By Rhys Thompson, Jan Kamer, Kenneth Er, Ng Boon Gee and Brett.C. Phillips

The Gardens by the Bay, a visionary cluster of three distinctive waterfront gardens in the heart of the city of Singapore, epitomizes the nation’s vision of ultimately being a city in a garden. Covering 101 hectares (249.57 acres) of land at Marina Bay, the site is comprised of the Gardens at Marina South (Bay South), Gardens at Marina East (Bay East) and Gardens at Marina Centre (Bay Central). When complete, the gardens will capture the essence of Singapore as a premier tropical garden city with an ideal environment in which to live and work, showcasing the application of cutting-edge innovations in environmentally friendly technology, sustainable natural resource management internationally.

Bay South is located in the heart of Singapore’s new downtown. With an area of 54 hectares (133.43 acres), it incorporates two conservatories covering around 2.5 hectares (6.17 acres); steel ’SuperTree‘ structures that are 30 to 50 meters (98.42 to 164.04 feet) tall and clad with vertical green and floral displays; a series of theme gardens; open green spaces; a network of paved areas and a lake system.

The lake system is about five hectares (12.35 acres) in surface area and two kilometers (1.24 miles) in length, and circulates around the eastern, southern and western boundaries of the gardens (see Figures 1 and 2). Given the integration of the lake to the adjacent marina reservoir, maintenance of water quality is a critical consideration in the design of the lake system. Singapore has a tropical climate of constant high temperatures (between 24.1 to 31°C [75.38 to 87.8°F]) and rainfall (1,118.9 to 3,452.4 mm [44 to 135 inches]) year-round, thereby resulting in relatively higher rates of nutrient infiltration and run-off. This, coupled with the use of fertilizers in the gardens, provides a challenging environment for the design of the lake.

Water quality modeling is undertaken as part of the design within an adaptive management framework. These investigations and resulting measures that were implemented are summarized here. They include source control, such as bioretention and permeable paving; lake management, such as pumping and aeration, and operational response actions.

Water-sensitive urban design strategies and water quality modeling
In the preliminary stages for the design of the lake, several water-sensitive urban design strategies were researched and incorporated into the design of the lake system. It is first necessary to ensure optimal hydrologic flow and capacity that balances drainage requirements and constant water level with aesthetic amenity. This covers detention time, water depth and flow velocity. The gardens also seek to minimize surface run-off through the provision of rainwater tanks and permeable paving in parking areas.

A crucial, key treatment of the lake system is in the incorporation of aquatic planting for the purpose of filtration and treatment of water, with a view to reduce total suspended solids, nitrogen and phosphorous levels, and chlorophyll-a (i.e., algal bloom and eutrophication). The plantings are contained in filter-bed systems (a series of cascading bioretention systems), and lake-edge filter beds (bioretention systems along the lake edge) that provide treatment of runoff directly into the lake from local  catchment. Maximal contact between water and the aquatic plantings is maintained by configuring the lake with a high length-to-width ratio. Two marginal wetlands are also incorporated, one at each terminal end of the lake system (one is in the form of a bog garden and another takes the form of a freshwater wetland). In particular, the freshwater wetland serves as a pretreatment system for water taken in from the marina reservoir.

Water quality modeling was undertaken to evaluate the effectiveness of these water-sensitive urban design strategies in the lake design. Operational norms of fertilizer application were input into the models. In doing so, water- sensitive considerations could be incorporated into the gardens’ future operational fertilization regimes.

To test the effectiveness of the strategies, two models were constructed to assess the expected water quality within the lake system; i.e., total suspended solids (TSS), total nitrogen (TN), total phosphorus (TP), chlorophyll-a and dissolved oxygen (DO).

Catchment runoff model
A catchment runoff model is developed using MUSIC (Model for Urban Stormwater Improvement Conceptualization). While the MUSIC model was developed in Australia, it had previously been used for catchment-based water quality modeling of the Eastern Catchment, where parameter values were developed for Singaporean conditions.

To derive the event mean concentrations (EMCs) of total nitrogen and total phosphorus loading for the landscaped areas within the gardens (termed ’garden areas’ here, it occupies 64 percent of the site), fertilizer application rates were taken to be higher in the initial four years of planting establishment than in the subsequent years of maintenance. The EMCs of TN and TP loading were then extrapolated from an earlier study of a relatively similar catchment in the north of Singapore, where two golf courses abut a reservoir. Similarly, the EMCs of TN and TP leaching were based on the sampling of runoff from the two golf courses in the north (see Table 1).

Leaching rates were determined known fertilizer loading rates and  typical golf course EMC values, which were typically between 5 to 10 percent (15 percent was adopted for this study, which was considered to be conservative). The leaching rates were neither a baseline nor a control set.

Estimation of EMCs and soil parameters for non-garden areas (e.g., paved areas, car parks, etc.) were based on the work undertaken by CPG Consultants and Cardno (2007), and those reported in guideline documents such as Australian Runoff Quality.

Figure 1: Gardens by the Bay study area
Figure 2: Bay South Gardens layout

Hydraulic and water quality model
A 1-D hydrodynamic model is established for the lake using the SOBEK modeling system (a software package for 1-D and 2-D simulation of flood forecasting, drainage systems, irrigation systems, sewer overflow, ground-water level control, river morphology, salt intrusion and water quality) from Deltares, the Netherlands. This model is capable of modeling both hydraulics and water quality processes occurring within the lake. Cross sections for the one-dimensional model are obtained from design drawings. Inflows to the hydrodynamic model were obtained from the MUSIC model. Estimates of the water quality concentrations of inflows from the marina bay reservoir are based on advice from the  Singapore Public Utilities Board (PUB). The efficiencies of the treatment strategies incorporated within the lake design are modeled for TSS, TN, TP, chlorophyll-a and DO.

In addition, pumping scenarios are incorporated into the model to simulate the adequacy of circulation and flushing of the lake. Pumping scenarios taken into account are pumping of water from the marina reservoir alone, and pumping from the marina reservoir and recirculating water within the lake system using irrigation pumps. Irrigation demands are pegged at a pumping rate of 42 kL/hr. (11,095 gph) over a 24-hour period. The second scenario reduces the demand for pumping water from the marina reservoir. A number of pumping rates are tested as a part of the investigations. It is to be noted that floating islands of aquatic vegetation were excluded from the analysis, as performance data of these features were not readily available. A key objective of the lake design is to maintain an operational water level of reduced level (RL) 101.8 m (333.98 ft.) for the purpose of aesthetics, with an upper limit of RL 102.1 m (334.97 ft.) and a lower limit of RL 101.6 m (333.33 ft.). Several pumping options extract water from the marina reservoir at a lower rate than the irrigation requirements. Under these scenarios, over a ten-year simulation period, the lake drops below the target water level on average 16 days each year. This is relatively rare and suggests that the design of the lake and pumps is adequate in maintaining the desired operational water level. There remains the option of reducing the irrigation-pumping rate or increasing the pumping rate of water from the marina reservoir during periods where the water level drops below the desired levels. Results of the modeling show that proposed strategies of incorporating aquatic plantings and minimization of surface run-off were able to effectively reduce levels of TSS, TP and TN.

At most locations, treatment measures achieve a minimum reduction in average annual TSS, TP and TN loads of 85, 65 and 45 percent respectively. Performance of treatment measures for TP is illustrated in Figure 3. These reduction efficiencies are well within the standards of the state of New South Wales in Australia. The predicted effectiveness of aquatic plants to reduce TSS, TN and TP is consistent with those found in other studies in the tropical and temperate systems, although this could vary with plant type and form.

Figure 3. Predicted effectiveness of treatment measures in reducing average annual TP loads Singapore’s tropical climatic conditions of high temperatures and rainfall and moderate solar radiation suggest a significant risk of algal blooms occurring within water bodies. It is, therefore, important to manage the lake system to reduce the risk of algal blooms. Management includes both the control of nutrients entering the lake system, as well as pumping rates within the lake system.

Chlorophyll-a results from the models indicate that the combination of pumping from the marina reservoir, together with recirculation pumping, results in the best outcomes for the lake, rather than the scenario of pumping from the marina reservoir alone, due to a lower reliance on the water quality within the marina reservoir. A comparison of monthly trends for chlorophyll-a, global radiation and rainfall based on modeling results is provided in Figure 4. Global radiation remains relatively constant throughout the year, varying between approximately 400 and 500mWh/cm2. Growth of chlorophyll-a is generally not limited within these radiation ranges. It is also apparent that the period of highest risk for elevated levels of chlorophyll-a in the lake system is the months of November, December and January, when rainfall peaks. This is not surprising since these are the months of highest run-offs. This suggests the need to be mindful of the timing of application of fertilizers, which can be built into the maintenance regime for the gardens in the future.

Dissolved oxygen
Long-term averages for DO levels do not differ significantly across the lake system. Different locations within the lake system, however, respond differently to runoff events. Southern areas are more likely to experience frequent low DO levels, as they respond quickly to any runoff event due to the small volume of water. Northern areas of the lake system contain larger volumes of water, which can buffer the impact of runoff events. Hence, those areas are less likely to experience low DO levels (see Figure 5 for summary of the DO levels across the lake). These results highlight the need for incorporation of aeration systems.

An adaptive management framework has been set up to facilitate the management of the lake system within the gardens. Such a framework enables field managers to better manage uncertainties in natural systems through the use of sensitivity analysis or modeling in decision-making, and the formulation of a monitoring framework that measures a set of pre-determined outcome indicators over time.

Intervention via management strategies within the system can then be undertaken with increased confidence. For such a framework to be successful, it must be implemented from the onset of the design process. In this case, the modeling study was undertaken in parallel with the design and development of the gardens lake system. A reiterative process (added water features and aeration systems were found necessary in consultation with the landscape architects) of modeling and design refinement ensues. Predictions of water quality indicators (i.e., TSS, TN, TP, chlorophyll-a and DO) are used to refine the design. The refined design is in turn fed back into the models for further validation. The predicted water quality indicators from the models are then used as benchmarks for the purpose of monitoring during the operation of the lake. Fixed locations within the lake system were established and modeled from the onset, and will continue to serve as sampling points for the purpose of monitoring in the future (see Figure 6).

One of the key features in the Bay South Gardens is a lake system that is integrated with the adjacent marina reservoir, and that has been designed with the implementation of a set of water-sensitive urban design strategies. The effectiveness of the proposed strategies and the results suggest that the strategies are largely effective in ensuring water quality in terms of TSS, TN and TP, but refinements will need to be made in order to enhance DO in the southern areas of the lake.

Operational adjustments will also need to be made in order to maintain the water level at the desired aesthetic levels at all  months times. Likewise, timing of fertilization will need to be considered in order to minimize the occurrence of algal blooms. Use of water quality modeling in the lake design is undertaken within an adaptive-management framework and will continue to set measurable indicators as benchmarks for subsequent monitoring. This will facilitate the operationalization and management of the lake system.


  1. Table 1: CPG Consultants, Singapore, 2002, Pte Ltd., Feasibility Study Seletar Serangoon Reservoir Scheme, Public Utilities Board, Singapore, unpublished.

About the authors
Rhys Thompson is an Environmental Engineer with Cardno, Australia, specializing in water quality, water resources, environmental economics and environmental design projects. He has extensive experience in water quality, including management of sampling programs, interpretation of water quality data and numerical modeling, as well as flood and floodplain management, water resource management, together with hydraulic and hydrologic investigations for surface and groundwater, urban, riverine and estuarine catchment studies, numerical model development, calibration and appraisal.

Jan Kamer is Senior Consultant with CPG Consultants, Singapore with over 20 year experience in environmental engineering and management. He was Design Manager for the feasibility study of the Punggol and Serangoon Reservoir Scheme for PUB and Project Manager for the Gardens by the Bay Water Quality Modeling Study. More recently, Kamer was involved as an Environmental Specialist with development of a Water Quality Management Master Plan for the Eastern Catchment for PUB. Overseas projects included hydraulic and water quality studies for urban developments in Vietnam, Middle East and China.

Kenneth Er, National Parks Board, General Manager for Gardens by the Bay, oversees development and operations. He is currently researching the response of plants to cool conditions under glass, and continues his research interests in conservation biology, with emphasis on the design and management of nature reserves within an urban landscape.

Ng Boon Gee, Assistant Director, (Development), National Parks Board and Dr. Brett C. Phillips, Director, Cardno, are contributing authors.


Comments are closed.