1. Lidar Data Applications for Natural Resource Management
Tom Bobbe, Mark Finco, Ken Brewer, Denise Laes
USDA Forest Service
Remote Sensing Applications Center
Salt Lake City, Utah
Geospatial 2007 Conference
Thursday - May 10, 2007

2. Presentation Outline Lidar system fundamentals
Resource management applications
Digital Terrain Models
Vegetation Models
Lidar applications in the Forest Service
Lidar acquisition specifications

3. Fundamentals of Lidar Lidar Basics:
Lidar = Light Detection And Ranging
Scanning Infrared Laser Rangefinder
thousand pulses per second result in typical point densities between 8 per 1-m2 to 1 per 4-m2 (called post spacing)
Multiple returns from a single pulse are possible
Coupled with IMU/GPS provides very accurate X,Y,Z point clouds (~15-cm in Z).
In the most general sense, the fundamental principals of lidar are very simple … and as the 1st speaker they should be presented very briefly …
Lidar is an acronym …
Lidar provides measurements from which images can be derived … not images themselves
At a lidar system’s heart is a laser rangefinder and a survey grade gps
… add a scanning system and lidar systems can acquire 100,000 or more height measurements every second
Most modern systems can collect multiple (often 4) height measurements from each pulse of the laser – point out example in figure of 2 returns
These height measurements are actually distance from laser measurements … coupled with GPS/IMU they create what we call DIGITAL TERRAIN MODELS (to differentiate them from digital elevation models)

4. Characteristics of Lidar Data
Point data, but …
Large volume of data
Assume: 1 to 4 pulses / m2
Assume: 2 returns per pulse
Assume: 6 values per return
Equals: 0.38 – GB per acre, or – TB per 10,000 acres
Because of data volume
Often standard GIS analyses don’t work
Require special pre-processing for analysis
Point data, but this is complicated by the volume of data that needs to be analyzed
Large volume of data
Assume: 1 to 4 pulses / m2
Assume: 2 returns per pulse
Assume: 6 attributes per return … e.g., x, y, z, I, angle, return #
Equals: 0.38 – GB per acre, or – TB per 10,000 acres
Because of data volume …
Often standard GIS analyses don’t work … or work so slowly that they aren’t practical
Many times tempted to move away from the vector format and quickly turn into a raster, but this amounts to throwing information away
Require special pre-processing for analysis. This includes tiling the data into smaller, more manageable, chunks
How big is your study area?
Consider these characteristics before diving in …
The image is of a study area in western Montana (Lubrecht Expt’l Forest) – 77k acres, 2km tiles for analyzability

5. Examples of Lidar Point Clouds
This lidar point cloud transect crosses a forest road
Points Colored by Height
Highest
Lowest
These are simply examples of what lidar point clouds look like colored by elevation.
The top image – lidar can be used as a profiler to investigate road cuts and gradients. Note that individual trees are visible within the point cloud.
The bottom image – Lidar data, however, is collected using either a fixed wing aircraft or helicopter. The resulting dataset can be viewed as a 3-d point cloud. Note in this forested image that there are some man-made structures which are visible under the canopy.
In this 3-D perspective of a lidar point cloud note the buildings

6. Multiple return lidar
All Returns
3rd Return
4th Return
2nd Return
1st Return
Multiple return lidar contributes to forest structure measurements
1st return is not just top of canopy
Last (4th) return is not just the ground
First analytical step typically filters ground returns from all returns
Figures Courtesy of PNW Seattle Laboratory

7. Primary Application – High Resolution DTM
The first application for lidar data was the creation of high resolution Digital Terrain Models
Note: we use the acronym DTM not DEM because DEM is a data format and is usually assumed to be a USGS product. DTM is a more generic term.
The largest federal user is FEMA for flood plain mapping
This slide visually demonstrates how the 3-d point cloud can be used to generate high resolution Digital Terrain Models (DTM).
The lidar points above the ground are typically discarded … vegetative overburden
The magic of lidar data analysis is the separation of ground returns from above ground returns
Click and the point cloud transitions to a shaded relief map
Note: it becomes obvious that this area is a road intersection. The main road also has a 2-track path or frontage road merging with it.

8. 10-m DEM / 1-m Lidar DTM Comparison
USGS 10-meter Digital Elevation Model (DEM)
Lidar-derived 1-m Digital Terrain Model (DTM)
Site A
Site A
Site B
Site B
This is a comparison of a 10-m USGS DEM and the 1-m Lidar-derived DTM
Differences are obvious. Micro-topography in the lower left hand side of the image is interesting.
Two sites – A and B – are going to be compared at larger scale in the next slide.
These high resolution DTMs seem to answer a lot of questions … but they also raise a few more
For example, if you have a 1-m DTM and there is a 36” stump on the ground … should it be included in the DTM as part of the “terrain”? (do we have a choice?)
New and important features are recognizable on the 1-meter digital terrain model (micro-hydrologic patterns, roads / trails, and other man-made features)

9. Comparison Areas USGS 10-meter Digital Elevation Model (DEM)
Lidar-derived 1-m Digital Terrain Model (DTM)
Site A
The comparison of site A shows that an earthen dam is easily discernable in the 1-m lidar DTM, but wholly un-recognizable in the 10-m DEM. Also note the road system leading to and away from the dam.
The site B comparison simply shows the micro relief in the 1-m DTM
Though interesting visual comparisons, these sites and other comparisons have practical implications for engineering applications. Consider, for example the improved cut-and-fill estimates and run-off/erosion calculations that could be created with 1-m DTMs.
Site B

10. DTM’s are just the beginning however …
Tools are being developed in the Forest Service and commercial sector to extract information about the vegetation
Individual Tree Measurements (potentially height, crown base height, crown diameter depending on crown spacing)
Canopy Height, Cover, Density
Vegetation Structural Characteristics
In addition to the bare earth DTM’s, the lidar data contains a wealth of information about the vertical distribution of vegetation (structure) on the landscape.
While these applications are relatively new (<5 years for areas of any size) they are rapidly being facilitated by new software being developed within the Forest Service and commercially
Some of the leaders in resource application development are right here in the USFS
PNW Seattle Lab – McGaughy and Reutebuch … application called FUSION for lidar data visualization and analysis
FUSION is distributed and supported by RSAC … (say more?)
RMRS researchers in Moscow, Idaho (and Others) are developing resource analysis methods and other applications …

11. Fusion Software
Developed by USDA Forest Service Pacific Northwest (PNW) Research Station (McCaughey, Reutebuch & Andersen)
Originally intended for PNW internal use
RSAC agreed to distribute and provide support for FS users
Capabilities include:
View lidar data quickly and easily
Handles almost any format of lidar data
Creates surfaces (bare earth models (DTMs), canopy surface models)
QA/QC of vendor-processed data
Easily measures heights of features
Large number of forestry-related measurements
And much more…
* One lidar analysis tool, FUSION, which was developed at USFS PNW in Seattle allows for (among many other things) the measurement of individual trees.

12. Fusion Tutorial

13. Lidar Tutorial

14. Vegetation information starts with DTMs
Raw lidar returns (Colored by height)
DTM
All returns normalized by the DTM
DTM extraction is a necessary 1st step in most vegetation analyses …
The image shows the progression from raw lidar points (top), to bare earth DTM (middle), to lidar points that are normalized based on the DTM … this changes the color scheme and you can see tall trees in the front and right of the image that were not visualizeable without the normalization.

15. USFS PNW’s FUSION Software
Individual Tree Measurements
FUSION facilitates many types of measurements …
One set of tools are provided to individual trees …
Other tools allow for the summarization of the lidar returns over larger areas

16. Lidar and ground measurements relationships
Strong relationships with ground measured variables
Height, Basal Area, Volume, Crown Bulk Density, etc.
Relationships verified by numerous researchers
McGaughy, Reutebuch & Andersen (USFS PNW)
Hudak and Evans (USFS RMRS)
Lefsky (Colorado State)
Evans (Mississippi State)
Wynne (Virginia Tech)
Popescu (Texas A&M)
Naesset (Norway)
Many others …
Dominant height (r 2 = 0.98)
Relationships between lidar data and ground-based forestry measurements are very strong …
Modeled height, BA, Volume, CBD and other structural variables are strongest
Some work (primarily at PNW) has been done to use the intensity attribute provide additional information (e.g., sub-lifeform)
These relationships have been confirmed by many researchers
... The next challenge is to use these relationships to meet an operational mapping objective
Figures Courtesy of PNW Seattle Laboratory

17. Lidar Applications in the USFS
Recent tally of lidar applications in the USFS (Lachowski and Reutebuch)
A recent study by RSAC and PNW reports on the known lidar applications / studies in the USFS
More detail and full report at

18. Lidar Mission Specifications
Wide lidar usage (in resource mapping) is just in its infancy
Like aerial photos – specifications are linked to information requirements
Currently no industry standards for specific applications
2 Areas to specify
Acquisition specs
Processing and Delivery specs
Government
Vendor
Specs
QA / QC
Setting specifications for lidar data acquisitions are not available as “boilerplate” for contracting …
1st step is to clearly understand and articulate (better yet write)
Experiences of research and pilot projects – and other agencies – can be used as guidance

19. Lidar Specifications – Acquisition
Acquisition Specifications
Point density (post spacing)
DTM -> based on vertical accuracy requirements
Vegetation Applications
1.5 point per square meter absolute minimum
4-6 points per square meter are preferable
Specify whether collected leaf on or leaf off
Multiple returns per pulse
Maximum 15-degree off nadir scan angle unfiltered data
Flight lines should have 50% “side lap” (30% minimum)
Cross flights for calibration
Attributes delivered: X, Y, Z, Intensity, Scan Angle, Return #
High resolution digital imagery (if possible)
Acquisition Specifications
1.5 point per square meter absolute minimum; 2-4 points per square meter are preferable (note: ask the vendor specify nominal point density, not "contour intervals")
Multiple returns per pulse (typically a maximum of 4; you want more than just first and last return lidar) – required if you need vegetation structure information (more than just top of canopy)
you may want to specify whether the lidar is collected leaf on or leaf off, depending on mission objectives … often leaf off in hardwood systems
Sidelap allows a more detailed analysis of the sub-top of canopy structure because you can look into the trees
maximum of 15-degree off nadir scan angle unfiltered data (i.e., unfiltered for ground returns) should be delivered in LAS format
one or more cross flights for calibration vertical and horizontal accuracy specification should be provided by vendor (typically ~15-cm vertical, 30-cm horizontal w/95% CI)
attributes in LAS file should be X, Y, Z, Intensity, Scan Angle, Return Number, and [Optional: GPS Time]

20. Lidar Specifications – Processing and Delivery
Vendor Processing and Delivery Specifications
Lidar data delivered in overlapping tiles
GIS dataset of the tiling system
GIS dataset of the flight lines
Report on GPS ground station locations
Geographic projection information (including vertical datum)
Heights should be orthometric heights
Report that lists all files delivered
Optional:
Tiled points filtered for bare earth returns
A high resolution DTM
Vendor Processing and Delivery Specifications
lidar data should be tiled into practical units, often 2-km x 2-km; a GIS dataset of the tiling system should be included
lidar data should be delivered with a GIS dataset of the flight lines used to collect the data, typical overlap is about 30%
a report on what GPS ground station locations were used to reference the data
all geographic projection information should be included; heights should be orthometric heights
a list of all files delivered to verify contract performance and full delivery
you may want to specify whether the lidar is collected leaf on or leaf off, depending on mission objectives
[Optional: Tiled points that have been filtered for bare earth returns]
[Optional: A high resolution digital terrain model (1-m or 2-m resolution)]
[Optional: High resolution photography (or digital equivalent) is a very valuable addition to the lidar data and can often be acquired at the same time]

21. Approximate Costs of Acquisition
Mobilization
$8k – $15k
Administration
Project and flight planning
Weather contingency
Pre-collection tasks
Basic Data Collection and Post-processing
Depends on study area size ($0.50 -$2.50/acre for 1M – 15k acres)
~$1/acre for a 250K acre project
Raw lidar data
Bare earth
First surface
These figures are presented very hesitantly … it is a dynamic market
One downside of lidar data is it’s relatively high cost. This too is changing, however.
Costs are a function of how much area is to be collected … and more so on the derived products you want the vendor to provide. Note that the third level of processing (above, Generating GIS Data Sets) is only available from a very limited number of vendors and are by no means standard products.
Lidar data consortiums are being created to collect lidar data over entire states –Carolinas, Wyoming, Nevada?, Washington?
The image shows the progression from raw lidar points (top), to bare earth DTM (middle), to lidar points that are normalized based on the DTM … this changes the color scheme and you can see tall trees in the front and right of the image that were not visualizeable without the normalization.
Advanced Processing
Additional $3 – $7/acre
Canopy cover
Tree height
Forest biomass
Other vegetation derivatives

22. Summary Lidar is an exciting (relatively) new technology
Provides measurements!
Vegetation structural information are its strengths
Existing research provides a strong foundation
Lidar processing requires special skills/tools
Data volume can be an issue
Specialized software (not just ESRI products) required for efficient large scale analysis
Lidar missions
Specifications becoming better understood
Still expensive, but costs coming down
Multiple resource applications & consortia allow for cost sharing