Rapidly rising transboundary atmospheric pollution from industrial and urban sources in Southeast Asia and its implications for regional sustainable development

Singapore (ca. 1° N, 103° E) is a highly urbanized island state with a land area of 721 km² (2018) that neighbours Indonesia and Malaysia and major international shipping routes (figure 1(a)). Singapore has a tropical climate, with abundant rainfall of 2166 mm yr⁻¹ (1981–2010 average), and is influenced by Northeast and Southeast monsoonal airflows [22]. The wind speed in Singapore increases from an overall average of 2.5 m s⁻¹ (1981–2010) to >10 m s⁻¹ during monsoon periods [23]. Singapore underwent rapid industrialisation and urbanisation from the 1960s to the late 1980s but has since transitioned towards a more services-based economy [24, 25].

Zoom In Zoom Out Reset image size

In the absence of natural lakes in Singapore, the country's oldest artificial reservoir [26] along with its catchment, a total of ca. 8 km² and part of the largest nature reserve in Singapore (table 1, figure 1(b), supplementary material (SM) figure 1 (available online at https://stacks.iop.org/ERL/15/1040a5/mmedia)), were selected as the principal focus of study. The main sources of water to the reservoir are direct rainfall and runoff from the catchment [27]. Water levels have been kept stable over time by actively discharging and pumping water during wet and dry periods, respectively [27]. Annual water inputs via direct rainfall and runoff from the catchment (ca. 11 million m³) are much larger than the reservoir capacity (ca. 3 million m³) [28]. The water residence time of the reservoir, therefore, may not exceed one year. There is negligible groundwater exchange to the reservoir [29] as a large part of the reservoir and its catchment sits on top of the Bukit Timah Granite formation, which has very low permeability [30]. Although the reservoir and its catchment are well protected today [26], the site has suffered from disturbances in the past, notably during World War II (1942–1945) [31].

Several reasons underpin the choice of Cr as a proxy of non-biomass burning-related TAP. First, Cr has well-documented biological toxicity, with the World Health Organization setting benchmarks for acceptable environmental concentrations of it [34, 35]. Second, although Cr can be released through natural processes, such as volcanic eruptions and soil erosion [36], a significant portion of Cr in the atmosphere is from anthropogenic activities (a global average of 59% in 1983) [37], notably the combustion of fossil fuels and manufacturing of steel [21]. Third, Cr tends to attach to airborne fine particulates and has a 7–10 d atmospheric residence time that enables long-distance transportation [38, 39]. Cr thus has the potential to act as a TAP, particularly in the context of Singapore, the territory of which immediately adjoins rapidly industrialising and urbanising parts of neighbouring Malaysia and Indonesia. Fourth, Cr is assumed to represent a broad range of potential transboundary atmospheric pollutants. This assumption is based on empirical evidence from the same study site in Singapore, from which it is clear that Cr concentrations vary over time in concert with other potential transboundary atmospheric pollutants, including lead, cadmium, zinc, copper and mercury [25, 40].

A catchment-reservoir mass balance approach depicting the movement of Cr in a hydrological system and depositions of Cr in sediments [41] was adopted. Following existing studies [32, 33, 42], we defined a 'catchment-reservoir system' (the sampling site), and a 'Singapore system', comprising all terrestrial areas of Singapore and hereafter referred to simply as Singapore, to construct a Cr budget (figure 2). Within the catchment-reservoir system, Cr inputs (from dry deposition (A_{DP_RE}), precipitation (W_PR), catchment runoff (W_RF), and pumped water (W_PT)) were assumed to be equal to the sum of Cr storage (in sediment (S) and reservoir water (W_RE)) and Cr outputs (through water overflow (W_OF) and withdrawal (W_WD)).

Zoom In Zoom Out Reset image size

Within Singapore, the total dry deposited Cr (A_DP) was assumed to be equal to the sum of dry deposited Cr from anthropogenic transboundary sources (A_{DP_TAPin}), anthropogenic local sources (A_{DP_LE}), and natural sources (A_{DP_BG}). Based on this framework, four steps were taken to determine the component of transboundary Cr pollution (SM section 1.1):

Step 1: Reconstruction of annual Cr fluxes in water (W) and sediments (S) in the catchment-reservoir system following approaches from Yang et al [43] and Sarkar et al [44]. The former was calculated for catchment runoff, precipitation, reservoir water, pumped water, withdrawal, and overflow (x) based on Cr concentrations in water (wc) and hydrological fluxes (Q) (equation 1, provided in full in SM equations 1 and 2). The latter was calculated for the whole reservoir from five sediment cores (j) using information including sediment Cr concentrations (sc), sedimentation rates ($\delta$), sediment focusing factors ($\gamma$), and represented reservoir areas (a) (equation 2, provided in full in SM equation 3).

$$\begin{equation}{W_x} = w{c_x} \times {Q_x}\end{equation} \tag{ 1 }$$

$$\begin{equation}S = \mathop \sum \limits_j s{c_j} \times {\delta _j} \times {\gamma _j} \times {a_j}\end{equation} \tag{ 2 }$$

Step 2: Reconstruction of the flux of total dry deposited Cr in Singapore (A_DP, equation 3) from the total dry deposited Cr in the catchment-reservoir system (A_{DP_RE}) through mass balance (equation 4) using data generated from Step 1. Calculations were based on the assumption that dry deposited Cr in the reservoir area (${a_{RE}}$) is representative of that across Singapore (${a_{SG}}$) (SM equations 4 and 5).

$$\begin{equation}{A_{DP}} = {A_{DP\_RE}} \times {a_{SG}} \div {a_{RE}}\end{equation} \tag{ 3 }$$

$$\begin{align}{A_{DP\_RE}} & = S + {W_{RE}} + {W_{WD}} + {W_{OF}} - {W_{PR}} - {W_{RF}} \nonumber\\ & \quad - {W_{PT}}\end{align} \tag{ 4 }$$

Step 3: Reconstruction of the flux of dry deposition of Cr from local sources in Singapore (A_{DP_LE}). Estimates were produced for three scenarios, which assume that the maximum annual contributions of local emissions to ${A_{DP}}$ from 1900 to 2017 are 75%, 50% and 25%, respectively, using 1972 as a reference year, when local emissions of Cr peaked [45] to determine the percentage of local Cr emissions that were deposited within Singapore (SM equations 6 and 7). Data on local Cr emissions in Singapore from 1900 to 2017 were compiled through a NEI analysis [45].

Step 4: Reconstruction of the flux of dry deposition of Cr from transboundary sources (A_{DP_TAPin}) through mass balance using data generated in Step 2 and 3 for Singapore (equation 5; SM equation 8). We assumed A_{DP_BG} equal to the mean value of A_DP from 1900 to 1920 when A_DP was in the lowest level.

$$\begin{equation}{A_{DP\_TAPin}} = {A_{DP}} - {A_{DP\_LE}} - {A_{DP\_BG}}\end{equation} \tag{ 5 }$$

Five sediment cores were collected from the reservoir in June 2017 using a UWITEC^TM gravity corer. The five cores were collected from the eastern (Core 1 and Core 2), central (Core 3 and Core 4), and western (Core 5) sections of the reservoir (table 1). All cores were subsampled at 1 cm vertical intervals upon collection. Following the United States Geological Survey water sampling protocol [46], water samples were collected from the reservoir (12 samples), main inflow streams (four samples), reservoir overflows (three samples) and rainwater (ten samples) from June to December 2017 (SM section 2.1, SM figure 1). It was not possible to sample the water that is occasionally pumped to the study site to adjust water levels in the reservoir because the inlet was inaccessible.

Historical economic, meteorological, and hydrological data were also collected from a variety of sources, including government statistical reports, third-party open databases, peer-reviewed publications, and local newspapers (SM section 2.2, SM figures 2 and 3).

Concentrations of Cr in sediment samples (mg Cr kg⁻¹) and water samples (µg Cr l⁻¹) were determined by inductively coupled plasma optical emission spectrometry and inductively coupled plasma mass spectrometry following protocols established by the United States Environmental Protection Agency [47, 48]. Triplicate measurements, replication tests and spike sample tests were also performed (SM section 2.3).

Chronological control was established for sedimentary evidence by the radiometric dating of sediment cores 1 and 3, based on activities of the isotopes ²¹⁰Pb [25]. The ages of material in cores 2, 4, and 5 were estimated through a process of sequence slotting involving comparisons across all five cores based on up-core variations in the abundances of spheroidal carbonaceous particles [49], organic-content based on loss-on-ignition analysis, and concentrations of zinc (SM figure 4). Core sequence slotting was performed using CPLslot software V.2.4 [50]. Chronological control also yields information on sedimentation rates and sediment focusing, both of which are useful in determining sediment Cr fluxes (SM section 2.4).

Contemporary Cr concentrations of water samples were used in the estimation of historical Cr fluxes due to a lack of historical measurements. Additionally, ${W_{PT}}$ was not included in the calculation of Cr budget owing to a lack of complete time-series records. Errors induced by using static concentrations and excluding ${W_{PT}}$ are discussed in SM section 4.1. Given that our calculation of Cr fluxes is based on a variety of data and associated expressions of error, the aggregated errors of Cr fluxes presented in this study only provide possible ranges of the mean fluxes without any further numerical estimation of probability.

Estimates of S are relatively low for the period 1900 to ca. 1920 at around 4 ± 3 kg Cr yr⁻¹. This period is also associated with the highest level of chronological uncertainty at ±11–17 years (figure 3(a), and SM sections 3.1–3.5, including SM figures 5–12 and SM tables 1–4). From the 1920s onwards, S increased and initially peaked at 10 ± 3 kg Cr yr⁻¹ in 1970 ± 6 years, following which it declined to a low of 7 ± 1 kg Cr yr⁻¹, a level similar to that of the 1950s, by the end of the 1990s. From ca. 2000 onwards, S increased again, peaking for a second time at 13 ± 2 kg Cr yr⁻¹ in 2017 ± 2 years.

Zoom In Zoom Out Reset image size

Estimates of W generally remained stationary over time, albeit with some interannual fluctuations (figure 3(a), SM figure 13, and SM table 5). Results show that ${W_{WD}}$ had the highest Cr flux at 8 ± 2 kg Cr yr⁻¹. By comparison, Cr fluxes of the other budget components, namely ${W_{OF}}$, ${W_{RE}}$, ${W_{RF}}$ and ${W_{PR}}$, ranged between 3 ± 1 kg Cr yr⁻¹ and 5 ± 1 kg Cr yr⁻¹. Estimates of W are associated with relatively small chronological errors as they were determined based on documentary sources of hydrological information that include clear calendar dates.

Measurements of Cr concentrations in reservoir water samples (mean 1 µg l⁻¹, SM section 3.6) were all significantly below the safety benchmark adopted in Singapore and recommended by the World Health Organization (50 µg l⁻¹) [34, 35]. Cr concentrations in the surface sediment samples (16–30 mg kg⁻¹, SM section 3.1.2) were also below the heavy metal benchmarks adopted in Canada (37 mg kg⁻¹) [54] and the United States (43.4 mg kg⁻¹) [55].

${A_{DP}}$ was at its lowest for the entire record between 1900 and 1920 at 5 ± 2 tonne Cr yr⁻¹ (figure 3(b)). This low-level was assumed to approximate to the natural background baseline and accords with a negligible intensity of Cr-emitting anthropogenic activities in both Singapore and SEA at the time; there was only one Cr emission source in Singapore before 1920, a privately-owned coal-fired power plant (capacity of 2 megawatts) [56] and there were no large-scale coal-fired power plants in SEA before 1960 [13]. From the 1920s onwards, A_DP slowly increased until the early 1940s, when an abrupt increase in fluxes from 6 ± 3 tonne Cr yr⁻¹ in 1942 ± 12 years to 10 ± 3 tonne Cr yr⁻¹ in 1946 ± 12 years occurred. This major change may have been linked to disturbances associated with World War II from 1942 to 1945 when bombing and associated civil strife and disruptions to water supplies occurred in Singapore [27]. From the 1950s, A_DP commenced a steady increase at an average rate of 2% per year, peaking at 12 ± 3 tonne Cr yr⁻¹ in 1970 ± 6 years. An abrupt deviation in steadily increasing A_DP occurred in the early 1960s possibly due to the effects of severe drought conditions that persisted in Singapore and neighbouring parts of the region from 1961 to 1963 [57] when annual precipitation and catchment runoff were ca. 20% lower than average [27] (SM figure 13). Following its ca. 1970 peak, A_DP declined slightly to 10 ± 2 tonne Cr yr⁻¹ in 2000 ± 2 years. Two obvious but short-lived declines in A_DP occurred during the 1970s, which could reflect the effects of two droughts recorded in Singapore in 1971 and 1976 (SM section 2.2.2). Noted that the drought-induced drops in A_DP might be overestimated based on equation (3), considering the reduced but unknown lake area over the duration of the drought. From ca. 2000 onward, A_DP started to increase for a second time, peaking at 16 ± 3 tonne Cr yr⁻¹ in 2017 ± 1 year.

A thorough discussion of local Cr emissions is presented in Chen and Taylor [45]. Local Cr emissions are associated with limited chronological errors as they were determined based on dated annual statistical reports and newspaper articles. In brief, a NEI analysis of Cr in Singapore from 1900 to 2017 showed that local emissions were relatively low (<1 tonne Cr yr⁻¹) from 1900 to the end of the 1950s. With the onset and rapidly increasing rate of industrialisation and urbanisation from the mid-1960s [24], local emissions increased sharply from 1 ± 0.3 tonne Cr yr⁻¹ in the early 1960s to an estimated peak of 61 ± 20 tonne Cr in 1972, following which they gradually declined to pre mid-1960s levels by the end of the 1980s. The decline likely results from a combined effect of increasingly stringent atmospheric emission regulations, notably the Clean Air Act of 1971 and its amendments [58], and a shift away from pollutive manufacturing industries to high-profit, low-pollution financial and services industries in Singapore [59].

Based on local Cr emissions, levels of ${A_{DP\_LE}}$ from 1900 to 2017 were estimated under scenarios 1, 2, and 3 (figure 3(b)). Results show that ${A_{DP\_LE}}$ peaked in 1972 at 8 ± 3 tonne Cr yr⁻¹ (Scenario 1), 5 ± 2 tonne Cr yr⁻¹ (Scenario 2) and 3 ± 1 tonne Cr yr⁻¹ (Scenario 3), or ca.12%, 8% and 4% of total local emissions, respectively. These estimated deposition rates are deemed to be good approximations of actual rates because Cr emitted to the atmosphere can be easily and rapidly transported beyond the borders of Singapore owing to the country's relatively limited territorial extent, monsoonal airflows that prevail throughout much of the year, and the potentially long atmospheric residence time of Cr.

${A_{DP\_TAPin}}$ was negligible prior to 1920, and largely (90–100%) originated from natural sources (figures 3(c) and (d)). From the 1920s onwards, ${A_{DP\_TAPin}}$ slowly increased from 0 ± 3 tonne Cr yr⁻¹ to 3 ± 3 tonne Cr yr⁻¹ by the end of the 1930s. Given that the lower bounds of the aggregated errors intersect with zero, the possibility remains that Singapore received little or no TAP until the end of the 1930s. In the 1940s, ${A_{DP\_TAPin}}$ fluctuated significantly, ranging from 1 ± 3 tonne Cr yr⁻¹ to 5 ± 4 tonne Cr yr⁻¹. This fluctuation could be attributed to disturbances associated with World War II and its immediate aftermath. Evidence of relatively low levels of Cr pollution from transboundary sources from the 1900s to 1940s accords with limited economic activity in neighbouring parts of SEA, as is evident in GDP per capita data [60]. In the 1950s, ${A_{DP\_TAPin}}$ slightly increased from 3 ± 3 tonne Cr yr⁻¹ to 5 ± 4 tonne Cr yr⁻¹, with a potential overestimation of 6% of ${A_{DP\_TAPin}}$in the 1950s arising from the adoption of static Cr water concentrations and the exclusion of W_PT (figure 4(a), SM section 4.1). This slow and small increase in TAP may reflect nascent economic development, post-World War II, in the form of the early stages of industrialisation and urbanisation in the region.

Zoom In Zoom Out Reset image size

Scenario 3 appears to result in the most reliable estimate of ${A_{DP\_TAPin}}$ from the 1960s to the 1980s, which indicates a stable level of deposition of around 4 ± 3 tonne Cr yr⁻¹. A series of three abrupt declines in ${A_{DP\_TAPin}}$ occurred from the early 1960s to the mid-1970s, possibly associated with the aforementioned droughts in the region in 1961–1963, 1971, and 1976. By comparison, Scenarios 1 and 2 indicate significant declines in Cr depositions over the whole period. These declines seem highly improbable, however, judging by the generally low GDP per capita of the majority of countries in the region throughout this period (figure 5(a)) and a positive correlation between pollutant emissions, e.g. Cr and CO₂, and early stages of economic development in ASEAN countries [45, 61]. Given that there is no evidence of climatic changes that could possibly alter the transportation process of Cr, scenarios 1 and 2 were rejected. Scenario 3 indicates that ${A_{DP\_TAPin}}$ was relatively low but fluctuated widely prior to the late 1980s and was not unduly influenced by the choice of chronological errors arising from radiometric dating (figure 4(b)). The low level of ${A_{DP\_TAPin}}$ presumably reflects comparatively attenuated levels of economic development in neighbouring parts of the region prior to the onset and the gathering pace of industrialisation in neighbouring parts of Malaysia and Indonesia in particular [62, 63]. By comparison, industrialisation commenced relatively early in Singapore (in the 1960s) characterised by export-oriented manufacturing [24]. Manufacturing output increased 68-fold in the three decades from 1960, peaking at ca. 25% of GDP by the end of the 1980s [59]. As a result, ${A_{DP\_LE}}$ contributed substantially to ${A_{DP}}$ from the 1960s to the 1980s, reaching a maximum of 25% in 1972 when local emissions peaked. ${A_{DP\_LE}}$ declined rapidly after the promulgation of the Clean Air Act in 1971, returning to a pre-industrial level and accounting for less than 1% of ${A_{DP}}$ by the end of the 1980s.

Zoom In Zoom Out Reset image size

From ca. 1990 onwards, ${A_{DP\_TAPin}}$ slowly increased to 6 ± 3 tonne Cr yr⁻¹ by the end of the 1990s. This increase coincided with the rapid development of export-oriented economies in many countries in SEA, including Singapore's immediate neighbours, Malaysia and Indonesia [64]. As a result, steel production [52] (figure 5(b)) and energy consumption increased two-fold in the 1990s [65]. Industrial development was more widespread in the region at the time, judging from the 44% increase of nitrogen oxide and 40% increase of sulfur dioxide emissions in SEA over the same decade [66, 67]. As opposed to the previously low and stable levels of deposition, this increasing trend since the 1990s provides unambiguous evidence of rising transboundary Cr pollution in Singapore.

${A_{DP\_TAPin}}$ rose rapidly from ca. 2000 at an average rate of about 3% per year, doubling to 11 ± 3 tonne Cr yr⁻¹, or 15 ± 4 kg km⁻² yr⁻¹, in 2017 ± 2 year, under all three scenarios. The estimate accords with the results of other studies in urban areas across the world, ranging from 10–63 kg km⁻² yr⁻¹ [68, 69] (SM section 4.2, SM table 6). The use of constant contemporary Cr water concentrations and the exclusion of W_PT have limited impacts on both the estimates (average overestimation of 12%) and the overall increasing trend (figure 4(a)). The contribution of ${A_{DP\_TAPin}}$ to ${A_{DP}}$ rose from ca. 50% in 2000 to ca. 70% in 2017, with natural sources accounting for the remainder as local emissions remained negligible. The rapid increase in ${A_{DP\_TAPin}}$ from 2000 to 2017, in agreement with the increasing regional mean concentrations of ambient PM_2.5 by 2.7% per year since 2000 [70], is amongst the highest in the world [71]. Simultaneously, Singapore experienced a decrease of 1% per year in mean visibility [72]. This decrease of visibility does not accord with the effective enforcement of stringent environmental regulations in Singapore [73], which has significantly alleviated local pollution, as evidenced by the low levels of local Cr emissions [45].

There are three possible reasons for the rapid increase in ${A_{DP\_TAPin}}$ after 2000. First, accelerating economic growth in the region (figure 5(a)), driven at least in part by a six-fold increase in foreign direct investment (figure 5(c)) [53]. These capital inflows were mainly oriented towards Cr-emitting industries, such as steel manufacturing (figure 5(b)), and energy production (figure 1(a)) in order to boost export-oriented industrialisation [74]. Second, the relocation and aggregation of industries in the context of free trade and globalisation. A possible case associated with this type of pollution is the Singapore-Malaysia-Indonesia Growth Triangle scheme, which involves locating labour-intensive and pollutive manufacturing factories in Indonesia and Malaysia, with cleaner high-tech and services industries based in Singapore [75]. Cr emissions generated from industrial centres located in neighbouring parts of Indonesia and Malaysia (figure 1(b)) are relatively easily transported to Singapore by the prevailing monsoon winds due to their close proximity [45]. Third, rising levels of maritime pollution. The port of Singapore ranks as the second largest in the world in terms of volume of container traffic [76]. The number of ships docking in Singapore in 2017 was 16 times the number recorded in 1985 [77, 78]. Ships emit substantial levels of pollution, including Cr [79–81], and are thus highly likely to have contributed to a rising trend of ${A_{DP\_TAPin}}$ in Singapore over the past two decades.