Diverting the tourists- a spatial decision-support system for tourism planning on a developing island

Impact of Tourism on Small Islands

  • Tourism has both positive and negative impacts on host countries.
  • The scale and intensity of these impacts depend on:
    • Extent of tourist site use
    • Resilience of ecosystems
    • Pattern and degree of facility development
    • Size of the country
  • Small developing countries experience more pronounced environmental and cultural effects from tourism than larger, developed countries.
  • Many small countries are islands relying on natural resources, climate, and landscape to attract tourists.
  • Sun, sea, and sand give islands a competitive edge in tourism, concentrating activities along coastal areas and causing changes in the natural environment.
  • The tourism industry is a main foreign exchange generator in these small countries.
  • Governments maximize tourism potential through promotion, facilities, and transport links [Briguglio & Briguglio, 1996].
  • Carrying capacity and limits on acceptable change are concepts used when analyzing tourism's environmental and societal impacts.
  • There isn't a universally valid capacity threshold for all countries or types of countries.
  • While there is no rigorous definition of carrying capacity, there is awareness that some maximum, tolerable degree of change exists for interrelated subsystems:
    • Economic
    • Environmental
    • Social
    • Cultural [Johnson & Thomas, 1996]
  • Tourist density (tourist arrivals per square kilometer) indicates tourist pressure on the physical environment.
    • 1990 average tourist densities:
      • Large territories (> 1,000,000 km^2): 0.93
      • Medium territories (1,000,000 km^2 - 100,000 km^2): 9.46
      • Small countries (< 100,000 km^2): 70.72
    • In 59 small states (< 10,000 km^2), tourist density was as high as 402.88.
  • Small countries' natural and cultural landscapes are under greater pressure than larger countries.
  • Tourist activities concentrated along coastal zones exert heavy pressure on the environment and local infrastructure.
  • Tourist-host contact is another capacity indicator.
  • Tourist-host ratio (tourist arrivals per 1000 host population) measures the frequency and intensity of tourist-host contact [Liu & Jenkins, 1996].
  • The tourist-host ratio can indicate the social consequence of tourism - the tourist-host relationship.
  • It may indicate the level of acceptance, tolerance, or resentment of the host population for tourists.
  • Tourist-host ratios are lower in large countries, indicating higher tourist-host interaction in smaller countries, leading to greater socio-cultural effects.

Overview of Mauritius

  • Mauritius is perceived as an upmarket tourist destination due to its tropical climate, attractive beach resorts, and stable political and economic conditions.
  • It is a small volcanic island in the southwest Indian Ocean (20° S 57° 30’ E), located 880 km off Madagascar and 2000 km off Africa.
  • The island is densely populated with one million people inhabiting a land area of 1865 km^2. The coastline is almost entirely fringed by coral reefs.
  • The land rises from the coastal plain to a central plateau, reaching a level of some 600 m.
  • Summer temperatures (November to April) may reach 35°C; winter temperatures (May to October) may reach 15°C.
  • Until recently, the Mauritian economy was largely dependent on the sugar sector.
  • Manufacturing has become the island’s leading sector in terms of export earnings and employment.
  • Tourism is the third most important economic sector.
  • Mauritius has air connections with most main European, African, and Asian cities, as well as Australia and other islands of the Indian Ocean.
  • The economy has been doing well for nearly a decade after a long period of recession.
  • Economic success has led to:
    • Increased employment level
    • Rise in the standard of living
    • Considerable pressure on housing, transportation, community facilities, and recreation
    • Intense competition for land among the industrial, agricultural, and tourism sectors.

Problem Definition

  • Mauritius has a coastline 322.5 km long.
  • The first hotel developers leased vast stretches of sandy beaches from the government.
  • Hotel sites occupy 41.9 km of coastal zones, and including bungalows, they occupy 29% of the coastal area.
  • Public beaches total 26.6 km in length, representing 8.2% of coastal land use.
  • Competition for access to sandy beaches arises among hotel developers, bungalow owners, and the public.
  • Pressure to build new hotels directly on beach frontages is relentless because tourism is a growth industry for the local economy and it is highly lucrative.
  • The main tourist zones are along the north, east, and southwest coasts where the best beaches are located.
  • There are practically no tourist zones inland.
  • The tourism industry is expected to expand, and tourist arrivals to increase in future years [Vision 2020, 1997].
  • Two options can meet the demands of this increase:
    • Expanding and further exploiting existing tourist zones
    • Exploiting alternative sites.
  • Further exploitation of existing zones will put more pressure on the already over-stressed regions along the coast.
  • The number of tourists visiting the island has rapidly increased from 292,000 in 1990 to 558,200 in 1998 [MTL, 1996; MTL, 1999a].
  • Similarly, tourist densities and tourist host-ratios have been rising quite rapidly over the past years (Table 1).
  • The intensification of tourist activities in local regions is detrimental to the environment.
  • The natural coastal environment is a primary resource of Mauritius, supporting sun-sea-sand tourism, including such activities as scuba diving, water skiing, snorkeling, deep-sea fishing, and motorboat racing, all which are, however, damaging to the coastal surroundings.
  • Fishing villages have been transformed into tourist recreation sites.
  • The large amount of waste (sewage, fuel emissions) generated by tourism-related activities is polluting the land, the air, and the sea.
  • Hotels tend to construct piers or jetties, which can severely interfere with the long shore movement of sand, creating sand erosion further down the coast, and which interfere with the free passage of the public up and down the coast.
  • Clearing of seaweed, coral, and other rocks close to the shore in lagoons has regularly been carried out to create suitable bathing areas or skiing lanes.
  • Pleasure craft contribute to the destruction of coral as a result of anchor damage.
  • Environmentalists and ecologists believe that already too much harm has been done to the environment and they insist that remedial action has to be taken.
  • Signs of environmental destruction and pollution will not only cause tourists to go to cleaner islands, but the island will irreparably lose its natural resources and its beauty.
  • Tourism depends for its success on the quality of the natural and human environment [Manning & Dougherty, 1995].
  • If tourism facilities and activities result in a high degree of environmental degradation, it is likely that tourism will decline.
  • The reduction in tourists will cause loss in national income that could result in curtailment of essential services, which will further alienate tourists.
  • Thus this industry has to be carefully maintained to obtain maximum benefits for least damage.
  • To accommodate the increase in tourist arrivals it seems to be more appropriate, therefore, to consider alternative sites and avoid further deterioration of existing tourist zones.
  • The choice of suitable alternative sites calls for sound planning techniques.

Spatial Planning

  • The space limitation of Mauritius warrants a rational approach to tourism development on the basis of long-term vision and the application of sound planning techniques [Ramsamy, 1994].
  • Spatial planning is essentially a decision-making process.
  • Relevant criteria have to be identified, analyzed, combined, and evaluated in order to meet specific objectives.
  • The process of evaluating several criteria is called Multi-Criteria Evaluation (MCE).
  • Multi-criteria methods provide a flexible way of dealing with land allocation decisions.
  • A problem faced in land use planning has been to match the resources attributes of an area with the appropriate activity [Boyd et al, 1994].
  • Geographic Information Systems (GISs) technology allows the matching of recreation potential with the characteristics of the regions.
  • The capability of a GIS to allow rapid modification, addition, or removal of constraints and to investigate the complex interrelations between the thematic layers is attractive for resource management and planning problems.
  • This dynamic tool for planners is capable of being readily adjusted as new data become available and as there is a change in requirements needs and preferences over time.
  • While lovely beaches do attract tourists to the island, the results of a survey conducted by the Ministry of Tourism & Leisure suggest that the prime motivation of tourists in choosing Mauritius was its tropical image (Table 2 - Survey: MTL 1999b).
  • Besides its beaches and water sports, other attractions such as tropical climate, cuisine, colorful festivals, natural sights, rare flora and fauna, historical sites, local handicrafts, and the welcome from the local people are highly rated by tourists.
  • Coupled with appropriate models, GIS can be used to provide a more holistic approach towards problem solving in which qualitative and quantitative information has to be processed.
  • It can also give a visual display of results in the form of graphs or maps, thus allowing rapid and efficient appraisal of results.
  • GIS can be used to promote participation at decision-making level.
  • Access to information need not be restricted to GIS scientists or policy makers but to all parties concerned - technical advisers, planners, interest groups and local people.
  • Thus user-friendly GIS can become a participatory and exploratory tool since discussions and negotiations are important aspects of decision-making [Carver et al, 1995].
  • Thus even if facilities are developed inland, or along the western and southwestern coasts, they will still be attractive to tourists.
  • The pressure on the northern and eastern coastal peripheries resulting from the concentration of visitors there can thus be alleviated or at least prevented from increasing.
  • To improve the usefulness of GIS as a decision-support tool, two needs are apparent.
    • First, decision-makers require methods that allow them easily to select alternatives most closely aligned with their priorities across a number of relevant criteria.
    • Second, it is necessary to recognize explicitly that most decision-making processes involve more than one participant.
  • Since problem solving is often characterized by multiple and conflicting objectives, methods that contribute toward consensus building are required.
  • Supporting decision-making in a spatial context is implicit in the use of GIS.
  • However, the analytical capability of a GIS has to be enhanced to solve semi-structured and unstructured problems characterized by multiple and conflicting objectives and criteria.
  • Thus integration of GIS with MCE techniques is being considered in the present work.
  • A prototype Spatial Decision Support System (SDSS), Spatial Multicriteria Evaluator (SpaME), for tourist resort location is being developed.
  • This SDSS will allow the user or decision-maker (DM) to be responsible for major inputs into a model developed using multi-criteria and multi-objective decision-making tools.
  • This can serve as an instrument of exploration with the user’s value systems and knowledge.
  • Users can thus interactively change or refine their inputs as the SDSS provides results from their previous inputs.
  • The aim of the SDSS is to combine spatial data from diverse sources in order to analyze interactions, to make optimum site selections with models and to provide support to DMs.
  • Decision-making is seen as a non-linear and dynamic recursive process.

Multicriteria Evaluation (MCE) Techniques

  • Spatial decision-making problems, such as site selection, require the DM to consider multiple and conflicting criteria and objectives.
  • Therefore a solution which is simultaneously best from all points of view does not exist.
  • Essentially, compromise alternatives and their rankings are generated according to their degree of attractiveness [Janssen & Rietveld, 1990].
  • MCE techniques aim at evaluating the multiple and conflicting criteria and objectives to identify ‘acceptable’ alternatives.
  • MCE is not a procedure through which pre-existing truths can be arrived at (or discovered), and it does not rely on a mathematical property of convergence, ie, the decision process does not automatically lead in a number of steps, to the optimum solution’.
  • In fact the final solution is more like a creation than a discovery.
  • In multi-criteria decision-making the principal aim is not to discover a solution but to construct or create something which is viewed as liable to help “an actor taking part in a decision process either to shape, and/or to argue and/or to transform his preferences, or to make a decision in conformity with his goals” [Roy, 1990].
  • MCE strategies can be classified into two categories depending on the level of cognitive processing demanded of the DM:
    • Compensatory (high processing)
    • Non-compensatory (reduced processing)
  • The compensatory approach is based on the assumption that the high performance of an alternative achieved for one or more criteria can compensate for the weak performance of the same criteria for other criteria.
  • The compensatory approach is cognitively demanding since it requires the DM to specify criterion priorities expressed as cardinal weights or priority functions.
  • Compensatory MCE techniques can be subdivided into additive techniques and the ideal point approach.
  • While the simplest approach is to use a linear utility function (weighted summation) there are also non-linear utility functions, such as multi-attribute utility theory (MAUT) by Keeney & Raiffa [1976].
  • Another method that is quite commonly used in MCE is concordance-discordance analysis.
  • Each pair of alternatives is analyzed for the degree to which one outranks the other on specified criteria.
  • Using a user-defined minimum concordance index and maximum discordance index, a dominance matrix can be calculated from the standardized evaluation matrix showing the outranking relationship of each alternative to the others [Carver, 1991].
  • Concordance-discordance analysis is computationally impractical in raster GIS as each cell in the grid is considered an alternative [Eastman et al, 1995, Jankowski, 1995].
  • In the ideal point approach the DM is asked to locate his ideal solution.
  • The distance between the ideal solution and each considered alternative is measured in order to arrive at a ranking of alternatives [Jankowski, 1995].
  • Under the non-compensatory approach, the alternatives are compared along the set of criteria without making intra-criterion compensations.
  • This technique leads to a stepwise reduction of the set of alternatives without trading off their deficiencies or their strengths along other evaluation criteria.
  • Boolean intersection results in very hard AND - a region is excluded from the result if a single criterion fails to meet the threshold.
  • Conversely, the Boolean union (OR) implements a very liberal mode of aggregation - a region is included in the result even if one criterion meets its threshold [Eastman, 1999].
  • Jankowski [1995] reviews the dominance, conjunctive, disjunctive and lexicographic non-compensatory techniques
  • A reduced level of cognitive processing is required from the DM when non-compensatory MCE techniques are employed, since he/she does not specify criterion priorities.
  • The disadvantage of these techniques is a potential for recommending an inferior alternative due to their reduced processing strategy [Jankowski, 1995].
  • The choice of the techniques is determined by the type of problem at hand, the type of data and the DM’s needs.
  • Hong & Vogel [1991] discuss how strategies are selected and implemented in accordance with the problem descriptions and decision needs specified by the DM.
  • Preference weights can be applied to individual criterion specified in the siting problem such that the sites identified possess an optimum or near optimum mix of characteristics.
  • One of the main drawbacks of the MCE is that many DMs would not have the appropriate background in mathematics to be able to completely appreciate what is happening.
  • This becomes particularly important when a DM may be called upon to justify a decision rule and results.
  • To improve comparisons between MCE techniques more than one method can be implemented in the same DSS.
  • Different MCE methods can be applied to allow DMs to explore the sensitivity of the different models.
  • A variety of different procedures exist for dealing with method sensitivity [Lodwick, 1989; Carver, 1991].
  • The MCE approach in which the DM varies the importance and selection of the criteria in an interactive way has a number of distinct advantages.
  • The technique permits the DM to interact with the decision rule so that he/she has a better appreciation of how changes in weights and scores will influence the decision outcome.
  • In addition, for different DMs the results will vary according to their objectives.
  • A comparison of the views of different users, hotel owners, environmentalists and members of public, can indeed be a very interesting exercise.

Implementation and Analysis Software, Hardware, and Data

  • A database was developed and used to generate scenarios for the identification of potential tourism zones taking into consideration the relevant criteria.
  • A conceptual model of the real world situation was created.
  • Decisions about the real world were made by referring to this model, which was much simpler than the real world because only pre-selected information considered to be relevant (and which was available) were included in this model.
  • For such a system to function satisfactorily, three important components had to be considered:
    • The computer software
    • The hardware
    • Proper organization
  • Initially the software package used was ARC/INFO, version 7.0 operating on a UNIX-based workstation.
  • Later all the work was implemented using ARCVIEW 3.1 because of its more user-friendly interface.
  • The scripts are written in the ARCVIEW programming language, ie, Avenue.
  • No relevant data were available in digital form and all the data was digitized manually.
  • Two maps were available for digitisation at scales of 1:300,000 and 1:100,000.
  • The 1:100,000 scale paper map of Mauritius, produced in 1994 by Ordnance Survey of Britain, was used as the data source for digitising the various data layers required.
  • As a first step a preliminary GIS model was constructed with available data.
  • Spatial analysis was carried out and scenarios developed.

Non-Compensatory Technique: Boolean Linear Combination

  • The first technique dealt with in this study was a non-compensatory one - the Boolean linear combination.
  • This is a common and simple spatial analysis.
  • The analysis was carried out in the vector mode.
  • The criteria and constraints relevant to the problem were defined and the data required to proceed with the analysis was identified.
  • Access characteristics were shown in terms of road access and the several categories of roads (Motorway, A-Road and B-Road) were digitized.
  • The areas that fell under certain buffer restrictions of the various roads were represented in a new layer (road buffer).
  • In terms of presence of dense residential areas and industrial zones, a buffer was drawn around each, ensuring that the potential sites would not be too close to already densely populated areas or existing tourist zones.
  • It was considered that no site could potentially be located on agricultural land or nature parks.
  • It was also deemed that tourist zones could not be located on mountain peaks.
  • The reasons being difficulty and cost of hotel construction, access problems and greater risk of damage during the cyclones.
  • The minimum land area requirement was another important criterion.
  • It has been estimated that the present 7,267 hotel rooms on the island can only be increased to 9,000 rooms.
  • Beyond this ‘green ceiling’, increased earnings will have to come not from higher numbers but from higher spending per visitor, with still higher standards of provision and a wider range of activities, including, perhaps, inland and eco-tourism [Vision 2020, 1997].
  • The approximate area requirement for a hotel project proposed by a Mauritian is a minimum of 2.5 acres; proposals by non-Mauritians must entail a minimum of 10 acres [Ministry of Tourism & Leisure, 1999].
  • The site would have to be a single parcel of land and have a surface area of at least 0.5 km^2 to accommodate 6 to 10 hotels with 100 to 150 rooms.
  • This surface area has been chosen so that a group of hotels can be built with all amenities and offer indoor and outdoor facilities.
  • Taking into consideration the above criteria, the following thematic layers were identified and digitized:
    • Access (Motorway, A-Roads, B-Roads)
    • Communities (Major Inhabited Areas, Industrial Communities and existing Tourist Zones)
    • Land Use (Agricultural Areas, Nature Parks, Reservoirs)
    • Landscape (Elevation)
    • Coast
  • The Boolean linear model is the traditional vector approach for converting the criteria into Boolean statements suitable for the decision under consideration.
  • The criteria having been defined as a set of deterministic rules, Boolean operators are applied to the set of input maps, and the output is a binary map because each site on the map is either appropriate (true) or inappropriate (false).
  • For example, the site must be:
    • Within 1500 m of existing A-Roads, B-Roads, Motorway AND
    • At least 1500 m away from existing Tourist Zones AND
    • At least 1000 m away from built-up areas AND
    • NOT on land over 600 m above sea level AND
    • NOT on Nature Parks AND
    • NOT on agricultural land AND
    • Its area should be ≥ 0.5 km^2
  • From this initial stage of the project, 46 potential sites were identified (Figure 2).

Analysis in Raster Mode

  • The non-compensatory technique used for the vector analysis was applied in the raster mode when DEM and slope grids of Mauritius were obtained from the Centre for Resource and Environmental Studies in Australia.
  • It was decided that all the analysis would be carried out consistently in one mode - raster mode with a cell size of 100 m x 100 m.
  • The results for the Boolean linear combination in the vector and raster mode matched.
  • Boolean linear combination is not a flexible technique and if an alternative does not meet some crisp threshold it is eliminated completely from further consideration.
  • Threshold values have to be defined and the use of such threshold values to map continuous variables on a nominal basis inevitably leads to substantial losses of information [Carver, 1991].

Compensatory Technique: Weighted Linear Combination

  • The second approach adopted was the application of weights to the evaluation criteria.
  • Traditionally most overlay procedures in GIS do not allow for the fact that the variables may not be equally important.
  • Preference weights can be applied to individual criterion specified in the siting problem such that the sites identified possess an optimum or near optimum mix of characteristics.
  • The task of identifying the best site is not merely a problem of finding the site that performs best on all criteria, since many of the specified criteria in this problem are conflicting.
  • For example, the site has to be close to power lines, telephone lines and the water supply, yet far from populated areas (accessibility to utilities is greater in urban areas).
  • Thus MCE techniques are necessary in this instance since they allow complex trade-offs to be made between conflicting criteria on the basis of specified preference weights.
  • The user or DM is given the flexibility to change the importance of each criterion depending on the current objectives.

Evaluation Matrix

  • The basic starting point of any MCE analysis is the construction of an evaluation matrix, the elements of which reflect the characteristics of the given set of choice alternatives on the basis of a specific set of criteria [Carver, 1991; Voogd, 1983].
  • In a spatial context, an alternative may refer to a region as an allocation for a certain activity.
  • If i is used to indicate an alternative and j to indicate a particular criterion, then p_{ji} can be interpreted as the effect of alternative i according to criterion j, where i = 1…I and j = 1…J. The effects of all alternatives according to all criteria can be summarised by the evaluation matrix, P:

P =
\begin{bmatrix}
p{11} & … & p{1J} \
… & … & … \
p{I1} & … & p{IJ}
\end{bmatrix}

  • The criteria and constraints are not set as discrete crisp variables but as continuous variables [Voogd, 1983].
  • Thus proximity to roads would not be treated as an all-or-none buffer zone of suitable locations but rather as a continuous expression of suitability according to a special numeric scale [Eastman, 1999].
  • In many analyses, especially those utilizing quantitative and mixed sources of data, some form of standardisation or criterion scores is necessary to enable meaningful comparisons to be made on the basis of criteria measured on different scales [Carver, 1991].
  • Several common standardisation techniques are available.
  • These are described by Voogd [1983] and include methods with an additivity constraint (ie, the standardised scores are based on raw scores divided by the sum of the raw scores), ratio scale properties and interval scale properties.
  • The most commonly used technique is linear scaling between the maximum and minimum values of the criterion -the linear resealing is to a consistent range (0 -1).
  • The resulting factors express the varying degrees of suitability for the decision under consideration.
  • For a factor i, if the highest original score is considered being best, the expression below is applicable, with the highest original score being resealed to 1 and the lowest resealed to zero.

Criterion Standardisation Score:

e{ji} = \frac{p{ji} - min \ p{ji}}{max \ p{ji} - min \ p_{ji}}

  • The expression can be modified to set the highest raw score as being worse, ie, being standardised to 0.
  • Each factor is then multiplied by a weight and summed to arrive at a multicriteria solution.

Suitability Calculation

Suitability = E'w

  • where E' is the transpose of the standardised evaluation matrix and w the weight.
  • The continuous suitability map has the same numeric range as the standardised factors if the weights that are applied sum up to 1.0 [Eastman, 1999].

\sum w_j \leq 1

\sum w_j = 1

  • The above procedure is the most prevalent MCE technique.
  • It is known as the Weighted Linear Combination [Voogd, 1983].
  • When Boolean constraints also apply, the procedure can be modified by multiplying the resulting suitability map by each of the constraints to ‘zero out’ unsuitable areas [Eastman et al, 1995].

Suitability with Constraints

Suitability = \sum Wj e{ji} \times \prod C_k

  • where wj is the weight of factor i, and e{ji} is the standardised criterion score of factor i, C_k is the criterion score of constraint k and \prod is the product.

Criterion Weights and AHP

  • In the weighted technique, the preferences of the DM were applied as weights to the evaluation criteria.
  • These weights have to be calculated each time before starting any weighted model.
  • Although a variety of techniques exist for the development of weights [Voogd, 1983], one of the most promising would appear to be that of pairwise comparisons developed by Saaty in 1977 in the context of a decision-making process known as the Analytical Hierarchy Process (AHP)[Eastman et al, 1995].
  • AHP basically extracts a relative weight for each objective by computing an eigenvector for that objective.
  • The eigenvector is computed from a square matrix that is populated with pairwise comparisons of the relative importance of each objective to the overall goal.
  • This means that the user has to provide these judgements for the model to be evaluated.
  • Assuming a matrix of size n, this would involve n(n-1) / 2 pairwise comparisons.
  • The derived weights lie in the range 0 to 1 and always add up to 1.
  • AHP is a logical framework, which allows better understanding of complex decisions by decomposing the problem in a hierarchical structure.
  • The incorporation of all relevant decision criteria, and their pairwise comparison allows the DM to determine the trade-offs among objectives.
  • Typically, the type of pairwise comparison is expressed in terms of:
    • Importance
    • Preference
    • Likelihood, depending upon the nature of the goal or objective
  • The actual elicitation of the data can be either:
    • Verbal
    • Graphical
    • Numerical
  • Saaty advocates a scale from 1 to 9 (Table 5).
  • If this limit is exceeded a subdivision is needed to make the analysis more manageable.
  • The AHP is essentially a ratio-based pairwise comparison, which means that if objective X is considered twice as important as objective Y, then it is understood that objective Y is assumed to be half as important as objective X.
  • While this does facilitate consistency among a pair of objectives, it does not guarantee consistency among a set of objectives, ie, if X is twice as important as Y, which is three times as important as Z, then a comparison between X and Z should yield a factor of 6 - which may not always emerge for the user.
  • To counter this, the AHP develops and reports an inconsistency index.
  • For an index greater than 0.1 it is recommended to reconsider the matrix [Saaty, 1990; Zahedi, 1986].
  • The smaller the index, the less inconsistent is the preference matrix.
  • Consistency must also be balanced with the reasonableness of the priorities at the validation stage [Xiang & Whitley, 1994].
  • Although consistency and reasonableness are often highly correlated with each other, there are situations where the decision-maker has strong, crisp, unalterable but inconsistent views [Keeney & Raiffa, 1976].
  • Sometimes excellent consistencies may yield unacceptable results.
  • The final result will be the one with which the DM is satisfied, based on his satisfaction with the priorities.
  • In this study the weights are derived using the AHP method.

Weighted Linear Combination: Implementation

  • The weighted linear combination procedure has been applied stepwise to the siting problem.
  • The criteria are not defined by crisp thresholds but by continuously varying scores, ie, the closer sites are to roads the more suitable they are; similarly, the farther away from urban and existing tourist zones the better, and the less elevated and the less steep the land the more suitable.
  • New grids were generated showing the linear distance decay from roads, urban areas and tourist zones.
  • To allow meaningful comparisons these three grids and the slope and elevation grids were standardised in a linear range (0 - 1).
  • For remoteness from urban areas and remoteness from existing tourist zones, the maximum (farthest) original criterion score was considered as being the best score and was re-scaled to one and the minimum (closest) re-scaled to zero.
  • For proximity to roads, land elevation and slope, the highest raw score was considered to be the worst score and was therefore standardised to zero and the lowest original score re-scaled to one.
  • The standardised grids were multiplied by the relevant criterion weights to obtain weighted standardised road proximity, urban remoteness, tourist zone remoteness, slope and elevation grids.
  • The grids were equally weighted (1/No. of criteria), ie, all weights equal to 0.2 (1/5).
  • Equal weights were used so that the results could be compared to those of the Boolean method.
  • The weighted grids were added to produce a summed grid.
  • As the potential site cannot be on agricultural land, urban areas and tourist zones, these constraints were applied on the summed weighted grid.
  • The suitability grid is one of continuous values, hence a threshold value is required to harden the results.
  • Logically, the highest scoring cells were the most ‘suitable’ ones.
  • Weighted scores above a threshold of 0.5 were arbitrarily considered to be potentially suitable.
  • The contiguous cells were grouped before selecting sites that have an area of at least 0.5 km^2.
  • Sixty potential sites were obtained (Figure 3).
  • The total number of suitable cells (50,955) was much greater compared to the number (20,310) resulting from the Boolean analysis.
  • There is no flexibility in the Boolean approach and any alternative not meeting any of the threshold values of any of the criteria are discarded.
  • On the other hand the weighted linear combination allows for trade-offs among all criteria scores for each alternative and hence the compensated total score is considered.
  • Each time the weights for the criteria change, depending on the DM’s objectives the suitability map changes.
  • For example, an environmentalist considers distance from existing tourist zones as being most important and the weights in Table 7 are determined after pairwise comparison of the criteria.
  • The inconsistency index is 0.05 and therefore the resulting weights are deemed not to be inconsistent.
  • Forty-three potential sites - (29,017 cells) resulted after application of these weights and taking a threshold score of > + 0.5 and area ≥ 0.5 km^2 (Figure 4).
  • All the sites are found far from existing tourist zones.
  • There are twenty-six inland sites (21,566 cells).
  • They are located in regions that have scenic views and good hiking terrain.
  • From the point of view of this DM (the environmentalist) these would be ideal for developing tourism sites with least negative effects on the environment.
  • Therefore weighted linear combination allows a specific decision to be reached by rank ordering the alternatives and selecting as many of the best-ranked areas as is required to meet the specific objective of the analysis.
  • When several users (hotel developers, environmentalists, members of the public) with differing objectives are involved, consensus can be reached by choosing the common sites arrived at by individual preferences, by calculating group weights (the geometric mean of individual weights), or by rank ordering the results for each user and for the group and making a visual comparison.
  • The model allows iterative runs with redefinition and reconsideration of weights until users are satisfied or reach an agreement.

Conclusions

  • In our opinion the Weighted linear combination technique has greater scope for spatial decision-making and site selection than the Boolean combination technique.
  • Weighted linear combination allows for the high performance of an alternative achieved on one or more criteria to compensate for the weak performance of other criteria.
  • In decision-making there are no crisp boundaries and trade-offs are common.
  • As the results show, the number of potential sites having the same scores are much higher using the compensatory approach than the non-compensatory one.