Considerations%20for%20Planning,%20Acquiring,%20and%20Processing%20LIDAR%20Data%20for%20Forestry%20Applications

1
Considerations for Planning, Acquiring, and
Processing LIDAR Data for Forestry Applications

• Robert J. McGaughey
• USDA Forest Service--PNW Research Station
• Hans-Erik Andersen
• University of Washington-Precision Forestry
  Cooperative
• Stephen E. Reutebuch
• USDA Forest Service--PNW Research Station

University of Washington-Precision Forestry
Cooperative
Pacific Northwest Research Station-Silviculture
and Forest Models Team
2
Outline

• Brief LIDAR overview
• LIDAR data characteristics deliverables
• QA/QC procedures
• Data processing
• Introduction of FUSION software
• Conclusions

3
LIDAR

• Light Detection and Ranging
• 4 types
• Atmospheric
• Continuous waveform
• Discrete return (profiling)
• Discrete return (scanning)
• Airborne Laser Scanning (ALS)
• Discrete return (scanning) mounted on aircraft

4
ALS System Components

• Scanning laser emitter-receiver unit
• Differentially-corrected GPS
• Inertial measurement unit (IMU)
• Computer to control the system monitor mission
  progress
• Interesting targets

5
Multiple Returns

• Many 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
• Many systems record the amount of energy
  reflected by target objects
• Intensity (near-infrared, 1064 nm)

6
Multiple Returns

• All returns (16,664 pulses)
• 1st returns
• 2nd returns (4,385 pulses, 26)
• 3rd returns (736 pulses, 4)
• 4th returns (83 pulses, lt1)

7
Comparison With Other Remote Sensing Technologies

• Active sensor
• Laser pulse is emitted and reflected energy is
  measured
• Passive systems rely on reflected solar energy
• Returns are actual measurements
• Range is computed based on round-trip travel time
  for laser energy
• Combined with accurate aircraft position and
  attitude to produce XYZ point measurement
• Small footprint at target
• 30-100 cm footprint at ground surface
• 4 pulses/m2 is common
• Multiple returns over porous targets

8
(No Transcript)
9
(No Transcript)
10
(No Transcript)
11
LIDAR Accuracy

• RMSE provided by LIDAR system manufacturers
• 10-15cm vertical
• 50-100cm horizontal
• Several studies provide independent verification
  of these values
• Ground survey points
• Tree heights
• Building heights
• Alignment of power transmission lines

12
Bare-Earth Accuracy

• Mean LIDAR DEM error
• 22 cm9-inch field boot height!
• Maximum errors
• 1.3 meter, -0.63 meter(4.3 ft, -2.1 ft)
• Error is not significantly affected by canopy
  density

13
Tree Height Accuracy

• Recent study uses total station to accurately
  measure tree heights in relatively open forest
• High density LIDAR (4 pulses/m2)

Narrow beam (0.3 mrad) Mean Std. Dev. Wide beam (0.8 mrad) Mean Std. Dev.
Douglas-fir -1.05 0.41 m -1.49 0.56
Ponderosa Pine -0.43 0.13 m -0.77 0.24
Overall -0.73 0.43 m -1.12 0.56 m
14
LIDAR Data Characteristics

• High spatial resolution
• Typical density is 0.5-6 pulses/m2
• 2-3 returns/pulse in forest areas
• Surface/canopy models typically 1-5m grid
• Large volume of data
• 5,000-60,000 pulses/hectare
• 12,500-150,000 returns/hectare
• 0.3-3.6 Gigabyte/hectare

15
Typical Mission Specifications
Topographic mapping Vegetation mapping
Scan angle 20 12
Flying height 2200 m 1200 m
Pulse rate 10-70 kHz 30-100 kHz
Beam footprint 66 cm 36 cm
Swath width 1600 m 510 m
Pulse spacing 1-3 m 0.2-0.8 m
16
LIDAR Contracting

• Area selection and specification
• Data specifications
• Deliverables
• Data acquisition and delivery
• QA/QC Assessment
• Data processing
• Final products

17
DeliverablesGeneral Considerations

• Delivery format
• Return data, surfaces, images, GIS coverages
• Can you read the format?
• Does the data contain all the information you
  need?
• Delivery media
• External hard drives are common but not easy to
  backup to more stable media
• You would like to be able to retrieve single data
  files from your backup
• What do you do with multiple DVDs?
• How do you know what has been delivered?
• Contracts need to include specific deliverables
  to help you assess overall data quality and
  completeness
• Deliverables based on the actual datanot just
  coverage area boundaries
• Do you have adequate storage space to move data
  onto faster devices?
• Disk space
• Bandwidth

18
DeliverablesCommon Data Products

• Metadata
• Flight information
• LIDAR system settings
• Data coverage
• Bare ground products
• Bare-ground returns
• Surface models
• Return data
• Coarse filtering to remove outliers
• Includes return number
• Always get the return data for a project
• Adds very little to contract cost
• Will cost you if you decide you want it later

19
DeliverablesForestry-Specific Products

• Canopy height models
• Normalized using bare-ground surface
• Filtered to remove buildings, powerlines, and
  other above-ground features
• Canopy cover maps
• Presence/absence of vegetation
• Vegetation density (percent cover)
• Geo-referenced LIDAR intensity images
• First return intensity value (reflected energy)
• Useful an image (B/W IR image)
• Useful as a layer for further analysis

20
QA/QC AssessmentDid you get what you ordered?

• Initial delivery
• Missing data
• Missing returns
• Misclassified returns
• Tile naming inconsistencies
• Contractor used custom software to produce ASCII
  formatted data
• Several flight lines were omitted
• They had no way to view the ASCII files
• They didnt know what they delivered
• Over 60 client-hours to sort out the problems
• Contractor made 3 deliveries over a 5-week period

21
QA/QC AssessmentQuality and Completeness
Total project area 47,818 ha (118, 111 acres)
Total returns 2.9 billion
Total file storage space (LAS binary format) 77 Gb
Data tiles (processing bins) 277
Tile size 1 km wide by 2 km tall (200 ha)
HTML report generated by FUSION-Catalog HTML report generated by FUSION-Catalog
22
Data ProcessingWhat do I do with all these
points?
Raw data are interesting to look at but require
extensive processing to create useful information
23
Data ProcessingBare-Earth Surface Model
24
Data ProcessingCanopy Surface Model
25
Final ProductsCanopy Height (Raw Data)
26
Final ProductsPercent Cover (2.5m grid)
27
Final ProductsLIDAR Intensity Image (1.25m grid)
28
Final ProductsLIDAR Intensity Image (5m grid)
1m grid
29
Coniferous/Deciduous Classification Using
Intensity Values (Raw Data)
30
FUSION Software
31
FUSION Software

• Displays several kinds of data
• Allows users to interactively select portions of
  large datasets for viewing
• Users can mine the data to discover new
  information
• Clips all data layers and makes them available
  for detailed 3D viewing
• Supports stereoscopic viewing
• Runs on a current hardware
• Available through RSACincluded on DVD you were
  given at registrationdemos on Thursday

32
Main System components

• FUSION 2D interface to several data types
• Allows extraction of data subsets
• Interacts with LDV to display samples
• Includes tools to develop images using LIDAR
  point cloud (colored by elevation or intensity)
• LDV 3D data visualization
• Very interactive
• Provides a variety of display options
• Allows direct measurement in data
• Provides structured measurement protocol for
  measuring tree attributes
• Provides analysis framework for prototyping
  analysis strategies

33
Command Line Components

• Catalog
• Creates QA/QC summary for project area
• Intensity image for project area
• Outlier analysis
• GroundFilter
• Filters all-return data to obtain bare-ground
  returns
• GridSurfaceCreate interpolates regular grid using
  bare-ground points
• VegetationMask (in progress)
• Identifies areas with vegetation cover
• Identifies and masks buildings, powerlines, and
  most man-made, above-ground features
• Provides mask for all vegetation analysis
• CanopyModel
• Creates canopy surface or vegetation height
  models
• PercentCover
• Computes canopy cover estimates
• Computes vertical structure indices (in progress)

34
Command Line Components

• ClipData
• Extracts subsets of return data
• Usually used to clip individual tree- or
  plot-level samples
• CloudMetrics
• Computes statistical metrics for data subsets
• Provides basis for plot-level regression with
  field plot data

35
Conclusions

• LIDAR can help foresters characterize spatial
  variability in forest conditions at resolutions
  beyond our wildest dreams
• Cost is decreasing while LIDAR system
  capabilities are increasing
• Analysis procedures are being defined and refined

36
Conclusions

• The number of commercial LIDAR providers is
  increasing
• More competition, more work, more interest in
  uses other than bare-earth modeling
• Large LIDAR acquisitions are underway
• Foresters and other resource specialists need to
  be at the table when decisions regarding data
  specifications are made
• Raw return data is always valuable even if the
  analysis tools and methods are not fully mature
• FUSION will be demonstrated Thursday