Analysis of Factors Influencing Illegal Waste Dumping Generation Using GIS Spatial Regression Methods

Abstract

Illegal municipal waste dumping practices in developing countries may be impacted by many factors such as socioeconomic, demographic, availability of waste collection facilities, recycling sites, and spatial characteristics. This study uses spatial regression analysis to identify which factors primarily impact illegal waste dumping practices. For this purpose, 8 variables explain the data for the 177 subdistricts used in the spatial regression analysis. This study used ordinary least squares (OLS) and geographically weighted regression (GWR) methods to build a regression model of the factors identified. OLS analysis showed that only elevation and population density were found to become determinants of illegal waste dumping activity based on spatial regression methods. Elevation above sea level is positively correlated while population density is negatively correlated with the number of illegal dumping generations. GWR shows a better statistical value than OLS, where the significance of the adjusted R-square increased from 0.24 to 0.61. This study may help reduce the number of illegal waste dumping practices, especially in a metro city context.

1. Introduction

Improper solid waste management practices, such as open burning, illegal disposal, and burying, are significant problems in many developing countries. Inadequate waste management treatment also increases the number of illegal waste disposal piles [1]. This condition can also be found primarily in rural areas where collection services are unavailable [2]. These practices are prohibited since they can emit significant amounts of greenhouse gas (GHG) emissions, such as CO₂, NO_x, and SO₂, and even short-lived climate pollutants (SLCPs), such as CH₄ and BC. SLCPs are powerful enough to increase the rate of climate change since they have a shorter lifetime in the atmosphere [3]. A recent study found that improper waste disposal in developing countries such as the Philippines can emit 363 kg CO₂ eq/tons of waste [4]. Moreover, improper waste disposal can also pollute the environment by releasing toxic materials and compounds that can accumulate in organisms [5].

Many factors contribute to higher illegal disposal rates. As stated by many researchers, the most common reason is poor waste management infrastructure, such as waste collection facilities and transportation [6,7,8]. Therefore, providing infrastructure may be a solution to reduce these business-as-usual (BAU) practices [9]. However, Sedova et al. found that illegal dumping behavior is also influenced by other factors such as education level, awareness, dumping cost, and income level [10]. Dumping costs are related to low-income communities. Communities tend to participate in illegal dumping practices rather than pay a certain amount of money [8]. Moreover, people who generate a large amount of waste based on their household size also tend to participate in illegal dumping practices [2]. These reviews present the possibility of other reasons behind the unlawful dumping behavior in some communities. Estimating the factors influencing illegal dumping generation can be essential for reducing human and environmental risks.

In most Indonesian cities, waste collection services can cover almost all residents; more than 80% of waste is collected and transported to landfill. However, at the national level, it is said that the collected waste is only less than 55%, and the rest can be subjected to an illegal disposal practice [11]. A recent study found that around 199.78 tons/day of waste, or 12.02% of the total waste generation, is illegally dumped in Semarang City [12]. Another report also found that around 20% of the waste generated in Padang City is disposed of improperly [13]. Therefore, the data on illegal dumping in Indonesia and maybe in other developing countries are scarce. Thus, determining the number of illegal waste dumping is a challenging task [14].

Spatial data analysis, such as spatial autocorrelation (SAR) and global weighted average (GWR), and nonspatial data analysis, such as ordinary least squares (OLS), are often found to solve environmental management problems such as waste management. These regression models include spatial dependence in the analysis [15]. Agovino and Musella have found several factors that may influence separation activity in mountainous areas [7]. Antczak and Keser et al. used the GWR model to predict several independent variables that may affect municipal waste production in Poland and Turkey [15,16]. OLS and SAR can be used to develop a global model appropriate for more extensive spatial data, such as a country or province, where the GWR can explore local variables. Therefore, good model validation will be established by employing GWR analysis in the local context [7].

This study aims to identify the factors that may influence the dumping practices in Semarang City using a geographic information system (GIS) by using subdistrict data collected from the municipal government of Semarang City. No previous researchers have worked on this topic, and most have used spatial analysis tools to find the determinants for waste generation and separation only [7,17,18,19]. The number of illegal dumping piles, considered illegal dumping generation, is estimated using Ramadan et al.’s study [1]. Thus, determinants such as land use, elevation, population, population density, household, sex ratio, disaster event, number of waste collection sites, and public facilities in each subdistrict are evaluated using the spatial technique. Section 2 of this paper presents the methodological details of the spatial analysis technique, while Section 3 consists of the Semarang City profile and the results of the spatial analysis. Section 4 is the discussion section, which suggests recommendations and is provided by compiling some findings from previous work to better understand the reduction of improper waste disposal in Semarang City. Last, Section 5 discusses the conclusions and policy implications. This study’s results may be helpful as a theoretical prediction related to illegally dumped waste generation.

2. Materials and Methods

2.1. Factors Influencing Illegal Dumping Practices

Most illegal dumping practices are caused by a lack of waste management infrastructure, such as waste transportation systems and collection facilities [2,6,10]. In rapidly urbanizing cities such as Indian cities, for example, the number of illegally dumped sites is increasing daily since waste collection sites cannot receive more waste. Municipal solid waste is left uncollected on every city corner [8]. Moreover, the government needs a faster response to provide waste collection infrastructure because of multilevel governance, budget structure, and public acceptance [6]. However, it is still predicted that the presence of waste collection facilities will reduce the rate of illegal dumping practices [7]. Other factors that may affect the number of improper waste disposal practices are the population and its density. The population indicates the volume of waste, which is also in line with the number of waste disposal practices [16]. However, as stated by Madden et al., higher population density indicates higher residential development and urban infrastructure development, and this condition will reduce the number of illegal dumping practices [20]. This hypothesis is also related to land use characteristics where higher activity will reduce the potential for illegal dumping practices [21]. Elevation above sea level is another possible factor that may increase the volume of illegal waste dumped. As Agovino and Musella stated, elevation indicates complex waste management problems. Sorting operations in mountainous areas are more complicated than in flat areas because collection points need to be reachable.

In the same way, housing dispersion, a characteristic of sloping areas, makes the collection system more unreachable, leading to higher illegal waste disposal practices [7]. The housing unit number is also predicted to be a significantly positively correlated determinant for improper waste disposal [20]. Housing indicates the number of occupants or the population, which also increases waste generation [16]. Gender is predicted to have a significant correlation with illegal dumping practices. As stated by Seng et al., male citizens tend to illegally dump waste rather than females. Females seem to appropriately dispose of their waste [2]. The number of disaster events also increases the possibility of illegal waste dumps. There is a positive relationship between the number of illegal waste dumps and the number of disasters. This condition may be due to the lack of infrastructure to prevent waste dumping from happening [22]. Public facilities such as markets, religious places, parks or green open spaces, and education facilities indicate the development of a region. Therefore, the more public facilities units, the less illegal waste dumping there will be [23]. In summary, Table 1 presents the predicted determinants or factors of improper waste disposal practices in Semarang City based on the summarized literature and the currently available data.

Table 1. Factors or determinants of improper waste disposal in Semarang City.

Variable	Code	Correlation	Description/Hypothesis	References
Waste collection site	WCS	Negative	Fewer waste collection sites will increase the possibility of illegal dumping practices.	[7,8]
Population	Pop	Positive	The bigger the population, the bigger the waste generation, leading to a higher possibility of illegal dumping practices.	[16,20]
Population density	PD	Negative	A more significant population density will increase the lower possibility of illegal dumping practices since the waste collection service is provided well.	[7,20]
Land use score	LUFINAL	Negative	The higher the activity, the lower potential for illegal dumping practices.	[21]
Elevation	ELEV	Positive	The higher the elevation, the higher possibility of illegal dumping practices.	[7]
Number of housing	HOUSE	Positive	The number of households indicates the higher activities in the area, thus increasing the number of illegal disposal sites.	[16,20]
Sex ratio	SR	Negative	The higher percentage of males in a subdistrict will increase the number of illegal dumping practices.	[2]
Disaster event	DE	Positive	The higher potential of disaster will lead to a higher possibility of illegal dumping practices.	[22]
Public facilities	CF	Negative	The more significant number of public facilities in an area, the less possibility of illegal dumping sites.	[23]

2.2. Spatial and Nonspatial Analysis

Data preparation and analysis were performed following previous studies by Rybova et al., Keser et al., Nazeer et al., and Antczak [15,16,24,25]. Details of the methodology can be seen in Figure 1. The research began with collecting data for determinants or explanatory variables that were to be analyzed spatially. As mentioned in Table 1, the explanatory and independent variables were gathered from several resources such as population density; sex ratio, number of housing; disaster events and public facilities from the Central Bureau of Statistics, Population and Civil Registration Agency of Semarang City; waste collection site number from the Environmental Services Agency of Semarang City; improper waste disposal estimation from Ramadan et al. [12]; and a Semarang City land use map from the Spatial Planning Agency of Semarang City. Elevation data were taken from Google Maps as the center of the subdistrict. The primary statistical data of the potential determinants or factors are shown in Table 2. The land use score (LUFINAL) was obtained by reclassifying the raster dataset of the Semarang City land use into 5 classes or values. The higher the value, the higher possibility that illegal waste disposal piles can be found easily. For instance, green open spaces, protected areas, and forests are valued as 1. Meanwhile, higher anthropogenic activity, such as households and marketplaces, are valued as 5. The land use score for each subdistrict can then be calculated using Equation (1).

$$L{U}_{Final}=(\sum A_{1}x 1)+(\sum A_{2}x 2)+\dots +(\sum A_{5}x 5)$$

(1)

where $\sum A_n$ is the total area for each land use score (km²). The highest score of $L{U}_{Final}$ indicates higher anthropogenic activities in the subdistrict. All data were prepared using a Microsoft Excel datasheet (.xls) and included in the spatial database using the “join and relate” tools in the ArcGIS 10.6 software package.

Figure 1. Research methodology, where OLS stands for ordinary least squares and GWR stands for geographically weighted regression.

Table 2. Basic statistical data information of the factors.

Variable	Unit	Max	Min	Mean	St. Dev
Waste collection site	unit	7	0	1.07	1.26
Land use score	km2	9,195,316	2,461,048	4,420,573	842,025
Elevation	m	349	0	75	99
Population density	people/km2	29,166	246	7758	5787
Sex ratio	-	1.09	0.87	0.98	0.04
Number of housing	unit	12,524	204	3099	2025
Disaster event	times	11	0	0.95	1.89
Public facilities	unit	215.00	9.00	50.72	30.17
Improper waste disposal weight	tons/day	11.14	0.03	1.13	1.51

The OLS regression model was applied as a nonspatial data analysis to estimate the regression coefficient. OLS is a preliminary regression analysis since it can provide a global model to find the appropriate independent variable or predicted process. The model is developed as shown in Equation (2) [24].

$$Y_i={\beta }_0+{\beta }_1{X}_{1i}+{\beta }_2{X}_{2i}+{\beta }_3{X}_{3i}\dots +{\beta }_k{X}_{ki}+{\epsilon }_i$$

(2)

where $Y_i$ is the observed dependent variable, $X_{1i}$, $X_{2i}, \dots X_{ki}$ are determinants or independent variables or factors, ${\beta }_k$ is the regression coefficient, and ${\epsilon }_i$ is the random error. The model performance was assessed by checking the multiple and adjusted R-squares. Global or local multicollinearity was also evaluated by considering the value of probability, robust probability, and variance inflation factor (VIF). The sign of the coefficient was also checked to understand the relationship between the determinants and the independent variable. A t-test was employed to define the relationship status, whether significant or insignificant. As a rule of thumb, the VIF values should be less than 7.5, as a higher VIF value indicates that the determinants have the same story as the dependent variable, so it should be removed from the factor list. The significance of the model was also assessed by using the Koenker (BP) statistic, where the p-value should be less than 0.05 to be determined as a significant model. Moreover, this analysis can be used to assess heteroscedasticity in the model.

Local geographically weighted regression (GWR) can be a suitable instrument for developing models with a dataset with a high degree of spatial heterogeneity [7]. The kernel type and bandwidth method for the developed GWR are adaptive and AICc. The adaptive kernel type is typically preferred when the observed spatial autocorrelation is clustered [25]. Lastly, Akaike’s information criterion (AICc) was used to compare the OLS and GWR models. A smaller AICc value will result in a better fit with the observed data. Before and after conducting the GWR analysis, the residual spatial autocorrelation was run. This assessment was used to predict whether the model developed from OLS and GWR is under-, in-between, or overpredictions. Moran’s I test measured the expressed pattern of the developed OLS and GWR models [25].

3. Results

3.1. Semarang City Waste Management Profile

Semarang is a coastal city that is also considered the only metro city in Indonesia. Semarang City comprises 16 districts that are divided into 177 subdistricts. Semarang City has an increasing GDP; from IDR 91.19 million in 2017, it increased until it reached IDR 123.98 in 2021. The average population density was 4552 people/km² in 2016 and is predicted to increase yearly. Because the population is growing, the significant amount of waste generated is also increasing yearly. As reported by Sadma and Syafrudin et al., the daily waste generation in Semarang City is around 5248 m³ or equal to 801.52 tons/day in 2018, where the waste transported to the Jatibarang landfill is only 88.50%. This waste is mainly dominated by organic waste, which accounts for 61.34% and plastic waste 16.34%. The recyclable waste mainly consisted of plastic (56%), paper (16%), metal (16%), glass (11%), and others (1%). The waste transportation can cover all districts in Semarang City. The vehicles used to transport the waste include 170 arm-roll and 41 dump trucks. The Jatibarang landfill is operated under a controlled landfill system, where waste is covered with a soil layer every 5–7 days after reaching 50 cm. This landfill covers 46 hectares that are divided into 3 zones, including 1 passive and 2 active zones, where a landfill gas (LFG) power plant is operated to extract and utilize methane gas as a power generation supply [26,27]. Table 3 summarizes the time series data of the Semarang City population, GDP, and waste management profile.

Table 3. Basic data of Semarang City profile.

Variable	2017	2018	2019	2020	2021	Sources
Population	1,658,552	1,668,578	1,674,358	1,685,909	1,687,222	Population and Civil Registration Agency (a)
GDP (IDR/million)	91.19	98.1	105.93	114.2	123.98	Central Bureau of Statistics
Waste handled (ton)	414,647.3	423,830.7	356,782.59	349,823.9	319,718.1	Open data of Semarang City
Waste generated (ton)	419,189.45	478,905.55	456,873.35	466,010.79	430,749.75	Open data of Semarang City (b)
Waste generation per capita (kg/cap/day)	0.69	0.79	0.75	0.76	0.70	((a/b)/365)*1000
Waste collected to landfill (%)	88	88.5	89	89.5	90	Open data of Semarang City
The waste bank operated (unit)	32	25	54	47	47	Open data of Semarang City
Waste recycled (m3)	4544	4645	4747	4852	4959	Open data of Semarang City

As can be seen in Table 3, the total waste collected from landfills also increases each year, indicating a higher collection rate compared to other cities in Indonesia. The amount of waste increased dramatically from 2017 to 2018 following the increase in GDP and population. Even though the GDP and population are always increasing yearly, waste generation per capita remained stable at 0.75–0.76 kg/cap/day from 2018 to 2020 and decreased in 2021. The reason behind these phenomena is still unknown. Moreover, the decrease in waste per capita in 2021 may be due to the pandemic situation, which led to the reduction of anthropogenic activity in Semarang City [28]. The recycling system is conducted mainly by informal recycling actors, which account for 8.54% of the total inorganic waste generated, such as waste banks (or community-driven material recovery facilities (CdMRFs)), street pickers, scavengers, and recyclers. The number of waste banks counted in the open data of the Semarang City website is also slightly similar to the number of waste banks reported by Budihardjo et al. [29]. Meanwhile, a lower recycling rate (around 0.18%) of formal recycling was performed in integrated temporary waste storages (ITWSs). There are only four ITWSs that worked normally in 2018, and this is the reason why the recycling rate is extremely low. The amount of waste treated is also increasing, which may help reduce SLCP emissions. According to Ramadan et al., illegal waste dumping accounted for 12.02% of the total generated waste in Semarang City, of which 9.70% was openly burned [12]. As shown in Figure 2, illegal waste generation is lower in the city center and increases in rural areas.

Figure 2. Improper waste disposal generation in Semarang City.

3.2. OLS Output

Among the eight factors mentioned, only two of them can pass the significance level assumption of the global OLS model (probability and VIF). As seen in Table 4, the elevation factor is positively correlated with the number of improper waste disposal practices, while the population density shows otherwise. Both variables provide VIF values less than 7.5, indicating the absence of multicollinearity between the variables. The regression model for estimating the number of improper waste disposal practices can be seen through Equation (3), where Elev refers to elevation (m), PD relates to population density (people/km²), and IWD is the number of improper waste disposal practices (tons/day).

$$\mathrm{IWD}=1.667615+0.002896 \mathrm{Elev}-0.000098 \mathrm{PD}$$

(3)

Table 4. Results of OLS estimations.

Variable	Coefficient	Standard Error	t-Statistics	VIF
Intercept	1.667615	0.241355	6.909	-
Elevation	0.002896	0.001177	2.461	1.404
Population density	−0.000098	0.000021	−4.763	1.404

The residual and predicted plot results show that the observed data may be over- or underpredicted. The residuals may have nonconstant variance because the histogram is not fitted, and the residuals plot has yet to get too close to the reference line. Figure 3 shows predicted improper waste disposal generation using the global OLS regression model. The spatial autocorrelation report also found that the model developed is significant at the value of p = 0.05. Therefore, the designed clustered pattern indicates an overprediction of the model. Thus, detecting heteroscedasticity and spatial autocorrelation results makes GWR the best candidate for improving this model.

Figure 3. Standard residual of OLS map based on elevation and population density factors.

3.3. GWR Results

As can be seen in Figure 4, the local R² values indicate the relationship between the local regression model and observed $Y_i$ values. The low values (light red) suggest that the local model predicted poorly with the observed $Y_i$ values. The map determines whether the prediction is well or poorly developed. The residual of the GWR model is observed to have a random distribution with a value of p = 0.4308. Therefore, the model computed using local GWR is specified.

Figure 4. Adjusted R-squared for GWR results.

The GWR model also has a better result compared to the OLS model (See Table 5). The residual squares have a mark of 124.69. At the same time, the effective number value is 39.02, far from the bandwidth of 3012.67. These large bandwidths indicate a similarity between the OLS and GWR models [25]. Moreover, the sigma value is 0.95, which is small and shows a preferable estimation of standard deviation for residuals. The AICc values for OLS are higher than GWR, indicating that the GWR models demonstrated a better fit than OLS. These results indicate a good improvement in the global model. Therefore, the R² value is 2.5 times higher than the OLS model, showing a good fit for the model developed.

Table 5. Final diagnostic tests.

Variable	OLS Result	GWR Result
R-squared	0.250	0.691
Adjusted R-squared	0.241	0.606
AICc	605.4840	512.167

4. Discussion

In this study, we found that improper waste disposal generation is influenced by elevation above sea level and population density. As mentioned by Agovino and Musella, the willingness of separating household waste collection is impacted by elevation. In mountainous municipalities, the difficulty of separation is higher since it is limited to transportation operations. There is also an increase in collection costs because of the longer distance to collection points. More complicated monitoring operations can also be found in mountainous areas. The distances between collection points and households might be higher since, in many mountain areas, house dispersion is high, which increases efforts to carry out any separation activities [7]. In the same way, people tend to burn or participate in other illegal disposal practices rather than better waste management practices in mountainous and rural areas [6]. This finding is consistent with the findings of Agovino and Musella, where people in higher-elevation or mountainous areas tend to participate in improper waste disposal practices. The northern area of Semarang City is a coastal-type area with an elevation range between 0 and 25 m, while the southwest is a mountainous area with a slope of 50–200 m. Topographic factors were also found to induce urban growth in Semarang. City. As mentioned by Sejati et al., areas with a high elevation need greater costs for urban development. Thus, low urban growth can promote illegal waste disposal and reduce the possibility of recycling practices [30]. The government of Semarang City should focus especially on this finding to accelerate the development of decentralized waste management systems in many mountainous subdistricts where formal waste management services cannot reach the location [29].

Population density is also an important factor that should be considered in reducing the number of illegal waste disposal practices. This factor has the same pattern as elevation. Mountainous areas have dispersed households, which also indicates a low population density [7]. When the population density is lower, people tend to burn and dispose of their waste in their backyards. Residents in these areas believe they still have a large area of land, but collection is limited. In many rural areas, illegal dumping practices are higher because of this reason [6]. Even though a higher population density indicates a higher waste generation rate, people will not try to illegally dispose of their waste. As mentioned by Šedová, a higher population density will lead to a higher potential for detecting illegal dumpers. Since waste dumping and open burning of waste are legally prohibited by the government, illegal dumpers in higher population densities will not participate in the practice so easily [10]. Therefore, this finding should also be considered by the district or local municipalities to prevent illegal dumping from happening in their area. In summary, mountainous and rural areas have the highest potential for illegal dumping practices. In this area, encouragement of waste reduction, promotion of reuse and recycling, and increasing public awareness will be essential to reduce the illegal dumping practice, thereby reducing the amount of environmental pollution [1,6,9].

5. Conclusions

Based on the results, there is evidence that some determinants provided in this study cannot significantly explain illegal waste dumping generation, such as the number of waste collection facilities, land use, sex ratio, number of housing, number of populations, and number of public facilities. There is, at least, more than a 35% possibility of other factors that can make a better model to predict illegal waste dumping practices. Other factors or determinants for the subdistrict, such as waste generation per capita, average wage, poverty rate, unemployment status, education level, and internet literacy, may increase the significance level for explaining illegal waste dumping frequency. Moreover, it is also interesting to note that the number of waste collection points is not the reason behind improper waste disposal practices. Future research or studies should also determine the impact of decentralized waste management in the form of waste banks or ITWSs and the recycling rate in each subdistrict on illegal waste dumping practices. Meanwhile, there are also challenges in performing spatial regression studies since all data mentioned are not available, so a more robust and detailed field study in this field should be conducted to fill the gap. Therefore, this research provides a benefit to reducing illegal waste disposal, especially in Semarang City and other relevant cities with similar morphological characteristics.

Through these results, some policy recommendations can be drawn in Semarang City. First, the Semarang City Government should accelerate the development of a decentralized waste management system, especially in mountainous or higher-elevation areas and considering the lower population density of Semarang City. The increasing attractiveness of community-based solid waste management systems, locally known as waste banks, can be an alternative to reduce illegal waste dumping. Second, increasing waste collection services is essential to reduce illegal waste dumping. Third, law enforcement and punishment should be carried out by applying legal and social pressure on illegal dumpers. These recommendations are necessary to reduce the number of illegal waste dumping.