1. Introduction
Waste refers to any substance requiring disposal, which includes unusable materials, worthless, defective and unwanted items. The site of the disposal of waste materials is called landfill. This site involves either collecting, sorting, processing, or recycling of wastes [
1,
2,
3]. Population has a direct relationship with waste production possesses, which contribute to environmental deterioration [
4]. Therefore, for each city, solid waste management is a crucial environmental challenge [
5].
Growing world population along with ever-increasing global urbanization has emerged as a major environmental concern in the 21st century. A direct correlation exists between population and waste quantity. Urbanites are generating more waste than ever before, as 56.2% of world population was residing in cities in 2020. It is estimated that by the middle of this century about 70% of people in the world will live in cities [
6]. In addition, proximity to manufacturing facilities and industrial plants in urban areas contribute to a large amount of waste representing a mix of ordinary garbage, termed Municipal Solid Waste (MSW). Electronic or e-waste, medical/health care waste (that became a serious issue during the COVID-19 pandemic) further adds to the problem of managing the increasing volume of waste [
6]. Despite the technological advances in converting waste to energy (WtE) and increasing recycling rates, landfills, as yet, are the predominant way of MSW disposal: For example, in 2019, China accounted for 45% landfill sites, where China produced more than 242 million tons of MSW [
7]. Yet, suitable sites for locating new landfills are getting scarce due to various geological, engineering, legal, and societal constraints. This paper is an attempt to provide a sound basis for landfill site (LFS) selection by using the most widely used criteria and subjecting them to rigorous Geographic Information System (GIS) methodologies. The article is intended to serve as a screening tool to select candidate sites that meet the known criteria to be followed by detailed site investigations. This approach would entail significant cost savings because field studies would be focused on only these promising (candidate) sites that have met the criteria, thereby substantially reducing time and expenses involved in site investigations.
Dozens of studies have been done to solve the waste disposal problem. Part of these studies applied various Multi-Criteria Decision-Analysis (MCDA) models to select suitable location for municipal solid waste (MSW) disposal sites [
8]. The most common MCDA models are: fuzzy Analytic Hierarchy Process (AHP) [
9], Technique for Order Performance by Similarity to an Ideal Solution (TOPSIS) [
10], Weighted Sum Method (WSM) [
11], and Weighted Product Method (WPM) [
12]. All these methods have been widely used in the field of MSW management [
13,
14,
15,
16,
17,
18].
A simple review of 27 high quality articles (
Table 1) selected from Scopus database and published recently, dealing with landfilling of MSW shows that more than 80% of these papers used slope gradient, distance to the villages, the towns and the cities, and distance to the road as important predictive factors (PFs) for LFS selection. More than 73% of these articles applied distance to surface water bodies as a predictive factor (PF), while >50% used lithology, soil, land use and land cover (LULC), groundwater depth, and distance to the airport as PFs. Elevation, distance to the active fault, distance to the powerline and distance to the agricultural lands were used less frequently (between 25% and 50%).
In this study, we identified suitable sites for LFS using GIS methods and prepared maps showing suitable LFSs for the Tanjero River Basin (TRB) in the Iraqi Kurdistan region. The aims of this paper were twofold: (1) to compare and evaluate the efficacy of five MCDA methods (Boolean Overlay (BO), WSM, WPM, AHP and TOPSIS); and (2) to find the most suitable site(s) for LFS in the TRB. For this purpose, we used 15 layers to assess methods’ performance. These thematic layers involve (1) lithology, (2) soil, (3) land cover, (4) distance to road, (5) slope gradient, (6) Topographic Position Index (TPI), (7) groundwater depth, (8) distance to the towns and the cities, (9) distance to the village, (10) distance to the active fault, (11) distance to the powerline, (12) distance to the surface water bodies, (13) distance to the agricultural lands, (14) elevation, and (15) distance to the springs.
Part of the TRB has been studied by [
19]. They used MCDA methods to identify seven suitable LFSs. However, we considered the entire TRB to give full evaluation of the whole basin. In addition, we expanded the factors used by [
19] via adding some important factors, such as distance to springs, distance to active faults, distance to agricultural lands, and Topographic Position Index (TPI). To the best of our knowledge, the TPI as a PF for LFS selection is being used first time in this study.
3. Predictive Factors
A several of appropriate factors must be taken into account to select the most suitable LFSs.
Table 1 shows the 27 reviewed high-quality articles published recently [
2,
3,
20,
21,
22,
23,
24,
25,
26,
27,
28,
29,
30,
31,
32,
33,
34,
35,
36,
37,
38,
39,
40,
41,
42,
43]. We depend on these articles to select the PFs dealing with landfilling. Fourteen PFs were selected (
Table 2; factors #1–5 and #7–15), which were applied >30% in the literature, and the rest factors were excluded (
Table 1). We used the TPI as a PF (
Table 2; factor no. 6) to select MSW landfill sites (LFSs) for the first time.
The PFs used to select the LFSs can be categorized as: hydrologic geologic, topographic, socio-economic and land use factors [
20]. We selected fifteen PFs as thematic layers (
Table 2).
The pixel of the layers were resized to 30 m spatial resolution. Two raster formats were used for the PFs (continuous and discrete;
Table 2). We converted the continuous factors to discrete factors by selecting multi threshold values based on our knowledge and background of the study area. The numbers and boundaries of categories can affect the results of the MCDA methods [
52]. We classified each continuous PF into five main groups, which are: most suitable, suitable, moderately suitable, less suitable, and not suitable. The classes in the discrete PFs are assigned to have the same five groups (i.e., most suitable, suitable, moderately suitable, less suitable, and not suitable). The weights of the five main groups are 1, 3, 5, 7, and 9, where the not suitable is 1 and the most suitable is 9.
3.1. Geological Factors
Geological factors significantly influence the seepage rate and flow direction of the leakage. Therefore, low permeability geological units were selected to mitigate the contamination risk resulting from the leakage [
20]. According to [
53], the study area is located within the Unstable Shelf (i.e., Imbricated Zone (IZ), and the High Folded Zone (HFZ)) and the Zagros Suture Zone (ZSZ) of the Zagros orogenic belt. This belt extends approximately 2000 km long, and trends in NW-SE direction from southern Iran through Iraq to SE Turkey [
54,
55,
56,
57,
58,
59].
We used the following three geological factors, which are lithological units, distance to active faults, and soil types. (1) Lithological units were obtained by scanning, georeferencing, and digitizing 1:250,000 scale geological quadrangle map of Sulaymaniyah [
60] and Khanaqin [
61]. The compiled map of the study area includes 21 lithological units (
Figure 2A). The ZSZ consists of five units, which consist of limestone, shale, radiolarian cherts, conglomerates, and basalt, ranging in age from Triassic to Late Cretaceous. The Unstable Shelf (USh) includes 11 units. These units mainly include limestone, besides minor amount of dolomite, marl, claystone, shale, conglomerate, siltstones, sandstones, gypsum, and bitumen, were deposited during the Triassic and the Middle Miocene periods. The ZSZ and USh are overlain by five types of Quaternary sediments. The Quaternary sediments include alluvial fans, depression fill, slope debris, and flood plains deposits.
(2) Distance to active faults, where faults are potential pathways for fluid migration and could also be seismically active. Highly faulted areas are not suitable for landfill siting and vice versa [
1]; the LFS should be located far from the active faults [
62]. We obtained the active faults by digitizing the two series of 1:250,000 scale above mentioned geological maps [
60,
61]. The TRB includes 148 fault segments. Eight of them are normal faults, while the rest are thrust faults (
Figure 2B). The total length of faults is ~161.5 km but most of them are <2 km in length. The major directions of the faults are NNW-SSE and WNW-ESE (
Figure 3A). The distance to the faults reaches 34.35 km. Several studies recommended that the landfill could not be located within 500 m of active faults [
3,
38,
63,
64].
(3) Soil types were extracted from the Harmonized World Soil Database (HWSD) [
65], which consist of a 1 km raster image. The LFSs should be located in areas of low permeability soil [
66] to prevent water entering the landfill from carrying dissolved polluted materials, causing serious contamination of groundwater system. Three types of soil are occurred in the TRB: leptosols, vertisols, and calcisols (
Figure 2C). The leptosols (thin soil with predominantly high infiltration rate) and calcisols (thick soil with predominantly moderately infiltration rate) are loamy soil, while the vertisols is thick light clay soil with predominantly low infiltration rate [
67,
68] (
Figure 3B).
3.2. Topographic (Morphological) Factors
We used three topographic factors, which are elevation, slope gradient, and Topographic Position Index (TPI).
(1) Elevation factor: high elevation lands are most suitable for LFSs than the low elevation lands in terms of flooding potential [
70]. But the drawback is high cost of MSW transportation in the high lands due to high runoff erosion, unstable slopes cuts for roads, all requiring frequent maintenance [
71]. In addition, the high lands maybe represent groundwater recharge zones [
72]. The range of elevation in TRB is between 423 and 2615 m (
Figure 4A). The highest suitability rank is assigned to moderate elevation lands (
Table 3).
(2) Slope gradient data were extracted from DEM. The significance of the slope gradient is in evaluating the stability of slopes and landslide potential and road failures for construction and operation of landfill. Lands having gentle slopes are more suitable than lands with a steep slope for landfill siting [
71]. The pixels with slope > 20° are unsuitable for MSW landfill [
8]. Slope gradients in TRB range between flat and 76.6° (
Figure 4B). We determined the slope gradient pixels in the 2°–12° [
3,
73] and 12.1°–15° range to be most suitable for landfill location. The horizontal, steep, and very steep slopes areas are moderate, less suitable, and not suitable sites, respectively (
Table 3).
(3) TPI, as a PF for landfill, is calculated using Equation (1) [
74], which calculates the fluctuation between the pixel elevation and the surrounded pixels average elevation using a pre-specific kernel-matrix (
M).
There is a relationship between the slope curvature and water infiltration to the groundwater where water infiltration increases with the concave up and decreases with the concave down. However, the slope curvature function considers a few pixels around the central point of the used Kernel (three by three pixels) [
75], which exhibits higher errors in terms of the occurrence of local depressions. Thus, some areas seem to be locally suitable for landfilling while, in fact, it is unsuitable for large-scale LFS selection. To overcome this challenge and propose more accurate sites to be used as landfilling, we proposed the use of TPI, which considers larger area and avoid local curvature.
Negative TPI means that the central pixel has elevation lower than the average surroundings pixels, while positive TPI means that the central pixel has an elevation higher than the average surroundings pixels [
52]. We computed TPI for the study area in ArcGIS software using a moving window of 50 pixels [
52]. The range of TPI in TRB is between −274.1 and 311.9 m (
Figure 4C). The highest rank of suitability is assigned to the areas within moderate topographic positions (i.e., 20.1–75.2 m;
Table 3).
3.3. Hydrogeological and Hydrological Factors
MSW landfills are a significant cause of groundwater pollution, so the depth to groundwater surface at a LFS is a very critical. The depth to groundwater in the study area is more than 10 m, which is suitable for landfilling [
76]. The available data of 243 boreholes obtained from Sulaymaniyah Groundwater Directorate is used [
77] to generate depth to groundwater map, the depth ranges between 0 to 159.4 m (
Figure 5A). In addition, data of surface water and springs (which will be mentioned below) as a zero-groundwater level are used. The groundwater in TRB classified as a fresh water [
78], the TDS content varies between <500 to 900 mg/L. Previous studies, such as [
19,
79,
80,
81,
82,
83,
84], have revealed that the ground and surface waters in the area are polluted with many organic and inorganic contaminants, and the problem is getting worse due to the prevailing draught condition in the area.
Water bodies and surrounding areas cannot be used as sites for MSW LFS due to high potential of direct contamination. [
3,
38,
85] stated that 500 m distance around the surface water is a fair buffer, while [
76] thought that the buffer zone should be >5000 m. Both the Darbandikhan Lake (with an area of 66.86 km
2) and the Tanjero River, located in the TRB, contain potable water with the TDS varying from 271 to 412 ppm, respectively [
86]. The distance to water bodies is up to 59.7 km (
Figure 5B), with water level of 485.06 m a.s.l.; [
87]).
Distance to springs is produced using 37 springs acquired from Sulaymaniyah Surface Water Directorate [
87]. According to [
3,
88], a minimum buffer of 300 m must be used around springs to determine MSW LFSs, while [
70,
89] stated that a MSW landfill should not be sited <500 m. The distance to springs areas in TRB reaches 28.9 km (
Figure 5C).
3.4. Socio-Economic Factors
This factor is included to evaluate potential impacts from landfill siting and to minimize economic and aesthetic deterioration. We used four socio-economic factors, which are: distance to towns and cities (m), distance to villages (m), distance to roads (m), and distance to powerlines (m). The MSW landfill must be located at a reasonable distance far from villages, towns, and cities due to health and public concerns [
66]. [
3,
88] reported appropriate distance of the landfill from villages, towns, and cities to be >1000 m. In the TRB, the farthest pixel from cities and towns is ~14.7 km away (
Figure 6A), while the farthest pixel from villages is ~5.7 km (
Figure 6B).
LFSs should be selected to be at a justifiable distance away from the roads to avoid negative aesthetic impacts [
24]. The researcher used different values as minimum distance to the road (m) for selecting the MSW landfill. They proposed 300 m [
30,
38], 500 m [
36,
37], and 1000 m [
66,
90] as buffer distance from MSW landfill. Accordingly, we eliminated sites within 1000 m of major roads. We believe that it will provide adequate buffer zone for noise, dust, etc. created from movement of garbage trucks, without requiring construction of new access roads for transportation and collection of solid wastes [
22,
27]. The farthest point from roads in the TRB is ~28.1 km (
Figure 6C).
As [
36] suggested, we used a buffer 300 m (4) distance to the powerline (m) as non-suitable areas for MSW LFSs. The farthest area from powerlines in the TRB is ~41 km (
Figure 6D). The shapefiles of the settlements, the roads and powerlines were obtained from [
91]. We calculated the Euclidean distances from existing villages, towns, cities, roads, and powerlines minimum distances from these features to each pixel in the TRB.
3.5. Land Use Factors
We used two land use PFs: (1) land use and land cover (LULC), and (2) distance to agricultural lands.
(1) The LULC map was supplied by Iraq Geological Survey and contains nine classes (
Figure 7A). It was produced using Landsat satellite data having 30 m spatial resolution. It has been validated by fieldtrip with an overall accuracy of ~93.60% [
92]. Several land covers are present in the study area and include: water bodies, urban and built-up land, vegetated land, harvested land, cultivated land, burnt land, carbonate rocks, conglomerates and gravels, and other clastic rocks [
92].
(2) The area under forest cover must be avoided for landfill siting because it negatively affects natural forest resources [
93]. We used distance to agricultural lands (m) to avoid selecting landfill in and near the vegetation cover. We calculated the Euclidean distances from vegetation cover distances within TRB (
Figure 7B). The Normalized Difference Vegetation Index (NDVI) was used to characterize the vegetation cover. We used the equation proposed by [
94] to calculate the NDVI. It was calculated after extraction of the reflectance (ρ) from the digital number (DN) of Landsat data [
38]. proposed that 300 m distance away from agricultural and forest lands could be acceptable for locating MSW landfill.
4. Suitable Landfill Site Selection Model
There is no agreement about a specific method that can be considered to be the most suitable for all types of decision-making technique [
95,
96,
97]. A big criticism of MCDM is the fact that various methods might obtain various results if used to the same issue [
98]. The definition of a suitable MCDM approach is thus not a simple task, and the focus should be on the precise determination of the approach [
95]. Available papers show huge practical applications of comparative analyses of various MCDM approaches [
9,
11,
12,
13,
14,
15,
16,
17,
18,
99]. We employed five methods to distinguish suitable locations for MSW LFSs. These methods are BO, WSM, WPM, TOPSIS, and AHP.
4.1. Boolean Overlay (BO)
BO is a simple method, widely used to determine suitable sites for solid waste landfills [
2,
16,
17,
31,
37,
38,
100,
101,
102,
103]. It is based on reclassifying multi-factors used to select LFSs into binary values (0, 1), where 1 and 0 are suitable (yes) and unsuitable (no) pixels, respectively [
31]. We prepared a final suitability map for landfilling by combining whole-created binary maps for the constraints using the Boolean overlay (AND operation), which is defined by Equation (2) [
17]:
where A and B are constant, and X is the factor
To be on the safe side for LFS selection, we used the maximum distance for various factors (
Table 4).
4.2. Weighted Sum Method (WSM)
WSM is a straightforward MCDM method, used for solid waste LFS selection [
37,
104,
105,
106]. This method considers that all factors have equal weight, which is one of its deficiencies [
107,
108]. Where the weight of the factors equals to each other. Firstly, we categorized each factor into five categories. These are 1, 3, 5, 7, and 9 for the not suitable, less suitable, moderately suitable, suitable, and most suitable for landfill location, respectively. The weight of these five categories was identified according to the possibility of air, water, and soil contamination for the surrounding areas of the proposed landfill. The relationship between LFSs and landfill PFs is shown in
Table 3. Column “Rank” shows the final WSM ranks for the factors used to select the LFSs. Following is a summation of whole PFs using the equation proposed by [
109,
110].
where
n is the number of factors,
is the actual value of the
ith of the
jth criterion and
is the weight of the
jth criterion.
4.3. Weighted Product Model (WPM)
WPM is very similar to the WSM, the essential variation is that instead of summation in the mathematical expression there is multiplication [
109,
110]. Similar to WSM, the weight of factors equals to each other. We used the same weight of the classes used in WSM (
Table 3 column “Rank”) to select the suitable LFSs. Equation (4) suggested by Bridgman (1922) was implemented [
109].
where
n is the number of factors,
is the actual value of the
ith of the
jth criterion (
Table 3),
is relative value, and
is the weight of the
jth criterion.
4.4. Technique for Order Performance by Similarity to an Ideal Solution (TOPSIS)
The alternative of choosing the shortest distance from the ideal best solution and the longest distance from the ideal worst solution in the TOPSIS method, makes this method suitable for LFS selection [
111,
112,
113]. This method is reliable because the decision makers may desire a decision not only on the most suitable LFSs but also to avoid unsuitable sites [
113]. Following steps that have been used to implement TOPSIS method [
71,
109,
110,
111,
112,
114,
115,
116,
117,
118,
119]. As a first step, we utilized the same weight of the classes in
Table 3 (column “Rank”) to build the TOPSIS method. Then, we normalized each PF
using Equation (5).
where
is pixel value.
Furthermore, we compute the suitability value of the PF
by calculating the mean of the weights given by previous eleven papers (
Table 5), and, hence, structured the normal weighting matrix
by multiplying the normalized PF by its weight Equation (6).
The Euclidean distance from the ideal best (
) and Euclidean distance from the ideal worst (
) value for each layer were calculated by using Equations (7) and (8), respectively. The final step was accomplished by calculating the performance score (
) using Equation (9).
4.5. Analytic Hierarchy Process (AHP)
The AHP method proposed by [
123]. It measures the index weight by comparing the PFs with each other [
124]. It is one of the most common approaches applied for LFS selection. The GIS environment was used to LFSs, ratings of each PF are provided on a five-point continuous scale (
Table 3 column “Rank”). While the suitability weight of the PF was computed by calculating the mean of the weights (
Table 5). This was based on a simple review of 11 papers that have applied these PFs for LFS selection. Map of suitable sites for LFS is computed by the raster overlay algorithm, using Equation (10) [
125]:
where
is the value of PF
i [where
i = (list of PFs in
Table 5)],
is the weight for PF
i, and
n is the number of PFs (
Table 3 column “Rank”). We correlated all PFs used by normalizing their scales and units following the common equation, Equation (11). The final normalized weights were computed in
Table 3 column “Normalized weight”.
where
is normalized value of pixel,
is the value of pixel,
is the minimum value of pixel and
is the maximum value of pixel.
The resulting maps using the WSM, WPM, TOPSIS and AHP methods were grouped into five classes, which are most suitable, suitable, moderately suitable, less suitable, and not suitable for landfill. We determined the final suitable LFSs based on the average weights of the landfill probability maps. The pixels that have average ranking ≥ moderate suitable were selected as suitable sites for landfill.
5. Results
Besides the BO map, which shows the suitable landfill locations within the TRB (
Figure 8), we generated four suitability maps for LFSs using WSM, WPM, TOPSIS, and AHP methods in the ArcGIS environment. In BO maps, the suitable landfill locations are presented in red color while the unsuitable locations appear in beige.
Figure 8A shows suitable landfill locations after combining all normalized binary maps using BO conditions, which are stated in
Table 4, while
Figure 8B exhibits the results without the distance to agricultural lands condition. Nearly all suitable sites are placed in the center and to the east of the TRB (
Figure 8A). The suitable locations represent 0.13% and 4.33% for the BO with and without agricultural lands condition, respectively.
The final spatial distribution for LFSs probability maps based on the WSM, WPM, TOPSIS, and AHP models were developed using 15 PFs. The weights of the PFs have been estimated by using the prediction model Equations (3), (4) and (7)–(10). For each method, each PF has a specific predictive weight, which differs from one model to another. Wide range of the predictive weights means high effectiveness of these factors for LFS selection. We classified the LFS selection map into five groups using equal intervals. We used the frequency threshold levels (i.e., 20, 40, 60, and 80%), representing “Not suitable”, “Less suitable”, “Moderate suitable”, “Suitable”, and “Most suitable”.
The distributions for LFS suitability of WSM and AHP are very close to each other (
Figure 9A,D). The WSM, TOPSIS and AHP landfill maps exhibit that their spatial distributions are somewhat similar, where more than 96.6% of “very high” and “high” probability classes are shared between these three methods. For the “very high” and “high” probability classes, the similarity between WPM and other methods is less than 11%. For these three models the suitable and most suitable areas for LFSs are located close to the watershed boundary of the TRB and those for “not suitable” and “less suitable” areas are placed in the center of the TRB (
Figure 9 and
Figure 10A).
We calculated average map of the WSM, TOPSIS, and AHP models (
Figure 10A). The WSM has been neglected from our consideration (See
Section 6.2). Based on
Figure 10A, eight best sites for landfill have been selected. Several areas appeared as suitable sites for LF, but most of them have small areas.
Figure 10B and
Table 6 show the locations of the eight suggested LFSs, with average landfill suitability weight ≥70% for the WSM, TOPSIS, and AHP models. These suitable sites are placed in the western part of the TRB, which have total area of 18.35 km
2.
Figure 11 shows the relationship between the suitability of the suggested sites and the PFs. All the eight suggested sites are lying out of [
19] studied area. They are located in suitable lithological units and far from springs, with lower suitability “but acceptable” distance to the villages. The most suitable landfill location is Site-6 while the least suitable location is Site-5.
7. Conclusions
The main aim of this paper was to recognize suitable landfill sites (LFSs) in the Tanjero River Basin (TRB) in the Kurdistan region, Iraq. In the current study, Boolean Overlay (BO) in addition to four Multi-Criteria Decision-Analysis (MCDA) models included Weighted Sum Method (WSM), Weighted Product Method (WPM), Analytic Hierarchy Process (AHP), and Technique for Order Performance by Similarity to an Ideal Solution (TOPSIS) were applied to enable combined the use of the 15 thematic layers. The distribution maps for LFSs probability from WSM and AHP are very close to each other; while the WSM, TOPSIS and AHP landfill results exhibit that their spatial distributions are somewhat similar; while the similarity between WPM and other methods is less than 11%. The accuracy of all methods was calculated, and the best accuracy was achieved by AHP method.
According to the results, the final suitable LFSs were identified by calculating average weights of the WSM, TOPSIS and AHP maps. Accordingly, the pixels weights that have suitable and very suitable ranks have been nominated for landfill.
To sum up, based on the final analyses, most of the suitable sites are located close to the TRB boundary. Eight suitable sites have been identified, that have the best condition for citing MSW landfills. These sites are situated in the western part of the TRB, and the most suitable site is Site-6 and the less suitable is Site-5. According to this research, the current location of the Sulaymaniyah dump is not suitable and its location might lead to pollution in the area. It is worth noting that ours is the first study to have used the Topographic Position Index (TPI) to select MSW LFSs. From 12 borehole dug in the TRB, a validation shows that the best model is AHP, where it has an inverse strong relationship (R2 = 0.78) with the TDS model. Finally, we recommend that the most suitable site among the eight determined sites for choosing landfill should be based on detailed on-site surface and subsurface investigations.