Tetsuhiko Yoshimura1 and Kouichi Kanzaki2
1 Assistant Professor, Forest Engineering, Faculty of Agriculture, Kyoto University, Japan.
2 Professor, Forest Engineering, Faculty of Agriculture, Kyoto University, Japan.
Abstract
In most mountains in Japan, it is very difficult to construct forest roads because the topography is very steep and slope failures often occur when constructing them. Therefore, it is very important to locate forest roads on stable slopes. In a former study, we made a quantitative risk assessment estimating the degrees of slope failure potentials using topographic maps. The objective of this study is to support decision-making for the layout of forest roads in mountainous areas based on the risk assessment. In particular, this study focused on decision-making based on the selection of passing points of forest roads using the fuzzy theory. The advantage of this method is that a computer can make the decision on where to lay out forest roads automatically as if it were a human being. It makes it much easier to plan forest roads in mountainous areas avoiding the danger that the areas considered collapse.
List of technical terms:
1. Risk assessment - It is very important to make a rational decision. By this method, the risk must be properly estimated in advance. In this study, the risk caused by constructing forest roads is employed for planning a forest road network.2. Slope failure potentials: The possibility of landslide occurrence.
3. Fuzzy expert system: This is the expert system using the fuzzy theory. The advantage of this system is that a computer can make a decision as if it were a specialist.
Introduction
In Japan, forest road networks are the essential basis of modern and sustainable forest production. However, one of the major problems that is that slope failures often occur when constructing forest roads especially in steep mountainous areas. Until now, little attention has been paid to the risk of slope failures when planning forest roads in those areas. We believe that it is most important to lay them out on stable slopes where slope failures rarely occur.
In a former study (Yoshimura et al. 1995), we made a quantitative risk assessment estimating the degrees of slope failure potentials using topographic maps. In addition, the degrees of slope failure potentials were calculated automatically using DEM (Yoshimura et al. 1996). It could be applied to GIS, which made it much easier to plan forest roads in steep mountainous areas.
Until now, planning of forest road networks has already been discussed many times, for example, Kanzaki (1966), Kobayashi (1975, 1980), Kitagawa (1972) and Sakai (1995). However, there is no precedent for a routing system based on the risk assessment as developed in this study which focuses on decision making based on the selection of passing points of forest roads using the fuzzy theory and the graph theory. It makes it possible for a computer to lay them out automatically as if it were a human being.
Method
1. Slope failure potentials calculated by using DEM
The factors used to estimate the degrees of slope failure potentials were inclinations, cross-sections of slopes, turning points of inclinations and catchment areas. When they were applied to GIS, it was necessary to obtain them automatically using DEM (Yoshimura et al. 1996, Takeuchi et al. 1992, Watanabe et al. 1993). Then, we calculated the slope failure potentials according to the method reported in a former study (Yoshimura et al. 1995). In this study, it was applied to Higashiyama mountains in Kyoto (Figure 1). The intervals of meshes were 20 m.
Figure 1. Slope failure potentials calculated automatically
2. Method of laying out a forest road
When a forest road network was planned in steep mountainous areas, we suggested that a forest road network consisting of vertical main roads and horizontal branch roads like a fish bone, was more suitable (Yoshimura et al. 1993, Ohashi 1992, Ohashi et al. 1989). In the present study, we developed a system for laying out a forest road automatically. This system was found suitable especially for a vertical main road. The methodology was as follows:
1. The meshes, where the degrees of slope failure potentials were below a constant value, were selected as possible passing points. In this study, that value was set at 0.0.2. The starting point and the ending point were decided at will.
3. A computer decided the routes by connecting some of those selected meshes. They often became the points where the directions of the routes were changed and were also suitable for the landings. This was a very important point.
4. When two passing points were connected, the methodology of selecting a next point was based on the fuzzy theory. Suppose you select point P' as the next point from P. The point which was evaluated the highest would be selected as P'. In the evaluation, the membership functions shown in Figure 2(a) ~(d) were used and then P' was evaluated by the following equations (1) ~ (3) (Tanaka, 1991):
IF X1 is A1i and ··· xn is AniTHEN yi = coi + c1ix1 + ··· + cnixn (i = 1, 2, ···, n) (1)
where i is a number of a rule, l is the total number of rules, AKi (k = 1, 2,..., n) is the fuzzy set, xk is a variable of input, yi is output from a rule i and cki is a parameter of a rule i. Then, the estimated values (y) are obtained as follows:
where
In this equation, m Aki (xk) is membership values of the fuzzy set Aki at xk. In this study, the equation (1) is expressed as follows:
IF x1 is "Big" THEN y1 = -50x1 (Very Small) (4)
IF x2 is "Big" THEN y2 = -100x2 (Very Small) (5)
IF x1 is "Medium or Small" and x2 is "Medium or Small" and x3 is "Big or Medium" and x4 is "Big or Medium"
THEN y3 = 0.3x1 - 10x2 + 0.5x3 + 3x4 (Big) (6)
IF x1 is "Medium or Small" and x2 is "Medium or Small" and x3 is "Big or Medium" and x4 is "Small"
THEN y4 = 0.2x1 - 15x2+ 0.3x3 + 2x4 (Medium) (7)
IF x1 is "Medium or Small" and x2 is "Medium or Small" and x3 is "Small" and x4 is "Big or Medium"
THEN y5 = 0.1x1 - 20x2 + 0.1x3 + x4 (Small) (8)
IF x1 is "Medium or Small" and x2 is "Medium or Small" and x3 is "Small" and x4 is "Small"
THEN y6 = -5x1 - 25x2 - 0.5x3 - 10x4 (Very Small) (9)
where:
x1 is the distance between P and P',
x2 is the gradient between P and P'.
The horizontal distance towards the point of the end (x3) and the vertical distance towards the point of the end (x4) are defined as:
x3 = d1 - d2 (10)
x4 = |d3| - |d4| (11)
where d1 and d2 are the horizontal distance between P and E (the point of the end) and between P' and E, respectively. Moreover, d3 and d4 are (E - P) and (E - P') in vertical levels, respectively. Here, the level of a mesh is defined as the average level of four corners.
Figure 2(a). Membership function. Distance between P and P'
Figure 2(b). Membership function. Gradient of a forest road
Figure 2(c). Membership function. Horizontal distance of the approach
Figure 2(d). Membership function. Vertical distance of the approach
5) We connected P to P' applying the Dijkstra method in the graph theory. It was already applied to the layout of a forest road by Kobayashi (1975) and was a very useful way to decide routes of forest roads. In this method, the value of each section between every two adjacent meshes was evaluated according to the following equation (Sakai et al. 1994):
E = W x F (12)
where E is the evaluated value, F is the average degree of slope failure potentials of two adjacent meshes, W is a coefficient which is defined as:
W = 1 (g < 8) (13)W = (g/8)2 (8 £ g < 24) (14)
W=¥ (g>=24) (15)
where g is the gradient between two adjacent meshes. We limited the maximum gradient of a forest road to 24 percent according to the results of the former experiment (Deki et al. 1990), in which it was concluded that the maximum gradient of a forest road for 4WD cars was 30 percent and a margin of about 5 percent was allowed for the cars' capacity to climb a hill, the conditions of the road surface and the psychological load of a driver.
Results
Figure 3 shows the forest road laid out by the above-mentioned expert system using the fuzzy theory. In this figure, the seven points A ~ G indicate the passing points determined using the fuzzy theory. In this figure, 'A' is the starting point and 'G' is the ending point, both of which were decided at will. The expert system selected the safety areas as the passing points of forest roads, which were approaching the point of the end. These passing points were also suitable for landings or forks of forest roads. In this process, the gradient of a forest road was also important.
After that, they were connected by using the Dijkstra method (Figure 4). As a result, the route decided by the expert system was rather good and seemed to be of practical use.
Conclusions
We have developed the expert system laying out a forest road automatically based on the risk assessment. In this way, the risk of slope failure occurrence by constructing forest roads would be minimized. In addition, if there was precious vegetation which should not be crossed, the degrees of slope failure potentials could be changed to make sure that the forest roads would avoid those areas. It can also be expected that the above-mentioned way of planning a forest road would contribute greatly not only to the prevention of landslides and soil disturbance but would also reduce construction and maintenance costs. In another paper, we will show the method of optimizing a forest road network including branch roads.
Acknowledgements
This study was conducted to realize the ideas that Mr Keizaburo Ohashi, a leading forest owner in the Osaka Prefecture, had about forestry management. We also thank Dr Kunihiko Numata, Dr Tetsurou Sakai and Mr Hajime Yamasaki (all of Kyoto University) for their assistance and advice.
Figure 3. Passing points determined using the graph theory
Figure 4. A forest road laid out by using the fuzzy theory
References
Deki, T., Furutani, S., Tanaka, Y., Suzuki, Y.& Kanzaki, K 1990. On the maximum gradient of forest road for 4WD car. J. Jpn. For. Soc. 72: 62-66.
Kanzaki, K. & Sakai, T. 1994. Method of planning a forest road caring for the prevention of slope failures. 2-45, Shinkou Kousaikai, Tokyo.1
Kanzaki, K. 1966. A planning of the forest road network by the theory of graphs. J. Jpn. For. Soc. 48: 365-371.
Kitagawa, K. 1972. Study on the automatic system of laying out forest roads (I) - A case study when longitudinal slopes are constant. Trans. 83rd Meet. Jpn. For. Soc.: 420-425.1
Kobayashi, H. 1975. Study on the automatic system of forest-road planning (III) - Layout of a forest road using Dijkstra method. Trans. 86th Meet. Jpn. For. Soc.: 463-464.1
Kobayashi, H. 1980. Studies on planning methods of forest road network - Forest road planning methods for maximum economic effect - Trans. 91st Meet. Jpn. For. Soc.: 459-460.
Ohashi, K. & Kanzaki, K. 1989. Manual on planning a forest road network in mountainous areas. Zenrinkyo, Tokyo. 173 pp.1
Ohashi, K. 1992. Management of thinning forestry based on a forest-road network. 187 pp. Zenrinkyo, Tokyo.1
Sakai, T. 1995. Two algorithms for forest road planning using digital terrain map (DTM). J. Jpn. For. Eng. Assoc. 10(3): 225-231.
Takeuchi, M., Mitsueda, K., Hasegawa, T., Sakai, S., Fujimoto, J. & Matsuba, K. 1992. Research on man-made forest in Kyoto University forest in Wakayama (1). Rep. Kyoto Univ. For. 23: 81-89.1
Tanaka, K. 1991. Introduction to the application of the fuzzy theory. Rassel sha, Tokyo. 206 pp.1
Watanabe, H. 1993. New air survey textbook. Japan Forest Technical Association, Tokyo. 264 pp.1
Yoshimura, T. & Kanzaki, K. 1993. On the selection method of a road network plan for steep terrain by AHP. Bull. Kyoto Univ. For. 65: 266-276.
Yoshimura, T. & Kanzaki, K. 1995. Method of planning a forest road network using the degrees of slope failure potentials (I) Estimation of the degrees of slope failure potentials with topographic maps. J. Jpn. For. Soc. 77:1-8.
Yoshimura, T. & Kanzaki, K. 1996. Method of planning a forest road network using the degrees of slope failure potentials.
1 The titles are tentative translations from Japanese by the authors of this paper.
