Chapter 12
Remote sensing applications
351
Remote sensing and GIS in forestry
Michael A. Wulder, Ronald J. Hall, and Steven E. Franklin
Remote sensing and GIS are complementary technologies
that, when combined, enable improved monitoring, map-
ping, and management of forest resources (Franklin 2001).
The information that supports forest management is stored
primarily in the form of forest inventory databases within a
tion, composition, and distribution of forest resources. As
one of the principal sources of forest management informa-
tion, these databases support a wide range of management
decisions from harvest plans to the development of long-
term strategies.
Historically, forest management inventories were primarily
for timber management and focused on capturing area and
volume by species. In the past decade, forest management
responsibilities have broadened. As a result, inventory data
requirements have expanded to include measures of non-
harvest related characteristics such as forest structure, wild-
life habitat, biodiversity, and forest hydrology.
The entire forest inventory production cycle, from plan-
ning to map generation, can take several years. Except for the
photo interpretation component, forest inventory produc-
tion is largely a digital process.
Operational level
inventories,
Operational level inventories, Operational level
based on both aerial photo interpretation and fi eld-sam-
pled measurements, provide location-specifi c information
required for harvest planning. Forest
management level
inven-
management level inven-management level
tories meet longer-term forest management planning objec-
tives. Though these levels differ in detail, they both require
information fundamentally based on forest inventory data.
A forest management inventory generalizes complex for-
est resource attributes into mapping units useful for forest
management. The types of attributes attached to individual
mapping units, or polygons, might include stand species com-
position, density, height, age, and, more recently, new attri-
butes such as leaf area index (Waring and Running 1998).
Much of the information collected for forest inventory is
generated by interpretation of aerial photographs at photo
scales of 1:10,000 to 1:20,000, depending on the level of
detail required. Other remote sensing sources such as air-
borne and satellite digital imagery have been valuable in
updating forest attributes such as disturbance, habitat,
and biodiversity. In providing more frequent information
updates, remotely sensed data can improve the quality of
forest inventory databases, thereby improving the resource
management activities they support.
The quality of photointerpreted data depends on the expe-
rience of the interpreters and the use of quality assurance pro-
cedures such as interpreter calibration and fi eld verifi cation.
Other factors can introduce inconsistencies that compromise
the quality of forest inventory data. For example, there may
be source data inconsistencies when aerial photography is
acquired on different dates or in different weather conditions
or inconsistencies in analysis when multiple contractors are
used. The quality of the resulting data may vary signifi cantly
within a map area. For example, information about distur-
bances related to fi re and insects may be inconsistent within
a map area because the aerial photography from which it was
interpreted was acquired in different years. Similarly, incon-
sistencies may occur at the edge of neighboring map sheets
because data was collected in different years or was produced
by different contractors.
Applications of remote sensing and GIS to forestry
The use of remote sensing by forest managers has steadily
increased, promoted in large part by better integration
of imagery with GIS technology and databases, as well
as implementations of the technology that better suit the
information needs of forest managers (Wulder and Frank-
lin 2003). The most important forest information obtained
from remotely sensed data can be broadly classifi ed in the
following categories:
• detailed forest inventory data (e.g., within-stand
attributes)
• broad area monitoring of forest health and natural
disturbances
• assessment of forest structure in support of sustainable
forest management
Detailed forest inventory data
Forest inventory databases are based primarily on stand
boundaries derived from the manual interpretation of aerial
photographs. Stand boundaries are vector-based depictions
of homogeneous units of forest characteristics. These stand
polygons are described by a set of attributes that typically
includes species composition, stand height, stand age, and
crown closure. Digital remotely sensed data can be used to
update the inventory database with change (e.g., harvest)
information for quality control, audit, and bias detection. It
can also add additional attribute information and identify
Remote sensing for GIS managers
352
biases in the forest inventory databases due to vintage, map
sheet boundaries, or interpreter preferences.
The objective of managing forests sustainably for multiple
timber and nontimber values has required the collection of
more detailed tree and stand data, as well as additional data
such as gap size and distribution. Detailed within-stand for-
est inventory information can be obtained from high-spa-
tial-resolution remote sensing data such as large-scale aerial
photography and airborne digital imagery. Two methods of
obtaining this information are
polygon decomposition
(Wul-
der and Franklin 2001) and
individual tree crown recognition
(Hill and Leckie 1999).
Polygon decomposition analyzes the multiple pixels rep-
resenting a forest polygon on a remotely sensed image to
generate new information that is then added to the forest
inventory database (see Wulder and Franklin 2001). For
example, a change detection analysis of multidate Landsat
Thematic Mapper satellite images can identify the areal extent
and proportion of pixels where conditions have changed.
Individual tree crown recognition is based on analyz-
ing high-spatial-resolution images from which characteris-
tics such as crown area, stand density, and volume may be
derived (Hill and Leckie 1999).
Forest health and natural disturbances
Fire, insects, and disease are among the major natural distur-
bances that alter forested landscapes. Timely update informa-
tion ensures inventory databases are current enough to support
forest management planning and monitoring objectives.
Insect disturbance
Among the insects that cause the most damage to trees are
defoliators and bark beetles (Armstrong and Ives 1995).
Damage assessment for these insects is typically a two-step
process that entails mapping the disturbed area followed by
a quantitative assessment of the damage to the trees within
the mapped areas.
Aerial sketch-mapping, where human observers manually
annotate maps or aerial photographs, has been the most fre-
quently used method for mapping areas damaged by insects
(Ciesla 2000). This process is costly, subjective, and spatially
imprecise. However, when augmented by ground survey
methods and the integrated analysis of remote sensing and
GIS, substantial benefi ts can be realized.
Insect damage causes changes in the morphological and
physiological characteristics of trees, which affects their
appearance on remote sensing imagery. Insect defoliation
causes loss of foliage that results in predictable color altera-
tions. For example, residual foliage after attack by spruce bud-
worm will turn the tree a reddish color
(figure x13_Wulder1)
.
The mountain pine beetle is a bark beetle that bores through
the bark and creates a network of galleries that girdle the tree
and cause the foliage to become a reddish-brown color. These
foliage loss and color changes often occur during a short time
period-this is the optimal time for detection by remote sens-
ing. Knowing the characteristics of a particular damage agent,
the most appropriate sensor characteristics and acquisition
times can be selected (see example by Hall below).
Integrated remote sensing and GIS analyses that support
insect damage monitoring and mitigation include:
• detecting and mapping insect outbreak and damage
areas
• characterizing patterns of disturbance relative to
mapped stand attributes
• modeling and predicting outbreak patterns through
risk and hazard rating systems
• providing data to GIS-based pest management decision
support systems
Fire
Fire is an ecological process that governs the composition,
distribution, and successional dynamics of vegetation in the
landscape (Johnson 1992). Knowledge of fi re disturbance is
necessary to do the following:
Figure x13_Wulder1
Landsat satellite classifi cation for spruce
budworm defoliation with fi eld photograph depicting red-colored trees
damaged by spruce budworm defoliation. (Location: Junction of Troy
Lockhart Kledo Creek and Alaska Highway, Fort Nelson, B.C.).
Source: ©Her Majesty the Queen in the right of Canada, Natural Resources Canada.
Chapter 12
Remote sensing applications
353
• understand fi re impacts on timber and nontimber
values
• defi ne salvage logging opportunities
• understand the effect of climate change and feedback
processes on forest fi re occurrence
• quantify the infl uence of fi re on regional, national, and
global carbon budgets (Kasischke and Stocks 2000).
To address this range of issues, foresters employ a multi-
tude of fi eld, global positioning system (GPS), and remote
sensing (airborne and satellite) methods and data sources.
Integrated remote sensing and GIS fi re support systems are
used in real-time, near real-time, and post-fi re applications.
For example, infrared and thermal-infrared cameras with
integrated GPS/INS (inertial navigation system) technolo-
gies can observe fi re hot-spots, active fi res, and fi re perim-
eters in real-time. Data on fi re location and size is sent from
the aircraft to fi eld-based systems from which precise direc-
tions can be given to water-bombers and fi refi ghting crews.
Near real-time remote sensing and GIS systems are generally
based on daily observations from coarse-resolution satellites
such as the AVHRR (1 km pixel) and MODIS (250 m to
1 km pixel) satellites. Daily hot-spot information identifi es
the occurrence of fi re activity over large areas and helps to
target locations to collect more detailed information. Post-
fi re applications largely entail mapping the extent of burned
areas from aerial photographs or satellite imagery and assess-
ing fi re damage to vegetation.
The Canadian Wildland Fire Information System (CWFIS)
and the Fire Monitoring, Mapping, and Modeling System
(Fire M3) are integrated remote-sensing- and GIS-based
systems providing nationwide coverage to support fi re man-
agement and global change research. NOAA AVHRR and
SPOT VEGETATION remote sensing products can be used
to monitor actively burning large fi res in near real-time
(fig-
ure x13_Wulder 2)
to estimate burned areas and model fi re
behavior, biomass consumption, and carbon emissions (Fra-
ser et al. 2000, Lee et al. 2002).
The rapid fi re detection and response system implemented
by the U.S. Forest Service Remote Sensing Applications
Center, in cooperation with NASA and the University of
Maryland, uses MODIS satellite imagery to identify hot
spots throughout the United States. MODIS Active Fire
Map products are compiled daily at 3:00 A.M. and 3:00 P.M.
mountain time and are available over the Internet approxi-
mately two hours later. In addition to forest fi re detection,
the center provides image data from several different sensor
sources in support of fi re response and post-fi re assessment
activities (Quayle et al. 2002, Orlemann et al. 2002). An
example of MODIS data for forest fi re detection is shown in
fi gure x7-6 in chapter 7.
Landscape ecology, habitat, and biodiversity
Sustainable forest management requires that landscape eco-
logical characteristics related to habitat and biodiversity be
included in forest inventory and certifi cation procedures
(Vogt et al. 1999). The characteristics of interest are (1) spa-
tial patterns within the landscape, (2) specifi c habitat-related
forest conditions, and (3) the ecological processes that link
spatial pattern, habitat, and ecosystem functioning.
Land-cover information is one example of spatial patterns
readily obtainable by classifying remotely sensed data. Other
useful datasets include forest canopy information (e.g., crown
closure or leaf area estimates), understory information (Hall
et al. 2000), and measures of the distribution and boundar-
ies of landscape units such as forest fragmentation (Debin-
ski et al. 1999). Remote sensing can provide repeatable and
consistent methods to develop these data layers such that
changes over time can be monitored and habitat models can
be developed and validated for individual species.
Habitat assessment is typically GIS-based; it involves select-
ing data layers likely to be of value in developing predictive
models for the occurrence and distribution of individual spe-
cies or species assemblages, as well as the identifi cation of spe-
cies useful as indicators of ecological condition (see example
by Franklin below). The use of remotely sensed data together
with other spatial datasets integrated within a GIS environ-
ment has greatly enhanced the habitat assessment process.
Figure x13_Wulder2
Sample of Canada-wide burn area mapping
from Fire M3 depicting an area in the Northwest Territories.
Source: ©Natural Resources Canada.
Remote sensing for GIS managers
354
Improvements in forest management also depend on
increased understanding of ecological processes within the
carbon, nutrient, and hydrological cycles. Remotely sensed
data provides key inputs to models of carbon fl ux, nutrient
uptake and the infl uence of fertilization, and drought and
water stress indicators (Lucas and Curran 1999).
Future directions of remote sensing in forestry
A key development in remote sensing has been the increased
availability of high-spatial- and high-spectral-resolution
remotely sensed data from a wide range of sensors and plat-
forms including photographic and digital cameras, video
capture, and airborne and spaceborne multispectral sen-
sors. Hyperspectral imagery promises to provide improved
discrimination of forest cover and physiological attributes.
Radar applications are being developed that penetrate the
forest canopy to reveal characteristics of the forest fl oor (dis-
cussed in chapter 8). New technologies such as LIDAR can
provide estimates of forest biomass, height, and the vertical
distribution of forest structure with unprecedented accuracy
(Lim et al. 2003). The use of advanced digital analysis meth-
ods and selective use of complementary data have provided
more detailed information about forest structure, function,
and ecosystem processes than ever before (Culvenor 2003,
Hill and Leckie 1999).
As the availability of multiresolution remotely sensed imag-
ery and multisource data increases, so will the capability to
generate timely and accurate maps of forest composition and
structure. Operational capabilities continue to improve for-
est attribute mapping with a precision commensurate with
forest management scales. This, in turn, will contribute to
efforts aimed at assessing the sustainability of our forests
through better forest practices and improved decision-mak-
ing in forest management.
Acknowledgments
I would like to thank Mark Gillis of the Canadian Forest Ser-
vice in Victoria, B.C., for valuable comments and suggestions.
Chapter 12
Remote sensing applications
355
Armstrong, J. A., and W. G. H. Ives, eds. 1995. Forest insect pests in Canada. Natural Resources
Canada, Canadian Forest Service, Science and Sustainable Development Directorate, Ottawa, Ont.
Ciesla, W. M. 2000. Remote sensing in forest health protection. United States Department
of Agriculture, Forest Service, Forest Health Technology Enterprise Team, Remote Sensing
Applications Center, Salt Lake City, Utah. Report, No. 00–03.
Culvenor, D. 2003. Extracting individual tree information: A survey of techniques for high spatial
resolution imagery. In
Remote Sensing of Forest Environments: Concepts and Case Studies
, ed. M.
Wulder, and S. Franklin, 255–277. Dordrecht, Boston, London: Kluwer Academic Publishers.
Debinski, D. M., K. Kindscher, and M. E. Jakubauskas. 1999. A remote sensing and GIS based
model of habitat and biodiversity in the Greater Yellowstone Ecosystem.
International Journal of
Remote Sensing
20: 3281–3291.
Remote Sensing 20: 3281–3291.Remote Sensing
Franklin, S. E. 2001. Remote sensing for sustainable forest management. Boca Raton, Florida: CRC
Press.
Fraser, R. H., Z. Li, and J. Cihlar. 2000. Hotspot and NDVI differencing synergy (HANDS): A
new technique for burned area mapping over boreal forest.
Remote Sensing of Environment
74:
Remote Sensing of Environment 74: Remote Sensing of Environment
362–376.
Hall, R. J., D. R. Peddle, D. L. and Klita. 2000. Mapping conifer under story within boreal
mixed woods from Landsat TM satellite imagery and forest inventory information.
The Forestry
Chronicle
76: 887–902.
Chronicle 76: 887–902.Chronicle
Hill, D. A., and Leckie, D. G., eds. 1999. International forum: Automated interpretation of high
spatial resolution digital imagery for forestry. Proceedings of a workshop held February 10–12,
1998. Natural Resources Canada, Canadian Forest Service, Victoria, B.C.
Johnson, E. A. 1992.
Fire and vegetation dynamics: Studies from the North American Boreal forest.
Cambridge, United Kingdom: Cambridge University Press.
Kasischke, E. S., and B. J. Stocks, eds. 2000.
Fire, climate change, and carbon cycling in the boreal for-
est.
New York: Springer-Verlag.
Lee, B. S., M. E. Alexander, B. C. Hawkes, T. J. Lynham, B. J. Stocks, and P. Englefi eld. 2002.
Information systems in support of wildland fi re management decision making in Canada.
Computers and Electronics in Agriculture
37: 185–198.
Computers and Electronics in Agriculture 37: 185–198.Computers and Electronics in Agriculture
Lim, K., P. Treitz, M. Wulder, B. St-Onge, and M. Flood. 2003. LiDAR remote sensing of forest
structure.
Progress in Physical Geography
27(1): 88–106.
Progress in Physical Geography 27(1): 88–106.Progress in Physical Geography
Lucas, N. S., and P. J. Curran. 1999. Forest ecosystem simulation models: The role of remote sens-
ing.
Progress in Physical Geography
23: 391–423.
Progress in Physical Geography 23: 391–423.Progress in Physical Geography
Orlemann, A., M. Saurer, A. Parsons, and B. Jarvis. 2002. Rapid delivery of satellite imagery
for burned area emergency response (Baer). In
Proceedings of the 9th Biennial Forest Service
Remote Sensing Applications Conference
, ed. J. D. Greer. Bethesda, Md.: American Society of
Photogrammetry and Remote Sensing.
Parsons, A., and A. Orlemann. 2002. Mapping post-wildfi re burn severity using remote sensing and
GIS. Annette Parsons, Andrew In Proceedings 2002 ESRI User Conference. ESRI. Redlands,
Calif. gis.esri.com/library/userconf/proc02/pap0431/p0431.htm
References
Remote sensing for GIS managers
356
Quayle, B., M. Finco, K. Lannom, R. Sohlberg, J. Descloitres, M. Carroll, J. Norton, and D.
Kostyuchenko. 2002. USDA Forest Service MODIS active fi re mapping and delivery. Proceedings
of RS2002: Ninth Biennial Forest Service Remote Sensing Applications Conference, San Diego.
Vogt, K. A., B. C. Larson, J. C. Gordon, D. J. Vogy, and A. Fanzeres. 1999.
Forest certification: Roots,
issues, challenges, benefits.
Boca Raton, Florida: CRC Press.
Waring, R., and S. Running. 1998.
Forest ecosystems: Analysis at multiple scales.
2nd ed. Academic
Press.
Wulder, M., and S. E. Franklin. 2001. Polygon decomposition with remotely sensed data: Rationale,
methods, and applications.
Geomatica
55(1):11–21.
Geomatica 55(1):11–21.Geomatica
Wulder, M., and S. Franklin, eds. 2003.
Remote sensing of forest environments: Concepts and case stud-
ies.
Dordrecht, Boston, London: Kluwer Academic Publishers.
Web sites
Canadian Wildland Fire Information System
cwfi s.cfs.nrcan.gc.ca/en/index_e.php
U.S. Forest Service Remote Sensing Applications Center Rapid Response Web site
www.fs.fed.us/eng/rsac/fi re_maps.html