We consider here the case study on housing development (HD) for territory of Bryansk region, Russia, as an example.
The description of this case study, approaches to its solution, and some results for the specific region and stakeholders'
interests and preferences are presented below.
General description of the Case Study (CSTr-1)
A group of stakeholders has a goal to find an area for HD in a given region. Stakeholders have an interest to find the
suitable area for HD in Novozybkovski administrative district in the Bryansk region of Russia. This district (about 50
km from south to north, and about 50 km from east to west) is mainly agricultural with more than 80 thousand residents. One problem that they may face is that the indicated district is situated in an area subjected to radioactive contamination as a result of the Chernobyl accident.
Invited experts have information about this district, including map of land-use and additional cartographic layers
(hydrological and road nets). We consider two possible approaches to investigation of the indicated problem.
The first approach (case study, CSTr-1.1) is based on implementation of the DECERNS
SDSS to realize the conjunctive screening process using only GIS functionality in DECERNS.
The second approach (CSTr-1.2) includes implementation of both GIS and
MCDA tools. Suitable lands as a result of the screening process form several alternative sites for land use, which are
analyzed using MCDA tools.
CSTr-1.1: Implementation of GIS functionality within housing development problem
The following criteria for HD were discussed and agreed by the facilitator and experts with the group of stakeholders:
- C1 - proximity to the settlements;
- C2 - proximity to major roads;
- C3 - proximity to major rivers or lakes/ponds;
- C4 - distance from rail roads.
Within this case study the conception of the criteria and constraints were specified:
- C1: towns/villages with the population more then 500 people
(i.e., not very small settlements) are considered, and (imposed constraint) 1≤ C1 ≤ 4 km;
- C2: roads with only blacktop are considered, and
0.08 ≤ C2 ≤ 2 km;
- C3: only large rivers and lakes/ponds are considered (the details of rivers/
lakes were discussed with the stakeholders), and 0.08 ≤ C3 ≤1 km;
- C4: distance from railroads, 0.2≤ C4 ≤ 3 km.
Additional constraints required that houses may not be built:
- on arable lands as well as on water-meadow pastures and hayfields (CN1),
and on areas of the given villages (CN2);
- on other unsuitable lands defined at the stage of constraint imposition (ravines,
farm lands, sandstones, burial grounds and some other types of land use) (CN3).
CSTr-1.1 solution
Our solution of the CSTr-1.1 is based on implementation of vector maps for the region under consideration.

Fig.1 Vector map of landuse for the district under consideration
The conjunctive screening process within this case study is based on realization of the following steps.
1.1 Preparing the sets of map objects/layers (settlements, rivers, roads,...) in accordance with the screening
criteria (for this, attributive data of the maps downloaded within the case study are used).
In this case study we use the following main map layers:
- Landuse layer; it contains data on land use polygons, including settlements, arable
lands, pastures and hayfields, forests, various 'dangerous' objects, etc.
(see the Polygon types);
- Rivers, roads, and rail roads layers
(see the River and Roads types).
Lets demonstrate one example, how the layer with the selected settlements can be created;
(read User Guide (UG), sections 2.4.1, 2.4.2.1-2.4.2.3).
For this we activate Landuse tab (in the right panel with the list of layers):
- corresponding tick should be added (next click takes off the tick),
- one click inside of the corresponding tab (then, may be with small delay,
depending on the Internet speed, this tab is getting grey; this means that the layer is active).
Double click of the tab invokes layer editor, see this option, UG-2.4.2.3.
Let's create the layer with the settlements in accordance with C1 criterion. For this we
click the Searching tab (below the map and map scale; see Searching functions,
UG-2.4.2.1). We see the list with attributive data for the activated layer (Fig.2),
we choose and double click TYPE string (TYPE Integer [1-40] in our case); we see TYPE
is appeared in the (SQL) searching line;
then we use the following clicks: = and type 1 (1 is the identifier for settlements,
see Polygon types);
then we click the string POPUL (in the list of attributes),
and then click the sign > (or ≥) and type 500 (the amount of persons in the
settlements according to the criterion C1;
then we click searching tab, and we obtain the result of searching: highlighted
objects/settlements.
Comment 1: In principle, the selection of big settlements may be organized using other approaches, eg, we could point
out/select several settlements and declare them as the selected settlements to be used with the C1 criterion; read
UG-2.4.2.1 Selection functions).

Fig.2 Selection of the settlements in the district with population >500 persons
1.2 Then, we create buffer zones (read the item buffering, UG-2.4.2.2) around the selected
objects, or, before that, we store these selected objects as a new layer: click Analysis tab and then click Cut selection, and
type the name of this new layer; then that layer will appears in the list of used layers (you can then create the appropriate
legend for this layer, using attributive info, read UG-2.4.2.3 legend editor).
Then, the buffer layer BL1=BL(r1<C1<R1) is created, using the function Buffering, UG-2.4.2.2.
The indicated steps are repeated for "implementation of other criterion requirements" (C2-C4).
As a result, we have buffer layers BLi=BL(ri<Ci<Ri), i=1,...,4.
1.3 Then, we use Overlay function(s), and find intersections of created buffer zones (i.e., we realize a conjunctive (&) screening):
L1=C&(1-4)=((BL4∩BL3)∩BL2)∩BL1
To see the result of each operation, click the tab Refresh (when creating overlays, computation are curried out in the asynchronous
mode, that means clients may use other operations (eg, layer editing, etc.) when overlays are computed.)
General recommendations for implementation of overlay operations:
Start with the layers, which lead to minimization of time and number of operations. Eg, BL4 is a buffer layer with a simple geometry, layers BL2 and BL1 are much more complex; thus, the indicated above sequence of intersections is more appropriate then, eg, ((BL1∩BL2)∩BL3)∩BL4
Start with the layers, which lead to minimization of time and number of operations. Eg, BL4 is a buffer layer with a simple geometry, layers BL2 and BL1 are much more complex; thus, the indicated above sequence of intersections is more appropriate then, eg, ((BL1∩BL2)∩BL3)∩BL4
Comment 2: If we should take into account a criterion constraints Ci> Ri (eg, for i=4),
we create buffer layer BL4=BL(0<C4<R4), and then, using results of screening for
criteria, C&(1-3), we create a layer C&(1-4)= C&(1-3)-BL4 using the subtraction function.

Fig.3 CSTr-1.1: Buffer zones for the selected settlements
1.4 The layer L2 is created taking into account the additional constraints CNi i=1-3, (see above),
then subtraction of constraint layers from layer L1 is implemented:
L2 =Lresult = L1-(∪L(CNi));
layers CNi are created using tables with data on map objects, searching operations, and cut selection operation; then we
use overlay operation to join these layers.
Thus, layer L= L(C(&; i=1-4) - CN(1-3)) is the resulting layer, which comprises all the lands suitable for HD within the
indicated problem, Fig.4.

Fig.4 CSTr-1.1: Suitable area for housing development (polygons), along with other layers
(roads, rail roads, and rivers), important for subsequent decision-making
Comment 3: To have (useful) attributive information for buffer polygons, we may, eg, intersect buffer layer Lresult,
with Land-use layer with splitting multi-polygons and inheritance of the attributive info using checkmarks for necessary
attributes in overlay window (in our case it is Cs137_2 (or Cs_99); we receive the resulting layer Lres+Attrib
(see the list of layers within this case study).
1.5 After a screening process (steps 1-4), the next step within the decision making is for the stakeholders to analyze
the resulting layer L=Lresult /Lres+Attrib using additional objective/ subjective criteria and preferences.
Then, eg, the stakeholders will select one or several of the most interesting lands for housing development.
In our case, experts agreed to select the most interesting/suitable polygons from each group of polygons of the layer
Lres+Attrib, based on analysis of different layers, including river, road, forest and other layers; these polygons form
the layer A(1-4), Fig.5.
The further step within the decision-making process may also include a new request by the stakeholders to continue evaluating
the suitable lands based on some additional criteria to find the 'optimal'/trade-off land(s) for HD.
Such an approach is used to form and analyze the case study CSTr-1.2.

Fig.5 CSTr-1.1/1.2: The resulting layer Lres+Attrib, additional layers (roads, rail roads, and rivers),
and the layer of selected polygons/alternatives A(1-4)
CSTr-1.2: Implementation of GIS and MCDA tools within housing development problem
Within the next iterations of the HD problem development and analysis, stakeholders extended the list of criteria and, working with a facilitator/analytic, agreed to use multicriteria methods for ranking alternatives.
The following criteria within the CSTr-1.2 are considered along with the constraints in addition to the indicated ones
for CS-1.1:
- C5 - level of radioactive contamination, Ci/km2;
(minimize) (C9≤20 Ci/km2);
- C6 - distance to hazardous objects (maximize)
(C6≥1 km);
within this case-study the following 'dangerous' objects are considered: stockyards and cattle-breeding farms, cattle burial ground,
sandstones and open pits, dumps, and cemeteries;
using searching functions we can create a layer with 'dangerous' objects.
- C7 - general (qualitative) assessment of the local landscape/site quality,
10 point scale, (maximize);
- C8 - cost (minimize), (land cost and all the expenses associated with
house building, depending on the site characteristics and its location).
CSTr-1.2 solution
2.1 Within the CSTr-1.2 we consider the criteria from CSTr-1.1: Ci, i=1,..., 4, the constraints CNk,
k=1,2,3, and the criteria Ci, i=5,...,8 indicated above.
Experts agreed with the conjunctive screening process for the criteria and constraints, implemented in the
CSTr-1.1. Stakeholders and experts consider the polygons of the layer A(1-4) as the alternatives Ai, i=1,...,4,
within the developed multicriteria problem.
2.2 Using attributive data for polygons Ai, distance function, and expert) assessments of expenses and
(on-site) evaluation of site quality, experts are ready to develop the multicriteria problem.
Within this case study the Top-down approach is used (see User Guide, sections 2.4.2.4.1,
and 2.4.2.4.2).
There are 2 possible approaches to form the set of alternatives in MCDA subsystem:
- without direct binding the alternatives with map objects (Indirect Binding
Alternatives, IBA);
- based on Direct Binding each alternative with the map object(s) (DBA)
(UG-2.4.2.4.3).
If we implement IBA, all the criterion values against each alternative are entered in the Performance Table (PT)
(see UG-2.4.3.1) manually with the use of expert judgments, GIS functions, model assessments,
etc.
To demonstrate spatial alternatives Ai on a map, we can present in this case the entire layer of alternatives
A(1-4); if we want to show the alternative/spatial object A1 (A2, and so on), we can create a layer
LAi with 1 alternative for each i=1,...,n
(for this we can use Split multi-polygon option when implementing overlay operation, see UG-2.4.2.2).
Within this case study we use DBA method. To realize binding spatial objects and alternatives (within the top-down approach)
read UG-2.4.2.4.1.
To start with MCDA tools read the section 2.4.3 MCDA module reference in the User Guide.
Right click on the TASK (root entry of Value Tree, VT, to be built) allows creating the criteria within multicriteria problem.
Criteria may be connected with map layers data (eg, attributive data with contamination of lands/polygons); these are so called map
based criteria.
Necessary information may not be found directly in map layers (eg., landscape quality, if detailed landscape maps/pictures are not in
the list of map layers and experts analyze such a criterion on-site;
or, eg, there is no vector layer with attributive data on land contamination, but there is a raster layer with contamination, and user
can add necessary info to the performance table as a result of a visual overlay of raster map and vector layer with investigated
objects/alternatives, Fig.6.

Fig.6 Raster layer of contamination and vector polygons
Within the CSTr-1.2 we use map based criterion C5 (radioactive contamination of the chosen spatial objects/alternatives)
(UG-2.4.2.4.3); as a result the criterion values for alternatives Ai,
C5(Ai), are automatically transferred to the Performance Table (PT).
The values of other criteria are inserted in the PT manually.
The criterion C6(Ai) is assessed through determination of the distance from the object Ai
to the closest object of the layer "Dangerous objects" Fig.7 (then these values are inserted manually).
The criterion values C7(Ai) and C8(Ai) are the result of
(on-site) expert evaluations.

Fig.7 Estimation of criterion C6(Ai) using distance function
Click MCDA tab to work with tools for multicriteria analysis (read UG-2.4.3).
Experts (the facilitator and experts) discuss the choice of one or several MCDA methods (UG-1.4)
for implementation within the case study under investigation.
Assigning method-specific parameters and functions (eg, value functions Vi(x) in MAVT, preference functions in
PROMETHEE, etc) and criterion weights are carried by experts in accordance with the existing recommendations [Refs: ], and
individual or group analysis of the specific multicriteria problem.
Comment 4: We recommend to start multicriteria analysis with MAVT (or TOPSIS) method. In this case the chosen 'mean
criterion values' and 'mean weights' will be a basis for subsequent analysis with the use of more sophisticated methods
(MAUT, ProMAA, F-MAVT). The basic characteristics of created within MAUT/ProMAA/F-MAVT distributed (random/fuzzy) values
are reflected in the PT (and some other internal forms) and can be compared with non-distributed/'mean' values used within
MAVT/TOPSIS/ PROMETHEE/[AHP].
Comment 5: Weight coefficients are associated/bound with each MCDA method and may differ (if it corresponds to experts
approach). Mean values of criteria for MAVT/TOPSIS/PROMETHEE are the same and are automatically transferred to other method
from this group being created in one of the method. The same approach has been realized for MAUT and ProMAA: the distribution
of criterion values assigned in one method may be used in another one; weight coefficient in MAUT are inherited from MAVT, but
may then be changed and differ from MAVT weights.

Fig.8 The performance table developed within the CSTr-1.2
After setting all the method-specific parameters [and weights] we can analyze the output results and implement uncertainty analysis; i.e., using the output forms and additional tools users can:
- analyze and compare the results of ranking alternatives for each method;
- implement weight sensitivity analysis for MAVT, TOPSIS, PROMETHEE, AHP, and MAUT;
- implement Value Function sensitivity analysis for MAVT, MAUT, ProMAA, and F-MAVT;
- in addition, user can develop a new scenario within the problem under investigation,
saving it in a new project, using other criterion data, distributions, weights, and model specific parameters (realization of
some steps within scenario planning approaches).
(see UG-2.4.3.2, 2.4.3.3; other sections with description of MCDA methods,
tools, and approaches to uncertainty analysis will be added to UG).

Fig.9 Results of ranking alternatives with the use of different MCDA methods
Based on the analysis of the output ranking and weight (and Value/Utility function) sensitivity analysis, stakeholders agreed
with experts recommendations to consider the alternative A1 as a trade-off one within this multicriteria problem.