The Decision-Making Problem: A Framework Challenge

The efficient development of information systems and, consequently, the efficient management of knowledge in these systems are highly dependent on a structured conceptual model. Models are always syntheses and cannot contain all aspects of reality. Then, a good conceptual model has to focus on relevant events and entities for being a proper synthesis. In this sense, the framework challenge is to reach the best systematisation of the decision-making possibilities, besides allowing the inclusion of new decision variables only possible to be identified with a structured problem overview (Figure 8.4).

Floods in Bessa streets (Nobrega, 2002)

Figure 8.3 Floods in Bessa streets (Nobrega, 2002).

Structuring level in a decision-making problem. (Adapted from Malczewski, 1999.)

Figure 8.4 Structuring level in a decision-making problem. (Adapted from Malczewski, 1999.)

In a GIS, conceptual modelling acts as a filter that helps to extract relevant information from a large and complex volume of data before running the spatial analysis. A GIS provides very efficient tools for modelling (Ghosh et al., 2011; Ghadiry et al., 2012), which can support water resources applications. In addition to the modelling tools already available in most existing GIS packages, the flexibility of some tools allows the generation of custom models, according to each problem-solving need. The following application is the result of a knowledge modelling process with the input of several consulted experts and does not require a specific GIS software package. It can be implemented in any GIS software and applied to any area with similar characteristics and problems.

The specialists have in-depth and ground-truth knowledge and expertise due to the long-term studies they have already done in the Bessa area. All the data collected during systematic interviews were turned into spatial themes to stimulate “a spatial problem solution”. Spatial thinking (N RC, 2006) is a concept based on three elements: space, tools of representation, and processes of reasoning (Wakabayashi and Ishi- kawa, 2011). After each interview, new spatialized data were modelled based on the experts’ exposure to the previous data and also to the GIS analysis possibilities pointed out during the interviews.

The measures and strategies for the problem came up from the exposure of experts to all available information (data structured) as well as the perception of the lack (or sometimes uncertainty) of information and their intuitive inferences regarding these missing data were considered highly valuable. Some essential aspects observed in those interviews:

i. The management strategies proposed were different for specialists exposed to the same amount and diversity of information;

ii. Even with vast accumulated knowledge about the area in question, the inferences about missing (or uncertain) data were supported and stimulated by the possibilities of data manipulation in the computational environment (GIS);

iii. Initially, the decision-making process was unstructured (data and expert sugges- tions/inferences), but gradually became structured. This feature is common in unstructured or semi-structured problems such as water-related problems.

Table 8.1 Conceptual modelling: description of the criteria selection, goals, and management measures

Central goal

Management measures

Promoting integrated water management (surface and groundwater) in the Bessa district (coastal area) to minimise the social, economic, and environmental impacts caused by frequent floods.

A Establish a diagnosis of flooding susceptibility,

identifying the “best” areas for occupation (areas with low susceptibility to flooding) and evaluating already occupied lots. Criteria: slope, proximity to water bodies (including the artificial channels), legal aspects, aquifer levels, land use, and topography.

В Encourage groundwater use to minimise aquifer outcrops. Pumping simulations with new wells (locations based on expert knowledge and urban planning guidelines) showing a possible aquifer depletion in the dry season. This depletion could be sufficient to contain the recharge of the rainy season, minimising the groundwater interference with the surface runoff. The saltwater intrusion must be the threshold for the simulations. Criteria: aquifer levels, land use, and topography.

C Promote the proper functioning of existing surface channels so that they continue to perform their draining function of both surface and groundwater. Therefore, consider the proximity of the channels to locate new wells according to the measure A diagnosis. Criteria: proximity to the channels and land use.

D Ensure a minimum amount of recharge areas to

promote sustainability in developing urban areas. Protected green areas may be created, as well as pervious pavements and surfaces, to minimize stormwater runoff and keep the recharge aquifer at an equilibrium level. Criteria: aquifer levels, topography, and land use.

Criteria Selection and Conceptual Modelling

The following table tries to make more accessible a synthesis of the criteria selection, goals, restrictions, and management measures included in the developed conceptual model (Table 8.1). One measure does not exclude another measure. Choosing one does not imply the exclusion of the other. On the opposite, it is possible to have the four management measures implemented in the GIS environment at the same time. Therefore, there is a priority suggestion in the conceptual model (А, В, C, and D).

Spatial Modelling: Management Measures Formulation

The spatial variability identified in all data makes it possible to convert the data into thematic layers. Map algebra models represent all decision-making processes, in addition to multicriteria decision analysis (MCDA) equations, weightage allocations, and spatial inferences using Boolean Logic, Fuzzy Sets, and site-suitability analysis.

Site-suitability analysis for the A.I sub-model (aquifer depth)

Figure 8.5 Site-suitability analysis for the A.I sub-model (aquifer depth): Linear function.

In the initial diagnosis of the area, the model proposes a ranking based on flooding susceptibility (A). This management measure is divided into four sub-models (A.l,

A.2, A.3, and A.4):

A.l. For occupied lots: (i) Lots occupied in already paved streets have a lower risk of flooding than others; (ii) Lots in very shallow areas (aquifers), with low slope values and very close to existing water bodies (including artificial ones such as the channels), have a higher risk of flooding.

A.2. For “islands” evaluation: Lots, occupied or empty, surrounded by very “shallow” areas (information provided by experts).

A.3. For legal restrictions: permanent preservation areas: (i) The Jaguaribe River (even the channelled part) and seasonal lakes are classified as water bodies that are 10 to 50 m wide (river width). So, the applicable law establishes as permanent preservation a 50 m marginal strip (riparian zones); (ii) The other streams are considered water bodies with less than 10 m width. The current legislation establishes as permanent preservation a marginal strip of 30 m.

A.4. For new occupations: (i) No way to occupy permanent preservation areas (making it a restriction criterion); (ii) Areas further from existing water bodies or steams, less shallow than others, with higher slope values and also far from the already identified areas in A.L, and close to a paved road (impervious and possible better access to the entrance) will be the best lots for occupation (must have a higher economic value).

Each sub-model implies some site-suitability analysis, and most of the GIS provides Fuzzy membership functions. Fuzzy Sets (Zadeh, 1965) have no definite boundaries. The probability of an element to belong to one set or class is gradual. The GIS uses Fuzzy Sets for site-suitability analysis. One pixel (or location) can be classified according to its suitability to a given hypothesis or analysis. In this application, a linear function (positive or negative) defines, on a scale from 0 to 100, how suitable one value is for the analysis goal. Figure 8.5 shows a sample for the aquifer depth analysis. The sub-model A.l considers that as much shallow the areas are as much they will be susceptible to flooding. Then, Figure 8.5 depicts a graphic representation of a monotonic decreasing linear function. In a spatial analysis approach, it is a simple site-suitability function applied to each pixel in the thematic layer.

The В management measure (Table 8.1) aims to induce an aquifer depletion to prevent overflow and to help in surface drainage in rainy seasons. It is divided into seven sub-models (B.l, B.2, В.З, B.4, B.5, B.6, and B.7). GIS visualisation and analysis tools supported most of the expert inferences.

B.l. Simulation of possible wells in the following inferred locations and assumptions: (i) Buildings with more than 20 households could use groundwater for non-potable uses (car and sidewalk washing, external cleaning, and garden irrigation); besides, the experts assure that there are already wells in some of those buildings (data missing from official records); (ii) Some churches in the neighbourhood already have wells (absent from official records); (iii) Parks (public green sites) could be water supply points for water tank trucks that regularly water the main avenues (plant beds); these trucks currently bring water from other places for this purpose; (iv) Car washing could save money and switch to groundwater pumping; (v) There are many “mini-farms” and luxurious mansions with huge garden areas; irrigation could use groundwater; also, experts state that some of these houses already have wells (missing from official records); (vi) Some of the largest empty lots may turn out to be public green spaces and therefore be a groundwater supply point for itself; (vii) Shallow areas are the same as already calculated in the previous sub-models (A.l).

B.2. Simulation of new wells in areas where the aquifer level is shallower in the driest period (less than 3 m): Consider previous spatial data provided by the “A” measure. B.3. Simulation of possible wells at the most frequent outcrop sites: (i) Adding a homogeneous recharge at the aquifers (adding the same rate level for all pixels) in the observed dry period; if the outcrops happen during the dry season, they will most likely happen in rainy periods; (ii) Repeat the procedure for different recharge values (100-400 mm) and check for new outcrop points or if it is spreading out. B.4. Near the identified outcrop points, locate possible wells.

B.5. Induce groundwater uses during the dry season, causing enough aquifer depletion to be fed by the natural recharge during the rainy season: the average value adopted was 10 m3/day, considering specific uses such as car washing, sidewalk washing, garden irrigation, and water tank truck supply.

B.6. Modifying the flow rates by use or location: (i) 20 nrvday for wells located in parks (existing and simulated); (ii) 20 m3/day in outcrop sites, and (iii) 10 m3/day in other simulated wells.

B. 7. Saltwater threshold: (i) Observe when the aquifer level is very low near the coastline; (ii) Analyse using streamlines in a groundwater flow model; (iii) Remove some wells to avoid saltwater intrusion and rerun the model; (iv) Find an equilibrium situation where the depletion is enough to support the surface drainage and does not cause saltwater intrusion.

In the C management measure, the idea was to use the current infrastructure (surface drainage channels) in the simulations. Then, new locations for wells are proposed surrounding the channels to improve their draining function.

C. l. Simulation of the existence of wells in the locations indicated by the A measure

considering the proximity of the channels (200 m).

C.2. Repeat the previous simulation using the saltwater intrusion as a threshold.

In the D management measure, there is a water sensitive land-use planning issue: ensuring the minimum recharge areas for aquifer balance and creating, at the same time, green areas in an urban space.

D.l. Identify the recharge areas pointed by the modelling in the A measure;

D.2. Identify the most suitable areas for aquifer recharge based on physical data such as soil permeability, slope, and precipitation.

D.3. Simulation of alternative recharge areas (land-use permits can make it possible): permanent green areas.

Spatial Modelling: Measures and Strategy Generation

Once spatial modelling is completed, the implementation phase is all about executing all the commands and steps represented in the model to obtain the expected results. However, often during GIS processing, some modelling flaws can be identified, which implies a return to the model for some review. Figure 8.6 is a graphic representation of the whole spatial modelling.

At this stage, there is, for the first time, the “spatial knowledge modelled”. There are layers with information from both sources: experts’ knowledge and collected data (ground-truth, official records, previous research, bibliography, and legislation permits). It is a blended approach: data-driven and knowledge-driven. Figure 8.7 shows the final maps as a result of each management measure proposed. All maps were presented to the authorities and experts (stakeholders) providing guidelines for new land- use laws and regulations in a water-sensitive way.

 
Source
< Prev   CONTENTS   Source   Next >