Estimating sediment reduction cost for low-volume forest roads using a Lidar-derived high-resolution DEM/Nesmenu mazinimo kainos skaiciavimas mazo intensyvumo misko keliuose, taikant aukstos rezoliucijos skaitmenini auksciu modeli (DEM) pagal lazerinio skenavimo technologija (Lidar).

Akay, Abdullah Emin ; Wing, Michael Gilbert ; Sessions, John 等

Zemes darbu izmaksu samazinasanas novertejums ar LiDAR tehnologijas palidzibu iegutajiem augstas izskirtspejas augstuma atzimju modeliem

Setete vahendamise maksumuse prognoos metsateede ehitamisel kasutades Lidar-andmetel pohinevat korgresolutsioonilist digitaalset korguste mudelit

1. Introduction

Designing an optimal forest road alignment involves economic and environmental considerations. Road construction and maintenance are generally the most expensive activities in the timber transportation process (Akay 2006) and typically surpass truck transportation costs. In addition, inadequately constructed and maintained forest roads have the potential to cause more environmental impacts than any other forest operation activity (Akay et al. 2008). Runoff from road construction activities removes forest vegetation from the road prism area, disturb the forest floor, and damage forest soil structure (Grace et al. 1998). Sediment delivered to streams from roads potentially leads to detrimental effects on water quality and aquatic life (Murphy, Wing 2005; Wing et al. 2000). Given these road construction considerations and increased public concerns about road effects on forest ecosystems, forest road managers have incentives to design economically viable and environmentally low-impact forest roads.

Because of the inherent trade-offs between economic and environmental considerations, selecting an optimal forest road alignment with the lowest total cost while protecting soil and water resources is a complex problem that is likely best addressed by computer-aided road design systems (Akay et al. 2005). Computer-aided road design systems generally employ mathematical optimization methods that allow users to examine a large number of feasible alternative alignments and then select an optimal solution from among the alternatives. There are a number of computer-aided forest road design systems that mainly search for the optimal road alignment with min road cost. There are, however, only a few studies that have sought an optimal road alignment that considers both economic and environmental constraints at the road project scale. Kirby and Rupe (1987) examined the cost of minimizing sediment considering road construction and harvest scheduling at the watershed scale using mixed integer programming, but did not consider trade-offs at the road project scale. Bettinger et al. (1998) developed a forest plan in which an algorithm determined the shortest travel path from each management unit to a mill. The road maintenance and obliteration choices were integrated with a harvest scheduling model and considered a sediment delivery constraint, however, the choices did not involve detailed road design.

Akay and Sessions (2005) developed a 3D forest road alignment optimization model (TRACER) that searches for the best vertical alignment that minimizes total road costs, while confirming sediment delivery to streams and driver safety. During the search process, vertical alignment alternatives were generated and evaluated using a hybrid simulated annealing/linear programming optimization technique. In a similar study conducted by Aruga et al. (2007), a forest road design model was developed to simultaneously optimize horizontal and vertical alignments of forest roads using a Tabu Search optimization technique. The model allowed users to select an optimal road alignment while constraining max allowable sediment yield. Application of this model led to reduced road construction costs and less sediment delivered to streams.

Forest road design models require a high resolution Digital Elevation Model (DEM) for accurate representation of the terrain. In recent years, Light Detection and Ranging (LiDAR) data based DEMs have been widely used in forest road design models (Akay, Sessions 2005; Aruga et al. 2005a, 2007). LiDAR technology, integrated with Global Positioning Systems (GPS), is a laser- based measurement system that calculates the three dimensional coordinates of objects based on laser pulse reflections (Akay et al. 2009). LiDAR sensors are mounted on aerial or terrestrial platforms. LiDAR operates by transmitting light pulses that travel until reaching an object and then are reflected back to the LiDAR sensor. The amount of time it takes the transmitted light pulse to return to the sensor is used to calculate a distance to the object. This distance is coupled with a GPS measurement to determine the three dimensional coordinates of the object. In forest areas, light pulses are reflected from different levels of vegetation canopy including top of vegetation surface (first return), intermediate surfaces (second and following returns), and the ground surface (last return) (Reutebuch et al. 2003). The first returns are used to generate a Digital Surface Model (DSM) of vegetation canopy (Takahashi et al. 2005) while the last returns provide high-resolution and accurate DEMs under forest canopy (Akay et al. 2009).

The primary objective in this study was to apply a 3D forest road optimization model to identify an optimal forest road alignment between two points given two application scenarios and based on a LiDAR-derived DEM. In the first scenario, a road alignment that minimizes the cost of road construction was developed. In the second, a road alignment that minimizes the delivery of sediment to surrounding streams was generated. In both scenarios, a 3D forest road alignment optimization model that was previously developed was employed (Akay, Sessions 2005). In order to estimate the cost of sediment reduction in forest road construction, the differences between the two scenario results regarding total cost and total sediment production were examined.

2. Material and methods

2.1. Study area and LiDAR data

The study area was selected from the McDonald-Dunn Research Forest of the College of Forestry at Oregon State University (OSU). The Research Forest is located about 15 min drive north of the OSU campus in Corvallis and consists of approx 4553 ha of predominantly forested land. Douglasfir (Pseudotsuga menziesii) and bigleaf maple (Acer macrophyllum) are the dominant trees with the presence of grand fir (Abies grandis), western hemlock (Tsuga heterophylla) and western redcedar (Thuja plicata) (Akay et al. 2012). The elevation in the three watersheds that span the forest ranges from 122 m to 664 m with an average ground slope of 26%. In the McDonald-Dunn Research Forest, forested lands are located in the upper elevations while there are small holdings of agricultural areas, rural residential areas and urban developments in the lower elevations.

LiDAR data from the McDonald-Dunn Research Forest was collected in April 2008 with a Leica ALS50 Phase II laser system (Akay et al. 2012). The data consisted of three datasets including raw point data (1st returns, last returns, and all returns), vector data (ESRI shape file format), and raster data (1 m ESRI GRIDS of bare earth (DEM) and highest hit (DSM), and 1/5 m GeoTIFF of intensity image). The data resolution (average number of pulses emitted by the laser system) was 10.0 points/[m.sup.2] and 1.1 points/[m.sup.2] for average first return and average last return (ground) densities, respectively. Based on ground-truth measurements using real-time kinematic GPS measurements, the vertical accuracy was 0.02 m at one-sigma absolute deviation.

[FIGURE 1 OMITTED]

A study area of 70 ha (1000x700 m) was selected from McDonald-Dunn Research Forest (Fig. 1). LiDAR data from within the study area were extracted using ArcGIS 9.2 software. Soil and stream data layers of the study area drawn from the McDonald-Dunn Research Forest database were converted into a raster format (1x1 m). The soil and stream layers were re-projected to a match the Universal Transverse Mercator projection used for the LiDAR data.

Following all data spatial data pre-processing, the LiDAR data (DEM), soil, and stream layers were converted into an ASCII format for input into the 3D forest road alignment optimization model.

2.2. Road alignment optimization model

The 3D forest road alignment optimization model implemented in this study, TRACER, is developed to assist road managers with rapid evaluation of alternatives for the most economical path selection problem (Akay, Sessions 2005). The model selects the best potential road location path that minimizes the sum of construction, maintenance, and transportation costs while satisfying design specifications. A modern optimization technique (simulated annealing) is implemented to search for the best path (Akay 2006). Simulated annealing guides the search for the best path using a neighbourhood search of incremental changes to the vertical and horizontal alignment. To minimize earthwork allocation costs for each alternative path, a sub-optimization problem using linear programming (Mayer, Stark 1981) is solved. The linear programming approach, rather than the conventional mass diagram, was chosen since it has the advantages of being able to consider various soil characteristics along the roadway and possible borrow and landfill locations. In order to read soil type data real-time, the model requires a soil type layer.

TRACER employs graphic routines (NewCyber3D, CA) to display high-resolution two and 3D images of the terrain in real-time, based on DEM data. For locating the initial path, intersection points are manually selected using a mouse interactively on the terrain image. After locating the initial path, the model automatically locates cross-sections, computes earthwork, and calculates the horizontal and vertical road alignment locations while considering road design specifications, environmental requirements, and driver safety.

Road design specifications within TRACER include geometric specifications (i.e. road gradient, curvature constraints, design speed, etc.), local site specifications (i.e. soil characteristics, stand data, etc.), and economic data (i.e. unit costs for road construction, maintenance, and transport activities). The environmental considerations that are addressed by TRACER include min allowable road grade for proper drainage, distance from streams, and max height of cuts and fills for soil protection. In addition, stopping sight distance on horizontal curves is applied within TRACER road design criteria in order to ensure driver safety. Additional detail and technical information regarding TRACER are also available in Akay and Sessions (2005) and Akay (2006).

2.3. Road sediment delivery prediction

For road sedimentation applications, TRACER implements the equations used in the GIS-based model (SED-MODL) that estimates the average annual volume of sediment delivered to a stream from road networks (Akay et al. 2008). SEDMODL estimates the sediment delivered to a stream from each road section using empirical relationships between road surfacing, road use, road template, road grade, vegetative cover, and delivery of eroded sediment to the stream channel. Total sediment delivered from each road segment (ton per year) is predicted from two potential road sediment sources: road tread and cut-slope.

Tread Sediment = [GE.sub.r] x [S.sub.f] x [T.sub.f] x [G.sub.f] x [P.sub.f] x [D.sub.f] x [L.sub.r] x RW, (1)

Cutslope Sediment = [GE.sub.r] x [CS.sub.f] x [h.sub.c] x [D.sub.f] x [L.sub.r], (2)

where [GE.sub.r]--geological erosion rate, kg/[m.sup.3]-yr; [S.sub.f]--surfacing factor; [T.sub.f]--traffic factor; [G.sub.f]--road grade factor; [P.sub.f]--precipitation factor; [D.sub.f]--delivery factor; [L.sub.r]--length of the road segment, m; RW--road width, m; [CS.sub.f]--cut-slope cover factor; [h.sub.c]--cut-slope height, m.

Factors used in the tread and cut-slope sediment formulas are obtained from look-up tables in the SEDMODL documentation, which are generated from previous research from regions within Idaho, Washington, and Oregon (Akay et al. 2008). The geological erosion rate is based on dominant lithology and age. The sediment model provides the user with the surfacing factors of various surface types such as gravel (0.2), pitrun (0.5), and native surface (1.0) (Akay, Sessions 2005). Traffic factors of various road classes are given based on the average measurements taken during road erosion inventory studies (Reid, Dunne 1984). Based on these studies, traffic factors of primary, secondary, and spur roads are suggested as 10, 2, and 1, respectively. The road slope factors are assigned to each road stage based on the road grade classes. For road stages with grade of less than 5%, 5% to 10%, and greater than 10%, the road grade factors are 0.2, 1.0, and 2.5, respectively (Reing et al. 1991).

The precipitation factor in SEDMODL is computed based on the average annual precipitation falling within a watershed basin (Reid, Dunne 1984). The sediment model computes the erosion delivery factor for each road stage based on the proximity of roads to streams. It is assumed that a road segment that delivers directly to streams results a delivery factor of 1.00 (i.e. at stream crossings). A road segment within 30 m and 60 m of a stream results a delivery factor of 0.35% and 0.10%, respectively (i.e. at roads parallel to streams). The road segments that are located further than 60 m do not deliver sediment to streams (i.e. sediment do not reach the stream). In order to compute the stream distance, the model database requires a spatial stream database. The cut-slope cover factor as a percent of vegetative or rock cover on cut-slopes is also included in the sediment prediction equation based on local conditions within the watershed. Road width, length of the road stage, and cut-slope height in the model are computed based on road template information (Akay 2006).

2.4. Road alignment application

TRACER was applied to a sample landscape area with the goal of locating a single-lane forest road connecting two points while taking into account road design constraints, environmental considerations, and transportation safety. A 3D image of the terrain was generated from the LiDAR-derived DEM. Primary road design constraints for input into the road design model were adopted based on standard forest road design practices in Pacific Northwest (Table 1) (Akay, Sessions 2005). The cost elements of road construction and maintenance were determined based on the "Cost Estimate Guide for Road Construction" prepared by the USDA Forest Service.

Within TRACER, once an initial path was manually generated on the 3D view of the terrain, the optimization algorithm searched for all the feasible vertical alignments near the initial path and selected the optimal alignment that minimized the total road cost. For each alternative road alignment, TRACER estimated the total sediment yield. The road alignment with the least amount of sediment delivered to streams was selected for comparison to the initial least cost road alignment.

3. Results and discussion

The soil and stream layers generated using ArcGIS 9.2 (Fig. 2). The study area consisted of silty clay (SC) loam, gravelly silty clay (GSC) loam, and very cobbly (VC) loam. The geologic age and lithology combination in the study area was Tertiary/Basalt based on the geologic maps. The LiDAR-derived in the study area ranged from 227 m to 594 m, with an average elevation of 391 m (Fig. 3). The average ground slope was found to be 43%.

The initial road alignment generated manually resulted in a total road cost of 13 815 EUR. Using TRACER to design the optimal road alignment with min cost reduced the total road cost to 12 687 EUR. Thus, the road alignment optimization model reduced the total road cost about 8.17%. This percent reduction in cost is similar to that determined by previous research. In a similar study where Genetic Algorithm (GA) and Tabu Search (TS) were applied to determine the optimum forest road alignment, Aruga et al. (2005b) found that GA and TS reduced the total road cost by 11.54% and 8.84%, respectively.

For the road alignment that minimized sediment delivery, the road length was found to be slightly longer (352 m) than that of the optimal road alignment (350 m), while the average road gradient reduced from 11.6% to 9.9%. The total road cost was computed as 14 955 EUR, which indicated that minimizing the sediment delivery increased the total road cost by 17.88%.

In both road alignments, it was required to locate the horizontal curves with the radius of 25.18 m. However, vertical curves were not required since the absolute value of the difference between grades was less than 5.0% along the roadway (Fig. 4). The average side slopes for both alignment scenarios were computed as approx 22.0%.

Detailed cost summary for the values of main cost components were calculated for both design scenarios (Table 2). The largest cost component for both road alignments was construction cost, followed by maintenance and transportation costs. For the optimal road alignment with min total cost, earthwork cost (39.24%) and surfacing costs (35.86%) were the largest components of total road construction cost.

[FIGURE 2 OMITTED]

[FIGURE 3 OMITTED]

[FIGURE 4 OMITTED]

[FIGURE 5 OMITTED]

The results indicated that the road alignment with min sediment yield increased the earthwork cost and surfacing cost by 17.52% and 37.02%, respectively, because of changes on road length and gradient. The road maintenance cost also increased in the min sediment delivery scenario by about 6.96% primarily due to the increased rock replacement cost. Finally, the proportion of transportation cost was 6.51% and 5.58% of the total costs in both road alignments, respectively.

Sediment delivery to streams is considered a critical indicator of environmental impact of forest road construction practices (Grace et al. 1998). The optimal road alignment which minimized costs delivered average annual sediment of 1.06 t (3.03 t/km) to the streams. In the SEDMODL, the total sediment delivered from each road segment originates from two sediment sources: tread and cut-slope activities (Akay et al. 2008). The amount of tread and cut-slope sediment in the optimum road alignment results was 0.97 t and 0.09 t, respectively.

For the road alignment scenario that minimized sediment delivery, the sediment delivered to the streams from the road section decreased to 0.77 t (2.19 t/km), a 27.36% decrease compared to the sediment delivered by the cost minimized alignment. According to SEDMODL formulas, a road alignment with a gradient greater than 10% produces 2.5 times more sediment than a road alignment with 5-10% road gradient. Thus, the optimum road alignment produced about 38% more sediment delivery than the sediment minimized road alignment, mainly due to steeper road gradient. The amount of tread and cut-slope sediment by the sediment minimized road alignment was 0.70 t and 0.07 t, respectively. The slightly longer road alignment, gentler road gradient, and additional earthwork of the sediment minimized road alignment reduced the sediment delivery to the streams by about 0.29 t, while the total road cost increased by 2269 EUR, over the cost-minimized road alignment (Fig. 5). Therefore, the unit cost of minimizing sediment delivery to the stream in this example is approx 7823 EUR per ton of sediment.

The road alignment optimization model searched for the optimum solution based on a high-resolution DEM that was derived from LiDAR measurements. The accuracy of the LiDAR data directly affects the performance of the optimization model employed by TRACER, especially in earthwork allocation process. Aruga et al. (2005a) reported that using a LiDAR-based high-resolution DEM provided more accurate results in earthwork computations than using a DEM generated by ground surveying equipment. In this study, a 1 m by 1 m resolution DEM of the study area was generated based on LiDAR dataset with the vertical accuracy of 0.02 m (Akay et al. 2012).

4. Conclusions

The TRACER forest road alignment optimization model, previously developed to assist road managers in designing a preliminary road alignment, was implemented to estimate sediment volume and the cost of sediment reduction in forest road construction activities. The optimization model was applied to consider two road alignment scenarios. The model initially searched for feasible alignment alternatives with the search tolerances and selected the optimal alignment that minimized the total road cost, while constraining road design parameters, environmental considerations, and stopping sight distance. Then TRACER was used to identify the road alignment that delivered the least sediment yield to streams, while considering the same initial constraints. A GIS-based sediment prediction model was integrated into TRACER to estimate the sediment delivered to the streams from the road section.

1. The results from the model application indicated that the total cost of road construction and maintenance activities in the alignment generated by the optimization model that minimized costs was 1128 EUR less than the initial alignment that was manually digitized within TRACER.

2. In the second part of the application, the sediment delivery decreased by 27.36%, while the total cost was 2269 EUR more than that of the optimum min cost alignment.

3. The cost of reducing annual sediment delivery from forest road to the streams is approx 7823 EUR.

4. It was not aimed to generalize the results from this application to other areas given the unique environmental characteristics associated with specified study area. However, it was found that the forest road alignment optimization model was an effective decision support tool to investigate the trade-offs between economic and environmental considerations in locating forest road alignments.

5. Thus, this approach has great potential to provide benefits for road managers in identifying and evaluating potential road alignment options in areas within their own jurisdictions that are vulnerable to excessive sediment production.

Caption: Fig. 1. McDonald-Dunn Research Forest and location of study area

Caption: Fig. 2. The soil and stream layers in study area

Caption: Fig. 3. 3D view of the study topography within TRACER's interface

Caption: Fig. 4. Ground profile and road profiles for both design scenarios

Caption: Fig. 5. The relationship trend between total road cost and sedimentation delivery

doi:10.3846/bjrbe.2014.07

Received 12 December 2011; accepted 9 September 2012

Acknowledgements

This study, funded by The Scientific and Technological Research Council of Turkey (TUBITAK) under the International Post-Doctoral Research Fellowship Programme-2219, is taken place in Oregon State University, Corvallis, Oregon, USA.

References

Akay, A. E.; Wing, M.; Sessions, J. 2012. Estimating Structural Properties of Riparian Forests with Airborne Lidar Data, International Journal of Remote Sensing 33(22): 7010-7023. http://dx.doi.org/10.1080/01431161.2012.697206

Akay, A. E.; Oguz, H.; Karas, I. R.; Aruga, K. 2009. Using LiDAR Technology in Forestry Activities, Environmental Monitoring and Assessment 151(1): 117-125. http://dx.doi.org/10.1007/s10661-008-0254-1

Akay, A. E.; Erdas, O.; Reis, M.; Yuksel, A. 2008. Estimating Sediment Yield from a Forest Road Network by Using a Sediment Prediction Model and GIS Techniques, Building and Environment 43(5): 687-695. http://dx.doi.org/10.1016/j.buildenv.2007.01.047

Akay, A. E. 2006. Minimizing Total Costs of Forest Roads with Computer-Aided Design Model, Academy Proceedings in Engineering Sciences (SADHANA) 31(5): 621-633. http://dx.doi.org/10.1007/BF02715918

Akay, A. E.; Sessions, J. 2005. Applying the Decision Support System, TRACER, to Forest Road Design, Western Journal of Applied Forestry 20(3): 184-91.

Akay, A. E.; Boston, K.; Sessions, J. 2005. The Evolution of Computer-Aided Road Design Systems, International Journal of Forest Engineering 16(2): 73-79.

Aruga, K.; Chung, W.; Akay, A. E.; Sessions, J.; Miyata, E. 2007. Incorporating Soil Surface Erosion Prediction into Forest Road Alignment Optimization, International Journal of Forest Engineering 18(1): 24-32.

Aruga, K.; Sessions, J.; Akay, A. E. 2005a. Application of an Airborne Laser Scanner to Forest Road Design with Accurate Earthwork Volumes, Journal of Forest Research 10(2): 113123. http://dx.doi.org/10.1007/s10310-004-0116-9

Aruga, K.; Sessions, J.; Akay, A. E. 2005b. Heuristic Techniques Applied to Forest Road Profile, Journal of Forest Research 10(2): 83-92. http://dx.doi.org/10.1007/s10310-004-0100-4

Bettinger, P.; Sessions, J.; Johnson, K. N. 1998. Ensuring the Compatibility of Aquatic Habitat and Commodity Production Goals in Eastern Oregon with a Tabu Search Procedure, Forest Science 44(1): 96-112.

Grace, J. M. III; Rummer, B.; Stokes, B.J.; Wilhoit, J. 1998. Evaluation of Erosion Control Techniques on Forest Roads, Transactions of the American Society of Agricultural Engineers 41(2): 383-391. http://dx.doi.org/10.13031/2013.17188

Kirby, M. W.; Rupe, J. B. 1987. The Cost of Avoiding Sedimentation, Journal of Forestry 85(4): 39-40.

Mayer, R.; Stark, R. 1981. Earthmoving Logistics, Journal of the Construction Division 107(2): 297-312.

Murphy, G.; Wing, M. G. 2005. Road Sediment Yields from Dispersed Versus Clustered Forest Harvesting Activity: a Case Study, International Journal of Forest Engineering 16(2): 65-72.

Reutebuch, S. E.; Mcgauhey, R. J.; Andersen, H. E.; Carson, W. W. 2003. Accuracy of Highlight LIDAR Terrain Model under a Conifer Forest Canopy, Canadian Journal of Remote Sensing 29(5): 527-535. http://dx.doi.org/10.5589/m03-022

Reid, L. M.; Dunne, T. 1984. Sediment Production from Forest Road Surfaces, Water Resources Research 20(11): 1753-1761. http://dx.doi.org/10.1029/WR020i011p01753

Takahashi, T.; Yamamoto, K.; Senda, Y.; Tsuzuku, M. 2005. Estimating Individual Tree Heights of Sugi (Cryptomeria japonica D. Don) Plantations in Mountainous Areas Using Small-footprint Airborne LiDAR, Journal of Forest Research 10(2): 135-142. http://dx.doi.org/10.1007/s10310-004-0125-8

Wing, M. G. 2000. Landslide and Debris Flow Influences on Aquatic Habitat Conditions, in The 4th International Conference on Integrating GIS and Environmental Modeling (GIS/ EM4): Problems, Prospects and Research Needs. September 2-8, Alberta, Canada.

Abdullah Emin Akay (1) ([mail]), Michael Gilbert Wing (2), John Sessions (3)

(1) Dept of Forest Engineering, Kahramanmaras Sutcu Imam University, 46100 Kahramanmaras, Turkey

(2,3) Dept of Forest Engineering Resources and Management, College of Forestry, Oregon State University, Corvallis, Oregon, 97331 USA

E-mails: (1) akay@ksu.edu.tr; (2) michael.wing@oregonstate.edu; (3) john.sessions@oregonstate.edu

Table 1. Primary road design constraints for road design
application

Constraints                                Values

Min radius of horizontal curve             18 m
Min length of vertical curve               15 m
Min value of differences between grades    5%
Min gradient (for drainage)                [+ or -] 2%
Max gradient                               18%
Max cut and fill height                    2 m

Table 2. Main cost components for two road design scenarios

Cost components      Optimal road     Alignment with
                    alignment, EUR   min sediment, EUR

Construction

Earthwork                4170              4901

Construction
Staking                  104                104

Clearing and
Grubbing                 578                595

Drainage                 395                503
Seeding and

Mulching                 325                338

Surfacing                3812              5223

Water Supply and
Watering                 846                741

Riprap                   399                399

Maintenance

Rock Replacement         361                447

Blading                   54                54

Culvert, Ditch,
Brushing                 816                816

Transportation           826                835