Use of remote sensing in research and industry:
an international perspective Michael Watt
Contributors - Erik Næsset – Norway, Matti Maltamo – Finland, Håkan Olsson
–Sweden, Mike Wulder – Canada, Ron McRoberts – USA, Terje Gobakken –
Norway, Russell Turner – Australia
.
Presentation Overview
• Overview of LiDAR and satellite imagery
• What is remote sensing being used for?
• Cost-benefit analyses
• Future applications
Light detection and ranging (LiDAR)
Active remote sensing technology
LiDAR emits ~ 200,000
pulses per second
Records up to 4 returns per pulse
Each return is converted to 3D point
1. Direct measurements Canopy height
2. Derived Canopy structure/density
3. Topographic measures
LiDAR - Trends
• Systems progressively improved from
dual return to multi-return with increased
laser strength and pulse densities.
• Lower point densities being flown
• Operationally used for inventory
purposes worldwide – wall to wall and
strip sampling
• Value recognised for terrain
characterisation
• Research focus moving towards
evaluating wood properties, LAI,
waveform LiDAR, single tree
delineation and modelling processes.
Mapping Applications
Source: Mark Forward. Nelson Forests Ltd.
Digital Terrain Models
• One of the most widely
used applications of LiDAR
• Can be used for :
Harvest planning
Road planning
Machine slope classing
ID hazards, e.g. escarpments
Erosion mapping
Source: Mark Forward. Nelson Forests Ltd.
Inventory
• Link LiDAR metrics to stand
dimensions
• Used to develop spatial maps
LiDAR mean height (m)
0 10 20 30 40
Tota
l ste
m v
olu
me (
m3 h
a-1
)
0
200
400
600
800
1000
1200
LiDAR 95th height percentile
0 10 20 30 40 50
Me
an t
op
he
ight
(m)
0
10
20
30
40
50
Source: Mark Forward. Nelson Forests Ltd.
Single tree delineation
Where are Lasers used Experimentally or Commercially in Forest Inventory
EUROPE AMERICA ASIA OCEANIA AFRICA
Finland Canada Japan Australia South Africa
Norway USA Nepal
New Zealand
Sweden Chile China
Austria Brazil
The Baltic States Uruguay
Bulgaria
Denmark
Germany
Poland
Portugal
Spain
Switzerland
UK
Italy
Extent of Use
• Operational stand based inventories have been
conducted in Norway, Finland and Sweden since
2002–2004.
• 2 Million ha inventoried in Finland
Stand management system developed 2005-10
LiDAR is about 1 pulse/m2.
• Surveying companies acquire LiDAR data and deliver
outputs describing forest inventory products.
• Stand metrics derived from LiDAR (0.5 – 1.0 pulse /m2)
for all forestry land in Sweden (23 Million ha)
Timeline to Implementation – Nordic Countries
Extent of Use – Australia
• Queensland, flood risk, coastal forests
• NSW, used operationally in
ca. 300,000 ha forest
Almost 1 M ha LiDAR coverage
• South Australia, ForestrySA
operationalised 2007
• Tasmania, FT undertook operational
trial 2007 by ‘double planning’ coupes
Operationalised 2009
Currently 700,000 ha LiDAR coverage
14/52
2010 National
Lidar Transects 25,000 km
Lidar metrics gridded to
25 x 25 m cells
> 20 Million lidar-plots
Source: Wulder, M.A.,
Canadian Forest Service
Volume of Forest Stands Costs & Error (precision): 2001
050
100150200250300350400450500
10 12 14 16 18 20 22
Co
sts
(U
SD
pe
r h
a)
Error (%)
Method 3
Method 1 Method 2
Laser
Method 1: Relascope plots
Method 2: Photogrammetry
Method 3: 15 sample plots per stand
Cost / Benefit Analysis
Early work demonstrated that LiDAR-assisted inventories were cost-effective
when compared to alternative methods
Forest inventory approaches – Case Study
(based on airborne laser scanning)
Source: Gobakken et al., 2012
Study area and datasets
• Field measurements
2407 trees measured
23 sample plots
• Airborne laser scanning
Optech ALTM 3100 EA
Source: Gobakken et al., 2012
Diameter distribution
18
Conclusions
• The value of the inventory information depends of what
the data are going to be used for.
• Analysis showed the prediction of diameter distributions
from LiDAR (ABA-DD) was the most cost effective
inventory method
• Further studies are needed to verify the results.
19
Satellite imagery
Satellite imagery - Trends
Typical Resource Applications
• Resource updates
• Growth variation mapping
• Repeat monitoring
• Image costs have dropped
• Data is available and frequent
• Resolution is appropriate for planning
• Fits into existing GIS systems
• Easy to implement
Landsat 30 m
High resolution 5 m
Three Levels: that interlink together to cover the rotation
1. Strategic Planning 2. Operational Planning 3. Precision Forestry (LiDAR) UAV?
5 m imagery Targeted coverage based on 1
Information Linkages
Increased Cost
Decreased update frequency
Approximate Data Costs vs. Detail
RapidEye – 5 m resolution
Wide area, daily imaging, data cost low. Ideal
for frequent monitoring & targeting use of
other options i.e. Level 2
Digital Globe – or aerial Imaging
Smaller area, 3-5 day imaging.
Ideal for mapping updates
LiDAR & CIR photography
Level 1: Strategic Planning
Level 2: Operational Planning
Level 3: Precision Forestry
1 cent / ha
33 cents/ ha tasked
16-19 cents/ ha archived
• Strategic use of different systems that result in cost savings through better information.
• Cost of the option. The decision maker should not pay for information that is more expensive
than the expected improvement in the value of the decision.
Data cost (USD)
~$ 4-10 /ha
Level
UAV?
Applications Satellite Imagery
Complements other technologies by identify areas where further
information is required.
• Plantation establishment success – i.e. has the crop established
• Variation in growth across plantations
• Monitoring of change caused by – i.e. fire, theft, termites, wind
damage or foliage disorders
• Evaluate progress of planned silvicultural operations and
harvesting
• Provides a cost-effective way of updating plantation mapping
Level 1: Strategic Planning – providing a cost effective overview
Overview
Data cost low at 1 cent/ha
• Highlights areas that
require further
attention. Due to fire,
disease,
encroachment or
plantation variation or
failure
• Enables long term
repeat monitoring
once GIS boundaries
are up-to-date
Disease Crop variation Fire
Failed areas
Examples: NZ, Aust. Brazil and China
Source: Watt P.J Indufor 2013
Detection of Harvesting and Thinning
The following slide transition shows harvesting and thinning operations as detected The first image is Jan 2012 and the second May 2012. From the images harvest progression is able to be monitored and estimated over this period.
P. radiata Australia
Source: Watt P.J Indufor 2012
Stand variation routine applied to
2 m DigitalGlobe imagery.
What does it provide?
• Semi-automated mapping of
plantation gaps and areas of
poor performance.
• Allows for improved
stratification
• Improves stocked area
estimates which lead to
improved woodflow / yield
projections.
Mapping Variation
Rubber Cambodia
Source: Watt P.J Indufor 2013
UAV – 90 day Monitoring Plantation Establishment
• Individual Tree identification
• Identifying weed areas
S. America Eucalypt plantation
Source: Watt P.J Indufor 2014
Using Sensefly UAV fitted with camera (5-7 cm) pixels. Routine used to firstly map canopies and then plot tree locations.
New Zealand Industry Where are we at? • Strong link between research and industry uptake
• LiDAR • First trialled MfE 2005, Ernslaw 2006
• Used operationally 2007 MfE for carbon
• 2012 onwards for inventory, Blakely Pacific, Nelson Forests, KT
• Road and harvest planning
• LiDAR-assisted inventories
• Satellite Imagery Mostly only trialled in NZ
• Mapping variability
• Forest changes – GIS quality checks, harvesting, large-scale
events
• Needle cast detection (proof of concept)
• Multispectral LiDAR
Colour LiDAR
Vertical profiles of indices
True LAI vs plant LAI
• Terrestrial Laser Scanning
Technology behind airborne LiDAR in terms of
integration
Portable instruments, hyperspectral, dual wavelength
Integration with aerial LiDAR – 3D forests
Future research directions
• Aerially derived Digital Surface Model
Require initial DTM from LiDAR
• Less expensive if aerial imagery for other purposes can
be used
Future research directions
• Unmanned Aerial Vehicles (UAV)
Sensor improvements, flight times increased and detection
routines improved
• Wood properties from LiDAR
• Linking information from harvesters to remotely sensed
data
Future research directions
Acknowledgements
• Pete Watt – New Zealand
• Erik Næsset – Norway
• Matti Maltamo – Finland
• Håkan Olsson – Sweden
• Mike Wulder – Canada
• Ron McRoberts – USA
• Terje Gobakken – Norway
• Russell Turner – Australia
http://www.scionresearch.com/gcff
Michael Watt
Research Leader – Forest Operations and Monitoring, Scion