Fp

RPC

a

ARR2A

KGSUG

1

die2etidi

(

1h

Ecological Indicators 37 (2014) 151– 160

Contents lists available at ScienceDirect

Ecological Indicators

jou rn al hom epage: www.elsev ier .com/ locate /eco l ind

actor analysis and geographic information system for determiningrobability areas of presence of illegal landfills

osa Jordá-Borrell ∗, Francisca Ruiz-Rodríguez, Ángel Luís Lucendo-Monederohysical Geography and Regional Geographic Analysis Department, Geography and History Department – Seville University [Universidad de Sevilla],/María de Padilla s/n, 41004 Seville, Spain

r t i c l e i n f o

rticle history:eceived 19 March 2013eceived in revised form3 September 2013ccepted 1 October 2013

eywords:eostatistical and geospatial methodologypatial patternsncontrolled/illegal landfillseographical factors

a b s t r a c t

The objective of this study is to develop a methodology for determining areas in which there is a dis-tinct probability of the presence of illegal landfills. This methodology is developed in three stages: (a)the application of factor analysis (FA) to identify relevant geographical factors (factor model); (b) theconstruction of a geostatistical model to calculate spatial patterns based on the identified factors; and (c)the integration of the geostatistical model into a geographic information system (GIS) to determine andlocate the illegal landfill sites (spatial model).

This methodology has proven to be valid because it confirmed that a verified population of illegallandfills (518) is not randomly distributed; instead, most of the illegal landfills (63.6%) are found in theareas of highest probability (over 36%). Additionally, the study confirmed that the application of thismethodology (FA and GIS) provides adequate results at the regional and local level.

The described method may also be applied to other spatial environments, as long as the necessarythematic and spatial data are available (although results would vary according to demographic, socio-economic, geomorphological, and environmental management characteristics).

Finally, the benefit of this methodology lies in the fulfilment of two necessary and sufficient demands.(a) The model does not arbitrarily include variables related to the probability of the presence of illegallandfills and considers those variables that have been shown via FA. (b) The variables included in the

spatial model are not considered to have the same importance.

Thus, the integration of FA and GIS offers an alternative tool to the application of multi-criteria eval-uation as this approach determines the criteria and their relative weights based on substantiated andnon-aprioristic indications. Moreover, the methodology used in this study enables the creation of mod-els because the GIS makes an excellent platform for the development, application, and validation of thesemodels.

. Introduction

The existence of illegal landfills continues to be a problem in theeveloped and peripheral countries of Europe. Research conducted

n Germany, Austria (Allgaier and Stegmann, 2006), Ireland (Doakt al., 2007), France (Biotto et al., 2009), Italy (Silvestri and Omri,008), Romania (Apostol and Mihai, 2011) and Serbia (Vasiljevict al., 2012) demonstrates the severity of this problem and indicateshat the analysis of illegal landfills is complicated. This complexity

s due to the lack of consistent and homogenous data (inventories,atabases, official statistics, satellite images. . .); thus the character-

stics and the spatial distribution of illegal landfills have been little

∗ Corresponding author. Tel.: +34 620191 928.E-mail addresses: borrell@Us.Es (R. Jordá-Borrell), fruiz@Us.Es

F. Ruiz-Rodríguez), angeluis@Us.Es (Á.L. Lucendo-Monedero).

470-160X/$ – see front matter © 2013 Elsevier Ltd. All rights reserved.ttp://dx.doi.org/10.1016/j.ecolind.2013.10.001

© 2013 Elsevier Ltd. All rights reserved.

studied. Furthermore, most landfill studies are centred on verifyingappropriate variables for locating controlled landfills.

In both types of studies a methodology integrating geographicinformation systems (GIS) and multi-criteria evaluation techniquesis applied. The aforementioned research identifies the factors andcriteria required to calculate the probability of an illegal landfilloccurring within a given territory and demonstrate that these land-fills are not randomly distributed in spatial terms (Doak et al., 2007;Silvestri and Omri, 2008; Biotto et al., 2009).

The objective of this paper is to develop a methodology foridentifying and locating areas with different probabilities of thepresence of illegal landfills. This methodology consists of threestages: (a) the application of factor analysis (FA) to identify geo-

graphical factors (factor model), (b) the definition of decision rulesfor calculating spatial patterns based on the factors (geostatisticalmodel) and (c) the integration of the geostatistical model into ageographic information system (GIS) (spatial model).

1 ical In

yaGitfDAaCbco1yluf

ri(lItplLcsp(a

2u

2

aaTd

2ei(sBKb

•

SK

a

For this purpose, we selected a series of vector coverages usingdatabases from different sources (Table 2) with scales rangingbetween 1:10,000 and 1:100,000.

3 The fieldwork comprised interviews with relevant experts and managers (localand regional government) and information gathering during visits to 110 illegallandfills distributed throughout the regional territory with the goal of identifying,validating, and gathering information on the illegal landfill terrain for the repre-sentative sample. For this purpose, a file of 45 variables was created, adjusting the

52 R. Jordá-Borrell et al. / Ecolog

This work, therefore, focuses on the development of a similaret alternative methodology to that used by the above-mentioneduthors. The methodology that we propose is similar because it usesIS as the basis of the entire process of spatial analysis and for locat-

ng probable landfill sites on the basis of the criteria that determinehese areas. And alternative because, among the existing methodsor obtaining the criteria (Analytic Hierarchy Process, AHP, or theelphi Method1), we have used the Principal Component Factornalysis (FA) with the goal of obtaining these criteria2 or factorsnd their relative importance or rank (eigenvalues). According tohang et al. (2008) and Kontos et al. (2005) the problem causedy the presence of illegal landfill sites is influenced by a series ofomplex factors that are not known a priori. An appropriate meansf identifying these factors is the use of AHP (Barredo and Bosque,995) or FA in order to subsequently define a spatial pattern anal-sis that specifies areas where the presence of illegal landfills isikely through the construction of a GIS (Barredo, 1996). Thus, these of FA as a criterion or factor selection method requires that theactors have normal characteristics.

The application and verification of this methodology was car-ied out with data from Andalusia (Spain) (Fig. 1). It must be takennto account that this region has a surface area of 87,598 km2

17.4% of Spain and 3.7% of the European Union) and a popu-ation of 8,437,681 inhabitants (INE Spanish National Statisticsnstitute, 2012) which comprises 17.87% of the Spanish popula-ion. The Autonomous Community of Andalusia is divided into eightrovinces and 770 municipalities of which some 150 have a popu-

ation above 10,000 inhabitants and contain 80% of the population.ocated in southern Europe at latitude 36–38◦45′ N, Andalusia isharacterized by its Mediterranean nature which gives it a con-iderable variety of landscapes and resources. From an economicerspective, it is one of the most backward regions of EuropeD 17,500 GDP per capita at current prices), and is far below the EUverage (D 23,500 500 per capita in 2009, according to EUROSTAT).

. Methodology for calculating probable areas ofncontrolled landfills

.1. Data, variable selection, and factor analysis

The object of analysis in this paper is the illegal landfill. This is waste deposit in an area without monitoring by the competentuthorities, and not the result of previous waste treatment systems.he Regional Government considers an illegal landfill to be whenumping covers an area of 2000 m2 or more.

The references on illegal landfills (Doak et al., 2007; Biotto et al.,009; Silvestri and Omri, 2008; Apostol and Mihai, 2011; Tasakit al., 2007, 2004) have used many geographical variables for stud-es to identify possible uncontrolled landfill sites. These variablesalways geo-referenced) match those selected by other authors totudy the optimal location of a landfill using GIS (Chang et al., 2008;asnet et al., 2001; Akbari et al., 2008; Sener et al., 2011, 2010;ontos et al., 2005). These variables are different in nature and cane grouped into:

Geophysical variables, such as geomorphological and lithologicalcharacteristics, vegetation, elevation, hydrology, types of land useand protected areas.

1 The Delphi method has mainly been used in studies on forest fires (Meddour-ahar et al., 2013), and in the formulation of pollution indices (Sharma et al., 2008;umar and Alappat, 2005).2 Criterion is a factor derived from factor analysis and has an eigenvalue above 1,

nd therefore in this study, factor is the same as criterion.

dicators 37 (2014) 151– 160

• Management and activity variables, such as accessibility and vis-ibility of illegal landfills, type of waste, administrative efficiency,existence of punitive policies and environmental culture.

• Socio-economic variables, such as the resident population, rent,road infrastructure and economic activities.

Fieldwork and interviews carried out with (16) technical expertsin waste management suggested the inclusion of new types ofvariables. This is the case of: distance measures of the physicalproximity of these landfills to different surrounding geographicalelements, references to soil use, the number of records opened bythe administration and the time dedicated to monitoring the ter-ritory. Therefore, due to the above reasons and the multivariatenature of the analyses for the determination and location of illegallandfills, we also believe that it is necessary to begin with the selec-tion of a wide and diverse enough set of variables to obtain a set ofrepresentative geographical factors.

It is important to stress that data for most of these variablesare not currently available. Thus, it has been necessary to conductfieldwork3 to obtain this information. Related social and economiccharacteristics of the municipalities in which are located and dis-covered illegal discharges that come from official statistical sources(Table 2) are appended to these data, given the lack of availabilityof these data for illegal dumping areas. In the end we selected atotal of 78 variables to perform statistical analyses (supplementarydata).

Supplementary material related to this article can befound, in the online version, at http://dx.doi.org/10.1016/j.ecolind.2013.10.001.

To do fieldwork, we calculated a sample of 110 cases dis-tributed across the eight Andalusian provinces in accordance withthe Nomenclature of Territorial Units for Statistics 3 (NUTS 34)These cases were selected through stratified sampling (size, dis-charge type, spatial distribution) from a population of all the illegallandfills detected via remote sensing and aerial photography (1751possible illegal landfills supplied by the Ambisat Company5). Aftercleaning up the data using different techniques and validating thedata via fieldwork, a population of 518 illegal landfills was deter-mined (Table 1).

When combining FA with the GIS techniques, it is necessary touse the same variables for the GIS as those used with the multi-variate technique but with geo-referenced information. This allowsto create a map for each variable distribution. We may performnot only pertinent spatial analysis using GIS to map the factorsdetermining the location of illegal landfill sites in Andalusia butalso analyze how different variables in the region are interrelated.

reference images (topographical and orthophotographic maps) and the photos takenin situ. Using these data, we obtained 78 variables (Supplementary data) to whichothers were added from the information sources listed in Table 2, with the goal ofcreating different databases for conducting the statistical analysis and building theGIS.

4 Level 3 (NUTS 3) in the Spanish state corresponds to the administrative divi-sion of the province. The Spanish State is organized territorially into municipalities,provinces and autonomous regions.

5 This research study on remote sensing as a tool for identifying landfills, namedthe VERTEL project, was conducted by the signatories to an R&D contract with theAmbisat Company (Madrid). The study was financed by the Andalusian TechnologyCorporation (CTA) in 2011 and 2012.

R. Jordá-Borrell et al. / Ecological Indicators 37 (2014) 151– 160 153

Atl

an

tic

Oc

ea

n

Medite

rranean S

ea

MOROCCO

PORTUGAL

ANDALUSIA

FRANCE

MADRI D

Fig. 1. Andalusian situation.

Table 1Illegal landfills by province (NUTS 3).

Province General total Sample

No. of landfills No. landfills per10,000 inhabitants

No. landfills per 1 km2 No. of landfills No. landfills per10,000 inhabitants

No. landfills per 1 km2

Almeria 147 2.09 59.7 31 0.44 283.0Cadiz 28 0.23 265.6 5 0.04 1487.2Cordoba 58 0.72 233.6 12 0.15 1129.2Granada 47 0.51 266.6 9 0.10 1392.3Huelva 86 1.65 118.0 18 0.34 563.8Jaen 57 0.85 236.6 12 0.18 1124.1

(tvfmv

yX

st

Malaga 44 0.27 166.1

Seville 51 0.26 275.3

Andalusia 518 0.61 168.5

For the FA6, we obtained 34 highly interrelated variablesTable 4), which constitute nine factors explaining 76.88% of theotal variance and showing high communality for almost all theariables. We used the method of principal components analysisor factor extraction and the quartimax method with Kaiser7 Nor-

alization for rotation. The resulting Kaiser-Meyer-Olkin (KMO)alue was 0.736, and the determinant was 1.41E-019 (Table 3).

6 The FA is published in Jordá et al. (2013): “Factores territoriales de localización caracterización de los vertederos incontrolados en Andalucía”, Scripta Nova Vol.VII (435), http://www.ub.edu/geocrit/sn/sn-435.htm.7 As the variables with larger communalities have a greater influence on the final

olution, it is convenient to apply Kaiser Normalization which consists in dividinghe square of each loading factor by the communality of the variable.

9 0.06 812.010 0.05 1404.2

110 0.13 793.4

The application of this technique has enabled us to show thefollowing: (a) which territorial variables affect the delimitation ofareas where illegal landfills are present (Andalusia); (b) what vari-ables are interrelated; and (c) that the importance or weight thateach factor has in the formation of these areas is indicated by thepercentage value of the variance expressed. Therefore, the obtainedfactors may be later used in a geostatistical model to evaluate theprobable presence of illegal landfills for each zone in Andalusia(Fig. 2).

2.2. Building a geostatistical model for calculating spatial

patterns (SP)

As a mathematical model, the results of the FA provide nei-ther the specific location of the areas defined by the different

154 R. Jordá-Borrell et al. / Ecological Indicators 37 (2014) 151– 160

nts of

ftiat

Fig. 2. Major geographical eleme

actors nor their size or shape. Consequently, it is convenient

o construct a geostatistical model to convert the factor modelnto concrete spatial information suitable for incorporation into

GIS. Thus, we may map and calculate the different areas wherehere is a distinct probability that illegal landfills will be located,

the illegal landfills environment.

on the basis of the factors identified by the FA in Andalu-

sia.

The geostatistical model is the quantitative decision rule that weuse to obtain a spatial value for calculating the magnitude of eachfactor (compensatory method). We calculated this value, which

R. Jordá-Borrell et al. / Ecological In

Table 2Thematic and spatial data sources.

Information type Source

Non-spatialInformation gathered during fieldworkData from the Multi-territorial InformationSystem of Andalusia (SIMA) compiled by theInstitute of Statistics and Cartography ofAndalusia. Regional Government of Andalusia(Spain)

SpatialCadastral information, General Directorate ofLand Use, Economic and Treasury Ministry.Government of SpainSpatial data from Andalusia (DEA100). Instituteof Statistics and Cartography of Andalusia.Regional Government of Andalusia (Spain)Andalusian Environmental Information

w(t

aamuwlp

S

wcvmq

V

w(tio

tI

vii) Factor 7 SP. Factor 7 refers to lithological and geomorphological

TT

Network (REDIAM). Environmental Council.Regional Government of Andalusia (Spain)

e call the spatial pattern based on median statistical parametersmean and mode). The SP values are representative of the size andype or dominant characteristic of each factor.

This geostatistical model is based on the contributions of Pozand Fernández (2010), Castro (2009), and Castano (2011). Theseuthors show different methods for building complex indices usingultivariate statistical analysis techniques. More importantly, they

se a principal components FA. Building on their contributions,e developed a formula for weighing the factors by adapting the

inear weighted sum model (Galacho and Ocana, 2006). Thus, theroposed geostatistical model may be expressed as:

Pi =n∑

f =1

(Efi vfi)

here SP is the spatial pattern for each factor, n is the number ofases, i is the value of each variable, factor or pattern, Efi is the eigen-alue (weight) of each factor, and Vfi is the average value (mean orode) of each variable making up each factor for generating the

uantitative value. This value is calculated as follows:

fi = meanmode

[(wf 1 × V1) + (wf 2 × V2)

+ (wf 3 × V3) + . . . + (wfi × Vi) + Remainder fi]

here wfi is the factor weight for each variable of each factorappearing in Table 4), Vi is the value of each variable composinghe factors for all the cases (110 illegal landfills), and Remainder fis the sum of the factor weights for the variables that are not a partf the factor model found.

To calculate the average value of each factor (Vfi), different sta-istical parameters were used according to the nature of the data.n the case of factors made up of metric variables, the mean was

able 3otal explained variance.a

Component/designation Initial eigenvalues

Total

F1. Socio-economic 8.95

F2. Land use 4.46F3. Proximity to municipal seat 2.304

F4. Waste type 1.938

F5. Management and accessibility 1.679

F6. Distance from highways and industrial zones 1.604

F7. Lithology/geomorphology 1.353

F8. Proximity to secondary roads 1.281

F9. Proximity to residential areas and waterways 1.035

a Utilizing the application on Statistical Package for the Social Sciences (SPSS), version

dicators 37 (2014) 151– 160 155

used, whereas for factors with categorical variables, the mode wasused.

The geostatistical model includes the weight of each of thefactors previously identified to rank the SPs. These weights wereassigned to each of the areas generated by these patterns. Applyingthis model to the factors, we obtained the spatial pattern for eachfactor [F1–F9]. Based on these patterns, we can calculate the areasbelonging to each factor using GIS (Table 4).

These SPs are as follows:

i) Factor 1 SP. Factor 1 defines the “socio-economic” pattern anddetermines that municipalities with the highest populationand socio-economic level will have the highest probability ofcontaining illegal landfills. After the application of a geosta-tistical model, this SP provides an index for each municipalitythat indicates the probability of the presence of illegal landfillsaccording to the “socio-economic” factor.

ii) Factor 2 SP. This pattern determines that the presence of ille-gal landfills is related to the “cadastral” (land use) factor. Itobtained a SP based on land use categories that are most oftenused as illegal landfills and described as “agrarian”, “unculti-vated”, and “rustic”.

iii) Factor 3 SP. The third pattern is associated with two metricvariables (distance to streets and to municipal seats) and onenon-metric variable (proximity to urban centres). The val-ues for calculating this SP are a mean distance to streets of1542.41 m and to municipal seats of 1594.09 m. For the “prox-imity to urban centres” variable, the most repeated value isthat illegal landfills are “close to urban centres”.

iv) Factor 4 SP. This pattern indicates that the majority of ille-gal landfills store, in general, “urban” and “construction anddemolition waste” (C&DW) and that the greater the surfacearea of the municipality, the greater the number of illegal land-fills. This SP determines a municipal index (which we havenamed the “waste generating capacity”) whose value is drawnfrom the size and number of illegal landfills found in eachterritory.

v) Factor 5 SP. The fifth SP pattern is defined by the type of “man-agement and accessibility” and indicates that illegal landfillareas are characterized by “easy access”, are found “withoutsupervision” and their visibility from the closest rural road ishigh (from 0 to 500 m).

vi) Factor 6 SP. This SP shows a probability for the presence of anuncontrolled landfill in those areas that are “far from highwaysand from industrial zones”. Specifically, the model enables acalculation of mean distances from these illegal landfills of 1.7and 1.3 km to highways and industrial zones, respectively.

units where illegal landfills appear. The values representativeof the pattern are “limestone and detrital materials”, “terracesor plains”, and “hillsides”.

Extraction sums of squared loadings Rotation sums of squared loadings% Variance % Accumulated

27.97 27.9713.98 41.907

7.199 49.1066.058 55.1645.246 60.4095.013 65.4224.229 69.6514.002 73.6533.235 76.888

18 (IBM SPSS Statistics).

156R

. Jordá-Borrell

et al.

/ Ecological

Indicators 37 (2014) 151– 160

Table 4Synthesis of the methodology developed.

Factor model Geostatistical model Spatial model

Factors (1st level ofrank)

Weight (%) Variables (2nd level ofrank)

Weight(correlation)

Statisticalparameter

Spatial pattern Spatial parameters (GIS)

Coverage/Analysis/Maps(partial)

Map/final

Socio-economic 36.3 PopulationRentService businessesConstructionbusinessesWasteTaxes (euros)IndustriesUrban wasteIllegal Landfill sizeInspection timeAgricultural businesses

0.9910.9870.9850.9830.9780.9770.9510.7900.7640.590

Mean Socio-economic weight(municipal index)

Municipality (DEA100)

Spatial join data

FACTOR 1ZONES

Spatial join analysis

Zones with probabilityof presence ofuncontrolled landfills

Land use 18.2 Property useType of cropType of propertyType of farm

0.919−0.8650.8640.838

Mode • Rural land• Uncultivated agrarianland

Cadastral (Land use)Select analysisFACTOR 2 ZONES

Proximity to municipalseats

9.4 Distance to streetsDistance to municipalseatsProximity to urbancentres

0.8460.754−0.708

MeanMode

≤1.5 km to streets≤1.6 km to municipalseats

Urban areas (DEA100)Buffer analysisFACTOR 3 ZONES

Main waste type 7.9 Construction anddemolition wasteUrban wasteNumber of illegallandfillsArea of municipality

0.8110.7580.7200.699

ModeMean

Waste generatingcapacity (municipalindex)

Municipality (DEA100)and Illegal LandfillsSpatial join dataFACTOR 5 ZONES

Management andaccessibility

6.8 VisibilityAccessManagement

0.8150.7240.723

ModeMedian

High visibility andaccess≤0.5-km distance

Illegal LandfillsBuffer analysisFACTOR 5 ZONES

Distance fromhighways andindustrial zones

6.5 Distance to highwaysDistance to industrialzones

0.8850.750

Median ≤1.7 km to highways≤1.3 km to industrialzones

Roads, Land use,Industrial states(DEA100)Buffer analysisFACTOR 6 ZONES

Lithology-geomorphology

5.5 Soil (lithology)Soils (geomorphology)

0.8770.795

Mode • Limestone anddetritus materials• Terraces or fields andhillsides

Geomorphology(REDIAM)Select analysisFACTOR 7 ZONES

Proximity to secondaryroads

5.2 Distance to roadsDistance to green areasDistance to paths

−0.8180.7180.675

Median ≤0.3 km to roads≤3.3 km to green areas≤0.6 km to paths

Roads, Paths, Naturereserves (DEA100)Buffer analysisFACTOR 8 ZONES

Proximity to residentialareas and waterways

4.2 Distance to residentialareasDistance to waterwaysAltitude

0.7990.7340.603

Mean ≤1.3 km to residentialareas≤0.6 km to waterways172.6-m altitude

Urban areas, Rivers,Altitude (DEA100)Buffer analysisSelect analysisFACTOR 9 ZONES

ical In

v

2

rmi(Tdb(7aa

cdcAw

tlws

•

•

•

•

•

alt

•

a

are found in municipalities with net rents above the Andalusianmean (almost 76 million). The high rent is clearly related to thesubstantial demographic and economic growth observed in aver-age Andalusian metropolitan areas and median cities. This rent

Table 5Number and percentage of illegal landfills in the sample by areas established accord-ing to their probability of containing illegal landfills.

R. Jordá-Borrell et al. / Ecolog

iii) Factor 8 SP. According to this factor, most illegal landfillsare found in “proximity to roads and paths” and “far fromgreen areas” (natural reserve, park). In calculating this SP, themean distance is 278.4 m to roads, 3.28 km to green areas, and625.4 m to paths.

ix) Factor 9 SP. This factor, designated “proximity to residentialareas and waterways”, indicates that the higher the terrainelevation and the greater the distance to a residential area andto a waterway, the lower the probability of finding an uncon-trolled landfill. This SP shows a mean distance of 1.36 km toa residential area and 607.62 m to a waterway and a meanaltitude of 172.6 m.

.3. Representation of probable areas of illegal landfills via GIS

The final step in the proposed methodology involves transfer-ing the geostatistical model to the territory by building a “spatialodel”. In this stage, the different SPs are represented cartograph-

cally using ArcGIS version 9.3, a GIS software package from ESRIEnvironmental Systems Research Institute, Redlands, CA, USA).he objectives are to calculate the areas that are defined by theifferent SPs, based on the factors, and to assign the SPs a proba-ility value for the presence of illegal landfills. As the total varianceweight) explained by the nine factors from the factor model was6.88%,8 we used the 76.88% value as the maximum probability on

scale from 0 to 100% for calculating the probability for each of thereas (Table 4).

To carry out this stage, we have used vector-based coveragesontaining spatial information related to the variables and factorsefined via FA and later converted to the SPs using the geostatisti-al model. Different analyses were applied using these values andnalysis Tools in ArcGIS 9.3 to calculate, for each factor, the areasith the probable presence of illegal landfills.

For the distance patterns (factors 3, 5, 6, 8, and 9), a “buffer”ype proximity analysis was first used for each one of the entities,eading to new polygonal coverages. Once these proximity areas

ere created for each entity, the areas were overlaid to obtain aingle area for each criterion using a “spatial join” analysis (Table 4).

The buffers for streets and municipal seats (pattern 3) generatedzones that, when joined, have a 9.4% probability of the presenceof illegal landfills.The buffer created for uncontrolled landfill visibility (pattern 5),especially from rural roads, generated areas with a 6.8% proba-bility.The buffers for highways and industrial zones (pattern 6) led toareas that together share a 6.5% probability.The buffers for roads, green and protected areas, and paths (pat-tern 8) produced areas with a combined probability of 5.25%.The areas generated by buffers created with mean distances toresidential areas and waterways (pattern 9) have a combined 4.2%probability.

In the case of patterns based on categories of factors (factors 2nd 7), the areas where the presence of uncontrolled landfills isikely were obtained by creating new polygonal entities based onhe “select analysis” tool (Table 4).

According to pattern 2, the areas in Andalusia with uncultivatedagricultural land have an 18.2% probability of the presence ofillegal landfills.

8 In the social sciences, it is considered sufficient for the factor model to explaint least 60% of the total variance.

dicators 37 (2014) 151– 160 157

• Pattern 7 shows that limestone and detrital soils together have a5.5% probability of the presence of illegal landfills.

Patterns 1 and 4 determine the probabilities for the presenceof illegal landfills for each municipality according to the indexgenerated by the geostatistical model and the weight of these pat-terns. Thus, because pattern 1 contributes a probability of 36.4%,each area or municipal surface would have a probability value thatvaries from a minimum of 0% for municipalities (with a lowerdemographic and economic weight); to a maximum of 36.4% formunicipalities with higher weights for these variables. Pattern 4assigns a maximum probability of 7.9% to each municipal area (tothe largest and those with the greatest number of uncontrolledlandfills) and a minimum of 0% to small municipalities withoutillegal landfills (Table 4).

Once the aforementioned analyses have generated probabilityareas for each pattern, we combine the patterns using “spatial join”analysis. This process produced a map (Fig. 3) that shows the totalprobability or Index of the Presence of Uncontrolled Landfills (IPUL)in each area of Andalusia.

3. Results

With the above methodology, we have used GIS to calculate theprobability of the presence of illegal landfills in each area of Andalu-sia (Fig. 3). The final objective of this method is to join (overlayon GIS) all the partial areas belonging to each factor according tothe results of the FA and the geostatistical model. Because SP 1 isobserved in all areas and has the highest weight, it becomes a keyelement for defining the limits of the probability intervals. Thus,we have established three probability intervals: (i) “low probabil-ity” less than or equal to 12% (one third of the probability value ofpattern 1); (ii) between 13 and 36% of probability (the maximumvalue of pattern 1); and (iii) greater than 36% (the value of pattern1 or different combinations of 1 with other factors).

As a result of this methodology, we made a map (Fig. 3) thatprovides accurate estimates of the areas with the greatest proba-bility of such illegal sites occurring. The uncontrolled landfills arenot located randomly or concentrated in low probability areas butare instead mostly found in the high probability areas (where 63.6%are found, almost two out of every three) (Table 5) (Fig. 3).

These results are corroborated by the validation of each one ofthe SPs configuring the probability areas.

Regarding the factor 1 SP, the validation procedures demon-strate that over 53% of the uncontrolled landfills are placed inmunicipalities with populations between 10,000 and 100,000inhabitants (44.07% of the Andalusian population in 2011) and onlyone third are located in areas with fewer than 10,000 inhabitants(20% of the population, 64% of the territory, and with municipalareas with a mean surface area of 90 km2 compared with an aver-age of 101 km2 in Andalusia). However, 56% of the illegal landfills

Probability of uncontrolled landfill presence Illegal landfills

No. %

Less than 12% (low probability) 22 3.4From 12 to 36% (medium probability) 153 33.0Over 36% (high probability) 343 63.6

Total 518 100.0

158 R. Jordá-Borrell et al. / Ecological Indicators 37 (2014) 151– 160

of the

pas

grttpAr

laTtab

5ivrap

tttTCotd

Fig. 3. Areas according to the probability

attern is particularly evident along the coast where expansion isssociated with the development of intensive agriculture, the con-truction sector, tourism, or medium-to-high-technology industry.

Thus, there is a positive correlation between the number of ille-al landfills and the increase in the population and the municipalental values. The causes of the correlation include cursory inspec-ion of the municipal areas by the local governments. Moreover,he Regional Government of Andalusia has implemented neither aunitive policy nor educational campaigns for waste management.dditionally, there is a significant lack of coordination betweenegional and local governments.

For the factor 2 SP, the validation task confirms that most illegalandfills are found in rural areas (60.23%), mainly on uncultivatedgricultural lands (60.23%), with only 7.07% on unproductive land.herefore, these areas receive less management (supervision) byhe appropriate authority (regional and municipal government),nd they constitute imperceptible areas for most of the populationy their low visibility.

Regarding the factor 3 pattern, the methodology’s results show2% of the illegal landfills are close to urban centres, but often

n marginal zones with poverty and delinquency (48.86%) or inacant lots (51.14%) due to the crisis in the construction sector inecent years. In neither case is there management by the municipaluthorities nor environmental awareness on the part of the localopulation.

The factor 4 SP indicates that illegal landfills are visible at lesshan 500 m from any rural road (93.18%) and that the primaryype of waste comes from construction and demolition (60% ofhe total), with urban waste accounting for only 35% of the total.oday, despite the application of punitive policies (especially for

&DW) by local government, permissiveness in the enforcementf the laws has been detected. Fieldwork done in this investiga-ion has shown that the management efficiency for urban waste isirectly related to the level of municipal development, the available

presence of illegal landfills in Andalusia.

financial resources, and especially the incorporation of environ-mental sustainability criteria in municipal development strategies.

The validation of the factor 5 pattern allows us to confirm thatmost of the illegal landfills are found less than 500 m from a track(87.73%), are active (yet unsealed), and unenclosed (53.6%). More-over, the inactive landfills have not been closed or eliminated,which is a result that indicates the absence of good managementby the local authorities or the regional government.

The area “far from highways and industrial zones” defined forpattern 6 contains 44.31% of existing uncontrolled landfills. Thevalidation of this SP is also associated with the level of environmen-tal awareness of the citizens and the municipal institutions. Thus,illegal landfills arise in territories with low visibility to the major-ity of the population and tend to be neglected by local authoritiescalled upon to eliminate them.

The verification of pattern 7 confirms that illegal landfills aremainly found over detrital land (50%), on limestone material(34.09%), and the rest (15.91%) on soils with a high degree of imper-meability (granite, phyllite, quartzite, and very pure clays). From ageomorphological perspective, uncontrolled landfills are situatedon soft rolling topography (21.59% on hillsides and mountains and20.45% on glacis) or flat terrain (32%), such as terraces, plains, mead-ows, and fluvial-tidal formations.

Because of their soil quality and smooth topography, theAndalusian detrital and limestone areas contain most of thepopulation and its associated productive activities (agriculture,construction, and tourism) and thus the majority of the illegallandfills. These areas are known as the Guadalquivir depression,the coast, and the Subbetic Mountains, in contrast with the SierraMorena, which has impermeable soils, is suitable for forests, and is

largely unpopulated.

The pattern 8 results indicate that 85.23% of the illegal landfillsare easily accessible via paths (32.2%) and secondary roads (49.1%),but most of the uncontrolled landfills are far from the green

ical In

pnaPmsirts

filasnati

4

aguiao(c

Graiceupeegv

aasmtgfivamm

isiibdmd

R. Jordá-Borrell et al. / Ecolog

rotected areas (87%). There is a negative correlation between theumber of illegal landfills and their presence in mountainous zonesnd areas of high ecological value, such as the Network of Naturalrotected Areas of Andalusia (RENPA), which combines all of theost representative ecosystems in Andalusia and benefits from a

pecial protective status based on regional, national, European andnternational regulations. Thus, the environmental policy of theegional government and the efforts of the national governmento prosecute offenders of environmental policy are efficient in thisetting.

The validation of pattern 9 shows that 77.27% of the illegal land-lls are found less than 1 km from a waterway, with 63.3% of these

andfills less than 1.3 km from a residential area and 61% at anltitude below the mean (172.6 m). This distribution is due to theignificant and diffuse expansion of the population in the most eco-omically dynamic municipalities in the region (metropolitan areasnd coastal zones), a phenomenon that occurs most intensively inhe municipalities with the largest surface areas, which are foundn the Guadalquivir Valley and on the coast.

. Discussion and conclusions

The model drawn up has been designed to conduct a spatialnalysis for identifying areas with probability of the presence of ille-al landfills whose unit of analysis is the illegal landfill, and whichses mainly geographical variables for this purpose. The interest

n this type of studies is also demonstrated by the scientific liter-ture available on illegal landfills in which we can find examplesf research carried out in countries with different characteristicsJapan, Italy, Romania) where geographical variables are used toonstruct different models to the one applied in this research.

The application of this methodology based on the use of FA andIS has provided adequate results at the regional and local level,

eaffirming that this combination of techniques is sufficient for thenalysis of phenomena that require a multi-scale study, as scales a proportion of reality, but it also represents a certain level ofonceptualisation of the problem. In this context, the methodologymployed uses variables at the local (characteristics of the area ofncontrolled landfill) and regional (rivers, roads, land use, naturalarks, etc.) scale. Thus, through the use of GIS, this methodologynables to analyze the complexity of identifying the probable pres-nce of illegal landfill sites using a two-scale approach: from theeneral to the specific, and from the regional to the local, or viceersa; and so determine the factors that influence each scale.

The method described may also be applied to other regionss long as the necessary thematic and spatial data are avail-ble, although the results could vary according to demographic,ocio-economic, geomorphological and environmental manage-ent characteristics. Likewise, this methodology may be used for

he analysis of other spatial problems apart from the issue of ille-al landfills (studies on areas where the risk of pollution, forestres, deforestation etc. is likely) that involve the analysis of manyariables, the use of large amounts of data and require differentnalytical tools. This methodology may also provide a solid infor-ation basis for future studies on environmental planning andanagement and, above all, for developing prevention plans.The main difficulty in implementing the method developed lies

n the availability of suitable variables on illegal landfills, as officialtatistics are not available. Hence, fieldwork is an essential prelim-nary task for obtaining the necessary data that are homogeneousn the same measurement scale (principally metric); and that may

e represented in a GIS (providing spatial coverage of the sameata). On the other hand, the need to carry out extensive fieldworkust be set against its high cost, depending on the territorial and

emographic size of the area under study.

dicators 37 (2014) 151– 160 159

As regard the advantage of this procedure, it undoubtedly liesin the fulfilment of two necessary and sufficient requirements:

- The method does not arbitrarily include variables related to theprobability of the presence of illegal landfills. Only those vari-ables that have been shown to be decisive in locating areas withthe probable presence of illegal landfills by using FA are consid-ered. Conducting this type of preliminary analysis is one of thepossible solutions, as it enables the identification of such vari-ables and, consequently, the model will only be composed ofthe key factors. Factors that explain the structure and the rela-tionships between geographical variables (physical environment,surrounding area, socio-economic, etc.) through the informationprovided by all cases simultaneously.

- The variables included in the spatial model are not considered tohave the same importance in determining the probability of thepresence of uncontrolled landfills. This methodology requires ageostatistical model to calculate the spatial pattern associatedwith each factor by using measures of central tendency (mean ormode). Thus, each factor is weighted based on its eigenvalue, bythe factorial loads and by the value of each of the variables thatmake up that factor. Therefore, the weights are not given subjec-tively in accordance with the existing bibliography, the opinionof ‘experts’ or the results of surveys.

To obtain new results in future studies on illegal landfills, thismethodology could be complemented by applying other tech-niques such as multiple regression in order to ascertain whether ornot any relationship exists between dependent variables and a setof independent variables. For example, the application of this tech-nique would provide guidance as to which geographical elementsneed to be combined so that there is a high likelihood of encounter-ing an illegal landfill. Furthermore, in order to advance research onillegal landfills it would be advisable to include new variables of thegeographical environment at the local scale with regard to social,cultural, political and economic aspects (environmental awareness,permissiveness in applying the law and punitive policies, wastemanagement, municipal resources, etc.) with the aim of analysingthe interrelationships between these variables and checking theirrelevance.

On the other hand, if the objective of the study is broadened toinclude knowledge of why landfills may appear in certain areas, orwhat features these areas have that encourage the location of Illegallandfills, it would probably be necessary to incorporate variableson the characteristics of the people and companies (agriculture,industry, services) that inhabit these places; and also to study therole of the public authorities in the management of landfills andthe application of laws and regulations in this respect. In this case,if the aim of the study was to explain the causes for the appearanceof landfills, the unit of analysis of the study would be the individualand/or the institution in order to verify whether the behaviour ofthese subjects is what causes the appearance of illegal landfills. Todo this, it would no longer be necessary to create a geostatisticalmodel or to use a GIS because it would not be a territorial study andit could be analyzed with those multivariate statistical techniquesmore appropriate to the subject of study.

In short, the application of a methodology that integrates theresults of multivariate analysis in the GIS (for the study of theidentification of probable areas of illegal landfills) provides an alter-native tool to the application of the techniques of multi-criteriaanalysis by determining the criteria and their relative weights fromfactors based on substantiated and non-aprioristic indications. As

a result, this methodology provides maps with previously lack-ing “information” on potential areas where uncontrolled landfillsmight be located depending on the relative importance of the fac-tors involved. In this way, this methodology enables the creation

1 ical In

oaa

R

A

A

A

B

B

B

B

C

C

C

D

E

G

Vasiljevic, T., Srdjevic, Z., Bajcetic, R., Miloradov, M., 2012. GIS and the ana-

60 R. Jordá-Borrell et al. / Ecolog

f models, because its capacity for storage, retrieval, and spatialnalysis makes GIS an excellent platform for the development,pplication, and validation of these models.

eferences

kbari, V., Rajabi, M.A., Chavoshi, S.H., Shams, R., 2008. Landfill site selection bycombining GIS and fuzzy multi criteria decision analysis, case study: BandarAbbas, Iran. World Appl. Sci. J. 3 (1), 39–47.

llgaier, G., Stegmann, R., 2006. Old landfills in the focus of the urban land man-agement. In: Workshop on landfill reclamation and remediation technologies,7–9. June, International Waste Working Group and University of Padova – Ed.(Cagliari, Italy: CISA Environmental Sanitary Centre).

postol, L., Mihai, F., 2011. The process of closing down rural landfills. Case Study:Neamt County. Present Environ. Sustain. Dev. 5 (2), 167–174.

arredo, J., Bosque, J., 1995. Modelado espacial integrando SIG y evaluación multi-criterio en dos tipos de datos espaciales: vector y raster. Estud. Geogr. LVI (221),637–663.

asnet, B., Apan, A., Raine, R., 2001. Selecting suitable sites for animal waste appli-cation using a raster GIS. Environ. Manage. 28 (4), 519–531.

arredo, J.L., 1996. Multicriteria Evaluation and Geographic Information Systems inLand Use Planning [Evaluación Multicriterio y Sistemas de Información Geográ-fica en la Ordenación del Territorio]. RA-MA Publishing, Madrid, pp. 264.

iotto, G., Silvestri, S., Gobbo, L., Furlan, E., Valenti, S., Rosselli, R., 2009. GIS, multi-criteria and multi-factor spatial analysis for the probability assessment of theexistence of illegal landfills. Int. J. Geogr. Inf. Sci. 2 (10), 1233–1244.

astano, E., 2011. Multidimensional indicator for life indicators [Indicador multidi-mensional de indicadores de vida.] Sociology in its settings [La Sociología en susescenarios.] 24., pp. 5–24.

astro, J.M., 2009. Indicators of Urban Sustainable Development. An Implementa-tion for Andalusia. Free electronic edition. [Indicadores de Desarrollo SostenibleUrbano. Una Aplicación para Andalucía. Edición electrónica gratuita.], Completetext at www.eumed.net/tesis/jmc/ (13.03.13).

hang, N., Parvathinathanb, G., Breeden, J.B., 2008. Combining GIS with fuzzy multicriteria decision making for landfill siting in a fast-growing urban region. J.Environ. Manage. 87, 139–153.

oak, M., Khan, S., Kelly, G., Silvestri, S., 2007. The use of remote sensing tomap illegal landfills at the border of Ireland/Northern Ireland. In: Proceedingsof 11th International Waste Management and Landfill Symposium, Ed. CISAEnvironmental Sanitary Centre, Cagliari, Italy, Sardinia, Italy, 1–5 October,pp. 747–748.

UROSTAT, 2012. Gross domestic product (GDP) at current marketprices by NUTS 2 regions, Complete text at http://appsso.eurostat.ec.

europa.eu/nui/show.do?dataset=nama r e2gdp&lang=en (18.02.13).

alacho, F.B., Ocana, C., 2006. Treatment using GIS and multicriteria evaluationtechniques of the general capacity of land for urban uses: residential andcommercial. Access to spatial information and new geographic technologies.National Congress for Geographic Information Technologies XII. [Tratamiento

dicators 37 (2014) 151– 160

con SIG y Técnicas de Evaluación Multicriterio de la capacidad de acogida delterritorio para usos urbanísticos: residenciales y comerciales. El acceso a lainformación espacial y las nuevas tecnologías geográficas. Congreso Nacionalde Tecnologías de la información Geográfica.XII.], Granada, pp. 1509–1525,http://age.ieg.csic.es/metodos/docs/XII 3/111%20-%20Galacho%20y%20Ocana.pdf (13.03.13).

INE [Spanish National Statistics Institute], 2012. Cifras oficiales dePoblación de los Municipios Espanoles. Revisión del Padrón Municipal,http://www.ine.es/jaxi/menu.do?type=pcaxis&path=%2Ft20%2Fe260&file=inebase&L=0 (01.02.13).

Jordá, R., Ruiz, F., Lucendo, A.L., 2013. Factores territoriales de localización y carac-terización de los vertederos incontrolados en Andalucía. Scr. N. Vol. XVII (435),http://www.ub.edu/geocrit/sn/sn-435.htm

Kontos, T.D., Komilis, D.P., Halvadakis, C.P., 2005. Siting MSW landfills with a spatialmultiple criteria analysis methodology. Waste Manage. 25, 818–832.

Kumar, D., Alappat, B.J., 2005. Evaluating leachate contamination potential of landfillsites using leachate pollution index. Clean Techn. Environ. Policy 7, 190–197.

Meddour-Sahar, O., Meddour, R., Leone, V., Lovreglio, R., Derridj, A., 2013.Analysis of forest fires causes and their motivations in North Algeria:the Delphi method. iForest (early view): e1–e8, http://www.sisef.it/iforest/contents/?id=ifor0098-006 (June, 2013).

Poza, C., Fernández, J., 2010. An approach to building a multidimensional povertyindicator. Which are the risk areas in Spain? [Una aproximación a la construcciónde un indicador de pobreza multidimensional. Cuáles son los focos de riesgo enEspana.]. J. Quant. Methods Bus Econ. [Revista de métodos cuantitativos para laeconomía y la empresa.] 10, 43–72, http://www.upo.es/RevMetCuant/art42.pdf(13.03.13).

Sharma, A., Meesa, S., Pant, S., Alappat, B.J., Kumar, D., 2008. Formulation of a land-fill pollution potential index to compare pollution potential of uncontrolledlandfills. Waste Manag. Res. 26 (5), 474–483.

Sener, S., Sener, E., Nas, B., Karagüzel, R., 2010. Combining AHP with GIS for landfillsite selection: a case study in the Lake Beysehir catchment area (Konya, Turkey).Waste Manage. 30, 2037–2046.

Sener, S., Sener, E., Karagüzel, R., 2011. Solid waste disposal site selection withGIS and AHP methodology: a case study in Senirkentâ Uluborlu (Isparta) Basin,Turkey. Environ. Monit. Assess. 173 (1), 533–554.

Silvestri, S., Omri, M., 2008. A method for the remote sensing identification of uncon-trolled landfills: formulation and validation. Int. J. Remote Sens. 29 (4), 975–989.

Tasaki, T., Kawahata, T., Osako, M., Matsui, Y., Takagishi, S., Morita, A., ShigekiAkishima, S., 2007. A GIS-based zoning of illegal dumping potential for efficientsurveillance. Waste Manage. 27, 256–267.

Tasaki, T., Matsui, Y., Kawahata, T., Osako, M., Takagishi, S., Morita, A., 2004. Analysisof geographic attributes and probabilities related to illegal dumping. Jpn. Soc.Waste Manage. Experts 15, 1–10 (in Japanese with English abstract).

lytic hierarchy process for regional landfill site selection in transitionalcountries: a case study from Serbia. Environ. Manage. 49 (2), 445–458,http://www.springerlink.com/content/p3342l5m07877313/fulltext.html(28.06.12).