Content uploaded by İsmail Rakıp Karaş
Author content
All content in this area was uploaded by İsmail Rakıp Karaş
Content may be subject to copyright.
USING HIGH-RESOLUTION DIGITAL ELEVATION MODEL
FOR COMPUTER-AIDED FOREST ROAD DESIGN
A. E. Akay
a,
*, I. R. Karas
b
, J. Sessions
c
, A. Yuksel
a
, N. Bozali
a
, R. Gundogan
d
a
Kahramanmaras Sutcu Imam University (KSU), Faculty of Forestry, Department of Forest Engineering, 46060
Kahramanmaras, Turkey - (akay, ayuksel, nbozali)@ksu.edu.tr
b
Gebze Institute of Technology (GIT) Department of Geodetic and Photogrammetric Engineering, 41400 Gebze,
Kocaeli, Turkey - ragib@penta.gyte.edu.tr
c
Oregon State University, College of Forestry, Department of Forest Engineering, Corvallis, OR 97331, USA -
john.sessions@cof.orst.edu
d
Kahramanmaras Sutcu Imam University (KSU), Faculty of Agriculture, Department of Soil Science, 46060
Kahramanmaras, Turkey - rgundogan@ksu.edu.tr
KEYWORDS: High resolution, DEM, Real-time, LIDAR, Three-dimensional, Forestry, Design, Cost
ABSTRACT:
The current forest road design systems are not capable of making computer-aided design judgments such as automated generation of
alternative grade lines, best fitting vertical alignment for minimizing the total road costs, or consider environmental impact. A three
dimensional (3D) forest road design model was developed as a decision support tool that provides a designer with a quick evaluation
of alternative road paths. In the model, initial trial routes are generated by “tracing” the possible paths using computer cursor on a
3D image of the terrain. The model integrates two optimization techniques (linear programming and a heuristic approach) for
selection of a vertical alignment with the lowest total costs, while conforming to environmental requirements. Improved 3D OpenGL
accelerator is used to display and render 3D images of the terrain in real-time, based on high-resolution Digital Elevation Model
(DEM) data. LIDAR (Light Detection and Ranging), one of the fastest growing systems in the field, can provide a high-resolution
and accurate DEM of forested areas. It is expected to provide even better accuracy in the near future. The development of the road
design model incorporating modern graphics capability, advanced remote sensing technologies, improved software languages,
modern optimization techniques, and environmental considerations will improve the design process for forest roads.
AUSZUG:
Gegenwärtige Waldweg-Konstruktionssysteme eignen sich nicht zur computergestützten Designkontrolle, wie z.B. automatisiertes
Erzeugung alternativer Gradlinien, optimaler vertikaler Ausrichtung zur Minderung der Gesamtkosten von Straßen oder
Berücksichtigung von Umwelteinflüssen. Ein dreidimensionales (3D) Waldweg-Designmodell wurde entwickelt als ein Werkzeug,
das den Entwickler durch schnelle Auswertung alternativer Straßenverläufe bei seinen Planungsentscheidungen unterstützt. Im
Modell werden Ausgangsprobewege erzeugt, indem man mögliche Wege mit dem Computer-Cursor auf einem 3D-Bild des
Geländes "verfolgt". Das Modell integriert zwei Optimierungstechniken (lineare Programmierung und heuristische Annäherung) zur
Auswahl einer vertikalen Ausrichtung mit den niedrigsten Gesamtkosten in Übereinstimmung mit den Umgebungsbedingungen. Ein
verbesserter 3D-OpenGL-Beschleuniger wird benutzt, um die 3D-Bilder des Geländes in Realzeit anzuzeigen und zu übertragen auf
Basis hochauflösender Daten des Digital-Aufzug-Modells (DEM). LIDAR (Light Detection and Ranging), eines der sich am
schnellsten entwickelnden Systeme auf diesem Gebiet, kann ein hochauflösendes und genaues DEM der bewaldeten Bereiche zur
Verfügung stellen. Es wird erwartet, noch größere Genauigkeit in naher Zukunft zu erreichen. Die Entwicklung des
Straßendesignsystems, das Leistungsfähigkeit moderner Graphikprogramme mit fortgeschrittenen Fernabfragetechnologien,
verbesserten Software-Sprachen, modernen Optimierungstechniken und Berücksichtigung von Umweltbedingungen miteinander
verbindet, wird eine Verbesserung des Designprozesses für Waldwege darstellen.
* Corresponding author.
1. INTRODUCTION
Forest road design is a complicated problem relating economic
and environmental considerations. Designers should evaluate
sufficient number of alternative routes to locate a final route
with the lowest total cost, while conforming to design
specifications and environmental requirements. The current
forest road design systems are not developed to provide a
designer with large number of alternative paths. They are
generally used to make the mathematical calculations required
in manual road design. Besides, they are not capable of
minimizing total cost of construction, maintenance, and
transportation costs, or aiming for least environmental impacts.
With the growing availability of high-resolution DEM data and
advanced features of high performance microcomputers,
designing alternative road paths in the office has become a more
realistic and practical exercise. It is difficult to generate DEM of
a forested area that represents the actual ground condition with
high accuracy. Airborne laser scanner technology generates a
high-resolution and accurate DEM of forested areas, using a
laser light detection and ranging (LIDAR) system (Reutebuch et
al., 2003). LIDAR has been used over twenty years in variety of
applications. Most recently, it has been used to provide DEMs
for large scale and high-accuracy mapping (Kraus and Pfeifer,
1998; Means et al., 2000).
LIDAR employs a powerful laser sensor positioned under an
aircraft. The laser sensor consists of a transmitter and receiver.
The distance from the sensor to points on the terrain below is
measured based on the location and altitude of the aircraft.
These two parameters are calculated using Global Positioning
System (GPS) and Inertial Navigation System (INS)
technologies. When a pulse hits the ground, the reflected light is
collected by a receiver. The first return may locate the top of a
tree, while the last return ideally locates the ground beneath the
tree canopy.
The spatial resolution of the DEM highly depends on the
ground cover. In a study conducted by Pereira and Janssen
(1999), it was found that the vertical accuracy of LIDAR in
open areas with hard surface is approximately 15 cm. However,
in areas with very tall and dense canopy, the photogrammetric
heights may have larger errors. Reutebuch et al. (2003)
indicated that over mature forested areas LIDAR provides
vertical accuracy of 25 cm, which is enough to generate an
accurate DEM. As one of the fastest growing remote sensing
technology, LIDAR is expected to provide even better accuracy
in the near future.
In order to assist designers in locating forest roads, various
computer-aided methods using DEMs have been introduced.
Reutebuch (1988) developed a computer program, ROUTES, to
estimate road gradient, length, and stationing along the possible
road alternatives using a DEM. In the program, the designer
could digitize the contours and use the digitizer puck to locate
the road on a large-scale contour map. Although the graphical
user interface (GUI) of ROUTES was primitive, the program
worked well and provided the designers with the ability of
quickly looking at alternative road locations at varying scales.
A computer program, PEGGER, was developed to automate
initial forest road design through the use of a Geographic
Information System (Rogers and Schiess, 2001). The
performance of the program relied on DEMs, which must
accurately represent the actual ground conditions. PEGGER
was a tool for quickly analyzing many road alternatives based
on a specified road gradient given by a user. It was not capable
of considering environmental and economic constraints such as
soil types, hydrology, property lines and slope classes.
However, forest road design using PEGGER with accurate
DEM data was expected to be more feasible and less time
consuming than the traditional road design methods.
Coulter et al. (2001) developed a method of forest road design
using high-resolution DEM data (1mx1m) from LIDAR. In the
method, road elevations were assigned to each pixel within the
road template to calculate earthwork from the difference
between road and surface elevations. This method was only
applicable to straight road segments and could not locate
horizontal or vertical curves. It also could not calculate total
road cost or consider environmental requirements. However, it
was probably the first method showing that using high-
resolution DEMs from LIDAR in forest road design may
significantly decrease design time and effort spent both in the
field and in the office.
A decision support system can be defined as an interactive
system that assists a user to conduct decision making tasks by
easily accessing decision models and data (Watson and Hill,
1983). Advances in the processing speed and real-time
rendering and viewing of three-dimensional (3D) graphics on
microcomputers permit locating a route interactively on a 3D
display of a ground surface generated by a high-resolution DEM
data. A 3-D forest road alignment optimization model,
TRACER, aided by an interactive computer system, was
developed as a decision support system. The model provides a
designer with a quick evaluation of alternative road paths to
locate the best path with minimum total road cost, while
considering design specifications and environmental
requirements. TRACER relies on a high-resolution DEM to
provide terrain data for supporting the analysis of road design
features such as ground slope, topographic aspect, and other
landform characteristics. In the model, two optimization
techniques are integrated: linear programming to minimize
earthwork allocation cost and Simulated Annealing to optimize
vertical alignment. In this paper, the features of the model were
described and an application was presented.
2. MATERIAL AND METHODS
2.1 Input Data
Input data include DEM and attribute data, road design
specifications, and environmental requirements. The DEM data
is a set of scattered metric data points (X and Y coordinates and
Z as elevation). In designing forest roads, the resolution of the
DEM should be in the range of 1.0 m to 3.0 m to represent the
actual terrain conditions (Akay, 2003). In this study, the DEM
data set of 2.0 m resolution is developed using LIDAR data set
collected from western Washington by Aerotec (1999). Bilinear
interpolarion method is used to extract ground elevations from
the DEM (Lemkow, 1977). First, the grid cell that contains the
horizontal position of the current point is determined based on
its coordinates, then elevation of this point is estimated by
interpolating between the elevations of four corners on this grid
cell. The attribute data, soil types and stream data, is
represented in the same format as the metric data points. It is
easy to incorporate other attribute data into the model if desired.
The road design specifications are geometric specifications,
local site specifications, and economic data. The geometric
specifications include road gradient, horizontal curves radius,
vertical curve length, distance between curves, and safe
stopping distance for driver safety. Road template
specifications, turnout dimensions, distance between road
sections, design speed, vehicle specifications, and traffic
volume are also included in the geometric specifications. Local
site specifications consist of swell and shrinkage factors of soils,
vegetation cover, geological data, stand data, and distance to
local resources of road construction materials. Economic data
include the unit costs for road construction, maintenance, and
transport activities. The environmental requirements considered
in this study are minimum allowable road grade for proper
drainage, minimum distance from the riparian zones and
minimum stream-crossing angle for stream protection, and
maximum height of cuts and fills for soil protection.
2.2 Displaying Terrain Image
The model uses graphics routines from NewCyber3D (2002) to
display high-resolution image of the terrain in 3D, based on
DEM data (Figure 1). The real-time 3D stereo display and
stereo image composition is also supported by NewCyber3D.
Above-below stereo display format is used to generate stereo
scenes, which requires liquid-crystal glasses and an infrared
emitter. The graphic programming runs in full-screen mode to
be compatible with this format. At the standard sixty fields per
Figure 1. Displaying 3D image of the terrain
second, scanning each image takes half the duration of an entire
field. Using a monitor operating at 120 fields per second, each
eye sees 60 fields of image per second, while the other 60 fields
are prepared for the other eye. Therefore, when the left eye can
see an image, the right eye cannot (Lipton and Meyer, 1984).
2.3 Interactive Features
The alternative forest road paths are selected through real time
interaction with 3D image of the terrain. The interactive features
of the model are provided by NewCyber3D (2002), using an
improved 3D OpenGL accelerator. The initial trial road path is
“traced” by locating a series of intersection points on the
terrain, using computer cursor (Figure 2). The model provides
the designer with the road geometry information and attribute
data in real time to locate control points with respect to road
design specifications and environmental requirements. If a
candidate intersection point is not acceptable by one or more
constraints, the model warns the designer by changing the
colour of the line between the previously selected intersection
point and the candidate intersection point. The designer can
zoom, pan, rotate, and scale the area in order to search
intersection points around the terrain. The model has various
interactive display features such as navigation control, bird-
view, and real-time flythrough.
Figure 2. Selecting intersection points of a trial road path
2.4 Calculating Horizontal and Initial Vertical Alignment
After locating the trial path, the model automatically calculates
the horizontal and initial vertical alignment considering road
design specifications. Road gradient is restricted by the
maximum allowable road grade considering the truck
performance. The gradient is also limited by the minimum
acceptable road grade to provide proper drainage.
In order to determine whether any type of curve is necessary,
the model calculates the difference between two consecutive
road grades (A) and horizontal deflection angle (∆) for each
intersection point along the roadway. In forest roads, it is not
necessary to locate a vertical curve if A is less than or equal to a
specified percentage of difference between grades (A
min
.), which
provides a log truck with a safe passage of the vertical curve. If
A is greater than A
min
and ∆ is zero, the model locates a vertical
curve. If A is less than or equal to A
min
and ∆ is greater than
zero, then the model locates a horizontal curve. Otherwise, a
straight segment (tangent) is located. If there is a case where A
is greater than A
min
and ∆ is greater than zero, the model warns
the designer to choose a different control point to avoid
overlapping of vertical and horizontal curves.
To ensure a safe roadway passage along the vertical curves, the
model is constrained to generate a minimum adequate curve
length, which allows a log truck to pass a curve without
bottoming out and provides safe stopping distance for driver
safety. To provide safe continuous operation along the
horizontal curve, the model is constrained minimum radius,
acceptable road grade on horizontal curve, and minimum safe
stopping distance.
2.5 Optimizing Vertical Alignment
After locating the horizontal alignment and initial vertical
alignment, the model computes the total cost of construction,
maintenance and transportation costs. The road construction
activities considered in this study are construction staking,
clearing and grubbing, earthwork allocation, drainage and
riprap, surfacing, water supply and watering, and seeding and
mulching. Road maintenance activities include replacing the
aggregate, grading, maintaining culverts, cleaning ditches, and
removing brush. Transportation cost varies with vehicle
performance, equipment costs, gradient, and curvature. The cost
of road design activities are estimated by multiplying their
average unit costs by the quantity of design parameters (e.g. m
3
,
m
2
, m).
The model searches for the optimum vertical alignment with
minimum total cost among the large number of alternative
alignments. Simulated Annealing (SA) algorithm, using a
neighbourhood search, is employed to guide the search for the
optimal vertical alignment. The model considers technically
feasible grades within the specified elevation ranges of the
intersection points. The SA algorithm was developed based on a
metallurgical technique of annealing, in which a solid material
is heated and cooled back slowly into an optimal state to
produce the best product (Beasley et al., 1993). In this study,
SA has been selected as an optimization technique because it
generally provides a near-optimal solution and it is easy to
implement into the model.
For each alternative vertical alignment, the model calculates
earthwork volumes using average end-area method, minimizes
Intersection
Points
Stream
Channel
Harvesting Unit
Boundary
Trial Road
Path
earthwork costs using linear programming (LP), and estimates
sediment production. The economic distribution of cut and fill
quantities is determined by implementing the LP method of
Mayer and Stark (1981). This method represents the earthwork
allocation better than the other methods due to considering
possible borrow and landfill locations and various soil
characteristics along the roadway. Besides, it provides the
optimal solution to the earthwork allocation problem. A linear
programming code, using the idea of simplex algorithm
(Bowman and Fetter, 1967), is developed to incorporate this
method into the model.
The average annual volume of sediment delivered to a stream
from the road segments is estimated by using the method of the
Geographic Information Systems (GIS) based erosion delivery
model, SEDMODL (Boise Cascade Corporation, 1999). Some
of the road erosion factors considered in this model include
geologic erosion rate, road surface type, traffic density, road
width and length, average road slope, average precipitation
factor, distance between road and stream, cut slope cover
density, and cut slope height. SEDMODL reasonably predicts
the sediment delivery and defines the road segments with high
sediment production.
To develop additional road alignment alternatives, various
feasible road paths can be traced out by the designer. For each
alternative, the model follows the same procedure to find the
optimal vertical alignment with minimum total cost. Therefore,
the designer can quickly generate many alignments and select
the optimal one among the alternatives in an efficient way.
3. RESULTS AND DISCUSSION
The model was applied to a study area of 55 hectares, located in
the Capitol State Forest, Washington, the USA. The high-
resolution DEM (1mx1m) of the study area was obtained from
LIDAR (Aerotec, 1999). The attribute data including soil,
hydrology, and geology data were provided by Washington
Department of Natural Resources. Road design specifications
(Table 1) and cost data in US dollars (Table 2) were obtained
from the local sources in Pacific North West (PNW) (Kramer,
200; USDA Forest Service, 1999).
Initial road path was generated by establishing six intersection
points on the 3D image of the terrain based on DEM data from
LIDAR, while considering road design specifications and
environmental requirements (Figure 2). The unit cost of this
initial path was $36.88/m. Then, using optimization techniques,
Table 1. The road specifications used in the example
Road Specifications Values
Road width
Cut slopes
Fill slopes
Minimum curve radius
Minimum length of a vertical curve
Minimum differences between grades
Minimum road grade
Maximum road grade
Minimum distance from road to streams
Maximum cut and fill height at centerline
Design speed
4 m
1:1
1.5:1
18 m
15 m
5 %
± 2 %
16 %
10 m
2 m
55 km/hr
Table 2. Local unit cost data for road construction in PNW
Cost Items Costs
Construction staking
Clearing and grubbing
Earthwork
Excavation
Haul
Embankment
Disposal
Borrow material
Excavation
Haul
Embankment
Surfacing
Base course
Traction surface
Drainage and riprap
Culvert
Riprap
Watering
Seeding and mulching
Material
Application
$ 778/km
$3700/ha
$1.6/m
3
$1.3/m
3
-km
$0.6/m
3
$0.1/m
3
$1.8/m
3
$1.3/m
3
-km
$0.6/m
3
$3.9/m
3
$11.8/m
3
$25 meter
$10/m
3
$3/kilo liter
$0.5/kg
$550 /ha
the model located the optimal vertical alignment with the
minimum unit cost of $27.74/m (Figure 3). Therefore, total road
cost was reduced about 25% by the optimization model. Total
amount of sediment delivered to streams from the road section
was estimated as 0.84 ton/km.
The results indicated that total construction cost was the largest
cost component, followed by maintenance and transportation
costs (Table 3). Within the construction cost components,
surfacing cost was the largest, followed by earthwork allocation
cost. Total cost of maintaining culverts and ditches and clearing
bushes was the largest cost component in the maintenance costs.
During the search process, the model calculated 147 feasible
solutions out of 1200 automatically generated vertical alignment
alternatives. The solution process took about 15 minutes. The
most of the computation time was spent on calculating
earthwork allocation using LP for each vertical alignment
alternative. The time spent on earthwork allocation increases as
the number intersection points along the roadway increase.
Table 3. Total cost summary table
Cost Components Costs ($)
Total Construction Cost
Earthwork Cost
Construction Staking Cost
Clearing Grabbing Cost
Drainage Cost
Seeding Mulching Cost
Surfacing Cost
Water Supply Watering Cost
Riprap Cost
Total Maintenance Cost
Rock Replacement Cost
Grading Cost
Culvert, Ditch, Brushing Costs
Total Transportation Cost
2109.09
177.29
737.05
406.85
369.99
3221.65
859.87
483.91
319.45
84.16
1060.62
1166.16
Figure 3. Optimal vertical alignment on 3D image of the terrain
Increasing the number of intersection points may also decrease
the driver safety and comfort due to frequent changes on the
roadway grade. However, using more intersection points would
reduce the earthwork volume that leads to significant reduction
in earthwork allocation cost and total construction cost since the
road profile becomes closer to the ground profile.
Profile and plan view of the road alignment was shown in
Figure 4 and Figure 5, respectively. The model required two
horizontal curves and one crest vertical curve along the
roadway. The radiuses of the horizontal curves (e.g. 62.9m and
50.9m) and the length of the vertical curve (66.23m) were
acceptable for safe traffic passage. The length of road section
was approximately 396 m with gradient of 2 to 12% (Figure 4).
4. CONCLUSIONS
Using high performance microcomputers, improved software
languages, advanced remote sensing technologies, modern
optimization techniques, and high-resolution DEM data has
significantly improved the designer efficiency in designing
preliminary forest roads in the office. In this study, a 3D forest
315
320
325
330
335
340
345
0 25 50 75 100 125 150 175 200 225 250 275 300 325 350 375 400
Road Length (m )
Elevation (m)
9%
12%
Vertical Curve
Length: 66.23 m
-2%
2%
5%
6%
Road
Profile
Ground
Profile
Figure 4. Profile view of the forest road
road alignment optimization model was developed as a decision
support system. Using this model, a designer can quickly
evaluate alternative paths and locate the best path with
minimum total road cost. In the model, an initial horizontal
alignment is located interactively on the 3D image of the terrain
generated based on a high-resolution DEM from LIDAR.
During the last couple of decades, LIDAR has been used in
variety of applications. Most recently, it has been used to
generate high-resolution DEMs of the forested areas with
sufficient accuracy. The model also relies on available GIS
layers of attribute data to represent topographic conditions.
Available GIS data collected from the forested areas currently
cannot represent the actual topographic condition with a high
accuracy; however, quality of GIS data is improving as remote
sensing technologies advance.
In the model, the optimal vertical alignment is determined
automatically, using the combination of two optimization
techniques: LP method for determining the economic
distribution of cut and fill quantities and SA algorithm for
locating the optimal vertical alignment. LP method guaranties
the global minimum cost for earthwork allocation problem,
while SA provides good/near-optimal solution for the optimal
vertical alignment selection problem.
The results from the brief example were instructive in
presenting how a decision support system equipped with
interactive features, advanced GIS and remote sensing
technologies, and environmental considerations can improve the
design process for forest roads. It provides a road designer with
a number of alternative alignments to evaluate quickly and
systematically. The model has several limitations for further
developments such as improving the graphic interface,
optimizing the horizontal alignment, and calculating earthwork
allocations where the unit costs vary with the quantity of the cut
and fill.
ACKNOWLEDGEMENTS
We would like to thank USDA Forest Services researcher, Steve
Reutebuch, who provided us with LIDAR data of the study
area.
500
525
550
575
600
625
650
675
700
215 240 265 290 315 340 365 390 415 440 465 490 515
X-Axis (m )
Y-Axis (m)
Curve
Radius: 62.9 m
Curve
Radius: 50.9 m
Beginning
Point
Ending
Point
Figure 5. Plan view of the forest road
Road
Sections
Optimal
Vertical
Alignment
REFERENCES
Aerotec, 1999. Airborne laser-scan and digital imagery. 560
Mitchell Field Road, Bessemer, AL 35022.
Akay, A.E., 2003. Minimizing Total Cost of Construction,
Maintenance, and Transportation Costs with Computer-Aided
Forest Road Design. Ph.D. Thesis, Oregon State University,
Corvallis, Oregon. 229 p.
Beasley, J., Dowsland, K., Glover, F., Manuel, L., Peterson, C.,
Reeves, C., and Soderberg, B., 1993. Modern heuristic
techniques for combinatorial problems. Halsted Press: an
Imprint John Wiley and Sons, Inc. New York. 320 p.
Boise Cascade Corporation, 1999. SEDMODL-Boise Cascade
road erosion delivery model. Technical documentation. 19 p.
Bowman, E.H., and Fetter, R.B., 1967. Analysis for production
and operations management. Irwin Series in Quantitative
Analysis for Business. Yale University. 870 p.
Coulter, E.D., Chung W., Akay, A., and Sessions J., 2001.
Forest road earthwork calculations for linear road segments
using a high resolution digital terrain model generated from
LIDAR data. Proc First Precision Forestry Symposium,
University of Washington, College of Forest Resources, Seattle,
Washington. pp.125-129.
Kramer, B.W., 2001. Forest road contracting, construction,
and maintenance for small forest woodland owners. Oregon
State University, Forest Research Laboratory, Research
Contribution 35. 79 p.
Kraus, K., and Pfeifer, N., 1998. Determination of terrain
models in wooded areas with airborne laser scanner data. ISPRS
Journal of Photogrammetry and Remote Sensing, Vol. 53, pp.
193–203.
Lemkow, D.Z., 1977. Development of a digital terrain simulator
for short-term forest resource planning. M.S. thesis, University
of British Columbia, Vancouver, B.C. 207 p.
Lipton, L. and Meyer, L., 1984. A flicker-free field-sequential
stereoscopic video system, SMPTE Journal. November. pp.
1047.
Mayer, R. and Stark, R. 1981. Earthmoving logistics. Journal of
Const. Div. 107(CO2):297-312.
Means, J.E., Acker, A.A., Fitt, B.J., Renslow, M., Emerson, L.,
and Hendrix, C., 2000. Predicting forest stand characteristics
with airborne scanning LIDAR. Photogrammetric Engineering
& Remote Sensing, Vol. 66, No.11, pp. 1367–1371.
NewCyber3D, 2002. GIS Solution Systems. 11 Temiessee St.,
Redlands, CA 92373.
Reutebuch, S.E., 1988. ROUTES: A Computer Program for
Preliminary Route Location. Pacific Northwest Research
Station: U.S. Department of Agriculture, Forest Service.
General Technical Report PNW-GTR-216. 18p.
Reutebuch, S.E., McGaughey, R.J., Andersen, H., and Carson,
W., 2003. Accuracy of a high-resolution digital terrain model
under a conifer forest canopy. [Abstract] Can. J. Remote
Sensing 29(5):527-535.
Rogers, L.R., and Schiess, P., 2001. Pegger & Roadview – A
new GIS tool to assist Engineers in operations planning.
Proceedings, International Mountain Logging and eleventh
Pacific Northwest Skyline Symposium, Dec. 2001. University
of Washington, Seattle, WA. pp. 177-183.
Pereira, L., and Janssen, L., 1999. Suitability of laser data for
DTM generation: a case study in the context of road planning
and design. ISPRS Journal of Photogrammetry and Remote
Sensing, Vol. 54, pp. 244–253.
USDA Forest Service, 1999. Cost estimate guide for road
construction Region 6. Cost Guide Zone 5, Davis-Bacon Area
5, Oregon. 98 p.
Watson, H. J., and Hill, M.M., 1983. Decision support systems
or what didn’t happen with MIS. Interface. 13(5):81-88.
