Trans-boundary variations of urban drought vulnerability and its impact on water resource management in Singapore and Johor, Malaysia

The hazard of drought, through accumulated deficits in precipitation (meteorological), streamflow (hydrological), and soil moisture (agricultural), poses a major risk across numerous regions affecting natural and human sectors [1, 2]. Investigating risks associated with changing drought frequencies and intensities is prevalent globally, especially under contexts of climate change [3]. A comprehensive analysis of these risks is fraught with complications, such as varying levels of confidence and likelihood over (a) global- and regional-scale drought trends, as well as (b) on attribution of human influence to recent droughts [4], and (c) from complex interplay in regional/local scale distinctions of demographic sensitivity and socio-economic adaptive capacities directly influencing vulnerability to drought [5].

In the rapidly developing Southeast Asian region, scant assessment exists of drought risk (i.e. the potential for consequences where something of value is at stake and where future outcomes are uncertain, recognizing the diversity of values) and vulnerability (i.e. the propensity to be adversely affected arising from the interactions of exposure, sensitivity, and adaptive capacity to hazards) on coupled natural-human systems [1]. Apart from increased risks to water (and food) shortages directly arising from reductions in precipitation and streamflow, concern also exists over increasing vulnerability of lowland dipterocarp forests from the confluence of drought and human activities (e.g. logging, fragmentation, and fire) [6–11]. This vulnerability is also compounded by slow drought recovery in low tropics relative to other regions [12]. Further, agricultural drought conditions enhances surface wildfire and generation of smoke aerosols observed regionally, resulting in heightened mortality and morbidity especially in the islands of Borneo and Sumatra [13, 14].

Concerns over current and future drought risks in Southeast Asia are justified given the synergistic regional projections of future urbanisation patterns and climate change. Rapid urban growth rates, coupled with uncertainties in precipitation distribution, affect future development and consequent policies within Southeast Asia vis-à-vis droughts e.g. the complications arising from transboundary water issues within the Mekong basin [15, 16]. Such drought policy dilemmas are also present between the city-state of Singapore and the state of Johor, Malaysia. A complex water policy relationship exists between these two entities, which manifests in the politics underpinning the 1962 Water Agreement entitling Singapore to draw, treat, and import water from a Johor catchment until 2061, while providing a fraction of the treated water (2% of supply) in return for Johor's municipal use. While there is much discussion examining the political aspect on this transboundary issue [17], little emphasis exists on examining how historical climate affects drought in this part of Southeast Asia, and its impact on ensuing policies with respect to past and present urban vulnerability.

As such, this paper examines temporal trends of drought, specifically arising from changes in climate over Singapore and Johor, and discusses how past variations in urban drought vulnerability affects municipal water policy within and between these areas. We examine changes to drought exposure, sensitivity and adaptive capacity within both regions through assessing climate and archival data related to drought risks. We conclude by examining varying vulnerabilities with respect to anticipated urban resilience, defined as how settlements effect policies that anticipate, absorb, accommodate, or recover from the effects of drought [4]. Analyses of regional drought vulnerability can be complicated by a dearth of data, especially on socio-economic aspects of coupled human-natural systems, and this study illustrates a potential approach to integrating data to dynamically assess vulnerability.

2.1. Geographical context

We investigate a region at the southernmost point of Peninsular Malaysia (figure 1). The Republic of Singapore is a small, heavily urbanised island city-state (~720 km² land area; 2015 population ~5.5 million) [18, 19]. Across the Johor Strait lies the state of Johor (~19 000 km²; 2015 population ~3.6 million) [20]. The majority of Johor's land-use is forested or cultivated for oil palm plantation, while urban areas are typically concentrated in its ten administrative district centres in the state; of note is that approximately half of the state population are concentrated in the state capital Johor Bahru [20, 21].

Zoom In Zoom Out Reset image size

Both regions experience Köppen Af climates, with high annual mean temperatures (~27 °C) and total precipitation (~2400 mm). Monthly precipitation is unevenly distributed, with two distinct monsoon seasons occurring during the Northern Hemisphere winter (December–February) and summer (June–August). Precipitation in this region largely occurs during the winter monsoon, especially in December and early January. In Johor, a precipitation gradient exists during this season with eastern coastal settlements receiving more rainfall relative to inland areas in the centre and West Johor [22]. Both regions are also strongly influenced by El Niño–Southern Oscillation events, with generally warmer and drier (cooler and wetter) conditions associated with El Niño (La Niña) conditions.

2.2. Water resource management in Singapore and Johor

We adopt a commonly-used systems framework defining urban vulnerability to drought as the combination of attributes of (a) exposure to drought, (b) demographic sensitivity to this hazard, and (c) the urban socio-economic capacity to adapt towards drought impacts (figure 2) [28, 29]. In this study, spatio-temporal changes in urban vulnerability is examined via corresponding variations in each attribute, and can be documented for both Singapore and Johor using a combination of available climate and archival data.

Zoom In Zoom Out Reset image size

To ascertain the exposure to drought in our study area, we quantified the precipitation deficit affecting watersheds by applying the Palmer Drought Severity Index (PDSI) [30]. PDSI is a standardised unitless index ranging from –10 (dry) to +10 (wet), and it accounts for moisture supply (precipitation) and demand (potential evapotranspiration) through inputs of local temperature data, soil conditions, and a physics-based water balance model. In general, a location's PDSI < −3 indicates severe drought.

We selected PDSI over other indices (e.g. precipitation-only Standard Precipitation Index, or percentiles of normal precipitation) as including temperature and soil moisture conditions as inputs enables assessing abnormalities of both water-demand and—supply with respect to drought [31]. PDSI has been applied in previous drought research within Southeast Asia [32–34]. Further, the region's low-latitude and low-elevation reduces a potential imprecision of PDSI arising from a key assumption that all precipitation is immediately available for supply as rainfall instead of snow. This assumption affects PDSI accuracy in high-latitude or -altitude regions where melting snowpack can reduce drought during water stressed periods in summer, but does not apply here.

Monthly PDSI was calculated from daily weather station temperature and precipitation data (table 1) using a custom MATLAB script [35]. The respective Available Water Content (AWC) used to calculate the Singapore and Johor soil moisture at field capacity are 350 mm m⁻¹ and 440 mm m⁻¹, which are typical values for the Rengam (Rengam-Jerangau) soil series found in Singapore (Johor) [36]. While we acknowledge a valid critique that autocorrelation may be present at monthly timescales of PDSI, we are confident the long duration of climate datasets used would not affect interpretation for observed drought conditions reported via PDSI.

While quantifying physical exposure to drought is achieved via PDSI, we utilise qualitative approaches to document societal attributes i.e. examining both demographic sensitivity, and each settlement's socioeconomic capacity to adapt to drought. This method is often utilised in the absence of quantitative data of sufficient quality or resolution e.g. data of daily water consumption or supply in settlements during severe drought events that are often unavailable for this region. As such, we adopt an alternative approach from the social sciences through examining historical media content, e.g. archived news media reports, which enhances vulnerability analyses by ascertaining crisis periods of droughts affecting the general public in a region, and corresponding societal responses to water scarcity [37]. Further, archival analysis of the effectiveness of historical socioeconomic policies to drought adaptation—after responding to crisis periods of drought—is an important tool towards planning for future adaptation to extreme weather events [38].

Zoom In Zoom Out Reset image size

To this end, we examined several publicly available archival media sources e.g. Singapore and Malaysian newspapers in which droughts were explicitly mentioned. We used the search terms 'drought' or 'dry spell' in the digital newspaper archives NewspaperSG and Factiva [39, 40]. NewspaperSG was used to search for Singapore-based newspaper articles prior to 1989, while Factiva was utilised to obtain English (Singaporean and Malaysian newspapers) and Bahasa Melayu (Malaysian newspapers) newspaper articles after 1989 (NB: one of the authors (CJC) is fluent in Bahasa Melayu). Articles reporting on drought/dry spell incidence, as well as associated impacts and responses, were noted from both archival datasets. These articles were grouped according to severe and/or extreme drought periods listed in the PDSI dataset, and their content categorised and analysed specifically for local and regional drought impacts and policy responses.

4.1. PDSI variations

There was considerable variability in ranges of derived PDSI (figure 3). The coastal stations of Changi (−7.26 to +5.31 (σ = 2.24)), and Mersing (−8.80 to +6.66 (σ = 3.02)), observed greater variability in their respective datasets compared to the inland Senai station (−4.59 to +5.19 (σ = 1.90)). Severe drought events (PDSI <−3) are labelled sequentially in figure 3 by station, and occurred more frequently in Changi and Mersing. These accounted for 11.4% (92 out of 804 months) and 17.8% (109 out of 612 months) in the respective Changi and Mersing datasets (figure 4). In contrast, severe drought only occurred at Senai for 38 out of 515 months (7.3%).

Zoom In Zoom Out Reset image size

Historical ONI (Oceanic Niño Index) data were obtained from NOAA's Climate Prediction Centre (NOAA, 2017) and plotted (figure 3). Significant correlation coefficients at p <0.05 levels were found between ONI and PDSI datasets at the coastal Changi (r = −0.46) and Mersing (r = −0.29) stations, which are more exposed to monsoonal precipitation, but not at the inland Senai station (r = 0.02); this significant inverse relationship is unsurprising given the strong influence of El Niño on synoptic climatology in Southeast Asia. More severe drought spells also overlapped with strong El Niño events. Notable episodes in 1982/83, 1997/1998, and 2015/2016 coincided with severe drought spells of considerable duration and intensity at all stations. Given the proximity of stations to each other, severe drought events often occurred simultaneously in two or three of the studied locations; a total of fifteen episodes, with the longest lasting over two years (June 2014 to June 2016) in Changi and Mersing (table 2).

Table 2. Occasions when ≥2 stations had monthly PDSI < −3.

Period	Station	Duration (Months)	Mean PDSI	Min PDSI
December 1982–August 1983	Changi	9	−5.13	−5.90
	Mersing	9	−4.64	−5.78
June 1990, August–September 1990	Senai	3	−3.31	−3.66
	Mersing	3	−3.22	−3.51
February 1991–April 1991	Changi	3	−3.48	−3.61
	Senai	3	−4.02	−4.59
July 1991	Changi	1	−3.79	−3.79
	Senai	1	−3.71	−3.71
	Mersing	1	−3.01	−3.01
August 1991	Changi	1	−3.17	−3.17
	Senai	1	−3.74	−3.74
Oct 1991	Changi	1	−3.65	−3.65
	Mersing	1	−3.00	−3.00
March 1992–June 1992	Changi	4	−3.77	−4.22
	Mersing	4	−3.81	−4.25
September 1992	Changi	1	−3.24	−3.24
	Mersing	1	−3.71	−3.71
August 1997–March 1998	Changi	8	−6.67	−7.26
	Mersing	8	−4.67	−5.71
April 1998–July 1998	Senai	4	−3.47	−3.55
	Mersing	4	−5.64	−6.87
September 1998–February 1999	Senai	6	−3.80	−4.32
	Mersing	6	−7.59	−8.24
August 2005–September 2005	Changi	2	−3.17	−3.89
	Mersing	2	−3.19	−3.28
June 2014–June 2016	Changi	25	−4.73	−6.36
	Mersing	25	−4.41	−6.67
July 2016–October 2016	Changi	4	−4.83	−4.96
	Senai	4	−3.78	−3.89
	Mersing	4	−8.54	−8.8030
November 2016–December 2016	Senai	2	−3.63	−3.71
	Mersing	2	−7.63	−8.29

There is evidence suggesting recent droughts have become more severe and prolonged, especially at Changi and Mersing. More extreme droughts (PDSI ≤ −6) have only occurred in the last two decades (figure 5), with the lowest PDSI of −8.80 recorded at Mersing in August 2016.

Zoom In Zoom Out Reset image size

4.2. Drought impacts

4.3. Responses to droughts

Both Singapore and Johor developed immediate- and long-term measures as responses to drought impacts (table 4). Immediate responses differed between regions, largely stemming from the varying severity of reported drought impacts. In Singapore, recent responses were largely limited to governmental advisories issued to consumers. In contrast, unconventional immediate responses were reported in Johor e.g. cloud-seeding operations to induce rain during severe droughts since the early 2000s. At the peak of the 2015/16 severe drought, Singapore turned provider when the Johorean government requested additional treated water (~21 MLD) to fulfil Malaysian transboundary demand.

The long-term strategy to reduce drought risks and vulnerabilities in both regions largely involves policies directed at water supply and infrastructure development. Singapore planned to increase water supply by increasing water catchment areas, developing sub-surface water supplies and spaces for storage, as well as exploring alternative sources of imported water [23, 43, 44]. Of particular importance is developing climate-resilient water resources unaffected by droughts. Presently, five NEWater plants exist in Singapore with combined production capacity of ~760 MLD, capable of meeting up to 40% of current demand. By 2061, projected NEWater supply will account for 55% of future water demand. Currently, two desalination plants operate in Singapore with a combined capacity of ~455 MLD (~25% of current water demand). Three more desalination plants are expected to begin operations by 2020. With a capacity of ~136 MLD each, projected desalinated water will meet up to 30% of Singapore's 2061 water needs (figure 6).

Zoom In Zoom Out Reset image size

Effective domestic water demand management was also applied in Singapore. Through a combination of economic and punitive policies, such as pricing water to discourage wastage, implementing a Water Conservation Tax, and prohibiting sales of water-inefficient plumbing fittings, toilet cisterns and washing machines (table 4), Singapore's per-capita domestic water consumption steadily decreased from 165 L day⁻¹ in 2003 to 151 L day⁻¹ in 2015. The stated policy aim is to reach 140 L day⁻¹ demand by 2030 [23].

Most of Johor's long-term drought response is strictly via intra-state infrastructure development to increase water supply, including (a) constructing coastal barrages preventing sea water intrusion into catchments, and (b) inter-basin transfer of water between rivers to enhance storage in existing reservoirs (table 4). Development of recycled and desalinated water production has also been considered, but concerns over commercial viability remain. Both national and state governments have also recognised the importance of water demand management [45]. Substantial evidence of these efforts, however, was not found.

Currently, drought vulnerability between Singapore and Johor appears different. Similar regional levels of exposure and demographic characteristics exist, but a lack of critical impacts from severe drought episodes, coupled with limited immediate-, and more diverse long-term responses on both water-demand and supply aspects, suggest that a high level of adaptive capacity and consequent low vulnerability to drought exists in Singapore (figure 2). Conversely, the greater number of water disruptions in Johor, larger variety of immediate responses to severe drought, and long-term responses concentrated on infrastructural development of water supply strongly suggest relatively higher vulnerability to drought in Johor exists due to lower adaptive capacity.

The important corollary from this present vulnerability analysis is how future uncertainty vis-à-vis drought risk for these settlements can be effectively managed. Perhaps the biggest challenge in reducing Singapore's drought risk is improving demand management through water conservation. Evidence exists of Singapore water users being aware of water conservation through perception surveys and by consistent decreases in domestic per-capita water consumption [46]. During drought episodes, however, water consumption has paradoxically increased by 2.6% and 5.0% in 1983 (C1) and in 2014 (C7) respectively [47, 48]. These reported consumption increases suggest that water use advisories during droughts may be ineffective in lowering demand.

Singapore's long-term policies to reduce its drought vulnerability through expanding its water resources are commendable, specifically with its successful development of NEWater and desalinated water sources. The production of both, however, is relatively energy-intensive [49]. Producing 1 m³ of recycled (desalinated) water requires 1.1 (3.5) kWh compared to 0.2 kWh for treating catchment water [50]. Using current technology, conservative projections indicate a quadrupling of energy consumption for producing desalinated water by 2060 [23]. Further, NEWater and desalinated water production requires careful management of rejected concentrates and brine effluents. Impacts of brine disposal to local environments from desalination to receiving waters have been well documented [51–53]. Wastes from recycled water production, on the other hand, contain high concentrations of various toxic compounds; many of which persist in the environment [54]. PUB claims that current recovery rates for recycled water is currently 75% [23]. While it is likely that current technologies will improve energy efficiency and effluent treatment, long-term questions remain on the economic and environmental viability of these two water sources.

NEWater in Singapore is safe for direct consumption, but it is produced mainly for non-domestic use as an indirect potable source; it is added to local reservoirs, which act as environmental buffers, and re-treated in waterworks [55]. Despite this inefficiency, the use of reclaimed water as an indirect rather than a direct potable source is unsurprising. Only one out of 3300 registered water reclamation projects worldwide, located in Windhoek, Namibia, is for direct potable use [56–58]. Widespread public scepticism towards drinking treated wastewater can be attributed to emotional grounds of disgust, a visceral psychological reaction known as the 'yuck' factor [59, 60]. As an approach to reduce drought vulnerability, it remains to be seen if Singapore can eventually complete the transition to direct supply of recycled water for potable domestic use.

Zoom In Zoom Out Reset image size

Presently, imported water from Johor fulfils ~60% of Singapore's total demand. Apart from drought, another pertinent risk is this source requires safeguarding against degradation. For Singapore, the risk of poor imported water quality is arguably larger than that of drought, especially as the Johor River is susceptible to contamination. In July 2016 and October 2017, high concentrations of ammonia were detected in the river that led to the temporary shutdown of PUB's water treatment plant [61–63].

Unlike Singapore, Johor is relatively water-rich. In 2010, total water demand for all sectors comprised only 22% of available surface water resources [45]. By 2050, total in-state water demand is expected to double, and even then, it still represents <40% of available resources (figure 7). The challenge for Johor, thus, is to effectively manage these resources and lower drought risk through complementary demand- and supply-side policies.

Of particular importance is protecting available surface water resources from contamination [64, 65]. Nearly all in-state rivers have documented degradation to water quality [66, 67]. The ammonia contamination within the Johor River in 2016 and 2017, which caused the temporary shutdown of PUB's waterworks, also disrupted operations at two Johor-managed waterworks; in 2016, ~600 000 Kota Tinggi consumers were affected [68]. In 2017, ~1.8 million residents in Johor Bahru also experienced supply disruptions [69]. Activities from intensive agriculture, particularly from oil palm plantations within water catchments, are also recognised threats that have resulted in significant fish die-off events [70, 71]. In the long-term, policies regulating Johor water pollution from agricultural or industrial activities could not only improve environmental quality but also lower drought vulnerability—with the proviso that these policies are effectively implemented.

Despite having surplus supplies, Johor should also consider policies minimising water loss from supply and conveyance, particularly in view of scarcity during recent severe drought spells. Clearing of forests in water catchments for farming activities potentially increases water loss via reducing surface and sub-surface storage capacities, and has been previously linked to water level decreases in local reservoirs [72–74]. Further, non-revenue water (NRW) is reportedly widespread in Johor [65, 75]. Although ranked third lowest among Malaysian states, Johor's NRW in 2016 is ~26% compared to Singapore's ~5% [44, 76]. As such, a combination of land-use management within catchments, and enhancements in water infrastructure and oversight against NRW could potentially reduce overall water supply loss and lower drought vulnerabilities.

Johor's larger per-capita water use relative to Singapore suggests that its demand-side management policies have potential for improvement [77]. Policy focus could shift from the supply-side towards reducing consumer demand through e.g. water pricing considerations, mandatory water conservation requirements, and societal engagement campaigns increasing awareness of water scarcity. If effectively implemented, these policies can complement supply-side initiatives to reduce Johor's drought vulnerability.

Improving overall local water governance to drought risks has transboundary benefits to Johor and Singapore. Supply disruptions arising from droughts can be reduced, likely diffusing potential political tensions between the two states as they have historically. With continued supply surpluses, Johor can continue selling water to Singapore beyond the expiration of the current agreement in 2061 if politically expedient to both parties. Singapore has made known its desire to extend the agreement, even with a higher water price [78]. Revenues generated from its sale, can in turn, be used to enhance water governance resilience in Johor.

Lastly, assessing drought risk and future drought vulnerability must also consider climate change impacts on temperature and precipitation baselines in Southeast Asia. While El Niño potentially influenced recent drought extremes, increasing global temperatures and variable precipitation regimes associated with climate change in this region could also be factors heightening hazard risk through changes in drought frequency and intensity in the region as summarised in [1].

Presently, clear variations in trans-boundary drought vulnerabilities between Singapore and Johor exist despite similar levels of exposure. Based on archival analysis, differences in vulnerability largely stems from variations in regional adaptive capacities. Both Singapore and Johor have historically, with varying success, applied different techniques and approaches over the short- and long-term in managing water resources with respect to drought risk. Discussing these present demand- and supply-side approaches reveals several ways in which reduction of local drought risk can be effected. A concern, which is beyond the scope of this study, is how stakeholders in each region should consider projected changes in drought intensity and frequency from global climate change, and how likely increases in future drought exposure will affect concomitant risk profiles for Singapore and Johor, as well as possibly increase potential strains in political relationships between these countries on this transboundary water resource management issue.

This research was directly supported by a NUS Research Grant R-603-000-214-490 (PI: WTLC). We acknowledge useful comments from Dr. Alan Ziegler and Dr. Robert Wasson during the development of this paper. Helpful suggestions from the editor and two anonymous reviewers towards improving this paper are also acknowledged.