GIS BASED DECISION SUPPORT SYSTEM FOR SEISMIC RISK IN BUCHAREST. CASE STUDY – THE HISTORICAL CENTRE

Because of the increasing volume of information, problem decisions tend to be more difficult to deal with. Achieving an objective and making a suitable decision may become a real challenge. In order to better deal with decision making, decision support systems (DSS) have been developed. The decision support systems (DSS) can be used in any kind of a decision-making process and are very suitable in situations that involve a lot of stakeholders and a large number of criteria. DSS offers support in the decision-making process (how a decision should be made), and it does not focus on the result (what decision that should be made). DSS can also involve a large number of stakeholders and criteria, in the same time. A limitation of this method is that, regardless of the mathematical results, the final decision has to be made by the decision maker. Depending on the nature of the decision problem, a decision maker can use decision support systems (DSS), if the decision problem is economic or technical, and spatial decision support systems (SDSS), if the decision maker is faced with a spatial decision problem. The main objective of the present study is to apply a spatial decision support system in order to find a suitable shelter in the historical centre of Bucharest City in the post-disaster phase, in case of an earthquake occurrence. The present work represents a first step in applying SDSS in the context of the seismic risk in Bucharest. For the present paper, the SMCE Module for ILWIS 3.4 was used. The method included the following steps: structuring the problem in a decision tree, applying standardization and weighting methods to the criteria, finding suitable alternatives and choosing one of the alternatives. The results show that several buildings can be used as a shelter and among these are ‘Creditul Roman’ Bank Palace, the National History Museum and the National Bank of Romania.


INTRODUCTION
Because of the increasing volume of information, problem decisions tend to be more difficult to deal with. Achieving an objective and making a suitable decision may become a real challenge. In order to better deal with decision making, decision support systems (DSS) have been developed. The decision support systems (DSS) can be used in any kind of a decision-making process and are very suitable in situations which involve a lot of stakeholders and a large number of criteria, in the same time. DSS offers support in the decision-making process (how a decision should be made) and it does not focus on the result (what decision that should be made) [1].
In DSS, the main challenge is problem structuring [2] and the understanding the characteristics of the problem, e.g. if we face an economic or a technical problem [3]. Meantime, there are two separate visions of DSS [4]. There is the objective rationality or substantive rationality, in which the decision process results for best solution/decision and emphasizes how a decision should be made, and there is the procedural rationality or the bounded rationality. In the last one, the decision maker searches for an optimal or satisficing solution and emphasizes how decision is made [5]. There are several frameworks for decision support systems [4,6,7]. For the present study, we have selected Sharifi's model (2004), which is an adaptation of the original Simon's model (1960). This model includes four main phases: (i) the intelligence phase, where the system is described and understood, then the problem is identified and the objectives are set; (ii) the design phase, in which feasible alternatives are generated after formulating and validating the model; (iii) the choice phase, in which the alternatives are being evaluated and one of them is selected and (iv) the implementation phase.
If DSS involves spatial problems, then it becomes a spatial decision support system (SDSS). Even though SDSS and GIS rely on GIS technology, they are not the same thing. SDSS is based on spatially referenced data, and GIS is based on spatial data [8]. SDSS is able to deal with complex spatial problems, proving a framework for integrating database management systems, graphical display, tabular reporting capabilities and the expert knowledge of decision makers (Table 1) [9]. Table. 1. Differences between SDSS and GIS [9]. SDSS GIS Flexible -support individual decision-making approaches. Chiefly designed for decision-making environment. Designed to solve ill-structured problems. Flexible, combining analytical models and data. Provides feasible alternatives. Iterative, integrative and participative. Can also include non-spatial data in the analyses.
Supports only cartographic display. Limited geographical information analysis. Too rigid to model (specific data formats, resolution). Limited tabular or diagram reporting. Shortage in analytical modelling techniques. Shortage in designing ill-structured problems.
There are many examples for the theory of SDSS [8][9][10][11][12][13] and the applications involved [5,[14][15][16][17][18]. Because of the complexity of the decision-making process, spatial multi-criteria decision analysis (SMCDA) have been developed. SMCDA is a process that combines and transforms geographical data (the input) into a decision (the output), constituting a series of procedures that include spatial data, the preferences of the decision maker, the data manipulation and the preferences regarding the specific decision rules [5]. For the present paper we used a more simplified Sharifi's model (2004), as seen in Figure 1. The structure of the paper follows the workflow of this model, presenting the problem structuring and the study area, the process of designing the alternatives, the process of decision making and the sensitivity analysis. Finally, some brief conclusions are offered.

PROBLEM STRUCTURING AND THE STUDY AREA
The main goal of this research is to find a suitable shelter in the historical centre to be used in the post-disaster phase if an earthquake occurred in Bucharest City ( Figure 2).

Fig. 2. Study area.
The reason for choosing the historical centre for this goal is because it is highly populated. There are not only residents, but many visitors and employees. Almost all buildings have ground floor pubs, clubs, restaurants or stores, which attract many visitors and need many employees for sustaining the touristic services.
The main problem is that the buildings are very old and damaged. Many buildings were constructed in the 19 th century and have experienced since then two world wars and several intense earthquakes, the last one in March 1977. Furthermore, most of these buildings have neither been consolidated nor properly maintained, making them extremely vulnerable to earthquakes [19].

DESIGN OF ALTERNATIVES
The SMCE module from ILWIS 3.4 was used in this decision-making process, and the multicriteria approach was selected, because of the problem complexity. We have identified four groups of factors for designing the alternatives: 1) the building characteristics and 2) function; 3) the population; 4) the building damages in case of an earthquake. Besides these groups of factors, we have also used two constraints: the state and the area of the building. It was considered that damaged buildings or buildings with an area less than 1,000 square meters cannot be used for shelters, and they were eliminated from further analysis ( Table 2). Economic losses All the above data were structured in a criteria tree. After problem structuring, every factor was standardized, resulting in values ranging from 0 to 1, depending on whether the factor represented a benefit (the more of it, the better for the goal) or a cost (the more of it, the worse for the goal) for the analysis.
The following equations were employed for value input standardization, depending on the nature of the criterion [21]: Benefit factor = value / maximum input value (1)

Cost factor = 1 -(value / maximum input value) + (minimum input value / maximum input value) (2)
We analyzed all the buildings having damage codes greater than 0 (used for damaged buildings or under construction) and areas larger than 1,000 square meters. The next step was to weight the factors and the group of factors. The method selected was the ranking order, setting the factors and/or the group of factors by their importance. From these 23 buildings, only 6 were selected as alternatives for the decision-making phase (Figure 3), based on the values obtained.

DECISION MAKING
The third phase of the analysis consists in evaluating the alternatives. For this phase, we also used the SMCE module of ILWIS software, but selecting the decision-making mode. A criteria tree was designed, with a different column for each alternative. The alternatives were evaluated based on three non-spatial criteria: area, suitability and no. of inhabitants and/or employees ( Figure 4). Standardization was applied, considering the area and suitability as benefits and the number of persons living in each building as a cost. The ranking order with expected value was selected for weighting. We considered that the most important criterion for choosing one alternative is the suitability, secondly it is the persons living and/or working in the building and thirdly the area of the building.
The results show that the highest value was obtained by building 4, which also has the highest suitability value. The lowest value was obtained by building 1, which has the second lowest suitability value, but has the highest number of persons living and/or working there.

SENSITIVITY ANALYSIS
Sensitivity analysis is performed to test the robustness of the analysis [13]. In decision-making problems, the biggest fear of the decision maker is if the decision is wrong. What if not all the criteria has been taken into account, or what if the standardization or weighting applied was not the most suitable? For avoiding making bad decisions, sensitivity analyses can be very helpful. They consist of changing one or more elements of the performed analysis. It can be used by changing the standardization or by changing the weighting method, or both. If the results are very similar to the original ones that means the analysis is very robust, so the decision is satisficing, under the given conditions. However, if the results differ with every small change that means that the analysis is very sensitive and should be reassessed.
For the present analysis, a sensitivity analysis was also applied ( Table 3). Because of the nature of the criteria, we can assume that the standardization is correct, so we can change only the weighting methods. Also, it was always assumed that suitability is the most important criterion.
In a first phase, it was considered that the number of inhabitants and/or employees is more important than the surface of the building. In this case, regardless the weighting method used, the results are very similar, showing the highest values for building 4.
Secondly, it was assumed that the area of the buildings is more important than the number of persons. In this case, for expected value and pairwise methods the results are similar, but for rank sum method, the highest value was obtained by building 6, with a small difference. As a conclusion, we can affirm that the analysis performed in this paper is a robust one, and the most suitable building that can be used in case of an earthquake occurrence is building 4 -'Creditul Roman' Bank Palace.
Even though the suitability for one alternative has a lower value, depending on the other criteria and on the weighting method applied, it can still become the most satisficing solution. In the present analysis, the most satisficing solution had also the highest suitability value, but if we observe buildings 1 and 6, we will see that the ranking resulted from the suitability analysis does not maintain in the decision making phase. Because building 1 has the greatest number of persons, at a considerable difference from the others it obtained the lowest values in the sensitivity analysis, despite the fact that it did not have the lowest suitability value. Similar, building 6 had the lowest suitability values, but did not obtain the lowest scores in the sensitivity analysis.

CONCLUSIONS
Decision support systems can be very helpful, especially in spatial planning, when a lot of data and of stakeholders are involved. A great advantage of these systems is that they can take into account, simultaneously, a large number of opinions, translated into 'what I want from this'. The stakeholders involved can differ significantly, they can be public authorities, private companies or NGOs, all involved at the same time, in the same decision-making problem. Also, because they do not focus on the result, on obtaining the 'best' solution, but on the decision-making process, they ensure transparency in decision-making. Using DSS we do not obtain best solutions, but satisficing solutions and are based on trade principle, from the stakeholder's point of view. By trading, one stakeholder can maintain one demand and give up another one, in perfect agreement with the other ones.
A limitation of these DSS is that they are semi-quantitative and they rely on human judgement. The decision maker has to be able to integrate all the demands of the stakeholders, to structure the problem, to find suitable criteria and data needed, to perform a suitable standardization and weighting of the factors or group of factors. Also, a sensitivity analysis is very useful for testing the robustness of the decision-making analysis.