water

Article The Application of Multi-Criteria Analysis Methods for the Determination of Priorities in the Implementation of Irrigation Plans

Barbara Karleuša 1,* , Andreja Hajdinger 2 and Lidija Tadi´c 2 1 Faculty of Civil Engineering, University of Rijeka, Rijeka 51 000, Croatia 2 Faculty of Civil Engineering and Architecture, Josip Juraj Strossmayer University of Osijek, Osijek 31 000, Croatia; [email protected] (A.H.); [email protected] (L.T.) * Correspondence: [email protected]; Tel.: +385-51-265-935



Received: 27 December 2018; Accepted: 5 March 2019; Published: 10 March 2019


Abstract: Irrigated agriculture has considerable impacts on the environment. To minimize negative effects and maximize positive effects, it is necessary to provide comprehensive analyses beyond the strictly technical domain. In this study, we apply a methodology for determining priorities in implementing irrigation plans using multi-criteria analysis methods on a specific case study area in the sub-catchment area of the Orljava River in Požega–Slavonia County, Croatia. Five potential irrigation areas (Orljava–Londža, Pleternica, Ovˇcare, Treštanovci, and Venje–Hrnjevac) were analyzed according to five selected criteria: environmental protection, water-related (four sub-criteria), social, economic, and time criteria with different criteria importance (weight). The aim of this study was to confirm the adequacy of using six multi-criteria analysis (MCA) methods (mostly used: PROMETHEE, AHP, ELECTRE TRI, and the less used: DEXi, PRIME, and PCA) in determining priorities for fulfilling irrigation plans, present models for preparation of the input data, apply certain methods, and compare the results on the selected case study area. The methods’ adequacy was confirmed during the research. Five of the six MCA methods identified the Ovˇcare area as the most appropriate for irrigation development (i.e., it has priority in implementing the irrigation plan). According to one (AHP) of the six methods, Orljava–Londža has more advantages over other areas. All MCA methods, except PCA, chose Venje–Hrnjevac as the least advisable (last to be implemented) alternative. Conclusions from this research confirm findings from recently published research regarding the application of MCA on water management problems.

Keywords: multi-criteria analysis; priority; irrigation plans implementation

1. Introduction Irrigated agricultural land comprises more than 250 million hectares all over the world, and is essential for ensuring food security and stimulating rural socio-economic development [1]. However, this land’s impact on water resources, soil, and the environment in general is unquestionable. Implementation of irrigation systems is a considerable investment involving operation and maintenance costs. Therefore, improving the performance of irrigation systems by increasing the efficiency and transparency of those systems is essential. Making decisions regarding the implementation of water management plans as irrigation system plans is a complex task due to the multiple objectives that must be satisfied, different and numerous criteria (economic, social, and environmental) and different measures (quantitative and qualitative) are used for objective fulfillment assessment with the involvement of multiple stakeholders [2,3]. Ranking alternatives to be implemented by priority, and in some cases selecting the first alternative to be implemented as a pilot project, represents this kind of decision-making problem [4,5].

Water 2019, 11, 501; doi:10.3390/w11030501 www.mdpi.com/journal/water Water 2019, 11, 501 2 of 20

The use of the traditional cost-benefit method, where only the quantified direct cost and benefits are incorporated in the analyses, is not sufficient when the decision involves consideration of variables that cannot be easily quantified into monetary units or if sufficient information for monetarization of all variables and criteria is not available [6]. In these situations, there are numerous procedures, classified as multi-criteria decision analysis methods (MCDM), which provide support to the complex decision-making process [7–9]. Multi-criteria analysis (MCA) provides a systematic methodology to integrate heterogeneous and uncertain information with cost-benefit information and stakeholders views in an understandable framework to rank project alternatives [10]. MCA is highly useful as a tool for project evaluation during the developing phase when decision makers do not have sufficient knowledge regarding details, but the importance of making the right decision is considerable. MCA has been used for analyses of various types of water management problems [11], for ranking and selection of: water management strategies [12,13], alternatives of water supply systems [14,15], desalination procedures for drinking water production [16], reservoir use alternatives [17], wastewater disposal locations [18], urban stormwater drainage management alternatives [19], locations for hydropower plants and dams [20], and alternatives for irrigation [4,5,21–24]. ELimination and (Et) Choice Translating Reality (ELECTRE TRI) method was applied for minimizing flood damages and minimizing damages in agricultural areas caused by deficits in water supply in Iran [25]. For effective management of irrigation networks in farms, based on the example of ranking and selecting check and intake structures for irrigation canals, in Iran, the Technique for Order of Preference by Similarity to Ideal Solution (TOPSIS) method was applied [26]. Often, multi-criteria methods are applied with GIS (Geographic Information System), as in site selection for certain types of fruit cultivation where multi-criteria analyses overlay method and Analytic Hierarchy Process (AHP) method with GIS were analyzed in Antalya, Turkey [27]. Evaluation of potential irrigation expansion using spatial fuzzy multi-criteria decision framework, which includes AHP in ARCGIS environment, was analyzed in the limestone coast region of South Australia [28]. MCA methods, such as Preference Ranking Organization METHod for Enrichment Evaluations (PROMETHEE), ELECTRE TRI, AHP, and Decision EXpert (DEXi), have been used to provide support decision making for defining priorities in the implementation of agricultural irrigation plans in different parts of Croatia [4,5,23]. A review of irrigation system performance evaluation using fuzzy set theory, direct measurement for indicators, AHP, and remote sensing was performed by Elshaikh et al. [29]. An overview of multi-criteria decision analysis in environmental sciences [10] shows that in water management, during the period 2000–2009 the most used methods were: Multi-Attribute Utility Theory (MAUT)/Multi-Attribute Value Theory (MAVT), AHP/Analytic Network Process (ANP), ELECTRE, and PROMETHEE. This is confirmed by more recent research [15,30–32]. If multiple methods are applied on the same decision making problem, then the methods that are used beside already mentioned are mostly: TOPSIS, Compromise Programming (CP), Simple product weighting, and Weighted average [10,31]. This led to the selection of six MCA methods: the three most used are PROMETHEE, AHP, and ELECTRE TRI; and three not so often used on this kind of problems are DEXi, Preference Ratios In Multi-attribute Evaluation (PRIME), and Principal Component Analysis (PCA), to be applied and compared on the specific case study in Croatia for defining priorities in implementation of the Irrigation plan of Požega–Slavonia County. The aims of this study are: confirm the adequacy of using all the selected methods, especially DEXi, PRIME, and PCA, when determining priorities in fulfilling irrigation plans on the Požega–Slavonia County irrigation plan, present the model for preparation of the input data for selected MCA methods application, use the selected MCA methods on the specific case study and compare the results. WaterWater2019 2019, 11, 11 501, x FOR PEER REVIEW 3 of8 20of 20

2. Materials and Methods 2. Materials and Methods 2.1. Description of Study Area 2.1. Description of Study Area Priorities in implementation of irrigation plans are defined on the case study area located in PožegaPriorities–Slavonia in implementation County (north of-eastern irrigation part plansof Croatia, are defined Figure on1), thein the case sub study-catchment area located area of in the Požega–SlavoniaOrljava River that County belongs (north-eastern to the Sava River part of Basin Croatia, [23,3 Figure3]. 1), in the sub-catchment area of the Orljava River that belongs to the Sava River Basin [23,33].

Figure 1. Study area: Požega–Slavonia County [23,33]. Figure 1. Study area: Požega–Slavonia County [23,33]. According to the spatial plan of the Požega–Slavonia County [34], agricultural land occupies the largest shareAccording in the to county. the spatial In this plan area, of wheat,the Požega maize,–Slavonia sugar beet, County tobacco, [34], grape,agricultural fruits, land and vegetablesoccupies the arelargest grown, share and agricultural in the county. land canIn this be divided area, wheat, into three maize, categories: sugar especiallybeet, tobac valuableco, grape, soil, fruits, valuable and soil,vegetables and other are cultivable grown, and soil. agricultural land can be divided into three categories: especially valuable soil,Considering valuable soil, the and cultivability other cultivable of agricultural soil. areas, it was found that of the total area of 894.9 km2, the cultivatedConsidering area covers the cultivability 782.9 km2 ofor agricultural 87.48% [33– 35areas,]. it was found that of the total area of 894.9 km2, theAccording cultivated to area data covers from 782.9 2003 [km36],2 or only 87.48% 1085.6 [3 ha3–3 were5]. irrigated in the Požega–Slavonia County, representingAccording only 2.55% to data of from its total 2003 area [3 suitable6], only for1085 irrigation..6 ha were This irrigated percentage in the is Požega higher than–Slavonia the average County, inrepresenting Croatia, which only amounted 2.55% of to its only total 0.86%, area but suitable it was stillfor irrigation. insufficient This to realize percent itsage own is potential.higher than the averageTo intensify in Croatia, the agricultural which amounted production, to only enable 0.86% development, but it was in still the insufficient area, raise the to standardrealize its and own qualitypotential. of life, a significant contribution can be expected if irrigation is applied to agricultural areas. AnalysesTo carriedintensify out the in agricultural the Požega production, area show thatenable the development total surface in where the area, irrigation raise the is requiredstandard isand 40,772quality hectares of life of, a grosssignif (29,327icant contribution ha net). can be expected if irrigation is applied to agricultural areas.

Water 2019, 11, x FOR PEER REVIEW 9 of 20

Analyses carried out in the Požega area show that the total surface where irrigation is required is 40,772 hectares of gross (29,327 ha net). These analyses have shown that the need for irrigation exists as a supplementary measure for the improvement of agricultural production. The amount of water from 1500 to 2500 m3/ha per year Water 2019, 11, 501 4 of 20 can meet the water needs of all crops. Areas that could potentially be irrigated are located in the central part of the Orljava River basin and amount to 29,237 ha. The entireThese study analyses area have is characterized shown that the by needsmall for quantities irrigation of existsgroundwater as a supplementary and high potential measure for for the buildingimprovement multipurpose of agricultural reservoirs production.and micro-reservoirs The amount; therefore, of water w fromater 1500for ir torigation, 2500 m 3~/ha77 mil. per yearm3, can can bemeet provided the water by reservoirs needs of allof crops.various Areas sizes that[34]. could potentially be irrigated are located in the central Thepart planning of the Orljava of irrigation River basin is based and amounton selecting to 29,237 water ha. resources that will provide the required quantitiesThe and entire can be study divided area into is characterized three phases. by The small first quantities phase involves of groundwater using exis andting high reservoirs, potential for constructionbuilding of multipurpose micro-reservoirs reservoirs, and preparation and micro-reservoirs; of pilot irrigation therefore, areas. water The for second irrigation, phase ~77 is the mil.m 3, developmentcan be provided of small byand reservoirs medium- ofsized various systems sizes based [34]. on the first stage knowledge and preparation for the performanceThe planning of larger of irrigation systems, is wh basedere coordination on selecting wateris needed resources in terms that of will aligning provide plans the with required otherquantities relevant factors. and can The be third divided phase into is threethe performance phases. The of first medium phase and involves large systems using existing [33]. reservoirs, construction of micro-reservoirs, and preparation of pilot irrigation areas. The second phase is the 2.2. Alternativesdevelopment and of Criteria small and medium-sized systems based on the first stage knowledge and preparation Plannedfor the performance irrigation areas of larger in the systems, 1st phase where represent coordination only a is small needed part in termsof the ofarea aligning of Požega plans– with Slavoniaother County relevant that factors. can be The irrigated. third phase Planned is the irrigation performance areas of medium(alternatives) and largein the systems 1st phase [33]. are Orljava–Londža (1), Pleternica (2), Ovčare (3), Treštanovci (4), and Venje–Hrnjevac (5) (Figure 1). 2.2. Alternatives and Criteria Based on analyses done in the study “Basics for irrigation in Požega–Slavonia County” [34], the most importantPlanned characteristics irrigation areas of inall the areas 1st phase(alternatives) represent were only aextracted small part by of authors the area for of Požega–Slavoniathis article researchCounty and thatare presented can be irrigated. in Figure Planned 2 and irrigation Table 1 [23] areas. (alternatives) in the 1st phase are Orljava–Londža Criteria(1), Pleternica development (2), Ovˇcare was (3), a ver Treštanovciy sensitive (4), process. and Venje–Hrnjevac To achieve sustainable (5) (Figure 1irrigation). systems, beside theBased criteria on that analyses include done natural in theand study technical “Basics potentials, for irrigation environmental, in Požega–Slavonia social, and economic County” [34], criteriathe were most selected. important Finally, characteristics authors defined of all areas five (alternatives) groups of criteria: were extracted “environmental by authors protection”, for this article “waterresearch-related”, and “social”, are presented “economic” in Figure, and2 and“time” Table criteria1[23]. (Table 2) [23].

TIME TO COMPLETE THE SYSTEM (years): RECREATION: • 3 to 5 • no • 3 • yes • 1 to 2

SOCIAL TIME CRITERION CRITERION

TYPE OF AREA: • drinking water protection zone ENVIRONMENTAL TOTAL COST PER (DWPZ) ECONOMIC • protected area (nature PROTECTION SELECTION OF THE IRRIGATION AREA AGRICULTURAL AREA: CRITERION • exact value (kn/ha) park, national park) CRITERION • does not belong to any of the specified

WATER RELATED CRITERION

DRAINAGE: CHANNELS: AREA TO BE STATUS OF • not existing • doesn't exist at all IRRIGATED: RESERVOIR • partially built • partially built • exact value (ha) CONSTRUCTION: • completely built/exist • planned • in construction • existing Figure 2. Selected criteria for MCA (Multi-Criteria Analysis). Figure 2. Selected criteria for MCA (Multi-Criteria Analysis). Criteria development was a very sensitive process. To achieve sustainable irrigation systems, beside the criteria that include natural and technical potentials, environmental, social, and economic criteria were selected. Finally, authors defined five groups of criteria: “environmental protection”, “water-related”, “social”, “economic”, and “time” criteria (Table2)[23]. The “environmental protection” criterion (e.g., “environment”) evaluates whether the agricultural area is in the drinking water protection zone—DWPZ (worst case)—other protected areas, e.g., nature park or national park, or does not belong to any of the specified (best case). The weight factor/importance for the “environmental protection” criterion is 20%. This information is qualitative but was easy to obtain via a comparison of available maps.

Water 2019, 11, 501 5 of 20

Table 1. Characteristics of all areas (alternatives) [23].

Alternative: Orljava–Londža Pleternica Ovˇcara Treštanovci Venje–Hrnjevac Investor/User Kutjevo Ltd. Hrvatski duhani Ltd. Kutjevo Ltd. Grbi´cLtd. Winery Enjingi Protection Area DWPZ DWPZ No No Nature Park Ownership Type Rent Management Rent Private, rent Private, rent Drainage Partially None Partially None None Channels Exists Partially Exists None None State of Drainage and Channel Network Existing Soy, lucerne and maize Tobacco Vine grafts, maize and Seed crops Vineyard and vine grafts sugar beet Planned Maize, sugar beet, soy, Tobacco Vine grafts, maize and Seed production Vineyards rapeseed sugar beet, industrial and

Agricultural Production seed crops Area (ha) 2500 300 200 150 80 Reservoir Londža—in construction Londža—in construction Kuštrevac—existing Kaptolka, Saraˇcevacand Kljunovac—planned Venjski—planned Possibility to Use Reservoir for Recreation Yes Yes No No No Water Demand (mil. m3/year) 6.250 0.75 0.5 0.375 0.2 Time to Complete the System (years) 3 to 5 3 to 5 1 to 2 3 3 Cost of Project Documentation Preparation 1.5 0.35 0.28 0.15 0.2 (mil. kn) Investment (mil. kn) 37.5 9 7 3.75 5.1 Total Cost per Agricultural Area (kn/ha) 15,600 31,167 36,400 26,000 66,250 Water 2019, 11, 501 6 of 20

Table 2. Characteristics of all alternatives sorted by selected criteria [23].

Criteria Water Related Time Social Economic Environmental Protection (Type) Drainage Channels Recreation Status of Reservoir Construction Area to Be Irrigated (ha) Total Cost per Agricultural Area (kn/ha) Alternatives Time to Complete the System (Years) Orljava–Londža DWPZ Partially Exists 2500 in construction Yes 15,600 3 to 5 Pleternica DWPZ None Partially 300 in construction Yes 31,167 3 to 5 Ovˇcare No Partially Exists 200 existing No 36,400 1 to 2 Treštanovci No None None 150 planned No 26,000 3 Venje–Hrnjevac Nature Park None None 80 planned No 66,250 3

The “water-related” criterion comprises four sub-criteria. The first is “drainage”. This sub-criterion evaluates whether the drainage from agricultural areas is partially built (best case), or non-existent (worst case). The weight factor for sub-criterion “drainage” (subsurface drainage) is 5%. The second sub-criterion is “channel” (surface drainage) and is evaluated depending on whether the channel networks are completely built/exist (best case), partially built, or do not to exist at all (worst case). The weight factor for “channel” network sub-criterion is 5%. Both of these sub-criteria assessments are qualitative and expert-opinion-based. The third sub-criterion is related to the size of the “area to be irrigated”. The larger the area is, the better the alternative. The weight factor for sub-criteria of the irrigated area is 10%. Assessment of irrigation areas has been done in detail in the plan [33], so this information enables the use of exact quantitative values. If a qualitative assessment is used, then the following principle is applied: three classes of areas are defined 0–149 (worst alternatives), 150–449, and over 450 ha (best alternatives). The fourth sub-criterion is related to “status of reservoir construction” ranking for an already existing reservoir that could be used for irrigation (best case), a reservoir for which construction has already started and a reservoir for which construction is only planned (worst case). This sub-criterion is qualitative. The total weight factor for this last sub-criterion is 10%. The total weight factor for the “water-related” criterion is 30%. The “social” criterion is related to the possibility of using reservoirs for recreation, i.e., whether reservoirs can be used for recreation (best case) or not (worst case). This sub-criterion is qualitative and expert-opinion-based. The weight factor for the “social” criteria is 5%. The “economic” criterion covers the total cost of the irrigation systems (the cost of designing the project and all needed documentation as well as the cost of building the complete system, in Croatian currency Kuna i.e., kn) divided by the size of the irrigated area. Assessment of irrigation areas has been done in detail in the plan; therefore, this information enables the use of exact quantitative values. If a qualitative assessment is done, the alternatives are sorted in three classes: less than 2,500,000 (best case); 2,500,000 to 5,500,000, and over 5,500,000 kn/ha (worst case). The weight for the “economic” criterion is 30%. The “time” criterion describes the time needed to build a system, which can be 1–2 years (best case), 2–3 years, and 3–5 years (worst case). This is a qualitative sub-criterion and expert-opinion-based. The weight factor for the “time” criterion is 15%. Although there are different techniques that can be used for elicitation of weight factors [37] in this paper weight factors were determined directly by authors as experts, considering awareness of the local population for environmental protection and its value, economic development and value of money, National Irrigation Strategy and County Irrigation Plan. The above explained weight factors represent Scenario 1 (Table3). For sensitivity analysis three additional weight factors scenarios (2, 3, and 4) were applied (Table3). Water 2019, 11, 501 7 of 20

Table 3. Scenarios for criteria weighting factors.

Water Related

Criteria Environmental Social Economic Time Area Drainage Channels Reservoir Status

Scenario 1 20% 5% 5% 10% 10% 5% 30% 15% Scenario 2 10% 5% 5% 20% 20% 5% 20% 15% Scenario 3 10% 10% 5% 10% 5% 5% 40% 15% Scenario 4 20% 5% 5% 10% 10% 10% 20% 20%

In comparison to Scenario 1: for Scenario 2 the “water related” criteria “area” and “reservoir status” were given greater importance (10% more), while the “environmental” and “economic” criteria were given less importance (10% less); for Scenario 3 the “water related” criterion “drainage” and “economic” criteria were given greater importance (5 and 10% more), while the “environmental” and “reservoir status” criteria were given less importance (10 and 5% less); and for Scenario 4 the “social” and “time” criteria were given greater importance (5% more), while the “economic” criterion was given less importance (10% less).

2.3. Multi-Criteria Analyses Methods In this study, six MCA methods were applied: PROMETHEE, AHP, ELECTRE TRI, DEXi, PRIME, and PCA, all of which are briefly described. Selected methods represent different MCA models. PROMETHEE and ELECTRE are outranking models, while AHP is a value measurement model based on pairwise comparison [31], DEXi combines “traditional” multi-attribute decision making with some elements of expert systems and machine learning [38], and PRIME incorporates the value tree analysis [39]. From ELECTRE family methods, the ELECTRE TRI as a sorting model was selected. PCA in many ways forms the basis for multivariate data analysis [40]. In this research, the main goal of PCA is to extract the most important characteristics of irrigation plans and compress the data set’s size by keeping only this important information. PROMETHEE are multi-criteria outranking methods used for partial (PROMEHEE I) or total pre-order of alternatives (PROMETHEE II) [41,42]. In PROMETHEE methods, the notion of criterion is extended by introducing the preference function, giving the preference of the decision maker for an alternative a as opposed to b, for each criterion. The preference value is between 0 and 1. The smaller the value, the higher the indifference of the decision maker, whereas the closer it is to 1, the higher his preference. In case of strict preference, the value of the preference function is 1. Six different types of criterion functions, for which the decision maker has to define a maximum of two parameters (indifference and preference thresholds), cover most of the cases that can happen in practice: usual criterion, quasi criterion, criterion with linear preference, level criterion, criterion with linear preference and indifference area, and Gaussian criterion. A valued outranking graph is considered by using a preference index. In this study, the PROMETHEE II method and usual criterion function were used. AHP is a priority method applicable to problems that can be represented by a hierarchical structure [43,44]. The top of the hierarchy is the goal, one level lower are criteria, and even lower sub-criteria. The lowest level is represented by alternatives. AHP method is conceived based on estimating relative priorities (weights) of criteria and alternatives on which pair-wise comparison criteria matrix and pair-wise comparison alternatives matrices (one matrix for each criterion) are generated. The matrices are normalized in order to calculate the priority weightings for criteria (the priority vector for criteria) and for alternatives on the basis of each criterion (the priority vector of alternatives for each criterion). Combining the criterion priorities and the priorities of each alternative relative to each criterion enables the development of an overall priority ranking of the alternative, which is termed the priority matrix. Water 2019, 11, 501 8 of 20

ELECTRE are multi-criteria optimization methods that enable the choice, ranking, and sorting of alternative problem solutions (depending of the ELECTRE version) considering decision-makers’ criteria and preferences [45,46]. The ELECTRE method was developed for a partial order in the set of solutions based on the decision maker’s preferences. A graph with nodes that represent possible solutions and a kernel that defines preferred solutions can be constructed. The ELECTRE method is useful in the case when criterion functions are weakly defined. The first developed method was ELECTRE I and on the basis of it, ELECTRE II, III, IV, and TRI followed. The ELECTRE TRI is a multi-criteria sorting method that assigns alternatives to predefined categories. The assignment of an alternative results from the comparison between the profiles that define the limits of the categories. The built outranking relation validates or invalidates the assertion of the alternative as to a predefined category. The ELECTRE TRI method gives the decision-maker a possibility to use pseudo-criteria with indifference and preference thresholds. The DEXi decision model is a qualitative decision model that divides the decision-making problem into less complex decision-making problems (sub-problems) [23,38]. Criteria are hierarchically organized and linked to the utility function that evaluates each criterion in relation to its goal in the hierarchy. Instead of numerical variables, DEXi uses qualitative variables, the values of which are usually presented in words. To display and evaluate the utility function, DEXi uses “if-then” decision-making rules. The utility function is defined throughout the hierarchy for each set of criteria, and the decision rule is described. The value of the aggregate criterion for each combination of input criteria is described and the relative importance of each criterion is expressed. Within each criterion, the ranking is performed depending on whether its value is good, neutral, or poor. PRIME is an analytic decision-making tool, where preferred data are given in intervals because the exact value is often not known, and it is easier to define a rough estimate [39,47]. In PRIME, the decision makers’ preferences are assumed to have an additive structure in which the total value of an alternative equals the sum of its attribute-specific scores. Determination of preferences can be score assessments, holistic comparisons, and weight assessments. PRIME uses the Swing Weighting Method, where the most important attribute is assigned with 100 points, and weights of other attributes are compared to the most important attribute in the interval from 0 to 100. PRIME has two styles to make weight assessments: bottom-up and top-down. In this study, the bottom-up weight assessment is used, where the decision maker compares sub-attribute weights of the main goal with each other. PCA is a multivariate technique that analyzes a data table in which observations are described by several inter-correlated quantitative dependent variables. The goals of PCA are to (a) extract the most important information from the data table, (b) compress the size of the data set by keeping only this important information, (c) simplify the description of the data set, and (d) analyze the structure of the observations and the variables [40]. PCA in many ways forms the basis for multivariate data analysis and it was first formulated in statistics by Pearson in 1901 [48]. There are many possibilities of using PCA; representation of spatial cause-effect data, in investigation of causal connections between spatial rainfall-runoff and rainfall-drought [49], as well as description of regional characteristics and differences among low-flow regimes at multiple hydrological stations, have been demonstrated [50].

3. Results and Discussion

3.1. PROMETHEE To apply the PROMETHEE method, data from Table2 were transformed into quantitative values for all criteria, on a scale of 1–3, where 1 is the lowest value of the assessment and 3 the highest, except for “social” criterion that had two values (1 and 2), “water-related”, “area” and “economic” criteria are already quantitative (Table4). Water 2019, 11, x FOR PEER REVIEW 14 of 20

To apply the PROMETHEE method, data from Table 2 were transformed into quantitative values for all criteria, on a scale of 1–3, where 1 is the lowest value of the assessment and 3 the highest, except for “social” criterion that had two values (1 and 2), “water-related”, “area” and “economic” criteria are already quantitative (Table 4).

Table 4. Quantitative assessment of alternatives by criteria. Water 2019, 11, 501 9 of 20

Water Related

Table 4. Quantitative assessment of alternatives bycriteria. Criteria

(Type) Criteria (Type) Water Related Time (min) (max) (max) (max) Social Area (max) Alternative (max) (max) (max) Status Status Economic Economic Drainage Channels Reservoir Environmental Economic (min) Time (max) Area (max) Social (max) Drainage (max) Channels (max) OrljavaAlternative–Londža 1 2 3 2500 Reservoir 2Status (max) 2 15,600 1 Environmental (max) Orljava–LondžaPleternica 1 11 2 2 3 2500300 22 22 15,60031,167 1 1 Pleternica 1 1 2 300 2 2 31,167 1 OvčareOvˇcare 3 32 2 3 3 200 200 33 11 36,40036,400 3 3 TreštanovciTreštanovci 3 31 1 1 1 150 150 11 11 26,00026,000 2 2 VenjeVenje–Hrnjevac–Hrnjevac 2 21 1 1 1 80 80 11 11 66,25066,250 2 2

The PROMETHEE II was applied using the usual preference function for all criteria. The results for criteria weights defined defined by S Scenariocenario 1 (Table 3 3)) areare presentedpresented inin FigureFigure3 a,3a, where where the the area area Ovˇcare Ovčare had the highesthighest rank/priority,rank/priority, followed by Orljava Orljava–Londža,–Londža, Treštanovci, and Pleternica. Figure 33bb shows thethe GeometricalGeometrical Analysis Analysis for for Interactive Interactive Decision Decision Aid Aid (GAIA) (GAIA) plane, plane, and and it can it becan observed be observed that, basedthat, based on the on direction the direction and length and length of the of decision the decision stick (red stick color), (red Ovˇcarecolor), Ovčare and then and Orljava–Londža then Orljava– shouldLondža haveshould priority have inpriority implementation in implementation of irrigation. of irrigation.

(a) (b)

FigureFigure 3. Results 3. Results of PROMETHEE of PROMETHEE II application II application for Scenario for Scenario 1: (a) 1:complete (a) complete ranking ranking scheme scheme of proposed of alternativesproposed, and alternatives, (b) GAIA ( andGeometrical (b) GAIA Analysis (Geometrical for Interactive Analysis Decision for Interactive Aid) plane Decision. Aid) plane.

Results forfor Scenario Scenario 2 ranked2 ranked alternatives alternatives Orljava–Londža Orljava–Londža and Ovˇcare and Ovčare in the first in the place, first followed place, byfollowed Pleternica by Pleternica and then Treštanovci; and then Treštanovci; for Scenario for 3 Orljava–Londža Scenario 3 Orljava was– rankedLondža in was the firstranked place, in followedthe first byplace, Ovˇcare, followed Treštanovci, by Ovčare, Pleternica; Treštanovci, for Scenario Pleternica; 4 Ovˇcare for was Scenario ranked 4 asOvčare first to was be implemented ranked as first followed to be byimplemented Orljava–Londža, followed Treštanovci, by Orljava Pleternica.–Londža, Venje–Hrnjevac Treštanovci, Pleternica. was the worst Venje alternative–Hrnjevac that was should the worst to be implementedalternative that last should in all scenarios.to be implemented last in all scenarios. 3.2. AHP 3.2. AHP The application of AHP was performed in two ways, the first using the direct assessment of each alternative by each criterion and each criterion in regard to the goal, and second by using pairwise comparison of criteria and alternatives. The direct assessment of each alternative by each criterion, and each criterion with regard to the goal, was based on data from Tables3 and4. Results for criteria weights defined by Scenario 1 (Table3) ranked the alternatives as per priority: the first alternative to be implemented was Orljava–Londža, the second alternative was Ovˇcare but with a very small difference, followed by Treštanovci, Pleternica, and Venje–Hrnjevac (Figure4). Water 2019, 11, x FOR PEER REVIEW 15 of 20

The application of AHP was performed in two ways, the first using the direct assessment of each alternative by each criterion and each criterion in regard to the goal, and second by using pairwise comparison of criteria and alternatives. The direct assessment of each alternative by each criterion, and each criterion with regard to the goal, was based on data from Tables 3 and 4. Results for criteria weights defined by Scenario 1 (Table 3) ranked the alternatives as per priority: the first alternative to be implemented was Orljava–Londža, theWater second2019, 11 , 501alternative was Ovčare but with a very small difference, followed by Treštanovci,10 of 20 Pleternica, and Venje–Hrnjevac (Figure 4).

(a) (b)

FigureFigure 4. AHP 4. AHP(Analytic (Analytic Hierarchy Hierarchy Process Process)) results results for direct for direct assessment assessment of each of eachalternative alternative by each by eachcriterion, and eachcriterion, criterion and with each regard criterion to the with goal regard for Scenario to the 1 goal: (a) forthe Scenariofinal ranking, 1: (a and) the (b final) graphical ranking, presentation and (b) of all alternativesgraphical’ presentationassessment by of each all alternatives’ group of criteria assessment. by each group of criteria.

ApplyingApplying criteria criteria weights weights defined defined by by S Scenarioscenarios 2 2 and and 3 3 resulted resulted in the same priority ranking as as forfor Scenario Scenario 1. 1. The The results results in in case case of of applying applying Scenario Scenario 4 4 criteria criteria weights weights differ differeded and and the the priority priority was was givengiven to to Ovčare Ovˇcare alternative, followed by Orljava Orljava–Londža.–Londža. PairwisePairwise comparison comparisonss ofof criteriacriteria withwith regardregard to to the the goal, goal, sub-criteria sub-criteria to to criteria criteria and and alternatives alternatives to toeach each criterion criterion according according to the to scalethe scale in Table in T5 ablewere 5 based were onbased data on from data Tables from2 andTable3, ands 2 and are presented3, and are presentedpartially in partially Table6 forin Table pair-wise 6 for pair comparison-wise comparison of all groups of all of groups criteria of to criteria the goal, to the all goal, water-related all water- relatedcriteria tocriteria the group to the criterion, group andcriterion, pairwise and comparison pairwise ofcomparison alternatives of to alternatives the criterion to “environment”, the criterion “environment”,as an example using as an criteria example weights using definedcriteria weights by Scenario defined 1. by Scenario 1.

TableTable 5. AHPAHP ( (AnalyticAnalytic Hierarchy Hierarchy Process Process)) pair pair-wise-wise comparison scale [4848].].

IntensityIntensity of Weight, of Weight Importance,, Preference Definition Definition Importance, Preference1 Equal importance (no preference) 1 3Equal Moderateimportance importance (no preference) (moderate preference) 3 5Moderate importance Strong importance (moderate (strong preference) preference) 7 Very strong importance (very strong preference) 5 9Strong importance Extreme importance (strong preference) (extreme preference) 7 2, 4, 6, 8Very strong importance (very Intermediate strong valuespreference) 9 Extreme importance (extreme preference) Table 6. Pair-wise2, 4, assessment 6, 8 of: all groups of criteria inIntermediate regard to the goal,values all water-related criteria to the group criterion “water-related criteria”, and all alternatives to the criterion “environment” based Tableon 6. Pair Scenario-wise 1. assessment of: all groups of criteria in regard to the goal, all water-related criteria to the group criterion “water-related criteria”, and all alternatives to the criterion “environment” based on Scenario 1. Goal Environment Water Related Social Economic Time Environment −2 5 −2 4 WaterGoal related Environment Water Related Social 5 Economic 1 Time 5 EnvironmentSocial −2 5 −2− 5 −43 Economic 5 Water related 5 1 5 Time Incon: 0.04 Social −5 −3 Water Related Drainage Channels Area Reservoir Economic 5 Drainage 1 −2 −2 Time Incon: 0.04 Channels −2 −2 WaterArea Related Drainage Channels Area Reservoir 1 DrainageReservoir 1 −2 Incon:−2 0.00 Environment Orljava–Londža Pleternica Ovˇcare Treštanovci Venje–Hrnjevac

Orljava–Londža 1 −6 −6 −5 Pleternica −6 −6 −5 Ovˇcare 1 5 Treštanovci 5 Venje–Hrnjevac Incon: 0.08 Water 2019, 11, x FOR PEER REVIEW 16 of 20

Water 2019Channels, 11, x FOR PEER REVIEW −2 −2 16 of 20 Area 1 ChannelsReservoir −2 Incon:−2 0.00 EnvironmentArea Orljava –Londža Pleternica Ovčare Treštanovci1 Venje–Hrnjevac OrljavaReservoir–Londža 1 −6 Incon:−6 0 .00 −5 EnvironmentPleternica Orljava– Londža Pleternica Ovčare−6 Treštanovci−6 Venje–Hrnjevac−5 OrljavaOvčare–Londža 1 −6 −61 −55 TreštanovciPleternica −6 −6 −55 VenjeOvčare–Hrnjevac 1 Incon:5 0.08

Water 2019Treštanovci, 11, 501 5 11 of 20 VenjeThe– rHrnjevacesults of using AHP with pairwise comparison of criteria and alternativesIncon: (for 0Scenario.08 1) ranked the Orljava–Londža alternative as the first to be implemented, Ovčare as the second, followed by Treštanovci,TheThe r resultsesults ofPleternica of using using AHP AHP, and with withVenje pairwise pairwise–Hrnjevac comparison comparison (Figure 5 ).of of criteria criteria and and alternatives alternatives (for (for Scenario Scenario 1) 1) rankrankeded the the Orljava Orljava–Londža–Londža alternative alternative as as the the first first to to be be implemented, implemented, Ovčare Ovˇcare as as the the second, second, followed followed byby Treštanovci, Treštanovci, Pleternica Pleternica,, and Venje–HrnjevacVenje–Hrnjevac (Figure(Figure5 5).).

(a) (b)

Figure 5. AHP results based on pairwise assessment of criteria and alternatives for Scenario 1: (a) the (a) (b) final ranking, and (b) graphical presentation of all alternatives assessment by each group of criteria. FigureFigure 5. 5. AHPAHP results results based based on on pairwise pairwise assessment assessment of of criteria criteria and and alternatives alternatives for Scenario 1 1:: ( (aa)) the the 3.3. ELECTREfinalfinal ranking, ranking, TRI and and ( b)) graphical graphical presentation presentation of all alternatives assessment by each group of criteria.criteria.

3.3. ELECTREELECTRE TRI TRI as a sorting method was applied on the same input data as those for the 3.3. ELECTRE TRI PROMETHEE method, i.e., data from Tables 3 and 4, to sort alternatives in three predefined categoriesELECTREELECTRE: first TRI TRIto be as as implemented, a a sorting sorting method method middle was was, and applied appliedlast to beo on nimplemented the the same same input input alternatives. data data as as Thethose those thresholds for for the the PROMETHEEPROMETHEEbetween categories method,method are, i.e., presentedi.e., data data from infrom T Tablesable Table 73, ands 34 in ,and to case sort 4 of, alternativesto Ovčare sort ,alternatives in inFigure three 6 predefined . in three predefined categories: categoriesfirst to be: implemented,first to be implemented, middle, and middle last to, and be implementedlast to be implemented alternatives. alternatives. The thresholds The thresholds between betwecategoriesen categories are presented are presented in TableTable7 in, and Table 7 in. Thresholds case7, and of in Ovˇcare, betweencase of in Ovčarecategories Figure,6 in.. Figure 6.

Criteria Table 7. ThresholdsWater Related between categories. Table 7. Thresholds between categories.

Criteria Water Related Economic Water Related Criteria Environment Social Economic Time

Environment Social (kn/ha)Time (ha)

Area (kn/ha)

Status

Area (ha) Economic Drainage Channels Threshold Environment Reservoir Social Time Drainage Channels Threshold Reservoir Status (kn/ha) (ha) Area First/MiddleFirst/Middle 2.5 2.5 2.52.5 2.5450 2.5 4502.5 Status 2.51. 1.55 25,00025,000 2.5 2.5 Drainage Channels ThMiddle/resholdLast Middle/Last 1.5 1.5 1.51.5 1.5150 1.5 1501Reservoir .5 1.51. 1.55 55,00055,000 1.5 1.5 First/Middle 2.5 2.5 2.5 450 2.5 1.5 25,000 2.5 Middle/Last 1.5 1.5 1.5 150 1.5 1.5 55,000 1.5

Figure 6.6. Visualization of alternative OvˇcareOvčare (red)(red) withwith categoriescategories thresholdsthresholds (blue)(blue) forfor ScenarioScenario 1.1.

TheFigure results 6. Visualization for criteria of weights alternative defined Ovčare by (red) Scenario with 1categories (Table3) thresholds are presented (blue) in for Table Scenario8. ELECTRE 1. TRI had two ways to rank the alternatives, using pessimistic or optimistic sorting. According to pessimistic sorting, there was no alternative in the first category (alternatives to be implemented first), whereas in the second, there was one, Ovˇcare, while all others were in the last category. For optimistic sorting, Orljava–Londža and Ovˇcare were sorted in the first category (alternatives to be implemented first), while Pleternica, Treštanovci, and Venje–Hrnjevac were to be implemented second. Sorting alternatives based on criteria weights defined in Table3 for Scenarios 2, 3, and 4 produced the same result. Water 2019, 11, x FOR PEER REVIEW 17 of 20

The results for criteria weights defined by Scenario 1 (Table 3) are presented in Table 8. ELECTRE TRI had two ways to rank the alternatives, using pessimistic or optimistic sorting. According to pessimistic sorting, there was no alternative in the first category (alternatives to be implemented first), whereas in the second, there was one, Ovčare, while all others were in the last category. For optimistic sorting, Orljava–Londža and Ovčare were sorted in the first category (alternatives to be implemented first), while Pleternica, Treštanovci, and Venje–Hrnjevac were to be implemented second. Sorting alternatives based on criteria weights defined in Table 3 for Scenarios 2, 3, and 4 produced the same result.

Water 2019, 11, 501 Table 8. Results of ELECTRE TRI for Scenario 1. 12 of 20

Category Pessimistic Optimistic First Table- 8. Results of ELECTRE TRI forOrljava Scenario– 1.Londža, Ovčare, Middle Ovčare Pleternica, Treštanovci, Venje–Hrnjevac Category Pessimistic Optimistic Orljava–Londža, Pleternica, Last - FirstTreštanovci, Venje -–Hrnjavac Orljava–Londža, Ovˇcare, Middle Ovˇcare Pleternica, Treštanovci, Venje–Hrnjevac Orljava–Londža, Pleternica, Last - It can be concluded thatTreštanovci, the first Venje–Hrnjavac alternative to be implemented suggested by ELECTRE TRI was Ovčare, followed by Orljava–Londža, and the last ones to be implemented were Pleternica, Treštanovci, and Venje–Hrnjevac. It can be concluded that the first alternative to be implemented suggested by ELECTRE TRI was Ovˇcare, followed by Orljava–Londža, and the last ones to be implemented were Pleternica, Treštanovci, 3.4. DEXi and Venje–Hrnjevac. DEXi method was applied on data from Tables 2 and 3, and the results for criteria weights defined3.4. DEXi by Scenario 1 are presented in Figure 7 [23]. DEXi software allows the display of results in multipleDEXi graphical method wasmodes. applied The first on data way from wasTables to display2 and the3, and final the evaluation results for in criteria which weights all the results defined of theby Scenarioanalysis we 1 arere presentedcombined (Figure in Figure 7a),7[ and23]. the DEXi recommended software allows alternatives, the display partially of results recommended in multiple, orgraphical not recommended modes. The we firstre waydisplayed. was to The display results the finalcan also evaluation be presented in which according all the results to a certain of the criterion,analysis were which combined also specify (Figure 7thea), andareas the tha recommendedt are recommended, alternatives, partially partially recommended recommended,, or or notnot recommended wereaccording displayed. to that The criterion. results canThe also results be presented can also accordingbe presented to a for certain two criterion,criteria. In which this case,also specifythe results the areaswere thatno longer are recommended, presented in partially line but recommended,were projected orat notthe recommendedpoint at the intersection according forto that both criterion. criteria (Figure The results 7b). The can presentation also be presented of results for for two three criteria. or more In thiscriteria case, is thedone results on a polygon. were no Anlonger ideal presented alternative in line would but wererepresent projected an unbroken at the point line at that the moves intersection at the for top both of the criteria polygon. (Figure If 7theb). Theresults presentation vary between of results good forand three medium, or more the criterialine is in is the done space on a between polygon. the An cent idealer and alternative the top wouldof the polygon,represent while an unbroken for the worse line that results, moves there at the is no top line, of the i.e., polygon. there is no If the shift results from varythe point between that goodindicates and themedium, center theof the line polygon. is in the space between the center and the top of the polygon, while for the worse results,An thereexample is no of line, results i.e., only there for is nothe shift area from Ovčare the has point been that presented indicates in the Figure center 8 of[23]. the polygon.

(a) (b)

FigureFigure 7. Presentation 7. Presentation of results of results of DEXi of DEXi application application for S forcenario Scenario 1: (a 1:) (resulta) result based based on onevaluation evaluation of ofall all criteria, and (bcriteria,) result andfor two (b) resultcriteria for [23]. two criteria [23].

WaterAn 2019 example , 11, x FOR of PEER results REVIEW onlyfor only for the area OvˇcareOvčare hashas beenbeen presentedpresented inin FigureFigure8 8[23 [23].]. 18 of 20

Figure 8. Result of alternative Ovˇcare analyses for Scenario 1 [23]. Figure 8. Result of alternative Ovčare analyses for Scenario 1 [23]. An example of results for only the area Ovˇcare has been presented in Figure8[23]. Based on the conducted analyses, using the DEXi software for Scenario 1, the area of Ovčare was recommended as the first in which the irrigation should be implemented (Figure 7a). The other three areas Orljava–Londža, Pleternica, and Treštanovci were partially recommended, while the area of Venje–Hrnjevac was not recommended (i.e., it is the last that should be implemented) [23]. Using Scenarios 2, 3, and 4 (from Table 3) the difference in the result was that both Orljava–Londža and Ovčare were recommended as the first in which the irrigation should be implemented.

3.5. PRIME The PRIME method was applied on data from Tables 2 and 3, and the results for criteria weights defined by Scenario 1 are presented in Figures 9–11. There are four ways to sort the results after importing the input data. The results are available in the following forms: value intervals, weights, dominance, and decision rules. A value interval represents the range of possible values, and every alternative has a value interval for each attribute, and each goal has its range of possible values. Form Figure 9, the alternative Venje–Hrnjevac was dominated by the alternatives Orljava– Londža and Ovčare because the least possible value of the latter was higher than the highest possible value of the former ones. The value interval did not tell which one of alternatives was the best, but it showed which had good chances. Dominance showed which alternative was preferred to another, for all the permissible combinations of preference information. Dominance is needed when the value intervals of two alternatives overlap, and a unique single alternative that would be preferred to the other may not exist. Absolute dominance means that the value intervals of the alternatives do not overlap and one alternative is preferred to the other without any doubt.

Figure 9. Range of possible values of alternatives for Scenario 1.

Dominance window contains a dominance matrix, as shown in Figure 10. A red dot in the matrix indicates that the alternative on that row is dominated by the alternative of that column, and the green dot means the opposite. Grey dots denote the diagonal of the matrix.

Water 2019, 11, x FOR PEER REVIEW 18 of 20

Figure 8. Result of alternative Ovčare analyses for Scenario 1 [23].

Based on the conducted analyses, using the DEXi software for Scenario 1, the area of Ovčare was recommended as the first in which the irrigation should be implemented (Figure 7a). The other three areas Orljava–Londža, Pleternica, and Treštanovci were partially recommended, while the area of Venje–Hrnjevac was not recommended (i.e., it is the last that should be implemented) [23]. Using Scenarios 2, 3, and 4 (from Table 3) the difference in the result was that both Orljava–Londža and Ovčare were recommended as the first in which the irrigation should be implemented.

3.Water5. PRIME2019, 11 , 501 13 of 20 The PRIME method was applied on data from Tables 2 and 3, and the results for criteria weights definedBased by onScenario the conducted 1 are presented analyses, in usingFigures the 9– DEXi11. There software are four for Scenarioways to 1,sort the the area results of Ovˇcare after importingwas recommended the input as data. the firstThe inresults which are the available irrigation in shouldthe following be implemented forms: value (Figure intervals,7a). The weigh otherts, dominancethree areas, Orljava–Londža, and decision rules. Pleternica, A value and interval Treštanovci represents were the partially range of recommended, possible values, while and the every area alternativeof Venje–Hrnjevac has a value was interval not recommended for each attribute, (i.e., it isand the each last goal that shouldhas its range be implemented) of possible values. [23]. Using ScenariosForm 2,Figure 3, and 9, 4 the (from alternative Table3) the Venje difference–Hrnjevac in thewas result dominated was that by both the Orljava–Londžaalternatives Orljava and– LondžaOvˇcare wereand Ovčare recommended because asthe the least first possible in which value the of irrigation the latter should was higher be implemented. than the highest possible value of the former ones. The value interval did not tell which one of alternatives was the best, but it 3.5. PRIME showed which had good chances. DominanceThe PRIME methodshowed was which applied alternative on data fromwas Tablespreferred2 and to3, andanother, the results for all for criteriathe permissible weights combinationsdefined by Scenario of preference 1 are presented information. in Figures Dominance9–11. Thereis needed are four when ways the tovalue sort theintervals results of after two alternativesimporting the overlap, input data.and a The unique results single are availablealternative in that the followingwould be forms:preferred value to the intervals, other may weights, not exist.dominance, Absolute and dominance decision rules. means A that value the interval value intervals represents of the alternatives range of possible do not values, overlap and and every one alternative is has preferred a value intervalto the other for eachwithout attribute, any doubt. and each goal has its range of possible values.

Water 2019, 11, x FOR PEER REVIEW 19 of 20

Water 2019, 11, x FOR PEERFigure REVIEW 9. Range of possible values of alternatives for Scenario 1. 19 of 20 Figure 10. DominanceFigure 9. Range matrix of of possible mutual values relationships of alternatives among alternativesfor Scenario for1. Scenario 1.

DominanceDecision rules window help the contains decision a dominance maker select matrix, the best as shown alternative. in Figure There 10. Aare red four dot decision in the matrix rules indicatesthat can thatapply the to alternative different situations:on that row maximax, is dominated maximin, by the central alternative values of, thatand column,minimax and regret. the greenMaximax dot meanssupposes the that opposite. the most Grey likely dots valuedenote lies the at diagonal or near theof the great matrix. bound of the value intervals, and selects the alternative with the greatest upper bound. Maximin supposes that the worst case for the alternative will happen and selects the alternative with the greatest lower bound of the value interval. Central values select the alternative with the greatest midpoint, while minimax regret takes a different approach and calculates the possible loss of value, which means that it selects the alternative with the least possible loss of value. These decision rules only give advice, and the decision makerFigures should10 10.. Dominance pick the matrix alternative of mutual that relationships they find best among (Figure alternatives 11). for Scenario 1.1.

Decision rules help the decision maker select the best alternative. There are four decision rules that can apply to different situations: maximax, maximin, central values, and minimax regret. Maximax supposes that the most likely value lies at or near the great bound of the value intervals, and selects the alternative with the greatest upper bound. Maximin supposes that the worst case for the alternative will happen and selects the alternative with the greatest lower bound of the value interval. Central values select the alternative with the greatest midpoint, while minimax regret takes a different approach and calculates the possible loss of value, which means that it selects the alternative with the least possible loss of value. These decision rules only give advice, and the decision makers should pick the alternative that they find best (Figure 11). Figure 11.11. Advised alternative for decision maker for Scenario 1.1.

FormUsing FigureScenarios9, the 1, alternative 2, and 4 (from Venje–Hrnjevac Table 3) all was the dominated results suggest by the that alternatives the first alternative Orljava–Londža to be andimplemented Ovˇcare because should the be leastOvča possiblere, followed value by of Orljava the latter–Londža, was higher Pleternica, than the Trešta highestnovci possible, and as value the last of one Venje–Hrnjevac. Results based on Scenario 3 suggest that the first alternative to be implemented should be Orljava–Londža, followed by Ovčare, Pleternica, Treštanovci, and Venje–Hrnjevac.

3.6. PCA Input data for PCA are adopted from Table 1 by introducing mark “2” instead of “best” and mark “1” instead of “worst” and based on Scenario 1 from Table 3. Application of PCA defined three key factors, and eachFigure of them 11. isAdvised a combination alternative of for several decision criteria maker (forTable Scenario 9). The 1. contribution of key factor F1 is 44% and it consists of two criteria: environmental protection and recreation. Contribution of keyUsing factor Scenarios F2 is 37%. 1, 2It, isand a combination4 (from Table of 3 four) all thecriteria: results drainage, suggest irrigation that the first area, alternative reservoirs ,to and be implementedtime for complet shouldion. Thebe Ovča thirdre, key followed factor byF3 Orljavahas a contribution–Londža, Pleternica, of 19% with Trešta thenovci contribution, and as the of twolast onecriteria: Venje cost/unit–Hrnjevac. area Results and channel based network.on Scenario 3 suggest that the first alternative to be implemented should be Orljava–Londža, followed by Ovčare, Pleternica, Treštanovci, and Venje–Hrnjevac. Table 9. Factor loadings according to PCA (Principal Component Analysis) for Scenario 1. 3.6. PCA F1 F2 F3 Input data for PCA areEnvironment adopted from Table−0 1.990 by introducing−0.126 −0 .069mark “2” instead of “best” and mark “1” instead of “worst” Drainageand based on Scenario0.072 1 from 0.942 Table 3.− Application0.329 of PCA defined three key factors, and each of themChannels is a combination of0.590 several − 0criteria.206 0.780(Table 9). The contribution of key factor F1 is 44% and it consistsIrrigation of two criteria: area environmental0.554 0.773 protection 0.309 and recreation. Contribution of key factor F2 is 37%. It is a combination of four criteria: drainage, irrigation area, reservoirs, and time for completion. The third key factor F3 has a contribution of 19% with the contribution of two criteria: cost/unit area and channel network.

Table 9. Factor loadings according to PCA (Principal Component Analysis) for Scenario 1.

F1 F2 F3 Environment −0.990 −0.126 −0.069 Drainage 0.072 0.942 −0.329 Channels 0.590 −0.206 0.780 Irrigation area 0.554 0.773 0.309

Water 2019, 11, 501 14 of 20 the former ones. The value interval did not tell which one of alternatives was the best, but it showed which had good chances. Dominance showed which alternative was preferred to another, for all the permissible combinations of preference information. Dominance is needed when the value intervals of two alternatives overlap, and a unique single alternative that would be preferred to the other may not exist. Absolute dominance means that the value intervals of the alternatives do not overlap and one alternative is preferred to the other without any doubt. Dominance window contains a dominance matrix, as shown in Figure 10. A red dot in the matrix indicates that the alternative on that row is dominated by the alternative of that column, and the green dot means the opposite. Grey dots denote the diagonal of the matrix. Decision rules help the decision maker select the best alternative. There are four decision rules that can apply to different situations: maximax, maximin, central values, and minimax regret. Maximax supposes that the most likely value lies at or near the great bound of the value intervals, and selects the alternative with the greatest upper bound. Maximin supposes that the worst case for the alternative will happen and selects the alternative with the greatest lower bound of the value interval. Central values select the alternative with the greatest midpoint, while minimax regret takes a different approach and calculates the possible loss of value, which means that it selects the alternative with the least possible loss of value. These decision rules only give advice, and the decision makers should pick the alternative that they find best (Figure 11). Using Scenarios 1, 2, and 4 (from Table3) all the results suggest that the first alternative to be implemented should be Ovˇcare, followed by Orljava–Londža, Pleternica, Treštanovci, and as the last one Venje–Hrnjevac. Results based on Scenario 3 suggest that the first alternative to be implemented should be Orljava–Londža, followed by Ovˇcare, Pleternica, Treštanovci, and Venje–Hrnjevac.

3.6. PCA Input data for PCA are adopted from Table1 by introducing mark “2” instead of “best” and mark “1” instead of “worst” and based on Scenario 1 from Table3. Application of PCA defined three key factors, and each of them is a combination of several criteria (Table9). The contribution of key factor F1 is 44% and it consists of two criteria: environmental protection and recreation. Contribution of key factor F2 is 37%. It is a combination of four criteria: drainage, irrigation area, reservoirs, and time for completion. The third key factor F3 has a contribution of 19% with the contribution of two criteria: cost/unit area and channel network.

Table 9. Factor loadings according to PCA (Principal Component Analysis) for Scenario 1.

F1 F2 F3 Environment −0.990 −0.126 −0.069 Drainage 0.072 0.942 −0.329 Channels 0.590 −0.206 0.780 Irrigation area 0.554 0.773 0.309 Reservoirs −0.534 0.793 0.293 Recreation 0.990 0.126 0.069 Cost/unit area 0.622 0.360 −0.695 Time to complete −0.534 0.793 0.293

The irrigation area of priority depends on the value of cos2. Squared cosines of the observations are presented in Table 10. The value of a contribution is between 0 and 1 and, for a given component, the sum of the contributions of all observations is equal to 1. The larger the value of the contribution, the more the observation contributes to the component. It is clear that Ovˇcarahas high priority mostly because of the contribution of F2 key factor. Water 2019, 11, x FOR PEER REVIEW 20 of 20

Reservoirs −0.534 0.793 0.293 Recreation 0.990 0.126 0.069 Cost/unit area 0.622 0.360 −0.695 Time to complete −0.534 0.793 0.293

The irrigation area of priority depends on the value of cos2. Squared cosines of the observations are presented in Table 10. The value of a contribution is between 0 and 1 and, for a given component, the sum of the contributions of all observations is equal to 1. The larger the value of the contribution, the more the observation contributes to the component. It is clear that Ovčara has high priority mostly Water 2019, 11, 501 15 of 20 because of the contribution of F2 key factor.

Table 10.10. Squared cosines of the observations for Scenario 1.1.

F1F1 F2F2 F3 F3 Orljava–LondžaOrljava–Londža0.558 0.558 0.1540.154 0.288 0.288 PleternicaPleternica 0.5490.549 0.0550.055 0.396 0.396 OvˇcareOvčare 0.3390.339 0.6180.618 0.042 0.042 TreštanovciTreštanovci 0.3660.366 0.5870.587 0.047 0.047 Venje–Hrnjevac 0.366 0.587 0.047 Venje–Hrnjevac 0.366 0.587 0.047

Locations Venje Venje–Hrnjevac–Hrnjevac and Treštanovci areare rankedranked on the second position dominantlydominantly u undernder the influenceinfluence of environmental protection criterion and criteria of F2 key factor. Pleternica is under the influenceinfluence of the F2 key factor, dominantlydominantly a channel network. All four locations have similar and rather closeclose rankings,rankings,but but Pleternica Pleternica has has been been evaluated evaluated as as the the least least advisable advisable location location even even though though the differencesthe differences are almostare almost negligible. negligible. The major major advantage advantage of of Ovˇcaralocation Ovčara location is the is timethe requiredtime required to complete to complete and the existingand the reservoir,existing whichreservoir is obvious, which fromis obvious the biplot from of the key biplot factors ofF1 key and factors F2 and F1 observed and F2 and locations observed (Figure locations 12). Locations (Figure Venje–Hrnjevac12). Locations Venje and– TreštanovciHrnjevac and are Treštanovci ranked on the are second ranked position on the second dominantly position under dominantly the influence under of environmentalthe influence of protection environmental criterion protection and criteria criterion of F2 and key criteria factor. of Pleternica F2 key factor. is under Pleternica the influence is under of the F2influence key factor, of the dominantly F2 key factor, the channeldominantly network. the channel All four network locations. All have four similar location ands have rather similar close rankings,and rather but close Pleternica ranking hass, but been Pleternica evaluated has as been the evaluated least advisable as the location. least advisable location.

Figure 12.12. Biplot of key factors and observed locations.locations.

Results presented in Figure 12 are based upon Scenario 1,1, and show a clearclear conclusion about the OvˇcareOvčare location as the most acceptable alternative. PCA was also applied on scenarios 2 to 4 and the conclusion was the same, there was nono changechange inin alternativesalternatives ranking.ranking.

3.7.3.7. Comparison of Results Previously described applications of six multi-criteria analysis (PROMETHEE, AHP, ELECTRE TRI, DEXi, PRIME, and PCA) in the Croatian case study showed similar results. According to criteria importance defined as Scenario 1 (Table3) the Ovˇcare area has been estimated as the first to be implemented (most recommended) by all methods, except AHP (Table 11). The reservoir for irrigation already exists, the time to complete the irrigation system is up to 2 years, the agricultural area is not in any restricted (protected) zone, the status of the drainage system is mostly built, and the cost is not the highest, although the reservoir is not used for recreation. Except for the PCA method, the least favorable location is Venje–Hrnjevac, mostly owing to its position close to the nature park. Water 2019, 11, 501 16 of 20

Table 11. Comparison of results from different MCA (multi-criteria analysis) methods for different criteria weight scenarios.

PROMETHEE AHP ELECTRE TRI DEXi PRIME PCA Alternative S1 S2 S3 S4 S1 S2 S3 S4 S1 S2 S3 S4 S1 S2 S3 S4 S1 S2 S3 S4 S1 S2 S3 S4 Orljava–Londža +/−+ + +/−+ + + +/−+/− +/− +/−+/−+/−+ + + +/−+/−+ +/−+/−+/−+/−+/− Pleternica +/−+/−+/−+/−+/−+/−+/−+/− − − − − +/−+/−+/−+/−+/−+/−+/−+/−− − − − Ovˇcare + + +/−+ +/−+/−+/−+ + + + + + + + + + + +/−+ + + + + Treštanovci +/−+/−+/−+/−+/−+/−+/−+/− − − − − +/−+/−/+−+/−+/−+/−+/−+/−+/−+/−+/−+/− Venje–Hrnjevac − − − − − − − − − − − − − − − − − − − − +/−+/−+/−+/− Green—alternatives with high priority (first to be implemented); Yellow—alternatives with medium priority; Red—alternatives with low priority (last to be implemented).

To test the robustness of MCA methods application, the sensitivity analyses was performed according to three additional criteria weight scenarios defined in Table3 and the results are presented in Table 11. The sensitivity analyses mostly confirmed Ovˇcare as the first alternative to be implemented. Exception were the results for the AHP method in case of Scenarios 1, 2, and 3, and in case of PROMETHEE, and PRIME methods for Scenario 3 where Orljava–Londža is suggested to be the first to be implemented. For Scenario 2 in case of the PROMETHEE method application and for Scenarios 2, 3, and 4 for DEXi both Orljava–Londža and Ovˇcare were ranked as first to be implemented. ELECTRE TRI and PCA ranked Ovˇcare as the first alternative to be implemented independently of the criteria weight scenario applied, while all methods except PCA pointed out Venje–Hrnjevac as the last alternative to be implemented. Most similar results were obtained by PROMETHEE, DEXi, and PRIME although their algorithms were different. In case of AHP application the differences in the final score between alternatives Ovˇcare and Orljava–Londža for all scenarios are small. ELECTRE TRI and PCA were the least sensitive to criteria weight changes. All selected MCA methods have demonstrated to be adequate for defining priorities in the irrigation plans implementation, and the following conclusions about their application on the case study were made. The PROMETHEE method demands all alternatives’ assessments to be quantitative and gives the criteria importance too, so the qualitative ones had to be transformed into quantitative ones. The method was simple to use and not time-consuming. The GAIA method gave additional information for decision making. For AHP application input data can be both quantitative and qualitative. In case of direct assessment with quantitative values, the use of the method was simple and not time-consuming. If pairwise comparison was applied, there was no need for quantitative data, but more time was needed for pairwise assessment of all alternatives by each criterion, and of criteria to the goal. Since ELECTRE TRI is a method that sorts alternatives in predefined categories, categories’ thresholds had to be defined. The method demands all alternatives assessments to be quantitative and gives the criteria importance too, so the qualitative values had to be transformed into quantitative ones. The method use was simple and not time-consuming. For DEXi decision model the quantitative assessment of alternatives had to be transformed into qualitative ones (i.e., words). Input data had to be hierarchically sorted for each criterion. It was simple to use, and allowed various display options for output data: graphical display for all alternatives or one alternative or a comparison of two criteria, with very good visualization. For PRIME method application quantitative assessment of alternatives was given in intervals. Input data must be compared with the most important attribute that may require slightly more time. Output data were a combination of graphical and tabular views, that are only advisory in nature, and the decision maker should pick the alternative he/she finds the best. PCA offers the possibility to combine several criteria and define key factors essential for specific process. A group of criteria were ranked according to their importance. In this way, some parameters could be neglected. PCA has good visualization and graphical presentation. Water 2019, 11, 501 17 of 20

Regarding the specific case study in Croatia, defining priorities in implementation of the irrigation plan of Požega–Slavonia County, more appropriate methods were the ones that have the possibility to use qualitative assessment of alternatives because in the planning phase the input data for the assessment of alternatives were not precise. Comparing results from this study with published research [51,52] confirmed that different multi-criteria decision making methods identify mostly the similar ranking among water management alternatives in case objective weights are assigned to the criteria. Selecting the MCA technique has proven in this research (including methods DEXi, PRIME, and PCA) to have lesser importance than the initial structuring of the decision problem (selection of criteria, generating alternatives, weighting the criteria and assessing the alternatives) as stated by Hajkowicz and Higgins in [51]. The difference occurs because the weights and the resulting distributions of the scores within the criteria do not have the same impact on all the methods [31]. The slight sensitivity to the weight assignment to the criteria [52] was observed in this case study too. It can be confirmed that the application of several methods, but also the sensitivity analysis, is necessary to check the consistency and increase the reliability of the results [31] for the analyzed case study.

4. Conclusions Currently, the irrigation of agricultural areas is essential to ensure food security and stimulate rural socio-economic development. In the 21st century, water resources will certainly be one of the major issues. Water consumption in irrigation is very large and very often non-sustainable. Improving the performance of irrigation systems by increasing the efficiency and transparency of those systems is essential. The proposed analysis should be done prior to the more detailed project development in order to avoid potential negative irrigation effects on environment and conflicts among water consumers. In the case study area of Požega–Slavonia county (Croatia), five potential irrigation areas were analyzed according to the defined criteria using six multi-criteria analysis methods PROMETHEE, AHP, ELECTRE TRI, DEXi, PRIME, and PCA. Five of the six methods identified Ovˇcare as the most appropriate area for an irrigation development case study for criteria weights defined by Scenario 1. The sensitivity analysis based on three additional criteria weight scenarios confirmed these conclusions with small variations, ranking Ovˇcare and Orljava–Londža both as alternatives that could be implemented first (PROMETHEE and DEXi), and in some cases pointed Orljava–Londža as the alternative that should be implemented first (PROMETHEE, AHP and PRIME). Five of the six methods have the same result for the least advisable alternative, Venje–Hrnjevac. The results show that selected methods can be applied for defining priorities in irrigation project development. The novelty in this research is the application of DEXi, PRIME, and PCA methods that were not used so often on this kind of problems. For the case study in Croatia, more appropriate MCA methods are the ones that can perform qualitative assessment of alternatives, since the input data for the assessment of alternatives were not precise. Initial structuring of the decision problem, especially the criteria’s definition, is the most sensitive part in the multi-criteria evaluation. The criteria should present realistic and objective characteristics of the study field in order to achieve results with an objective and valuable decision. As the weighting factors in this research were defined by authors (considering awareness of the local population for environmental protection and its value, economic development and value of money), although the sensitivity analysis was done, further research should involve stakeholders in the definition of weight factors. The selected methods confirmed to be a useful tool for informing decision-makers about the sustainability of catchment scale water use in a relatively simple way and help them to develop strategies for further improvement of water resources management in irrigation districts. Water 2019, 11, 501 18 of 20

Author Contributions: Conceptualization, B.K.; methodology, B.K., A.H., and L.T.; formal analysis, B.K., A.H., and L.T.; investigation, B.K., A.H., and L.T.; resources, B.K.; data curation, B.K., A.H., and L.T.; writing—original draft, B.K., A.H., and L.T.; writing—review/editing, B.K.; visualization, B.K., A.H., and L.T.; supervision, B.K.; project administration, B.K.; funding acquisition, B.K. Funding: The research for this article and the publication of this article were funded by the University of Rijeka within projects: “Development of New Methodologies in Water and Soil Management in Karstic, Sensitive and Protected Areas” (13.05.1.3.08) and “Implementation of Innovative Methodologies, Approaches and Tools for Sustainable River Basin management” (UNIRI-TEHNIC-18-129). Conflicts of Interest: The authors declare no conflict of interest. The funders had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript; or in the decision to publish the results.

References

1. Masia, S.; Sušnik, J.; Marras, S.; Mereu, S.; Spano, D.; Trabucco, A. Assessment of Irrigated Agriculture Vulnerability under Climate Change in Southern Italy. Water 2018, 10, 209. [CrossRef] 2. Đorđevi´c,B. Vodoprivredni sistema; Nauˇcnaknjiga: Beograd, Serbia, 1990. 3. Margeta, J. Osnove Gospodarenja Vodama/Basis of Water Management; Građevinski fakultet: Split, Croatia, 1992. 4. Karleuša, B.; Ožani´c,N. Određivanje prioriteta u realizaciji vodnogospodarskih planova/Determination of priorities during realization of water management plans. Građevinar 2011, 63, 2. Available online: http://www.casopis-gradjevinar.hr/archive/article/371 (accessed on 1 November 2018). 5. Karleuša, B.; Ožani´c, N.; Deluka-Tibljaš, A. Improving decision making in defining priorities for implementation of irrigation plans using AHP methodology. Tech. Gazette 2014, 21, 673–680. Available online: https://hrcak.srce.hr/index.php?show=clanak&id_clanak_jezik=182226 (accessed on 11 January 2018). 6. Tiwari, D.N.; Loof, R.; Paudyal, G.N. Environmental-economic decision-making in lowland irrigated agriculture using multi-criteria analysis techniques. Agric. Syst. 1999, 60, 99–112. [CrossRef] 7. Opricovi´c, S. Višekriterijumska optimizacija/Multi-criteria optimization; Nauˇcna knjiga: Beograd, Yugoslavia, 1986. 8. Triantaphyllou, E. Multi-Criteria Decision Making Methods: A Comparative Study; Kluwer Academic Publishers: Dordrecht, The Netherlands, 2000. 9. Figueira, J.; Greco, S.; Ehrogott, M. Multiple Criteria Decision Analysis: State of the Art Surveys. In International Series in Operations Research & Management Science; Springer: New York, NY, USA, 2005; Volume 78. 10. Huang, I.B.; Keisler, J.; Linkov, I. Multi-criteria decision analysis in environmental sciences: Ten years of applications and trends. Sci. Total Environ. 2011, 409, 3578–3794. [CrossRef][PubMed] 11. Hajkowicz, S.; Collins, K. A Review of Multiple Criteria Analysis for Water Resource Planning and Management. Water Resour. Manag. 2007, 21, 1553–1566. [CrossRef] 12. Raju, K.S.; Pillai, C.R.S. Multi-criterion decision making in river basin planning and development. Eur. J. Oper. Res. 1999, 112, 249–257. [CrossRef] 13. Duckstein, L.; Treichel, W.; El Magnouni, S. Ranking Ground-water Management Alternatives by Multi-criterion Analysis. J. Water Resour. Plan. Manag. 1994, 120, 546–565. [CrossRef] 14. Paneque Salgado, P.; Corral Quintana, S.; Guimaraes Pereira, A.; del Moral Ituarte, L.; Pedregal Mateos, B. Participative multi-criteria analysis for the evaluation of water governance alternatives. A case study in the Costa del Sol (Malaga). Ecol. Econ. 2009, 68, 990–1005. [CrossRef] 15. Sapkota, M.; Arora, M.; Malano, H.; Sharma, A.; Moglia, M. Integrated Evaluation of Hybrid Water Supply Systems Using a PROMETHEE–GAIA Approach. Water 2018, 10, 610. [CrossRef] 16. Srđevi´c,B.; Jandri´c,Z.; Potkonjak, S. Vrednovanje alternativa korištenja akumulacije pomo´cuanalitiˇckog hijerarhijskog procesa. Jugoslavenski ˇcasopis:Vodoprivreda 2000, 32, 237–242. 17. Afify, A. Prioritizing desalination strategies using multi-criteria decision analysis. Desalination 2010, 250, 928–935. [CrossRef] 18. Margeta, J. Izbor optimalnog rješenja dispozicije otpadnih voda Poljiˇckerivijere/Selection of the optimal solution for waste water disposal of Poljiˇckariviera. Vodoprivreda 1986, 18, 79–87. 19. Martin, C.; Ruperd, Y.; Legret, M. Urban stormwater drainage management: The development of a multicriteria decision aid approach for best management practices. Eur. J. Oper. Res. 2007, 181, 338–349. [CrossRef] Water 2019, 11, 501 19 of 20

20. Margeta, J.; Mladineo, N.; Petriˇcec,M. Višekriterijsko rangiranje lokacija za male hidroelektrane/Multicriteria ranking of locations for small hydropower plants. Građevinar 1987, 39, 239–244. 21. Raju, K.S.; Pillai, C.R.S. Multicriterion decision making in performance evaluation of an irrigation system. Eur. J. Oper. Res. 1999, 112, 479–488. [CrossRef] 22. Goncalves, J.M.; Pereira, L.S.; Fang, S.X.; Dong, B. Modelling and multicriteria analysis of water saving scenarios for an irrigation district in the upper Yellow River Basin. Agric. Water Manag. 2007, 94, 93–108. [CrossRef] 23. Karleuša, B.; Hajdinger, A.; Tadi´c,L. Use of Multicriteria Analysis Method DEXi to Define Priorities in Implementation of Irrigation Plans. Proceedings 2018, 2, 663. [CrossRef] 24. Assefa, T.; Jha, M.; Reyes, M.; Srinivasan, R.; Worqlul, A.W. Assessment of Suitable Areas for Home Gardens for Irrigation Potential, Water Availability, and Water-Lifting Technologies. Water 2018, 10, 495. [CrossRef] 25. Malekmohammadi, B.; Zahraie, B.; Kerachian, R. Ranking solutions of multi-objective reservoir operation optimization models using multi-criteria decision analysis. Expert Syst. Appl. 2011, 38, 7851–7863. [CrossRef] 26. Hosseinzade, Z.; Pagsuyoin, S.; Ponnambalam, K.; Monem, M.J. Decision-making in irrigation networks: Selecting appropriate canal structures using multi-attribute decision analysis. Sci. Total Environ. 2017, 601–602, 177–185. [CrossRef][PubMed] 27. Selim, S.; Koc-San, D.; Selim, C.; San, B.T. Site selection for avocado cultivation using GIS and multi-criteria decision analyses: Case study of Antalya, Turkey. Comput. Electron. Agric. 2018, 154, 450–459. [CrossRef] 28. Chen, Y.; Paydar, Z. Evaluation of potential irrigation expansion using a spatial fuzzy multi-criteria decision framework. Environ. Model. Softw. 2012, 38, 147–157. [CrossRef] 29. Elshaikh, A.E.; Xiyun, J.; Shi-hong, Y. Performance evaluation of irrigation projects: Theories, methods, and techniques. Agric. Water Manag. 2018, 203, 87–96. [CrossRef] 30. Montazar, A.; Gheidari, O.N.; Snyder, R. A fuzzy analytical hierarchy methodology for the performance assessment of irrigation projects. Agric. Water Manag. 2013, 121, 113–123. [CrossRef] 31. Tscheikner-Gratl, F.; Egger, P.; Rauch, W.; Kleidorfer, M. Comparison of Multi-Criteria Decision Support Methods for Integrated Rehabilitation Prioritization. Water 2017, 9, 68. [CrossRef] 32. Sun, H.; Wang, S.; Hao, X. An Improved Analytic Hierarchy Process Method for the evaluation of agricultural water management in irrigation districts of north China. Agric. Water Manag. 2017, 179, 324–337. [CrossRef] 33. Hidroprojek-ing, D.O.O.; Hidroing, D.O.O. Osnove navodnjavanja na podruˇcjuPožeško-slavonske županije/Basics for irrigation in Požega-Slavonia County; Osijek: Zagreb, Croatia, 2005. 34. Prostorni plan Požeško-slavonske županije. Spatial plan of Požega-Slavonia County; Požeško-slavonski službeni glasnik 5/02, 5A/02, 05/2015; Požega-Slavonia County: Požega, Croatia, 2015. 35. Razvojna strategija Požeško-slavonske županije za razdoblje 2011–2013/Development strategy of Požega-Slavonia County; Požega-Slavonia County: Požega, Croatia, 2011. 36. Kantoci, D. Navodnjavanje/Irrigation. Glasnik zaštite bilja 2012, 35, 66–72. Available online: https://hrcak. srce.hr/index.php?show=clanak&id_clanak_jezik=240422 (accessed on 10 May 2018). 37. Riabacke, M.; Danielson, M.; Ekenberg, L. State-of-the-Art Prescriptive Criteria Weight Elicitation. Adv. Decis. Sci. 2012, 276584. [CrossRef] 38. Rozman, C.;ˇ Pažek, K. Introduction to DEXi multi criteria decision models: What they are and how to use them in agriculture? Agricultura 2012, 9, 23–30. Available online: http: //www.agricultura-online.com/portal/issues/50-issue-16/153-introduction-to-dexi-multi-criteria- decision-models-what-they-are-and-how-to-use-them-in-agriculture (accessed on 23 November 2017). 39. Gustafsson, J.; Salo, A.; Gustafsson, T. PRIME Decisions: An Interactive Tool for Value Analysis; University of Technology, System Analysis Laboratory: Helsinki, Finland, 2011. 40. Abdi, H.; Williams, L.J. Principal Component Analysis. Wiley Interdiscip. Rev. Comput. Stat. 2010, 2, 433–459. [CrossRef] 41. Roy, B.; Vincke, P.; Mareschal, B. How to select and how to rank project: The PROMETHEE method. Eur. J. Oper. Res. 1981, 24, 228–238. [CrossRef] 42. Brans, J.P.; Vincke, P. Preference Ranking Organization Method: The Promethee Method for MCDM. Manag. Sci. 1985, 31, 647–656. [CrossRef] 43. Saaty, T.L. The Analytic Hierarchy Process, 2nd ed.; RWS Publications: Pittsburg, PA, USA, 1996. 44. Saaty, T.L. Fundamentals of Decision Making and Priority Theory; RWS Publications: Pittsburg, CA, USA, 1994. Water 2019, 11, 501 20 of 20

45. Mousseau, V.; Slowinski, R. Inferring an ELECTRE TRI Model from Assignment Examples. J. Glob. Optim. 1998, 12, 157–174. [CrossRef] 46. Mousseau, V.; Slowinski, R.; Zielniewicz, P. ELECTRE TRI 2.0 a Methodological Guide and User’s Manual; Universite Paris-Dauphine-Paris-Francuska and Poznan University of Technology: Piotrowo, Polland, 1998. 47. Gustafsson, T. A Comparison of PRIME Decisions with Other Tools for Decision Analysis; University of Technology, System Analysis Laboratory: Helsinki, Finland, 2000. 48. Wold, S.; Kim Esbensen, K.; Paul Geladi, P. Principal component analysis. Chemom. Intell. Lab. Syst. 1998, 2, 37–52. [CrossRef] 49. Rao, A.R.; Burke, T.T. Principal Component Analysis of Hydrologic Data. In Integrated Approach to Environmental Data Management Systems; Harmancioglu, N.B., Alpaslan, M.N., Ozkul, S.D., Singh, V.P., Eds.; NATO ASI Series (Series: 2: Environment); Springer: Dordrecht, The Netherlands, 1997; Volume 31. 50. Daigle, A.; St-Hilaire, A.; Beveridge, D.; Caissie, D.; Benyahya, L. Multivariate analysis of the low-flow regimes in eastern Canadian rivers. Hydrol. Sci. J. 2011, 56, 51–67. [CrossRef] 51. Hajkowicz, S.; Higgins, A. A comparison of multiple criteria analysis techniques for water resource management. Eur. J. Oper. Res. 2008, 184, 255–265. [CrossRef] 52. Yilmaz, B.; Harmancioglu, N.B. Multi-criteria decision making for water resource management: A case study of the Gediz River Basin, Turkey. Water 2010, 36, 5. [CrossRef]

© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).