Assessment of heavy metal levels in an urban river in the Philippines using an unconstrained ordination- and GIS-based approach: evidence of the return of past pollution after the 2013 Typhoon Haiyan (Yolanda)

Toxic metal species are a globally recognized environmental concern because they pose a pressing threat to the aquatic environment and human life with their persistence and toxicities (Kumar et al. 2019, 2021a). As a result of anthropogenic activities, they seep into bodies of water through point sources, such as industrial, municipal, and domestic wastewater effluents, and through secondary pathways via surface runoff, erosion, or atmospheric deposition (Pericherla et al. 2020; Ali et al. 2022). The affected aquatic biota is vulnerable, especially benthic organisms since trace metals from the sediments accumulate in the food chain (Khan et al. 2015). When the physicochemical and hydrodynamic conditions change in bodies of water, sediment heavy metals are resuspended resulting in dispersion and secondary pollution (de Souza Machado et al. 2016).

Rapid industrialization exacerbates heavy metal pollution in aquatic ecosystems (Kumar et al. 2020a). This phenomenon is particularly true for densely populated cities in developing countries that do not adequately implement municipal waste management systems to prevent the indiscriminate discharge of wastes into rivers and lakes (Haghnazar et al. 2021; Kumar et al. 2021b; Bhat et al. 2022). The majority of studies on heavy metal pollution in urban areas in the Philippines focused on Manila (Prudente et al. 1994; Gorme et al. 2010; Paronda et al. 2019) and the City of Meycauayan (Bulacan), the latter being considered one of the most polluted cities in the developing world (Diwa et al. 2021a). Moreover, the impact of urbanization on heavy metal pollution in the aquatic environment outside these urban hotspots has been hardly addressed (Decena et al. 2018) except for investigations focused on the impact assessment of mining activities in the provinces, particularly in the Mindanao and Palawan group of islands (Cabuga et al. 2020; Ebol et al. 2020; Diwa et al. 2022).

Tacloban City, a city outside of the National Capital Region, posted one of the highest urbanization rates in the country (https://www.tacloban.gov.ph/). Alongside its rapid development, informal settlements proliferated on vacant government lots along riverbanks and creeks. In a 2011 survey, 38% of the total population in 37 barangays were informal settlers (Rahman & Ferdous 2011). However, when Typhoon Haiyan (Yolanda) struck in 2013, Tacloban became the worst-hit region in the Philippines. All the informal settlements along the coastal areas and riverbanks were wiped out. During the first years after the devastation, the homeless survivors moved to emergency and temporary housing tenements built by non-governmental organizations, international donors, and the government. Although there were programs for socialized permanent housing settlements in safer locations by the government, the affected residents were reluctant as the locations were too far from their former urban and coastal livelihoods. As a result, many houses from the informal sector were rebuilt from the debris of the disaster (Hara et al. 2018).

Heavy metal contamination from the Mangonbangon river, which cuts across Tacloban City, was assessed in 2016 by Decena et al. (2018). Incidentally, the study provided a snapshot of the rebirth of heavy metal pollution as sampling was implemented three years after the devastation by Typhoon Haiyan. Ten trace metals were detected and the magnitude of the average concentrations followed the order: Fe>Mn>Zn>Cu>Cr>Ni>Co (Decena et al. 2018). From the multivariate analysis, Cu, Cr, Ni, and Fe were in the same cluster that indicated anthropogenic sources of heavy metals. This study aims to re-visit the 2016 Mangonbangon river data and dissect nuances of source apportionment by applying a recently developed GIS-based geostatistical modeling technique (Gu & Gao 2019). We hypothesized that the apportionment profile may differ and can add new information to assist the local government in developing targeted and practical strategies to address the toxic heavy metal sources. This study is also timely and relevant given the renewed calls for rehabilitation and preservation of the Mangonbangon river as new structures encroach on the aquatic systems and other dangerous areas in the city.

The 4-km Mangonbangon river, which traverses informal settlements and commercial areas in the city, flows from an upstream wetland and drains into the San Juanico Strait. Tacloban city, a first-class highly urbanized city and the provincial capital of Leyte Island, has 251,881 inhabitants and a population density of 2,377 km² based on the 2020 population and housing census (Philippine Statistics Authority 2020). According to the Corona classification system, the region has a Type II climate characterized by an absence of a dry season and heavy rainfall from November to January. The average annual temperature is 27.2 °C, with the highest monthly temperature of 31.4 °C occurring in May and August, and the lowest being at 22.7 °C in February. Farming (rice, coconut, and sugarcane), fishing, and mining are the main economic activities in the region. Forestry workers and farmers account for around 20% of the workforce. Despite being a coastal city, only 3.6% of the workforce in Tacloban works in the fishing sector (Marteleira et al. 2017; Decena et al. 2018).

Several models following the United States Environmental Protection Agency (US EPA) guidelines were used to assess the human health risks associated with heavy metal exposure (Means 1989). The Hazard Index (HI) and Lifetime Cancer Risk (LCR) estimate the non-carcinogenic and carcinogenic risks due to heavy metal exposure, respectively. The risk models differentiated adults from children because of their differing physiology and behavior relevant to health risk estimates (Hu et al. 2017).

We applied the newly developed method of UOGM to identify the sources of heavy metal pollution in the Mangonbangon river. The approach followed three successive steps. First, the Enrichment Factor (EF) was calculated to differentiate heavy metals of anthropogenic and natural origins. Second, the source patterns of the heavy metals were determined using a non-metric multidimensional scaling (NMS). And lastly, to spatially evaluate anthropogenic influences, the NMS scores were interpolated and visualized. The results were visualized using weighted colors of the NMS scores to elucidate the spatial variation of the heavy metal pollutants since interpolation is not possible in a linear body like a stream or river.

Using Fe as a normalizer, the EFs of the heavy metals from the sediments are presented in Decena et al. (2018). NMS was performed using Bray-Curtis distance measure at 30 random initial starting configurations using the vegan package of R (ver. 4.0.4). Like the Principal Component Analysis (PCA), NMS is a multivariate analysis that reduces the dimensionality of the dataset (Dexter et al. 2018). The position of the objects in the reduced dimensions is useful in identifying the underlying patterns and similarities between the species. However, in contrast to PCA, NMS uses a rank-order (non-metric) relationship between inter-sample dissimilarity and inter-sample distance instead of a linear combination to produce an ordination (Fasham 1977). Since NMS does not assume a linear relationship, the distance measured is suitable for any data, and the method can also determine the best positions of the objects in the reduced dimensions. Unlike other ordination techniques, NMS employs ‘stress’ or the measure of similarity between the best configuration and the original multidimensional space to evaluate the fit of the ordination (Dexter et al. 2018).

Table 1 shows the descriptive statistics for heavy metals from surface sediments of the Mangonbangon River sampled in 2016. The mean heavy metal concentrations were in the following order: Fe > Mn > Zn > Cu > Cr > Ni > Co (Decena et al. 2018). The sediment materials obtained from the 14 sampling stations along the river showed low variations in Fe, Zn, and Mn concentrations. For the remaining heavy metals, particularly Co, there was a high CV % observed which indicates anthropogenic origins since previous studies have shown that heavy metals with high variations in concentrations are usually associated with human activities (Zhou & Wang 2019). Recent comparisons of heavy metal concentrations from sediments of different rivers showed that levels of Cr and Ni at the Mangonbangon river did not exceed the values from the Tigris (Turkey), Ganga (India), and Korotoa (Bangladesh) rivers. However, there were higher levels of Co and Ni in the Mangonbangon river relative to that of estuaries in Malaysia (Asare et al. 2022). The Mangonbangon River also has more than thrice the Ni as in the Meycauayan River, one of the world's top 30 dirtiest river systems (Diwa et al. 2021a).

Heavy metals have toxicological effects on humans as Co, Cr, and Ni identified in the Mangonbangon River are known carcinogens (Briffa et al. 2020). With the high population density along the riverbank due to informal settlements, it is critical to assess the heavy metal risk exposure of the population. We, therefore, calculated the Hazard Index (HI) and Lifetime Cancer Risk (LCR) values. These risk models have been extensively used to assess health risks from toxic elements in several media and types of exposures, i.e., ingestion, inhalation, and dermal contact (Weissmannová et al. 2019).

On the other hand, LCR is a measure of the potential for the population to develop cancer as a function of both metal toxicity and exposure doses. An LCR value that exceeds 10⁻⁴ is associated with a high risk of developing cancers. As seen in Figure 2(c), the LCRs for both adults and children are below 10⁻⁴, indicating no carcinogenic risk to the population. Moreover, our results show the sampling sites with the highest LCRs for adults and children: MS04, MS06, MS07, MS08, DS09, DS10, DS11, and DS14 (Figure 2(c)). It is noteworthy that these sites were in the river's midstream and downstream segments and where the informal settlements dominate. Localized small-scale point sources, like metal works, auto shops, gas stations, and commercial areas, are also present (Decena et al. 2018). Among the three dominant carcinogenic metals, the mean CR contribution of the heavy metals to the LCR for both adults and children is Co (60.7%) > Ni (22.8%) > Cr (16.5%), as shown in Figure 2(d). The genotoxic activity of Co is generally associated with its ability to generate reactive oxygen species (ROS) via a Fenton-like reaction. Co is also known to modify various cellular proteins, like zinc finger motifs, disrupting DNA metabolic activities and genome stability (Lison et al. 2018).

To distinguish whether the carcinogenic heavy metals are from anthropogenic or natural sources, we applied UOGM by first calculating the enrichment factor (EF) for each heavy metal. EF is a popular tool used to assess the intensity of anthropogenic pollutant deposition on soils and sediments. It is regarded as a good indicator of enrichment by anthropogenic influences because it allows the comparison of heavy metal concentrations from the sediment versus the background (Barbieri 2016). Background concentrations are geogenic natural content in the soil before anthropogenic inputs (Cicchella et al. 2005). The criteria used by other investigators to infer heavy metal apportionment are as follows: <0.5, natural; 0.5≤EF≤1.5, either natural or anthropogenic; and > 1.5, mainly anthropogenic (Gu et al. 2016).

To shed light on the nature of the heavy metal sources, we finally used GIS to understand the spatial variation in NMS ordination. Since interpolation was not possible for a linear body like a river, we used circles showing the weights of the influences of the two dimensions identified by NMS. Figure 4 illustrates the different areas along the river as influenced by NMDS 1 and NMDS 2. We noted a striking relationship between the NMS ordination distribution with dwelling density and types of activity at the sample sites. NMDS 1 consistently influenced sampling stations with high dwelling density and informal settlements. Through Google Map visualization of the sampling areas (upstream: US03; midstream: MS04, MS05; downstream: DS09, DS10, DS11, DS12, DS13), we notice the high density of informal settlements along the downstream segment. Given the lack of municipal services and the challenges by the informal settlers in Tacloban city (Yee 2018), we assume that a considerable amount of sewerage and garbage drains from these areas. Thus, it was highly probable that Cr, Cu, and Ni, associated with NMDS 1, are from the municipal wastes coming from high-density dwelling areas along the river (Ishchenko 2019). Meanwhile, we observed a generally decreasing trend in NMDS 2 from the upstream to downstream sampling stations. The NMDS 2-associated heavy metals (Co, Mn, and Zn) can be ascribed to agricultural activities from upstream wetlands and localized small-scale industrial activities along the river. These heavy metals are usually in trace amounts in phosphate fertilizers (El-Taher & Althoyaib 2012) and steel production (Horn et al. 2021). The observed diminishing trend for NMDS 2 from upstream to downstream stations supports the hypothesis that these heavy metals originated from point sources such as upstream farms and localized sources, like metalworks and auto shops near the river.

To extract interpretable patterns from the heavy metal data in the Mangonbangon river, Decena et al. (2018) employed both clustering and linear multivariate analyses. PCA was used to visualize the relationship among heavy metals in a continuous vector space by creating a subspace that captures the salient features of the data set. In doing so, there can be instances where data dimensionality may be insufficiently low with information compression (Camacho et al. 2020). Thus, in principle, such a method may not always be robust. Consequently, the previous investigators failed to pinpoint the specific sources despite attributing the heavy metals (Cr, Cu, Mn, Ni, and Zn) to anthropogenic activities (Decena et al. 2018). Our analysis significantly differed from the results in Decena et al. (2018) since we were able to show that Co was a product of anthropogenic activities in specific areas along the river. From these findings, we also highlighted the robustness of the UOGM.

NMS combined with GIS is a compelling approach since insights from spatial information and geographic features aid the increased data dimensionality from NMS. However, this approach also has limitations. A major drawback is that NMS sometimes fail to find the best configuration after getting stuck on local minima. However, this technical issue can be surmounted by computing for multiple ordinations. Given the superior performance in identifying sources of environmental pollution, there are only a handful of published examples of NMS (Gu et al. 2018; Gu & Gao 2019).

Lastly, we also noted two critical insights from our results. First, as Tacloban City became one of the worst-hit regions when Typhoon Haiyan (Yolanda) passed the Visayas on 8 November 2013, the entire city was destroyed. With winds reaching 315 km/h, the super typhoon decimated all residential areas as it was the strongest and most destructive to have made landfall in Philippine history (Santos 2013). With Tacloban city given a chance to re-build (or ‘re-boot’) itself and strictly implement its urban settlement and zoning programs, the heavy metal profiles of the Mangonbangon river three years after the Typhoon Haiyan (Yolanda) indicate the return of anthropogenic activities and the ‘age-old’ pollution problems. Second, despite the low EF of Co, its source identification is not trivial because it had the most significant contribution to the LCR in the population. Note that Co metal and its compounds are used in the production of alloys, cemented carbides, and hard metals. It is also used as catalysts, drying agents in paints, and additives in animal feed and pigments. This case, so to speak, allowed us to identify the ‘gorilla in the room’ of heavy metal toxicants.

There is an enrichment of heavy metals in the surface sediments of the Mangonbangon River. Using UOGM, we deduced that the present heavy metal pollution in the river is caused mainly by municipal discharge and agricultural activities. Heavy metals such as Cr, Cu, and Ni are linked to municipal discharges, while Co, Mn, and Zn are attributed to agricultural and industrial activities upstream. With the more precise data on source apportionment of heavy metal pollution in the river at its earliest stage, the analytical approach and our results can guide the Department of Environment and Natural Resources-Environmental Management Bureau at Region VIII in its Mangonbangon River Water Quality Improvement Program (EMB 8 Conducts Advocacy Campaign in Mangonbangon River Communities [press release] 2019). Similarly, environmental groups can be better guided in implementing their river-clean-up drives by better addressing the root causes of the pollution (Leyte Normal U leads rehab of Tacloban's Mangonbangon River [press release]. Philippine News Agency December 19 2020 2020). The application of indigenous and industrially relevant phytoremediators in the vacant spaces and riverbanks, such as bamboo (Chua et al. 2019; Go et al. 2021), or other rooted macrophytes in sediment may help in the mitigation of the river pollution (Kumar et al. 2020b).

All relevant data are included in the paper or its Supplementary Information.

The authors declare there is no conflict.