1. LiDAR Enhanced Forest Inventory
FAIB/Tolko/BCTS
Good afternoon
First I would like to thank you for inviting us. My colleague Ann Morrison was supposed to give this talk but she could make it. I am glad to be here as I think this group is pretty much linked to our group as we have common interest in geomatics.
Xiaoping Yuan
For the meeting in Kamloops
on May 2, 2016

2. Outline
Team: Xiaoping Yuan (PL), Chris Butson, James Wang, Chris Mulvihill, Geoff Quinn, Rob Kennett/Jamie Skinner
Background
LiDAR EFI basics
Okanagan/Kamloops LiDAR EFI 2015/16
New LiDAR data 2015

3. Background
- LiDAR research and activities across the country and world
- TFL 18 LiDAR Pilot 2006
- UBC (Alex Creek) LiDAR EFI 2014
- Northern Vancouver Island LIDAR EFI 2014/15
- Okanagan/Kamloops LiDAR EFI 2015

4. LiDAR Enhanced Forest Inventory
Why LiDAR?
A new way of acquiring inventory data
Automated, objective and modeling approach vs human, subjective, and interpretation
Comparative cost
Better and higher resolution inventory
Misperceptions

5. Input Data

6. LiDAR
LiDAR can provide highly accurate 3-dimensional characterization of forest canopy, sub-canopy, vegetation and terrain.
Forest stand attributes include:
stand height
Density/crown closure
basal area
diameter
volume
biomass

7. Input Data
Sampling design: representative plots representing all the possible variations of the inventory area
Tree/Vegetation cover measured, GPS positions
Aerial photos, existing inventory, and other auxiliary data

8. Output Data Modeling and predictions
Analysis Steps
Prediction options: traditional multivariate regression and non-parametric estimation
Gridded raster predictions
Generalization to polygons
Validation
Attribute Estimation/prediction
Area-based approach
Individual tree based approach

9. Okanagan/Kamloops LiDAR EFI
Partnership, LiDAR data acquired and owned by BCTS/Tolko
FAIB, under the data sharing agreement, for EFI
LiDAR area: 367,000 ha, completed 317,000ha (Area 1 and 2)
Objectives of the LiDAR EFI:
Design, test, and complete an operational inventory
Develop standards/specifications for operational inventory
Investigate methods for improving the traditional inventory (such as VRI)

10. Methodology Pre-inventory analysis and stratification
Ground sampling design and data collection and compilation
LiDAR metrics and prediction modeling
LiDAR EFI products and integration with the traditional inventories

11. Methodology
Pre-inventory analysis (VRI and LiDAR canopy) – Stratification (4)
Ground sampling (229 plots)
Predictions
Products: 20m raster, generalize to VRI 2015
Enhanced spatial and new inventory 2016

12. LiDAR Enhanced Forest Inventory Flowchart
Acquisition
Processing
Acquire LiDAR
Pre-process
Normalized and
classified point cloud
Convert to .laz
Retile and buffer
Remove duplicates
Create 1m DEM QA
Create 1m CHM
Normalized CHM
VRI 2014
BEC/TRIM
Generate 0.04ha
plot metrics
Generate 20m
metrics
Pre-inventory
analysis
Stratification
Sample plot
selection
Orthophoto
Ground data
collection
Ground data
compiled
Statistical
models
20m raster
predicted ATT
Metrics indices
Ground and modeling
Sampling Design
Landsat 2014
And 2015
Object segments
kNN/RF
processing
Processing for
VRIMS
VRI 2014
Processing for
VRIMS
VRI upgraded
& publishing
New LEFI
for VRIMS
Integration
Enhancement
It is a new inventory tool!

13. Results Stratification and Ground Sampling Stratum BEC Zone
Leading Species
Area %
Target # Plots*
Final # Plots 1
ICH and IDF FD 29 57 61 2
Others 27 55 69 3
ESSF and MS BL 21 41 43 4 23 47 60
Total
100
200
233

14. 2. Variables to be predicted
Variable name
Units
Utilization
Description
Ba_Ha
m2/ha
4.0 cm
Basal Area/Ha of live trees DQ cm
Quadratic mean Dbh of live trees
Calculated as sqrt(BA/TPH/ ).
Vht_Wsv
Whole stem volume/ha of live trees
Vha_nwd
Whole stem volume/ha of dead trees
Vol_nwb
12.5 cm
Net merchantable stem volume. Whole stem volume net of stump and top, Cruiser Decay, Waste and Breakage
ht_lor_l M
lorey height live trees
The basal area-weighted average height
3. Parametric vs Non-parametric
Traditional regression modeling
kNN and Random Forest imputations

15. 4. LiDAR data processing Original DEM Re-tile Normalized CHM
CHM 1m filled

16. 5. Statistical Analysis: parametric
Lorey Height by Strata
Volume by Strata

17. Volume against p90 without stratification
Circled for >100m3/ha of dead volume

18. Lorey Height: predicted vs actual by strata and combined
Basal Area: predicted vs actual by strata and combined
Whole Stem Volume: predicted vs
actual by strata and combined

19. 6. Statistical Analysis: non-parametric
Basal Area comparison
Whole Stem Volume comparison

20. Summary 2015/2016 Ground sampling, measured, and compiled 223 plots
LiDAR CHM, QA, and metrics
The 20m raster predicted BA, Quadratic diameter, gross/net volume, dead gross volume.

21. Summary 2016/2017 Integration with VRIMS
Enhanced spatial with new attributes
Evaluation and validation
Cost and benefit analysis
Satellite canopy evaluation
Implementation
Final summary report