A GIS-based multicriteria model for the evaluation of territorial accessibility

Transcription

1 Sustainable Development and Planning II, Vol A GIS-based multicriteria model for the evaluation of territorial accessibility J. F. G. Mendes, D. S. Rodrigues & R. A. R. Ramos Department of Civil Engineering, University of Minho, Portugal Abstract In most of the land-use planning and management processes location analysis is present and often plays a major role. Accessibility models were developed by many authors as a way to evaluate how easy or difficult it can be to link origins and destinations, adopting different formulations. In this paper, a multicriteria approach for the evaluation of accessibility is presented. The model was developed and implemented within a GIS context and applied to three different typologies of problems: general accessibility evaluation; accessibility evaluation in the context of industrial location; and accessibility evaluation of a University campus. Keywords: accessibility, multicriteria evaluation, GIS. 1 Introduction Land-use planning is a decision-making process that often involves selection, evaluation and combination of several factors. Some of these factors are closely related to the accessibility of the potential location alternatives, which stresses the interest of evaluating accessibility. The concept and evaluation of accessibility have been discussed for almost two hundred years. In one of the most interesting texts about accessibility, Hoggart [1] sustains that accessibility is associated with the interpretation, implicit or explicit, of the easiness of reaching spatially distributed opportunities. This means that accessibility depends not only on the location of opportunities but also on the easiness of overcoming the spatial separation between individuals and specific places. In the same line, Ingram [2] defines accessibility of a place as its characteristic (or advantage) regarding the overcoming of any form of resistance

2 796 Sustainable Development and Planning II, Vol. 2 to the movement. Ingram distinguishes between relative accessibility, which regards the degree of connection between two points on a surface (or network), and integral (or global) accessibility, which regards the degree of connection between a point and all the other points on a surface (or network). The way accessibility is evaluated depends on the purpose or objective to be achieved. Morris et al. [3] present an extensive classification and formulation of measures for relative and integral accessibility. This last class includes: measures of separation between all the points; measures of separation incorporating the effect of distance; measures of separation incorporating network capacity restrictions; and complex measures of separation and supply/demand. More recent contributions proposed accessibility measures that somehow can be framed in the classification of Morris et al. [4, 5, 6, 7]. 2 Multicriteria accessibility evaluation The model proposed in this research stands on a measure of separation incorporating the effect of distance. The principal theoretical points and assumptions in this model regarding envisioning accessibility include: i) Accessibility evaluation is related to a certain objective/purpose; in this case we are concerned with accessibility evaluation for industrial location purposes. ii) The accessibility index is a result of the combination of distances to a set of key-destinations, which can be particular points (e.g. facilities), lines (e.g. roads), or areas (e.g. industrial clusters); iii) Key-destinations are related to the objective/purpose and can have different priorities (weights); iv) Key-destinations can be reached through road and/or off-road travelling, each one with different resistance to movement (friction); v) Cost-distances to a key-destination are a result of the combination of actual distances with the friction surface; vi) Off-road friction is a function of the slope; vii) Cost-distances to key-destinations can be normalised through fuzzy set functions that, after weighting, represent their contribution to the accessibility index. Denoting the fuzzy set membership function applied to cost-distances by f(c ij ), and the weight of key-destination j by w j, the accessibility of a location A i is given by equation (1): A i = f ( cij ). wj (1) j Points i, for which accessibility is measured, depend on the way space is modelled. For a network, the node points are considered; for a continuous space (surface), grid points are considered. In particular, when a raster model of the space is used, considered points are pixels of the raster image and depend on the adopted grid resolution. Equation (1) is essentially a Weighted Linear Combination, one of the aggregation procedures available in the context of multicriteria evaluation [8].

3 Sustainable Development and Planning II, Vol A very important component of a multicriteria evaluation model concerns the priorities attached to the various criteria, i.e. the values of the weights w j in equation (1). The objective of developing weights is to quantify the relative importance of criteria to one another, in terms of their contribution to an overall accessibility index. Among many methods to derive weights established and used by different authors, two are most commonly used [9]: the n-points scale (originally seven-points scale, as introduced by Osgood et al. [10]); and a more complex method called Pairwise Comparisons, which was developed by Saaty [11] in the context of a decision making process known as Analytical Hierarchy Process (AHP). Because of different scales upon which criteria are measured, it is necessary to standardise them before aggregation. The process of standardisation is essentially identical to that of fuzzification in fuzzy sets [12]. The objective is to transform any scale to a comparable one measured according to a standardised range (e.g. 0-1). In our case, the result expresses a membership grade that ranges from 0.0 to 1.0, indicating a continuous variation from non-membership (no accessibility) to complete membership (maximum accessibility), on the basis of the criterion (distance) being fuzzified (Figure 1). 1 a 0 b Figure 1: Fuzzy set membership function. Depending on the nature of the criterion being fuzzified, different fuzzy functions can be selected. Among the most used are: Sigmoidal (S-shaped), J-shaped, Linear and Complex [9]. When fuzzifying distance variables, the sigmoidal monotonically decreasing function (Figure 1) is one of the most used, for which membership grade µ (i.e. standardised value) is given by: µ = cos 2 ( α) α = ( d da) /( db da) π / 2 (2) when d > db, µ = 0 ; d < da, µ = 1 Control points a and b (Figure 1) are critical points that should be set for each particular situation, considering their inherent meaning. 3 A GIS-based multicriteria accessibility model The formal model presented in the previous section can be implemented within a GIS environment, making use of map algebra functions.

4 798 Sustainable Development and Planning II, Vol. 2 The first step of the model is to create the cost-distance maps for each of the key-destinations. The flowchart of Figure 2 shows the sequence of operations required. On one hand, a slope map is derived from a digital elevation model (DEM); then, a reclassification operation produces the off-road friction map. On the other hand, the road friction map is derived directly from a road map. Overlay of these two friction maps result in the final friction surface, which is combined with each one of the key-destination maps to give the cost-distance maps. Digital Elevation Model calculate Slope Map Road Map reclassify reclassify Off-road Friction Map Road Friction Map overlay Key Destination Map Friction Surface calculate Cost-distance Map Figure 2: GIS model to create cost-distance maps. Having the cost-distance maps for each key-destination, the multicriteria procedure is implemented following the flowchart of Figure 3. The sequence of operations starts with the standardisation (i.e. the application of the selected fuzzy set functions) followed by the weighting. Afterwards, the weighted standardised cost-distance maps are overlaid (added) to give the final accessibility map. In order to use the accessibility evaluation model established in this paper, the model must be "customised" to the particular context under study. This means: (i) to identify the set of key-destinations; (ii) to establish the weights for each key-destination; (iii) to identify the fuzzy set functions to be used; (iv) and to set the control points a and b for the fuzzy set functions. Next sections show three applications of the model corresponding to three typologies of problems. 4 Application 1: industrial location in the northwest of Portugal In the context of an industrial location study for the Northwest of Portugal, a panel of 25 industrial entrepreneurs was interviewed about the relevant criteria

5 Sustainable Development and Planning II, Vol involved in their decision-making processes when locating new industrial units. Criteria were divided in three main classes: criteria associated with industrial activity; criteria associated with administrative and social-economical options; and criteria associated with physical planning. Among the factors of the first criteria class, the accessibility ones were considered the most important. Cost-distance Cost-distance Cost-distance Map for Map for Map for Key Dest. 1 Key Dest. 2 Key Dest. 3 standardise standardise standardise Standardised Standardised Standardised Cost-distance Cost-distance Cost-distance Map 1 Map 2 Map 3 * weight 1 * weight 2 * weight 3 Weighted Stand. Weighted Stand. Weighted Stand. Cost-distance Cost-distance Cost-distance Map 1 Map 2 Map 3 overlay + Accessibility Map Figure 3: GIS-based multicriteria accessibility model. Results of the interviews, regarding accessibility, are presented in Table 1, which answers to the four points stated earlier in the previous section. Weights were derived through a pairwise comparison procedure undertaken with each entrepreneur, followed by average and normalisation. The sigmoidal (monotonically decreasing) fuzzy function adopted was almost unanimous, and control point distances are an average of the entrepreneurs opinions. The accessibility evaluation model, specified this way, was then used in order to create an accessibility map (for industrial location purposes) of a Portuguese municipality named Valença, located in the Northwest of the country close to the Spanish border. Table 1: Industrial location: key-destinations, weights and fuzzy function data. Key-destination Weight Fuzzy function Dist. a (Km) Dist. b (Km) Main road 0,41 Sigmoidal 0,0 3,5 Motorway junction 0,24 Sigmoidal 0,0 27,4 Truck freight terminal 0,16 Sigmoidal 0,0 10,4 Railway freight terminal 0,11 Sigmoidal 0,0 20,1 Seaport freight terminal 0,06 Sigmoidal 0,0 51,9 Airport freight terminal 0,02 Sigmoidal 0,0 69,9

6 800 Sustainable Development and Planning II, Vol. 2 The final accessibility map is presented in Figure 4, where the accessibility scores were adjusted to a scale. The reclassified image (Figure 4-b) shows a score distribution of irregular rings centred in the Northwest part of the study area with two deformation sources due to the presence of the South motorway junction and the presence of the hill area (b) Reclassified (a) Continuous scale Figure 4: Accessibility map of Valença. 5 Application 2: regional accessibility evaluation in Cavado Valley, Portugal In order to evaluate the general accessibility of the region of Cavado Valley, in the Northwest part of Portugal (Figure 5), key-destinations were chosen and grouped after the Cavado Valley Strategic Plan (1995), as follows: (1) Major cities, with national relevance; (2) Minor cities, with regional relevance; (3) Cities in Spain relevant to the region; (4) Sea ports and airports. Table 2 presents key-destinations, weights and fuzzy function data. Figure 6 presents the distribution of the general accessibility index of the Cavado Valley Region. Urban areas and major roads are shown in black. It is quite obvious that the eastern part of the region has lower levels of accessibility as it corresponds to hilly areas where the road network is quit insufficient. 6 Application 3: accessibility evaluation in the University of Minho Campus In order to evaluate the accessibility of the University of Minho Campus, in Braga, Portugal, key-destinations were chosen and grouped according to their

7 204 Sustainable Development and Planning II, Vol functionality within the campus as follows: (1) Pedagogical facilities, Schools, Institutes and Departments; (2) Support Services; (3) Accesses and parking. Table 3 presents the key-destinations groups, weights and fuzzy functions data. N MONTALEGRE ESPOSENDE 13 IC BARCELOS A3/IP1 VILA VERDE AMARES A3/IP BRAGA 101 TERRAS DE BOURO POVOA DE LANHOSO VIEIRA DO MINHO Kilometers (a) Portugal (b) Cavado Valley Figure 5: Cavado Valley: location and road system. Table 2: Cavado Valley: key-destinations, weights and fuzzy function data. Keydestinations Weight Major cities 0,33 Minor cities 0,25 Cities in Spain 0,21 Sea ports Airports 0,21 Weight Guimarães 0,25 Porto 0,25 Viana do Castelo 0,25 Vila Real 0,25 Chaves 0,20 Felgueiras 0,20 Póvoa de Varzim 0,20 Vila do Conde 0,20 V. N. Famalicão 0,20 Vigo 0,50 Orense 0,50 Porto de Leixões 0,20 Porto de V. Castelo 0,20 Porto de Vigo 0,20 Aeroporto do Porto 0,20 Aeroporto de Vigo 0,20 Fuzzy function Sigmoidal Dist. a (Km) Dist. b (Km) 0 60 Sigmoidal 0 90 Sigmoidal Sigmoidal The campus buildings and walking network can be seen in Figure 7. Figure 8 presents the distribution of the accessibility to Support Services in the Campus. As expected, the concentration of Support Services in the central part of the campus unbalances the accessibility over the campus as marginal areas have accessibility levels considered too low.

8 802 Sustainable Development and Planning II, Vol. 2 N 0,00 0,06 0,13 0,19 0,25 0,31 0,38 0,44 0,50 0,56 0,63 0,69 0,75 0,81 0,88 0,94 1, Kilometers Figure 6: Cavado Valley: General accessibility index. Table 3: University Campus: key-destinations, weights and fuzzy function data. Key-destinations Weight Fuzzy function Dist. a (m) Dist. b (m) Pedagogical facilities, Schools, Institutes and Departments 0,37 Sigmoidal 0 Varies from 119 to 356 m Support Services 0,34 Sigmoidal 0 Accesses and Parking 0,29 Sigmoidal 0 Varies from 138 to 406 m Varies from 109 to 295 m Rede pedonal Metros Meters Figure 7: University Campus: buildings and walking network. Figure 8: University Campus: accessibility to Support Services index.

9 Sustainable Development and Planning II, Vol Conclusions In this paper a multicriteria accessibility evaluation model was developed within a GIS context and specified for three different types of problems. The proposed model calculates an accessibility index given by the weighted summation of cost-distances to a number of key-destinations. Relevant elements in this model include: The calculation of cost-distances making use of a friction surface that represents the resistance to movement, combining road and off-road frictions. This approach overcomes the insufficiencies of the models that consider Euclidean distances or, alternatively, network distances. The standardisation of cost-distances using fuzzy set membership functions that, when calibrated, represent much better the effect of distance in the evaluation. The combination of cost-distances taking into account the relative weight of key-destinations in the evaluation. The implementation in a GIS environment, taking advantage of the map algebra and visualisation toolbox. This configuration is particularly useful when a raster data model is used, as it allows the calculation of the accessibility for each pixel, resulting in a continuous accessibility surface. References [1] Hoggart, K., Transportation Accessibility: some references concerning applications, definitions, importance and index construction. Council of Planning Librarians: Exchange Bibliography 482, Monticello, [2] Ingram, D.R., The Concept of Accessibility: a search for an operational form. Regional Studies, 5(2), , [3] Morris, J.M.; Dumble, P.L.; Wigan, M.R., Accessibility Indicators for Transport Planning. Transportation Research. Part A, Policy and Practice, 13(2), , [4] Mackiewicz, A.; Ratajczak, W., Towards a New Definition of Topological Accessibility. Transportation Research. Part B, Methodological, 30(1), 47-79, [5] Love, D.; Lindquist, P., The Geographical Accessibility of Hospitals to the Aged: a Geographical Information Systems Analysis within Illinois. Health Services Research, 29(6), , [6] Geertman, S.C.M.; Van Eck, J.R.R., GIS and Models of Accessibility Potential: an Application in Planning. International Journal of Geographical Information Systems, 9(1), 67-80, [7] Allen, W.B.; Liu, D.; Singer, S., Accessibility Measures of U.S. Metropolitan Areas. Transportation Research. Part B, Methodological, 27(6), , [8] Voogd, H., Multicriteria Evaluation for Urban and Regional Planning. Pion, London, 1983.

10 804 Sustainable Development and Planning II, Vol. 2 [9] Mendes, J.F.G., Decision Strategy Spectrum for the Evaluation of Quality of Life in Cities. In Foo Tuan Seik, Lim Lan Yuan and Grace Wong Khei Mie (eds.), Planning for a Better Quality of Life in Cities, NUS, Singapore, 35-53, [10] Osgood, C.E.; Suci, G.J.; Tannenbaum, P.H., The Measurement of Meaning. University of Illinois Press, Urbana, [11] Saaty, T., A scaling method for priorities in hierarchical structures. Journal of Mathematical Psychology, 15, , [12] Jiang, H.; Eastman, J.R., Application of Fuzzy Measures in Multi-criteria Evaluation in GIS. International Journal of Geographical Information Science, 14(2), , 2000.