LIDAR%20Data%20Visualization

1
LIDAR Data Visualization
Click on image
2
Forest Measurement and Monitoring using
High-Resolution Airborne LIDAR
Hans-Erik Andersen Precision Forestry Cooperative
University of Washington, Seattle Steve
Reutebuch Bob McGaughey USDA Forest Service
PNW Research Station Seattle, Washington
3
UW PFC LIDAR Research

• Evaluate LIDAR for
• Terrain mapping under canopy
• Forest inventory
• Forest structure analysis (canopy gaps, canopy
  cover)
• Canopy fuel mapping
• Comparison to field estimates at plot- and
  individual tree-level

4
Airborne Laser Scanning (LIDAR) System Components

• Active sensor emits 40,000 150,000 infrared
  laser pulses per second
• Differentially-corrected GPS
• Inertial measurement unit (IMU)
• Computer to control the system monitor mission
  progress
• Interesting targets

5
Multiple Returns

• Most laser systems can record several returns for
  each pulse
• Multiple returns occur when the laser beam is
  only partially blocked
• Part of the laser energy is reflected back to the
  sensor
• The remaining laser energy continues downward
• Up to 5 returns per pulse
• Typically only 2-3 returns
• Typically 1 -10 measurements per m2 or 4,000
  40,000 measurements per acre
• Most systems record the amount of energy
  reflected by target objects
• Intensity (near-infrared 1064 nm)

6
Airborne Laser Scanning (LIDAR) Technology

• Acquires 1-5 reflections (returns) per pulse
• Typically 1 -10 measurements per m2 or 4,000
  40,000 measurements per acre
• Data delivered as XYZ points in a data cloud
• Direct measurement of 3-D structure
• Terrain
• Forest vegetation
• Infrastructure

Adapted from Lefsky et al. (2002)
7
LIDAR-derived Bare-Earth Surface Model
8
How Accurate is LIDAR Terrain Model?
9
LIDAR Terrain Mapping in Forests
10
LIDAR Ground Accuracy

• Mean LIDAR DEM error ? 0.22 m Field boot
  height!
• St. Dev. LIDAR DEM error ? 0.24 m
• Maximum errors
• 1.31 m, -0.63 m
• Not significantly affected by canopy density

Source Reutebuch, S.E., McGaughey, R.J.,
Andersen, H.-E., and Carson, W.W. 2003. Accuracy
of a high-resolution lidar terrain model under a
conifer forest canopy. Canadian Journal of Remote
Sensing 29(5)527-535
11
LIDAR-derivedCanopy Surface Model
12
LIDAR-derivedCanopy Height Model
13
LIDAR-based canopy cover estimation
An estimate of canopy cover is generated from
first-return LIDAR points
LIDAR first returns in canopy (13) Total
LIDAR first returns (20)
65

14
LIDAR-based Canopy Cover Estimation
15
Plot-level LIDAR forest measurement

• Plot-level LIDAR metrics can be used to estimate
  forest inventory parameters
• Dominant height, basal area, stem volume,
  biomass, canopy fuel variables
• Multiple regression models used
• Independent variables based upon vertical
  distribution of LIDAR vegetation heights
• y f (hmax, hmean, hcv, h25, h50, h75, h90,
  LIDAR density)

Source Andersen, H.-E., S.E. Reutebuch, and R.J.
McGaughey. 2005. Forest measurement and
monitoring using high-resolution airborne lidar.
In Productivity of western forests A forest
products focus. Harrington, C.A., and Schoenholtz
(eds.) PNW-GTR-642, PNW Research Station,
Portland, OR.
16
Plot-level LIDAR forest measurement Capitol
Forest, WA

• 99 plots established at Capitol Forest across
  range of stand ages
• Plot-level forest inventory variables estimated
  using regression models

• Forest measurements in 0.2-ac. plot,
• mature (70 yr.) stand

17
Plot-level LIDAR forest measurement Capitol
Forest, WA

• Stem-mapped tree crowns within plot,
• mature (70 yr.) stand

18
Plot-level LIDAR forest measurement Capitol
Forest, WA

• Distribution of LIDAR point data within plot,
• mature (70 yr.)stand

19
Plot-level LIDAR forest measurement Capitol
Forest, WA

• Distribution of tree crowns and LIDAR point data
• within plot, mature (70 yr.) stand

20
Plot-level LIDAR Dominant HeightCapitol Forest

• (R 2 0.96 RMSE1.9 RMSEcv 2.5)

LIDAR-derived (x) vs. field (y) Line shows 11
relationship
21
Plot-level LIDAR Stem Basal AreaCapitol Forest

• (R 2 0.91 RMSE1.02 RMSEcv 1.03)

22
Plot-level LIDAR Stem VolumeCapitol Forest

• (R 2 0.92 RMSE7.4 RMSEcv 7.7)

23
Plot-level LIDAR Tree BiomassCapitol Forest

• (R 2 0.91 RMSE 43.5 RMSEcv 44.1)

24
Automated landscape mapping of forest attributes

• Basal Area

25
Automated landscape mapping of forest attributes

• Tree Biomass

26
LIDAR canopy fuel mapping

• Crown fires pose significant threat to forests
  and communities in western US
• Assessment of crown fire risk is a priority for
  many resource managers
• Accurate canopy fuel data needed to support fire
  behavior modeling and fuel management programs

27
Plot-level LIDAR canopy fuel measurement

• Plot-level LIDAR metrics can be used to estimate
  canopy fuel parameters (Capitol Forest test site)
• Canopy height (R20.98)
• Canopy base height (R20.77)
• Canopy bulk density (R20.84)
• Canopy fuel weight (R20.86)
• All measures are hi-res, geospatial data at
  landscape-level ? GIS layers of fuel variables

Source Andersen, H.-E., R.J. McGaughey, and S.E.
Reutebuch. 2005. Estimating canopy fuel
parameters using LIDAR data. Remote Sensing of
Environment 94441-449.
28
LIDAR Canopy Height GIS Layer
29
LIDAR Canopy Fuel Weight GIS Layer
30
LIDAR Canopy Bulk Density GIS Layer
31
Automated LIDAR individual tree recognition
32
Rigorous assessment of LIDAR tree height
measurements

• Previous studies show high correlations between
  field- and LIDAR-derived tree height measurements
• Field tree ht. measurements are relatively
  imprecise
• Information on absolute accuracy of tree ht.
  measurements from LIDAR is needed for forest
  inventory
• Influence of laser beam diameter, species (pine
  vs. Doug-fir), and DTM error
• Comparison to conventional field methods (Impulse
  laser clinometer)

33
Rigorous assessment of LIDAR tree height
measurements

• Study carried out in Fort Lewis
• Military Reservation, WA

• Extremely accurate tree ht. measurements acquired
  using surveying instruments (total station)
  least squares adjustment

• Average error of tree top measurements from
  survey 2 cm!

• Lidar tree hts. measured in FUSION software

34
Rigorous assessment of LIDAR tree height
measurements

• Tree heights from narrow beam LIDAR (-0.73 0.43
  m) more accurate than wide beam (-1.12 0.56 m)

• LIDAR hts. for Ponderosa pine (-0.43 0.13 m)
  more accurate than for Douglas-fir (-1.05 0.41
  m)

• Heights from conventional field techniques
  (-0.27 0.27 m) more accurate than LIDAR (-0.73
  0.43 m)

Andersen et al., Can. J. of Remote Sensing, In
review
35
LIDAR-based Species Class Recognition

• Near IR intensity of laser reflections related to
  species type (esp. in leaf-off conditions)
• Infrared reflection stronger from live foliage
  than from branches and stems
• In leaf-off conditions, LIDAR intensities are
  higher from conifer foliage than hardwood crowns
• Preliminary analyses indicate that
  differentiation between conifer species is
  possible

36
Hard/softwood determination using LIDAR active
infrared intensity data
Leaf-on orthophoto
LIDAR IR Intensity
Leaf-off hardwoods dead trees
Conifers/evergreens
37
Automated species recognition using LIDAR IR
intensity data

• Individual tree crowns can be classified into
  conifer/hardwood classes

Orthophoto
Classified Individual Tree Crowns
(CONIFER/HARDWOOD)
Segmented Individual Tree Crowns
38
University of Washington LIDAR intensity
image 6-20 pts/square meter Acquired March 17,
2005
39
All returns colored by intensity
40
All returns colored by intensity Above ground
objects
41
Monitoring Growth with LIDAR
Individual Tree LIDAR Datasets
2003
1999
1998
42
LIDAR-based measurement of growth

• Multitemporal, high-density LIDAR data can be
  used to measure the growth of individual trees
• Individual tree heights measured in 1999 and 2003
  LIDAR datasets using automated methods
• Height growth map generated for large area

43
LIDAR-based measurement of growth (cont.)
1999 2003
44
LIDAR-based measurement of growth (cont.)

• Difference represents individual tree height
  growth from 1999-2003

• Enables detailed, spatially explicit analysis
  of site quality and productivity

45
Current Research LIDAR forest sampling

• 121 FIA plots on Kenai Peninsula covered by 2004
  LIDAR flight
• Plot-level variables estimated with high-density
  LIDAR
• UW PFC working with PNW-FIA to develop field
  survey protocol for measurement of accurate plot
  locations

46
Possibilities for Future Research

• Further development and validation of regression
  models using independent LIDAR field data
• Application of methodology in different forest
  types
• Douglas-fir/ponderosa pine forest in eastern
  Cascades, WA
• Spruce/birch forest on Kenai Peninsula, Alaska
• Chaparral and mixed-conifer in Southern
  California
• Fusion of high-resolution multispectral imagery
  and 3-D lidar data

47
Questions Discussion