A Geographic Information System-Based Indicator of Waste Risk to Investigate the Health Impact of Landfills and Uncontrolled Dumping Sites

Uncontrolled and poor waste management practices are widespread. The global health impact of hazardous waste exposure is controversial, but the excess of some diseases appears to be consistent. The Geographic Information System (GIS, ESRI Inc., Rome, Italy) method used to estimate the waste risk exposure, in an area with many illegal waste dumps and burning sites, is described. A GIS geodatabase (ESRI ArcGIS format) of waste sites’ data was built. A municipal GIS-based indicator of waste risk (Municipal Risk Index: MRI) has been computed, based on type and quantity of waste, typology of waste disposal, known or potential environmental contamination by waste and population living near waste sites. 2767 waste sites were present in an area 426 km2 large. 38% of the population lived near one or more waste sites (100 m). Illegal/uncontrolled waste dumps, including waste burning areas, constituted about 90% of all sites. The 38 investigated municipalities were categorized into 4 classes of MRI. The GIS approach identified a widespread impact of waste sites and the municipalities likely to be most exposed. The highest score of the MRI included the municipalities with the most illegal hazardous waste dumps and burning sites. The GIS-geodatabase provided information to contrast and to prosecute illegal waste trafficking and mismanagements.


Introduction
Waste management is a worldwide problem. At the Sixth Ministerial Conference on Environment and Health of the 53 countries of the World Health Organization (WHO) Regional Office for Europe, waste and contaminated sites were declared among the priority areas for the environmental policy agenda to reach the goals of 2030 Sustainable Development [1]. Municipal and industrial waste disposals contributed to soil and groundwater contamination in about 38% of the contaminated sites in Europe [2]. In the US, the Environmental Protection Agency's National Priority List (NPL) included, in January 2007, 1240 hazardous waste sites. Waste storage/treatment/disposal were present in 31.5% of NPL sites across the country, representing the main activities in contaminated sites [3]. In middle-low income countries, the burden of diseases of waste-related exposures is increasing and not sufficiently recognized [4]. In Asia, 679 areas contaminated by hazardous waste were identified in seven countries [5]. In Africa, the WHO included hazardous waste among the first three main environmental risk factors for the health of the population [6].
Uncontrolled and poor hazardous waste management practices are widespread in some areas of both industrialized and middle-low income countries. In addition, illegal waste practices and The available data on waste disposals, uncontrolled and illegal waste dumps and burning waste sites located in the study area were collected in a GIS geodatabase (ESRI ArcGIS format), in different polygonal or point features, depending on the source of data. Data enclosed by the Prosecutor during judiciary inquiries were considered. Databases of Regional Agencies and Institutes, available at December 2017-January 2018, were consulted. Data on waste disposals present in Campania Region Agency for Environmental Protection (ARPAC) and Zooprophylactic Institute of Southern Italy The available data on waste disposals, uncontrolled and illegal waste dumps and burning waste sites located in the study area were collected in a GIS geodatabase (ESRI ArcGIS format), in different polygonal or point features, depending on the source of data. Data enclosed by the Prosecutor during judiciary inquiries were considered. Databases of Regional Agencies and Institutes, available at December 2017-January 2018, were consulted. Data on waste disposals present in Campania Region Agency for Environmental Protection (ARPAC) and Zooprophylactic Institute of Southern Italy (Istituto Zooprofilattico Sperimentale del Mezzogiorno: IZSM) database were enclosed. In addition, waste treatment plants and disposals reported on the Campania Region website [23], were considered. Moreover, data available at the Department of Environment and Health of the National Institute for Health (ISS) database, elaborated during previous monitoring activities and Projects, particularly in the "Land of Fires", were included. In summary, the considered data sources included the waste sites present in the territory of 38 municipalities at study, in the 2008-2017 period. Uncontrolled and illegal waste management practices have been occurring in the area since the late 1980s [22] and significant remediation actions have not been carried out in the area, at the time of the beginning of the present investigation.
Initially, 3461 sites were mapped, considering all data sources. The layers were created based on the data source and the type of waste site. All layers, built on the vector data model, have geometry and attributes. The sites present in more than one layer have been left in a single layer in which the attributes available in the other data sources have been transcribed. After dropping the repeating sites and a thorough validation, 2767 out of 3461 waste sites were selected for the present investigation ( Figure 2).  [23], were considered. Moreover, data available at the Department of Environment and Health of the National Institute for Health (ISS) database, elaborated during previous monitoring activities and Projects, particularly in the "Land of Fires", were included. In summary, the considered data sources included the waste sites present in the territory of 38 municipalities at study, in the 2008-2017 period. Uncontrolled and illegal waste management practices have been occurring in the area since the late 1980s [22] and significant remediation actions have not been carried out in the area, at the time of the beginning of the present investigation. Initially, 3461 sites were mapped, considering all data sources. The layers were created based on the data source and the type of waste site. All layers, built on the vector data model, have geometry and attributes. The sites present in more than one layer have been left in a single layer in which the attributes available in the other data sources have been transcribed. After dropping the repeating sites and a thorough validation, 2767 out of 3461 waste sites were selected for the present investigation ( Figure 2).
. Analytical data on specific environmental contaminants and chemical agents present in waste were available for about 1% of the sites. Considering the low number of sites for which this information was available, estimating the exposed population based on models of environmental contamination by chemicals emitted/released by waste sites was not possible.
Because of this limitation, a municipal hazard risk for waste was elaborated on the basis of the population living in proximity of waste sites and the estimated hazard of the waste, following the steps described below. The ascertained presence of contaminants at high concentration (such as reported in the considered database), when available, was considered in the judgment of the potential health impact of the waste site. Analytical data on specific environmental contaminants and chemical agents present in waste were available for about 1% of the sites. Considering the low number of sites for which this information was available, estimating the exposed population based on models of environmental contamination by chemicals emitted/released by waste sites was not possible.
Because of this limitation, a municipal hazard risk for waste was elaborated on the basis of the population living in proximity of waste sites and the estimated hazard of the waste, following the steps described below. The ascertained presence of contaminants at high concentration (such as reported in the considered database), when available, was considered in the judgment of the potential health impact of the waste site.

Hazard Risk Level Index, by Waste Site
In the first step, a Hazard Index (HI) was attributed to each of the 2767 waste sites on the basis of the information available for all sites: modality of waste disposal (i.e., illegal burning sites and dumps, controlled landfills and treatment plants, temporary storage), characteristic of the site and the type of waste. The alpha-numeric HI is composed by a number, based on potential environmental impact of the different types of waste disposals/sites (based on expert judgment), and a letter, corresponding to the type of waste.
A panel of three experts evaluated the potential environmental impact of each category of site, on the basis of the likelihood of environmental contamination, considering the probability of leaching and/or runoff in groundwater and of emission of contaminants in air and in soil. In 1% of the sites for which the presence of contaminants (organic/inorganic) was reported, the corresponding HI value has been raised up. F *: Non-hazardous waste in controlled storage sites and disused industries and quarries without any information.
In addition, the letter "G" was attributed to the activities without supposed waste release (e.g., heaps of inert waste, less than 10,000 mc, no defined treatment plants, productive activities): these sites were not considered in the waste hazard risk estimate. Table 1 shows the devised site-specific Hazard Index (HI) with the corresponding waste site, and the criteria for the attribution. Considering the sites located in the area at study, eleven combinations of the alpha-numeric hazard index (HI) were identified in the study area: 5A, 4A, 4B, 3B, 2B, 2C, 1C, 1D, 1E, 1F and 1F * ( Table 1).

Conversion of Alpha-Numeric Index to Numeric Hazard Risk Level
The next step was to derive a hazard quantification (Hazard Risk Level: HRL) from the alpha-numeric index. The experts' panel attributed a numeric HRL to each HI, considering that the magnitude of the impact of the sites depends on the combination between the type of waste (the letters of the HI index) and the modalities of disposal and characteristics of the site (the number of HI). The attribution of the numeric correspondence was based on the likelihood of contamination by toxic agents and their potential hazard for the population. The most weight was attributed to the numeric part of HI, one order of magnitude moving from one to the next of the five levels, while letters were transformed in numbers without any amplification through different levels ( Table 2). The conversion of the numeric part to ten powers was applied in such a way that the presence in one area of multiple low-hazard sites produced a hazard level lower than one high-hazard site (i.e., an illegal burning waste site is considered to be far more impacting than many controlled landfills of inert materials). The value of HRL is the product between the two numeric correspondences (i.e., an HI score equal to 5A corresponds to a hazard risk level of 60,000:10,000 * 6).

Municipal Waste Risk Index
The final aim of this study is to estimate the human health risk of populations residing in the study area due to the waste sites, at the municipal level. To reach this objective, the next step was to consider the HRL of the sites and the population living in the areas potentially impacted by the waste sites.
A new polygonal layer was therefore generated by the merge geoprocessing tool, preserving geometries and HRL of each waste site for all layers. The point layers have been previously transformed into hexes with 100 m apothem. Figure 2 shows the map of the waste sites.
In order to identify the population living in the areas impacted by the waste site, a circular buffer of 100 m radius around the features of each waste site was generated, first without the dissolve option and subsequently with this option. The intersection of these two layers generated the break buffer overlaps-up ( Figure 3).   B  5  3  100  C  4  4  1000  D  3  5 10,000 E 2 The final aim of this study is to estimate the human health risk of populations residing in the study area due to the waste sites, at the municipal level. To reach this objective, the next step was to consider the HRL of the sites and the population living in the areas potentially impacted by the waste sites.
A new polygonal layer was therefore generated by the merge geoprocessing tool, preserving geometries and HRL of each waste site for all layers. The point layers have been previously transformed into hexes with 100 m apothem. Figure 2 shows the map of the waste sites.
In order to identify the population living in the areas impacted by the waste site, a circular buffer of 100 meters radius around the features of each waste site was generated, first without the dissolve option and subsequently with this option. The intersection of these two layers generated the break buffer overlaps-up ( Figure 3).  The choice of a 100 m radius, much lower than what has been used in some comparable studies (1-2 km), is motivated by the high density of waste sites in the study area (around 3000 sites in an area of about 426 km 2 ). Buffers of larger breadth have caused an overlap of the impacted areas and the whole area would appear to be affected by waste impact, preventing a distinction of the areas at different degrees of waste impact. For each municipality, we considered the waste site buffers (100 m) laying in the municipality, even if the centroid was located in other municipalities of the area at study.
The resulting layer was combined (union geoprocessing tool) with the layer of the 2383 census tracts (Figure 1) with the data relating to the resident population previously associated (Census 2011).
The union between the two layers (waste sites and census-tracts) generated about 75,000 polygons, with many overlaps despite the small size of the buffers. Territorial overlaps net, the number of polygons has fallen to around 26,000 (Figure 4).
The choice of a 100 m radius, much lower than what has been used in some comparable studies (1-2 km), is motivated by the high density of waste sites in the study area (around 3000 sites in an area of about 426 km 2 ). Buffers of larger breadth have caused an overlap of the impacted areas and the whole area would appear to be affected by waste impact, preventing a distinction of the areas at different degrees of waste impact. For each municipality, we considered the waste site buffers (100 m) laying in the municipality, even if the centroid was located in other municipalities of the area at study.
The resulting layer was combined (union geoprocessing tool) with the layer of the 2383 census tracts (Figure 1) with the data relating to the resident population previously associated (Census 2011).
The union between the two layers (waste sites and census-tracts) generated about 75,000 polygons, with many overlaps despite the small size of the buffers. Territorial overlaps net, the number of polygons has fallen to around 26,000 (Figure 4). A multi-code HRL (equal to the sum of HRL) was attributed to the areas influenced by more than one site, with an ad hoc Visual Basic software (Microsoft Office, Istituto Superiore di Sanità Licence, Roma, Italy). The population living in the areas impacted by waste was estimated on the basis of the density of population in the census-tract where the polygon falls.
For each polygon, a Risk Index (RI) was computed: where S is the surface of the polygon, HRL is the hazard risk level index of the waste site, or the multicode HRL of the waste sites, lying in the polygon, Sc is the surface of the census-tract, P is the population residing in the census-tract and S / Sc × P is the estimated population residing in the polygon. RI is proportional to the population living in the census-tract: if the buffer falls in an inhabited census-tract, the RI is equal to 0. A multi-code HRL (equal to the sum of HRL) was attributed to the areas influenced by more than one site, with an ad hoc Visual Basic software (Microsoft Office, Istituto Superiore di Sanità Licence, Roma, Italy). The population living in the areas impacted by waste was estimated on the basis of the density of population in the census-tract where the polygon falls.
For each polygon, a Risk Index (RI) was computed: where S is the surface of the polygon, HRL is the hazard risk level index of the waste site, or the multi-code HRL of the waste sites, lying in the polygon, Sc is the surface of the census-tract, P is the population residing in the census-tract and S/Sc × P is the estimated population residing in the polygon. RI is proportional to the population living in the census-tract: if the buffer falls in an inhabited census-tract, the RI is equal to 0. Subsequently, for each municipality, the areas influenced by one or more waste sites, with the corresponding HRL and living population (site-specific RI), were considered.
A waste Risk Index at the municipal level (Municipal Risk Index: MRI) was computed, summing up the scores of all areas (polygons) comprising the municipality: where p is the number of polygons lying in the municipality, and RI p is the Risk Index of polygons lying in the municipality. Finally, municipalities were categorized into four classes of risk (1-low to 4-high), on the basis of MRI, using Jenks' method (Natural breaks), that maximizes homogeneity within groups and variance between groups. The categorization of municipalities in 4 MRI classes was evaluated as the most appropriate, with respect to 5 or more classes, to distinguish the municipalities at the highest MRI, after a sensitivity analysis considering the distribution of municipalities and residing population by MRI classes.

Results
The study area, constituted by 38 municipalities, is 426 km 2 large. In this area, on the basis of the available data at January 2017, 2767 waste sites were mapped. The GIS approach allowed us to estimate, in the whole study area, 354,845 people (38% of the total population) living close to one or more waste sites (in a buffer of 100 m). In the whole area, there is a high environmental pressure by waste sites and likely human exposure to a variety of agents, including toxic ones. Table 3 reports the distribution of waste sites by HI score and municipality.  In bold-the waste sites in overall area at study.
The most represented HI score class of waste sites (33% of all waste sites) is that of HI = 1F, corresponding to the waste sites with unlikely release of hazardous substances (controlled urban waste landfills, treatment plants of non-hazardous waste and illegal heaps of undefined waste), followed by the group of waste burning sites (HI = 5A, 23%) and by dumping sites with hazardous wastes and documented or potential contamination of soil (HI = 3B, 20%). Considering the HI score classes corresponding exclusively to illegal waste sites (5A, 4A, 4B, 3B) and the uncontrolled waste disposals or heaps included in other HI score classes, on the basis of their hazard level, all illegal and uncontrolled waste sites represent about 90% of the waste sites present in the study area. All municipalities have more than one waste site. The municipalities with the highest number of waste sites are Giugliano in Campania and Caivano (628 and 282 waste sites, respectively) and Casavatore is the municipality with the least number of waste sites (3 sites). Twenty-seven percent (178 sites) of all waste burning sites (HI:5A) are located in Giugliano in Campania. Table 4 shows the surface impacted by waste, the population living in these areas and the corresponding MRI class, by municipality.  Considering the area of impact (100 m buffer) of waste sites falling in municipal areas, even if the centroid of the site was located outside the municipality, 7.9% (4510 inhabitants) of the municipal population and 1.3% of municipal surface represented the lowest percentage of the municipal population living in one or more waste site buffers and of the impacted area, respectively (Table 4). Figure 5 shows the geographical distribution of MRI classes by municipality. Class 4 includes the two municipalities with the highest waste risk for the population, while municipalities of MRI class 1 are the least impacted with respect to those at study, even if there is a waste risk exposure.
Int. J. Environ. Res. Public Health 2020, 17, 5789 13 of 17 Considering the area of impact (100 m buffer) of waste sites falling in municipal areas, even if the centroid of the site was located outside the municipality, 7.9% (4510 inhabitants) of the municipal population and 1.3% of municipal surface represented the lowest percentage of the municipal population living in one or more waste site buffers and of the impacted area, respectively (Table 4). Figure 5 shows the geographical distribution of MRI classes by municipality. Class 4 includes the two municipalities with the highest waste risk for the population, while municipalities of MRI class 1 are the least impacted with respect to those at study, even if there is a waste risk exposure.

Discussion
The adopted GIS approach [24,25] identified areas with different levels of waste risk exposure for the population, also in the lack of analytical data on environmental contamination.

The Main Results
The whole area at study is impacted by waste sites and 354,845 people (38% of the population residing in the study area) live within 100 m from one or more waste sites. Also, the municipalities least impacted by waste sites include a percentage of the population living near waste sites (at least around 8%), thus at risk of exposure. The municipalities with the highest MRI correspond to those with the highest number of burning waste sites and illegal hazardous waste dumps. In these areas, the implementation of remedial actions appears to be a priority. The results reported in the GISdatabase permit, on the one hand, the Prosecution Office to modulate in the most updated way the investigations and their priorities, and, on the other hand, Public Health Authorities to schedule convenient precautionary strategies.

Discussion
The adopted GIS approach [24,25] identified areas with different levels of waste risk exposure for the population, also in the lack of analytical data on environmental contamination.

The Main Results
The whole area at study is impacted by waste sites and 354,845 people (38% of the population residing in the study area) live within 100 m from one or more waste sites. Also, the municipalities least impacted by waste sites include a percentage of the population living near waste sites (at least around 8%), thus at risk of exposure. The municipalities with the highest MRI correspond to those with the highest number of burning waste sites and illegal hazardous waste dumps. In these areas, the implementation of remedial actions appears to be a priority. The results reported in the GIS-database permit, on the one hand, the Prosecution Office to modulate in the most updated way the investigations and their priorities, and, on the other hand, Public Health Authorities to schedule convenient precautionary strategies.

Selection of Waste Impact Area Buffer
A conservative method was developed, in order to highlight the most exposed populations. The radius (100 m) used to identify the population exposed to waste sites is relatively short with respect to the one of 1-2 km used in other similar investigations [9]. The rationale for choosing a 100 m radius buffer was illustrated in the Methodology Section. The choice was based on a priori considerations. One alternative approach might have been that of selecting the radius after testing the effects of the adoption of different values for this parameter. Given the context and the aims of the study, namely, to provide elements for priority evaluation in adopting site-specific remedial actions, it was agreed to follow the a priori rather than a posteriori approach. Moreover, the application of a low radius waste impact area allowed us to distinguish the subareas at different levels of waste impact and to highlight the highest impacted areas, preventing false positives.

Limits of the Adopted Method
Some limitations of the adopted approach have to be discussed. The present investigation focused on the sites where mismanagement and illegal dumping and burning of waste cause a potentially dangerous exposure for the population and did not consider other potential sources of contamination. Some industrial emissions might represent a source of environmental pollution in the areas, but this issue was not an object of the request of the Prosecutor.
The municipal level of the waste indicator and the use of the distance from the waste sites as a proxy of environmental exposure, in particular, deserve some considerations. Firstly, the attribution of the same indicator of waste risk to all people living in a municipality could represent a bias in the exposure assessment process. Considering the percentage of the population living near waste sites with different levels of impact, in the computation of the indicator, though, could reduce the distorsion. The indicator of waste risk was computed at the municipal level because of the availability of health outcomes data (mortality, hospitalization) at this same geographical level, that will be used in the epidemiological study to assess the association between waste exposure and occurrence of specific diseases. The investigations carried out at the population level, applying appropriated methods, provide useful indications for public health interventions to be implemented at the community level [26]. On the other hand, the findings, informative at the population level, may not be used to infer individual exposure levels, which would lead in this case to the so-called "ecological fallacy". In the cases of availability of outcome data at smaller areas (e.g., census-tract) or at the individual level, the described GIS method and the criteria adopted to define the potential health impact of waste sites, could be still applied.
Second, in order to identify the exposed population, the distance of the residence from a waste site was used. In the lack of information on environmental contaminants, the distance from the source of environmental contamination may represent a reasonable indicator of environmental exposure [11][12][13][14][15][16][17]. In the contexts where large areas and populations are affected by mixtures of unknown substances potentially emitted or released, carrying out biomonitoring studies and applying dispersion models can be problematic [18]. Furthermore, because of the lack of analytical environmental data, the indicator was based on the estimated toxicity of waste and on the likelihood of the exposure to emitted/released toxic substances. Detailed environmental data, including biomonitoring data, when available, might enrich the indicator, by making it more accurate and less vulnerable to random misclassification that implies underestimation of risk.
Considering the possible health impact of electric and electronic waste disposals [8,[27][28][29], we point out that no explicit information regarding the presence of this type of waste is available in our dataset. Despite this, in view of the illegal and uncontrolled management of waste in the study area, the presence of e-waste, that was not considered, could not be ruled out. In addition, considering that the impact areas of waste sites located in neighboring municipalities, but excluded by the study area, were not considered, the exposed population at waste risk could be underestimated.
Lastly, a limitation of the present investigation could be represented by the lack of information on the running period of each site, partially due to the informal and uncontrolled nature of the practices. Most waste sites identified in the study area exert their polluting actions starting from the early 1990s. Illegal trafficking and dumping of hazardous waste by crime organizations in a sub-area of Campania Region including that of the present study, since the end of the 1980s, are documented, on the basis of crime organizations' exponents' statements and judiciary investigations [22].

The Developed MRI
The meaning of the different classes of MRI deserves specific considerations. The MRI developed in this study updates the previously computed indicator [30] used in an earlier geographic mortality study [21]. The new indicator considers a smaller territory (38 versus 196 municipalities), but a higher number of waste sites (2767 versus 226), including the burning waste sites not previously investigated. The integration of several data sources on the waste sites present in the study area, including those enrolled by judiciary authorities, provided an updated and a more detailed and exhaustive database.
The Hazard Index, and therefore the classes of the related MRI, is based on expert judgment with rare availability of measurements of pollutants to validate the choice of the categories. This process might be unable to quantify differences in terms of specific hazards. However, it allows us to distinguish among different likely hazardous exposure scenarios and to highlight municipalities differently impacted by waste, when looking at the health profile of populations residing in these areas. HRL score classes represent a "relative" score, pointing out municipalities with the highest waste impact among those at study, and the municipalities included in the lowest HRL score class (class 1) are also impacted by waste sites ( Table 4). The present categorization in 4 classes makes out, in particular, the municipalities at the highest waste impact (4 and 3 MRI classes), the principal aim of the present investigation. A different categorization of the absolute value of MRI is possible.

The contributions of the Present Investigation
Notwithstanding the above-described limitations, the present study highlighted the need for interventions of environmental remediation in the whole area, and, primarily, in the municipalities with the highest MRIs. In particular, illegal and uncontrolled waste dumping and burning, still ongoing after more than thirty years, have to be urgently stopped.
The GIS indicator, built as described above, will be applied to search correlations with health outcomes in ad hoc epidemiological investigations. These studies will quantify the health impact of waste sites, focusing on the diseases that recognize exposure to waste among their risk factors, and will identify the diseases, also in the childhood and adolescent population, requiring the implementation of appropriate health surveillance and healthcare actions.

Conclusions
The present investigation was inspired by and originated from a formal request to the North Naples Prosecutor Office to the National Institute for Health, to identify the municipalities at the highest waste impact in an area characterized by a widespread waste mismanagement and illegal practices. The ad hoc-developed GIS-approach, notwithstanding the limitation discussed in the paper, identified the municipalities affected by different waste impact levels.
In the whole area, there is a high environmental pressure by waste sites and likely human exposure to a variety of agents, including toxic ones. Thirty-eight percent of the population lives within 100 m from one or more waste sites. Illegal and uncontrolled waste sites, including 653 waste burning sites, represent about 90% of all 2767 waste sites present in the area. Municipalities with different levels of predicted waste exposure risk for the population were identified with the GIS approach.
On the basis of the present results, environmental remediation actions and stopping still ongoing illegal and poor waste management practices are urgently needed. The data reported in the GIS system provide useful information to the Prosecutor, in order to contrast illegal waste mismanagements and to prosecute the criminal acts regarding waste trafficking and management. The MRI will be applied in further epidemiological study on the correlation between waste impact and the occurrence of specific diseases, in the population residing in the study area.
The GIS method described in the present paper could also be usefully applied in the lack of detailed environmental data and analytical information on waste sites. These situations could be common in similar contexts, of informal and illegal waste management, in specific areas of both industrialized and middle-low income countries.