

BER

SPEC

IES

FRO

M L

IDAR

DA

TA –

CAS

E ST

UDY

FE

BRU

ARY

14, 2

017

27

Page 2: Case Study #3 Identifying Tree Species From€¦ · Case Study #3 Identifying Tree Species From Geiger-Mode LiDAR Data: Compatibility and Accuracy Assessment CAPITOL STATE FOREST

Project Area and Species

FEBRUARY 14, 2017 28

Project area 19.5 ha (48.1 acres) • On the western edge of

the Capitol State Forest • Flown August 2016 (leaf-

on) at a height of 8,000 ft AGL.

• 25 ppm • Four species found by the

field crew on site but only three in suitable amounts

Red Alder Douglas Fir Hemlock Cedar

Page 3: Case Study #3 Identifying Tree Species From€¦ · Case Study #3 Identifying Tree Species From Geiger-Mode LiDAR Data: Compatibility and Accuracy Assessment CAPITOL STATE FOREST

Project Timeline

FEBRUARY 14, 2017 29

Nothing like a deadline!

Jan 3 Conference call Harris, Object Raku, Forsite Consultants

Jan 4 Statement of Work drafted,

discussed, revised

Jan 10 Forsite field

crew on site to collect ground

truth

Jan 13 First ground

samples rec’d by Object Raku

Jan 19 Forsite field

crew completes collection

Jan 23 Full collection

set rec’d by Object Raku

Jan 30 Object Raku completes

ground truth processing

Feb 1 Object Raku completes

calibration tests

Feb 10 Object Raku completes

accuracy testing

Page 4: Case Study #3 Identifying Tree Species From€¦ · Case Study #3 Identifying Tree Species From Geiger-Mode LiDAR Data: Compatibility and Accuracy Assessment CAPITOL STATE FOREST

Ground truth sample collection

FEBRUARY 14, 2017 30

Field samples collected TSI ground truth

• Over 1,000 collected

GPS points for trees at and beneath canopy level.

• 501 samples trees processed as ground truth for use in TSI.

• Measured heights for 100 trees had an absolute error margin of 1.3m against an average height of 28m

Page 5: Case Study #3 Identifying Tree Species From€¦ · Case Study #3 Identifying Tree Species From Geiger-Mode LiDAR Data: Compatibility and Accuracy Assessment CAPITOL STATE FOREST

Ground truth calibration

FEBRUARY 14, 2017 31

If Incorrect, selected:

GT TSI Correct Inc

TSI Select Over or Under % Correct % Precision FD CE HW DR

FD 256 260 246 10 4 96.1% 94.6% FD N/A 0 6 4 CE 5 4 4 1 -1 80.0% 100.0% CE 1 N/A 0 0

HW 119 111 104 15 -8 87.4% 93.7% HW 10 0 N/A 5 DR 120 125 116 4 5 96.7% 92.8% DR 3 0 1 N/A

500 500 470 30 94.0%

Species Legend

CW Western Red Cedar

FD Douglas Fir

HW Western Hemlock

DR Red Alder

Page 6: Case Study #3 Identifying Tree Species From€¦ · Case Study #3 Identifying Tree Species From Geiger-Mode LiDAR Data: Compatibility and Accuracy Assessment CAPITOL STATE FOREST

32

Height Volume

Metrics & Measurements

SLOPE (%) ASPECT (deg) MIN AVERAGE MAX MIN AVERAGE MAX 3.5 48.5 164.8 0 188 315

NUMBER OF POINTS PER TREE MIN AVERAGE MAX 176 2,362 7,460

HEIGHT DISTRIBUTION (m) 5 To 9 10 To 14 15 To 19 20 To 24 25 To 29 30 To 34 35 - UP 3.6% 4.6% 10.0% 13.6% 18.2% 35.6% 14.5%

CANOPY AREA DISTRIBUTION (m2) 0 To 9 10 To 19 20 To 29 30 To 39 40 - UP 34.7% 57.9% 7.2% 0.2% 0.0%

HEIGHT (m) ELEVATION (m) MIN AVERAGE MAX MIN AVERAGE MAX 6.0 27.7 44.0 23.2 64.5 111.8

LIVE CANOPY DISTRIBUTION 0 To 9 10 To 19 20 To 29 30 To 39 40 To 49 50 To 59 60 To 69 70 To 79 80 To 89 90 - 100 1.1% 6.0% 36.7% 37.0% 11.8% 4.8% 1.3% 0.7% 0.7% 0.1%

Page 7: Case Study #3 Identifying Tree Species From€¦ · Case Study #3 Identifying Tree Species From Geiger-Mode LiDAR Data: Compatibility and Accuracy Assessment CAPITOL STATE FOREST

Experiment set-up

FEBRUARY 14, 2017 33

Ground Truth by Group # Experiment sub-areas • The douglas fir,

hemlock, and alder were divided up into three sample groups.

• We then created 12 sub-areas to better manage the analysis and highlight any trends.

Page 8: Case Study #3 Identifying Tree Species From€¦ · Case Study #3 Identifying Tree Species From Geiger-Mode LiDAR Data: Compatibility and Accuracy Assessment CAPITOL STATE FOREST

Accuracy Testing – Sample 1 on Sample 2

FEBRUARY 14, 2017 34

Species Legend

FD Douglas Fir

HW Western Hemlock

DR Red Alder

Areas 1 2 3 4 5 6 7 8 9 10 11 12

Sample 1 ground truth trees available

to be used in the TSI model

Sample 1-FD 6 7 1 8 1 1 10 13 6 17 10 14

Sample 1-HW 3 4 5 1 4 17 2 4 1 1 1 1

Sample 1-DR 4 5 7 9 7 2 4 1 0 0 5 0

Sample 2 - FD 4 4 1 6 0 0 6 9 4 11 6 9 Sample 2 - HW 2 3 3 1 3 11 2 1 0 0 0 0 Sample 2 - DR 3 3 5 6 4 1 3 0 0 0 4 0

AVG HGT 21.1 18.3 17.5 26.0 26.1 29.1 30.6 33.8 36.8 32.3 31.7 30.0 AVG ELEV 106.9 103.6 98.6 66.9 70.7 46.7 41.7 75.1 83.2 65.1 65.5 77.2

AVG INTENSITY 8.0 10.5 10.0 7.5 7.4 6.8 6.8 7.0 6.8 6.4 6.8 8.8

# > 80%

AVG PTS PER TREE 3,162 2,725 2,777 3,171 3,267 3,627 3,717 4,548 5,482 3,917 4,066 3,582 Areas

AVG FD HW DR Model 1 2 3 4 5 6 7 8 9 10 11 12 7 84% 83% 77% 93% 1b-304CNSR_GRP2 80% 70% 89% 77% 75% 75% 91% 100% 100% 73% 80% 100%

Page 9: Case Study #3 Identifying Tree Species From€¦ · Case Study #3 Identifying Tree Species From Geiger-Mode LiDAR Data: Compatibility and Accuracy Assessment CAPITOL STATE FOREST

Accuracy Testing – Sample 1 & 2 on Sample 3

FEBRUARY 14, 2017 35

Areas 1 2 3 4 5 6 7 8 9 10 11 12

Sample 1 & 2 ground truth trees available to be used in the TSI

model

Sample1+2-FD 10 11 2 14 1 1 16 22 10 28 16 23

Sample1+2-HW 5 7 8 2 7 28 4 5 1 1 1 1

Sample1+2-DR 7 8 12 15 11 3 7 1 0 0 9 0

Sample 3 - FD 5 8 1 6 0 0 10 14 8 17 11 14 Sample 3 - CE 0 0 0 0 0 0 0 0 0 0 0 0

Sample 3 - HW 3 4 6 1 5 16 3 4 1 1 1 1

AVG HGT 21.1 18.3 17.5 26.0 26.1 29.1 30.6 33.8 36.8 32.3 31.7 30.0 AVG ELEV 106.9 103.6 98.6 66.9 70.7 46.7 41.7 75.1 83.2 65.1 65.5 77.2

AVG INTENSITY 8.0 10.5 10.0 7.5 7.4 6.8 6.8 7.0 6.8 6.4 6.8 8.8

# > 80%

AVG PTS PER TREE 3,162 2,725 2,777 3,171 3,267 3,627 3,717 4,548 5,482 3,917 4,066 3,582 Areas

AVG FD HW DR Model 1 2 3 4 5 6 7 8 9 10 11 12 7 80% 75% 82% 89% 23-482CNSR_GRP3 85% 88% 85% 75% 82% 89% 64% 76% 88% 72% 79% 80%

Species Legend

FD Douglas Fir

HW Western Hemlock

DR Red Alder

Page 10: Case Study #3 Identifying Tree Species From€¦ · Case Study #3 Identifying Tree Species From Geiger-Mode LiDAR Data: Compatibility and Accuracy Assessment CAPITOL STATE FOREST

FEBRUARY 14, 2017 ILMF 2017: IDENTIFYING TIMBER SPECIES FROM LIDAR DATA – CASE STUDY 36

Cam Brown, MF, RPF Strategic Planning Forester Salmon Arm, BC Office: 250-832-3366 (ext 220) Cell: 250-833-6631

Mike Parlow Managing Director Qualicum Beach, BC Cell: 250-954-7850