Airborne LiDAR Data Acquisition for Forestry Applications Mischa Hey WSI (Corvallis, OR)
WSI Services Corvallis, OR Airborne Mapping: Light Detection and Ranging (LiDAR) Thermal Infrared Imagery 4-Band Multi-Spectral Imagery Bathymetric/Sonar Geodetic Survey Analysis: Forest Inventory and Vegetation Analysis Automated Feature Extraction. Water Quality Modeling. Fish & Wildlife Habitat Assessments.
Forest Research Affiliations Oregon State University USFS PNW, Forest Sciences Laboratory, Corvallis. University of Washington Precision Forestry Cooperative. USFS Rocky Mountain Research Station Moscow, ID. US Forest Service Research Lab, Portland and Seattle. Panther Creek LiDAR Research (BLM, EPA, Private Industry)
Presentation Outline LiDAR Concepts Forest Applications - Timber inventory, terrain mapping, roads, stream networks. Flight timing Acquisition specifications Processing needs Product development
Airborne Light Detection and Ranging (LiDAR) Lat/Long/EL (from GPS) Pitch (from IMU) Roll (from IMU) Heading (from IMU) Scan Angle Range and Intensity GPS Base
Airborne LiDAR Terms Laser Wavelength Field-of-View (FOV) Pulse Density- - Emitted pulses from sensor per unit area Return Density- - Pulses returning to sensor per unit area Discrete Return- - Individual return from emitted pulse Full Waveform- - Digitized entire wave returning to sensor Back-Scatter Intensity - Reflected energy from pulse
Acquisition Control Airborne instrumentation - GPS accuracy (PDOP, constellations) - IMU accuracy (drift, line length) - Base length < 13miles Ground control network - Monument occupation - Control point distribution - Considerations: access, security, sky visibility
Relative Accuracy (Line-to-Line Calibration) Individual flight-line swaths are spatially integrated. Good relative accuracy is essential for vegetation analysis. Relative accuracy is good QC measure.
Absolute Accuracy
Point classification Calibrated Points Automated algorithms Manual interpretation Ground Classified Points Full Feature Classification Accurate classification is essential for model development.
False vegetation from offset
Acquisition Timing When to go and why
Timing Considerations Leaf-on vs Leaf-off - Leaf-on => better canopy surface, spectral info from intensity - Leaf-off => increased canopy penetration, better ground model, hardwood/conifer distinction Climate and Other Factors - Snow, clouds, fog, smoke - Think broad patterns not specific days Find the balance: Leaf off, low flow, over 2,500 ft in PNW is tricky.
Acquisition Specifications The important numbers
Acquisition Specifications Side-lap and FOV - Decreases shadowing - Consistent point distribution Flight Line 2 Flight Line 1 Point density - Dictates resolution of information available - Higher density => increased ground returns, increased canopy detail (8 pts/sqm) - Ground return density can be 1/10 th the native density a heavily forested environment.
Point Density: More Points are Better 40pts/sqm 20pts/sqm 10pts/sqm 4pts/sqm 2pt/sqm
Real world examples Contracted LiDAR Survey Settings & Specifications Sensor Survey Altitude (AGL) Target Pulse Rate Sensor Configuration Laser Pulse Diameter Field of View GPS Baselines Leica ALS60 800 m 106 khz Single Pulse in Air (SPiA) 19 cm 28⁰ 13 nm GPS PDOP 3.0 GPS Satellite Constellation 6 Maximum Returns 4 Intensity ALS 60 LiDAR sensor install 8-bit Resolution/Density Average 8 pulses/m 2 Accuracy RMSE Z 15 cm Delivered Absolute Accuracy Relative Accuracy Sample 1,301 points 86 surfaces Average -0.001 m 0.034 m Median 0.000 mt 0.034 m RMSE 0.023 m 0.033 m Classification 1σ 0.024 m 0.016 ft 2σ 0.046 m 0.031 m Point Density First-Return 10.75 points/m 2 Ground Classified 6.03 points/m 2
Processing Considerations Treating the data right
Processing Considerations LiDAR classification - Ground, vegetation, building, utilities - High, medium, low vegetation - Water surface, bridges/culverts
Intensity Normalization Receiver auto-gain-control (AGC) Laser power emission variations Atmospheric transmissivity Laser Angle of incidence Line to Line Inconsistency Streaking is Occurring Throughout Image
Raw Intensities
Normalized Intensity
Products What should you buy
Products - Basic Point Cloud - Classified and Calibrated Points (LAS) Surface models - Bare earth DEM, Canopy DSM, Canopy height ndsm - Contours (requires smoothing tolerance) Intensity Image (normalized) Report and Metadata!
Products - Advanced Feature extraction - Road networks - Stream networks - Hydro breaklines - Building footprints Feature analysis - Stand delineation and characterization - Individual tree inventory and attribution
Network extraction
The Future Mapping yesterday s tomorrow today
Full Waveform Discrete Return: Capture only the exact time of the peaks of independently-recognized return pulses. Most current systems record up to 4 returns. However, new systems are starting to have more dynamic return recording. Full Waveform (FWF): the entire return signal is measured, allowing capture of subtle deviations in the shape of the reflected as compared to the shape of the outbound laser pulse
Full Waveform Green LiDAR point cloud highlighting 7 returns digitized from 1 outgoing pulse using Riegl s online waveform processing
Full-Waveform Considerations Advantages Detection of pulse stretching (return pulse wider than laser pulse) indicating: Low vegetation on ground, indicating need to adjust point elevation downward Improved classification by using combination of return pulse width and spatial context Indication of biomass by evaluating area contained under the pulse shape. Discrete Returns Considerations Massive storage requirements often require subsampling or switching drives during flight. More difficult to perform accurate geocorrection of the continuous wave-form. Limited software tools.
Multi-wavelength LiDAR Applications: o Forestry: Potential for species delineation using return intensity information. o Stream/Riparian: shallow water bathymetric data for surface water modeling, wetlands, and habitat assessment. Green laser NIR laser Green laser NIR laser
Summary Not all LiDAR is created equal. - High density, high accuracy are paramount. Consider all desired applications. - Get the most from your data. Talk to your vendors and outside experts. - Find a trusted source and be specific about your goals Cost and quality are tightly correlated. - Cheaper data is cheaper for a reason. 2007 LiDAR 8 pulses/sq m 2005 LiDAR 2-3 pulses/sq m
Mischa Hey WSI- Corvallis, OR mhey@wsidata.com www.wsidata.com