Assessing the Impact of Industrial Wastewater Reuse As a Demand-Side Management Strategy for Potable Water Deficit in Selangor, Malaysia

Assessing the impact of industrial wastewater reuse as a demand-side management strategy for potable water deficit in Selangor, Malaysia

By Nini Fatahna Muhamad Sopian

Submitted in fulfilment of the requirements for the degree of Doctor of Philosophy

Centre for Environment and Sustainability Faculty of Engineering and Physical Sciences University of Surrey October 2018

Summary

Policy makers are now re-aligning their thinking towards demand-side management (DSM), where for the industrial sector specifically, strategies have been emphasising water consumption reduction through conservation, increased efficiency, and wastewater recycling. Although DSM-based strategies have been seen as a viable option to delay the need for large capital investment and to avoid irreversible environmental impact, their implementation at a large scale have been largely unknown as is implementation has generally been reported at a small-scale. As such, this research attempts to assess the effectiveness of implementation of decentralised wastewater reuse facilities in industrial parks as a DSM-based strategy targeted at the manufacturing sector.

The research first assessed the vulnerability of three reservoirs in Selangor – namely Semenyih, Klang Gates, and Langat dam – against water deficiency events using a modelling approach to analyse their: i) reliability, via the use of an inflow-demand reliability index, and ii) resilience, through a water supply resilience index. Total potential potable water savings were then estimated for all industrial parks in the area served by each reservoir and subsequent potable water treatment plant. The impact of the potential water savings through reuse of wastewater in industrial parks within an overall water supply system was then evaluated.

The feasibility of this DSM-based strategy was assessed in terms of its technical, economics, and environmental impact. Four treatment trains corresponding to the two cases of i) potable and ii) non-potable reuse were assessed in terms of their removal efficiencies necessary to comply with the current potable and non-potable water quality standards. The financial cost of each treatment train was determined for both capital and operational costs, while the environmental assessment was assessed using a Life Cycle Assessment (LCA) approach based on several environmental impact categories.

Findings for the three dams indicated varying level of vulnerability, with Klang Gates dam being the most vulnerable, while Semenyih the least. For Klang Gates and Langat dams, the resultant IDR and WSR indexes achieved similar conclusions; both indices resulted in almost constant negative values for both dams, suggesting that an effective DSM-based strategy could possibly delay the need for implementation of an SSM strategy. For Semenyih dam, it was observed that the system is approaching system vulnerability, where lower IDR and WSR values were observed in the later years as compared to historical scenarios. Here, it can

be hypothesised that a well-planned DSM-based strategy might eliminate the need for a SSM strategy.

It was found that there was minimal number of industrial parks within the Klang Gates water supply area, further reducing the feasibility of wastewater reuse to reduce the vulnerability of this WSS. Conversely, it was estimated that, for Langat and Semenyih WSS, approximately 308,903 and 430,281 m3/month of water consumed by the premises within the identified IPs. This translates to approximately 266,430 and 374,009 m3/month of water savings if 90% of the total demand was discarded as effluent, and treated at 95% recovery rate. The savings calculated for this reuse option could contribute to a savings of 5.5% (8.8 MLD) and 12.2% (12.47 MLD) for the Langat and Semenyih dam respectively. These accumulated savings had a different impact on each dam’s vulnerability. For the Langat dam, recalculation of the IDR and WSR values with the savings indicated little change to the system’s overall vulnerability. In contrast, recalculations for the Semenyih dam resulted in positive changes for the WSR specifically, indicating reduced vulnerability of the system.

The feasibility of four treatment trains corresponding to both reuse options for their technical, financial, and environmental dimensions were considered. The general findings in all three dimensions reinforced the option of non-potable reuse, specifically the NP-1 treatment train, which resulted in the lowest cost while providing the highest technical capability to remove contaminants from the industrial wastewater to meet both non-potable and potable water reuse standards. Additionally, from the environmental perspective, the resultant water for reuse had a lower human toxicity impact without an additional blending requirement, though this produced a higher terrestrial eco-toxicity impact due to the contaminants removal to landfill.

On the other hand, it was also illustrated via this research that the additional financial resources, institutional reconfiguration of existing water service industry, as well as current awareness of the industrial players itself will limit the effective implementation of the identified strategy. As such, it was concluded that, although the use of industrial wastewater for manufacturing sector is technical, financial, and environmental feasible, its deployment however, is improbable against the current organizational structure and current awareness of the users.

Declaration of originality This thesis and the work to which it refers are the results of my own efforts. Any ideas, data, images, or text resulting from the work of others (whether published or unpublished) are fully identified as such within the work and attributed to their originator in the text, bibliography or in footnotes. This thesis has not been submitted in whole or in part for any other academic degree or professional qualification. I agree that the University has the right to submit my work to the plagiarism detection service TurnitinUK for originality checks. Whether or not draft have been so-assessed, the University reserves the right to require an electronic version of the final document (as submitted) for assessment above.

Nini Fatahna Muhamad Sopian

15th October 2018 ______

III

Table of Contents

Summary ...... I Declaration of originality ...... III Table of Contents ...... IV List of Tables ...... VII List of Figures ...... IX List of Acronym ...... XI List of Notation ...... XIII Acknowledgements ...... XVI Chapter 1 : Introduction ...... 1 1.1. Malaysia’s socioeconomic transition ...... 1 1.2. Top-down approach for water supply imbalance: SSM vs DSM ...... 2 1.3. Capturing the full extent of market-based policies ...... 5 1.4. Research rationale and focus ...... 7 1.5. Aim and objectives of research ...... 8 1.6. Overview of thesis ...... 9 Chapter 2 : Benchmarking of current water supply system’s vulnerability ...... 12 2.1. Introduction ...... 12 2.2. Aim and research questions for Chapter 2 ...... 15 2.3. Literature review ...... 16 2.3.1. Vulnerability, resilience, and reliability of a WSS, and its linkages to WSS performance ...... 16 2.3.2. Methods of estimating inflow into reservoirs ...... 19 2.4. Methodology ...... 22 2.5. Results and discussion ...... 31 2.5.1. Reservoir inflow estimation...... 31 2.5.2. Inflow, outflow, and water level ...... 32 2.5.3. IDR/WSR scatterplot for reservoir vulnerability assessment ...... 35 2.5.4. Detailed IDR and WSR analysis for Semenyih, Klang Gates, and Langat dams .. 40 2.6. Conclusion ...... 44 Chapter 3 : Estimation of potential water savings from industrial premises ...... 46 3.1 Introduction ...... 46 3.2 Aim and research questions for Chapter 3 ...... 51 3.3 Literature review ...... 52

3.3.1. Estimation of industrial potable water consumption ...... 52 3.3.1.1 Econometric models...... 52 3.3.1.2 Unit water demand ...... 54 3.4. Methodology ...... 57 3.4.1. Industrial park water balance ...... 59 3.4.1.1. Water demand estimation (inflow) - estimation of total potable water consumption for industrial premises...... 59 3.4.1.2. Industrial park clustering ...... 62 3.4.1.3. Wastewater effluent estimation (outflow) ...... 63 3.4.2. Projected saving at dam and potable water treatment plant ...... 63 3.5. Results and discussion ...... 64 3.5.1. Industrial park clustering and water consumption estimation ...... 64 3.5.2. Water reuse potential at industrial park ...... 71 3.5.3. Impact of water saving on water supply system ...... 74 3.6. Conclusion ...... 77 Chapter 4 : Technical, financial, and environmental cost of wastewater reuse options . 79 4.1 Introduction ...... 79 4.2 Aim and research questions for Chapter 4 ...... 81 4.3 Literature review ...... 81 4.3.1 Wastewater reuse design for industrial wastewater effluent ...... 81 4.3.1.1 Advance wastewater treatment for industrial wastewater ...... 81 4.3.1.2 Industrial wastewater reuse treatment trains for direct non-potable industrial water reuse ...... 85 4.3.1.3 Industrial wastewater reuse treatment trains for direct potable industrial water reuse 88 4.3.2 Cost estimation for treatment trains ...... 90 4.3.3 Life cycle assessment for wastewater treatment and reuse ...... 92 4.4 Methodology ...... 95 4.4.1 Treatment train removal efficiency assessment ...... 95 4.4.2 Cost estimation of treatment train ...... 98 4.4.3 LCA methodology ...... 99 4.4.3.1 Goal and scope ...... 99 4.4.3.2 Life cycle inventory ...... 100 4.5 Results and discussion ...... 103 4.5.1 Effectiveness of contaminant removal using identified treatment trains ...... 103 4.5.1.1 Non-potable reuse ...... 103

4.5.1.2 Potable reuse ...... 107 4.5.2 Assessment of cost function utilisation for treatment train cost estimation ...... 112 4.5.3 LCA interpretation ...... 115 4.5.3.1 Non-potable reuse ...... 115 4.5.3.2 Potable reuse ...... 122 4.5.4 Discussion and comparison of non-potable and potable reuses ...... 129 4.6 Conclusions ...... 131 Chapter 5 : Overall assessment of decentralised wastewater treatment plant implementation for industrial reuse ...... 133 5.1. Introduction ...... 133 5.2 Comparison of DWWTP implementation with other DSM-based approach ...... 133 5.3 Conclusion ...... 154 Chapter 6 : Conclusions and future work ...... 155 6.1. Review of findings ...... 155 6.2. Wider implication of research ...... 156 6.3. Future research avenues ...... 157 References ...... 159 Appendix A: Incidents of water rationing as reported in various tabloids ...... 180 Appendix B: NAM model using MIKE Hydro Basin software ...... 182 B.1 Methodology ...... 182 B.2 Results and findings ...... 184 Appendix C: List of parameters monitored and upper limit from Fifth and Seventh Schedule ...... 196 Appendix D: Maximum allowable limit for non-potable and potable water reuse ...... 198 Appendix E: Coefficient accuracy assessment ...... 202 E.1 Overall profile for light, medium, and heavy industry ...... 202 E.2 Coefficient accuracy assessment ...... 206 Appendix F: Results for potable and non-potable reuse ...... 212 Appendix G: Profile of interviewed companies and summary of initial findings ...... 220 Appendix H: Database for Chapter 4 ...... 224

List of Tables Table 2-1: Percentage of growth in the number of connections, and volume of treated water for 2013 to 2014 (Malaysian Water Association (MWA) 2015) ...... 13 Table 2-2: Summary of Semenyih, Klang Gates, and Langat dam ...... 22 Table 2-3: Summary of inflow/outflow of Semenyih, Klang Gates, and Langat dam...... 32 Table 3-1: Overview of water supply system ...... 57 Table 3-2: Characteristic and potable water consumption for individual industrial park ...... 69 Table 3-3: Baseline and 15th year projection for cumulative water consumption by all IPs in Langat and Semenyih WSS ...... 72 Table 3-4: Total estimated baseline and 15th year projected water saving from non-potable reuse for Semenyih and Langat WSS ...... 72 Table 3-5: Total estimated baseline saving from potable reuse for Semenyih and Langat WSS ...... 73 Table 3-6: Total estimated 15th year projected saving from potable water reuse for Semenyih and Langat WSS ...... 73 Table 4-1: Unit processes in non-potable and potable treatment trains ...... 103 Table 4-2: Compliance to non-potable reuse limit by NP-1 and NP-2 ...... 104 Table 4-3: Compliance to potable reuse limit by P-1 and P-2 ...... 110 Table 4-4: CAPEX, OPEX, and unit cost of the four treatment trains ...... 112 Table 4-5: CAPEX, OPEX, and unit cost for Semenyih WSS ...... 114 Table 4-6: Life cycle assessment (LCA) for the production of 1m3 of treated mixed industrial wastewater for non-potable water reuse ...... 116 Table 4-7: Life cycle assessment (LCA) for the production of 1m3 of treated mixed industrial wastewater for potable water reuse ...... 123 Table 5-1: Cost and capacity comparison between DWWTP, New WSS, and New Ponds projects . 134 Table 5-2: Comparison of 5 selected states with regards to potable water growth rate, percent used by the manufacturing sector, and the industrial water tariff (in MYR) ...... 136 Table 5-3: Comparison of techno-economic feasibility between DWWTP, New WSS Project, and New Ponds Project ...... 142 Table 5-4: Policies regarding water and energy efficiency (extracted from Adham, Merle and Weihs (2013)) ...... 149 Table 5-5: Comparison of organizational, policy, and guideline requirements between DWWTP, New WSS Project, and New Ponds Project ...... 152 Table B-1: Rainfall and evaporation stations within SF 3118445 catchment ...... 185 Table B-2: Calculation of average observed evaporation data for station EV 3117370 ...... 186 Table B-3: Stations used to fill in rainfall gaps for station RF 3118102 ...... 187 Table B-4: Result of double-mass curve for each station ...... 187 Table B--5: Rainfall and streamflow data completeness and streamflow rating curve check ...... 188 Table B-6: Weightage for rainfall stations in SF 31183445 catchment ...... 191 Table B-7: Accuracy assessment of model validation for year 2014 - 2016 ...... 192 Table E-1: Number of consumer points (CP) used for coefficient estimation ...... 202 Table E-2: Descriptive statistics for light, medium, and heavy industry ...... 203 Table E-3: Correlation between area and water consumption for light, medium and heavy industry 204 Table E-4: Industrial class range ...... 206 Table E-5: Accuracy comparison between hierarchical clustering and overall overage coefficient for light industry ...... 210 Table E-6: Accuracy comparison between hierarchical clustering and overall overage coefficient for medium industry ...... 210 Table E-7: Accuracy comparison between hierarchical clustering and overall overage coefficient for heavy industry ...... 211

VII

Table F-1: Estimated baseline and 15th year projected water saving from non-potable reuse by individual IP ...... 212 Table F-2: Estimated baseline saving from potable reuse from individual IP ...... 213 Table F-3: Estimated 15th year projected saving from potable water reuse from individual IP ...... 214 Table F-4: Estimated baseline saving at WSS from non-potable water reuse ...... 216 Table F-5: Estimated 15th year projected saving at WSS from non-potable water reuse ...... 216 Table F-6: Estimated baseline saving at WSS from potable water reuse (m3/month)...... 217 Table F-7: Estimated baseline saving at WSS from potable water reuse (m3/month)...... 217 Table F-8: Estimated 15th year projected saving at WSS from potable water reuse (m3/month) ...... 218 Table F-9: Estimated 15th year projected saving at WSS from non-potable water reuse (%) ...... 219 Table G-1: Profile of interviewed companies and summary of initial findings ...... 220 Table H-1: Removal efficiency for unit processes ...... 224 Table H-2: Literature sources for removal efficiency ...... 226 Table H-3: Cost function for CAPEX and OPEX ...... 230 Table H-4: References for CAPEX and OPEX...... 232 Table H-5: Look-up-table (LUT) for Ozone ...... 232 Table H-6: Look-up-table (LUT) for UV ...... 232 Table H-7: ENR and BLS value for base year conversion ...... 232 Table H-8: Inventory for coagulation/flocculation/sedimentation ...... 233 Table H-9: Inventory for GAC and chlorination ...... 234 Table H-10: Inventory for NF/RO (construction phase) ...... 234 Table H-11: Inventory for MBR (construction phase) ...... 235 Table H-12: Inventory for DAF (construction phase)...... 235 Table H-13: Inventory for SF, MF, UF, and UV (construction phase) ...... 236 Table H-14: Inventory for NP-1 and NP-2 treatment train (use phase) ...... 239 Table H-15: Inventory for P-1 and P-2 treatment train (use phase) ...... 240

VIII

List of Figures Figure 1-1: Breakdown of GDP by State in 2014 (Department of Statistics, 2015) ...... 1 Figure 1-2: Water minimization for industrial sector ...... 6 Figure 1-3: Overview of research concept ...... 10 Figure 2-1: Illustration of decreasing head room for Selangor (National Water Service Commission (SPAN), 2016b) ...... 14 Figure 2-2: Reservoir water level profile for Sungai Selangor Dam from 2010 to July 2016 (SPAN) 15 Figure 2-3: Diagram for the variables included in the calculation of IDR and WSR for a given reservoir ...... 23 Figure 2-4: Example of 푄푒푠푡 calculation ...... 25 Figure 2-5: Programming flowchart for inflow calculation estimation, subsequent IDR/WSR calculation, and Monte Carlo simulation ...... 29 Figure 2-6: Illustration of IDR/WSR scatterplot indicating a singular event's vulnerability ...... 30 Figure 2-7: Calculation of vulnerability of each singular event relative to the arithmetic mean of the 1998 cluster ...... 30 Figure 2-8: Sample of data smoothing effect for Semenyih Dam (1/1/2010 - 1/1/2016) ...... 31 Figure 2-9: Monthly inflow, monthly water demand, and dam capacity for Semenyih dam ...... 34 Figure 2-10: Monthly inflow, monthly water demand, and dam capacity for Klang Gates dam ...... 34 Figure 2-11: Monthly inflow, monthly water demand, and dam capacity for Langat dam ...... 34 Figure 2-12: IDR/WSR scatterplot for Semenyih dam ...... 36 Figure 2-13: IDR/WSR scatterplot for Klang Gates dam ...... 36 Figure 2-14: IDR/WSR scatterplot for Langat dam ...... 37 Figure 2-15: Vulnerability curve for Semenyih dam ...... 38 Figure 2-16: Vulnerability curve for Klang Gates dam ...... 38 Figure 2-17: Vulnerability curve for Langat dam ...... 39 Figure 2-18: IDR and WSR plot for Semenyih dam ...... 41 Figure 2-19: IDR and WSR plot for Klang Gates dam ...... 43 Figure 2-20: IDR and WSR plot for Langat dam ...... 44 Figure 3-1: Diagram of a complete WSS ...... 57 Figure 3-2: Overall methodology for Chapter 3 ...... 58 Figure 3-3: Estimation of total potable water consumption for industrial premises ...... 59 Figure 3-4: Mandatory vegetation buffer zone for industrial park...... 62 Figure 3-5: Example of industrial premises clusters for a formal IP configuration (left, CH_IP-2) and freeform IP configuration (right, CH_IP7) ...... 66 Figure 3-6: Comparison of IDR/WSR values for Langat dam at base year (2016) ...... 75 Figure 3-7: Comparison of IDR/WSR values for Semenyih dam at base year (2016) ...... 75 Figure 4-1: System boundary for LCA ...... 100 Figure 4-2: Treatment train diagram for NP-1 ...... 103 Figure 4-3: Treatment train diagram for NP-2 ...... 103 Figure 4-4: Effect of decrease in overall removal efficiency on compliance (NP-1) ...... 105 Figure 4-5: Effect of decrease in overall removal efficiency on compliance (NP-2) ...... 106 Figure 4-6: Treatment train diagram for P-1 ...... 108 Figure 4-7: Treatment train diagram for P-2 ...... 108 Figure 4-8: Effect of decrease in overall removal efficiency on compliance (P-1) ...... 111 Figure 4-9: Abiotic depletion result for the Construction Phase for non-potable reuse...... 117 Figure 4-10: GWP result for (Operation Phase) for non-potable reuse ...... 119 Figure 4-11: Human toxicity result (Operation Phase) for non-potable reuse ...... 120 Figure 4-12: Terrestrial ecotoxicity result (Operation Phase) for non-potable reuse ...... 121

Figure 4-13: Abiotic depletion result (Operation Phase) for potable reuse ...... 125 Figure 4-14: Comparison for GWP (Operation Phase) for potable reuse ...... 126 Figure 4-15: Comparison for human toxicity (Operation Phase) for potable reuse ...... 127 Figure 4-16: Comparison of blending ratio of effluent on human toxicity impact category for potable reuse ...... 128 Figure 4-17: Comparison for terrestrial ecotoxicity (Operation Phase) for potable reuse ...... 129 Figure 5-1: Extract on policy for water pollution in the National Policy on the Environment (Adham, Merle and Weihs, 2013) ...... 148 Figure 5-2: Illustration of the synergy between the water-related stakeholders, industry related stakeholders, and the industrial parks for DWWTP implementation...... 153 Figure B-1: Double-mass curve between station RF 3118102 and RF 3118103 ...... 188 Figure B-2: Double mass curve for SF 2917401 and SF3118445 ...... 190 Figure B-3: Streamflow rating curve for SF 3118445 ...... 191 Figure B-4: Observed water level and streamflow for SF 3118445 (1986-2016) ...... 193 Figure B-5: Result of calibration for SF 3118445, period: 2014-2016 ...... 194 Figure E-1: Water intensity/area scatterplot - light industry ...... 205 Figure E-2: Water intensity/area scatterplot - medium industry ...... 205 Figure E-3: Water intensity/area scatterplot - heavy industry ...... 206

List of Acronym

AC Activated carbon AIS Syarikat Air Selangor C/F/S Coagulation/Flocculation/Sedimentation CAPEX Capital investment COD Chemical oxygen demand CP Consumer point DAF Dissolved air flotation DoE Department of Environment, Malaysia DSM Demand-side management DWWTP Decentralised wastewater treatment plant EDC Endocrine disrupting chemicals/compounds EIP Eco-industrial park GAC Granular activated carbon GHG Greenhouse gas GIS Geographical Information System GWP Global warming potential ICI Industrial, commercial, and institutional sector IDR Inflow-demand reliability index IP Industrial park JPBD Selangor Town and Country Planning Department LCA Life cycle assessment MBR Membrane bioreactor MF Microfiltration MLD Million litre per day MWA Malaysian Water Association NAM Nedbør-Afstrømnings-Model NP Non-potable NS condition Non-satisfactory condition

O3 Ozone OPEX / O&M Operation and maintenance cost P Potable PST Primary sedimentation tank RO Reverse osmosis SA Service area SCP Sustainable consumption and consumption SDG Sustainable development goal SF Sand filtration SPAN National Water Service Commission SSM Supply-side management UF Ultrafiltration UV Ultraviolet WHO World Health Organization WSR Water supply resilience index WSS Water supply system WTP Water treatment plant WWTP Wastewater treatment plant

XII

List of Notation

Blnr,o,p Potable water blending ratio r for industrial park o for reservoir n in month p; m3/month

C1 Material requirement for smaller scale plant; g/unit 3 c1 Capacity for smaller scale plant; m /month

C2 Material requirement for smaller larger plant; g/unit 3 c2 Capacity for larger scale plant; m /month

Cbase Cost of unit process in base year; USD

Ccurrent Cost of unit process in current year (2016); USD

Cfinal Final contaminant concentration; mg/l

Cinitital Initial contaminant concentration; mg/l d(j) Duration of the jth failure event; minute or hour

3 Dp-1,n Total water demand at month p-1 for reservoir n; m /month i Time duration, ranging from month 1 to m; month

Ixbase ENR (CAPEX) or BLS (OPEX) index in base year; unitless

Ixcurrent ENR (CAPEX) or BLS (OPEX) index in current year (2016); unitless

Lx Water losses from reservoir for time period x (i.e. seepage and evaporation); m3/day M Number of failure events; unitless m Selected time frame (6 months for within year dam); number/count mblended Contaminant concentration in wastewater, post blending; mg/l meffluent Contaminant concentration in wastewater effluent; mg/l mpotable Contaminant concentration in potable water; mg/l

Ny Population per category in the water system service area in time period y; unitless 3 Omin Reservoir minimum operational storage; m 3 Ox Reservoir release during time period x; m /day P Measure of factor prices of all input; USD

3 pdis Volume of water effluent generated prior to discharge; m or Gal pe Pricing of energy; USD 3 pin Volume of water obtained from external sources; m or Gal pl Pricing of labour; USD

XIII

pm Pricing of material; USD 3 prer Volume of water recirculated; m or Gal 3 ptrt Volume of water treated prior to use; m or Gal pw Pricing of water; USD px Pricing of capital; USD Q Quantity of water as derived demand; m3/time

3 Q out Monthly water demand at month t; m 3 Qest Total estimated water demand during projected time frame m (6 months); m 3 Qin Monthly inflow into reservoir at month t; m 3 Qx Reservoir inflow during time period x; m /day

Qy Total system water use in time period y; unitless qy Unit-use coefficient of a customer category in the water system service area in time period y; unitless

Rup Removal efficiency for unit process up; expressed as fraction S Cost shares; USD

3 Stt Reservoir storage at month t, where t = 13,…,m; m

Svgsnp,o,n,p Water savings from non-potable np reuse in industrial park o in reservoir n in month p; m3/month

Svgspt,o,n,p Water savings from potable p reuse in industrial park o for reservoir n in month p; m3/month T Total number of time intervals; minute or hour t Time step, ranging from 13 months for within year dam, to m; number/count v(j) mean value of deficit event j; unitless

3 vblended Total wastewater volume, post blending; m 3 veffluent Wastewater effluent volume; m 3 vpotable Potable water volume; m 3 Vx Initial storage volume for time period x; m /day 3 Vx+1 Final storage volume for time period x+1; m /day

WSnp,n,p Total water savings from non-potable np reuse for reservoir n in month p; m3/month 3 WSpt,n,p Total water savings from potable pt reuse for reservoir n in month p; m /month x Initial time for reservoir inflow estimation; day

XIV

Y Output; number Z All other independent variable

αt Inflow-demand reliability (IDR); unitless

βt Water supply resilience (WSR); unitless Δx Time step of x to x+1; day

Acknowledgements

Foremost, I would like to thank Allah for granting me this priceless opportunity to complete my studies and pursue my dream. I would like to express my deepest gratitude to both of my supervisors - Jonathan Chenoweth and Richard Murphy - for guiding me through this journey, especially during those times where I lost sight of the way. Without your feedback and mentorship, I would not have been able to complete this research.

I would like to express my thanks to the Commonwealth Scholarship Committee for the financial support, and SIRIM Berhad for their assistance in making this research possible. I would also like to thank all the agencies for allowing me to use their data for this research: Syarikat Bekalan Air Selangor (SYABAS), Syarikat Air Selangor (AIS), Jabatan Perancangan Bandar dan Desa (JPBD), National Hydraulic Research Institute of Malaysia (NAHRIM), Jabatan Pengairan dan Saliran (DID) and National Water Service Commission Malaysia (SPAN).

I am forever grateful to Sunitha, Melissa, and Elva for their support, especially in times where I didn’t think I would be able to get through this journey. Words cannot express how blessed I am to have met all of you, and I sincerely thank you from the bottom of my heart for being there for me. I would also like to thank the rest of my friends who have helped me during my studies, especially Anne, Ana, Shiqin, Anastasia, Paschalena, and Abber.

My deepest thanks to my mum and sisters for their constant motivation, and for manning the fort back home so I am able to peruse my dreams. Though there are times when I drive you up the wall with my constant chatter regarding my research, all of you kept listening anyway – you have no idea how much that helped. This thesis is dedicated to you, and Abah – I hope I made you proud.

XVI

Chapter 1 : Introduction

1.1. Malaysia’s socioeconomic transition

During the tabling of the Sixth Malaysia Plan in 1991, the then Prime Minister of Malaysia, Tun Dr Mahathir Mohamad, announced the creation of Vision 2020, where it was targeted that Malaysia will be classified as a developed country by 2020. In his speech, he emphasised the need for a competitive, self-sustaining economy that is resilient, robust, and dynamic through a “diversified and balanced economy, with a mature and widely based industrial sector, a modern and mature agriculture sector, and an efficient and productive and an equally mature service sector” (Mohamad, 1991).

Over a series of transformations to diversify the economy, Malaysia’s GDP has migrated from dependence on the agriculture sector to the current economic trend of dependence on the manufacturing and service sectors, which, in 2014, jointly contributed 76% of total GDP as compared to the agriculture sector of approximately 9% (Department of Statistics Malaysia, 2015). It is expected that the manufacturing and service sectors will continue to be the key drivers of economic growth in the future (Bank Negara Malaysia Malaysia, 2015), although the manufacturing sector’s share of the GDP is expected to decline (Economic Planning Unit (EPU) Malaysia, 2015). The high contribution from the manufacturing and service sectors can be observed in the 2014 GDP breakdown for all states illustrated in Figure 1-1 below.

Figure 1-1: Breakdown of GDP by State in 2014 (Department of Statistics, 2015)

The majority of the players in the manufacturing and service sectors are concentrated in four main conurbations – Selangor (including Kuala Lumpur and Putrajaya), Penang, Johor, and Pahang – as they contain the necessary talent and knowledgeable workers, and supporting physical infrastructure and resources (economic, human, and natural) for business growth, which is essential for drawing in investments for the country (Economic Planning Unit (EPU) Malaysia, 2010; Satterthwaite, McGranahan and Tacoli, 2010). Furthermore, these areas tend to contain a concentration of the nation’s economic, government, and commerce activities, as well as providing links between cities, states, regions, and international borders (Department of Economic and Social Affairs, Population Division, 2014).

In tandem with this, the shifting of the economy has also seen changes in the distribution of population, concentrating more within these conurbations. Consistent with the global trend, the increase in the number of people living in urban areas has been driven by the increase in the opportunity for investment and employment, where the majority of the service and manufacturing activities are located (Satterthwaite, 2007). As such, it is projected that 75% of 32 million of Malaysia’s expected population will reside within these conurbations by 2020.

However, the growth in both urbanization rate and the increase in contribution of the manufacturing and service sectors to the GDP have led to several problems – one of which is the increasing pressure on previously abundant natural resources (in this instance, water specifically) to support rapid domestic and industrial growth. This had led to water imbalance issues in the country in general, and the four main conurbation areas specifically. For example, the Malaysia Economic Report had reported an increase in water demand by 1.2% in 2015, primarily due to improved access to water supply, and an increased number of new residential, industrial, and commercial areas (Ministry of Finance Malaysia, 2015). Another example is a projected increase of 44% of water use in 2025 as compared to a 2005 baseline in the South Johor Economic Region, requiring approximately RM 770 million in investments to upgrade existing water supply infrastructures (Khazanah Nasional Berhad, 2006). In order to ensure continued growth of the country, it is imperative that these water imbalance issues be mitigated.

1.2. Top-down approach for water supply imbalance: SSM vs DSM

Literature addressing strategies for reducing the risk of water imbalance issues similar in Malaysia has been focusing on increasing water supply reliability, either by increasing water

supply availability (supply-side management, SSM) or reducing water demand (demand-side management, DSM) (Mens, Gilroy and Williams, 2015). Conventional efforts had favoured SSM by building large infrastructures to capture, store, and redirect available raw water for treatment (Dawadi and Ahmad, 2013). However, as freshwater became less accessible or available, additional investments were needed to extract less accessible sources of freshwater, where the financial resources were primarily attained through public investments (Barbier, 2015). In addition, infrastructures built to extract these freshwater resources have resulted in irreversible social, ecological, and environmental costs (Vedwan et al., 2008; Wang et al., 2011), which has overshadowed any economic-wide productivity gains (Barbier, 2015). Wang et al. (2011) had exemplified these points through a case study of Yulin city. On the one hand, the Chinese government has emphasised the importance of securing adequate water supply for the country through consideration of various water transfer projects. On the other hand, it was noted that these transfer projects had required higher investments, leading to higher marginal cost of treated water, as well as several negative environmental impacts, resulting in difficulties to achieve sustainable development for the country.

Due to this, the implementation of DSM has been advocated over SSM. In contrast to the latter, DSM emphasises the reduction of water demand by incorporating elements of technical, economic, administrative, financial and social measures to regulate water use and disposal (UNEP, 2011). Additionally, implementation of DSM has been highlighted as a more attractive approach to governments, as it delays the need for large capital investments for the expansion of a water infrastructure. This is then able to be used to provide more financial resources to cover the cost of the implementation of a comprehensive DSM programme (Vairavamoorthy, Gorantiwar and Pathirana, 2008).

Although DSM applications for water management have been observed to be gaining momentum, its implementation is not without challenges. The first challenge is to quantify the impact of potential DSM strategies onto the current water supply system (WSS), and incorporating those results into an operational concept, fit for policy and management purposes (Klein, Nicholls and Thomalla, 2003). This had been emphasised through the views of planners and policy makers, who require a balance between practical applications of research to meet political needs (Beveridge and Monsees, 2012), as well as a more straightforward estimation of uncertainties and risk (Mastrandrea et al., 2010) which can be seamlessly integrated into existing procedure (Moss and Schneider, 2006). From a WSS’s perspective, this primarily points to a need to determine a method of quantification which

incorporates the root cause of system failure to aid in the development of effective laws and regulation. Such method would also need to assess and illustrates the impact of any changes into a system to allow for tangible means of quantifying the effectiveness of the DSM strategy to reduce system vulnerability. Method such as the use of sustainability index (e.g. Mayer, 2008; Sandoval-Solis et al., 2011; Safavi, Golmohammadi and Sandoval-Solis, 2016), mathematical modelling (e.g. Dawadi and Ahmad, 2013; Asefa et al., 2014; Soundharajan, Adeloye and Remesan, 2016), and interaction-based modelling such as agent-based modelling (e.g. Berglund, 2015; Xiao, Fang and Hipel, 2018) have been explored to address this issue.

The second challenge is to develop policies to support the implementation of the identified strategies (Gourbesville, 2008; Blackman, 2009). Although the aforementioned quantification method had managed to give insights to the savings required to decrease the likelihood of water imbalance, they had rarely consider specific policies or adaptation strategies to realise these savings (Olmstead, 2014; Jayarathna et al., 2017); these strategies are typically address in separate but parallel bodies of literature and are rarely integrated. These strategies, as underlined by Kampragou, Lekkas and Assimacopoulos (2011), are based on several core principles, namely water conservation enforcement through legislation, reinforcement using economic instruments, infrastructure or technology investment, and awareness. In contrast to domestic sector, top-down DSM strategy targeted at industrial sector specifically had mainly focused on the use market-based policies, defined as the use of market-based signal to encourage desired outcome of decision-making, either through financial reward for positive behaviour or compensation for undesirable behaviour (Cantin, Shrubsole and Aït-Ouyahia, 2005). This approach – which utilises measures such as tradeable permits and tax incentives or rebates – has been advocated as a more encouraging method for a given business to employ the most efficient means possible in achieving the final goal (Gunatilake and De Guzman, 2008), as it enables them to adhere to compliance using means that are within the limits of their available resources.

Although that is the case, arguments have been made that the adaptation of market-based policies might not be an effective instrument to tackle this issue for developing countries. In a comprehensive review by Bell and Russell (2002), it has been underlined that one of the key factor is the underestimation of the overarching cost of compliance or implementation of the identified policies. Here, they had emphasized the importance of a realistic policy which considers the regulatory body’s capabilities to implement and enforce, as well as the financial

and human resources required to reach compliance. These findings echoed that of Stavins (2001), in which he had underlined the importance of flexibility, simplicity, and role of monitoring and enforcement, in addition to consideration of the capabilities of the private sector needed to enable market-based policies to work. Consideration should incorporate the other principles noted above, for example, the establishment of mechanism to facilitate and monitor water use and trade (enforcement), as well as expansion of current water infrastructure (investment) to support the policy (Cantin, Shrubsole and Aït-Ouyahia, 2005). Tortajada and Joshi (2014) had exemplified this; in their review of the role of institutions, laws, and regulation of water quality management in Singapore, they had emphasised the importance of a flexible laws and regulation, coupled with stringent enforcement of its law through constant and consistent monitoring, in addition to investment of infrastructure development to compliment the policies.

1.3. Capturing the full extent of market-based policies

Critical to the success of a market-based policy is the estimation of the price elasticity of water to the user to gauge the response of the users to price. Price elasticity is defined as percent change in (water) demand divided by the percent change in price (Varian, 2010). Literature examining the price elasticity of the domestic sector have noted it to be relatively inelastic (e.g. Inman and Jeffrey, 2006; House-Peters and Chang, 2011), while industrial water elasticity was observed to be higher than the former, though it varies with industrial type (e.g. Olmstead, 2014; Renzetti, 2015). Thus, it can be hypothesised that market-based strategies targeted at the industrial sector might have a higher chances of achieving the overall goal of water use reduction. On the other hand, arguments have also been made regarding the risk of its implementation – especially with regards to increasing water tariffs – as there is a potential risk of industrial relocation (Olmstead and Stavins, 2009); this is particularly true for countries with large number of firms relying on international market as it decreases product competitiveness (Cantin, Shrubsole and Aït-Ouyahia, 2005).

In lieu with this, other complementary bottom-up strategies implemented to reach a common goal of mitigating water related issues might increase the effectiveness of this policy. Castonguay et al. (2018) had illustrated this concept using rainwater harvesting system for residential dwellings; they had noted the importance of multiple policy instrument used

simultaneously to trigger the uptake of the system to achieve the aim of water savings in Australia.

For industrial or manufacturing sectors specifically, bottom-up strategies have been observed to focus on encouraging the industry players to increase water efficiency (e.g. internal recirculation or recycling, and process retrofitting) (e.g. Chungsiriporn, Prasertsan and Bunyakan, 2006; Barrington, Prior and Ho, 2013), or use of alternative water sources (e.g. groundwater or treated wastewater) (e.g. Tan, Manan and Foo, 2007; Agana, Reeve and Orbell, 2013). Literature regarding this approach has mainly focused on the implementation at firm/plant level (intra-plant), or of a more concentrated effort implemented within an industrial park (inter-plant). Extensive research have been done for both, where literature on achieving this target can be grouped into three categories; quantification, simulation, and implementation, as summarised below in Figure 1-2.

Figure 1-2: Water minimization for industrial sector Although there is a growing body of literature which focuses on the three categories in Figure 1-2 for the industrial sector, it has been emphasised that the realisation of these potential water savings remains a challenge. Such challenges highlighted were the inability to successfully apply the technique at operational level due to insufficient expertise (Dunn and Wenzel, 2001), and higher cost as compared to the added benefits, where the technology transfer needs to be financially justifiable (Bates et al., 2008). Subsequently, it has also been noted that companies might not have access to the necessary funds, or are not willing to invest in the identified solution due to high cost of technology implementation and low water tariffs (Wenzel et al., 2002), despite an attractive payback period of the potential solution

(Dunn and El-Halwagi, 2003). As such, these barriers would need to be taken into consideration during identification of the complementary bottom-up strategies alongside the implementation of estimating the full extent of the market-based policy.

Furthermore, although cost and savings from implementation of water savings strategies have been explored and documented, its impact at a broader scale have received little attention (Gurung et al., 2016) – even less so for the industrial sector. For the latter, these strategies have been rigorously studied at a firm level as exemplified above, while the scaling up to a broader context remain unexplored, primarily due to the lack to high resolution data (Reynaud, 2003).

1.4. Research rationale and focus

Based on the review of the current status of the WSS in Malaysia, it can be seen that Malaysia faces increased risk of experiencing socioeconomic drought1 in the future, due to population and industrial growth, which are expected to increase demand for potable water in the future. The combination of these issues place additional pressure on the current WSS.

From the top-down perspectives, the implementation of DSM-based strategies have been noted to be more favorable as compared to SSM strategies, where the use of market-based policies have been highlighted as one of the means to execute this strategy. In this sense, the industrial and manufacturing sectors, which have been noted to be more responsive to price as compared to the domestic sector, might be more susceptible to this strategy. It has been underlined that the use of a complementary bottom-up strategy to enable the industry players to adhere to the market-based strategies might aid in increasing the effectiveness of its implementation. In turn, the cost of the complementary strategy would need to be incorporated into the estimation of the overall cost of compliance in order to justify its implementation. Furthermore, the effectiveness of its implementation to achieve the common goal of reducing water imbalance would also need to be considered.

1 the inability of water supply system (WSS) to meet the demand of an economic good, such as drinking water, irrigation, or hydroelectric power (Wilhite and Glantz, 1985; Heim, 2002; Mishra and Singh, 2010; Van Loon, 2015). Assessment within this context has often been associated with human activities affected by drought, including losses and benefits with respect to the local and regional economy (Changnon, 1987).

In order to first estimate the total cost of compliance, the challenge is to examine potential bottom-up strategies and link its impact to a WSS to determine its technical, financial, and environmental feasibility. Literature linking potential bottom-up approach which examines all three impacts have been rigorously studied either at in individual (e.g. firm) or cluster (e.g. industrial park) level, and have shown encouraging result in terms of potential water savings in general. However, the scaling up of these estimations to a border context of a total WSS remains unexplored though it is an important consideration as execution of these bottom-up strategies at a small scale might not be justifiable as it may not result in a positive outcome to elevate the current water imbalance issue in a WSS. Additionally, the main challenge faced by the industry players is still realizing these estimated savings, due to, in part, the cost, and another, lack of expertise (e.g. Dunn and Wenzel, 2001; Wenzel et al., 2002; Bates et al., 2008); these limitations highlighted by the industrial players would need to be taken as a limiting factor when identifying an appropriate complementary strategy.

1.5. Aim and objectives of research

In order to address the highlighted challenges, this research attempts to assess the feasibility of industrial wastewater reuse as a demand-side management strategy to alleviate a water supply imbalance issue, using Selangor, Malaysia as a case study. In order to address this overarching aim, two objectives were considered, namely:

OB 1: To assess the technical, economical, and environmental feasibility of industrial wastewater reuse for the manufacturing sector

OB 2: To analyze the overarching financial resources, and administrative and social measures required for its implementation

The following objectives were addressed through the execution of three phases, designed to evaluate the impact of small scale implementation (industrial wastewater reuse) against a large scale system (water supply system). The three phases considered were:

Phase 1: To establish a benchmark of the current water supply systems against conditions of water imbalances by taking into account the supply, demand, and storage capability of the system

Phase 2: To assess the impact of potential water savings generated from a complementary strategy targeted at the manufacturing sector onto the overall WSS to alleviate water imbalance issues

Phase 3: To evaluate the technical, financial, and environmental feasibility of deploying complementary strategy identified by taking into consideration the industry player’s narratives

For this research specifically, a water supply system (WSS) is defined as a structure which supplies potable water to users in a service area supplied via an upstream components of a dam and water treatment plant. Three separate WSS – namely Klang Gates, Langat, and Semenyih WSS – were used as case studies. All three systems are located within Selangor, Malaysia, where the state was used as a case study because:

i. There are a high volume of potable water consumption by the industrial sector, as reported by The Malaysian Water Association (MWA) (2015) ii. It has been known to face water supply interruptions in the past (as illustrated in Appendix A) iii. Selangor has been identified as one of the main conurbation areas of the country and is expected to see a further increase in population and industrial growth in the future

1.6. Overview of thesis

Each of the phases underlined in subsection 1.5 above were further explored and discussed in greater detail in Chapter 2, 3, and 4 for phase 1, 2, and 3 respectively, while Chapter 5 discussed the feasibility of the implementation of industrial wastewater reuse as a DSM- based strategy from the technical, financial, and environmental perspective, as compared to other SSM-based strategies – an overview of the research concept is illustrated in Figure 1-3 below. As each phases covers specific theme to address the overall aim, all three chapters were designed as a stand-alone structure consisting of specific research questions, where relevant literature according to the theme is critically reviewed, leading to the development of the methodology, and followed by the results and discussion of the resultant findings.

Figure 1-3: Overview of research concept

Chapter 2 focuses on benchmarking the historical and current vulnerability of the three dams which corresponds to each of the three WSS mentioned above. In the context of this research specifically, a function-oriented definition of vulnerability was adopted, emphasising on the ability of a WSS to deliver a consistent and reliable water service (Carter, Tyrrel and Howsam, 1999; Montgomery, Bartram and Elimelech, 2009). The vulnerability of each dams were assessed in terms of the inflow’s reliability to cater to demand, as well as the reservoir’s resilience to provide adequate buffer in events of low flow into the system. Furthermore, the reliability and resilience were associated with the need for a supply-side management (SSM) based strategy or a demand-side management (DSM) based strategy.

Potential water savings by the manufacturing sector within the service area of each WSS were then estimated in Chapter 3. Total potential potable water savings were then estimated for the area served by each reservoir and subsequent potable water treatment plant. The impact of the potential water savings within an overall WSS was then evaluated. Finally, the impact of the saving onto the dam and potable water treatment plant were assessed.

The final phase assessing the technical, financial, and environmental feasibility of the identified strategy is presented in Chapter 4. The financial cost was determined in terms of both capital and operational costs, while the environmental assessment was assessed using a Life Cycle Assessment (LCA) approach based on several environmental impact categories. The results are discussed in terms of a holistic view for the overall WSS.

The findings obtained throughout the thesis are compiled and discussed in the context of the second objective as well as the overall aim of this research in Chapter 5. Finally, Chapter 6 concludes the thesis, summarizing the main finding, the contribution of the research, and potential avenues for future research.

Chapter 2 : Benchmarking of current water supply system’s vulnerability

2.1. Introduction

As introduce in the previous chapter, any strategies considered to a decrease water supply imbalance need to be assessed against the WSS itself in order to capture the affect they will have to mitigate the problem. As such, this chapter focuses on benchmarking the three WSSs considered in this study by taking into consideration factors which cause the water imbalance incidents, and further assesses the potential impact of SSM and DSM strategies on the system.

With the exception of Kelantan, the sources of treated water for all states in Malaysia are heavily dependent on river flows, accounting for 82% of total supply as compared to water from storage dams (17%) and groundwater (1%) (SPAN, 2015). The direct use of river flows is due to the accessibility of the resource; there are over 150 river systems in Malaysia (Abdullah and Mohamed, 1998). Nevertheless, dams are critical to the water treatment system as they regulate river flows into water treatment plants, amongst other functions (Subramaniam, 2004).

As previously highlighted, the increasing demand for treated water resulting from industrial expansion and subsequent population growth in urban areas has placed substantial pressure on water operators to ensure reliable and uninterrupted treated water supply to consumers. However, they are faced with two major challenges: an increase in demand for potable water, and a decrease in raw water supply volume.

i. Increase in demand

Water operators monitor the number of consumers and demand for potable water via the number of connections to the water distribution network. Table 2-1 summarises the growth in the number of connections, and volume of potable water consumption from 2013 to 2014 for the domestic and non-domestic sectors in Malaysia. From this, it can be seen that East Malaysia has a higher growth rate for both the number of connections and volume (domestic) as compared to Peninsular Malaysia, indicating rapid development in the area.

Table 2-1: Percentage of growth in the number of connections, and volume of treated water for 2013 to 2014 (Malaysian Water Association (MWA) 2015)

Connections (% growth) Volume (% growth) Domestic Non-Domestic Domestic Non-domestic - 7.0 - 4.0 Minimum 1.6 (Perak) 2.2 (Kedah) (Terengganu) (Terengganu) Maximum (East 6.1 (Labuan) 19.3 (Sabah) 32.0 (Sabah) 12.0 (Sabah) Malaysia) Maximum (Peninsular 4.24 12.7 5.6 (Kelantan) 25.0 (Johor) Malaysia) (Kelantan) (Kelantan) Average 3.1 5.4 8.3 3.6

The increase in total demand – especially in Kuala Lumpur and Selangor – have been noted to affect the raw water reserve margin and simultaneously decreasing the available head room of available treated water, as illustrated in Figure 2-1 for the whole of Selangor. The National Water Service Commission, SPAN (2016) has reported that the total raw water reserve margin for the two states is one of the lowest in the country. Currently, it was reported to be at 8.68%, as compared to the previous head room of 2.16%, where the 6.52% increase in reserve head room was due to the completion of the Pahang-Selangor water transfer project in 2014, which cost approximately RM 2.48 billion (USD 0.6 billion). Furthermore, a similar condition was reported for the potable water margin; it was noted that several of the 34 water treatment plants serving the consumers are overloaded, producing approximately 20% above their design capacity and have a potable water head room of 0.3% (The Malaysian Water Association (MWA), 2015).

Figure 2-1: Illustration of decreasing head room for Selangor (National Water Service Commission (SPAN), 2016b)

ii. Decrease in raw water availability

From the supply side, the availability of water in Malaysia’s river basins has been affected due to the change in the rainfall pattern. Although the total volume of precipitation has remained relatively constant over the years, there has been a shift in recent years in the seasonality of rainfall compared to the historical trend. For example, it has been observed that there has been a decreasing trend in total rainfall and frequency of wet days during the southwest monsoon, while an increasing trend has been observed instead for these two parameters during the northeast monsoon (Suhaila et al., 2010). Furthermore, Shaaban et al. (2007) in an earlier study reported that the variation between high and low river flows will become more pronounced as compared to the historical record, although the volume of annual rainfall is expected to remain relatively constant within all sub-regions. In addition, Daud (2015) reported that there has been a change in rainfall pattern over catchment basins, where increased rainfall is reported to fall outside of the water treatment plant catchment area.

The above issues noted in the literature are consistent with the challenges highlighted by the water operators, where they observed a shift in current precipitation patterns over the catchment basin in terms of temporal variability and intensity as compared to historical patterns. In addition to this, higher demand for potable water from users has led to a higher

rate of raw water withdrawal from the dams. These two factors ultimately influence the time needed for the dams to refill. This can be observed through the water level profile, as illustrated in Figure 2-2, for the Sungai Selangor Dam.

Figure 2-2: Reservoir water level profile for Sungai Selangor Dam from 2010 to July 2016 (SPAN)

2.2. Aim and research questions for Chapter 2

From the above, it can be concluded that both the increase in demand and decrease in supply have affected the capability of the current WSS to provide adequate water to the users. In tandem with this, it is imperative that the two parameters underlined are considered in order to assess the impact of DSM strategies implementation, in addition to the capabilities of the dam itself to provide adequate storage of the raw water, where it is deemed satisfactory (S) if it is able to meet the demand of the users, and non-satisfactory (NS) otherwise. As such, the overall aim of the first part of this research is to establish a benchmark of the current reservoir, specifically in terms of the reliability of inflow into the reservoir to cater for the demand, the resilience of the reservoir to provide for adequate storage, and the overall vulnerability of the system against water supply inadequacy. Thus, the following research questions were considered:

RQ 1: What is the current trend of each reservoir in terms of trend of inflow and outflow, and its impact on water level?

RQ 2: What is the relationship between inflow, outflow, and storage with conditions of low vulnerability for each reservoir?

Here, this study adopts a performance-based definition of vulnerability, defined as a reservoir’s inability to provide adequate potable water to meet the user’s demand. Simulations to determine the performance of a WSS was assessed based on this definition, where it is considered highly vulnerable when the dam is not able to meet the demand of the users, and low vulnerability otherwise. For the former case, this could either be due to a combination of low inflow (supply) and high outflow (demand), or low raw water storage in the reservoir.

2.3. Literature review 2.3.1. Vulnerability, resilience, and reliability of a WSS, and its linkages to WSS performance

The performance of a WSS has been typically assessed through its ability to meet user’s demand, where it is deemed satisfactory (S) if stored water volume is higher than or equal to demand, and non-satisfactory (NS) otherwise (Kundzewicz and Kindler, 1995; Asefa et al., 2014; Fabre et al., 2015). Its assessment has been done using various criteria; of which, the three most commonly used is reliability, resilience, and vulnerability (Srdjevic, Medeiros and Faria, 2004). Hashimoto et al (1982) was amongst the first to introduce the use of the three criteria to assess different aspects of a WSS. Over the years, it had been applied in various research assessing a WSS’s performance, such as by Vogel & Bolognese (1995) who had formulated a general approach to assess overall behaviour of over-year dam system, and Jain & Bhunya (2008) who had investigated the applicability of the three criteria in understanding the behaviour of a multipurpose dams. Regardless, this concept has primarily been applied to identify and quantify undesirable situations which pose a risk towards a water resource system as a whole (Asefa et al., 2014). Although there have been numerous definitions for reliability, resilience, and vulnerability, definitions introduced by Hashimoto et al. (1982) have been the most comprehensive and most commonly referenced (Srdjevic, Medeiros and Faria, 2004); defined as below (Kjeldsen and Rosbjerg, 2005):

Reliability (Rel): Frequency or probability of a system being in a satisfactory state, and estimated using Eq. 2.1:

Eq. 2.1

Where:

푑(푗) − 푑푢푟푎푡푖표푛 표푓 푡ℎ푒 푗푡ℎ 푓푎푖푙푢푟푒 푒푣푒푛푡, 푚푖푛푢푡푒 표푟 ℎ표푢푟

푀 − 푛푢푚푏푒푟 표푓 푓푎푖푙푢푟푒 푒푣푒푛푡푠, 푢푛푖푡푙푒푠푠

푇 − 푡표푡푎푙 푛푢푚푏푒푟 표푓 푡푖푚푒 푖푛푡푒푟푣푎푙푠, 푚푖푛푢푡푒 표푟 ℎ표푢푟

Resilience (Res): Describes how quickly the system is likely to recover from failure, measure as the inverse of the mean value of the time or duration the system is spent in a NS state (Eq. 2.2).

Eq. 2.2

Where:

푑(푗) − 푑푢푟푎푡푖표푛 표푓 푡ℎ푒 푗푡ℎ 푓푎푖푙푢푟푒 푒푣푒푛푡, 푚푖푛푢푡푒 표푟 ℎ표푢푟

푀 − 푛푢푚푏푒푟 표푓 푓푎푖푙푢푟푒 푒푣푒푛푡푠

Vulnerability (Vul): Likely magnitude of a failure, or how significant the likely consequences of failure might be. Hashimoto, Stedinger and Loucks (1982) had estimated vulnerability as the mean value of deficit events v(j) in the following equation (Eq. 2.3).

Eq. 2.3

Where:

푣(푗) − 푚푒푎푛 푣푎푙푢푒 표푓 푑푒푓푖푐푖푡 푒푣푒푛푡푠, 푢푛푖푡푙푒푠푠

푀 − 푛푢푚푏푒푟 표푓 푓푎푖푙푢푟푒 푒푣푒푛푡푠, 푢푛푖푡푙푒푠푠

The concepts of reliability, resilience, and vulnerability have also been used as indicators to link the performance of a reservoir with climate change impacts. For example, Mohammed and Scholz (2017) used the three indicators to assess the performance sensitivity of Dokan

reservoir with respect to climate change; they observed a decline in the reservoir’s reliability and increase in resilience due to dam inflow decrease. Safavi, Golmohammadi and Sandoval- Solis (2016) combined the three criteria into a single sustainability index to evaluate the impact of SSM, DSM, and a combination of the two management strategies on the sustainability of Zayandehrud basin; it was found that the combination of both scenarios resulted in a continuous reliability and sustainability of the system.

Each of the abovementioned criteria has been used to assess different parts of a water resource system and is complementary to one another. In a study by Kjeldsen and Rosbjerg (2005), it has been highlighted that of all possible combinations of the three criteria, integration of resilience and vulnerability assessment of a system has been preferred as it is able to capture the non-stationary (i.e.: subject to a trend) circumstances of a WSS and its corresponding demand as there exist a strong correlation between both criteria (i.e.: system with low vulnerability will also have a high degree of resilience). This is in line with the definition above and the literature, where it can be generalised that vulnerability assessment is viewed as a damage-oriented approach, where the impact of a system’s shortcomings is translated into issues of concern which can be mitigated though various strategies to reduce its vulnerability, as compared to a problem-oriented approach of assessment of the reliability and resilience of a WSS.

Although the aforementioned three criteria have been widely used to measure the performance of a WSS, they have been typically reported in the literature as a single value. For example, reliability of a reservoir is expressed as percentage of time or volume where the reservoir is able to meet the demand of users over the period of assessment, while resilience is often expressed in terms of duration (e.g.: days or months) where the system is able to recover from the disruption. However, these means of calculating the WSS’s performance are not able to indicate the onset of NS occurrence, nor do they explicitly indicate the root cause of the NS condition (e.g.: either due to decrease in inflow or increase in demand) - rather they account for the number of times the system has fail to meet the demand.

Due to this, a different approach for quantifying system resilience and reliability, and their linkages with vulnerability is considered; specifically that which was proposed by Mehran, Mazdiyasni and Aghakouchak (2015). In their study, the definition and calculation of resilience and reliability is associated with the causes of a system failure, where a WSS’s reliability is determined by the adequacy of inflow into a reservoir relative to the demand as

well as its ability to ensure that the volume of water the reservoir is maintained, while resilience is defined as the ability of a reservoir to provide sufficient stored water to the users in the event of low inflow into a system. As such, a system is considered to be vulnerable in the event where the reservoir’s resilience and reliability are low. In comparison with the method described by Hashimoto, this method is able to distinguish the onset of a NS condition by specifically examining the two components which have been known to lead to system failure (i.e.: raw water supply, and demand). Moreover, because of the method’s ability to distinguish the causes of failure, policies relating to the reduction of its vulnerability can be tailored to the cause. For example, if the system is deemed to have low resilience (i.e.: inability of a reservoir to cope with high demand), a DSM strategy would be preferred, as it allows for a decrease in water demand, and delaying of large capital investments; while a low reliability of a system could point to a need to increase the capacity of the system via SSM strategies (e.g.: water transfer scheme).

2.3.2. Methods of estimating inflow into reservoirs

Due to high cost of maintenance and monitoring of instruments in remote, mountainous areas, the majority of rivers and tributaries in the world are ungauged. As such, estimating inflow for an ungauged catchment remains one of the bigger challenges faced in hydrological modelling. Methods for estimation of these ungauged catchments have been generally classified into two streams of techniques; hydrological-based (or physically-based) modelling, and data-driven modelling. Hydrological-based modelling refers to methods or models that were developed based on theoretical insights and physical processes within a catchment, whereas data-driven models had emphasised the use of mathematical equations and information which can be extracted from the data without focusing on the physical interaction of the hydrological process (Solomatine and Ostfeld, 2008).

The former group of method for estimating inflow can be further be aggregated into two groups of models; conceptual lumped type models, and distributed models. The main difference between these two groups of models is the incorporation of spatial-based information into the estimation of inflow – conceptual lumped models typically assumes a spatially singular entity of a catchment or basin by transforming rainfall excess into an outflow hydrography, while distributed hydrological models integrate spatial distribution of influencing parameters (e.g.: rainfall distribution and intensity) into a computational

algorithm for the model to predict corresponding responses (Xu, 2002). Xu (2002) highlighted that the choice between the two types of model depends on the analysis conducted – while a conceptual lumped model has the capability to simplify complex hydrological model, distributed models have been known to be able to capture impacts of spatial-related factors, such as land-use change, and pollutants and sedimentation dispersions; as such, the type of model chosen would need to take into consideration the overall objective of the research. However, the use of conceptual lumped models has generally been preferred over distributed model for streamflow estimation as the less complex lumped model have been shown to produce in equally reliable result (Yadav, Wagener and Gupta, 2007).

Conceptual lumped models have largely been used in regionalized techniques to estimate model parameters for inflow, where regionalised techniques have been defined as the process of transferring hydrological information from a gauged catchment to an ungauged catchment of similar climate, geology, topography, and vegetation cover (Razavi, Coulibaly and Asce, 2013). Various models have been developed and are available for use (e.g.: Hydrologiska Byrans Vatenbalansavdelning model (HBV), Topographic Model (TOPMODEL), Nedbør- Afstrømnings-Model (NAM)), where the choice of models used has been dependent on factors such as ease of model use (simplicity), data requirements, code availability, experience of use, as well as the extent of dissemination of model in the modelling community (Hrachowitz et al., 2013), amongst others. However, it has been highlighted that the choice of models used has little differences in the performance of the various models (Chiew, 2010), and that simple models should be and have been preferred (Razavi, Coulibaly and Asce, 2013), as complex models risk the chance of over-parameterisation, preventing the complex models from reaching their potential performance level (Perrin, Michel and Âassian, 2001).

In contrast to hydrological-based modelling, data-driven modelling has been driven by advanced modelling procedures which emphasised the use of stochastic modelling (e.g.: autoregressive (AR), autoregressive moving average (ARMA), and autoregressive integrated moving average (ARIMA)) and/or computational intelligence (CI) methods (e.g.: artificial neural network (ANN), fussy systems, evolutionary computing (i.e.: genetic algorithm, GA), and adoptive neural-based fuzzy inference system (ANFIS)) (Solomatine and Ostfeld, 2008). There have been several advantages in the use of CI methods as compared to stochastic modelling; specifically the ability of CI methods to account for the non-linear relationship amongst variables, resulting in a more robust and accurate prediction in a simulation (Raman

and Sunilkumar, 1995). For example, Kisi (2005) compared the use of AR method with ANN to estimate streamflow in Gila River, Arizona, and found results obtained using ANN were more accurate as compared to AR, given the use of a similar dataset. Wang et al. (2009) also concluded that the use of ANFIS and genetic programming (GP) were able to generate a relatively high accuracy for streamflow forecasting at Lancangjiang and Wujiang River in China; however they showed a reduction in prediction accuracy in the event of high variability within the data. Thus, it can be concluded that the use of data-driven modelling has been able to result in superior performance, although it was debated whether the use of the model to increase simulation accuracy aided or provided new insight to further understanding catchment or watershed processes (Govindaraju, 2000).

There is a trade-off for both types of models discussed above; on the one hand, it has been emphasised that hydrological-based models are sensitive to data quality, where errors in data, impaired streamflow and snowmelt observations, and human errors have been known to influence the accuracy of the model (Razavi, Coulibaly and Asce, 2013). However, due to the underlying physical and theoretical insights incorporated into the model structure, hydrological-based modelling requires a relatively smaller sample set of data with a shorter timeframe to calibrate and validate the model (Hrachowitz et al., 2013). On the other hand, the use of data-driven models allowed the use of smoothing techniques to minimise or remove the presence of noise within a dataset, enabling the model to adequately capture inherent relationship between model variables. However, due to data dependence, relatively larger samples have been required to account for data variability, which have been known to influence the accuracy of data-driven models (Londhe and Charhate, 2010), as it is not able to accurately extrapolate input outside the range of those used for model training (Wu, Chau and Li, 2009).

Although the methods and techniques discussed above have their advantages and drawbacks, the underlying accuracy and reliability of resulted simulation has been shown to be heavily dependent on the quality and availability of data used in the simulation. This presented a drawback in estimating discharge in a highly data scarce area or region, in which another method had been used instead; water balance modelling. Inherently, it can be classified as a data-driven approach, whereby the determination of inflow into a system has been determined by the input/output data collected, without considering the physical or hydrological interaction of a catchment. In a study comparing the accuracy of a conceptual lumped model against a simple water balance method, Chiew (2010) had found that the

simulated result for estimating streamflow data using the latter method performed as well as the former model. From the perspective of this research specifically, the use of water balance modelling has been successfully applied to predict the inflow into a reservoir (e.g.: Vining and Vecchia, 2007; Werner, Prowse and Bonsal, 2015) such as by Güntner et al. (2004), who used this concept to simulate the dynamic of a multi-system reservoir in Brazil.

2.4. Methodology

In Chapter 1, it was noted that three dams in Selangor were used as case study; Semenyih dam, Klang Gates dam, and Langat dam. Of the three dams, Klang Gates is a non-regulating reservoir (direct abstraction), while Langat and Semenyih are regulating reservoirs. The differences between the two reservoirs are the methods in which stored water are directed into a WTP. For regulating reservoirs, feeding of the stored water to a WTP was done via a river network; water from the dam is released to ensure that the water level or flow of the downstream river is adequate for potable water abstraction. In contrast, a direct supply reservoir delivers the raw water directly to a WTP via pipes. Details regarding each dam are as below (Table 2-2).

Table 2-2: Summary of Semenyih, Klang Gates, and Langat dam

Semenyih Klang Gates Langat Longitude 101.88 101.75 101.89 Latitude 3.08 3.24 3.21 Start year of operation 1985 1959 1979 Type Earth fill Reinforced Earth fill concrete dam Maximum release (m3/s) 600 376.6 510 Minimum release (m3/s) 1.37 0.32 1.01 Maximum level (m) 113.9 96.65 220.96 Minimum operational level (Critical 95.8 84 204.21 level) (m) Environmental Flow (MLD) 118 28 87 Flood control level (m) 112.5 94.488 220.96 Catchment area (km2) 56.7 77.16 41.45

Gross storage (MCM) 62.6 32 37.48 Active storage (MCM) 60.4 22.6 36.77 Water surface area at full supply level 3.6 2.7 2.2 (km2)

As discussed in section 2.3.1 above, the methodology for this study adapted those developed by Mehran et al. (2015) to establish a baseline of the current reservoirs in Selangor. It was chosen specifically for its emphasis of the three components which have a direct impact on a reservoir system – namely the system inflow, outflow, and storage – by estimation of each system’s reliability through the calculation of inflow-demand reliability index (IDR), and its resilience via calculation of its water supply resilience index (WSR). The former takes into consideration the availability of inflow into and corresponding outflow out of a system, to assess for its sufficiency to satisfy water demand regardless of storage within a reservoir. In contrast, WSR assesses its resilience by taking into account the monthly inflow, monthly water demand, monthly storage, and total water demand within the projected time frame (6 months) to determine if the capacity of the stored water volume is able to satisfy the demand in the event of low flows. A simplified diagram illustrating the variables included in the calculation of IDR and WSR (i.e. inflow, outflow, and reservoir storage) is appended in Figure 2-3 below.

Figure 2-3: Diagram for the variables included in the calculation of IDR and WSR for a given reservoir

Eq. 2.4

Where:

훼푡 − 퐼푛푓푙표푤 − 푑푒푚푎푛푑 푟푒푙푖푎푏푖푙푖푡푦 (퐼퐷푅); 푢푛푖푡푙푒푠푠

푄푖푛 − 푀표푛푡ℎ푙푦 푖푛푓푙표푤 푡표 푡ℎ푒 푟푒푠푒푟푣표푖푟 푎푡 푚표푛푡ℎ 푡 (푖 푟푎푛푔푒푠 푓푟표푚 푚표푛푡ℎ 1 푡표 푚); 푚3

푚 − 푆푒푙푒푐푡푒푑 푡푖푚푒 푓푟푎푚푒 (6 푚표푛푡ℎ푠 푓표푟 푤푖푡ℎ푖푛 − 푦푒푎푟 푑푎푚); 푛푢푚푏푒푟/푐표푢푛푡

푄푒푠푡 − 푇표푡푎푙 푒푠푡푖푚푎푡푒푑 푤푎푡푒푟 푑푒푚푎푛푑 푑푢푟푖푛푔 푝푟표푗푒푐푡푒푑 푡푖푚푒 푓푟푎푚푒 (6 푚표푛푡ℎ푠); 푚3

푡 − 푇푖푚푒 푠푡푒푝, 푟푎푛푔푖푛푔 푓푟표푚 13 푚표푛푡ℎ푠 푓표푟 푤푖푡ℎ푖푛 − 푦푒푎푟 푑푎푚 푡표 푚; 푛푢푚푏푒푟/푐표푢푛푡

Eq. 2.5

Where:

훽푡 − 푊푎푡푒푟 푠푢푝푝푙푦 푟푒푠푖푙푖푒푛푐푒 (푊푆푅)

3 푆푡푡 − 푅푒푠푒푟푣표푖푟 푠푡표푟푎푔푒 푎푡 푚표푛푡ℎ 푡, 푤ℎ푒푟푒 푡 = 13, … , 푚; 푚

3 푂푚푖푛 − 푅푒푠푒푟푣표푖푟 푚푖푛푖푚푢푚 표푝푒푟푎푡푖표푛푎푙 푠푡표푟푎푔푒, 푚

3 푄푖푛 − 푀표푛푡ℎ푙푦 푖푛푓푙표푤 푡표 푡ℎ푒 푟푒푠푒푟푣표푖푟 푎푡 푚표푛푡ℎ 푡, 푚

3 푄푒푠푡 − 푇표푡푎푙 푒푠푡푖푚푎푡푒푑 푤푎푡푒푟 푑푒푚푎푛푑 푑푢푟푖푛푔 푝푟표푗푒푐푡푒푑 푡푖푚푒 푓푟푎푚푒 (6 푚표푛푡ℎ푠), 푚

3 푄표푢푡 − 푀표푛푡ℎ푙푦 푤푎푡푒푟 푑푒푚푎푛푑 푎푡 푚표푛푡ℎ 푡, 푚

Calculation of IDR and WSR was done using Eq. 2.4 and Eq. 2.5 respectively. For this research, estimation of the total water demand (푄푒푠푡) was done based the data obtained from the water operators regarding the historical water released from the reservoir. This differs from the original estimation for the total water demand as underlined in the methodology,

where the water demand from the reservoir throughout the study duration was done based on the data from the first year. For this research, a moving 6-month window was used for the obtained historical dataset in order to obtain a continuous sequence of 푄푒푠푡; a block period of 6 months was used instead of 12 months as all three reservoirs were classified as within-year dam. A graphical example of its calculation is illustrated in Figure 2-4 below, where the calculation for 푄푒푠푡 for the first 6 months (m = 1) is highlighted in yellow, while the calculation for m = 6 is highlighted in blue. The corresponding IDR was then calculated by subtracting the inflow (푄푖푛) with the calculated 푄푒푠푡 for a block period of 6 months, before dividing it by the total number of months considered (6 months). The same method was used to calculate the corresponding WSR values for each month, with the incorporation of the historical values of the reservoir water storage volume.

Figure 2-4: Example of 푄푒푠푡 calculation The original methodology required that the calculated IDR and WSR be normalised using an approximation equation. However it was omitted from this research’s method, primarily due to the overall aim of this research, which is to determine the impact of water minimization practice from the industrial sector to the overall performance of an existing WSS. Although the normalisation of the two indicators could prove useful in allowing for fair cross comparison between multiple systems, it would not, however, be able to accurately measure any impact of changes made to the system. This is attributed to the method of normalizing of the indicator; the normalisation method ranks each event based on severity of the IDR or WSR values, essentially adjusting and re-assigning the calculated value into a common scale. As such, any changes to both indicators would be re-adjusted to fit the scale, resulting in other events which are not classified as ‘high vulnerability’ events being classified as such.

Based on the method of calculation, positive values for IDR indicates that the inflow is able to cater for the demand and maintain the water volume for a 6-month buffer, while positive values for WSR signals that the reservoir is able to provide adequate supply to the users in the event of low flow or dry spell. Here, positive values for both indicators suggest that the system is stable, with low vulnerability to non-satisfactory-related events. However, negative values for WSR suggest that the dam is no longer able to ensure a 6-month buffer (low resilience), while a negative value for IDR indicates that the demand outweighs the supply (low reliability); a combination of these occurrence suggest a high vulnerability scenario for the dam.

The use of IDR and WSR result for each dam also suggest which of the two strategies (SSM or DSM-based strategy) is more appropriate to address the challenges faced by the system to decrease the dam’s vulnerability. Negative values for WSR with positive IDR values indicate a potential need for a SSM-based strategies over DSM, as the root cause of the system’s vulnerability remains the inability of the dam to provide adequate buffer for the system; in this case, even if the demand was managed adequately to match the supply, future occurrence of low inflow could disrupt this balance, leading to a NS condition. On the other hand, negative values for IDR with positive values for WSR suggest the opposite need, where the implementation of a DSM-based strategy could further lower the system vulnerability without the need for a SSM-based strategy implementation, as the dam is more than adequate to provide the necessary buffer. In cases where both values are negative, it could point to the need of both strategies, although SSM-based strategies should supersede DSM; though a DSM-based strategy could assist in delaying NS conditions and, indirectly, prolonging the need of a SSM-based strategy implementation, the system is still in need of system expansion to fully address the issue of the system’s high vulnerability.

In order to calculate the IDR, the inflow into all three reservoirs was first calculated. Initial attempt to estimate the inflow was done using the Nedbør-Afstrømnings-Model (NAM) model (Goh, Zainol and Mat Amin, 2015). However, due to the low data quality of available streamflow data, calibration of the model resulted in low accuracy and thus, it was not used for the IDR and WSR calculation; the findings from this attempt are presented in Appendix B. As the data obtained from the water operators regarding the water storage and operation were of a higher quality, a water balance model was used instead. Inflow into each reservoir was calculated based on the following equation (Eq. 2.6).

푉 −푉 푄 = 푂 + 푥+1 푥 + 퐿 Eq. 2.6 푥 푥 ∆푥 푥

Where:

푥 − 퐼푛푖푡푎푙 푡푖푚푒 푓표푟 푟푒푠푒푟푣표푖푟 푖푛푓푙표푤 푒푠푡푖푚푎푡푖표푛; 푑푎푦

훥푥 − 푇푖푚푒 푠푡푒푝 표푓 푥 푡표 푥 + 1; 푑푎푦

3 푄푥 − 푅푒푠푒푟푣표푖푟 푖푛푓푙표푤 푑푢푟푖푛푔 푡푖푚푒 푝푒푟푖표푑 푥; 푚 /day

3 푂푥 − 푅푒푠푒푟푣표푖푟 푤푎푡푒푟 푟푒푙푒푎푠푒 푑푢푟푖푛푔 푡푖푚푒 푝푒푟푖표푑 푥; 푚 /day

3 푉푥 − 퐼푛푖푡푖푎푙 푠푡표푟푎푔푒 푣표푙푢푚푒 푑푢푟푖푛푔 푡푖푚푒 푝푒푟푖표푑 푥; 푚 /day

3 푉푥+1 − 퐹푖푛푎푙 푠푡표푟푎푔푒 푣표푙푢푚푒 푑푢푟푖푛푔 푡푖푚푒 푝푒푟푖표푑 푥 + 1; 푚 /day

퐿푥 − 푊푎푡푒푟 푙표푠푠푒푠 푓푟표푚 푟푒푠푒푟푣표푖푟 푓표푟 푡푖푚푒 푝푒푟푖표푑 푥 (푖. 푒. 푠푒푒푝푎푔푒 푎푛푑 푒푣푎푝표푟푎푡푖표푛); 푚3 /day

Compared to the physical-based model, the water-balance equation relies on the operational data of each reservoir, and is heavily dependent on the accuracy of the measurements. As a result, errors and uncertainties in the measurements of each parameters have been known to result in negative values (Deng, Liu, Guo, et al., 2015). Sources of error includes systematic errors of the instruments where they were not frequently calibrated or maintained, random errors during manual data-recording by operators, change in practice for measurement recording (e.g.: change in instrument, or change in datum for stage measurement), or errors during water level to volume conversion. Due to this, the method described by Deng, Liu, Liu, et al. (2015) was adopted using data smoothing and positive adjustments to minimise the likelihood of negative values.

Accordingly, all data were manually checked for inconsistencies, using basic descriptive statistics (average, standard deviation, and minimum and maximum values). Then, data smoothing was done using a moving average method by means of a 3-day window to minimise any inherent errors present in the data. Subsequently, the estimated inflow into the reservoir was then analysed using SPSS software to determine the frequency distribution to further determine possible outliers caused by random errors. The streamflow was then adjusted using positive increment, where the value was adjusted by using the minimum value of the calculated streamflow.

Finally, sensitivity analysis using the Monte Carlo simulation was performed using Excel to analyse the uncertainties caused by errors in calculation of inflow using the underlined method. Calculation for IDR and WSR was done for one month using aggregated inflow values randomly generated by the software. Iteration of the simulation was done until the standard deviation of the calculated values converged. The above explained methodology is simplified in the programming flowchart, appended in Figure 2-5.

Calculated IDR and WSR values for all three dams were compared to those obtained for the year 1998; this year was primarily selected as a baseline year as it was noted throughout various literature and press releases as the year where of longest water rationing period occurred (April, 1998 – September, 1998) as well as having one of the biggest impact on users. Furthermore, all three reservoirs recorded their lowest water level in recent history of below 40%. This period was used as a baseline to determine the cluster of NS incidents for all three dams, where all of the IDR/WSR value throughout the study period was compared against.

A scatterplot illustrating the distribution of the calculated monthly IDR/WSR values were generated in order to determine the vulnerability of the system. In this sense, a singular event over the study period is classified having a ‘high vulnerability’ in instances where both the IDR and WSR values are in the lower left quadrant of the scatterplot, as illustrated in Figure 2-6. Additionally, all singular events over the study period were assessed for its vulnerability relative to the 1998 incident. This was done via the comparison of the IDR/WSR value of each event over the study time period (i.e. singular event) against the centroid of the 1998 cluster (i.e. arithmetic mean of 1998 cluster), as illustrated in Figure 2-7 below. Here, change in a dam’s vulnerability was measured as a distance from each singular event to the arithmetic mean of the 1998 cluster, where the smaller the distance of the singular event to the arithmetic mean indicates a higher vulnerability.

Figure 2-5: Programming flowchart for inflow calculation estimation, subsequent IDR/WSR calculation, and Monte Carlo simulation

Figure 2-6: Illustration of IDR/WSR scatterplot indicating a singular event's vulnerability

Figure 2-7: Calculation of vulnerability of each singular event relative to the arithmetic mean of the 1998 cluster

2.5. Results and discussion 2.5.1. Reservoir inflow estimation

During the data collection period, historical data regarding each reservoir’s operation were obtained from the water operator, which includes information regarding the daily storage volume, water released, overflow, and environmental flow. Evaporation rates over the reservoir were estimated using the average monthly evaporation from station EV 3117370 obtained from the Department of Irrigation and Drainage (DID), and using pan coefficient of 0.90 (Jabatan Pengairan dan Saliran Malaysia, 1976). However, data regarding reservoir seepage were not available and thus omitted from the calculation.

The combined use of the moving average smoothing technique and positive adjustment method successfully minimised the occurrences of negative values of inflow, as illustrated in Figure 2-8 for the Semenyih Dam. However, there were some dates which resulted in significantly high values (both negative and positive) of inflow; these values were assumed to be resulted from random errors during recording of data by the operators and have been excluded. As the number of these data were extremely small (Table 2-3), the exclusion of these numbers did not have any significant impact on the IDR and WSR calculation. Summary of relevant statistics of the calculated inflow, observed outflow, and number of omitted data is summarised in Table 2-3 below.

14,000

13,000

12,000

11,000

Storage(MG) 10,000

9,000

8,000 1/1/2010 1/1/2011 1/1/2012 1/1/2013 1/1/2014 1/1/2015 1/1/2016 Date

Storage (Raw) Storage (Smoothed)

Figure 2-8: Sample of data smoothing effect for Semenyih Dam (1/1/2010 - 1/1/2016)

Table 2-3: Summary of inflow/outflow of Semenyih, Klang Gates, and Langat dam

Semenyih Klang Gates Langat Period of historical data collected 1997 - 2016 1998 - 2016 1997 - 2016 Number of omitted data (days) 9 6 0 Maximum estimated inflow (m3/s) 28.5 36.9 16.5 Average estimated inflow (m3/s) 11.0 2.8 5.1 Minimum observed outflow (m3/s) 1.5 0.9 1.1 Maximum observed outflow (m3/s) 26.6 2.8 10.0 Average observed outflow (m3/s) 2.0 2.2 2.5

Finally, in order to assess the impact of uncertainties on the calculated inflow using the water balance equation, sensitivity analysis using Monte Carlo simulation was carried out using Excel. As highlighted in subsection 2.4, the number of iterations for the simulation was done in an exploratory method, where the simulation was initially run with 20,000 iterations. However, convergence of both the average IWR and WSR values, and the standard deviation was reached at 10,000; as such, the simulation used an iteration of 10,000.

Results of the sensitivity analysis of the data samples indicated that the use of approximate inflow data had not significantly affected the calculation of both IDR and WSR. It was found that changes in daily inflow between ± 50% had only resulted in an increase of the calculated IDR and WSR by 0.306 and 0.123 respectively. The small change of the calculated indicators could be attributed to the aggregation method used, where the daily inflow data was ultimately accumulated into monthly inflow, which was further assessed from a time frame of 6 months. Due to this, the method for calculating the IDR and WSR could be considered robust, allowing the use of approximate estimated inflows into a reservoir to calculate the two indicators.

2.5.2. Inflow, outflow, and water level

The estimated inflow (inflow), observed release to WTP (outflow), and water level of each of the three reservoirs were plotted and are illustrated in Figure 2-9, Figure 2-10, and Figure 2-11 for Semenyih, Klang Gate, and Langat dam respectively. Overall, all three reservoirs showed a similar rise and fall of inflow into the system, only differing in magnitude. For all

three dams, the period of 2002 to 2005, and 2013 to 2016 showed a lower than average inflow into the system, while other years indicated a higher than average inflow. This is consistent with the report of water-related incidents reported in the local press.

For Semenyih (Figure 2-9) and Langat (Figure 2-11) dams specifically, the outflow from the reservoirs was known to depend on the ability of the downstream catchment to provide adequate water resources to each WTP; in the instances where the streamflow is low, only then will water be released from the dam. As such, an oscillating signal of inflow-outflow was observed, whereby a high outflow was perceived for periods of low inflow into the system, and vice versa. Although that is the case, there were also several instances where the high outflow matched that of the inflow. From this, it can be seen that instances where the outflow is higher or equal to the inflow into the system will result in a decline of water level in the reservoir. Additionally, it was also observed that the reduction in inflow and increase in outflow for both systems was slightly higher in magnitude in year 2013-2016 as compared to 2002-2005. The reduced inflow into the system is consistent with the statement made by the water operators, who had noted the change in inflow into the system as compared to historical patterns. As for the increased magnitude for the outflow over time, it could be due to a combination of several factors namely reduction of streamflow of the downstream rivers which feeds into the WTPs (indicated by the reduction of inflow into the reservoir, as noted by both the water operator as well as the historical data presented), an increase in water abstraction along the rivers within the downstream catchment by other sources (i.e.: premises which obtained license for direct abstraction from the river), a change in the dam’s water management practices, or an increase in overall water demand as compared to previous years. Nevertheless, due to the increase in magnitude for both the inflow and outflow as compared to historical condition, the system had become more unstable, resulting in a more rapid change in the reservoir’s water level over time.

5000 120 4500 100

4000 3500 80 3000 2500 60 2000 40

1500 Dam capacity (%) capacity Dam 1000

Inflow/outflow (MDG) Inflow/outflow 20 500 0 0

Estimated monthly inflow (MDG) Observed monthly water demand (MDG)

Average monthly inflow (MGD) Average monthly water demand (MGD)

Dam Capacity (%)

Figure 2-9: Monthly inflow, monthly water demand, and dam capacity for Semenyih dam

14000 120

12000 100

10000 80 8000 60 6000 40

4000 Dam capacity (%) capacity Dam

Inflow/outflow (MLG) Inflow/outflow 2000 20

0 0

Estimated monthly inflow (MLD) Observed monthly water demand (MLD)

Average monthly inflow (MLD) Average monthly water demand (MLD)

Dam Capacity (%)

Figure 2-10: Monthly inflow, monthly water demand, and dam capacity for Klang Gates dam

18000 110 16000

14000 70 12000 50 10000 30 8000 10 6000 -10

4000 (%) capacuty Dam Inflow/outflow (MLD) Inflow/outflow 2000 -30 0 -50

Estimated monthly inflow (MLD) Observed monthly water demand (MLD)

Average monthly inflow (MLD) Average monthly water demand (MLD)

Dam Capacity (%)

Figure 2-11: Monthly inflow, monthly water demand, and dam capacity for Langat dam

In comparison, Klang Gates dam (Figure 2-10) exhibits a more check-and-balance condition, where the outflow responded to the condition of both the inflow and the water level of the reservoir. As it is the only source of raw water for the Bukit Nanas and Wangsa Maju WTPs, it was imperative that the water level be conserved in order to ensure that the reservoir is able to meet the 6 months demand. Due to this, it can be seen that there is a reduction in the outflow following the months of low inflow or low water level in the reservoir. Similar to Semenyih and Langat, this system also became relatively more volatile in 2002-2006 and 2013-2016 as compared to other years. Comparing between the two periods, a change in practice can be seen, where the system could no longer afford to reduce the outflow in 2012- 2016 to the same degree as 2002-2006. In the former years, the system had adjusted the water outflow in accordance to the inflow in attempt to maintain the water level in the reservoir; this had resulted in only one instance where the water level had dropped below 60% - an ‘alert’ level identified by the water operator, indicating that the system is approaching low storage level. However, in the latter years, although a similar pattern in below average inflow was perceived, the output remained consistent, only reducing for a short period of time in 2014. As a result, the water level in the dam had dropped below the 60% level twice within 4 years. This change in pattern could point to the possibility of an increase in water demand in 2012-2016 as compared to 2002-2006.

2.5.3. IDR/WSR scatterplot for reservoir vulnerability assessment

A scatterplot of IDR-WSR was generated to examine the exact relationship between the inflow-demand (IDR) and storage (WSR), with the 1998 incident of water rationing in Selangor used as a reference point. All data plotted in the IDR-WSR scatterplot were clustered into two categories, namely:

Low vulnerability – high (positive) values of IDR and WSR, which indicate low probability or no threat of possible water rationing

High vulnerability – low (negative) values of IDR and WSR, with a high probability of water rationing

Figure 2-12: IDR/WSR scatterplot for Semenyih dam

Figure 2-13: IDR/WSR scatterplot for Klang Gates dam

Figure 2-14: IDR/WSR scatterplot for Langat dam

Results of the scatterplot for Semenyih, Klang Gates, and Langat are shown in Figure 2-12, Figure 2-13, and Figure 2-14 respectively. Isolated readings of both IDR and WSR for the 1998 event for all three reservoirs were successfully done based on the scatterplot of IDR/WSR. For each plot, a cluster of values at the bottom left of the plot corresponds to the conditions that were observed during the 1998 water rationing occurrence, thus underlining the circumstances of high vulnerability of each dam. The recorded lowest values of IDR and WSR suggest that high vulnerability conditions in all three dams can be linked to both the low inflow-outflow relationship as well as the inability of the system to meet the demand.

Post 1998, there were several other incidents where the tabloids have reported major water supply-related problems in Malaysia although to a lesser degree than the 1998 incident. Reports of the cause of water rationing then were primarily attributed to the low storage capacity for all three dams. The result of the scatterplots above mirror these reports, where it can be seen that there were data clusters of low IDR/WSR readings similar to that in 1998. Interestingly, Klang Gates dam recorded an even lower reading of both IDR and WSR in 2016 as compared to 1998, while Langat dam had several instances in 2005 and 2015 where both indicators matched those in 1998.

7.00

6.00

5.00

4.00

3.00

2.00

1.00 Distance Distance toartihmatic mean, unitless

0.00

1/1/2006 1/1/1998 1/9/1998 1/5/1999 1/1/2000 1/9/2000 1/5/2001 1/1/2002 1/9/2002 1/5/2003 1/1/2004 1/9/2004 1/5/2005 1/9/2006 1/5/2007 1/1/2008 1/9/2008 1/5/2009 1/1/2010 1/9/2010 1/5/2011 1/1/2012 1/9/2012 1/5/2013 1/1/2014 1/9/2014 1/5/2015 1/1/2016 1/9/2016 Date

Figure 2-15: Vulnerability curve for Semenyih dam

1.40

1.20

1.00

0.80

0.60

0.40

0.20

0.00

-0.20 Distancee Distancee arithmaticto mean,unitless

-0.40

1/6/2014 1/6/1998 1/2/1999 1/6/2000 1/2/2001 1/6/2002 1/2/2003 1/6/2004 1/2/2005 1/6/2006 1/2/2007 1/6/2008 1/2/2009 1/6/2010 1/2/2011 1/6/2012 1/2/2013 1/2/2015 1/6/2016

1/10/2001 1/10/1999 1/10/2003 1/10/2005 1/10/2007 1/10/2009 1/10/2011 1/10/2013 1/10/2015 Date

Figure 2-16: Vulnerability curve for Klang Gates dam

4.50

4.00 3.50 3.00 2.50 2.00 1.50 1.00 0.50

0.00 Distance Distance toarithmetic mean,unitless

-0.50

1/9/2016 1/1/1998 1/9/1998 1/5/1999 1/1/2000 1/9/2000 1/5/2001 1/1/2002 1/9/2002 1/5/2003 1/1/2004 1/9/2004 1/5/2005 1/1/2006 1/9/2006 1/5/2007 1/1/2008 1/9/2008 1/5/2009 1/1/2010 1/9/2010 1/5/2011 1/1/2012 1/9/2012 1/5/2013 1/1/2014 1/9/2014 1/5/2015 1/1/2016 Date

Figure 2-17: Vulnerability curve for Langat dam

Figure 2-15, Figure 2-16, and Figure 2-17 each illustrates the vulnerability curve of the Semenyih, Klang Gates, and Langat dam respectively. Overall, it could be observed that the Klang Gates dam is the most vulnerable to NS event as compared to the Semenyih and Langat dam, where the range of distance of all the singular events to the arithmetic mean of the 1998 cluster for this dam is the smallest (i.e. maximum distance of approximately 1.2). Comparatively, the maximum distance observed for Semenyih dam was the highest of approximately 6.31, with Langat dam resulting in a maximum distance of approximately 4.1. Interestingly, several of the calculated distance for several of the singular events outside of the 1998 events for Klang Gates dam specifically resulted in multiple negative values, which signal that the calculated IDR and WSR values for those incidents were lower than calculated for the 1998 cluster.

The above findings further emphasise the probability of increased susceptibility of the three dams to NS condition over time. The difference in response between Semenyih and Langat/Klang Gates dam could be due to the larger size and higher volume of active storage of Semenyih dam as compared to the other two. Similar to the 1998 incident, IDR and WSR indicators for Langat and Klang Gates followed the same trend in 2016 (Klang Gates dam) and in 2015 (Langat dam), recording the lowest reading for both indicators, indicating high vulnerability for the dam during these years as compared to the 1998 incident.

Another observation was the distance of the cluster between the 1998 incident with other years. Comparing between the three dams, only scatterplot for the Semenyih dam resulted in a clear segregation between the two clusters, further indicating the stability of this system compared to former years. In contrast, the cluster for both the Klang Gates and Langat dam were intermixed between all the years considered, suggesting a worsening condition as compared to the historical situation.

2.5.4. Detailed IDR and WSR analysis for Semenyih, Klang Gates, and Langat dams

Results for the calculated IDR and WSR for Semenyih, Klang Gates, and Langat dam are illustrated in Figure 2-18, Figure 2-19, and Figure 2-20 respectively. Comparing the calculated IDR and WSR for all three dams, the interaction between the two indicators differ from one another despite some similarities in dam classification (direct vs regulated) and size. This is further discussed below for each dam.

i. Semenyih dam

WSR for Semenyih dam showed a consistently higher reading than IDR (Figure 2-18); the calculated values for WSR rarely dropped below zero – this suggests that the dam is fully capable in providing an adequate buffer during dry spells. This is consistent with the overall pattern of water level in the dam, which only showed one event where the water level was below the warning level of 60%. In comparison, the calculated IDR value fluctuated between positive and negative, indicating several occasions where the inflow could not keep up with demand. Similar to WSR, readings for IDR were lowest in 1998; cross-referencing with Figure 2-9, this was primarily attributed to the consistently higher water release from Semenyih dam for several months relative to the average, as compared to the inflow into the system.

A decline of both IDR and WSR values could be observed for the two periods of reduced inflow and increase in outflow; specifically in 2002-2005 and 2013-2016. The WSR for Semenyih showed a decrease in reservoir resilience in the latter period as compared to the former by an order of magnitude; although the dam capacity for the period of 2013-2016 did not drop below 60%, the calculated WSR indicated several instances where the dam's resilience was low (below zero). Coupled with low values of IDR, this suggests that the

system was approaching similar conditions to which water rationing had been carried out in 1998. However, the main difference between the 2013-2016 period as compared to the 1998 event was that the constant reading of slightly above zero in former period is due to a prolong episode of low inflow and high outflow, as compared to the instantaneous release of high outflow from the dam in 1998. This, in turn, led to longer instances of below zero readings in 2013-2016.

As indicated in the methodology section, the relatively consistent positive values for WSR while a fluctuating IDR values suggest that the dam is capable to avoid high vulnerability instances, though results also suggest that the system is approaching failure. In this case, the implementation of an affective DSM-based strategy could aid in further decreasing the system’s vulnerability, thus avoiding the need for system expansion.

4 4

3 3

2 2

1 1

0 0

demand reliability, IDR reliability, demand -1 -1

- Inflow -2 -2 WSR resilience, storage Water

-3 -3

-4 -4

IDR WSR

Figure 2-18: IDR and WSR plot for Semenyih dam

ii. Klang Gates dam

Interestingly, the IDR and WSR profile for Klang Gates dam (Figure 2-19) indicates a sensitive system which is heavily dependent on the inflow/outflow relationship. As compared to Semenyih dam, the interaction between the IDR and WSR for Klang Gates dam is the opposite, where the calculated IDR values are constantly higher than WSR. In terms of WSR, it could be observed that the values were mostly negative, even though the reservoir was full,

potentially indicating that the reservoir was not capable to meet the 6 months demand in the case of low flow (small reserve margin). In contrast, the IDR readings for Klang Gates dam were consistently in the positive range. For the latter, this could be linked to the type of dam itself (direct abstraction). Due to the direct link between the dam and WTP, it allowed the water operator to control the demand relative to the supply; this was highlighted in the previous section, where it could be observed that the outflow was controlled in order to maintain adequate buffer. Due to this, it could be concluded that the Klang Gates dam is more dependent on the inflow/outflow management than the ability of the reservoir to provide adequate buffer. On the other hand, the size of the reservoir could also be a factor which influences the instability of the reservoir’s reliability, where the demand had simply surpassed the buffer’s capacity; this is reflected in the rapid fluctuation between positive and negative values for WSR as compared to Semenyih dam.

For the timeframe of 2002-2006, results of IDR and WSR as compared to the physical water level in the dam were contradictory; both the IDR and WSR had suggested the system is more than capable to supply adequate supply to the users. However, the water level in the Klang Gates dam was relatively low (approximately 60%); this could be due to the reduction in water demand, which dropped to well below the average in order to allow for the system to recover (to increase water level in dam) and compensate for the low inflow into the system. As such, both indicators resulted in positive readings even though the water level was relatively low. Comparing this period with that of 2012-2016, the IDR and WSR values were the opposite. Although this period also indicated several instances of low water level in the dam, both IDR and WSR values were well below zero, indicating that the system is no longer able to meet the demand, both from the standpoint of inflow/outflow relationship as well as reservoir storage. Thus, specifically for this system, it can be concluded that both DSM and SSM strategies needs to be considered in order to decrease its vulnerability.

1 1

0.8 0.8

0.6 0.6

0.4 0.4

0.2 0.2

demand reliability, IDR reliability, demand 0 0

- Inflow -0.2 -0.2 WSR resilience, storage Water

-0.4 -0.4

-0.6 -0.6

IDR WSR

Figure 2-19: IDR and WSR plot for Klang Gates dam

iii. Langat dam

Although Langat dam has the same classification (regulating reservoir) as that of Semenyih dam, it is more comparable to the Klang Gates dam in terms of size and capacity; it was observed that the resultant IDR and WSR relationship is a mixture of the two. In terms of response, it is similar to Klang Gates dam where the change in signal has a sharper incline/decline. However, changes in the positive/negative response for IDR and WSR is more gradual, similar to Semenyih dam.

From Figure 2-20, it can be seen that the IDR and WSR curve for Langat dam interchanges with one another, where there are instances where the IDR is higher than WSR and vice versa. Although this is the case, the value of IDR and WSR typically mimics one another (i.e.: when IDR is negative, WSR will tend to be negative as well); this is similar to the response exhibited in Klang Gates dam, further reinforcing the fact that the size of the reservoir impacts the likelihood of high vulnerability occurrences. As compared to Klang Gates dam, the output to the three WTPs cannot be as easily controlled, rather it depends on the condition of the downstream basin; this could be the reason why the oscillations between positive and negative for IDR values are more prominent in Langat dam as compared to Klang Gates dam. In the events where the dam was required to release water to supply downstream WTPs, the IDR resulted in negative readings, indicating that the inflow is no longer able to support the

outflow. In these instances, the resilience of the dam is crucial in order to ensure continuous supply to the users; however the resilience changed over time. Between the periods of 2002- 2005, the WSR indicates that the dam had approximately equal stretches of period of positive and negative values, signalling that the dam had begun to approach system failure. In 2013- 2016, similar conditions of high output/low input were observed, where the IDR response is similar to the 2002-2005 period resulting in a mostly negative reading for IDR. However, it could be observed that the WSR mostly fell below zero, indicating that the reservoir is no longer able to cope with the high demand. Based on this, similar to Klang Gates dam, both DSM and SSM strategies are needed to decrease Langat’s dam vulnerability.

2 2

1.5 1.5

1 1

0.5 0.5

0 0

demand reliability, IDR reliability, demand -0.5 -0.5

- Inflow -1 -1 WSR resilience, storage Water

-1.5 -1.5

-2 -2

IDR WSR

Figure 2-20: IDR and WSR plot for Langat dam

2.6. Conclusion

In order to assess the impact of water minimization from the industrial sector to alleviate water imbalance issues on existing WSS, each of the reservoirs’ vulnerability to non- satisfactory condition was first established. The methodology as introduced by Mehran, Mazdiyasni and Aghakouchak (2015) was used, where the calculation of the inflow-demand reliability (IDR) and water supply resilience (WSR) was done for the Semenyih, Klang Gates, and Langat dam. This research expanded on the original research by introducing the clustering method to determine the vulnerability of each reservoir as compared to the known

water rationing incident in 1998. Here, the cluster of IDR/WSR value for the 1998 water rationing incident was identified through the establishment of the IDR/WSR scatterplot, where the modified methodology was able to isolate values relating to the aforementioned period. Furthermore, distance of each singular event throughout the study period to the arithmetic mean of the 1998 cluster was also calculated in order to determine the magnitude of the reservoirs’ vulnerability.

From the analysis, it can be seen that all three reservoirs have shown an increase in system vulnerability, where the root cause identified for all three reservoirs was the increase in demand, as well as the relatively smaller reservoir size for Klang Gates dam and Langat dam specifically. As for the increase in demand, this can be observed through the increase in volume of water release for Semenyih (Figure 2-9), and Langat dam (Figure 2-11), and via the change in water restriction practices in 2013-2016 for Klang Gates dam (Figure 2-10). Comparing between the three dams, it can be seen that Semenyih dam has the lowest vulnerability as compared to Klang Gates dam and Langat dam, indicating that the size of the storage does play an important role in ensuring the sustainability of a WSS. However, both Klang Gates dam and Langat dam have had periods where the reservoir is at full capacity, suggesting that both dams have the capability to supply adequate water to the users if the demand could be reduced via affective implementation of DSM strategies.

Calculation of both IDR and WSR for each reservoir supported these findings, where values of both IDR and WSR were found to decrease in the recent years (2013 – 2016) as compared to historical scenarios, indicating an increase in system vulnerability. Benchmarking the current timeframe with that of the 1998 event, it was discovered that, for both Klang Gates dam and Langat dam, there were several other occurrences of low reliability (IDR) and low resilience (WSR) which mimics the conditions of the 1998 event. For Semenyih dam, there was no other timeframe which falls into the low reliability/low resilience cluster, indicating the system’s low vulnerability against NS events; however, long-term pattern for both indicators suggest that the system is approaching conditions of system failure, possibly leading to increase system vulnerability if no changes are made to the current water management. As such, it can be concluded that a combination of SSM-based and DSM-based strategies are needed to decrease Klang Gates dam and Langat dam’s vulnerability, while an implementation of effective DSM strategy for Semenyih dam should be sufficient to further increase its stability.

Chapter 3 : Estimation of potential water savings from industrial premises

3.1 Introduction

In Chapter 2, it was illustrated that the three reservoirs are vulnerable to non-satisfactory (NS) events, and that demand-side management (DSM) based strategies targeted at various users within the water supply system (WSS) might be able to alleviate this vulnerability. In Chapter 1, it was noted that the full cost of compliance to implement a market-based policy will need to consider other complementary bottom-up strategies to increase the effectiveness of the aforementioned policy. For Malaysia specifically, the government has identified several complementary strategies targeted at the industrial sector specifically, namely the use of alternative water sources and water efficiency audits (National Water Service Commission (SPAN), 2016a). In this chapter, the impact of one of the DSM-based strategies proposed – specifically the use of alternative water sources – is assessed, where the impact of potential water saving to the overall WSS is further evaluated. Subsequent technical feasibility, financial cost, and environmental impact will be further explored in Chapter 4.

Alternative sources for water reuse have been considered from the perspective of domestic wastewater as well as industrial wastewater. For the former, the piping and distribution cost from the domestic WWTP to the individual IPs could be costly due to the distance, potentially rendering this option unfeasible. In this sense, the use of industrial wastewater generated and reused within the facility would require a shorter piping and distribution network, resulting in smaller cost as compared to the former option.

Water minimization using industrial wastewater has been studied from the perspective of both intra (within)-plant, as well as inter (between)-plant in the industrial sector. However, several major limitations have been reported regarding actual implementation of intra-plant approaches, as noted previously in Chapter 1. As such, the application of inter-plant approaches through the concept of industrial ecology have been seen as a more viable way of overcoming these obstacles, as they offer a mean to combine efforts by these firms towards a common goal. Originating from the concept of industrial symbiosis, industrial ecology has been known to engage industries which traditionally operated in isolation, towards optimizing their total material cycle via physical exchange of minerals, energy, water, and by-products (Chertow, 2000). This concept has been the foundation of the Eco-Industrial Park (EIP), which emphasised the development and management of an industrial park towards achieving

a high level of environmental, economic, and social benefits through mutual collaborations and synergies between companies (Taddeo, 2016).

The implementation of EIP could be achieved through the conversion of existing industrial parks. Here, it is defined as specific areas allocated for industrial facilities, separated from a densely populated and urbanised area (Ahmed, 2012). According to Lambert & Boons (2002), IPs can be categorised into two types; industrial complexes, primarily consisting of heavy industries which are mutually interrelated, and mixed industrial parks, which typically host a variety of small- and medium-sized enterprises (SMEs). Due to the close proximity of the industrial facilities to each other within an IP, the implementation of industrial ecology has been seen as a viable concept for reducing an industry’s water consumption (Aviso, 2014).

Literature published has generally been observed to study inter-plant water minimization approaches within an IP via direct or indirect integration. The concept of direct integration between facilities has been of treated wastewater distributed from one plant (source) to another (sink) via a cross-plant pipeline, while indirect integration has focused on a localised or decentralised reuse system, shared between a cluster of plants (each plant acts as both sink and source) (Chew et al., 2008; Lim and Park, 2010). For example, in a study by Boix et al. (2012), they attempted to compare various configurations of both direct and indirect water reuse systems using Mixed-Integer Linear Programming (MILP), and found that the best configuration was an EIP with direct integration. In contrast, Chew et al. (2008) found that an implementation of a centralised utility hub containing a reuse unit showed a lower freshwater and wastewater flow rate in the overall water network. Furthermore, they also noted that the proposed configuration improved the overall water network practicability and flexibility, and was capable of serving a greater number of plants with individual water networks. In another study which considered both water optimization and piping cost, Alnouri et al. (2015) demonstrated that utilization of both an on-site and off-site reuse system yielded the most economical option as well as utilising the least volume of freshwater and generating the lowest volume of wastewater discharge.

Successful implementation of inter-plant water minimization approaches have been reported to rely on several factors, similar to EIP implementation. In a critical review of the success and limiting factors of an EIP realization, Sakr et al. (2011) identified six factors: symbiotic business relationship, economic value added, awareness and information sharing, policy and regulatory framework, institutional and organizational setup, and technical factors; of the six,

the symbiotic business relationship between companies has arguably been seen as a critical element. Good relationships and mutual trust between the companies have been seen as key factors due to the significant investments expected of each company, as well as being the foundation which allows for confidential information between firms to be exchanged. Gibbs and Deutz (2007) further emphasised this; they noted that differences between organizational cultures within the EIP could result in a behavioural barrier, leading to a higher resistance which might overshadow any economic gain from resource exchange. Elabras Veiga and Magrini (2009) also underlined this factor as one of the threats in establishing EIPs in Brazil, noting that businesses tend to be unwilling to depend on, or participate in any knowledge sharing with other businesses due to this lack of trust and good relationships.

In this regards, indirect inter-plant water minimization approach could aid in bridging trust between all players within an IP. The configuration proposed by Alnouri, Linke and El- Halwagi (2015) above was specifically considered for this research, where both on-site (private wastewater treatment plant (WWTP) for each firm) and off-site (localised or decentralised) wastewater reuse treatment facility were considered for individual IP; there are several reasons behind this. First, with the establishment of an off-site wastewater reuse facility outside the boundary of the firms, the element of trust would then be transferred to the owner of the facility (which could either be privately or publically owned) instead of between firms, thus overcoming the perceived risk of information leak regarding confidential operational processes between competing businesses. Second, it allows for a more flexible and practical network configuration between each premise and the treatment facility, allowing for natural evolution of plant ownership within an IP over time. Third, the establishment of a decentralised WWTP reduces the need for individual investment by each firm to install isolated advanced treatment systems in order to comply with water reuse standards.

In addition to this, the proposed configuration also takes advantage of the previously established regulation set by the Department of Environment (DoE), whereby it was stated that each premise is required to pre-treat their industrial effluent to an upper limit set forth in the Environmental Quality Act 1974 (Fifth, Seventh, and Eighth Schedule). Industrial premises are expected to comply with the concentration limit of Standard A if the effluent is discharged into any inland waters within any catchment areas, or the limits of Standard B otherwise (Department of Environment Malaysia, 2009). By assuming that each premise adhere to this regulation, the high variation inherently found in all effluent discharged by an

industrial premise due to the differences in processes (Gurnham, 1965) can now be streamlined. As such, the reduced contaminants concentration could potentially decrease operation and maintenance cost of a designed WWTP by reducing the use of electricity, chemicals, and equipment maintenance and change. The list of all parameters monitored, as well as the upper limit for each parameters in the Fifth and Seventh Schedule are listed in Appendix C.

Similar to the water minimization approach discussed above, the reuse of treated wastewater has also been known to be done via indirect or direct reuse; however, their definition differs slightly. Here, indirect industrial wastewater reuse refers to the discharge of the treated wastewater into an environmental buffer (e.g. streams or river) prior to its reuse, downstream of the buffer; in contrast, direct reuse omits this buffer, where instead the treated wastewater is consumed directly by the user (Asano et al., 2007). Streamlining the various definitions discussed in the previous section with the overall objective of this research, this scope of this study is confined to assess the application of an indirect inter-plant water minimization approach for direct industrial wastewater reuse.

From the end-use perspective, the standard to which the water quality of the treated industrial wastewater would need to conform to is highly dependent on its use, where it would need to conform to drinking water standard for potable water reuse, or irrigation water standard for non-potable agriculture water reuse (Sadr et al., 2015). The reuse of industrial wastewater has predominantly been reported for non-potable water reuse, either for non-potable domestic use (e.g. toilet flushing), agriculture or landscape irrigation purposes, or for industrial consumption (e.g. cooling) (Bixio et al., 2008). For example, Adewumi, Ilemobade and Van Zyl (2010) reported the use of treated wastewater in Johannesburg for both communal toilets and landscape irrigation with no records of public health related incidents. Similar uses were reported by Piadeh, Alavi Moghaddam and Mardan (2014), of treated industrial wastewater for irrigation purposes in Iran.

Public health is paramount; both potable and non-potable standards monitor closely for both pathogenic and chemical contaminants (Asano et al., 2007). There are various internationally recognised standards used as benchmarks for both types of water end uses such as World Health Organization (WHO) guidelines for drinking water quality (World Health Organization, 2008), USEPA national primary drinking water regulation (Lee and Tan, 2015), and national drinking water guidelines developed by various countries (e.g. Ministry of

Health Malaysia, 2010; Singapore Public Utility Board, 2016) for potable drinking water standard, while for non-potable water reuse for irrigation, the WHO Guidelines for the safe use of wastewater, excreta, and greywater in agriculture (WHO, 2006; Sadr et al., 2015), California code of regulation for water recycling criteria (California Department of Public Health, 2014), and the ISO 16075: 2015 series for the guidelines for treated wastewater use for irrigation projects (International Organisation for Standardisation, 2015). The main differences between the standards for both end uses are the parameters and the maximum allowable concentration of contaminants monitored; the former standards are set to a higher quality than the latter.

For both end uses, although pathogenic contaminants have been considered a primary hazard for domestic wastewater reuse, it has been argued however that the risk in industrial wastewater reuse has been principally due to chemical contaminants – specifically, heavy metals, organic compounds, and other toxic substances – as industries utilise a higher volume of chemicals (Mohsen and Jaber, 2003; Carr, 2005; WHO, 2006); an example of this is endocrine disrupting chemicals/compounds (EDCs). EDCs have been recently noted to be present – albeit in lower concentrations than in natural hormones – in untreated sewage effluent (Toze, 2006). However, these compounds have also been known to be higher in industrial effluent; for example, in a study in Toronto, Canada, it was found that 82% of 97 industrial wastewater samples discharged a median concentration of 1056 μg/L of a type of EDC, nonylphenol ethoxylates (NPEO) – well above the city’s limit of 10 μg/L (Lee et al., 2002).

Due to the higher organic and inorganic concentration as well as the presence of trace elements of both compounds, it has been emphasised that the inclusion of industrial wastewater for reuse has either been discouraged (Carr, 2005; du Pisani, 2006) or minimised (WHO, 2006). Furthermore, the regulations for industrial effluent discharge do not monitor for majority of the parameters specified in the drinking water standard for potable water reuse, adding to the risk of high concentrations of unknown toxic substances in the discharged effluent. Consequently, literature regarding industrial wastewater reuse has been more commonly inclined towards non-potable reuse, predominantly due to the unknown risk of the toxicological effect of chemical contaminants present in the wastewater if used instead for potable reuse (Tchobanoglous, Butron and Stensel, 2004).

Development for non-potable wastewater reuse standards and guidelines have generally been for agriculture and irrigation applications (Kellis et al., 2013). Although limited, there are several standards which cater specifically for industrial wastewater reuse; some examples are by the Public Utility Board (PUB) of Singapore (Singapore Public Utility Board, 2009), various states in the U.S.A. (U.S. Environmental Protection Agency, 1992), and the Iranian Strategic Planning and Supervision Division (Piadeh, Alavi-moghaddam and Mardan, 2018). Compared to the limits specified by standards and guidelines for agriculture and irrigation use, those which were tailored for industrial use placed heavier weight on the wastewater effluent’s physical and chemical characteristics as compared to its biological characteristic, as these chemical constituents will have detrimental impacts on industrial instruments and equipment (e.g. corrosion) (Asano et al., 2007).

For this research, the use of the Malaysian drinking water standard, which has been developed in accordance with WHO drinking water standard was used as the benchmark for potable water reuse (Ministry of Health Malaysia, 2010). However, a combination of Singaporean and Iranian standards for industrial water quality were used for non-potable water reuse where the lower limit for each parameter of the two standards was used, as well as WHO Guidelines for the safe use of wastewater, excreta, and greywater in agriculture was used (for trace elements limit), as there is no universally excepted standard for industrial wastewater reuse; parameters and maximum allowable limit of contaminants for each parameters are summarised in Appendix D.

3.2 Aim and research questions for Chapter 3

The second part of this research aims to quantify the potential savings that could be expected from the use of alternative water supply at industrial park level (IP) via a decentralised wastewater treatment plant, and measure its impact to elevate water imbalance occurrences in the three WSSs. For this research specifically, the term decentralised wastewater treatment plant is defined as a treatment facility that treats wastewater in close proximity of the source (Libralato, Volpi Ghirardini and Avezzù, 2012). Accordingly, the following objectives were considered:

RQ 3: What is the water consumption and subsequent wastewater effluent generation from the industrial sector within an IP?

RQ 4: What is the potential estimated water savings from multiple IPs within a service area (SA) which can be realised by means of a decentralised wastewater reuse system?

RQ 5: What is the potential impact of water savings within a SA onto its corresponding WSS?

3.3 Literature review 3.3.1. Estimation of industrial potable water consumption

As compared to the residential sector, water use in the industrial sector differs drastically. It has primarily been used for energy or vapour production, refrigeration, direct input for production of output, cleaning, and other sanitary uses – although the distribution of usage within the confine of a firm’s boundary differs from one to another (De Gispert, 2004). Gleick et al. (2003) indicated that industrial, commercial, and institutional (ICI) water use can be classified into six broad end uses – sanitation, cooling, landscaping, process, kitchen, and laundry – and that the mix of uses and quantity differs greatly depending on industry type and activities. Due to this large variation in water related activities, literature focusing on the estimation of industrial water demand has been scarce at best, predominantly due to the difficulty in collecting appropriate data from industry players (Reynaud, 2003). In general, there have been two broad approaches to estimating industrial water use reported in the literature – econometric and unit water demand.

3.3.1.1 Econometric models

From econometric perspective, water demand is a derived demand based on the price and cost shares of water and other input to produce an output. For both industrial and commercial demand, it follows the Eq. 3.1, where Q is the quantity of water as a derived demand, P is a measure of all factor prices of all input, S is factor cost shares, Y is the level of output, and Z represent all other independent variable (Worthington, 2010).

푄 = 푓(푃, 푆, 푌, 푍) Eq. 3.1

Further expanding Eq. 3.1, the prices (P) which determine the output production can further be estimated via cost function. Typically, cost function for an output (Y) is dependent on the pricing of several variables, namely the capital (푝푥), labour (푝푙), materials (푝푚), energy (푝푒), and water (푝푤), expressed in Eq. 3.2:

퐶(푌) = 퐶(푝푥, 푝푙, 푝푚, 푝푒, 푝푤) Eq. 3.2

Renzetti (1992) further examine the cost of water usage on the overall output production, by decomposing water use into four separate components (Eq. 3.3), namely quantity of water obtained from external sources (푝푖푛) , water treatment prior to use (푝푡푟푡) , water recirculation (푝푟푐푟), and wastewater effluent treatment prior to discharge (푝푑푖푠):

퐶(푝푤, 푌) = 퐶푤(푝푖푛, 푝푡푟푡, 푝푟푐푟, 푝푑푖푠) Eq. 3.3

Using the cost function appended in Eq. 3.3 estimation, price elasticity of water relative to the output production was then estimated. The majority of the literature estimating industrial water demand presented their findings through the calculation of water demand price elasticity. Findings reported in literature resulted in varying figures, depending on the manufacturing activities. For example, Reynaud (2003) estimated a network water price elasticity of -0.10 to -0.79 for France, depending on the type of industry considered, while Ku and Yoo (2011) estimated a range of 0.02 to 0.04 for Korea. Due to the difference in reported figures, this resulted in a lack of consensus on the range of value of water demand price elasticity (Reynaud, 2003).

The strength in using water demand price elasticity lies in its ability to predict the change in water demand relative to the change in water price, as well as its impact on the overall yield of a firm. Furthermore, it has been illustrated that, as compared to the residential and agriculture sectors, industrial water consumption is relatively responsive to price (Cantin, Shrubsole and Aït-Ouyahia, 2005), although the effect is limited, as water, in general, only accounts for a limited portion of their overall total cost of production (Renzetti, 2005). For example, Duport and Renzetti (2001) illustrated that an increase in a non-water input variable (e.g.: labour, capital, and energy) has been shown to result in a reduction in internal water recirculation. Worthington (2010) also interpreted the effect of elasticity in this manner; he highlighted that due to the output elasticity of several industrial users in various sectors, water use will likely to increase more than proportionately with increasing output. Thus, this method is more suitable in assessing the impact of water price change on the efficiency of water use; as emphasised by Worthington and Hoffman (2008).

On the other hand, the use of cost function also poses a challenge. In order to estimate the water demand for each firm, input data of a company’s yield, capital, labour, materials, and energy will be required. However, these data have been typically aggregated at a national level, as it is difficult to collect at a finer resolution (Reynaud, 2003) due to the variation in

water usage, which has been linked to the level of technology and water recycling used (De Gispert, 2004).

As such, the use of econometrics for the purpose of this research specifically is a conundrum. On the one hand, it is able to provide a stable, long-term forecast for industrial water demand by assessing the interaction between output, capital, labour, materials, and energy consumption at an average value. On the other hand, in order to assess water demand at a firm/factory level (operational level), it would require substantial financial and human resources to collect and manage the database. Furthermore, data that are available tend to be aggregated at national or state level to ensure data anonymity, which pose an even greater challenge in conducting an assessment at an operational level.

3.3.1.2 Unit water demand

Due to the limitation of the econometric approach to estimate industrial water consumption from an operational level, another approach of modelling industrial water consumption was considered. As highlighted in subsection 3.3.1.1, the strength in the use of econometric models lies in their ability to predict changes in water demand relative to changes in water price. Thus, enabling the estimation of industrial water demand at a national level or state level, as well as evaluating the impact assessment of possible changes in demand due to change in water tariff. However, for this research specifically, there is a need to assess water demand at a premise level, in order to further estimate the feasibility of a decentralised WWTP as a means for a water minimization strategy. As such, the use of spatial based information is crucial, as it has been known to provide information regarding the variation in water consumption within and between various users of potable water, across a geographic focus area (House-Peters and Chang, 2011).

Due to this, a multidisciplinary approach was considered; one which integrates the use of data from multiple resources by linking spatial-based datasets (i.e.: land-use classification, customer geo-location) with other non-spatial datasets (e.g. production and socio- demographic) of users. For the industrial sector specifically, linkages of non-spatial datasets regarding industrial water consumption have been joined together using water bills and the Standard Industry Classification (SIC) code for each premises as means to cluster the range in which water is consumed. This dataset is then linked to a spatial-based dataset such as land use information using the parcel or lot code obtained from the town planners, or geocoded

address generated by the water operators themselves (Holmes et al, 2005). For example, Dziedzic et al. (2015) incorporated water billing records with land use and demographic data to assess water consumption trends in Ontario, Canada., while Mamade et al. (2014) assessed average demand pattern using water billing data at a district metring area level and various socio-demographic data (i.e. building, dwelling, and family size) in Portugal.

There is a possibility to integrate the use of econometric models with the multidisciplinary approach described above to estimate a firm’s water demand. However, data requirements to execute the econometric model are intensive and not readily available; this has been reflected in the scarce body of literature dedicated to estimating industrial water demand (Reynaud, 2003). From this perspective, the use of unit water demand analysis was used instead. Donkor et al., (2014) highlighted that, despite the growing number of publications emphasising on more sophisticated modelling approaches such as univariate time series analysis and artificial neural network to estimate urban water demand, in practice, utility managers and town planners often favour the use of this method to estimate overall water consumption within a customer category (i.e. residential, non-residential). This has predominantly been due to the extensive input variables required by other modelling approaches, ones that are neither collected, monitored, used, nor made available to both parties at a finer resolution. Accordingly, unit water demand analysis has been defined as the product of a unit-use coefficient of a customer category (coefficient) multiplied by the population of the category, using the following mathematical expression (Eq. 3.4) (Billings and Jones, 2011):

푄푦 = 푁푦 ∗ 푞푦 Eq. 3.4

Where:

푄푦 – Total system water use in time period y, unitless

푁푦 – Population per category in the water system service area in time period y, unitless

푞푦 – Unit-use coefficient of a customer category in the water system service area in time period y, unitless

The use of number of employees as the unit-use coefficient has typically been used to estimate industrial water demand. This can be linked back to the econometric modelling where it has been shown that the number of employees is a substitute for water demand, where a higher water demand with larger a number of employees have been observed (Duport

and Renzetti, 2001; Reynaud, 2003; De Gispert, 2004). Furthermore, it has been highlighted that that a larger employment requirement tend to reflect a larger scale of production, leading to a higher requirement for water (Dziegielewski et al., 2000; Kieger, Krentz and Dziegielewski, 2015). Gleick et al. (2003) calculated a water-use coefficient of gallon per employee used per day to estimate the total water use for industry in California. Heberger, Donnelly and Cooley (2016) also underlined the use of the number of employees for industrial water demand forecast, while Kowalski et al. (2011) has utilised this method to estimate the volume of freshwater abstracted for several manufacturing processes in the United Kingdom. However, although information regarding the number of employees has been reported to the government for monitoring purposes, the release of their statistics has typically been aggregated at a state or national level.

Another method that has recently been explored is the calculation of the water-use coefficient via physical representation of a premise; specifically using the building footprint. The rationale behind this branch of research is to link water use intensity with a physical structure of a firm which is able to be measured using remotely-sensed data. Here, an indirect correlation between number of employees and physical space required for a firm has been reported, where it was underlined that larger employment has been associated with a larger floor area, which have been known to increase the need for water-consuming equipment such as cooling towers (Pryor, 1972; Dziegielewski et al., 2000). Dziedzic et al. (2015) also indicated a medium to strong correlation (0.5 to 0.99) between water consumption and building space, property area, and unit count for the industrial, commercial, and institutional (ICI) sectors in Canada. In a series of published work Morales, Martin and Heaney (2009), Morales and Heaney (2010, 2014), and Morales et al. (2011) illustrated a methodology to estimate ICI sectors water demand based on heated area of a premise using publically available data in Florida.

Comparing between the use of the two coefficients, Kieger, Krentz and Dziegielewski (2015) highlighted that the choice of employment or building footprint as a coefficient for unit water demand calculation depends highly on data availability. They demonstrated the use of both variables to estimate ICI water demand for several states in U.S.A. and highlighted that both means were equally effective, although the preferred variable is still a direct measure of function (i.e. production yield). However, the limitation to its use has been highlighted, where it is only able to account for water-use intensity, as oppose to variation in water use (Dziedzic et al., 2015). As such, it is not able to account for impact of technological evolution, change

in efficiency, or a change in water tariff (Hagen et al., 2005), unless the coefficient is recalculated using an updated dataset. In addition, while this method has been preferred by the water operators and town planners, it has been noted that minimal work has been done to assess for its accuracy to predict industrial water demand (Donkor et al., 2014).

3.4. Methodology

In order to estimate the impact of any potential savings from industrial premises onto a corresponding WSS, the location of each premise within the study area was segregated to each service area. Here, the definition of WSS is further expanded to include service areas, connected via a potable water distribution system; a complete diagram of the WSS definition used in this study is illustrated in Figure 3-1 below.

Figure 3-1: Diagram of a complete WSS As emphasised in the previous part, three WSSs were used as case studies, where the breakdown of the reservoir, WTP, and service area which makes up each WSS is summarised in Table 3-1. An overview of the overall methodology for Chapter 3 is illustrated in Figure 3-2, where it will be further elaborated in the following subsections.

Table 3-1: Overview of water supply system

Reservoir Water treatment plant Service area Semenyih Sungai Semenyih Hulu Langat Sepang Petaling

Kuala Langat Klang Gates Bukit Nenas Bukit Nenas Wangsa Maju Wangsa Maju Langat Hulu Langat (Phase 1 & 2) Petaling Jaya Cheras, Batu 11 Cheras Kuala Lumpur

Figure 3-2: Overall methodology for Chapter 3

3.4.1. Industrial park water balance 3.4.1.1. Water demand estimation (inflow) - estimation of total potable water consumption for industrial premises

In order to estimate the total potable water consumption for each industrial premise within a service area, the coefficient method based on water intensity was used. This was primarily due to data constraints, as data required to be input into an econometric model or the coefficient method using number of employees was not made available at per premise level. An overall procedure for this subsection of the methodology is illustrated in Figure 3-3 below.

Figure 3-3: Estimation of total potable water consumption for industrial premises

Estimation of potable water consumption per premise in each service area was done primarily using data from two sources; water operator (Syarikat Air Selangor, AIS), and the town planners (Selangor Town and Country Planning Department, JPBD). Upon retrieval of the data from both sources, several limitations of the data were revealed, namely:

- AIS: data regarding water user accounts were only partially geocoded (i.e.: only 80% of their total accounts had information regarding their coordinates); water users were only segregated using basic tariff classification (‘Residential’ and ‘Industrial’) and no further detailed information regarding type of industrial activities carried out at premises categorized as industrial tariff was available; only one year of billing information was accessible due to an outdated database management system; and the spatial-based water user accounts data (AIS-GIS database) and the non-spatial based information for each accounts (Oracle database) were not linked with one another - JPBD: Land parcel data (JPBD-GIS database) for industry was only segregated by type of industry (‘Light’, ‘Medium’, and ‘Heavy’), and no structural information was available for each premises (e.g.: building footprint, year built, property code, etc.)

Due to the disaggregation of the various data sources, several steps were taken to link the three databases together in order to estimate premise level water consumption. Here, the challenges were twofold; first was to match partially geocoded spatial water user account data (AIS-GIS database) to the billing information (Oracle database) provided by the water operator, and second was to join industrial water users information received from AIS to that of the industrial land use information obtained from JPBD. The procedure regarding data treatment for the AIS and JPBD datasets is further described below.

i. Data treatment procedure for AIS dataset Data from AIS were received in two formats, where the billing information (Oracle database) for all accounts with tariff as ‘industry’ were given in Excel format (*.xlxs), while the spatial-based data (consumer points, CP) regarding the location of each accounts were given in ArcGIS shapefile format (*.shp). In order to join the spatial-based information with that of the billing information, coordinates for each CPs were extracted and added into the attribute table in the GIS file, where it was then exported into Excel file. Primary ID linking these two datasets were the account number; as such, joining of the two datasets for all CPs was done by cross- referencing the billing information dataset with the spatial information dataset, using

the account number as the linking variable. Finally, CPs containing data from both databases were re-projected into ArcGIS 10.5 software, to generate a point-based location of each industrial premise.

ii. Data treatment procedure for JPBD dataset Data received from JPBD consisted of current land use information regarding each land parcel in Selangor, where all land parcels which were classified as ‘Industrial’ were extracted from the overall database. Then, the extracted land parcel polygons were overlayed with the service area polygon provided by AIS, to determine the parcels which fall within the study area. These polygons were then overlayed with high resolution satellite imagery that was part of the ArcGIS package in order to visually capture the buildings residing on each land parcel. In order to eliminate the possibility of artifacts, the base year used for the image was consistent as the year in which the data collection was conducted (2014), as well that the image was accurately georeferenced (RMSE < 4 m). Finally, each land parcel polygon was resized according to the premise or premises located within the boundary.

iii. Merging of AIS and JBD database The CP points were then overlayed with the resized land parcel polygons in order to combine both attribute tables into a single table. This was done by merging the two attribute tables using the ‘spatial join’ function in ArcGIS software. Attributes for each CP were only merged to the corresponding land parcel if it was located completely within the polygon. Furthermore, the use of ‘many-to-one’ rule was used to merge multiple CPs which was contained within a single polygon; this was a common occurrence for premises with multiple water meters, either due to segregated billing records according to production line, or due to multiple firms or companies operating within a single premise. Finally, the combined GIS attribute table were exported to excel format for further analysis.

Based on the above steps, two sets of data were generated; premises with known and unknown water consumption information. In order to estimate the water consumption for premises without billing records, a water intensity coefficient was used, generated based on all premises with billing records. Here, the methodology for the estimation of the water intensity coefficient adapted those developed by Morales, Martin and Heaney (2009), and

Kieger, Krentz and Dziegielewski (2015). Furthermore, as highlighted in the literature review section, there has been no research assessing the accuracy of the use of this method to predict industrial water consumption; as such, an accuracy assessment of the use of the coefficient was done in order to determine the limitation of this method using the steps underlined by Donkor et al. (2014).

The entire population of the premises were categorised based on the industrial classification class/type determined by JPBD; either light, medium, or heavy industry. Furthermore, for each industrial type, data were further divided randomly into three sets, where 50% of the samples were used to estimate the average coefficient (set A), 40% were used for per premise validation of the estimated coefficient (set B), and 10% were used to estimate the accuracy of the coefficient to predict water consumption for an industrial park (set C).

3.4.1.2. Industrial park clustering

In order to determine existing industrial parks residing in each service area (SA), clustering of individual premises was done using data obtained from AIS. Identification of potential industrial parks was done based on the attributes of known premises, where keywords relating to the term ‘industrial park’ in both English and the native language (Bahasa) were used to isolate relevant premises. Then, buffer analysis was done based on the identified premises in order to find the natural division between clusters. The buffer distance utilised was a maximum of 30 meters from each premises; this distance was chosen based on the vegetation buffer zone required to be established around existing industrial parks, as mandated by Selangor town planners for heavy industry (Jabatan Perancangan Bandar dan Desa, 2016) as illustrated in Figure 3-4.

Figure 3-4: Mandatory vegetation buffer zone for industrial park

Source: Jabatan Perancangan Bandar dan Desa, 2016

3.4.1.3. Wastewater effluent estimation (outflow)

Data regarding actual effluent released by specific industrial premises were not available as they were classified as confidential by Department of Environment, Malaysia (DoE). As such, a range of effluent discharged volume was estimated from water demand, whereby it was calculated as a percentage of 30% to 90% of water demand; this range was identified during the interview sessions with various industry players, where the lower and upper range was identified as the smallest and largest volume of effluent discharge by the interviewed companies, as appended in Appendix G. Furthermore, the 15-year projection of water demand was also simulated of a 2% per annum increment from base year (2016), as identified by the water operators.

3.4.2. Projected saving at dam and potable water treatment plant

Calculations of the projected for non-potable water savings per year were done based on the estimated wastewater effluent generated as outlined in section 3.4.1.3 above, where it was assumed that the wastewater treatment facility had a 95% efficiency to treat industrial effluent wastewater for non-potable water reuse. Similarly, water savings for potable water reuse were also calculated using the same assumption, with an additional subtraction of blending ratio of 2%, 15%, and 35% with potable water as currently utilised by Singapore, Texas, and Namibia respectively (Rodriguez et al., 2009; Tchobanoglous et al., 2011; Gerrity et al., 2013; Vincent et al., 2014). Equations for water saving estimation for non-potable and potable water are illustrated in Eq. 3.5 and Eq. 3.6 respectively.

표 푊푆푛푝,푛,푝 = 퐷푝−1,푛 + (퐷푝−1,푛 × 2%) − ∑1(푆푣푔푠푛푝,표,푛,푝 × 95%) Eq. 3.5

Where:

푊푆푛푝,푛,푝 − Total water savings from non-potable (np) reuse for reservoir n in month p; m3/month

3 퐷푝−1,푛 − Total water demand at month p – 1 for reservoir n; m /month

푆푣푔푠푛푝,표,푛,푝 − Water savings from non − potable 푛푝 reuse in industrial park 표 for reservoir 푛 in month 푝; m3/month

표

푊푆푝,푛,푝 = 퐷푝−1,푛 + (퐷푝−1,푛 × 2%) − ∑(푆푣푔푠 푝푡,표,푛,푝 × 95%) − 퐵푙푛푟,표,푝 1

Eq. 3.6

Where:

푊푆푝푡,푛,푝푡 − Total water savings from potable pt reuse for reservoir n in month p; m3/month

3 퐷푝−1,푛 − Total water demand at month p − 1 for reservoir n; m /month

푆푣푔푠푝,표,푛,푝 − Water savings from potable 푝 reuse in industrial park 표 for reservoir 푛 in month 푝; m3/month

퐵푙푛푟,표,푝 − Potable water blending ratio r for industrial park o for reservoir 푛 in month p; m3/month

Additionally, projected savings at the water treatment plant were also calculated, where two scenarios were considered. Presently, the current potable water treatment plants were reported to be overloaded and producing potable water at a higher volume than designed (National Water Service Commission (SPAN), 2016b). Hence, the two scenarios considered were:

S1: A new water treatment plant is built in a location close by, and is abstracting raw water from the same river and dam as the current water treatment plant. As such, all identified industrial parks and premises were assumed to consume potable water from the current water treatment plant as opposed to the new WTP.

S2: Current water treatment plant is upgraded to meet the projected demand.

Water savings calculations at the water treatment plant level were calculated using the same equation as above (Eq. 3.5 for non-potable water reuse and Eq. 3.6 for potable water reuse).

3.5. Results and discussion 3.5.1. Industrial park clustering and water consumption estimation

With regards to the identified IPs within the service areas, several observations were noted. One of the observations was that, although there were some industrial premises clustered

together in pattern similar to that of an industrial park, upon closer inspection, these premises were identified to be shop lots, used for commercial activities such as offices and automotive workshops; these premises were not included in the analysis, primarily due to the wastewater collection system installed. For manufacturing premises specifically (i.e. factories), treated wastewater from the compound is discharged via open drainage and into inland waterways, hence emphasizing the importance of the industrial premises to comply with the limit set in the Fifth, Seventh, and Eighth Schedules in the Environmental Quality Act 1974. Conversely, for premises not classified as industrial compounds such as those intended for commercial and service-based activities, all wastewater is routed into the closed domestic wastewater treatment system; as such, wastewater generated from these compounds are not able to be rerouted into a decentralised wastewater system as proposed by this research.

Additionally, it was also observed that there were multiple known industrial premises with large water consumption situated outside of any IP; this could be due to several reasons, such as that the establishment of these industrial premises preceded the formation of a nearby IP, or that the industrial premises were intentionally located next to a natural body of water (i.e. river) for raw water abstraction for private consumption due to a high volume of water required for their operation. In this case, although the combined savings from these premises could contribute further to the overall water savings from the industrial sector as a whole in addition to those saved via the IPs, efforts to achieve these saving can only be realised through efforts made by specific companies rather than a combined effort achieved via a decentralised wastewater reuse treatment plant. Furthermore, the mechanism required to realise the savings from individual premises differs slightly from those for the latter strategy, such as the financial aid or incentives required to offset capital investment made by individual companies, as compared to incentives for companies in the IPs to procure recycled wastewater via the decentralised wastewater reuse facilities. As the focus of this research is more on the former objective, emphasis was given to the identification of IPs and the corresponding water consumption and potential saving from these clusters of industrial premises instead of individual premises.

Different characteristics of the premises found within various parks could hint of different means of IP formation and management. The more uniform sizes and systematic building configuration of the premises in some of the parks such as CH_IP-2 and HL_IP-1 suggest that their development was more formal in nature, where the creation of the IP was previously planned and subsequently executed. Furthermore, this could suggest the existence

of a private entity managing the park. Conversely, a more freeform configuration as those found in PJ_IP-3 and CH_IP-7 could point to the later establishment of an ‘industrial park’ due to the approximation of several factories to one another, or a segregated space allocated for industrial premises specifically. For the latter, it could indicate the possibility to establish a similar type of industrial park for premises which are relatively close together, but are not currently labelled as one, and could be under the management of local government. An example of the differences between these two configurations can be seen in Figure 3-5 below, where the cluster on the left illustrates a more formal IP (CH_IP-2), while the cluster on the right illustrates a more freeform IP (CH_IP-7).

Figure 3-5: Example of industrial premises clusters for a formal IP configuration (left, CH_IP-2) and freeform IP configuration (right, CH_IP7)

Clustering of industrial premises in all three service areas had yielded 26 unique industrial parks spread over two specific service areas. Interestingly, although Klang Gates dam had indicated a sensitive system which has a high potential in reducing water shortage risk by means of DSM-based strategy, no industrial park which could positively assist in reducing the overall water consumption within the system could be identified in the corresponding service areas (Bukit Nenas and Wangsa Maju). Assessing the current land use pattern in both

service areas for the Klang Gates WSS, a majority of land use was classified as residential and commercial areas (e.g. shopping complexes, offices, and hotels) as compared to industrial premises. Although there were several industrial premises identified within the two service areas, they were not however situated within a formal industrial park, and hence, outside the scope of this research due to the rationales outlined in subsection 3.5.1. As such, for this water supply system specifically, although a DSM-based strategy could aid in reducing the risk of non-satisfactory conditions in the WSS, targeting the industrial sector – specifically manufacturing – would not significantly change the current and future water resource challenges faced in this system.

As for both Langat and Semenyih dams, the number of industrial parks identified was higher in the latter system as compared to the former, although the largest industrial park identified was in the former system; in total, 8 industrial parks were identified in the Langat WSS, while 18 industrial parks were isolated in the Semenyih WSS. Overall, most of the industrial premises in each park had more premises with known water consumption as compared to the unknown; approximately 9 of the 26 parks had less than 50% unknown premises, with only 3 of the 9 parks having more than 100 unknown premises within the IP; this indicates that the estimation of water consumption for majority of the IPs are actual water consumption. For the unknown premises, it was found that the hierarchical clustering (HC) method resulted in an overestimation of approximately 19±41% using the HC method when applied to premises in an IP. Although the use of the estimated method had overestimated the water consumption, the majority of the premises in an IP were of actual consumption, thus providing a relatively accurate snapshot of the total water consumption within a service area. A detailed assessment of the method used to estimate the unknown water consumption of a premise is appended in Appendix E.

For the Langat system specifically, it can be observed that the highest estimated water consumption was identified for CH_IP-6 (n = 1,469) in the Cheras service area of approximately 165,000 m3/month (5,500 m3/day) of water consumed, while the lowest was for CH_IP-1 (n = 70) in the same service area of approximately 950 m3/month (32 m3/day). On the other hand, the water consumption of the IPs identified within the Semenyih dam were somewhat consistent, where majority of the IPs had showed an estimated water consumption of above 10,000 m3/month (333 m3/day); however, the range in water consumption was large, with the highest consumption of approximately 73,740 m3/month (2,458 m3/day, PJ_IP-8, n = 709) and lowest of 1,070 m3/month (36 m3/day, PJ_IP-3, n = 7).

A summary of the identified IPs as well as the estimated water consumption per industrial park was illustrated in Table 3-2 below.

Comparing between the two IPs with the smallest water consumption, it was further observed that premises in PJ_IP-3 consumed a higher volume of potable water as compared to CH_IP- 1, despite the number of premises in the latter IP being 10 times higher than the former. It was further observed that each of the known premises in PJ_IP-3 consumed water at a higher level than those in CH_IP-1, suggesting a vastly different types of activities carried out within these premises, although all the premises were similarly classified as medium industries. This reflect the observations found in the literature, which emphasised the large variance in industrial water consumption, depending on industry type, water use practices, and activities conducted within each premises (e.g. Gleick et al., 2003; De Gispert, 2004). This, in turn, points to a large uncertainty regarding water consumption estimation for the unknown premises due to the “black box” assumption concerning the activities conducted within these premises; to account for this, an upper and lower boundary was also estimated alongside the average.

Table 3-2: Characteristic and potable water consumption for individual industrial park

- -

n n

/month /month /month /month /month

3 3 3 3 3

monthly monthly monthly

premises premises premises

19.95%), 19.95%), 41.55%),

Reservoir

% known known %

Unknow

m m m m m

Ave actual actual Ave

(+41.55%), (+41.55%),

upper limit limit upper

Est average Estaverage

(multiplier), (multiplier),

% unknown unknown %

consumption consumption

lower limit ( limit lower

consumption, consumption,

Total adjusted Totaladjusted

consumption ( consumption

Total, adjusted, adjusted, Total, adjusted, Total,

Industrial park, park, Industrial

Service area, SA area, Service

Known premises Known average monthly monthly average

CH_IP-1 70 0 100 0 954 - 954 954 954 CH_IP-2 8 47 15 85 1,228 5,478 5,614 3,792 7,436

CH_IP-3 196 130 60 40 10,240 44,059 45,509 30,855 60,163

Cheras CH_IP-4 244 0 100 0 73,082 - 73,082 73,082 73,082

CH_IP-5 415 95 81 19 12,614 3,701 15,577 14,346 16,808 Langat CH_IP-6 1067 402 73 27 108,240 70,339 164,547 141,151 187,942 CH_IP-7 0 20 0 100 - 2,192 1,755 1,026 2,484 Kuala Lumpur KL_IP-1 75 13 85 15 3,830 910 4,574 4,265 4,883 HL_IP-1 0 258 0 100 - 51,068 40,880 23,894 57,865 HL_IP-2 114 93 55 45 122 31,731 25,523 14,969 36,076 Hulu Langat HL_IP-3 639 195 77 23 38,632 12,325 48,499 44,399 52,598 HL_IP-4 77 436 15 85 2,516 28,180 25,074 15,701 34,446

HL_IP-5 0 132 0 100 - 83,093 66,516 38,879 94,153

PJ_IP-1 359 121 75 25 15,908 6,118 20,806 18,771 22,841 PJ_IP-2 198 75 72 28 12,078 14,232 23,471 18,738 28,205

PJ_IP-3 3 4 43 57 479 734 1,066 822. 1,310 Semenyih PJ_IP-4 31 8 80 21 2,029 1,479 3,213 2,721 3,704 Petaling PJ_IP-5 129 39 77 23 4,544 3,664 7,478 6,259 8,697 PJ_IP-6 526 104 84 16 12,297 4,623 15,998 14,461 17,536 PJ_IP-7 404 150 73 27 27,188 9,938 35,144 31,838 38,449 PJ_IP-8 442 267 62 38 23,166 63,173 73,737 52,725 94,749

- -

/month /month /month /month /month

3 3 3 3 3

monthly monthly monthly

premises premises premises

19.95%), 19.95%), 41.55%),

Reservoir

% known known %

Unknown Unknown

m m m m m

, adjusted, adjusted, ,

Ave actual actual Ave

(+41.55%), (+41.55%),

upper limit limit upper

Est average Estaverage

(multiplier), (multiplier),

% unknown unknown %

consumption consumption

lower limit ( limit lower

consumption, consumption,

Total adjusted Totaladjusted

consumption ( consumption

Total adjusted, Total,

Industrial park, park, Industrial

Service area, area, Service

Known premises Known average monthly monthly average

SE_IP-1 0 12 0 100 - 13,568 10,861 6,348 15,374

SE_IP-2 13 15 46 54 1,959 3,292 4,595 3,500 5,690 Sepang SE_IP-3 56 11 84 16 2,590 362 2,879 2,759 3,000

SE_IP-4 97 282 26 74 3,782 17,412 17,721 11,929 23,512 Semenyih SE_IP-5 193 49 80 20 10,717 4,073 13,978 12,623 15,332

3.5.2. Water reuse potential at industrial park

Each IP was first assessed individually and detailed results are appended in Table F-1 Appendix F for potable water consumption estimation. The results suggest that there were several IPs which had small water demands, translating to even less industrial effluent which can be reused. In total, at 90% wastewater discharge, 7 of the 26 IPs had a daily total wastewater discharge of below 170 m3/day; the 170 m3/day minimum discharge was as reported by Piadeh, Alavi-moghaddam and Mardan (2018). It could be seen that the lowest flow for Langat WSS was 27.2 m3/day (CH_IP-1), while for Semenyih WSS, 30.4 m3/day (PJ_IP-3). However, this number had decreased to 5 IPs after the 15-year projection (CH_IP- 1, CH_IP-7, PJ_IP-3, PJ_IP-4, and SE_IP-3). For the premises of less than 170 m3/day of wastewater discharge, it could be hypothesised that the activities within these IP are those with small water consumption such as packaging or assembling. Thus, although there might be a slight increase in water demand in 15 years, the projected wastewater discharge would still be minimal; as such, the construction of a decentralised wastewater reuse facility would not be feasible as the cost of establishing the treatment plant might outweigh the benefit. Thus they were omitted from the total water savings estimation, highlighted in red in the subsequent results presented in Table F-1 in Appendix F. On the other hand, the volumes of wastewater effluent in all of the other IPs were between 300 – 2000 m3/day (baseline), which then grew to approximately 400 – 6,000 m3/day at the 15th year mark. This suggest that, at an IP level, the majority of the IPs in both WSS have the potential to sustain the deployment of the proposed wastewater reuse facilities, in terms of influent into the system, as well as the corresponding treated output, which can be sold back to the premises in the IPs.

As a whole, total potable water consumed within the Semenyih WSS by the IPs was slightly higher than in the Langat WSS, with the former consuming approximately 437,440 m3/month (14,581 m3/day) and the latter 311,610 m3/month (10,387 m3/day). This, in turn, accounted for 34.3% and 15.6% of the non-domestic raw water withdrawal from the Semenyih and Langat dam respectively. Based on this, the projected 15-year potable water consumption for all IPs at a rate of 2% increase per annum was also calculated, resulting in a total of 579,100 m3/month (19,303 m3/day) for Semenyih dam, and 415,741 m3/month (13,858 m3/day) for Langat dam. The adjusted, lower, and upper estimation for both the baseline year (2016) and the 15th year projection is summarised in Table 3-3 below.

Table 3-3: Baseline and 15th year projection for cumulative water consumption by all IPs in Langat and Semenyih WSS

Baseline (m3/month) Y15 (m3/month) Adjusted Lower Upper Adjusted Lower Upper Langat 311,610 269,469 353,752 415,741 362,670 476,104 Semenyih 437,440 321,335 553,539 579,100 432,475 744,991

Table 3-4: Total estimated baseline and 15th year projected water saving from non-potable reuse for Semenyih and Langat WSS

Baseline, Adjusted (Non-potable) Projected Y15, Adjusted (Non-potable) (m3/month) (m3/month) WSS Effluent, Effluent, Effluent, Effluent, Effluent, Effluent, 30% 60% 90% 30% 60% 90% Langat 88,037 176,075 264,111 118,486 236,972 355,459 Semenyih 122,629 245,259 367,888 165,042 330,087 495,128

As previously mentioned in the methodology section, the wastewater effluent in each IP was estimated based on a range of probable discharge volumes of all premises as a whole; Table F-1 in Appendix F summarises the effluent discharge for 30%, 60%, and 90% from the total water demand for individual IPs, while Table 3-4 above summarises the total effluent discharge for the two WSSs for non-potable reuse. For both non-potable and potable water reuse, it could be observed that it would be more beneficial for each IP if 90% of wastewater effluent is discharged into the open water without any internal reuse, as it would provide the maximum effluent volume into the decentralised wastewater treatment plant (Table 3-4). Overall, both the Langat and Semenyih WSS could potentially save approximately 264,111 m3/month (8,880 m3/day) and 367,888 m3/month (12,263 m3/day) of potable water respectively if all of the identified IPs collectively implement non-potable water reuse, assuming the reuse of 90% of wastewater effluent at 95% efficiency. In contrast, approximately 88,037 m3/month (2,935 m3/day) and 122,629 m3/month (4,088 m3/day) of water saving could be expected for Langat and Semenyih respectively if only 30% of the wastewater was assumed to be discarded and reuse at 95% efficiency.

Table 3-5: Total estimated baseline saving from potable reuse for Semenyih and Langat WSS

Savings - Baseline, Adjusted (Potable) (m3/month)

Effluent, 30% Effluent, 60% Effluent, 90%

WSS

nding

Blending 2/98 , Blending 15/85 , Blending 35/65 , Blending 2/98 , Blending 15/85 , Blending 35/65 , Blending 2/98 , Blending 15/85 , Ble 35/65 , Langat 2,641 13,206 30,813 5,282 26,411 61,626 7,923 39,617 92,439 Semenyih 3,679 18,394 42,920 7,358 36,789 85,841 11,037 55,183 128,761

Table 3-6: Total estimated 15th year projected saving from potable water reuse for Semenyih and Langat WSS

Savings - Y15, Adjusted (Potable) (m3/month)

Effluent, 30% Effluent, 60% Effluent, 90%

WSS

2/98 2/98 2/

15/85 35/65 15/85 35/65 15/85 35/65

Blending, Blending, Blending, Blending, Blending, Blending, Blending, Blending, Blending, Langat 3,555 17,773 41,470 7,109 35,546 82,940 10,664 53,319 124,411 Semenyih 4,951 24,756 57,765 9,903 49,513 115,530 14,854 74,269 173,295

Conversely, the potential savings realised from potable water reuse is substantially smaller than non-potable reuse due to the need for blending with other water sources; Table 3-5 and Table 3-6 above summarise the calculated savings for the baseline and the 15-year projection for potable water reuse. Taking the maximum available effluent which could be generated for each IP (90% effluent discharge from estimated consumption), only a maximum of 35% of the treated wastewater could be successively blended with other water sources – in this case, potable water. In this instance, saving calculated for Langat and Semenyih WSS was 92,439 m3/month and 128,761 m3/month respectively for the base year. Here, it could be seen that the cumulated water savings for Langat WSS was similar to those savings generated by the non-potable water reuse at 30% effluent discharge, while for Semenyih, it was equivalent to savings at 60% effluent discharge. The higher estimated volume for Semenyih was due to the number of IPs isolated in this WSS; there were twice as many IPs in Semenyih as compared to the Langat WSS. However, for the other two blending options, the accumulated savings were minimal, especially for 2/98 blending ratio; even at the maximum generated effluent of 90%, the savings had only accumulated to 7,923 m3/month and 11,037 m3/month for Langat and Semenyih WSS respectively. Thus, it can be concluded that the use of the 35/65 blending ratio was preferable as it was able to offer the maximum potential savings for both WSS.

3.5.3. Impact of water saving on water supply system

Although substantial savings can be done at an IP if 90% of the effluent is discharged and subsequently reused, the effect of savings at a WSS level was observed to be minimal, as illustrated in Table F-4 to F-9 in Appendix F. Estimation was further done through the lenses of both the dams as well as the water treatment plant; the latter was due to the raw water supplied to the potable water treatment plant only accounting for 40% and 16% of the total water intake to the WTP for Langat and Semenyih WTP respectively.

Overall, the estimated water savings from all IPs at a dam level were calculated to accumulate to 5.5% (8.8 MLD) and 12.2% (12.47 MLD) for Langat and Semenyih dam respectively for non-potable water reuse at the baseline year (2016), while the projected water savings had also resulted in minimal savings of only 0.7% and 0.4% increase by the 15th year mark. Conversely, due to the blending with treated water, potable reuse had resulted in savings of approximately 1.9% and 7.8% were seen for the Langat and Semenyih WSS respectively for 35% blending of 90% water effluent discharge, with an increase of 0.07% and 0.27% by the 15th year.

The estimated savings impacted the two dams differently. In the previous chapter, it was illustrated that both DSM and SSM strategies could aid in reducing Langat dam’s vulnerability, while an effective DSM-based strategy could prolong the need for SSM-based strategy for Semenyih dam. The IDR and WSR scores for both dams were recalculated for the base year (2016), taking into account the potential savings appended in subsection 3.5.2. For Langat dam, it could be observed that the expected savings produced minimal impact on decreasing its vulnerability as illustrated in Figure 3-6 below. Although there was some shifting of the calculated IDR/WSR value, the IDR calculation still resulted in negative readings for the year, signalling that the inflow was still not able to meet the outflow despite the 5.5% reduction in non-domestic consumption. Additionally, the WSR readings primarily generated negative values, further indicating that the dam is no longer able to provide adequate storage. Thus, the combination of both IDR and WSR here suggests that the situation had past the point of this particular DSM-based strategy having any impact on decreasing the dam’s vulnerability. This, in turn, points to two other options; either implement an SSM-based strategy such as increasing the dam’s storage capacity or increasing the inflow via inter basin water transfer, or couple this specific DSM-strategy with those targeted at other sectors – namely, domestic, commercial, and service sectors.

0.3

0.2

0.1

0 -1.4 -1.2 -1 -0.8 -0.6 -0.4 -0.2 0 -0.1

-0.2

-0.3

Water storage resillience, WSR -0.4

-0.5

-0.6 Inflow-demand reliability, IDR

Without savings With savings

Figure 3-6: Comparison of IDR/WSR values for Langat dam at base year (2016)

1.4

1.2

0.8

0.6

0.4

0.2

Water storage resillience, WSR 0 -3 -2.5 -2 -1.5 -1 -0.5 0 -0.2

-0.4 Inflow-demand reliability, IDR

Without savings With savings

Figure 3-7: Comparison of IDR/WSR values for Semenyih dam at base year (2016)

Conversely, for Semenyih dam, the savings further solidify the dam’s ability to provide satisfactory storage for this WSS, as illustrated in Figure 3-7 above. It could be observed that the potential savings of 11.1% from the IPs further increase the WSR values, where two of the previously negative readings resulted in positive values after accounting for the savings. Although the IDR value also showed some changes, all of the readings still resulted in primarily negative values, indicating that the demand still outweighs the supply. In this case, if this DSM-based strategy was coupled with other strategies targeted at other sectors, it could further increase the system’s resilience and, to a degree, its reliability, and thus, decrease its vulnerability. In the long run, this could eliminate the need for a SSM-based strategy.

From the perspective of the WTP however, the savings expected were significantly lower, as appended in Table F-4 to Table F-9 in Appendix F; savings via non-potable reuse could only ‘free up’ approximately 2% of the produced potable water for other users. This provided an additional extension for one year for both WTP or an additional 38,445 and 53,970 people in the Langat and Semenyih service areas respectively, if the demand grows at a rate of 2% per annum and at a constant per capita consumption of 231 litre per capita per day (The Malaysian Water Association (MWA), 2015). For the 15th year projection, percentage saving for both water supply systems did not vary greatly, where for the Langat WSS, savings of 3% and 2.3% were observed for S1 and S2 respectively, while for Semenyih WSS, savings of 2.3% and 1.8% were observed. Translated, this had allowed the accumulated savings to serve an additional 13,298 and 16,922 people in the Langat and Semenyih WSS.

Additionally, similar to the situation observed for the dams, the water savings generated from potable water reuse also indicated a small saving of 0.78% and 1.21%, allowing to serve 13,456 and 34,248 people for Langat and Semenyih WSS. As for the 15th year projection for both scenarios, a slight increase in savings was noted, where it increased to 1.04% (S1) and 0.8% (S2) for Langat WSS, and to 1.62% (S1) and 1.25% (S2) for Semenyih WSS, allowing for a total of 18,110 and 46,093 additional people to be served by the two WSS.

The above results further underline the benefit of non-potable water reuse as a water minimization strategy for the manufacturing sector. Although the savings were initially estimated to be able to provide an additional 38,445 and 53,970 people in the Langat and Semenyih service areas respectively, they could be further increased if the per capita water consumption for the state was reduced. In the event where it was reduced to a conservative

estimation of 180 l/cap.day, this number would then be increased to 49,338 and 69,262 people in Langat and Semenyih service area respectively for the baseline scenario.

3.6. Conclusion

The impact of water minimization of all industrial parks onto the Klang Gates, Langat, and Semenyih water supply system was assessed. In this chapter, the research introduced a new means to incorporate the use of land use information from the local town planners with that of metering information from the water operators to identify clusters of industrial parks in the study area. Identification of industrial parks within all service areas was done using a spatial clustering method; in total, 26 isolated industrial parks were identified, although it was further found that no IPs resides in the Klang Gates WSS. Due to this, it can be established that, although Klang Gates WSS could benefit in the implementation of DSM-based strategies, those targeted at industrial sector would result in no change in the WSS vulnerability against non-satisfactory events.

This research had estimated the volume of water demand from each industrial park within the service area for all three dams via a combined use of 1-year billing information from known premises as well as the use of water coefficient for premises with unknown water consumption. The incorporation of spatial-based information for all the identified industrial premises as well as industrial parks introduced in this study had enabled this research to measure the impact of potential savings from industrial wastewater reuse to a specific WSS. It was found that for Langat and Semenyih WSS, most of the identified IPs have a high potential for generated wastewater discharge reuse, especially in the case of 90% effluent discharge. Non-potable water reuse resulted in a higher saving than potable water reuse, due to the required blending with potable water for the later reuse option. However, assessing the total savings generated for both reuse options with respect to the overall system, it was found that it resulted in differing impact for both the dams, while resulting in minimal impact at the potable water treatment plant. At dam level, it was found that non-potable reuse had resulted in 5.5% and 12.2% raw water saving for Langat and Semenyih dam respectively. This in turn, did not result in any significant change in vulnerability to Langat dam, while it successfully decreased Semenyih dam’s vulnerability.

On the other hand, an observed value of 2.2% and 1.7% savings were estimated for the Langat and Semenyih WTP respectively, indicating minimal impact; this amount did not

significantly change in the 15th-year projection for both WSS. Water saving for potable water reuse resulted in an even significantly smaller impact; it was found that at 35/65 blending ratio of 90% effluent discharge utilisation, it resulted in a 1.9% and 7.8% saving for Langat and Semenyih dam respectively, while a small saving of approximately 1% was estimated at WTP level for either system.

Re-assessment of the vulnerability of both Langat and Semenyih dam was also done, incorporating the estimated water savings generated from industrial wastewater reuse. In this chapter, both the IDR and WSR values were recalculated for the scenario of 90% wastewater reuse of non-potable water standard in order to determine the impact of the implementation of industrial water reuse as a DSM strategy to reduce water imbalance issues. Overall findings for both systems resulted in an increment of both IDR and WSR values for both WSS. However, it was observed that for the Langat WSS, the estimated savings had resulted in minimal impact on decreasing the overall vulnerability of the system as both the IDR and WSR values were still negative. As such, it was concluded that the implementation of industrial wastewater reuse would not result in significant change to minimise water imbalance issue for the Langat WSS. On the other hand, for the Semenyih WSS, it was found that the resultant recalculation of the IDR and WSR values had indicated a slight decrease in the system vulnerability. This was observed through the shifting of both the IDR and WSR values towards a more positive value for several of the data points in the 2016 baseline scenario. Due to this, it can be concluded that the savings generated from industrial wastewater reuse in the Semenyih WSS could result in reducing the water imbalance issue for this system.

Chapter 4 : Technical, financial, and environmental cost of wastewater reuse options

4.1 Introduction

In Chapter 1, it was argued that the likelihood of adoption of a bottom-up strategy to complement a market-based policy depends on the industrial player’s capability to implement the solution. As a result, the use of decentralised wastewater treatment plants for industrial wastewater reuse was proposed to meet the challenges faced by the players, as presented in Chapter 3. The uptake of this reuse option is still determined by the player’s perspectives; in order to capture this, preliminary semi-structured interviews were conducted with several industrial players. The respondents were asked about their views regarding the government’s strategy to implement DSM for potable water consumption, specifically the impact of restructured water tariff on their business, increasing water efficiency within their operations, and use of alternative water sources for their production2.

Generally, the respondents were aware of the fact that water related issues in the country could be an emerging threat to their business and how they could have a negative impact on their reputation; this was seen as the main driver for them to consider investing in technologies to secure their water supply and adopt sustainable practices for water utilisation. Nevertheless, several challenges were raised by the respondents for the technology implementation, specifically:

- Water quality of alternate water sources

‘Quality of treated wastewater effluent to be used must be similar to tap water quality. For the paint industry, we are concerned about microbiological contamination within the paint, especially mould and fungus. Even though we are currently adding additives to control fungus activities within the wastewater, contaminants can still be present; which is why we are currently not recycling [wastewater effluent]’

Industry player, RI2

2 Details regarding the respondents are presented in Appendix G, including coding system used for identifying different respondents

- Increased cost of production (due to high capital cost and maintenance cost of new technology)

‘Due to the nature of the product, it is not feasible to reuse wastewater as it has high levels of contaminants such as algae and suspended solids and would incur a high cost to treat the wastewater to the same standard as treated water. Treated water available is cheap and our product is very competitive; any additional cost for manufacturing would translate to higher product price, making us unable to compete with other countries such as Indonesia and Thailand’

Industry player, RI5

- Limited space to install new technology on existing premise.

‘[Our] current plant does not have a water recycling system; there are a few factors which need to be taken into consideration, one of which is that we have limited space on the premise. However, another plant we have in Johor Baharu has implemented wastewater recycling, where some of the wastewater discharged is reused for irrigation’

Industry player, RI3

Furthermore, efforts to increase their water use efficiencies are limited to activities which do not interfere with current manufacturing process as they are unwilling to compromise their current system. For example, RI6 highlighted the following statement:

‘The system design is fixed. The system is costing us over RM 8 million over. Once we have really have set up this so called running system, water efficiency will be set at that level continuously until the system retires’

Industry player, RI6

Based on the findings of the interviews, there appears to be a gap between what is strategized by the government, versus what can be implemented by industry players. Although the government has identified restructuring of water tariffs and the use of alternate water sources as key strategies to decrease the industrial sector’s dependencies on potable water, the impact on industry players will primarily be on the cost of production, thus decreasing the likelihood of the implementation of the identified strategies.

From the industrial player’s perspective, securing adequate potable water supply was underlined as the most important factor; it can be seen that several companies have invested a

significant amount of financial resources to reduce the impact of future water rationing on their operations. However, based on the interviews, there was no evidence to indicate their willingness to extend this effort beyond securing their potable water needs; the biggest obstacle remained the low water tariff, where it is seen as the major factor deterring the industry players from adopting advanced technology to harness alternative water resources.

Results from the initial finding suggest that in order for the proposed wastewater reuse facility to be considered by the industry players, it would need to satisfy the following conditions:

i. Alternative water sources would need to be of the same standard as potable water supply ii. Be cost competitive iii. Would need to be readily and easily adapted to their current operational procedure

4.2 Aim and research questions for Chapter 4

The third phase of the research focused on assessing the technical, financial, and environmental feasibility of either non-potable or potable water reuse, while taking into consideration the restriction underlined by the industry players. Therefore, the following objectives were considered:

RQ 6: What is the technical feasibility of identified treatment options in terms of their contaminant removal capabilities to meet non-potable or potable water reuse standard?

RQ 7: What is the corresponding cost of implementation and subsequent unit cost for both reuse options?

RQ 8: What are the environmental impacts of implementing the identified treatment options?

4.3 Literature review 4.3.1 Wastewater reuse design for industrial wastewater effluent 4.3.1.1 Advance wastewater treatment for industrial wastewater

In order to adhere to strict reuse standards, it has been argued that the use of membrane filtration systems – specifically reverse osmosis (RO) and nanofiltration (NF) – has been

preferred (Shannon et al., 2008) as they provide a practical means to remove contaminants to a high degree at a reasonable cost (Asano et al., 2007). Furthermore, they have a higher removal rate for low molecular weight trace pollutants, high quality effluent including low organic concentration, high removal of microbes and viruses without added chemical disinfection, and modularity, which enables their integration with new and existing systems (Liu, Kanjo and Mizutani, 2008; Taheran et al., 2016). Other highlighted advantages of membrane filtration systems as a whole include higher energy efficiency (Zhang et al., 2009), higher capability to produce more consistent permeate water qualities (Bennett, 2005), as well as reduced footprint and sludge production (Lin et al., 2012).

The primary disadvantage of using either membrane technologies remains the cost (Judd, 2016) – specifically during the operation and maintenance (O&M) phase of the system. It has been highlighted that, the O&M cost of a membrane system is higher than the conventional activated sludge process (CASP), primarily due to the high electricity requirement associated with the technology (Cote, Masini and Mourato, 2004; Bolzonella et al., 2010). Furthermore, the O&M cost is also associated with the occurrence of high membrane fouling – defined as accumulation of constituents in the feed stream on the membrane (Tchobanoglous, Butron and Stensel, 2004) – and low permeate flux (Fu and Wang, 2011); as fouling occurs on the membrane surface, higher energy expenditure is required to maintain its efficiency, or more frequent membrane cleaning and replacement is required to maintain high quality permeate (Miller et al., 2014).

Due to this, it has been emphasised that the selection of pre-treatment technologies prior to the use of a membrane system is crucial to reduce the effect of membrane fouling, and to further aid in removal of trace constituents. Furthermore, a disinfection process as an added subsequent unit process is also necessary to reduce the likelihood of exposure to potential waterborne diseases from pathogenic organisms. In terms of the former objective, treatment trains that include the usage of NF/RO have done so with the combination of other types of membrane filtration systems, such as microfiltration (MF), ultrafiltration (UF), and membrane bioreactor (MBR); adsorption using materials such as activated carbon; or oxidisation process, such as photocatalytic, ozone, and ultraviolet (UV) (Gogate and Pandit, 2004; Sipma et al., 2009; Alturki et al., 2010; Fu and Wang, 2011; Katsou, Malamis and Loizidou, 2011)

The use of MBR systems in particular has been seen as an attractive option for NF/RO hybrid systems as they combine elements of a CASP system – specifically suspended biomass – with an immersed MF or UF membrane, eliminating the need for a secondary clarifier normally found in a CASP system (Shannon et al., 2008). Due to this, it has been noted that a MBR system is able to produce a higher effluent quality – where an improved removal of 10 – 15% was observed with a MBR system as compared to CASP (Bolzonella et al., 2010) – with a smaller physical footprint (Nguyen et al., 2013). From the perspective of contamination removal, literature assessing the removal efficiency of a MBR-NF/RO hybrid system has illustrated a wide array of contaminants found in both domestic and industrial wastewater. For example, Alturki et al. (2010) demonstrated a 95% removal below analytical detection limit of 40 trace organic compounds using a MBR-RO system; they highlighted that the use of an MBR system had complemented the RO system by removing hydrophobic and biodegradable trace organic compounds due to the enhanced residence time in the biological reactor. Subsequently, residual hydrophilic trace organic compounds from MBR process were then removed by the RO membrane. However, they had noted that this combination was not as effective for chlorinated organic pesticides. Similar findings were also reported by Nguyen et al. (2013); they noted that removal of trace organic compounds is dependent on both adsorption of the compounds to the suspended solid in the mixed liquor and biological degradation in the reactor, of which, in turn, are dependent on the compound’s molecular structure. Although most trace organic compounds are able to be degraded by biomass in the MBR system, trace organic compounds containing chlorine are particularly resistant to biological degradation, leading to low removal efficiency. Due to this, although the use of the RO system was able to increase the overall removal efficiency from the wastewater, it was also noted that the use of a UV oxidation method would provide a more effective removal.

The high removal rate of heavy metals were also reported in the literature using the MBR system; removal of heavy metal elements such as cadmium (Cd), chromium (Cr), copper (Cu), nickel (Ni), arsenic (As), lead (Pb), and zinc (Zn) have typically been studied; however the removal efficiencies of these elements varies. For example, Bolzonella et al. (2010) reported a low removal of As (31 – 43%), and a low to high removal efficiency of Co (62%), Cu ( 49 – 94%), Ni (48 – 65%), and Pb (16-88%) using an MBR system. Katsou, Malamis and Loizidou (2011) also presented similar findings; without the addition of aluminosilicate mineral used to reduce heavy metal concentrations in the permeate, they reported a low removal for Ni (31 – 58%), medium to high removal efficiency for Cu (71 – 86%) and Zn (66

– 85%), while a high efficiency for Pb (90 - >99.9%). Although the use of MBR was able to reduce the concentration of most heavy metals, there were some elements in which low removal efficiency was observed; for Ni in particular, Katsou, Malamis and Loizidou (2011) further indicated that, although the use of MBR could reduce most of the concentration of heavy metal in industrial wastewater to an acceptable level, it could not reduce Ni concentration to meet any of the reuse criteria set forth by USEPA. In this instances, the use of NF/RO filtration would complement this deficiency, where, the removal of Ni was observed to be between 96.8 – 99.7% using an RO system, and a 89 – 98% efficiency using NF system (Coman, Robotin and Ilea, 2013). The removal of As has also been a challenge due to the inherent chemistry of the compound in water where it exists in multiple forms; in ground and surface water for example, the two predominant forms of As are arsenate (V) and arsenite (III) (Ning, 2002). Although low removal efficiency was reported using MBR systems (Bolzonella et al., 2010), the use of NF/RO systems can compensate for this inefficiency as it has been illustrated that NF/RO technologies have a removal efficiency of 89-90% for NF (Nguyen et al., 2009) and 88 – 99% for RO (Ning, 2002).

Treatment options other than MBR have also been proposed in tandem with the use of NF/RO technologies, such as advanced oxidisation processes, adsorption using activated carbon (AC), electrocoagulation, and flocculation. The use of electrocoagulation, for example, have been exemplified with regards to contaminants removal, such as arsenic, selenium, and phosphorous (e.g. Emamjomeh and Sivakumar, 2009; Martínez-Villafañe, Montero-Ocampo and García-Lara, 2009; Garcia-Segura et al., 2017; Song et al., 2017). For example, Hansen et al. (2019) illustrated the removal efficiency of 90% for selenium in petroleum refinery wastewater, though they highlighted the constraint regarding competition between species affecting it removal efficiency.

Although the aforementioned processes have been illustrated to have a high removal efficiency in the highlighted literatures, its application remains at small or pilot scale, with little literature reporting its implementation at a larger scale. For example, for advanced oxidation processes, although their use has been proven effective at bench or pilot scale, it has been argued that implementation at a large scale is largely unknown due to limited research, and might not provide the same removal efficiency, or require a higher knowledge to be implemented at an operational level (Gogate and Pandit, 2004). For example, the usage of oxidation processes for removal of trace organic matter has been argued to be uneconomical in large scale wastewater treatment applications, in addition to not being able

to fully degrade complex compounds (Gogate and Pandit, 2004). Alternatively, advanced oxidation process such as O3 and UV have typically been utilised as disinfection process instead, where chlorination has conventionally been used as well (Asano et al., 2007). Similar challenges have been reported for the use of AC for heavy metal removal; Fu and Wang (2011) noted that, although its usage is recognised as a more effective means of removing heavy metal contaminations, the disadvantage mainly lies in the cost of AC itself as it has been noted that the price of AC is increasing due to depleting sources of commercial coal- based activated carbon. They also indicated analogous challenges to the use of flotation technologies for heavy metal removal in terms of cost, although the high cost of implementation is throughout its lifespan (both capital and O&M) rather than due to the source itself.

As such, an effective treatment train that is to be implemented within an industrial park thus depends on several factors other than contaminants removal efficiency, namely cost (capital cost (CAPEX) and operational cost (OPEX)), with technology availability, availability of technical expertise, land availability further influencing its uptake. However, for this research, only the CAPEX and OPEX will be taken into consideration. Furthermore, treatment trains which have been implemented at a large scale, either at an industrial park (IP) or national level are prioritised over conceptual design due to arguments highlighted previously.

4.3.1.2 Industrial wastewater reuse treatment trains for direct non-potable industrial water reuse

While the majority of the literature focuses on wastewater reuse optimization for industrial application at an IP level, the treatment trains considered have usually been represented as a black box; literature which considers specific treatment trains for decentralised wastewater treatment facilities for industrial wastewater reuse in an IP to date is limited. An example is a study by Sadr et al. (2015), in which they attempted to determine the preferred treatment options for various water reuse scenarios using multi-criteria analysis (MCA). Ten criteria were used, encompassing costs, energy consumption, environmental impact, community acceptance, adaptability, ease of deployment and level of complexity, land requirement, and water quality and reliability to assess both direct and indirect potable/non-potable water reuse in developed and developing countries. Scenario 3 in particular looked into the use of

municipal wastewater for non-potable water reuse in a water-stressed developing community, where the following treatment train was found to be suitable:

Treatment train, direct non-potable reuse (DNPR): Primary treatment + CASP (Anoxic + Aerobic) + MF/UF + Disinfection

Similarly, Piadeh, Alavi-moghaddam and Mardan (2018) compiled several real cases of treatment trains in Iran to assess for their sustainability when deployed in developing countries. Using a modified analytic hierarchy process (AHP) as an evaluation method, they identified the most sustainable configurations based on 32 indices of economical, technological, environmental, social, and safety aspects of each configuration; they found that economic and technologic criteria to be the most crucial role when a treatment train is to be implemented as a new practice. However, they noted that, due to the lack of measurement instruments to monitor trace organic and inorganic contaminants, they did not manage to assess the removal efficiency of the treatment trains for trace organic and inorganic constituents; instead, their design parameters only considered the discharged rate, COD, TSS, and TDS. They found the following configuration to be preferable:

Treatment train, DNPR (Iran): Sand filtration (SF) + MF + UF + RO

Tong et al. (2013) estimated the potential environmental impact of a newly built water recycling plant in Jiangsu, China, in which the plant was designed to produce 4,000 m3/day of deionised water for the EIP. The water recycling plant received treated wastewater from a decentralised WWTP in the EIP itself, where wastewater from chemical, grain, and oil industries was treated using a secondary treatment based treatment train of a CASP in a sequencing batch reactor (SBR) and an anoxic/oxic (AO) process, followed by a biological fluidized bed (BFB). The main objective of the study was to assess the environmental impact of wastewater recycling in the EIP, and although they found water reuse to be beneficial from the environmental perspective, they did not, however, assess the removal efficiency of contaminants in the wastewater to adhere to water reuse limits. Nevertheless, the treatment train designed for the EIP is below:

Treatment train, DNPR (China): Continuous membrane filtration (CMF) + two step RO + Electrodeionization (EDI)

Singapore has also provided alternative water resources by treating municipal wastewater for non-potable reuse (Singapore Public Utility Board, 2011). Termed NEWater, it was offered

as an alternative water resource for non-potable industrial use for wafer fabrication plants, power generation and petrochemical industries, in addition to commercial and public buildings for cooling towers (Lee and Pin Tan, 2016). Although the use of NEWater is primarily for direct non-potable reuse, it is also used for indirect potable reuse, as it is discharged into an existing reservoir for further integration with raw water prior to treatment for potable water use (Seah et al., 2008). Furthermore, it has been emphasised that NEWater consistently complies with drinking water standards set by WHO and is considered safe for human consumption (Gerrity et al., 2013). The treatment train to produce NEWater is below:

Treatment train, DNPR (Singapore): Primary sedimentation tanks (PST) + MBR + RO + UV

Of the four treatment trains, the one proposed by Sadr et al. (2015) was primarily designed for agricultural application instead of industrial reuse. Furthermore, this treatment train is a conceptual design. As such, this treatment train might not be suitable for this research. On the other hand, although the configuration reported by Tong et al. (2013) was designed for implementation in an actual EIP, it did not include biological treatment; instead, this was compensated for with a two-step RO treatment. Due to this, the configuration might be more prone to fouling if deployed for a Malaysian effluent scenario, which could translate to a higher O&M cost. Furthermore, although the RO system is known to have a high removal efficiency, its usage via this configuration might not reduce constituent concentrations to the limit specified for this research without additional pre-treatment prior to the RO system.

The treatment train utilised in Singapore has been proven to remove contaminants to a high degree. However, it was designed to serve large users (national level); this might point to high capital and O&M cost, which might not be financially feasible if it were to be replicated at an IP level. Conversely, the configuration reported by Piadeh, Alavi-moghaddam and Mardan (2018) is similar to Singapore’s design, with added treatment of SF. Although the use of SF might be better than only employing a primary sedimentation tank, it might not be necessary. However, as compared to the former two configurations, the latter two trains seem to offer the best coverage in terms of contaminant removal, with an added advantage of being readily installed in other countries; due to this, the treatment train reported by Piadeh, Alavi- moghaddam and Mardan (2018) and Singapore Public Utility Board (2011) are considered as potential design options for direct non-potable wastewater reuse at IP levels.

4.3.1.3 Industrial wastewater reuse treatment trains for direct potable industrial water reuse

In comparison to non-potable reuse, literature relating to potable reuse (both direct and indirect) emphasises the importance of multiple barrier systems to reduce the probability of drinking water contamination from pathogens and harmful organic and inorganic contaminants to the greatest practical sense (World Health Organization, 2008). Due to this, although the implementation of indirect potable reuse is common, direct potable reuse is extremely limited with only a handful of cases being reported globally. This is primarily due to the health risk of directly linking treated wastewater sources with drinking water supply without an additional environmental buffer, and to a degree, due to the public perception of the source of water itself (Binz, Razavian and Kiparsky, 2017).

In order to minimise human exposure to pathogens and contaminants, several measures can be taken such as source control, the use of environmental buffer for indirect potable use or blending for direct potable reuse, and a combination of multiple treatment processes where each treatment provide a specific level of contamination reduction (Asano et al., 2007). The use of the first measure – source control – has been emphasised greatly by systems currently in operation. An example of this can be observed through the oldest existing cases of direct potable reuse in Windhoek, Namibia. This system complied with the above objective through the use of three barriers, comprising of non-treatment (management), treatment, and operation barriers. Throughout it operation lifespan (1968 – present), it has been reported that no adverse health outcomes have been associated with the consumption of drinking water produced by this system (Rodriguez et al., 2009). However, it has been emphasised that one of the most critical elements of the system is the strict separation of domestic and industrial wastewater, where any industrial wastewater discharged is treated separately in another central treatment plant (du Pisani, 2006; Lahnsteiner, Van Rensburg and Esterhuizen, 2017). The Public Utility Board (PUB) of Singapore has also underlined the importance of source separation, in which they have limited industrial effluent into their wastewater treatment to a maximum of 15% (Seah et al., 2008). In both of these cases, the emphasis on industrial effluent separation from municipal wastewater is in contradiction with the interest of this research as the wastewater effluent considered here is purely industrial wastewater. As such, there is a risk that the use of these existing treatment trains might not be able to cope with the high level of organic and inorganic contaminants inherently found in industrial wastewater, leading to the system not being able to comply with the limits set forth by drinking water standards.

Another means of minimizing contaminants is the use of an environmental buffer; however, for direct potable reuse, this element is eliminated from the overall process where, instead, treated wastewater is blended with either raw or treated water (Leverenz, Tchobanoglous and Asano, 2011; Gerrity et al., 2013). Blending is a crucial element for direct potable reuse due to the lack of environmental buffer required for the dilution of dissolved solids and effluent organic matter in the recycled water (Lahnsteiner, Van Rensburg and Esterhuizen, 2017). This practice can be observed in Namibia, in addition to Big Spring, Texas (Tchobanoglous et al., 2011; Gerrity et al., 2013); the treatment train employed in Namibia has a recycled water contribution of no more than 35%, while Texas has a maximum value of 15% (Rodriguez et al., 2009; Tchobanoglous et al., 2011; Gerrity et al., 2013). Additionally, although Singapore’s reuse practice can be considered as indirect potable reuse, they have employed the same practice; approximately 2% of treated wastewater is being injected to Singapore’s reservoir for further potable water treatment (Vincent et al., 2014).

The final measure involves the combination of treatment processes to reduce wastewater contaminants to comply with a specific standard. In all three configurations, the overall treatment train is divided into two parts; wastewater treatment train prior to raw surface water blending, followed by a conventional drinking water treatment subsequent to the blending. Between the three configurations, the treatment train in Namibia places more emphasis on the disinfection process post blending, while Texas and Singapore have split the two stages of disinfection between pre- and post-blending processes. For Namibia specifically, this intensive design could be to compensate for shortcomings of the Gammams WWTP. In addition, although treatment trains for Texas and Singapore are similar, slight differences could be observed; in addition to those in Texas omitting the use of the environmental buffer that is exploited in Singapore, the use of MF (pre-blending) and granular media filtration (post-blending) in the former configuration as compared to MBR (pre-blending) and sand filtration (post-blending) in the latter was also observed.

Treatment train, direct potable reuse (DPR) (Namibia): Secondary domestic effluent blended with raw water + Coagulation/Flocculation (C/F) + Dissolved air flotation (DAF) +

Rapid Sand Filtering (RSF) + O3 + Biological and Granular AC (B/GAC) + UF + Chlorination and stabilization + Distribution

Treatment train, DPR (Texas): Disinfected tertiary effluent + MF + RO + UV/H2O2 + Blending with raw surface water + Coagulation/Flocculation/Sedimentation + Granular media filtration + Chorine + Distribution

Treatment train, indirect potable reuse (IDPR) (Singapore): Primary settlement tank (PST) + MBR + RO + UV + Blending + Coagulation/Flocculation/Sedimentation (C/F/S) +

SF + Chlorine/O3 + Distribution

The treatment train of all three cases discussed previously are as above. Comparing between the three, the one utilised in Namibia differs slight from that of Texas and Singapore in terms of complexity and contaminant removal efficiency due to differences in technology application. Conversely, although the treatment train in Texas was designed specifically for direct potable reuse, little literature has been published regarding its efficiency and effectiveness of contamination removal as compared to Singapore. As this treatment train is similar to that of Singapore, in which the extension of the treatment train post-blending is similar to a conventional drinking water treatment, Singapore’s treatment train will be compared with that of the Namibia treatment train for direct potable reuse suitability.

4.3.2 Cost estimation for treatment trains

Costings for wastewater treatment have been noted to depend heavily on several factors, in which the three that are most criticised are arguably the capital investment (CAPEX), operation and maintenance (OPEX), and land cost. The components considered for the three factors are as follows (Economic and Social Commission for Western Asia, 2003):

Capital investment (CAPEX): cost of the unit treatments in the treatment train (inclusive of the corresponding ancillary equipment), piping, instrumentation and controls, pumps, installation, engineering, delivering, as well as any contingency cost of establishing the facility

Operation and maintenance (OPEX or O&M): maintenance, taxes and insurance, labour, energy, treatment chemicals, and any residual maintenance

Land cost: cost associated with procurement of the total land required for the facility, typically based on the equipment’s dimensions

Although all of these cost components influences the pricing of potential wastewater for reuse, it has been noted that literature focusing on cost estimation of wastewater treatment and water reuse projects have been limited at best (Guo, Englehardt and Wu, 2014), mainly due to difficulty in obtaining relevant data (Hernandez-Sancho, Molinos-Senante and Sala- Garrido, 2011). Similar to the method used to estimate potable water consumption described in Chapter 3, both capital and O&M costs have usually been estimate via the use of cost function; the most common approach to its development has arguably been through statistical and mathematical methods. Here, the actual or estimated capital and/or operating cost have been associated with the main variable of a wastewater treatment facility (e.g. per capita or treated volume of wastewater) via fitting the best curve through the data points collected (Koenig, 1973; Hernández-Sancho et al., 2015). For example, the Economic and Social Commission for Western Asia (2003) compiled 20 data points relating to the installation cost for RO filtration plant per treated volume of wastewater in US to obtain the corresponding cost function, while Guo, Englehardt and Wu (2014) developed a cost function of several unit treatments using published data from several countries.

In the context of wastewater treatment costings specifically, literature has been observed to quantify the cost function via the use of inverse, logarithmic, power, linear, or polynomial functions (Gillot et al., 1999; Nogueira et al., 2009). While several studies have reported the use of the power law to be applicable in representing the wastewater capacity as a function of capital, and O&M cost (e.g. Feng and Chu, 2004; Friedler and Pisanty, 2006; Nogueira et al., 2009), others have illustrated the use of linear fit for their data points (e.g. Ahmed, Tewfik and Talaat, 2002). The resulting choice of best fit differs according to data used in the study as a means of data extraction and factors included in the cost estimates vary between authors (Tsagarakis, Mara and Angelakis, 2003).

For CAPEX specifically, the use of power functions have further resulted in the six-tenth rule typically used by water planners, in which the unknown cost estimation of a re-sized plant is estimated from a known plant using Eq. 4.1 (Economic and Social Commission for Western Asia, 2003). Here, the power coefficient (α = 0.6) was assumed based on the observation of median installed investment exponents from various studies of installed wastewater treatment plant of varying capacities; it was noted that the corresponding exponents had a close to normal distribution, resulting in a median of 0.68 and a standard deviation of 0.13 (Koenig, 1973). Although Eq. 4.1 has been utilised by planners, it has been noted however, that the use of power coefficient differs according to unit processes utilised.

0.6

퐶푎푝푎푐푖푡푦푛푒푤 푝푙푎푛푡 퐶표푠푡푛푒푤 푝푙푎푛푡 = 퐶표푠푡푒푥푖푠푡푖푛푔 푝푙푎푛푡 × ( ⁄ ) 퐶푎푝푎푐푖푡푦푒푥푖푠푡푖푛푔 푝푙푎푛푡

Eq. 4.1

The use of method to estimate wastewater treatment facility costs highly depends on data availability. For example, for CAPEX, although the use of the six-tenth rule has been preferred by the water operators due to its ability to scale up/down a pre-designed wastewater facility using a known cost estimation primarily from a vendor’s quotation, due to the calculation method, it is more applicable in the instances where the circumstances of building the new plant is similar to an existing plant. In other words, the use of this method might not be applicable for projects which are in different countries or regions, unless there are strong justifications for the similarities between the two projects. On the other hand, the use of the cost function developed using data from other countries also presents the same challenge as well. For example, in the latter case, Friedler and Pisanty (2006) noted variances in construction cost in Israel as compared to USEPA due to reasons such as differences in living standards which leads to lower construction cost, as well as differences in process efficiency and cost of specific technologies between the two countries. However, as compared to the six-tenth rule, the development of a cost function generally takes into consideration a wider range of projects with different sets of construction environments and conditions, either within the same country, or across countries, as illustrated by the Economic and Social Commission for Western Asia (2003) and Guo, Englehardt and Wu (2014) above. Thus, the use of a generalised cost function allows for a more robust estimation of the overall cost, as compared to a more rigid estimation of the six-tenth rule. Furthermore, although it has been cautioned that the use of cost functions to estimate both capital and O&M cost is an approximation of the actual cost as the developed functions are not universal, it has been recognised as an acceptable means for screening and scaling of available treatment processes, rather than the use of actual cost determination (Hernández-Sancho et al., 2015).

4.3.3 Life cycle assessment for wastewater treatment and reuse

Assessing the environmental impact of wastewater reuse depends heavily on the inventory data. In this sense, a majority of the literature indicated a small, but not insignificant, impact on of the construction phase as compared with the use or operational phase (e.g. Renou et al.,

2008; Fang, 2010), although this might not be the case for passive technologies such as constructed wetlands and trickling filters which have low electricity and chemical consumption during their use phase (Vlasopoulos et al., 2006). For example Stokes and Horvath (2006) illustrated that construction accounted for 4% to 9% of the total global warming potential (GWP) of several alternative water supply systems consisting of a desalination plant, and recycled water via a nearby wastewater treatment plant, as compared with an operation phase contribution of 60% to 91%. A similar range was also reported by Igos et al. (2014) in a study that was specifically designed to assess the impact of the construction phase as compared with the operational phase of several unit processes for drinking water production, and illustrated a range of 4% to 12% contribution of the construction phase on several impact categories, including fossil depletion, climate change, and human toxicity, in which the steel content was the main contributor to the environmental impact categories.

Concurrent with the above findings, a large amount of literature reported the significant impact of electricity consumption (e.g. Sombekke, Voorhoeve and Hiemstra, 1997; Mohapatra et al., 2002; Pillay, Friedrich and Buckley, 2002), while activated carbon (when utilised) (e.g. Igos et al., 2014; Zepon Tarpani and Azapagic, 2018) and chemical consumed such as lime and alum also significantly contributes to the total environmental impact (e.g. Bonton et al., 2012; Tong et al., 2013). For the former, this is particularly true for treatment trains which require a higher contaminant removal rate due to stringent requirements for reuse applications. In these instances, it has been noted that the use of advanced treatment technologies such as MBR and membrane filtration technologies have been preferred, resulting in higher electricity consumption during the use phase (e.g. Ortiz, Raluy and Serra, 2007; Hospido et al., 2012a; O’Connor, Garnier and Batchelor, 2014). The electricity grid mix, in this case, was observed to play a significant role in further amplifying this result – for example, Bonton et al. (2012) illustrated a smaller impact for GWP when using the Quebec electricity grid mix which has a high percentage of hydroelectric power, as opposed to a higher impact if the grid mix has a higher percentage of coal. On the other hand, when simulating the use of nuclear energy, its result for resource depletion was significantly larger than the use of hydroelectricity due to the uranium consumption.

In terms of impact category, the two indicators which have been commonly reported with regards to wastewater treatment and reuse have been GWP and eutrophication (Corominas et al., 2013). For the former specifically, it’s mainly due to its linkages with the high electricity

consumption; Rodriguez-Garcia et al. (2011), for example, attributed the increase in GWP impact to the larger aeration period required to decrease the COD concentration in the effluent. They also noted an increase in GWP due to higher electricity use in several tertiary treatments for wastewater reuse; similar findings were reported by Pasqualino, Meneses and Castells (2010). Although that was the case, sludge disposal was also highlighted to be one of the contributing factors to GWP when included in the system boundary; O’Connor, Garnier and Batchelor (2014) illustrated that, when sludge management was considered, the contribution of sludge to the overall GWP is higher than electricity, except in instances where RO was included in the treatment train.

For eutrophication, its association was mainly with the wastewater effluent, relating to the total nitrogen, total phosphorus, and, to a degree, COD (Amores et al., 2013). For wastewater reuse specifically, lower eutrophication impact was reported as compared to no-reuse scenario - however, the removed impact from the nitrogen and phosphorus was transferred to sludge to increase the impact on other indicators (Tong et al., 2013). Pintilie et al. (2016) exemplified this, where they indicated an increased impact in most of the indicators except eutrophication and total toxicity due to the lower discharge of phosphorus via wastewater effluent.

Additionally, the assessment of toxicity-related impact categories was also observed to be relevant to this study specifically as the source of wastewater reuse is industrial effluent; however it has been noted that there is an uncertainty in the characterisation of heavy metal (Pizzol et al., 2011a). This is due to the different number of metals, as well as the inherent differences in the fate and exposure modelling considered, depending on the LCIA model used (Renou et al., 2008). As such, compared with the GWP and eutrophication indicators where most models will generate a relatively consistent result, the use of different models for toxicity-related impact could generate a different result even if applied to the same inventory (Dreyer, Niemann and Hauschild, 2003). Pizzol et al. (2011) emphasised this; in a study which assessed nine different LCIA methodologies, they found no consensus between the models when assessing the impact of various metals on human health due to the reasons given above.

4.4 Methodology 4.4.1 Treatment train removal efficiency assessment

For this research, the parameters used to assess the effectiveness of the identified treatment trains were limited to those that are common between the effluent discharge limits in Appendix C and potable/non-potable water limits specified in Appendix D. This is because for contaminants that are reported in Appendix C and not in Appendix D, there is no means to determine the initial concentration of the contaminants in the wastewater and their subsequent removal to comply with the potable/non-potable water standard, while for the reverse, there is no means to determine if the resulting concentration after removal would be of any impact on the users.

In order to estimate the efficiencies of the identified treatment train options, literature reporting the removal efficiency and its reference for each constituent were compiled, as shown in Table H-1 and H-2 in Appendix H. Initial screening of the current estimate for the removal efficiency for each parameter using a specific treatment option was first done via recently published review papers. Subsequently, in-depth review of literature reporting the removal efficiency of each parameter obtained through experiments and optimisations of unit process was undertaken to further assess their suitability and efficiencies for use in the present research. In this regard, when available, literature which reported the use of industrial wastewater was prioritised over those using synthetic wastewater or groundwater, as the latter usually focus on single species contaminant removal. Furthermore, the multiple contaminants that are inherent in mixed industrial wastewater might affect the removal efficiency of other species. For example, contradictory results for the removal efficiency of manganese in the + 2+ presence of NH4 and Fe were noted by Bruins et al. (2014), where a high removal + efficiency for manganese was observed in instances of high NH4 removal, while instead low removal of manganese was observed in the presence of high iron concentration. The latter specifically is due to the competing nature of Fe2+ ions with Mn2+ ions for adsorption sites; as the former is oxidised first, available active sites on filter media is first covered by high iron hydroxides/oxides concentration, leading to limited active sites to adsorb Mn2+.

Additionally, priority was also given to literature which reported the use of commercially available equipment and materials without modification or enhancements – for example, literature which used commercial AC were preferred over modified/impregnated/low-cost alternative AC. This is primarily because the use of modified/enhanced materials/systems are

typically not implemented at large/industrial scales; as such, the use of the removal efficiencies reported by this type of literature might not capture the setting typically encountered at the latter scale.

Although the overall removal efficiency of each contaminant for the identified trains could be estimated using this method, it is critical to note that that this estimation via summation of the removal efficiency of each unit process is only an estimation and may not reflect the in- practice removal efficiency in actual implementation of the system which will depend on the continuous, practical optimization of each unit process in the treatment train. This, in turn, depends on various conditions such as the physical property of the unit process (e.g. pressure, temperature, and pH) and effluent characteristic, composition, and concentration (e.g. Shu and Chang, 2004; Ipek, 2005; Nguyen et al., 2009; O ’connor, Garnier and Batchelor, 2014). Furthermore, although the compilation of the removal efficiency prioritised those using mixed wastewater, the majority of the literature reports the removal of single species of contaminant in the wastewater. In reality, multiple contaminant species can be expected to be present in actual industrial wastewater. This was demonstrated by Bernal-Martínez et al. (2010), in which they used industrial effluent collected from an industrial park and reported a lower removal efficiency for COD using ozone treatment due to the complex mix of compounds present in the wastewater sample. Due to this, the optimization of the system based on a mixed industrial wastewater would need to be done to precisely determine a contaminant’s removal efficiency. Due to the complexity in modelling the removal of multiple contaminants using the identified treatment trains, the optimization of the multiple treatment trains falls outside of the scope of this study and the working assumption for the modelling was that the system would perform at an optimized condition with a constant removal efficiency as reported in the literature. The compiled removal efficiency for all unit processes are summarised in Table G-1 and G-2 in Appendix G.

The performance of each treatment train was assessed via the ability of each unit process to remove the listed contaminants, calculated via the percent removal of a specific unit process. The contaminant concentration remaining after each unit process in a given treatment train was calculated by multiplying the incoming concentration by (1 – removal efficiency). Specifically, the final contaminant concentration of all contaminants was calculated using Eq. 4.2 below.

퐶푓푖푛푎푙 = 퐶푖푛푖푡푎푙(1 − 푅푢푝)(1 − 푅푢푝−1) … (1 − 푅1) Eq. 4.2

퐶푓푖푛푎푙 − Final contaminant concentration, mg/l

퐶푖푛푖푡푖푎푙 − Initial contaminant concentration, mg/l

푅푢푝 − Removal efficiency for unit process up, expressed as a fraction

For potable water reuse, contamination concentration removal calculations differed slightly from that for non-potable water reuse due to the blending of wastewater effluent with potable water. The blending element was necessary in order to comply with the multiple barrier concept underlined in subsection 4.3.1.3 above. Here, several blending ratios of wastewater/potable water were considered similar to those discussed in the literature; specifically 2/98, 15/85, and 35/65, following Singapore, Texas, and Namibia respectively. Additionally, considering the location of the industrial parks in the service area to be within urban area with limited access to rivers or groundwater, only blending with potable water was considered for this research. As such, recalculation of contaminant concentration with additional potable water was done via Eq. 4.3 below. Final contaminant removal was then calculated via Eq. 4.2 above.

푚푒푓푓푙푢푒푛푡푣푒푓푓푙푢푒푛푡 + 푚푝표푡푎푏푙푒푣푝표푡푎푏푙푒 = 푚푏푙푒푛푑푒푑푣푏푙푒푛푑푒푑 Eq. 4.3

푚푒푓푓푙푢푒푛푡 − contaminant concentration in wastewater effluent; mg/l

3 푣푒푓푓푙푢푒푛푡 − wastewater effluent volume; 푚

푚푝표푡푎푏푙푒 − contaminant concentration in potable water; mg/l

3 푣푝표푡푎푏푙푒 − potable water volume; 푚

푚푏푙푒푛푑푒푑 − contaminant concentration in wastewater, post blending; mg/l

3 푣푏푙푒푛푑푒푑 − total wastewater volume, post blending; 푚

Moreover, each of the treatment trains were also assessed for the lowest possible efficiency reduction in order to assess the minimum required efficiency to comply with the limits set in Appendix D. In this regard, systematic reductions of percent removal efficiency of each unit process in the treatment train relative to the reported efficiencies were carried out. The system was considered to be ineffective in the event where the reduction in efficiency can no longer meet the limits specified.

4.4.2 Cost estimation of treatment train

The method specified by Hernández-Sancho et al. (2015) was used to estimate the cost for the four identified treatment trains utilised in this research. Accordingly, the use of cost functions to estimate the CAPEX and OPEX of each unit processes was done based on existing literature, as appended in Table H-3 in Appendix H. In order to ensure consistency, the cost functions used have mainly been compiled exclusively from the United States of America, primarily due to the extensive literature available from that country. The land cost, however, was omitted from the cost calculation, as it was assumed that the land within each IP is within the ownership of the IP management itself, and thus would require minimal expenditure to acquire the land.

Although the resulting cost estimate was true in the year in which it was developed, several pieces of literature (e.g. U.S. Environmental Protection Agency, 2000, 2007; Hernández- Sancho et al., 2015) indicated the importance of updating the cost to reflect current values. Due to this, both OPEX and CAPEX were further normalised from the base year referenced in the literature to 2016 using Eq. 4.4 below. As previously noted, since the cost functions were from the U.S.A. specifically, the Engineering News Record (ENR) Construction Cost index and U.S. Bureau of Labour Statistic (BLS) Labour Cost/Index were used to adjust the calculated CAPEX and OPEX respectively (Qasim, 1999). Both the ENR and BLS index used for this research are summarised in Table G-7 in Appendix G.

퐼푥푐푢푟푟푒푛푡 퐶푐푢푟푟푒푛푡 = 퐶푏푎푠푒 × ( ⁄ ) Eq. 4.4 퐼푥푏푎푠푒

Where:

퐶푐푢푟푟푒푛푡 − Cost of unit process in current year (2016); 푈푆퐷

퐶푏푎푠푒 − Cost of unit process in base year; 푈푆퐷

퐼푥푐푢푟푟푒푛푡 − ENR (CAPEX) or BLS (OPEX) index in current year (2016); unitless

퐼푥푏푎푠푒 − ENR (CAPEX) or BLS (OPEX) index in base year; 푢푛푖푡푙푒푠푠

For CAPEX specifically, the resulting updated cost was further annualised without amortization, as it is assumed that the capital investment for the wastewater treatment facility will be borne by the government, specifically the Pengurusan Aset Air Berhad (Water Asset Management Company, PAAB) as a result of the water service industry restructuring.

However, as compensation, the OPEX of each wastewater treatment facility included a lease rental of a rate of 4% per annum of the CAPEX (The Malaysian Water Association (MWA), 2015). Due to this, calculation of the internal rate of return (IRR) was also omitted as it was assumed that the implementation of the corresponding wastewater treatment facility was to address the issue of water deficit in Selangor by the government rather than as a means to generate profit. Additionally, the cost of potable water used for blending purposes for potable water reuse was also included in the calculation of OPEX at a rate of RM 2.27/m3 for Selangor (The Malaysian Water Association (MWA), 2015).

Subsequent to this, the calculated CAPEX and OPEX was then converted to the Malaysian Ringgit (MYR) for the year 2016, using the Purchasing Power Parity (PPP) conversion factor published by The World Bank. The PPP was used over the exchange rate as the former was calculated by taking into consideration the exchange rates, inflation, as well as cost of living between countries, using the U.S.A. as the base country. The use of PPP essentially enables the comparison of the same good or service in different economies, and although other factors have been known to affect price comparison such as tax incentives, income level, and price structure of the same good or service, in principle, PPP will be able to provide equivalent utility comparison between two countries (World Bank, 2015). The PPI index used was 1.425.

4.4.3 LCA methodology 4.4.3.1 Goal and scope

The LCA assessment was conducted in line with the principles underlined in BS EN ISO 14040 and 14044 document. The goal of this research was to compare the environmental performance of four treatment trains to produce recycled wastewater for industrial reuse, corresponding to both non-potable and potable water reuse. However, due to the differences in the water quality produced by both type of reuse options, the resulting environmental impact cannot be considered equal, and thus they are interpreted as two separate scenarios. Thus, the functional unit (FU) considered for this study was:

“The production (outflow) of 1 m3 of treated mixed industrial wastewater for non-potable (NP), and potable water (P) reuse at Industrial Parks in Selangor, Malaysia”.

Based on the literature review discussed in subsection 4.3.3, it can be observed that the impact from the construction and operational phases, as well as the sludge disposal each contributes to different impact categories; as such, all three components were considered in the scope of the LCA study. Additionally, the scope of this study was from cradle to gate, without taking into consideration the decommissioning of the treatment train. As such, the boundary was confined to flows relating to the unit processes only, without taking into account the distribution system of the wastewater into, and treated wastewater out of the treatment plant. A diagram of the system boundary is illustrated in Figure 4-1 below.

Figure 4-1: System boundary for LCA

4.4.3.2 Life cycle inventory

As the four treatment trains were conceptual in nature, there was no site-specific, primary data collected for any of the configurations. Due to this, data from multiple literature sources were compiled for each of the unit processes in the four treatment trains, where the most significant input and output flows within the system boundary were considered in the inventory.

For the inventory relating to the construction materials of each unit processes, data reported in the literature were scaled up using the ‘economy of scale’ relationship (Zepon Tarpani and Azapagic, 2018) using Eq. 4.5 below. The scaling of the wastewater treatment plant was done

100

based on the capacity for the largest industrial park previously identified in Chapter 3 – specifically CH_IP-6 – which would have a capacity of 5,500 m3/day. Priority were given to literature which had reported the quantity for steel and concrete, as these two materials were identified to influence the results of the LCA for the construction phase (Igos et al., 2014). Inventory data for the construction phase, as well as the lifespan associated with each components reported in the literature for all the unit processes in each of the treatment trains are illustrated in Table H-8 to Table H-13 Appendix H.

0.6 푐2 퐶2 = 퐶1 ( ) Eq. 4.5 푐1

Where:

퐶1 and 퐶2 are the material requirements for smaller and larger scale plants, respectively

푐1 and 푐2 are the capacity for smaller and larger scale plants, respectively

In terms of data for the operational phase, when available, literature which reported the use of full-scale operational data were prioritised over those compiled from other studies, modelled, or those obtained from pilot scale experiments, similar to the compilation of removal efficiency. If data were not available or of poor quality, literature which reported a similar influent quality as this study were then considered. If neither type of literature was found, data from pilot-scale and modelled data were used instead, in addition to average value of a general assumption for a wastewater treatment plant reported from Qasim (1999) or Tchobanoglous, Butron and Stensel (2004). For the later source, it is typically the case for electricity consumption of motors or pumps for specific unit process as the electricity consumption reported in literature are usually the total electricity consumption for a reported treatment train.

Data from all literature were adjusted to reflect the inflow into the system. In instances where the inventory was reported for the outflow (e.g. per m3 of potable water produced), the data were back-calculated to reflect the inflow. The average value of all reported quantities of material and electricity consumption compiled from various literature sources were then used as input for impact assessment calculations at the midpoint level.

All unit processes in each treatment train was assumed to be operating at a 1% loss, with the exception of ozone, UV, and chlorination. Wastes generated for each of the unit processes were modelled based on the removal efficiencies reported. The sludge produced from the

101

DAF and the sedimentation tank were calculated based on the suspended solids and heavy metal removal and was assumed to be disposed to landfill (O’Connor, Garnier and Batchelor, 2014); a similar assumption was also used for the activated carbon used in the GAC process, and all removed contaminants by the membrane filtration technologies (Tangsubkul et al., 2005; Tong et al., 2013). It was assumed that no wastes were generated from ozone, UV, and chlorination.

In terms of the Life Cycle Impact Assessment (LCIA) method used, this research used the CML 2001 method. It was previously underlined that the CML 2001 method placed a heavier emphasise on heavy metals that were considered for this study to calculate the toxicity- related categories, while resulting a comparable result for global warming potential (GWP), eutrophication, and acidification impact categories as other methods (Dreyer, Niemann and Hauschild, 2003; Renou et al., 2008). On the other hand, Pizzol et al. (2011b), in a later study, highlighted that the use of USEtox had been preferred over other methodologies due to its consistency, transparency, and reliability in estimating toxicity-related categories specifically. However, USEtox only considers the human toxicity and freshwater ecotoxicity for its midpoint impact category, with the latter being reported in terms of PMF m3-day/kg emission, as compared to 1,4-DBeq/kg emission for CML 2001 and ReCiPe. For this research in particular, the contaminants removed from the wastewater were assumed to be disposed in landfill; as such, the primary impact from the contaminant was considered to be terrestrial ecotoxicity, with freshwater and marine ecotoxicity secondary, as a result from the potential leaching of the sludge to both compartments via surface runoff. Due to this, CML 2001 was also used to assess the ecotoxicity-related categories (terrestrial, freshwater, and marine), as the USEtox had no basis for comparison for the terrestrial and marine ecotoxicity, owing to the different unit being used for the freshwater ecotoxicity as compared to CML 2001 and ReCiPe. Additionally, the human toxicity impact category was also used to assess the final effluent being reused as well as the impact of contaminant disposed to landfill, as it is being directly consumed by the users of the system for the former instance (in case of potable water reuse); the CML 2001 method was also used for this impact category as the list of substances covered by this method was more in line with this research as compared with USEtox.

Inventory data for the operational phase of all the unit processes in all treatment trains are compiled in Table H-14 and Table H-15 in Appendix H. The LCA calculation for the treatment trains were done using SimaPro v8 software, and utilised the Ecoinvent 3 database, using the unit process datasets, and the default allocation method (alloc, def).

102

4.5 Results and discussion 4.5.1 Effectiveness of contaminant removal using identified treatment trains 4.5.1.1 Non-potable reuse

Calculation of contaminants removal was first assessed for non-potable reuse using the treatment trains identified in the literature review section above (4.3.1.2), where NP-1 replicated the treatment train in Singapore (Figure 4-2), while NP-2 was the treatment train utilised in Iran (Figure 4-3), as summarised in Table 4-1 below. Table 4-2 illustrates the contaminants removal according to the reported removal efficiency; overall, both treatment trains were able to remove all contaminants to the level well below those specified by the limit for non-potable water reuse summarised in Appendix C. Interestingly, assuming that all premises complied with the discharge limits, NP-1 had even managed to reduce all contaminants to meet the potable water standard, while NP-2 had met all limits except aluminium and phenol. Although, for the latter, this could be due to the compiled database itself; limited literature were found which had reported the removal of the two species specifically for the other unit processes in NP-2. For NP-1, the results were within the range reported by Singapore’s National Water Agency for the NEWater quality (Singapore Public Utility Board, 2009). However, for NP-2, it was noted that the treatment train design did not take into consideration most of the pollutants specified due to lack of sufficient measurements (Piadeh, Alavi-moghaddam and Mardan, 2018).

Table 4-1: Unit processes in non-potable and potable treatment trains Reuse option Treatment Train Unit Processes NP-1 PST + MBR + RO + UV Non-potable reuse NP-2 SF + MF + UF + RO

P-1 C/F/DAF + SF + O3 + GAC + UF + Chlorination Potable reuse P-2 NP-2 + C/F/S + SF + Chlorination + O3

Figure 4-2: Treatment train diagram for NP-1

Figure 4-3: Treatment train diagram for NP-2

103

Table 4-2: Compliance to non-potable reuse limit by NP-1 and NP-2

NP-1 NP-2 Influent, Std. Limit, Limit, Parameter Unit Final Comp, Comp, Final Comp, Comp, B NP P Concentration NP P Concentration NP P BOD5 at 20°C mg/L 50 5 - 1.4x10-1 Yes Yes 3.7x10-1 Yes Yes COD mg/L 200 10 - 3.8x10-1 Yes Yes 4.7x10-1 Yes Yes Suspended mg/L 100 5 - 3.5x10-1 Yes Yes 8.7x10-1 Yes Yes solids Cadmium mg/L 0.02 0.01 0.003 3.0x10-6 Yes Yes 1.0x10-5 Yes Yes Arsenic mg/L 0.1 0.1 0.01 5.7x10-4 Yes Yes 1.0x10-3 Yes Yes Copper mg/L 1 0.2 1 6.0x10-3 Yes Yes 2.4x10-8 Yes Yes Manganese mg/L 1 0.2 0.1 7.5x10-3 Yes Yes 3.9x10-6 Yes Yes Nickel mg/L 1 0.2 0.02 2.5x10-3 Yes Yes 1.4x10-6 Yes Yes Zinc mg/L 2 2 3 5.1x10-3 Yes Yes 1.8x10-6 Yes Yes Iron (Fe) mg/L 5 5 0.3 7.1x10-3 Yes Yes 6.5x10-5 Yes Yes Aluminium mg/L 15 5 0.2 6.6x10-2 Yes Yes 8.4x10-1 Yes No Selenium mg/L 0.5 0.02 0.01 1.0x10-3 Yes Yes 1.0x10-3 Yes Yes Fluoride mg/L 5 1 0.6 1.4x10-1 Yes Yes 1.4x10-1 Yes Yes Colour ADMI 200 20 15 1.5 Yes Yes 2.4x10-1 Yes Yes

104

Comparatively, NP-2 managed to remove most of the heavy metal contaminants to a higher degree than NP-1; this likely due to the unit processes deployed in the treatment trains. For NP-2, unit processes employed focused more on the usage of membrane filtration technologies (MF/UF/RO), while NP-1 only employed RO in combination with a biological treatment technology (MBR) which typically utilised UF membrane within the MBR system. Although in this case the use of membrane filtration technologies showed higher contaminants removal, it was also illustrated here that the use of a combined MBR/RO system was also capable of meeting the limit with potentially lower resource consumption than is typically associated with the former technologies (e.g. Fu and Wang, 2011; Miller et al., 2014). Furthermore, the use of MBR as compared to multiple membrane filtration technologies have the added advantage of biological treatment which is not presented in NP-2, enabling NP-1 to remove a higher degree of organic matter potentially present in the mixed industrial effluent; this could be the case for wastewater generated by food and beverage manufacturing. Additionally, the combined use of RO/UV unit processes in NP-1 also allowed for a higher degree of pathogen removal, although this parameter was not assessed here due to not being monitored by the Environmental Quality Act 1974.

100

10 Decrease Decrease inoverall removal efficiency, % 0 BOD COD SS Cd Cu Mn Ni Al Se F Colour

Compliance Non-compliance

Figure 4-4: Effect of decrease in overall removal efficiency on compliance (NP-1)

105

100

10 Decrease Decrease inoverall removal efficiency, % 0 BOD COD SS Cd Cu Mn Ni Al Se F Colour

Compliance Non-compliance

Figure 4-5: Effect of decrease in overall removal efficiency on compliance (NP-2) On the other hand, it was also demonstrated that the use of multiple membrane filtration technologies similar to those deployed by NP-2 is potentially more robust in the event of reduction in removal efficiencies. This can be observed in the results summarised in Figure 4-4 and Figure 4-5 above for NP-1 and NP-2 respectively. Overall, both systems performed similarly for most of the parameters, where the more notable difference observed was for heavy metal contaminants. For both treatment trains, the weakest link was Se; it was further observed that neither system was able to remove the contaminant effectively if the removal efficiency dropped below 3%. This is mostly due to the inability of most of the unit processes to effectively remove the contaminant. According to the literature, the RO system was demonstrated to be effective in removing Se from wastewater at full scale; however, all other unit processes in both treatment trains were reported to be ineffective or provide insignificant removal on its own due to the molecular weight and size of the contaminant (Sandy and DiSante, 2010). Due to this, it is imperative that the RO system be adequately maintained in order to ensure satisfactory removal of Se from the wastewater. However, this could pose a challenge due to competition with other species such as arsenic, phosphate, and silicate, as well as the overall condition of the system (Santos et al., 2015), both of which could drastically reduce the system’s efficiency for Se removal. As such, this contaminant might be a hindrance for industrial parks with heavy emphasis of electronic, rubber, and stainless steel and copper coating manufacturers, as they are known to have high usage of selenium (Kapoor, Tanjore and Viraraghavan, 1995). In this sense, the incorporation of electrocoagulation in

106

particular might be able to increase the removal efficiency of this contaminant in particular, as illustrated by Hansen et al. (2019) in subsection 4.3.1.1., where they had exemplified a removal efficiency of 90% for selenium in the case of petroleum refinery wastewater. However, as noted previously, the use of this technology would need further research in order to fully capture the feasibility of its use at a large scale.

Further comparing the two configurations, it can be observed that the use of a multiple membrane system in NP-2 results in a more robust design for the removal of most of the heavy metal contaminants. With the exception of Al and Se, it was observed that the system was still able to remove the contaminants in the event where the percentage removal efficiency of each unit process decreased to below 50%. As three of the four unit processes in NP-2 reported consistently higher capabilities for heavy metal removal than the MBR/RO hybrid in NP-1, a significantly higher decrease in efficiency would still result in a higher combined removal than the latter system, thus enabling the former configuration to have a higher tolerance for system inefficiency. On the other hand, in order to maintain the high removal efficiency of the overall system, it is essential that fouling be controlled via consistent backwashing or replacement of the UF and RO membrane in order to maintain the system, or risk a decline in contaminant rejection rate; as iterated in multiple literature sources and the discussion in subsection 4.3.1.1, this could lead to a higher O&M cost for the former or a higher electricity consumption to maintain filtration efficiency. In this sense, NP- 1 could be the more inexpensive option as the system only utilised two membrane filtration technologies while simultaneously adhering to the specified limit. However, its lower tolerance would require a close monitoring of both the wastewater influent concentration as well as consistent maintenance of the system.

4.5.1.2 Potable reuse

Similarly, calculation for contaminants removal for potable water reuse was assessed for the treatment train P-1 (Figure 4-6) and P-2 (Figure 4-7), as summarised in Table 4-1 above. However, potable water reuse has the added element of blending with piped potable water which acts to reduce all of the contaminants concentration in the wastewater via dilution. The wastewater to potable water ratio also greatly influenced the compliance with the drinking water quality standard, Table 4-3 summarises the results of the treatment trains for all three blending ratios. Overall, it was found that for the 2/98 wastewater/potable water ratio, both

107

treatment trains managed to meet the limit for all contaminants even if the reported efficiencies for each unit process dropped to below 50%. However, for a blending ratio of 15/85 and 35/65, although interestingly P-2 had managed to still meet the limits, P-1 had only managed to do so for selected pollutants. For the former system specifically, it was mainly due to the highly efficient contaminant rejection rate in the NP-1 treatment train described in subsection 4.5.1.1; even without blending, the system had managed to meet the limit specified by potable water reuse standard.

Figure 4-6: Treatment train diagram for P-1

Figure 4-7: Treatment train diagram for P-2

For the P-1 treatment train, a similar pattern for selenium was observed as for non-potable water reuse trains. For ratio of 15/85 and 35/65, none of the unit processes were able to remove Se from the incoming wastewater, nor did the blending managed to dilute it enough to meet the required standard. In this case, it could be attributed to the fact that the treatment train was not designed to receive any industrial effluent into the system in the first place, as emphasised in most of the literature (e.g. du Pisani, 2006; Lahnsteiner, Van Rensburg and Esterhuizen, 2017); this was reflected in the low removal efficiency with regards to some of the heavy metal contaminants.

In terms of robustness, the P-1 system showed a low tolerance for aluminium, fluoride, and phenol, where the system was not able to remove these contaminants according to the limit if the reported removal efficiencies decreased by more than 40% for 15/85 ratio, and 20% for 35/65 ratio (Figure 4-8). For aluminium and phenol, the system is required to remove a high percentage of the contaminants from the wastewater; approximately 94% for the former and 99% of the latter. As such, although the treatment train has a relatively high removal efficiency for these two pollutants, there is still a small margin of error in the event of fluctuation in initial contaminant concentration or decrease in system efficiency.

108

On the other hand, although fluoride only required approximately 54% removal from blended wastewater, the unit processes in this treatment train reported a lower removal efficiency. For DAF specifically, although there are literature reporting a high removal using this technology (e.g. Chuang, Huang and Liu, 2002), contradictory results were reported when mixed industrial effluent was used. This was possibly due to the vast array of contaminants found in the wastewater, leading to higher competition between the contaminants species and reducing the rejection rate, as the case described by Chuang et al. (2006); similar reasoning is also attributed for the use of B/GAC process. Furthermore, the use of UF had only managed to reduce a small percentage of residual F concentration, as it has been noted that this technology was not efficient for ion removals due to its lower molecular size (Ndiaye et al., 2005; dos Santos Bazanella et al., 2012; Liu and Liu, 2016). Conversely, for P-2, it had the advantage of the RO system prior to blending, which has been noted as one of the more efficient technologies for removing the pollutant (Ndiaye et al., 2005; Mohapatra et al., 2009). As such, even though the technologies subsequent to NP-1 did not significantly add to its removal, the F concentration after blending was diluted enough to meet the drinking water quality standard.

It can be concluded that the blending ratio of wastewater to potable water does play a significant role in enabling the treatment trains to meet the potable water reuse limit. However, the analysis suggests that there is a risk to increasing the ratio of wastewater to beyond 35%. For P-1, it is not feasible as there was a specific contaminant that did not meet limit (Se) even with lower blending ratio, while for removed contaminants, there is a possibility of non-compliance in instances where removal efficiency decreases below a certain limit for several of the pollutants. However, for P-2, it might be feasible as all the contaminants were removed well below the required limits; even Se – which presented the highest risk as the P-2 had no removal of the contaminant – was able to meet the limit with a margin of 50% with blending.

109

Table 4-3: Compliance to potable reuse limit by P-1 and P-2 Wastewater Blending, 2/98 Blending, 15/85 Blending, 35/65

concentration

1 2 1 2 1 2

------

tion,

e water water e

Unit

1 2 1 2 1 2

------

P P P P P P

Contaminant

Potabl

concentration

after NP after

2, Final conc. Final 2, conc.

1, Standard B Standard 1,

Compliance, P Compliance, P Compliance, P Compliance, P Compliance, P Compliance, P Compliance,

Final concentration, Finalconcentration, Finalconcentration, Finalconcentration, Finalconcentra Finalconcentration, Finalconcentration, BOD5 at mg/L 50 1.4x10-1 - 1.1x10-4 Yes 2.8x10-5 Yes 8.9x10-4 Yes 2.2x10-4 Yes 2.3x10-3 Yes 5.8x10-4 Yes 20°C COD mg/L 200 3.8x10-1 - 4.5x10-3 Yes 4.0x10-4 Yes 3.7x10-2 Yes 3.2x10-3 Yes 9.5x10-2 Yes 8.4x10-3 Yes Suspended mg/L 100 3.5x10-1 0 1.6x10-4 Yes 1.6x10-5 Yes 1.3x10-3 Yes 1.3x10-4 Yes 3.4x10-3 Yes 3.4x10-4 Yes solids Mercury mg/L 0.05 8.0x10-5 0.001 2.8x10-9 Yes 5.0x10-4 Yes 2.8x10-9 Yes 4.7x10-4 Yes 2.8x10-9 Yes 4.1x10-4 Yes Cadmium mg/L 0.02 3.0x10-6 0.003 1.8x10-8 Yes 1.4x10-7 Yes 3.1x10-8 Yes 1.3x10-7 Yes 5.7x10-8 Yes 1.1x10-7 Yes Arsenic mg/L 0.1 5.7x10-4 0.01 1.3x10-4 Yes 5.0x10-4 Yes 2.8x10-4 Yes 4.6x10-4 Yes 5.5x10-4 Yes 4.1x10-4 Yes Cyanide mg/L 0.1 3.8x10-4 0.07 2.0x10-3 Yes 1.9x10-3 Yes 2.2x10-3 Yes 1.8x10-3 Yes 2.7x10-3 Yes 1.5x10-3 Yes Lead mg/L 0.5 7.0x10-5 0.01 0 Yes 0 Yes 0 Yes 0 Yes 0 Yes 0 Yes Copper 6.0x10-3 1 1.0x10- Yes 3.5x10-5 Yes 1.1x10-10 Yes 3.3x10-5 Yes 1.2x10- Yes 2.8x10-5 Yes mg/L 1 10 10

Manganese mg/L 1 7.5x10-3 0.1 1.7x10-9 Yes 8.6x10-7 Yes 3.7x10-9 Yes 8.1x10-7 Yes 7.4x10-9 Yes 7.1x10-7 Yes 2.45x10 Nickel mg/L 1 0.02 4.4x10-9 Yes 3.9x10-4 Yes 2.0x10-8 Yes 3.7x10-4 Yes 4.9x10-8 Yes 3.3x10-4 Yes -3 Zinc mg/L 2 5.1x10-3 3 9.7x10-7 Yes 3.0x10-1 Yes 9.9x10-7 Yes 2.8x10-1 Yes 1.0x10-6 Yes 2.4x10-1 Yes - - -3 8.9x10 -5 -10 -5 5.3x10 -5 Iron mg/L 5 7.1x10 0.3 11 Yes 6.5x10 Yes 2.4x10 Yes 6.1x10 Yes 10 Yes 5.3x10 Yes

Aluminium mg/L 15 6.6x10-2 0.2 1.6x10-4 Yes 8.2x10-3 Yes 8.6x10-4 Yes 8.0x10-3 Yes 2.1x10-3 Yes 7.6x10-3 Yes Selenium mg/L 0.5 1.0x10-3 0.01 6.8x10-3 Yes 5.0x10-3 Yes 3.1x10-2 No 4.7x10-3 Yes 7.5x10-2 No 4.2x10-3 Yes

110

Fluoride mg/L 5 1.4x10-1 0.6 9.4x10-2 Yes 1.8x10-2 Yes 1.8x10-1 Yes 1.7x10-2 Yes 3.5x10-1 Yes 1.6x10-2 Yes Phenol mg/L 1 0 0.002 5.1x10-4 Yes 2.0x10-6 Yes 1.7x10-5 Yes 1.8x10-6 Yes 1.7x10-5 Yes 1.6x10-6 Yes Free mg/L 2 4.8x10-2 5 2.5 Yes 2.5 Yes 2.5 Yes 2.3 Yes 2.4 Yes 2.0 Yes Chlorine Colour ADMI 200 1.5 15 1.3x10-4 Yes 1.7x10-1 Yes 3.1x10-4 Yes 1.6x10-1 Yes 6.5x10-4 Yes 1.4x10-1 Yes

Figure 4-8: Effect of decrease in overall removal efficiency on compliance (P-1)

111

4.5.2 Assessment of cost function utilisation for treatment train cost estimation

Results for the financial assessment – both CAPEX and OPEX – for all four treatment trains are presented in Table 4-4 below. Based on these the costs of potable water reuse are similar for both treatment trains, while a substantial difference could be seen for non-potable water reuse.

Table 4-4: CAPEX, OPEX, and unit cost of the four treatment trains

Overall costs in MYR P-1 P-2 NP-1 NP-2 Total CAPEX (MYR) 341,221,054 344,993,877 231,905,683 152,873,643 Annualised CAPEX (MYR/Year) 11,712,038 11,862,360 8,589,052 7,290,478 Annualised OPEX (MYR/year) 71,841,903 67,313,148 16,350,412 37,153,688 Per unit cost (MYR m3/day) 8.70 8.10 2.60 4.60

For CAPEX specifically, the differences between the two types of wastewater reuse could be due to the number of unit processes utilised in the treatment trains for potable water reuse as compared to non-potable reuse. For the former, the higher number of unit processes were to ensure the removal of contaminants to an acceptable level (du Pisani, 2006); this naturally translated to a higher CAPEX. However, although the unit processes utilised in P-1 and P-2 differs from one another, the total CAPEX for both options were broadly similar. Conversely, comparing between NP-1 and NP-2 for non-potable reuse, despite the latter treatment train having a higher number of advanced treatment technologies, the CAPEX for NP-1 was higher than NP-2. Upon closer inspection, this was due to the higher cost of the MBR system. This is in line with published literature highlighting the high CAPEX (and subsequent OPEX) as one of the major disadvantage of the MBR system as compared with conventional systems (U.S. Environmental Protection Agency, 2007) owing to the additional installation of a biological tank for the former unit process (Judd, 2017).

A similar trend for OPEX was also observed for both water reuse options, where higher OPEX was observed for potable reuse as compared with non-potable reuse; here, the OPEX for the former was calculated to be approximately double the latter. This was primarily attributed to the additional cost of potable water used during blending. As previously emphasised, blending is used to simulate the dilution factor found in indirect reuse in order to decrease the risk of using wastewater as a source of potable water; this was validated in subsection 4.5.1.2 where the blending ensured the treatment train met the requirement even

112

though the treatment train did not significantly add to the removal of some of the contaminants. This requirement added an additional MYR 14.2 million per year to the OPEX cost for non-potable reuse. In this instance, the use of other sources of water such as ground or river water instead of potable water would reduce approximately 17% and 27% of the total OPEX costs for P-1 and P-2 respectively. However, depending on the quality of the raw water used, it could also result in a potentially higher influent contaminant level – for example, a higher concentration of arsenic could be expected if groundwater was used instead.

The overall annual OPEX for P-1 was approximately MYR 4.5 million more than for P-2. Aside from the lease and potable water cost, the highest O&M cost was, interestingly, due to the sand filtration process, potentially due to the high cost of sand considered in the cost function. Additionally, although most of the unit processes in both treatment trains were similar, it was also observed that the major differences were the use of DAF and B/GAC in P- 1, while P-2 used MBR and RO. Assessing these four unit processes in the same capacity, the highest O&M cost was observed to be for B/GAC. In this case, it could potentially be due to the associated high cost of AC where various literature have emphasised this due to material scarcity (e.g. Fu and Wang, 2011). However, this is not certain as there are also literature which noted the cost savings in the utilisation of BAC in place of RO, particularly in terms of annual electricity (e.g. Gerrity et al., 2014).

Conversely, the difference for OPEX for non-potable water reuse was substantial of approximately MYR 21 million higher for NP-2 than NP-1, despite a higher CAPEX for the latter. The literature emphasises i) the high O&M cost for most of the unit processes found in NP-2 (e.g. Cote, Masini and Mourato, 2004; Bolzonella et al., 2010; Judd, 2016), ii) the higher electricity requirement for RO and, iii) higher maintenance resources required to reduce the occurrence of fouling in the MF and UF unit processes, all of which translated into a higher OPEX for NP-2.

The subsequent unit cost for potable water reuse was observed to be higher than for non- potable reuse. As summarised in Table 4-4 above, the calculated cost for potable water reuse was observed to be four times higher than the current potable water tariff for industrial use of MYR 2.27/m3, where a cost of MYR 8.70 and MYR 8.10 was calculated for P-1 and P-2 respectively. As illustrated previously, the use of the 35/65 blending ratio of wastewater/potable water was determined to be the most feasible as it result in the highest potential savings; however, this had effectively reduce the feasibility of deploying the P-1

113

treatment train for potable water reuse as it was not able to meet the required limit. Here, it was further demonstrated that it does not make economic sense to produce potable water using this treatment train, as it would result in a price of RM 8.70/m3 compared with RM 2.27/m3 of potable water. Even in the event where the calculated OPEX was overestimated by 30% due to the difference in several of the cost components such as electricity and labour cost, the subsequent unit cost for P-1 was only able to be reduced to MYR 7.30/m3. On the other hand, the use of the combined treatment train of P-2 resulted in a technically sound option for potable water reuse as it was able to remove the contaminants to the limits specified in Appendix D; however, the high cost of this option as compared to the current water tariff is likely to be a barrier to its uptake by the industry players.

Conversely, the use of wastewater for non-potable reuse seems to be a feasible option, especially from NP-1 where it resulted in a unit cost of MYR 2.60/m3 as compared with MYR 4.60/m3 for NP-2. Additionally, in comparison with potable water reuse, it was previously demonstrated that both treatment trains managed to remove the contaminants to the level specified. It was previously speculated that NP-1 could potentially meet those requirement by using fewer resources that NP-2, and this was further confirmed via the estimated cost of operation here.

The unit cost appended in Table 4-4 was calculated in the event where the treatment trains are implemented in all of the IPs identified. However, as previously illustrated in Chapter 3, the implementation of this strategy might not result in positive outcome of reducing the WSS’s vulnerability if implemented in the Langat system. Thus, the result was recalculated for the Semenyih system specifically, and it can be observed that there is a slight increase in unit costs for all treatment trains though the same pattern as the above findings remains true: these results are presented in Table 4-5 below.

Table 4-5: CAPEX, OPEX, and unit cost for Semenyih WSS

Semenyih WSS P-1 P-2 NP-1 NP-2 Total CAPEX (MYR) 227,125,746 219,957,209 147,008,158 97,805,580 Annualised CAPEX (MYR/Year) 7,848,872 7,141,953 5,421,184 4,645,743 Annualised OPEX (MYR/year) 45,507,666 42,089,696 10,267,386 23,664,067 Per unit cost (MYR m3/day) 9.50 8.70 2.80 5.00

114

4.5.3 LCA interpretation 4.5.3.1 Non-potable reuse

Result of the LCIA for non-potable water reuse treatment trains are presented in Table 4-6 below. The results generated for all four treatment trains were compared to the LCA results for drinking water from surface water in order to gauge the performance of the treated wastewater against potable water. Generally, it was observed that the construction phase resulted in minimal contribution to all impact categories other than abiotic depletion, in which the construction phase contributed approximately 34% and 35% for NP-1 and NP-2 respectively (compared with less than 3% contribution in the rest of the impact categories). For impact categories other than abiotic depletion, the contribution was similar to those reported in the literature, where the construction phase generally has a small impact. However, for abiotic depletion specifically, the result here was slightly higher than that reported in the literature.

Overall, there were no major differences between the two treatment trains for non-potable reuse for the operational phase; both were observed to be comparable in all impact categories, with a difference of only approximately 1 to 2% between them. This was not surprising as both treatment trains had similar removal efficiencies for most of the contaminants considered, as well as relatively similar total electricity consumption of all unit processes. For the electricity consumption specifically, both treatment trains deployed the use of several advanced treatment processes which have been noted to be energy intensive, resulting in a higher contribution from the operational phase. Results generated mirrored those reported and discussed in subsection 4.3.3, noting the substantial impact of electricity consumption (e.g. Sombekke, Voorhoeve and Hiemstra, 1997; Mohapatra et al., 2002; Pillay, Friedrich and Buckley, 2002), where it could overshadow the contribution of other materials and resources consumed during both the construction and operational phases. Although this is the case, it was observed that values for toxicity-related impact categories were primarily driven by the contaminants disposed to landfill; this is also in line with reported literature when sludge disposal was included within their system boundary (Lemos et al., 2013).

115

Table 4-6: Life cycle assessment (LCA) for the production of 1m3 of treated mixed industrial wastewater for non-potable water reuse

Drinking water, water purification NP-1 NP-2 treatment, production mix, at

plant, from surface water, RER U

Impact Unit

category

Construction Operation % Construction, % Operation, Construction Operation % Construction, % Operation, Total -7 -7 -7 -7 -6 AD kg Sbeq 3.3x10 6.4x10 33.5 66.5 2.7x10 5.7x10 31.8 68.2 1.5x10 -2 -2 -4 GWP kg CO2eq 4.5x10 3.18 1.4 98.6 1.7x10 2.6 0.6 99.4 6.3x10 -9 -8 -10 -8 -11 ODP kg CFC-11eq 2.1x10 8.6x10 2.4 97.6 7.6x10 7.4x10 1.0 99.0 1.2x10 -1 1 -1 1 -5 HT Kg 1,4 DBeq 1.3x10 1.6x10 0.8 99.2 1.6x10 1.6x10 1.0 98.7 2.7x10 -2 -2 -7 F-ET kg 1,4-DBeq 3.0x10 4.0 0.7 99.3 3.0x10 4.09 0.7 99.2 3.2x10 1 4 1 4 -3 M-ET kg 1,4-Deq 7.7x10 1.2x10 0.7 99.4 4.3x10 1.1x10 0.4 99.4 3.3x10 -4 -1 -4 -1 -7 T-ET kg 1,4-DBeq 2.3x10 3.4x10 0.07 99.9 2.3x10 3.6x10 0.1 99.9 2.2x10 -5 -4 -6 -4 -8 PO kg C2H4eq 1.7x10 6.2x10 2.7 97.3 6.9x10 5.0x10 1.4 98.7 5.3x10 -4 -2 -5 -2 -6 Ac kg SO2eq 1.9x10 1.5x10 1.3 98.7 8.7x10 1.2x10 0.7 99.3 1.6x10 -5 -3 -5 -3 -7 Eu kg PO4---eq 5.0x10 4.4x10 1.1 98.9 2.4x10 3.7x10 0.6 99.4 4.5x10 Note:

Abiotic depletion, AD; Global warming, GWP; Ozone layer depletion, ODP; Human toxicity, HT; Freshwater aquatic ecotoxicity, F-ET; Marine aquatic ecotoxicity, M-ET; Terrestrial ecotoxicity, T-ET; Photochemical oxidation, PO; Acidification, Ac; Eutrophication, Eu

116

Based on the overall results, the major contributors to four impact categories (abiotic depletion, global warming potential (GWP), human toxicity, and terrestrial ecotoxicity) are discussed in detail in the following subsections according to their importance in either the Construction or the Operation phase. This focus is due to the relevance of resource consumption used during the Construction phase, electricity consumption during the Operational phase, and the impact of contaminants retained in the final effluent and those removed from the wastewater and disposed to landfill.

i. Abiotic depletion – Construction phase (Figure 4-9)

It is noted that the largest impact on abiotic depletion was the 66% contribution which is mostly attributed to electricity consumption (specifically the use of copper in the electricity production). However, the construction phase is unusual with regard to most impact categories in that it accounted for 34% of the impact on Abiotic Depletion. The drivers for this 34% differ according to the materials used for each unit process. The total impact score for abiotic depletion for NP-1 and NP-2, as well as the breakdown from each unit processes (UPs in Fig 4-4) in both treatment trains for the construction phase are shown Figure 4-9 below.

7.00E-07

6.00E-07

5.00E-07

4.00E-07

/m eq

3.00E-07 kg Sb kg

2.00E-07

1.00E-07

0.00E+00 NP-1 NP-2 NP-1: NP-2: NP-1: NP-2: NP-1: NP-2: NP-1: NP-2: PST SF MBR MF RO UF UV RO Total UP-1 UP-2 UP-3 UP-4

Figure 4-9: Abiotic depletion result for the Construction Phase for non-potable reuse

117

The process which generally had the highest contribution was the use of steel in the unit processes, similar to what has been reported in the literature (e.g. Igos et al., 2014). However, for the MBR process specifically, which has the largest contribution to the total score it was found that, although steel was also included in the inventory, the highest contributor for this impact category was from the use of synthetic rubber, specifically, the use of zinc concentrate.

ii. Global warming potential (GWP) – Operation phase (Figure 4-10)

NP-1 exhibited a higher total electricity consumption resulting in a higher GWP than NP-2 (Figure 4-10). Additionally, the use of a primary sedimentation tank (UP-1) also contributed to the total GWP for the treatment train due to the methane emission of the open tank, although this contribution is minimal as compared with the electricity demand, where the contribution of coal from the Malaysia’s electricity grid mix dominated the GWP score.

Of the eight unit processes considered in both treatment trains, the highest electricity consumption was observed to be from the MBR process; the relatively high electricity requirement is recognised in multiple literature focusing on MBR specifically (e.g. Ortiz, Raluy and Serra, 2007; Hospido et al., 2012) . Although this process was noted to be able to produce a higher effluent quality from a smaller physical footprint as compared with the conventional activated sludge (CAS) process, the trade-off for this technology remains its higher energy consumption. On the other hand, it has been noted that a lower value of 0.4 to 0.8 kwh/m3 can be achieved if the system is fully optimised, although this value has been known to be difficult to achieve (Krzeminski, van der Graaf and van Lier, 2012; Barillon et al., 2013). To further assess the impact of electricity consumption of MBR specifically on the overall GWP score, a sensitivity analysis was conducted. In this instance, if a reduction of 30 to 50% of the electricity consumption from the average 1.8 kwh/m3 can be achieved, a reduction of approximately 16 – 26% of the total GWP for NP-1 (approximately 0.46 – 0.83 3 kg CO2 eq.) can be expected, resulting in a total GWP of 2.35 kg CO2 eq. per m of treated wastewater (in the case of 50% electricity reduction). For the latter, this would then result in a

lower value for NP-2 of 2.6 kg CO2 eq.

118

3.50E+00

3.00E+00

2.50E+00

3 m

/ 2.00E+00 2eq

1.50E+00 kg CO kg

1.00E+00

5.00E-01

0.00E+00 NP-1 NP-2 NP-1: NP-2: NP-1: NP-2: NP-1: NP-2: NP-1: NP-2: PST SF MBR MF RO UF UV RO Total UP-1 UP-2 UP-3 UP-4

Figure 4-10: GWP result for (Operation Phase) for non-potable reuse iii. Human toxicity – Operation Phase (Figure 4-11)

Human toxicity was assessed from two perspectives; the overall, aggregated human toxicity score for the Operation Phase (includes emissions from background data, and emissions to soil due to the contaminant disposal) and the human toxicity score for the resulting treated wastewater for industrial reuse. The differentiation of the two was mainly to distinguish between indirect emissions from the background data and long-term emissions from contaminant extracted from the wastewater and their subsequent disposal via landfill (aka returned to soil) with that of the more direct impact of the use of treated wastewater, where it also consists of residual contaminants within the treated wastewater though at reduced concentrations as specified in the reuse limit.

Of the eight heavy metal contaminants considered in the study for the non-potable reuse, only selenium, lead, nickel, cadmium, copper, zinc, and phenol were taken into account in the CML 2001 method to calculate human toxicity, with selenium having the largest characterization factor of 2.8E4. Here, both NP-1 and NP-2 used RO as one of the unit processes in the treatment train, which has been noted as one of the best technologies available to remove selenium. Interestingly, although NP-2 managed to remove most of the heavy metals to a higher degree than NP-1, the human toxicity score for NP-1 was slightly higher than the former, as illustrated in Figure 4-11 via the darker coloured columns. This is observed for UP-2 (MBR for NP-1 and MF for NP-2) for both treatment trains; although NP-

119

2 removed a higher concentration of nickel, the total score for NP-1 was higher. Upon closer inspection, this was due to the higher electricity consumption of the latter treatment train, in which selenium was also emitted during the electricity production.

The relatively lower removal efficiency of NP-1 was also reflected in the slightly higher human toxicity score for the resulting treated wastewater than for NP-2. As compared with the human toxicity calculation for the use phase, those calculated for the treated wastewater only considered the residual contaminants left in the effluent to be reuse in the industrial parks, as demonstrated by the two lighter coloured columns on the right side of Figure 4-11 below. Comparing between the two treatment trains, NP-1 was noted to have a slightly lower removal efficiency for several of the heavy metals under consideration. However, the relatively higher score was due to the larger characterization factor for nickel, resulting in a higher score for NP-1, where the final nickel concentration in NP-1 (2.45x10-3) was three times higher than NP-2 (1.40x10-6). This indicates that, although the human toxicity potential of both treatment trains were somewhat similar, the utilisation of the effluent from NP-2 represented a slightly lower impact in the LCA than NP-1, based on the contaminants being considered in this study.

18.00 0.0574

kg 1, 4 1, kg 16.00

0.0573 - 14.00 DB

0.0572 eq

12.00 /m

for treated wastewater treated for 10.00 0.0571

8.00 0.057

for contaminantfor removal

6.00 /m

eq 0.0569

DB 4.00 - 0.0568

2.00

kg kg 1, 4 0.00 0.0567 NP-1 NP-2 NP-1: NP-2: NP-1: NP-2: NP-1: NP-2: NP-1: NP-2: NP-1 NP-2 PST SF MBR MF RO UF UV RO Total UP-1 UP-2 UP-3 UP-4 Treated wastewater

Figure 4-11: Human toxicity result (Operation Phase) for non-potable reuse

120

iv. Terrestrial ecotoxicity – Operation Phase (Figure 4-12)

Compared with the human toxicity impact category, the score for terrestrial ecotoxicity was dominated by the disposal of contaminants to soil, where the substance with the largest characterization factor was mercury and nickel, as compared with selenium in the previous category. The result for this terrestrial ecotoxicity category was the opposite of that for the human toxicity category of the effluent generated by these treatment trains, where NP-1 resulted in a higher score than NP-2, whereas for terrestrial ecotoxicity NP-2 resulted in a higher value than NP-1 (Figure 4-12). This is not surprising, as the contaminants that were discarded from the effluent were eventually disposed of in landfill (i.e. entered the terrestrial environment), resulting in a higher score for terrestrial ecotoxicity.

The highest contribution to the overall terrestrial ecotoxicity scores was from MBR and RO unit processes in NP-1, and RO in NP-2. Discounting the removal of selenium (as both treatment trains had removed this contaminant to the same degree due to the presence of RO in both treatment trains), the removal of nickel from NP-2 was far higher than NP-1 in UP-2, resulting in a higher score for the former. The residual nickel that was not removed from UP- 2 in NP-1 however, was then removed in the RO process (UP-3) along with other substances such as cadmium and selenium, which further drove up the score for this unit process for NP- 1.

4.00E-01

3.50E-01

3.00E-01

3 2.50E-01 /m

2eq 2.00E-01

kg CO kg 1.50E-01

1.00E-01

5.00E-02

0.00E+00 NP-1 NP-2 NP-1: NP-2: NP-1: NP-2: NP-1: NP-2: NP-1: NP-2: PST SF MBR MF RO UF UV RO Total UP-1 UP-2 UP-3 UP-4

Figure 4-12: Terrestrial ecotoxicity result (Operation Phase) for non-potable reuse

121

4.5.3.2 Potable reuse

For the potable treatment trains, a wastewater/potable water blending of 35/65 was used in preference to the other two blending ratios, as it had the potential to result in the highest water saving when taking into consideration the efficiency of each unit process. Furthermore, the differences in blending ratio only generally impacted the toxicity-related impact categories, while resulting in consistent results for all other categories due to similar resource and electricity consumptions. For the 35/65 blending ratio, the higher wastewater ratio resulted in higher score for the toxicity-related impact categories.

Table 4-7 below summarises the LCIA result for both treatment trains for potable water reuse, with P-2 being an extension of the previous NP-1 treatment train. As compared to the results generated for scenario 1, contribution from the construction phase as compared to the operational phase mirrored those reported in the literature review; resulted in a minimal contribution of less than 13% for all of the impact categories.

For abiotic depletion, although the contribution of the construction phase for P-2 was observed to be approximately 13% most of this contribution was ‘inherited’ from NP-2 (where it accounted for 67% of NP-2’s total score). However, for the use phase, a different pattern was observed for both treatment trains. As compared with results generated for the non-potable treatment trains, which showed the major contribution of electricity during the Operation Phase, result for abiotic depletion for scenario 2 for both potable treatment trains suggested a heavier impact of materials to the total score (e.g. alum).

Further comparison between P-1 and P-2, showed higher scores for P-2 for the toxicity- related impact categories, while similar scores occurred for other impact categories for the Operation Phase. This was, again, attributed to the higher removal efficiency for the combined P-2 and NP-1 treatment trains as compared to P-1. Similar to the non-potable reuse scenario, results for the abiotic depletion, GWP potential, human toxicity, and terrestrial ecotoxicity impact categories are discussed in further detail in the following subsections due to the relevance for resource and electricity consumption, as well as the impact of contaminant retained in the final effluent and those removed from the wastewater and disposed to landfill.

122

Table 4-7: Life cycle assessment (LCA) for the production of 1m3 of treated mixed industrial wastewater for potable water reuse

Drinking water, water purification P-1 P-2 treatment, production mix, at plant,

from surface water, RER U

Impact category Unit

Construction Operation % Construction, % Operation, Construction Operation % Construction, % Operation, Total -7 -6 -7 -6 -6 AD kg Sbeq 1.2x10 1.1x10 9.7 90.4 2.3x10 1.4x10 14.4 85.6 1.5x10 -3 -2 -4 GWP kg CO2eq 9.7x10 1.9 0.5 99.5 2.5x10 1.9 1.3 98.7 6.3x10 -10 -8 -9 -8 -11 ODP kg CFC-11eq 4.7x10 6.5x10 0.7 99.3 1.8x10 6.3x10 2.8 97.2 1.2x10 -2 -2 -5 HT Kg 1,4 DBeq 5.8x10 1.8 3.2 96.8 6.3x10 6.4 1.0 99.0 2.7x10 -2 -2 -7 F-ET kg 1,4-DBeq 1.3x10 1.6 0.8 99.2 3.0x10 2.1 1.4 98.6 3.2x10 -3 M-ET kg 1,4-Deq 17.8 3,176.2 0.6 99.4 45.8 5,450.1 0.8 99.2 3.3x10 -4 -1 -3 -7 T-ET kg 1,4-DBeq 1.1x10 2.1x10 0.1 99.9 1.1x10 1.3 0.1 99.9 2.2x10 -6 -4 -5 -4 -8 PO kg C2H4eq 4.8x10 3.8x10 1.3 98.8 1.0x10 3.9x10 2.6 97.4 5.3x10 -5 -3 -4 -3 -6 Ac kg SO2eq 4.6x10 9.2x10 0.5 99.5 1.1x10 9.3x10 1.2 98.8 1.6x10 -5 -3 -5 -3 -7 Eu kg PO4---eq 1.5x10 2.7x10 0.6 99.5 3.9x10 2.8x10 1.4 98.6 4.5x10 Note:

123

i. Abiotic depletion – Operation Phase (Figure 4-13)

Generally, it was observed that the total score for P-2 was higher than P-1 for this impact category (Figure 4-13). As discussed in the previous section, the contribution of the construction phase to the overall score was minimal as compared to the Operational Phase at 9% and 13% for P-1 and P-2 respectively. For P-1, the highest contribution was from the ultrafiltration unit process, followed by the tanks in the coagulation, flocculation, and sedimentation unit process, in which both unit processes use steel; while for P-2, as noted previously, the majority of the contribution was from the NP-1 treatment train.

For the operational phase, the majority of the unit processes indicated the largest contribution to the total score due to electricity consumption, similar to the score obtained for non-potable reuse. However, for the two unit processes which utilised alum (specifically P-1/UP-1 and P- 2/UP-2), the contribution from alum superseded the electricity contribution to the overall score, specifically due to the zinc concentrate extraction during alum production. A sensitivity analysis for the alum consumption was also done, where an arbitrary reduction of 50% was assumed; a reduction of 21% from the total score can be expected in this instance.

Comparing between the results here and the literature, though the use of coagulants have also been highlighted as contributing factors to this impact category, it was expected that the use of activated carbon (AC) might have resulted in a higher contribution to the overall score. However, it was observed here that the influence of AC was not as great as has been reported in the literature (e.g. Igos et al., 2014; Zepon Tarpani and Azapagic, 2018), as illustrated in Figure 4-13 below for UP-2 in the P-2 treatment train. One potential reason for this could be due to the relatively smaller quantity of AC used in the present. However, the contribution of AC to the overall total score was found to be minimal; even an increase in the total AC consumption by 50% only resulted in an increase of 0.4% of the total score. The minimal impact of AC could partially be due to Malaysia’s grid mix, where the main contribution to the score for electricity utilisation was from coal, overshadowing the contribution of the AC used in UP-4.

124

1.60E-06

1.40E-06

1.20E-06

3 1.00E-06

m / /

eq 8.00E-07

kg kg Sb 6.00E-07

4.00E-07

2.00E-07

0.00E+00

SF SF

O3 O3

P-1 P-2

GAC

NP-1

C/F/S

C/F/DAF

Chlorination Chlorination Total UP-1 UP-2 UP-3 UP-4 UP-5 UP-6

Figure 4-13: Abiotic depletion result (Operation Phase) for potable reuse

ii. Global warming potential (GWP) – Operation Phase (Figure 4-14)

The GWP scores for both P-1 and P-2 were observed to be relatively similar (Figure 4-14). Although P-2 had more advance treatment technologies employed via NP-1 resulting in the largest contribution of GWP for P-2, P-1 also utilised an ultrafiltration technology which contributed largely to the overall score. Additionally, UP-1 and UP-4 unit processes in P-1 also influenced the overall score; the former being also due to electricity consumption (both during the Operation phase and production of alum), and the latter due to a combination of energy consumption during AC production, as well as the coal that was used to produce the AC itself.

125

2.50

2.00

m 1.50

/ 2eq

1.00 kg kg CO

0.50

0.00

SF SF

O3 O3

P-1 P-2

GAC

NP-1

C/F/S

C/F/DAF

Chlorination Chlorination Total UP-1 UP-2 UP-3 UP-4 UP-5 UP-6

Figure 4-14: Comparison for GWP (Operation Phase) for potable reuse iii. Human toxicity – Operation Phase (Figure 4-15 and Figure 4-16)

Figure 4-15 summarises the human toxicity impact category for both P-1 and P-2. P-2 had a substantially higher score than P-1; this is primarily due to the use of the NP-1 treatment train prior to P-2. As compared with the potable reuse scenario, NP-1 did not have the benefit of blending with potable water, and thus having a higher initial contaminant concentration prior to entering the system. However, the NP-1 treatment train managed to remove most of the contaminants largely due to the number of advanced treatment processes deployed in the configuration (most notably the removal of selenium via the RO process), subsequently increasing its indirect human toxicity impact via its potential exposure of humans via the soil compartment through the disposal of the contaminants to landfill, reflecting the higher contribution from NP-1 for human toxicity for the P-2 treatment train.

In contrast, the lower score for P-1 was primarily due to the lower removal efficiency for selenium relative to P-2, which was identified as the substance which has the highest characterisation factor of the contaminants being considered in this study; P-1 only managed to remove approximately 35% of the total selenium that entered the system. The chemical precipitation via UP-1 had largely assisted in the removal of this substance in addition to most of the other heavy metal contaminants, resulting in the relatively higher human toxicity score for this unit process. The contribution from other unit processes for P-1 was observed to

126

be a combination of other contaminants removed as well as the electricity consumed during the process.

3 /m

4 DB DB eq - 3

kg kg 1,4 2

SF SF

O3 O3

P-1 P-2

GAC

NP-1

C/F/S

C/F/DAF

Chlorination Chlorination Total UP-1 UP-2 UP-3 UP-4 UP-5 UP-6

Figure 4-15: Comparison for human toxicity (Operation Phase) for potable reuse Assessing the resultant treated wastewater, it was found that the blending ratio greatly influenced the final contaminant concentration found from P-1 specifically, while having minimal impact for the P-2 treatment train. P-2 consistently produced a lower score than P-1 (Figure 4-16), where, the human toxicity score for P-1 was four times higher than P-2 in the case of the 35.65 blending ratio. This, again, was attributed to the residual contaminant found in the treated wastewater – specifically selenium – in the former treatment train. As highlighted in Chapter 3, a blending of 35% wastewater with potable water did not managed enough dilution to enable the unit processes in P-1 to remove selenium to meet the required limit for potable reuse; this, in turn, resulted in the significantly higher impact for P-1. In comparison, when a blending ratio of 2/98 was used, the treatment train managed to meet the standard; this was also reflected in the lower score for human toxicity where it was comparable with those for P-2. However, as previously argued, the use of this blending option is inconsistent with the overall aim of reducing potable water consumption by the manufacturing sector, especially when the efficiency of the unit processes was taken into consideration.

Although the blending with potable water assisted P-2 to meet the potable water reuse standard, it did not play such a critical role as for P-1. As noted in subsection 4.5.3.1, the NP-

127

1 treatment train was able to meet the potable water standard even without subjecting the resultant wastewater to blending and the subsequent processes in P-2. Due to this, the difference in blending ratio did not produce such a large difference for the human toxicity impact category.

4.5

3.5

3 /m

eq 2.5

DB - 2

kg kg 1,4 1.5

0.5

0 Blending ratio - 2/98 Blending ratio - 35/65 Blending ratio - 2/98 Blending ratio - 35/65 P-1 P-2

Figure 4-16: Comparison of blending ratio of effluent on human toxicity impact category for potable reuse

iv. Terrestrial ecotoxicity – Operation Phase (Figure 4-17)

The results for terrestrial ecotoxicity (Figure 4-17) complemented those of the human toxicity for the resultant effluent for both treatment trains as the substances taking precedence for this impact category were mercury and nickel, as compared to selenium and mercury for the previous category. Consistent with the human toxicity category, the higher removal of most of the contaminants found in P-2 was due to the bulk of their removal by NP-1, as compared to the rest of the unit processes in the treatment train, with the rest of the unit processes in the configuration contributing to the removal to a lesser degree due to the lower contaminant concentration after blending.

As for P-1, it was observed that most of the contaminants were removed in the UP-1 (coagulation/flocculation/DAF), with the residual contaminants being removed via the rest of the unit processes. Similarly, the removal of selenium and nickel was found to play the major

128

role in driving the score for P-1, where, aside from GAC which was able to only remove 5% of the selenium (UP-4), selenium was exclusively removed via UP-1.

1.40E+00

1.20E+00

1.00E+00

3 /m

eq 8.00E-01

DB - 6.00E-01

kg kg 1,4 4.00E-01

2.00E-01

0.00E+00

SF SF

O3 O3

P-1 P-2

GAC

NP-1

C/F/S

C/F/DAF

Chlorination Chlorination Total UP-1 UP-2 UP-3 UP-4 UP-5 UP-6

Figure 4-17: Comparison for terrestrial ecotoxicity (Operation Phase) for potable reuse

4.5.4 Discussion and comparison of non-potable and potable reuses

In this chapter, both treatment options were assessed for their viability in terms of their technical, financial, and environmental merit as a DSM-based strategy for the manufacturing sector to aid in reducing the current WSS vulnerability in Selangor. Analysis of both treatment options suggested non-potable reuse to be a more effective means to do this as compared with potable reuse in all three aspects.

From the technical and financial perspective, potable reuse was not seen as a viable option. In order to minimise harmful exposure, it required the added element of blending of wastewater with potable (or other sources of) water; this resulted in a maximum of 35% water savings regardless of any type of water used for blending. However, it was also observed that the 35/65 bending option did not manage to remove several key pollutants – particularly selenium – to meet the potable water standard due to the limitations of the unit processes in both treatment trains. This, in turn, led to a further decrease in the overall water savings to only 2% as the 2/98 blending ratio was the only option which could effectively meet the limits specified. Furthermore, due to the number of unit processes used in this reuse option, it also resulted in a significantly higher unit cost as compared to the current water potable water tariff.

129

In comparison, non-potable reuse was perceived to be able to potentially realise the water savings calculated in Chapter 3. As this reuse option did not need to account for the blending ratio, 95% of the total estimated wastewater generation (90% of total water consumption) was potentially treated for reuse, resulting in far higher savings if it was fully utilised. Additionally, the use of NP-1 was also observed to be able to meet the potable reuse standard without the additional blending step, though it is not able to be classified as potable reuse due to the absence of the blending step recommended by WHO. However, for NP-1, due to the minimal number of unit processes used in this treatment train, also presented a risk; it was observed that this configuration is not as robust and could fail to meet the limit if the influent concentration was increased significantly, or if the predicted removal efficiency decreased. In this sense, the overall ecology in the industrial park in terms of effluent concentration from the premises as well as the efficiency of the wastewater treatment facility itself need to be closely monitored in order to adhere to the reuse standard. Furthermore, this option also has the added benefit of a lower unit cost due to the small number of unit processes deployed in this configuration.

From the environmental perspective, literature highlighted the trade-off between meeting a higher reuse standard, with higher resource consumption – specifically electricity – due to either additional treatment processes or the use of advanced treatment technologies such as membrane filtration (e.g. O’Connor, Garnier and Batchelor, 2014). Here, a similar trend was observed. As an example, the NP-1 treatment train (without the P-2 extension) was compared to the P-1 configuration; the use of NP-1 as basis for comparison was due to its meeting the potable water standard without the dilution process deployed by P-1. Assessing the NP-1 treatment train to the same standard as P-1 (i.e. potable reuse), the former required a higher electricity demand, though the overall configuration was much simpler than P-1 in terms of number of unit processes used. In this case, the complexity of P-1 (which favoured the use of more conventional technologies) relied heavily on the blending step as a compensation for weaker removal effectiveness, resulting in this configuration not being able to meet the set limit without utilising the 2/98 blending ratio. While this, in turn, resulted in a lower AD and GWP score for P-1, it also resulted in a significantly higher HT score for the resultant treated wastewater.

Another pattern identified was the burden shifting between the HT impact category for the resulting treated wastewater, with that of terrestrial (and, indirectly, freshwater and marine) ecotoxicity. In complying with the strict reuse standard, the burden was observed to shift

130

from the former category to the latter as various contaminants were extracted from the wastewater and discarded to the landfill. In this case, NP-1, which had a higher removal rate for most of the contaminants (especially selenium which greatly influenced the score for both impact categories), was observed to result in a higher score for terrestrial ecotoxicity as compared with P-1, while simultaneously yielding a lower impact for human toxicity for the reuse of the water .

Using the HT impact factor as a basis of comparison for the environmental impact, as well as taking into consideration the technical feasibility and financial cost of the system, it could be argued that the use of the NP-1 treatment train is superior to P-1 as a DSM-based strategy due to i) its simplicity, ii) it was able to generate a lower impact with lower resource consumption (abiotic depletion) and greenhouse gas emissions (GWP potential), iii) with higher contaminants removal, and iv) at a substantially lower cost. Moreover, NP-1 also presented the possibility of utilising this treatment train to generate output water which adheres to the potable reuse standard. However, for the possibility of actual use of such water for potable use, there still remains the factor of the ‘unknown of the unknowns’. For this research specifically, only a small number of contaminants with a pre-determine concentration were assessed; in reality, there exist a possibility of other potentially hazardous contaminants that could be present in the incoming wastewater that are neither considered nor monitored by the regulatory bodies (e.g. Department of Environment, DoE). In this instances, the simplicity of the NP-1 treatment train could backfire as only a limited number of unit processes are deployed and there is no contingency plan if the existing unit processes were not able to adhere to the expected removal efficiency (e.g. due to premises not complying with their discharge limits, or reduced efficiencies due to competition between contaminant species). In this instance, the use of the blending step prior to entering the treatment train might act to minimise this risk, with an added advantage of potentially prolonging the lifespan of the MBR and RO processes due to the dilution factor. Another consideration could be the addition of another advanced treatment option such as MF to further increase the removal efficiency of the treatment train; however, this could translate into higher costs and environmental burdens.

4.6 Conclusions

The technical feasibility, financial cost, and environmental impact of four treatment trains corresponding to non-potable and potable reuse were assessed. For non-potable reuse options,

131

it was found that both NP-1 and NP-2 managed to treat the wastewater effluent according to the non-potable water reuse standard. Additionally, it was also found that NP-1 was capable of meeting the potable water reuse standard without the need for further blending with potable water. Comparatively, NP-2 did not managed to meet the standard for aluminium and phenol, although this would not be an issue if potable water blending was used. Regarding potable water reuse, both P-1 and P-2 managed to meet the limit if only 2% of the discharged wastewater effluent was utilised (2/98 blending ratio). However, if the blending ratio was increased to 15/85 and 35/65, the P-1 treatment train failed to meet the requirement for selenium, although this was met for P-2 because the NP-1 pre-treatment in this train prior to the blending did remove all of the contaminants according to the potable water standard. It was further found that although it was able to be removed in the NP-1, NP-2, and P-2 treatment trains, the removal of selenium was observed to be a sensitive to the removal efficiency of the RO technology as other unit processes were not able to remove the contaminant as effective as the former technology, thus making the system vulnerable in instances of decreased efficiency of the unit process to below 96% in the three treatment train options.

The issue with selenium influenced the environmental assessment for the reuse options, specifically the human toxicity impact category. Comparing between the two reuse options, the inability of the P-1 treatment train to remove selenium resulted in a significantly higher score for the HT impact category as compared to all other treatment trains; this is especially true for the blending ratio with higher wastewater contribution (35/65). On the other hand, it was also observed that the lower HT score for P-2, NP-1, and NP-2 resulted in a concomitant high terrestrial ecotoxicity score due to the assumption that all contaminants removed are disposed in landfill, indicating a burden shifting from the former category to the latter. This, in turn, further emphasise the need to dispose of the contaminants appropriately.

From the financial perspective, the non-potable reuse resulted in a significantly lower unit cost as compared with the potable reuse option; this was due to the higher number of unit processes in the latter option, as well as the element of blending with potable water. Comparatively, NP-1 resulted in a similar cost to the current industrial water tariff, suggesting that this treatment option might be a good candidate to encourage the uptake of alternative water supplies by the industrial players.

132

Chapter 5 : Overall assessment of decentralised wastewater treatment plant implementation for industrial reuse

5.1. Introduction

In Chapter 2, 3, and 4, it was illustrated that the use of industrial wastewater for the manufacturing sector was technically, financially, environmentally feasible to implement, if the non-potable reuse option was considered. Furthermore, it was also illustrated that this option is heavily dependent on the critical mass of industrial premises within the service area, as well as the volume of wastewater discarded from each of the premises. As illustrated in Chapter 3, the use of non-potable reuse for the manufacturing sector had only managed to reduce the vulnerability of the Semenyih WSS against NS events, while had resulted in minimal impact to the Klang Gates and Langat dam.

Although that is that is the case, as highlighted in Chapter 1, consideration of a DSM-based strategy implementation should also take into consideration factors beyond the technical, financial, and environmental factors (UNEP, 2011). Other consideration such as the institutional structure of the existing water sector, as well as the readiness of the manufacturing sector itself should be incorporated and evaluated in order to determine the effectiveness of its implementation – this is the focus of this chapter, in addition to assessing the effectiveness of the DWWTP implementation with regards to other SSM-based approach.

5.2 Comparison of DWWTP implementation with other DSM-based approach

Ultimately, this research sought to determine the feasibility of industrial wastewater reuse in a strategy based on decentralised wastewater treatment plants (DWWTP) at industrial parks as a DSM-based approach to minimise the water imbalance issues. In order to put the proposed strategy into perspective as well as to gauge its feasibility, it is here compared to two SSM water projects that have been approved for budget allocation tabled during the recent Malaysian budgets. It has to be noted, however, that detailed information regarding these two projects are sparse and only available at a generic level. The first was presented in the 2017 Malaysian budget of approximately RM 732 million (here called New WSS), allotted to increase and upgrade a WSS for 22,412 people (estimated for 5,200 houses at an average household size of 4.31 (Department of Statistics, 2010)) (Prime Minister’s Office, 2017), while the second is the allocation of an additional RM 1.3 billion in the 2018 budget

133

for an off-river raw water storage (here called New Ponds), aimed at supplying approximately 3000 MLD of raw water upon completion of 5 ponds (Prime Minister’s Office, 2018).

For the former case, it is planned for a community within a rural area, requiring the establishment of a relatively small WSS through the conventional means of securing raw water supply via the construction of a new dam, potable WTP, and corresponding distribution network. In comparison, the latter New Ponds project is planned at a relatively large scale for a rapidly developing city where raw water supply was no longer readily accessible, thus requiring the refurbishment of existing mining ponds for supply augmentation. The construction of these ponds is expected in two phases, with the first phase currently under construction with a budget of RM 700 million and expected to supply approximately 300 MLD of raw water – for comparison, information found regarding the completed phase of the project will be used. A summary and comparison of the above projects with regards to those proposed in this thesis is as below (Table 5-1).

Table 5-1: Cost and capacity comparison between DWWTP, New WSS, and New Ponds projects DWWTP (NP-1) New WSS Project New Ponds Project Water supply 15 decentralised Conventional WSS: 3 retention ponds system wastewater Dam-conventional located near river for treatment plants: potable water raw water PST-MBR-RO-UV treatment plant- abstraction for potable water additional water distribution system storage Cost (CAPEX) 147 million 732 million 700 million (MYR/USD) (36 million) (177 million) (170 million) Capacity (MLD) 12.5 203 300 Population served 53,970 people 22,412 people 758,442 people Buffer time 6.3 years (short 50 years (long-term) > 50 years (long- term) term) Implementing state Selangor (2%) Pahang (1.2%) Selangor (2%) (potable water demand growth rate, 2016)

3 Estimated based on domestic demand of 5.2 MLD for 22,412 people, and 3.7 MLD for non-domestic demand (41.6%), projected over 50 years lifespan of WSS at 1.2% demand growth rate

134

The treatment train proposed in this research is comparable to the New WSS project in terms of capacity and population served, though the NP-1 treatment train requires lower capital investment. Calculation based on the savings suggests that the DWWTP system is able to decrease the vulnerability of the Semenyih WSS and provide a buffer of approximately 6.3 years at a constant demand growth rate of 2%. Here, it can be observed that if the growth rate can be reduced by 0.5% through other DSM-based strategies such as reducing per capita water consumption or by reducing the rate of non-revenue water (NRW), it can further increase the buffer period to approximately 8.3 years. A similar increase in the buffer period can also be illustrated if the savings estimated at the Semenyih WSS can be replicated at the same location as the New WSS project; here, its implementation is estimated to be able to provide a buffer time of approximately 10 years at a growth rate of 1.2%. In this case, a similar 2 year extension in buffer time to Semenyih for the New WSS could be achieved if this demand growth rate can further be reduced by 0.2%.

In both cases, although the proposed DWWTP strategy was illustrated to be able to stall the need for the implementation of other large scale SSM strategies at a relatively low financial cost, it is only able to provide relief on a relatively short term time scale, even when taking into account decreased potable water demand growth through other means e.g. water conservation strategies. However, it also provides an additional security net to stave off potential non-satisfactory incidents while the introduction of unavoidable SSM strategies is taking place. The implementation of large scale SSM strategies – specifically the more conventional construction of a new dam or inter-basin water transfer systems – require detailed planning which considers the environment, social, and land issues all of which translate into long timeframes to execute (ASM, 2016). Additionally, though the capacity of the decentralised systems are by no means large, they enable complementary SSM strategies to be implemented on a relatively short time scale than if this strategy is not implemented, thus minimising potential negative environmental impacts, such as those relating to land use change.

Based on the discussion above, it can be observed that the estimation of buffer time offered by DWWTP implementation depends heavily on both the concentration of industrial premises located within an IP and the quality of the resulting wastewater generated from those premises, as well as the overall potable water demand growth rate in the corresponding WSS. In this case, the implementation of this solution in states outside of Malaysia’s three major conurbation areas (i.e. Selangor, Pulau Pinang, and Johor) might be feasible as they are

135

not expected to have a high potable water demand growth rates similar to those in the three aforementioned states. States such as Melaka and Negeri Sembilan – both of which have been highlighted previously as having water rationing incidents in their recent history (see Appendix A) and which have a relatively lower growth rate than Selangor could be suitable, as these states also have appreciable manufacturing sectors. Furthermore, for the two states, the industrial water tariff is comparable to the unit price of the treated wastewater calculated in Chapter 4, thus potentially requiring lower incentives to encourage the industry players to adopt the offered water source. A comparison of the two states relative to Selangor is summarised in Table 5-2 below.

Table 5-2: Comparison of 5 selected states with regards to potable water growth rate, percent used by the manufacturing sector, and the industrial water tariff (in MYR)

Potable water Used in Industrial water growth rate manufacturing tariff (MYR/m3 of sector potable water) Selangor 2.0 % 29.5 % 2.27 Pulau Pinang 2.5 % 44.1 % 1.36 Johor Bahru 2.2 % 30.7 % 2.93 Melaka 1.7 % 40.4 % 2.13 Negeri Sembilan 1.6 % 41.7 % 2.64

Conversely, the proposed DWWTP strategy may have limited opportunities in conurbation areas (i.e. Selangor, Johor Bahru, and Pulau Pinang), even when there is a large share of manufacturing sector, such as for Pulau Pinang. For this state specifically, although the DWWTP strategy might be feasible due to a larger volume of wastewater that could be collected from the industrial premises, it might not be as attractive as compared to the development of large scale infrastructure such as those of the New Ponds project. Compared with the effort required to implement multiple small scale DWWTP facilities which can only offer relief for a relatively short period, the execution of large scale SSM project - specifically the New Ponds Project - is a comparatively easy and quicker means of securing a large volume of raw water supply for the rapidly growing demand despite the higher initial capital cost; this is discussed further below in the following subjections. In fact, the water tariff is much lower in Pulau Pinang, making the DWWTP option either less attractive to the

136

industry players especially due to the non-potable reuse option, or requiring a higher financial allocation for subsidies or tax incentives from the government to encourage its adoption.

Although the assessment of the DWWTP conducted in this research may have a number of advantages, realising the estimated potential water savings at a large scale would also require additional techno-economic, environmental, awareness, and institutional perspective as discussed in the subsequent sections.

i. Techno-economic perspective

This consideration stems from the reuse option itself. Throughout this thesis, this research has indicated that, technically, the NP-1 treatment train was able to meet both the potable and non-potable water reuse limits. The former reuse limit was considered as it was preferred by the industry players as introduced in Chapter 4. Although the potable reuse option takes into consideration the industry players’ narrative, it was further illustrated that this option was likely to be unattractive to the policy makers as it only result in an insignificant increase of the reservoir’s resilience, and does not aid in mitigating the water imbalance issue within the WSS as illustrated in Figure 3-6 and Figure 3-7 for Langat and Semenyih WSS respectively. For the latter, as underlined in Chapter 3, the limitation is due to the blending step required to ensure minimal human exposure to potential contaminants and pathogens, resulting in a maximum of 35% savings from this reuse option. As the study assumed the use of potable water for blending purposes due to the distance of the IPs from river networks, this blending resulted in an insignificant impact on decreasing the water imbalance issue in the WSS.

The above limitation can be reduced however, if other sources of raw water not within the current WSS catchment is utilised instead (i.e. groundwater). However, this would further increase the unit cost of the treated wastewater due to additional infrastructure require to abstract the raw water. For groundwater specifically, additional assessment regarding the availability of the source would need to be conducted to determine long-term viability. A detailed groundwater assessment might not be available and would require additional financial and technical resources to execute. Furthermore, from a technical viewpoint, the optimisation of the treatment train would need to be done to account for possible contaminants that are present in the groundwater (e.g. arsenic). Nevertheless, the higher unit cost attributed to both the CAPEX and OPEX could be a major deterrent for industry if the use of industrial wastewater for potable reuse is considered; for the former, due to the higher

137

number of unit processes in the treatment train to account for contingencies to limit human exposure to contaminants, while for the latter, due to blending with other source of water.

On the other hand, the use of non-potable reuse for industrial application was illustrated to result in substantial savings in potable water for the overall WSS. Although its implementation would allow the policy makers to meet the objective of reducing the occurrence of non-satisfactory conditions, the non-potable reuse option may have a lower likelihood of being implemented simply due to the industry player’s perception of preferring the treated wastewater to be of potable standard, despite this not being technically necessary for the manufacturing processes (e.g. cooling).

One reason for this preference of the industry players may well be the requirement for a separate piping system to ensure continued segregation of the treated non-potable wastewater from the potable water supply at their premises. Although the resultant treated wastewater also complied with the potable water reuse standard and not having any adverse effect of the manufacturing process, direct human consumption is discouraged due to the risk of contaminants that are either not monitored or not regulated being present in the treated wastewater - the element of the ‘unknown of the unknown’. as emphasised throughout this thesis. Although the installation of the separate piping could be well within the technical capability of the implementing firm given that there is a clear guideline for its installation, the financial cost required to construct the piping system could be high because of the need for retrofitting of the current structure (Koppol et al., 2003). One means of overcoming the latter challenge could be though the use of economic instruments such as tax incentives or subsidies; either of which will further increase the overall cost of adoption from the government’s perspective. However, the cost for piping installation could be reduced if the DWWTP is implemented in a new industrial park, as the second pipe system would be installed concurrently with the traditional potable water piping system. The issues regarding separate piping would also impact the use of this non-potable treated wastewater for other users outside of the IPs such as for commercial or irrigation use if longer piping systems were required, as it leads to higher piping cost costs.

A further technical consideration is the identification of the industrial activities within each IP. Although the identified treatment train was able to meet the established limits, the system is sensitive to some contaminants that could be discharged within an IP such as selenium and fluoride. Both elements have been noted to be issues for industrial activities such as

138

electronic, rubber, and stainless steel and copper coating manufacturers. Interestingly, electric and electronic manufacturing has been identified as one of Malaysia’s National Key Economic Areas (NKEA4), where processes such as the manufacturing of semiconductors, solar modules, LED products, and household appliances have been identified as generating the largest investment for the country (Malaysian Administrative Modernisation and Management Planning Unit, 2018). On the one hand, the increase in these types of industrial activities indicate large potential for wastewater reuse due to the inherently high volume of water consumed and consequently discharged by the manufacturing processes, assuming limited internal reuse by the premises. On the other hand, the increase of these types of manufacturing processes also increases the risk of system failure to adhere to the reuse limits in the event that premises struggle to comply with the discharged limit utilised in this study. For example, in the case of the manufacturing of photovoltaic solar cells, the wastewater generated has been noted in the literature to contain a high concentration of fluoride (500- 2000 mg/l) (Drouiche et al., 2013), which is several hundred times higher in magnitude than the discharge limit considered in this research (5 mg/l). In the event that premises do not comply with the 5 mg/l discharge limit due to possible internal wastewater treatment plant malfunction or breakdown, the treatment train proposed would not be able to comply with the limit, assuming the removal efficiency is maintained as modelled.

It has been noted that many small and medium scale industries have difficulty in complying with the specified discharge limits, leading to higher contaminant concentrations (e.g. cyanide, iron, and mercury) being discharged into surface water (Muyibi et al., 2008). This could be due to the limitation faced by the firm such as limited access to laboratories for consistent and continuous monitoring of their wastewater (Rajaram and Das, 2008), or that firms prefer to pay the fines imposed rather than adhere to the regulation due to the lower cost than a technology upgrade (Muyibi et al., 2008). This issue then highlights the criticality of stringent enforcement and consistent monitoring of compliance of discharge limits by industrial premises; one means of doing so is through the use of a real time monitoring systems implemented alongside the development of the wastewater treatment facility. The overall cost and structure was not considered in the calculation of the unit cost of treated wastewater in the present study, pointing to a need for further research to incorporate its

4 defined as “drivers of economic activity that has the potential to directly and materially contribute a quantifiable amount of economic growth to the Malaysian economy” (Malaysian Administrative Modernisation and Management Planning Unit, 2018)

139

potential cost into the overall CAPEX and OPEX estimation. The costs and complexity of such monitoring systems could be a further barrier to industry players and policy makers.

Additionally, it could be seen that more in-depth, site specific studies would need to be done in order to determine the precise circumstance within each park in order to further optimise the treatment train identified in this study, as well as to determine possible contingency plans in the event of possible non-compliance of premises with discharge limits. In this instance, availability of higher resolution data regarding industrial activities within an IP could further assist in optimising the treatment train’s capability to comply with the reuse limit without the use of company specific data, further increasing the accuracy of the initial assessment of its feasibility from a top-down perspective. This, in turn, could provide the funding body with the information to obtain a more realistic cost estimate associated with its deployment. However, through the experience gathered during this research, as well as those experiences highlighted in the literature in previous chapters, higher resolution data regarding industrial activities coupled with the geographical location of the premises is typically not available, due to data confidentiality, as well as the high cost of capital investment of both the infrastructure, data collection and validation process for the latter. In this case, the development of other top-down methodologies to relate operational data with that of spatially-related information such as the coefficient methodology utilised in this research would aid in further bridging this gap. However, as illustrated in this study, further development is required to increase accuracy

Comparatively, the New WSS and the New Ponds projects discussed earlier in this chapter could be perceived as more feasible to implement than the DWWTP. A key influencing factors could be the water quality itself as, compared with the DWWTP, both projects offer the option of potable water. This, in turn, translates to higher acceptance by the industry players since they are not required to make any changes in their manufacturing process. Furthermore, as both of these two projects are essentially based on the same principle as conventional supply augmentation via large storage (reservoir for New WSS, old mining pond conversion for New Ponds), the technical expertise required to implement either projects should already be available in the country though for New Ponds, additional technical requirement might be essential to reduce the probability of heavy metal accumulation in the stored raw water. Additionally, the unit cost of produced water would be the same as the current water tariff as the processes from both projects would be an extension of the existing potable water supply system. In the case where the current water tariff is

140

increased to reflect full cost of operation from the current 78% (Tan, 2012), the revised tariff for Selangor would change to MYR 2.77 – a rate lower than the estimated unit cost of the treated IP wastewater.

In terms of timeline, the implementation of New WSS project would take considerably longer time to execute due to the environment, social, and land issues, as well as additional time needed to collect the raw water to be stored in the system. Comparatively, the New Ponds project, though encountering similar challenges regarding land issues and raw water accumulation, would require substantially less time to implement than New WSS. According to a published report for New Ponds, the first construction phase of the project of refurbishing three existing mining pond started early 2014. However, it was further reported that there was a significant delay in its completion due to issues regarding land acquisition from the State government (Selangor State Assembly, 2014). Furthermore, it was also reported that the two ponds that have been completed did not managed to supply the predicted quantity of raw water due to a lack of groundwater (Selangor State Assembly, 2015), thus further delaying or limiting its contribution. As of 2017, it was reported that the third phase is still under construction and is estimated to be completed by end of 2018, while the completed two ponds are now supplying approximately 190 MLD of raw water of the expected 300 MLD (Selangor State Assembly, 2017).

Additionally, a similar risk to the DWWTPs with regards to the non-compliance with quality standards could also be encountered by both the New WSS and New Ponds projects due to other non-point pollution along the river network due to illegal or other anthropogenic activities. These have been reported by the general media in the country previously, resulting in interrupted water supply to users (e.g. Leoi Leoi, 2015; Adilah and Menon, 2016; The Straits Times, 2016). Furthermore, low flow incidents leading to reduced dilution capability could also cause both of the latter systems to be compromised, resulting in the shutdown of the corresponding water treatment plant (National Water Service Commission (SPAN), 2015).

From the techno-economic perspective, it can be argued that the New WSS and New Ponds projects would be more favourable to the policy makers, with implementation of New Ponds being slightly more advantageous due to the shorter construction timeframe. It was shown in this research that although the techno-economics of the DWWTP treatment train itself can be sound and attractive, a variety of additional technical considerations surrounding its implementation (quality of input water, monitoring, dual piping etc.) could discourage

141

implementation. A summary of the techno-economic comparison between the three projects is given in Table 5-3 below.

Table 5-3: Comparison of techno-economic feasibility between DWWTP, New WSS Project, and New Ponds Project DWWTP New WSS Project New Ponds Project Water quality Non-potable Potable Potable Capital Medium High High investment Unit cost of 2.80 2.805 2.80 produced water (MYR/m3) Technical Medium to high due to Low due to Low to medium as the difficulty several advance unit conventional processes option is only processes in treatment already installed in the providing raw water train, and optimization country storage, similar to of process conventional reservoir, though a higher technical knowledge is required to reduce risk of heavy metal contaminant to the stored raw water Likelihood of Medium to low due to High – requiring no High – requiring no adoption by additional installation change to current change to current industry of dual piping system, manufacturing process manufacturing process players as well as non-potable reuse option Length of Short to medium due Long due to more Short to medium due construction to complexity of dual stringent to potential issue phase piping installation in environmental, social, regarding land

5 Assuming 22% increase from current industrial water tariff of MYR 2.27

142

existing parks, and and land issues, acquisition and additional optimization securing CAPEX prior accumulation of raw process of treatment to construction, water storage train construction of reservoir itself, and accumulation of raw water Risk of system Low to medium, Low to medium, Low to medium, failure depending on industry depending on effluent depending on effluent premises’ ability to released from non- released from non- comply to wastewater point sources, and low point sources, and low discharge limit flow incident leading flow incident leading to reduced dilution to reduced dilution capability of the river capability of the river

ii. Environmental perspective

The stringent enforcement and monitoring required of the reused wastewater also resulted in another long-term benefit for the government, specifically environmental benefits achieved through avoiding the discharge of pollutants in the wastewater effluent to surface waters, as illustrated from the impact assessment in Chapter 4. As previously highlighted, there are still incidents of industrial effluent discharged into the rivers feeding into the water treatment plant, leading to water disruption incidents. To a degree, the freshwater and marine ecotoxicity impact scores in the LCA illustrate the reduction in these impacts by the removal of the heavy metal contaminants from the wastewater. However, the present research has almost certainly not managed to capture the full extent of how these savings might impact the overall water quality of the river; it is recommended that this is further examined in future work.

On the other hand, due to the higher electricity consumption required by the advanced unit processes in the treatment train, it was observed that the global warming potential (GWP) score for this option was significantly higher than the potable water produced via conventional means. Comparatively, the use of the non-potable water reuse resulted in a 3 GWP score of 3.18 kgCO2eq/m , while potable water processed via conventional means

143

3 generated a GWP score of 0.0007 kgCO2eq/m (Ecoinvent 3.3, 2014). In this case, the higher GWP score is attributed to the higher contribution from the use of coal in Malaysia’s grid mix. In a ‘best case’, hypothetical scenario of the conversion of coal to nuclear power, coupled with further optimization of electricity consumption of the treatment train, the GWP score 3 was calculated to be 1.07 kgCO2eq/m . However, that GWP score was still well above that for conventional treatment for potable water.

The higher GWP score could impact the decision of both the policy makers as well the industry players, especially if the targets set by both parties’ centres on the reduction of their carbon-related emissions. Taking Malaysia as an example, the government has pledged to reduce the nation’s greenhouse gas emission by up to 40% by 2020. Although that is the case, the implementation of the treatment train would increase electricity demand, and indirectly, the greenhouse gas emissions (GHG) of the industrial sector as a whole by approximately 2%6, assuming the energy consumption from other processes remains constant. This suggests that if the government adopts industrial wastewater reuse, effort to align this strategy with those relating to energy management such as those focusing on energy efficiencies would need to be done.

Additionally, from a firm’s perspective, the use of the treated wastewater might increase the carbon footprint of their products coming from water use by a factor of three, depending on the volume of substituted water used during their production. As the use of carbon footprint labelling is one of the more common means to communicate a firm’s or organisation’s effort regarding implementation of sustainable consumption and production (SCP) practices, the increased GHG could further deter them from integrating the use of the treated wastewater. For example, aligning this with the United Nations Sustainable Development Goal (UN-SDG)

7 8 9 , the use of CO2 emission per unit of value added is specifically used to measure target 9.4 .

Comparing the DWWTP to the New WSS and New Ponds projects, the hotspots for both projects are not linked with their electricity use as both remain relatively passive during their operational phase. For New WSS specifically, it has been noted that the carbon and methane emissions associated with its GWP are related to its construction due to the flooding of the

6 i.e. total electricity consumption by the industrial sector in 2016 = 5,822 ktoe / 6.77E7 kWh, and total electricity required for treatment train implementation = 111.78 ktoe/1.3E6 kWh 7 SDG goal 9: Built resilient infrastructure, promote inclusive and sustainable industrialization and foster innovation 8 9.4 target: By 2030, upgrade infrastructure and retrofit industries to make them sustainable, with increased resource-use efficiency and greater adoption of clean and environmentally sound technologies and industrial processes, with all countries taking action in accordance with their respective capabilities

144

above and below ground biomass (Deemer et al., 2016). However, it was also highlighted that its indirect, social implication is also substantial due to the relocation of human settlements, as well as land use change and loss of access to natural resources (Vio De, Ribeiro and Anderi Da Silva, 2010; Zarfl, Lumsdon and Berlekamp, 2014).

For New Ponds, there are currently limited studies which assess the environmental impact of its use, highlighting the need for a more in-depth work regarding this issue; the environmental impact of this project is discussed here from a theoretical perspective. Although the concept of supply augmentation via ponds is similar to a conventional reservoir, the aforementioned issues regarding the environmental impact associated with the construction of a new dam might are probably not applicable here as the ponds are essentially refurbished mining ponds with small presence of terrestrial organic content and no human settlement issues during its conversion. Additionally, depending on the allocation between the previous mining activities and the water augmentation scheme, any environmental impacts from the construction of the ponds could be small or insignificant, when the lifespan of the mining and supply augmentation activities are taken into consideration. Instead, the impact would be from the construction of the piping and pumping system (both materials and activities), the dredging of the pond to increase its depth, and transportation of materials and sand into and out of the construction site, as well as any energy used during its operational phase (e.g. energy for aeration). The heavy metals content in the raw water stored was assessed prior to its use and the levels found to be well below the Malaysian limit for untreated raw water for potable water production so no long term adverse impact is indicated from its use (Madzin, Fitri Shai-in and Mohd Kusin, 2015; Kusin et al., 2016; Mohd Kusin et al., 2016). This, in turn, would not greatly affect the toxicity-related impact categories for the New Ponds project due to both the lower heavy metal concentration, as well as further dilution of its concentration via river to the WTP.

As such, it can be concluded that, strictly from a perspective of carbon-related emission, the use of water for the New WSS and New Ponds projects is more beneficial for both the industry players and policy makers. However, if other considerations are taken into account – such as land use change and social implications – the use of the DWWTP might be more advantageous as it avoids the need for intensive land exploitation as well as having minimal social impact. Furthermore, it would also contribute towards realising another SDG goal –

145

specifically goal 129 – where the industry players are encouraged to reduce their hazardous waste generation to reduce impact to human health.

iii. Awareness

DWTTP was illustrated in the previous subsections to reduce the amount of freshwater used by manufacturing industry, and was able to provide a buffer time to prolong the need for SSM strategy implementation. On the other hand, it was also argued that its use had increased the carbon-related emissions due to the increased use of electricity. As such, it points to the need to increase the scope of the general public’s – and, to a degree, the policy maker’s – awareness regarding a more inclusive view of environmental performance indicators. In this case, the use of other types of environmental labelling beyond those centred on GHG emissions (i.e. carbon footprint labelling, CFP) should be emphasised and made aware to the public and industry players. Mechanisms of eco-labelling schemes, such as EU flower (EU), Nordic swan (Scandinavia), Blue Angel (Germany) (Horne, 2009), and Eco-label (Malaysia) should be emphasised instead as it considers a broader scope of criteria beyond those focusing on carbon-related emissions in the document relating to product certification. In this instance, the use of DWWTP would have an advantage over the use of both New WSS and New Ponds, as allows the firms to reduce their freshwater consumption, as well as reducing their wastewater discharge into the river. Additionally, the campaign would also need to emphasise the indirect benefits of its adoption such as access to a consistent and reliable water source – from both the quality and quantity aspect – especially during time of water rationing and cuts (Dickinson et al., 1995). For this specifically, it could be seen as a large benefit for the industry players especially those relying heavily on water use during their manufacturing process as it would impact their productivity and production cost (MIDA, 2014). As such, in order to decrease their risk against instances of water rationing, the industry players might be willing to purchase the treated wastewater despite the higher cost of water supply

Although that is the case, through the awareness campaign highlights the abovementioned advantages, it does not guarantee its effectiveness; for example, Adham, Merle and Weihs

9 Sustainable consumption and production, target 12.4: by 2020, achieve the environmentally sound management of chemicals and all wastes throughout their life cycle, in accordance with agreed international frameworks, and significantly reduce their release to air, water and soil in order to minimize their adverse impacts on human health and the environment

146

(2013) noted that despite conducting multiple awareness campaigns and activities to educate the public, consumers and businesses lack the right mind-set to support the government’s objectives for SCP implementation. In this instance, understanding the drivers of industry players to adopt the use of treated wastewater for non-potable reuse could be the key to tailor the awareness campaign to further increase their uptake of the option; for example, with regards to green manufacturing, it has been noted that larger industry players would react positively towards its adoption as compared to SME as it offer the former a competitive edge while not the latter (Seth, Rehman and Shrivastava, 2018). As such, it can be observed that the implementation of the New WSS and New Ponds project could be seen as a more favourable alternative solution to mitigate the current water imbalance issues as compared to the DWWTP as it does not significantly rely on the effectiveness of an awareness campaign.

iv. Institutional perspective

Another aspect which has been noted in various literature highlighted in Chapter 1 is the institutional structure surrounding the implementation of such schemes, both from the bottom-up and top-down direction. Here, existing policies are reviewed in order to gauge the readiness of existing government institutions to adopt this solution: however, since policy are not the focus of this research, only a general review was conducted.

From the top-down standpoint, although there are currently existing strategies, policies, and frameworks enabling the implementation of the DWWTP, it was observed that its integration spans two set of stakeholders, each with their own set of interests. It is possible to view the relevant stakeholders as water-centric stakeholders, focusing on water resource management, and industrial-centric stakeholders who concentrate on increasing the industrial sector’s productivity. Due to the difference in interests, existing strategies, policies, and frameworks formulated by both set of stakeholders can fall into different themes; Sustainable Consumption and Production (SCP) for the industry-related stakeholders, and water demand management (WDM) for water-related stakeholders. In order to compare existing policies and frameworks designed by both stakeholders, the reports published by Akedemi Sains Malaysia regarding the National Integrated Water Resource Management Plan (Abdullah et al., 2016; Akademi Sains Malaysia, 2016) were used as a reference for the water-centric stakeholders, while the Sustainable Consumption and Production in Malaysia report (Adham, Merle and Weihs, 2013) was referred for the industry-centric stakeholders.

147

For the water-centric stakeholders, it was observed that the strategies planned for WDM are focused on reducing the non-revenue water; water tariff adjustment; the promotion of reducing, recycling, and reuse of water; and adoption of technologies and incentives to achieve more efficient use. For water use reduction, recycling, and reuse specifically, it was observed that it was still considered to be in the R&D stage with limited integration into existing policies; this is even more prominent with regards to industrial wastewater reuse and recycling. The source of wastewater considered was limited to domestic wastewater without consideration of industrial wastewater, though the target end user and use remains the same (i.e. non-potable industrial reuse). This could be due to the limited information regarding industrial wastewater processing and reuse. For example, as observed in the Malaysia DoE’s Environmental Quality Act, the regulating bodies place a much higher emphasis on the quality of the discharged wastewater (which is heavily monitored) as compared with the quantity (quantity of effluent is scarcely monitored beyond flowrate). Additionally, as compared with domestic wastewater where the collection and treatment is more centralised, the discharge of industrial effluent is dispersed and occurs outside the network of the existing domestic wastewater treatment system. As such, the information regarding the potential volume of wastewater effluent which could be recycled and reused is not readily accessible, requiring further effort from the policy makers to capture its overall potential for non-potable industrial wastewater reuse. It seems, therefore, rather overlooked as a source for industrial reuse.

For the industry-centric stakeholders, it was observed that policies identified regarding SCP were observed to be more inclined towards increasing production efficiencies. However, it was observed that policies formulated placed a greater emphasis on decreasing raw material and energy consumption, rather than reducing potable water use. Polices relating to water resource specifically were formulated with regards to controlling pollution discharge, as exemplified in Figure 5-1 below, extracted from the National Policy on the Environment (NPE).

Figure 5-1: Extract on policy for water pollution in the National Policy on the Environment (Adham, Merle and Weihs, 2013)

148

In contrast, although policies formulated had emphasise water use efficiencies to decrease water consumption, it was found that corresponding incentives to be non-specific, without clear definition, action plans, or goals, as compared to energy, where a more formal structure and clear direction towards increasing its efficiencies was observed; Table 5-4 below illustrates the differences in several of the policies formulated. The stronger emphasis on increasing energy efficiency as compared with water could be due to the government’s commitment to reduce CO2 emissions.

Table 5-4: Policies regarding water and energy efficiency (extracted from Adham, Merle and Weihs (2013))

Developing a long-term strategy for water resource management to achieve water security (10MP10, p.281) Promote culture of conservation and efficiency in energy and water use (10MP,

p.132)

... review tax incentives, such as tax breaks for buildings and designs that are Water environmentally friendly, incorporating green design elements like solar panels for heating, rain water harvesting facilities and water conservation features (10MP, p.279) Bonus feed-in tariff rate criteria [energy efficiency] ... Use of gas engine technology with electrical efficiency of above 40% … Use of steam-based electricity generating systems with overall efficiency of above 14% (REA11, p.5145) Setting of minimum energy performance standards [(MEPS)] for appliances and development of green technologies (10MP, p.113) Developing and enforcing regulations especially on energy efficiency in buildings for new developments (10MP, p.132)

The Government will make a move to gradually rationalise the subsidy on electricity Energy to create an incentive for both industries and consumers to adopt more energy- efficient practices (ETP12, p.190) To this end, a green public procurement policy shall be put in place by October 2011, to give preference to local producers, establish buying guidelines for eco- labelled products and specify the required energy efficiency certification for specific products (ETP, p.417)

10 10MP – 10th Malaysian Plan 11 REA – Renewable Energy Act 2011 12 ETP- Economic Transformation Programme

149

Interestingly, Adham, Merle and Weihs (2013) had highlighted an existing financial incentive to assist in the implementation of wastewater reuse, where firms undertaking wastewater recycling activities are eligible for tax allowance; specifically the Waste Eco Park (WEP) incentive offered by MIDA. However, in the guideline pertaining to claiming of this incentive, there were no specifications regarding the minimum target required of the applicant, only general statement of the applicant needing to incorporate the use of a wastewater treatment facility (MIDA, 2016). In comparison, a clear maximum limit of total processed waste disposed to landfill was defined (i.e. 30%). The differences in specification further accentuate the shortcomings above as there were no specific targets which need to be met by the policies relating to SCP implementation regarding reduction of potable water use. As such, it could be inferred that the economic incentives offered were only used as a tool to encourage water savings behaviour by the industry players, rather than as a means to aid reducing their dependencies on potable water.

As such, it can be seen that although the main objectives of both stakeholders are geared towards sustainable water use, policies formulated by both set of stakeholders was aligned to their respective interest. It was observed that those formulated by the water-centric stakeholders had focused on increasing the supply with minor incorporation of the action from the users (demand), while industry-centric stakeholders had concentrated on reducing the demand without assessing its impact on the supply. In this sense, further integration of existing policies from both the WSM and SCP perspective would need to be done in order to further drive the uptake of industrial wastewater reuse as a means to alleviate water imbalance. From the WSM perspective, involvement of water-centric stakeholders is crucial to synchronise savings from DSM-based strategy to large scale SSM system to monitor and maintain the overall water balance of a catchment. While from the SCP perspective, involvement of industry-centric stakeholders is essential to assist in the communication and dispersion of information to the industry players to increase the likelihood of its uptake.

Additionally, the introduction of a management body to bridge the gap between the two sets of stakeholders focusing on the implementation of the DWWTP in the IPs itself might be ideal. From the bottom-up perspective, the identification and formation of an organizational or management setup which focus on the operation and maintenance of the decentralised facility, as well as the synergy between players within the IPs and the policy makers is essential to ensure effective implementation. This concept is similar to those underlined with regards to the execution of eco-industrial park (EIP) such as those noted by Behera et al.

150

(2012), Cawley (2016), and Taddeo (2016), where they noted the importance of a champion or manager responsible for the implementation of an EIP to provide a strong direction for and facilitated good communications between all parties involved.

For Malaysia, this concept is similar to those advocated through the Waste Eco Park (WEP) initiative offered by MIDA, where an appointment of a WEP operator is required to oversee the operation and maintenance of the park (MIDA, 2016). However, it was observed that the initiative that is offered to a single entity overseeing the operation in a single IP, which could result in inconsistent degree of efficiency and quality of supply to the consumers if multiple IPs are established; a problem encountered by the water service industry in Malaysia. In this case, the existing structure pertaining to the current potable water management could be replicated, where a singular company or organisation is tasked to oversee its operation and maintenance. Furthermore, the existence of a singular entity would be more effective in its liaison between both set of stakeholders, as well as collaborating with the state government and local council. A simplified diagram summarising the players and general scope of responsibility for each of the set of stakeholders are illustrated in Figure 5-2 below.

In addition to the changes required of the current institutional structure, significant expansion of the current guidelines would also need to be done. In terms of the non-potable reuse option, currently, there is no guideline, specifying the technical requirements of its use regarding issues such as quality, restriction of use, delivery system, monitoring, and risk assessment. For example, the component of dual piping system required to maintain separation between potable and non-potable water is not typically installed in Malaysia (Oh et al., 2018) and should be specified in the guideline. Another aspect which requires further consideration is the reuse water quality itself, where limits regarding various parameters similar to those compiled in this thesis (Appendix D) would need to be considered. In this case, guidelines that have been established internationally such as those utilised in Australia, US, Israel (Garcia and Pargament, 2015), and China (Lyu et al., 2016), or a recently published ISO standard regarding centralised urban water reuse system (ISO 20760-1:2018) could be used as basis for its development.

The above discussion indicates that there are considerable changes required of the current organizational structure and subsequent policies and framework. In comparison with the implementation of the New WSS and New Ponds projects, it could be concluded that the adoption of these two projects are highly likely to be preferred because this utilises the same

151

organizational structure as the current water service industry, requires few changes and is familiar to implement from the government side as well as the industry players. Furthermore, as both of the two projects supply potable water, no changes are required of existing guidelines regarding potable water production. A summary of comparison of the organizational structure between the three projects is given in Figure 5-2 below.

Table 5-5: Comparison of organizational, policy, and guideline requirements between DWWTP, New WSS Project, and New Ponds Project

DWWTP New WSS Project New Ponds Project High - requires the formation of a new Low - utilises Low - utilises management structure of centralised existing existing institutional operation and management of the institutional structure

facilities, integration of two separate structure change

set of stakeholders to support new Organizational management structure,

High, as existing policies would need Low, requiring no Low, requiring no

to be streamlined changes to the changes to the Policy Policy change current policies current policies High, requiring the formation of new Low, requiring no Low to medium, guideline regarding standards of non- change to existing possibly needing to

potable water reuse, dual piping guidelines and upgrade the current change Guideline Guideline system standards raw water standard

152

Figure 5-2: Illustration of the synergy between the water-related stakeholders, industry related stakeholders, and the industrial parks for DWWTP implementation

153

5.3 Conclusion

This research has assessed the viability and the trade-offs in using DWWTP to address water imbalance issues, and the requirements necessary to ensure effective implementation. Based on the above arguments, it can be concluded that, although the proposed DWWT can be technically, financially, and environmentally feasible to aid in alleviating the water supply imbalance issue, its implementation is improbable at a larger scale against the background of the current existing outlook and organizational structure, and it would likely be discounted over other supply-side management strategies. This is primarily due to the other considerations outside of the execution of the option itself.

For both Malaysia and other countries facing similar situations e.g. Thailand, Indonesia, and Cambodia, the option of deploying the DWWTP will likely be conditional on their priorities, which can be linked to two SDGs – Goal 6: clean water and sanitation, and goal 12: sustainable consumption and production. Although this research illustrates the capability of deploying the DWWTP to assist in achieving the targets set by goal 6, it has also illustrated here that it requires substantial effort in ensure effective implementation – efforts which are either not currently being considered, or are currently being considered but require time to fully execute.

In this sense, countries which need to focus on providing adequate water supply to its people as well as to ensure that there are adequate supplies of water for continued industrial growth will tend to favour (where possible) large sale supply augmentation projects similar to those of New WSS or New Ponds, or the use of desalination projects for countries with restricted or limited access to other types of raw water sources. The choice of supply augmentation projects, though requiring higher capital cost and coming with relatively high environmental damage, enables the countries to secure a large volume of raw water for a long period of time, thus ensuring uninterrupted water supply for continued economic growth. Additionally, the implementation of the options discussed in this chapter requires effort only by the existing stakeholders governing water resource management in the country. As compared with the DWWTP strategy, which, at its core is more in line with targets and thinking of goal 12, a more dispersed effort is required over a larger set of stakeholders to drive the necessary organizational change and awareness campaign.

154

Chapter 6 : Conclusions and future work

6.1. Review of findings

Natural endowment of plentiful freshwater from rainfall does not necessarily assure a continuous water supply to consumers. As a country, such as Malaysia, moves away from dependency on its agriculture sector, the industrial sector is expected to consume higher volumes of water. Furthermore, the industrial sector and the population will continue to converge in the major cities, leading to a rapid increase in water use and demand, creating further stress on water resources (Adnan, 2013).

This research attempted to assess the feasibility of deploying decentralised wastewater treatment plants on a large scale for industrial wastewater reuse as a strategy to alleviate this water imbalance issue. Its practicability was assessed from the standpoint of the technical, economical, and environmental feasibility of the DWWTP deployment, as well as the overarching financial resources, and administrative and social measures required for its implementation.

From the first standpoint, it was illustrated that, although implementation of SSM strategies are unavoidable in the long-run, integration of DSM-based strategies could be implemented alongside the SSM-based strategies to delay the need for WSS expansion as they are able to decrease water imbalance issue of an existing water supply system. Furthermore, assessment of the financial, technical, and environmental merits of the DWWTP deployment was found to be comparable to other SSM-based strategy (i.e. conventional WSS of a dam and WTP, and raw water retention ponds).

The external financial resources, and administrative and social measures required for its implementation was further explored in Chapter 5. Here, it was argued that although the identified treatment train itself was feasible, additional efforts required for its implementation would be substantial as compared to other SSM-based strategies. Several challenges were highlighted, specifically those focusing on the additional financial resources required to support its implementation, as well as the awareness and organizational inclusiveness from multiple stakeholders. It was found that, although the implementation of the DWWTP itself could be achieved if these constrains were addressed, it would take substantial time and effort to realise, which might deter the policy makers to adopt this strategy as a means to mitigate water imbalance issues.

155

As such, it can be concluded that, although the execution of DWWTPs at a large scale is able to reduce an existing WSS’s vulnerability and is feasible in terms of it technical, financial, and environmental merits, its deployment however, is improbable against the current organizational structure and current awareness of the users.

6.2. Wider implication of research

Although this study used Selangor as a case study, the findings generated from this study can be extrapolated to other countries of similar economic and geographic characteristics as they will likely face similar problems as highlighted throughout this thesis. In terms of urban water supply, the increased intensity of the manufacturing sector and subsequent population migration is expected to increase pressure on the existing water supply system to provide adequate water supply to its users. This, in turn, points to a need for a more efficient water resource consumption.

For the manufacturing sector specifically, water minimization could be encouraged through the implementation of green manufacturing, where it encourages the use of strategies and techniques to increase eco-efficiency (Deif, 2011). In terms of green manufacturing and its relation with water use minimization, it is often given less priority due to the relatively low cost of water as compared to other inputs in the overall value chain (i.e. energy and material consumption). Although the increase in water tariff could be a driver of green manufacturing implementation by the industry players, it also has the added risk of business relocation to other neighboring countries which offers a more competitive tariff and tax incentives as a strategy to attract FDIs into their country (Shuhaimen et al., 2017). In this aspect, this research offers a means of assessing the implementation of DWWTP as a means of encouraging the uptake of green manufacturing. As it was designed here to have minimal financial and technological requirements from the users, it can be seen as a low risk strategy for the industrial players to increase their corporate green image. Although this comes at a higher financial cost to the government, this could provide a better incentive to the industry players to avoid potential relocation, as it offers for a more consistent water source which is not influenced by water disruption incidents, as well as potentially attracting other investors who place a higher priority on increasing their green corporate image via green manufacturing implementation.

156

The issue of increased carbon-related emission typically associated with electricity utilization is also of concern, both from the policy maker’s as well as the manufacturers’ perspectives. Internationally, a greater emphasis has been given to energy reduction due to it is associated with carbon related emissions and its impact of a more global scale. In contrast, water related issues are more local in nature. Due to this, the policy makers are now faced with another tradeoff of prioritizing global versus local interests. In this instance, the implementation of DWWTP can be argued to go against the international interest of reducing carbon-related emissions, due to the higher electricity consumption, though offering benefits such as those highlighted throughout this thesis. With respect to this, this research offers a means of assessing the tradeoffs between water savings and energy consumption for the implementation of DWWTP and other supply augmentation projects.

6.3. Future research avenues

One of the main strength of this research is the utilisation of existing data that is typically collected by data scarce countries to estimate the industrial water consumption and subsequent wastewater generation via the use of a coefficient. However, due to the low resolution of the data used in this study, it resulted in large uncertainties during its estimation. Although the use of the coefficient would greatly assist in a coarse estimation of potential water savings, it would be beneficial to increase the accuracy of the method so that a more accurate estimation of the potential water savings could be estimated. In this case, higher resolution data regarding industrial water use which incorporates the Standard Industry Classification (SIC) code could be beneficial as it allows for a more accurate representation of water use according to specific industrial activities. The use of these detailed coefficients by developing countries however could be limited as these data are rarely collected, as exemplified through this research.

The research done around the assessment of reservoir vulnerability also highlights opportunities to incorporate climate change projections to further assess its impact on the water inflow into the reservoir and rivers. While this research provided a means of water balance accounting at a reservoir level, it assumed that the inflow into the reservoirs and rivers would be stationary. In this case, taking into consideration the effect of climate change could provide a more vigorous projection of the raw water supply into the WSS, and could provide a more comprehensive understanding of the WSS vulnerability, both in terms of

157

supplying adequate water to the users as well as the impact of DWWTP implementation of mitigating non-satisfactory conditions.

Finally, this research illustrated the potential of industrial wastewater reuse as a means to increase freshwater availability for other water users; however, it assumed minimal internal reuse or increase in process efficiency for water minimization. Although this was done to take into consideration the current limitation faced by the industry players, future works could incorporate the likelihood of altered behavioural change due to an increase in awareness by the industry players. In this instance, the impact of reduced water consumption by each premise via the abovementioned strategies would need to be incorporated into the methodology presented in this study, to re-evaluate the financial, technical, and environmental feasibility of the combined effort for industrial water use minimization, as well as the changes required of the current organizational structure to ensure effective implementation.

158

References

Abdullah, K. and Mohamed, A. (1998) ‘Water–A situation appraisal and possible actions at the community level’, in Seminar on “Local Communities and the Environment II”, Environment Protection Society Malaysia, Petaling Jaya, pp. 24–25. Abdullah, S., Chand, F., Zakaria, S. and Loganathan, P. (2016) Transforming the Water Sector: National Integrated Water - Resources Management Plan: Strategies and Road Map. Kuala Lumpur, Malaysia. Abou-Elela, S. I., Nasr, F. A., Ibrahim, H. S., Badr, N. M., Raziq, A. and Askalany, M. (2008) ‘Pollution prevention pays off in a board paper mill’, Journal of cleaner production, 16, pp. 330–334. doi: 10.1016/j.jclepro.2006.07.045. Abu Bakar, A. F. and Abdul Halim, A. (2013) ‘Treatment of automotive wastewater by coagulation-flocculation using poly-aluminium chloride (PAC), ferric chloride (FeCl3), and aluminium sulphate (alum)’, AIP Conference Proceedings, 1571(1), pp. 524–529. doi: 10.1063/1.4915854. Adewumi, J. R., Ilemobade, A. A. and Van Zyl, J. E. (2010) ‘Decision support for the planning of integrated wastewater reuse projects in South Africa’, Water Science and Technology: Water Supply, 10(2), pp. 251–267. doi: 10.2166/ws.2010.231. Adewumi, J. R., Ilemobade, A. A. and Van Zyl, J. E. (2010) ‘Treated wastewater reuse in South Africa: Overview, potential and challenges’, Resources, Conservation and Recycling, 55(2), pp. 221–231. doi: 10.1016/J.RESCONREC.2010.09.012. Adham, K. N., Merle, K. and Weihs, G. (2013) Sustainable Consumption & Production in Malaysia: A Baseline Study on Government Policies, Institutions and Practices. 1st edn. Putrajaya: Economic Planning Unit (EPU) Malaysia. Adilah, A. and Menon, J. (2016) Syabas: Pollutants from industrial park caused shutdown of water treatment plant | Malaysia | Malay Mail, Malay Mail. Available at: https://www.malaymail.com/s/1217165/syabas- pollutants-from-industrial-park-caused-shutdown-of-water-treatment-p (Accessed: 5 September 2018). Adnan, H. (2013) Water , Food and Energy Nexus in Asia and the Pacific. Available at: http://www.worldwatercouncil.org/fileadmin/world_water_council/documents/programs_hydropolitics_sdgs/W ater-Food-Nexus Report.pdf. Agana, B. A., Reeve, D. and Orbell, J. D. (2013) ‘An approach to industrial water conservation - A case study involving two large manufacturing companies based in Australia’, Journal of Environmental Management, 114, pp. 445–460. doi: 10.1016/j.jenvman.2012.10.047. Agrawal, N. and Desmukh, T. S. (2016) ‘Rainfall Runoff Modeling using MIKE 11 Nam – A Review’, IJISET - International Journal of Innovative Science, Engineering & Technology, 3(6). Available at: www.ijiset.com (Accessed: 5 July 2017). Ahmed, K. (2012) A Sourcebook of Pollution Management Policy Tools for Growth and Competitiveness. 71608. Washington, DC. Available at: http://siteresources.worldbank.org/ENVIRONMENT/Resources/Getting_to_Green_web.pdf (Accessed: 7 September 2016). Ahmed, S. A., Tewfik, S. R. and Talaat, H. A. (2002) ‘Development and verification of a decision support system for the selection of optimum water reuse schemes’, Desalination, 152, pp. 339–352. Available at: www.elsevier.com/locate/desal (Accessed: 14 March 2018). Akademi Sains Malaysia (2016) Strategies to Enhance Water Demand Management in Malaysia. Kuala Lumpur, Malaysia. Al-Zoubi, H., Ibrahim, K. A. and Abu-Sbeih, K. A. (2015) ‘Removal of heavy metals from wastewater by economical polymeric collectors using dissolved air flotation process’, Journal of Water Process Engineering, 8, pp. 19–27. doi: 10.1016/j.jwpe.2015.08.002. Alnouri, S. Y., Linke, P. and El-Halwagi, M. (2015) ‘A synthesis approach for industrial city water reuse networks considering central and distributed treatment systems’, Journal of Cleaner Production, 89, pp. 231– 250. doi: 10.1016/j.jclepro.2014.11.005.

159

Alturki, A. A., Tadkaew, N., Mcdonald, J. A., Khan, S. J., Price, W. E. and Nghiem, L. D. (2010) ‘Combining MBR and NF/RO membrane filtration for the removal of trace organics in indirect potable water reuse applications’, Journal of Membrane Science, 365, pp. 206–215. doi: 10.1016/j.memsci.2010.09.008. Amores, M. J., Meneses, M., Pasqualino, J., Antón, A. and Castells, F. (2013) ‘Environmental assessment of urban water cycle on Mediterranean conditions by LCA approach’, Journal of Cleaner Production, 43, pp. 84– 92. doi: 10.1016/J.JCLEPRO.2012.12.033. Ansari, R. and Sadegh, M. (2007) ‘Application of activated carbon for removal of arsenic ions from aqueous solution’, E-Journal of Chemistry, 4(1), pp. 109–108. Available at: file://homes.surrey.ac.uk/home/Journals/2018/829187.pdf (Accessed: 26 February 2018). Aoudj, S., Drouiche, N., Hecini, M., Ouslimane, T. and Palaouane, B. (2012) ‘Coagulation as a Post-Treatment Method for the Defluoridation of Photovoltaic Cell Manufacturing Wastewater’, Procedia Engineering Procedia Engineering Procedia Engineering, 33, pp. 111–120. doi: 10.1016/j.proeng.2012.01.1183. Araby, R. El, Hawash, S. and Diwani, G. El (2009) ‘Treatment of iron and manganese in simulated groundwater via ozone technology’, Desalination, 249, pp. 1345–1349. doi: 10.1016/j.desal.2009.05.006. Asano, T., Berton, F. L., Leverenz, H. L., Tsuchihashi, R. and Tchobanoglous, G. (2007) Water reuse: issues, technologies, and applications. First. Edited by L. S. Hager. U.S.A.: Metcalf & Eddy. Asefa, T., Clayton, J., Adams, A. and Anderson, D. (2014) ‘Performance evaluation of a water resources system under varying climatic conditions: Reliability, Resilience, Vulnerability and beyond’, Journal of Hydrology, 508, pp. 53–65. doi: 10.1016/j.jhydrol.2013.10.043. Aviso, K. B. (2014) ‘Design of robust water exchange networks for eco-industrial symbiosis’, Process Safety and Environmental Protection, 92(2), pp. 160–170. doi: 10.1016/j.psep.2012.12.001. Azimi, A., Azari, A., Rezakazemi, M. and Ansarpour, M. (2017) ‘Removal of Heavy Metals from Industrial Wastewaters: A Review’, ChemBioEng Reviews, 4(1), pp. 37–59. doi: 10.1002/cben.201600010. Bank Negara Malaysia Malaysia (2015) Annual Report 2014. Kuala Lumpur, Malaysia. Barbier, E. (2015) ‘Water and growth in developing countries’, in Dinar, A. and Schwabe, K. (Kurt A. . (eds) Handbook of water economics. Edward Elgar Publishing, pp. 500–512. Barillon, B., Ruel, S. M., Langlais, C. and Lazarova, V. (2013) ‘Energy efficiency in membrane bioreactors’, Water Science and Technology, 67(12), pp. 2685–2691. doi: 10.2166/wst.2013.163. Barrington, D. J., Prior, A. and Ho, G. (2013) ‘The role of water auditing in achieving water conservation in the process industry’, Journal of Cleaner Production, 52, pp. 356–361. doi: 10.1016/j.jclepro.2013.03.032. Barrios-Martinez, A., Barbot, E., Marrot, B., Moulin, P. and Roche, N. (2006) ‘Degradation of synthetic phenol- containing wastewaters by MBR’, Journal of Membrane Science, 281, pp. 288–296. doi: 10.1016/j.memsci.2006.03.048. Bates, B. C., Kundzewicz, Z. W., Wu, S. and Palutikof, J. P. (2008) Climate Change and Water, Climate change and water. doi: 10.1016/j.jmb.2010.08.039. Behera, S. K., Kim, J.-H., Lee, S.-Y., Suh, S. and Park, H.-S. (2012) ‘Evolution of “designed” industrial symbiosis networks in the Ulsan Eco-industrial Park: “research and development into business” as the enabling framework’, Journal of Cleaner Production, 29, pp. 103–112. doi: 10.1016/j.jclepro.2012.02.009. Bell, R. G. and Russell, C. (2002) ‘Environmental policy for developing countries’, Issues in Science and Technology, 19(3), pp. 63–70. Available at: file://homes.surrey.ac.uk/home/Journals/2017/Bell and Russell, 2002 - Environmental policy for developing countries.pdf (Accessed: 24 June 2017). Bennett, A. (2005) ‘Membranes in industry: facilitating reuse of wastewater’, Filtration & Separation, 42(8), pp. 28–30. doi: 10.1016/S0015-1882(05)70658-3. Berglund, E. Z. (2015) ‘Using Agent-Based Modeling for Water Resources Planning and Management’, Journal of Water Resources Planning and Management, 141(11), p. 04015025. doi: 10.1061/(ASCE)WR.1943- 5452.0000544. Bernal-Martínez, L. A., Barrera-Díaz, C., Solís-Morelos, C. and Natividad, R. (2010) ‘Synergy of

160

electrochemical and ozonation processes in industrial wastewater treatment’, Chemical Engineering Journal, 165, pp. 71–77. doi: 10.1016/j.cej.2010.08.062. Beveridge, R. and Monsees, J. (2012) ‘Bridging parallel discourses of Integrated Water Resources Management (IWRM): institutional and political challenges in developing and developed countries’, Water International, 37(7), pp. 727–743. doi: 10.1080/02508060.2012.742713. Bhattacharjee, S., Datta, S. and Bhattacharjee, C. (2007) ‘Improvement of wastewater quality parameters by sedimentation followed by tertiary treatment’, Desalination, 212, pp. 92–102. doi: 10.1016/j.desal.2006.08.014. Billings, R. B. and Jones, C. V. (2011) Forecasting urban water demand. American Water Works Association. Binz, C., Razavian, N. B. and Kiparsky, M. (2017) ‘Of Dreamliners and Drinking Water: Developing Risk Regulation and a Safety Culture for Direct Potable Reuse’, Water Resources Management, pp. 1–15. doi: 10.1007/s11269-017-1824-1. Bixio, D., Thoeye, C., Wintgens, T., Ravazzini, A., Miska, V., Muston, M., Chikurel, H., Aharoni, A., Joksimovic, D. and Melin, T. (2008) ‘Water reclamation and reuse: implementation and management issues’, Desalination, 218(1–3), pp. 13–23. doi: 10.1016/J.DESAL.2006.10.039. Blackman, A. (2009) Alternative Pollution Control Policies in Developing Countries Alternative Pollution Control Policies in Developing Countries: Informal, Informational, and Voluntary. Washington, D.C. Available at: https://core.ac.uk/download/pdf/9304592.pdf (Accessed: 29 June 2017). Boix, M., Montastruc, L., Pibouleau, L., Azzaro-Pantel, C. and Domenech, S. (2012) ‘Industrial water management by multiobjective optimization: from individual to collective solution through eco-industrial parks’, Journal of Cleaner Production, 22, pp. 85–97. doi: 10.1016/j.jclepro.2011.09.011. Bolzonella, D., Fatone, F., Di Fabio, S. and Cecchi, F. (2010) ‘Application of membrane bioreactor technology for wastewater treatment and reuse in the Mediterranean region: Focusing on removal efficiency of non- conventional pollutants’, Journal of Environmental Management, 91, pp. 2424–2431. doi: 10.1016/j.jenvman.2010.07.024. Bonton, A., Bouchard, C., Barbeau, B. and Jedrzejak, S. (2012) ‘Comparative life cycle assessment of water treatment plants’, Desalination, 284, pp. 42–54. doi: 10.1016/J.DESAL.2011.08.035. Brik, M., Chamam, B., Schöberl, P., Braun, R. and Fuchs, W. (2004) ‘Effect of ozone, chlorine and hydrogen peroxide on the elimination of colour in treated textile wastewater by MBR’, Water Science and Technology, 49(4), pp. 299–303. Brik, M., Schoeberl, P., Chamam, B., Braun, R. and Fuchs, W. (2006) ‘Advanced treatment of textile wastewater towards reuse using a membrane bioreactor’, Process Biochemistry, 41, pp. 1751–1757. doi: 10.1016/j.procbio.2006.03.019. Bruins, J. H., Vries, D., Petrusevski, B., Slokar, Y. M. and Kennedy, M. D. (2014) ‘Assessment of manganese removal from over 100 groundwater treatment plants’, Journal of Water Supply: Research and Technology - AQUA, 63(4), pp. 268–280. doi: 10.2166/aqua.2013.086. California Department of Public Health (2014) ‘Regulations Related to Recycled Water’. California, p. 81. Available at: https://www.waterboards.ca.gov/drinking_water/certlic/drinkingwater/documents/lawbook/RWregulations_201 40618.pdf (Accessed: 23 December 2017). Cantin, B., Shrubsole, D. and Aït-Ouyahia, M. (2005) ‘Using Economic Instruments for Water Demand Management: Introduction’, Canadian Water Resources Journal, 30(1), pp. 1–10. Available at: http://www.tandfonline.com/doi/pdf/10.4296/cwrj30011?needAccess=true (Accessed: 2 October 2017). Carr, R. (2005) ‘WHO guidelines for safe wastewater use - more than just numbers’, Irrigation and Drainage, 54, pp. 103–111. doi: 10.1002/ird.190. Carré, E., Beigbeder, J., Jauzein, V., Junqua, G. and Lopez-Ferbery, M. (2017) ‘Life Cycle Assessment Case Study: Tertiary Treatment Process Options for Wastewater Reuse’, Integrated Environmental Assessment and Management, 13(6), pp. 1113–1121. doi: 10.1002/ieam.1956. Carter, R. C., Tyrrel, S. F. and Howsam, P. (1999) ‘The Impact and Sustainability of Community Water Supply

161

and Sanitation Programmes in Developing Countries’, Water and Environment Journal, 13(4), pp. 292–296. doi: 10.1111/j.1747-6593.1999.tb01050.x. Cashman, S., Gaglione, A., Mosley, A. and Al, E. (2014) Environmental and Cost Life Cycle Assessment of Disinfection Options for Municipal Drinking Water Treatment. Washington, DC. Cashman, S., Gaglione, A., Mosley, J., Weiss, L., Hawkins, N., Ashbolt, N., Cashdollar, J., Xue, X., Ma, C. and Arden, S. (2014) Environmental and Cost Life Cycle Assessment of Disinfection Options for Municipal Wastewater Treatment. Washington, DC. Available at: file://homes.surrey.ac.uk/home/Journals/LCA/EPA600R-14377-REV.PDF (Accessed: 17 May 2018). Castonguay, A. C., Urich, C., Iftekhar, M. S. and Deletic, A. (2018) ‘Modelling urban water management transitions: A case of rainwater harvesting’, Environmental Modelling & Software, 105, pp. 270–285. doi: 10.1016/J.ENVSOFT.2018.05.001. Cawley, R. (2016) The management of eco-adaptation by small and medium-size enterprises and the networking role of eco-industrial parks. 1603. London. Available at: http://www.regents.ac.uk/media/2407881/rwpbm1603- cawley-r.pdf (Accessed: 8 September 2016). Changnon, S. A. (1987) Detecting drought conditions in Illinois. Champaign, Illinois. Cheremisinoff, N. P. (2002) ‘What Sand Filtration Is All About’, in Handbook of Water and Wastewater Treatment Technologies. 1st edn. Massachusetts, U.S.A.: Butterworth-Heinemann, pp. 235–267. Available at: https://app.knovel.com/web/view/khtml/show.v/rcid:kpHWWTT005/cid:kt003XD9H3/viewerType:khtml/root_ slug:7-what-sand-filtration-is-all-about/url_slug:what-sand-filtration?kpromoter=federation&b-toc- cid=kpHWWTT005&b-toc-root-slug=&b-toc-url-slug=what-sand-fi (Accessed: 19 February 2018). Chertow, M. R. (2000) ‘Industrial Symbiosis: Literature and Taxonomy’, Annual Review of Environment and Resources, 25(3), pp. 13–17. Chew, I. M. L., Tan, R., Ng, D. K. S., Foo, D. C. Y., Majozi, T. and Gouws, J. (2008) ‘Synthesis of Direct and Indirect Interplant Water Network’, Industrial & Engineering Chemistry Research, 47(23), pp. 9485–9496. doi: 10.1021/ie800072r. Chiew, F. H. S. (2010) ‘Lumped conceptual rainfall-runoff models and simple water balance methods: Overview and applications in ungauged and data limited regions’, Geography Compass, 4(3), pp. 206–225. doi: 10.1111/j.1749-8198.2009.00318.x. Chuang, S.-H., Chang, W.-C., Chang, T.-C. and You, S.-J. (2006) ‘Improving the Removal of Anions by Coagulation and Dissolved Air Flotation in Wastewater Reclamation’, Environmental Technology, 27(5), pp. 493–50. doi: 10.1080/09593332808618663. Chuang, T.-C., Huang, J. and Liu, J. C. (2002) ‘Treatment of Semiconductor Wastewater by Dissolved Air Flotation’, Journal of Environmental Engineering, 128(10), pp. 974–980. doi: 10.1061/͑ASCE͒ 0733- 9372͑2002͒128:10͑974͒. Chungsiriporn, J., Prasertsan, S. and Bunyakan, C. (2006) ‘Minimization of water consumption and process optimization of palm oil mills’, Clean Technologies and Environmental Policy, 8(3), pp. 151–158. doi: 10.1007/s10098-005-0002-y. Ciardelli, G., Corsi, L. and Marcucci, M. (2000) ‘Membrane separation for wastewater reuse in the textile industry’, Resources, Conservation and Recycling, 31, pp. 189–197. Available at: www.elsevier.com/locate/resconrec (Accessed: 22 February 2018). Colla, V., Annunziata Branca, T., Rosito, F., Lucca, C., Padilla Vivas, B. and Men Endez Delmiro, V. (2016) ‘Sustainable Reverse Osmosis application for wastewater treatment in the steel industry’, Journal of Cleaner Production, 130, pp. 103–115. doi: 10.1016/j.jclepro.2015.09.025. Coman, V., Robotin, B. and Ilea, P. (2013) ‘Nickel recovery/removal from industrial wastes: A review’, ‘Resources, Conservation & Recycling’, 73, pp. 229–238. doi: 10.1016/j.resconrec.2013.01.019. Corominas, L., Foley, J., Guest, J. S., Hospido, A., Larsen, H. F., Morera, S. and Shaw, A. (2013) ‘Life cycle assessment applied to wastewater treatment: State of the art’, Water Research. Pergamon, 47(15), pp. 5480– 5492. doi: 10.1016/J.WATRES.2013.06.049.

162

Cote, P., Masini, M. and Mourato, D. (2004) ‘Comparison of membrane options for water reuse and reclamation’, Desalination, 167, pp. 1–11. Available at: www.elsevier.com/locate/desal (Accessed: 3 January 2018). Covaliu, C. I., Matei, E., Georgescu, G., Mălăeru, T. and Ùtefan Biriú, S. (2016) ‘Evaluation of powdered activated carbon performance for wastewater treatment containing organic and inorganic pollutants’, Environmental Engineering and Management Journal, 15(5), pp. 1003–1008. Available at: http://omicron.ch.tuiasi.ro/EEMJ/ (Accessed: 19 February 2018). Cui, J., Wang, X., Yuan, Y., Guo, X., Gu, X. and Jian, L. (2014) ‘Combined ozone oxidation and biological aerated filter processes for treatment of cyanide containing electroplating wastewater’, Chemical Engineering Journal, 241, pp. 184–189. doi: 10.1016/J.CEJ.2013.09.003. Cyr, P. J., Suri, R. P. S. and Helmig, E. D. (2002) ‘A pilot scale evaluation of removal of mercury from pharmaceutical wastewater using granular activated carbon’, Water Research, 36, pp. 4725–4734. Available at: https://ac.els-cdn.com/S0043135402002142/1-s2.0-S0043135402002142-main.pdf?_tid=2dbc903e-13ea-11e8- 920e-00000aacb362&acdnat=1518875903_abc540a4430c1866a9f0e86291253b8f (Accessed: 17 February 2018). Daud, M. A. (2015) ‘Challenges in managing water resources, Malaysian experience’, in 2015 JWWA Annual Conference and Exhibition, Water Associations Meeting and International Forum on Water Supply. Available at: http://www.jwwa.or.jp/jigyou/kaigai_file/seminar_05/F-03_MWA(F).pdf (Accessed: 11 January 2016). Dawadi, S. and Ahmad, S. (2013) ‘Evaluating the impact of demand-side management on water resources under changing climatic conditions and increasing population’, Journal of Environmental Management, 114, pp. 261– 275. doi: 10.1016/j.jenvman.2012.10.015. Deemer, B. R., Harrison, J. A., Li, S., Beaulieu, J. J., Delsontro, T., Barros, N., Bezerra-Neto, J. F., Powers, S. M., Santos, M. A. Dos and Vonk, J. A. (2016) ‘Greenhouse Gas Emissions from Reservoir Water Surfaces: A New Global Synthesis’, BioScience, 66, pp. 949–964. doi: 10.1093/biosci/biw117. Deif, A. M. (2011) ‘A system model for green manufacturing’, Journal of Cleaner Production, 19(14), pp. 1553–1559. doi: 10.1016/J.JCLEPRO.2011.05.022. Deng, C., Liu, P., Guo, S., Wang, H. and Wang, D. (2015) ‘Estimation of nonfluctuating reservoir inflow from water level observations using methods based on flow continuity’, Journal of Hydrology, 529, pp. 1198–1210. doi: 10.1016/j.jhydrol.2015.09.037. Deng, C., Liu, P., Liu, Y., Wu, Z. and Wang, D. (2015) ‘Integrated Hydrologic and Reservoir Routing Model for Real-Time Water Level Forecasts’, Jurnal of Hydrologic Engineering, 20(9). doi: 10.1061/(ASCE)HE.1943- 5584.0001138. Department of Economic and Social Affairs, Population Division, U. N. (2014) World Urbanization Prospect: The 2014 Revision, Highlights, United Nation. New York. doi: 10.4054/DemRes.2005.12.9. Department of Environment Malaysia (2009) ‘Environmental Quality Act 1974, Environmental Quality (Industrial Effluent) Regulation 2009’. Putrajaya: Department of Environment Malaysia, p. 50. Available at: https://www.doe.gov.my/portalv1/wp- content/uploads/2015/01/Environmental_Quality_Industrial_Effluent_Regulations_2009_-_P.U.A_434- 2009.pdf (Accessed: 23 December 2017). Department of Statistics Malaysia (2015) Annual National Accounts: Gross Domestic Product (2010-2014). Kuala Lumpur, Malaysia. doi: 10.1017/CBO9781107415324.004. DHI (2003) ‘MIKE 11: A modelling system for Rivers and Channels - User Guide’, DHI Software. DHI Water & Environment, p. 430. Available at: https://www.tu-braunschweig.de/Medien- DB/geooekologie/mike11usersmanual.pdf (Accessed: 11 July 2017). DHI Water & Environment (2008) McKenzie River MIKE BASIN Model. Available at: file://homes.surrey.ac.uk/home/MBMACANGEMODEL.pdf (Accessed: 11 July 2017). Dickinson, D. A., Draper, C. W., Saminathan, M., Sohn, J. E. and Williams, G. (1995) ‘Green product manufacturing’, AT&T Technical Journal, 74(6), pp. 26–35. doi: 10.1002/j.1538-7305.1995.tb00263.x. Dominguez-Tagle, C., J. Romero-Ternero, V. and Delgado-Torres, A. M. (2011) Boron removal efficiency in

163

small seawater Reverse Osmosis systems, Desalination. doi: 10.1016/j.desal.2010.07.028. Donkor, E. A., Asce, S. M., Mazzuchi, T. A., Soyer, R. and Roberson, J. A. (2014) ‘Urban Water Demand Forecasting: Review of Methods and Models’, Journal of water resources planning and management, 140(2), pp. 146–159. doi: 10.1061/(ASCE)WR.1943-5452. Dreyer, L. C., Niemann, A. L. and Hauschild, M. Z. (2003) ‘Comparison of Three Different LCIA Methods: EDIP97, CML2001 and Eco-indicator 99 Does it matter which one you choose?’, The International Journal of Life Cycle Assessment, 8(4), pp. 191–200. doi: 10.1065/Ica2003.06.115. Drouiche, N., Djouadi-Belkada, F., Ouslimane, T., Kefaifi, A., Fathi, J. and Ahmetovic, E. (2013) ‘Photovoltaic solar cells industry wastewater treatment’, Desalination and Water Treatment, 51(31–33), pp. 5965–5973. doi: 10.1080/19443994.2012.763217. Dunn, R. F. and El-Halwagi, M. M. (2003) ‘Process integration technology review: background and applications in the chemical process industry’, Journal of Chemical Technology & Biotechnology, 78(9), pp. 1011–1021. doi: 10.1002/jctb.738. Dunn, R. F. and Wenzel, H. (2001) ‘Process integration design methods for water conservation and wastewater reduction in industry’, Clean Products and Processes, 3(3), pp. 307–318. doi: 10.1007/s100980100121. Duport, D. P. and Renzetti, S. (2001) ‘The Role of Water in Manufacturing’, Environmental and Resource Economics, 18(4), pp. 411–432. Dziedzic, R., Margerm, K., Evenson, J., Karney, B. W. and Asce, M. (2015) ‘Building an Integrated Water– Land Use Database for Defining Benchmarks, Conservation Targets, and User Clusters’, Journal of water resources planning and management, 141(4). doi: 10.1061/(ASCE)WR.1943-5452.0000462. Dziegielewski, B., Kiefer, J. C., Opitz, E. M., Porter, G. A., Lantz, G. L., DeOreo, W. B., Mayer, P. W. and Nelson, J. O. (2000) Commercial and Institutional End Uses of Water. U.S.A. Available at: http://www.waterrf.org/PublicReportLibrary/RFR90806_2000_241B.pdf (Accessed: 2 October 2017). Ecoinvent 3.3 (2014) ‘Tap water {RoW}| tap water production, conventional treatment | Alloc Def, U’. Economic and Social Commission for Western Asia (2003) Wastewater treatment technologies: a general review. New York. Available at: http://www.igemportal.org/Resim/Wastewater Treatment Technologies_ A general rewiev.pdf (Accessed: 4 April 2018). Economic Planning Unit (EPU) Malaysia (2010) Rancangan Malaysia Kesepuluh (Tenth Malaysia Plan). Putrajaya, Malaysia. doi: 10.1007/s13398-014-0173-7.2. Economic Planning Unit (EPU) Malaysia (2015) Rancangan Malaysia Kesebelas (Eleventh Malaysia Plan). Putrajaya, Malaysia. Elabras Veiga, L. B. and Magrini, A. (2009) ‘Eco-industrial park development in Rio de Janeiro, Brazil: a tool for sustainable development’, Journal of Cleaner Production, 17(7), pp. 653–661. doi: 10.1016/J.JCLEPRO.2008.11.009. Emamjomeh, M. M. and Sivakumar, M. (2009) ‘Review of pollutants removed by electrocoagulation and electrocoagulation/flotation processes’, Journal of Environmental Management. Academic Press, 90(5), pp. 1663–1679. doi: 10.1016/J.JENVMAN.2008.12.011. Fabre, J., Ruelland, D., Dezetter, A. and Grouillet, B. (2015) ‘Simulating past changes in the balance between water demand and availability and assessing their main drivers at the river basin scale’, Hydrology and Earth System Sciences, 19(3), pp. 1263–1285. doi: 10.5194/hess-19-1263-2015. Fang, L. L. (2010) Life Cycle Assessment of TRENS as a water scarcity mit- igation option for Copenhagen Exploration, modelling and management of groundwater-dependent ecosystems in karst. Technical University of Denmark. Available at: https://ars.els-cdn.com/content/image/1-s2.0-S0043135415302827-mmc1.pdf (Accessed: 26 May 2018). Feng, X. and Chu, K. H. (2004) ‘Cost Optimization of Industrial Wastewater Reuse Systems’, Process Safety and Environmental Protection, 82(3), pp. 249–255. doi: 10.1205/095758204323066019. Friedler, E. and Pisanty, E. (2006) ‘Effects of design flow and treatment level on construction and operation costs of municipal wastewater treatment plants and their implications on policy making’, Water Research.

164

Pergamon, 40(20), pp. 3751–3758. doi: 10.1016/J.WATRES.2006.08.015. Fu, F. and Wang, Q. (2011) ‘Removal of heavy metal ions from wastewaters: A review’, Journal of Environmental Management, 92, pp. 407–418. doi: 10.1016/j.jenvman.2010.11.011. Garcia-Segura, S., Eiband, M. M. S. G., de Melo, J. V. and Martínez-Huitle, C. A. (2017) ‘Electrocoagulation and advanced electrocoagulation processes: A general review about the fundamentals, emerging applications and its association with other technologies’, Journal of Electroanalytical Chemistry. Elsevier, 801, pp. 267–299. doi: 10.1016/J.JELECHEM.2017.07.047. Garcia, X. and Pargament, D. (2015) ‘Reusing wastewater to cope with water scarcity: Economic, social and environmental considerations for decision-making’, Resources, Conservation and Recycling, 101, pp. 154–166. doi: 10.1016/J.RESCONREC.2015.05.015. Gerrity, D., Owens-Bennett, E., Venezia, T., Stanford, B. D., Plumlee, M. H., Debroux, J. and Trussell, R. S. (2014) ‘Applicability of ozone and biologically activated carbon for potable reuse’, Ozone: Science and Engineering, 36(2), pp. 123–137. doi: 10.1080/01919512.2013.866886. Gerrity, D., Pecson, B., Trussell, R. S. and Rhodes Trussell, R. (2013) ‘Potable Reuse Treatment Trains throughout the World’, Journal of Water Supply: Research and Technology—AQUA, 62(6), pp. 321–338. doi: 10.2166/aqua.2013.041. Gibbs, D. and Deutz, P. (2007) ‘Reflections on implementing industrial ecology through eco-industrial park development’, Journal of Cleaner Production, 15(17), pp. 1683–1695. doi: 10.1016/j.jclepro.2007.02.003. Gillot, S., De Clercq, B., Defour, D., Simoens, F., Gernaey, K. and Vanrolleghem, P. A. (1999) ‘Optimization of wastewater treatment plant design and operation using simulation and cost analysis’, in Proceedings 72nd Annual WEF Conference and Exposition. New Orleans, USA. Available at: https://s3.amazonaws.com/academia.edu.documents/41506314/Optimization_Of_Wastewater_Treatment_Pla20 160124-28484- k54k7i.pdf?AWSAccessKeyId=AKIAIWOWYYGZ2Y53UL3A&Expires=1523348717&Signature=kY6WfOth vgewHbBfQeOZgBT5Tbo%253D&response-content-disposition=inline%25 (Accessed: 10 April 2018). De Gispert, C. (2004) ‘The economic analysis of industrial water demand: a review’, Environment and Planning C: Government and Policy, 22, pp. 15–30. doi: 10.1068/c20s. Gleick, P. H., Haasz, D., Henges-Jeck, C., Srinivasan, V., Wolff, G., Cushing, K. K. and Mann, A. (2003) Waste Not, Want Not: The Potential for Urban Water Conservation in California. Oakland, California. Gogate, P. R. and Pandit, A. B. (2004) ‘A review of imperative technologies for wastewater treatment I: oxidation technologies at ambient conditions’, Advances in Environmental Research, 8, pp. 501–551. doi: 10.1016/S1093-0191(03)00032-7. Goh, Y. C., Zainol, Z. and Mat Amin, M. Z. (2015) ‘Assessment of future water availability under the changing climate: case study of Klang River Basin, Malaysia’, International Journal of River Basin Management, pp. 1–9. doi: 10.1080/15715124.2015.1068178. Goher, M. E., Hassan, A. M., Abdel-Moniem, I. A., Fahmy, A. H., Abdo, M. H. and El-sayed, S. M. (2015) ‘Removal of aluminum, iron and manganese ions from industrial wastes using granular activated carbon and Amberlite IR-120H’, The Egyptian Journal of Aquatic Research, 41(2), pp. 155–164. doi: 10.1016/J.EJAR.2015.04.002. Gourbesville, P. (2008) ‘Challenges for integrated water resources management’, Physics and Chemistry of the Earth, 33, pp. 284–289. doi: 10.1016/j.pce.2008.02.002. Govindaraju, R. S. (2000) ‘Artificial neural networks in hydrology. II: Hydrological application’, Journal of Hydrologic Engineering, 5(2), pp. 124–137. Available at: http://ascelibrary.org/doi/pdf/10.1061/(ASCE)1084- 0699(2000)5:2(124) (Accessed: 8 July 2017). Gunatilake, H. and De Guzman, F. D. (2008) Market-Based Approaches for Managing the Asian Environment: A Review. 124. Philippines. Available at: https://www.adb.org/sites/default/files/publication/28374/economics- wp124.pdf (Accessed: 26 June 2017). Güntner, A., Krol, M. S., Carlos De Araújo, J. and Bronstert, A. (2004) ‘Simple water balance modelling of surface reservoir system in a large data-scarce semiarid region’, Hydrological Sciences Journal, 49(5), pp. 901–

165

918. doi: 10.1623/hysj.49.5.901.55139. Guo, T., Englehardt, J. and Wu, T. (2014) ‘Review of cost versus scale: water and wastewater treatment and reuse process’, Water, Science, and Technology, 69(2), pp. 223–234. doi: 10.2166/wst.2013.734. Gupta, V. K., Ali, I., Saleh, T. A., Nayak, A. and Agarwal, S. (2012) ‘Chemical treatment technologies for waste-water recycling—an overview’, RSC Advances, 2(16), p. 6380. doi: 10.1039/c2ra20340e. Gurnham, C. F. (1965) Industrial wastewater control: A texbook and reference work. First. Edited by R. F. Baddour. Illinois: Academic Press. Gurung, T. R., Stewart, R. A., Beal, C. D. and Sharma, A. K. (2016) ‘Investigating the Financial Implications and Viability of Diversified Water Supply Systems in an Urban Water Supply Zone’, Water Resources Management, 30, pp. 4037–4051. doi: 10.1007/s11269-016-1411-x. Hagen, E. R., John Holmes, K., Kiang, J. E. and Steiner, R. C. (2005) ‘Benefits of iterative water supply forecasting in the Washington D.C., Metropolitan Area’, Journal of the American Water Resources Association (JAWRA), 41(6), pp. 1417–1430. Available at: https://www.potomacriver.org/wp- content/uploads/2014/12/HagenJAWRA05.pdf (Accessed: 3 October 2017). Hansen, H. K., Peña, S. F., Gutiérrez, C., Lazo, A., Lazo, P. and Ottosen, L. M. (2019) ‘Selenium removal from petroleum refinery wastewater using an electrocoagulation technique’, Journal of Hazardous Materials. Elsevier, 364, pp. 78–81. doi: 10.1016/J.JHAZMAT.2018.09.090. Hashimoto, T., Stedinger, J. R. and Loucks, D. P. (1982) ‘Reliability, resiliency, and vulnerability criteria for water resource system performance evaluation’, Water Resources Research, 18(1), pp. 14–20. doi: 10.1029/WR018i001p00014. Heberger, M., Donnelly, K. and Cooley, H. (2016) A Community Guide for Evaluating Future Urban Water Demand A Community Guide for Evaluating Future Urban Water Demand. Oakland, California. Available at: www.pacinst.org (Accessed: 3 October 2017). Heim, R. R. (2002) ‘A review of twentieth-century drought indices used in the United States’, Bulletin of the American Meteorological Society, pp. 1149–1165. doi: 10.1175/1520-0477(2002)083<1149:AROTDI>2.3.CO;2. Hernández-Sancho, F., Lamizana-Diallo, B., Mateo-Sagasta, J. and Qadir, M. (2015) Economic valuation of wastewater - The cost of action and the cost of no action. Available at: https://wedocs.unep.org/bitstream/handle/20.500.11822/7465/- Economic_Valuation_of_Wastewater_The_Cost_of_Action_and_the_Cost_of_No_Action- 2015Wastewater_Evaluation_Report_Mail.pdf.pdf?sequence=3&isAllowed=y (Accessed: 10 April 2018). Hernandez-Sancho, F., Molinos-Senante, M. and Sala-Garrido, R. (2011) ‘Cost modelling for wastewater treatment processes’, Desalination, 268(1–3), pp. 1–5. doi: 10.1016/J.DESAL.2010.09.042. Horne, R. E. (2009) ‘Limits to labels: The role of eco-labels in the assessment of product sustainability and routes to sustainable consumption’, International Journal of Consumer Studies, 33(2), pp. 175–182. doi: 10.1111/j.1470-6431.2009.00752.x. Hospido, A., Sanchez, I., Rodriguez-Garcia, G., Iglesias, A., Buntner, D., Reif, R., Moreira, M. T. and Feijoo, G. (2012) ‘Are all membrane reactors equal from an environmental point of view?’, Desalination, 285, pp. 263– 270. doi: 10.1016/J.DESAL.2011.10.011. House-Peters, L. and Chang, H. (2011) ‘Urban Water Demand Modeling: Review of Concepts, Methods, and Organizing Principles’, Water Resources Research, 47, pp. 1–15. doi: 10.1029/2010WR009624. Hrachowitz, M., Savenije, H. H. G., Blöschl, G., Mcdonnell, J. J., Sivapalan, M., Pomeroy, J. W., Arheimer, B., Blume, T., Clark, M. P., Hrachowitz, C., Blöschl, H. H. G., Mcdonnell, G., Sivapalan, J. J., Pomeroy, M. and Clark, T. (2013) ‘A decade of Predictions in Ungauged Basins (PUB)—a review’, Hydrological Sciences Journal – Journal des Sciences Hydrologiques Hydrological Sciences Journal, 58(586), pp. 1198–1255. doi: 10.1080/02626667.2013.803183. Igos, E., Dalle, A., Tiruta-Barna, L., Benetto, E., Baudin, I. and Mery, Y. (2014) ‘Life Cycle Assessment of water treatment: what is the contribution of infrastructure and operation at unit process level?’, Journal of Cleaner Production, 65, pp. 424–431. doi: 10.1016/J.JCLEPRO.2013.07.061.

166

Inman, D. and Jeffrey, P. (2006) ‘A review of residential water conservation tool performance and influences on implementation effectiveness’, Urban Water Journal, 3(3), pp. 127–143. doi: 10.1080/15730620600961288. International Organisation for Standardisation (2015) ISO 16075-1:2015 - Guidelines for treated wastewater use for irrigation projects -- Part 1: The basis of a reuse project for irrigation. Available at: https://www.iso.org/standard/62756.html (Accessed: 23 December 2017). Ioannou-Ttofa, L., Foteinis, S., Chatzisymeon, E. and Fatta-Kassinos, D. (2016) ‘The environmental footprint of a membrane bioreactor treatment process through Life Cycle Analysis’, Science of The Total Environment, 568, pp. 306–318. doi: 10.1016/J.SCITOTENV.2016.06.032. Ipek, U. (2004) ‘Phenol removal capacity of reverse osmosis with and without pre-treatment’, Filtration & Separation, 41(7), pp. 39–40. Available at: https://ac.els-cdn.com/S0015188204003210/1-s2.0- S0015188204003210-main.pdf?_tid=91ace862-127a-11e8-834f- 00000aab0f26&acdnat=1518718007_92afbdebb578dd8d53f30febf186d57d (Accessed: 15 February 2018). Ipek, U. (2005) ‘Removal of Ni(II) and Zn(II) from an aqueous solution by reverse osmosis’, Desalination, 174, pp. 161–169. Available at: www.elsevier.corrdlocate/de (Accessed: 15 February 2018). Jabatan Pengairan dan Saliran Malaysia (1976) Evaporation in Peninsular Malaysia. Kuala Lumpur, Malaysia. Jabatan Perancangan Bandar dan Desa, J. (2016) ‘Manual garis panduan dan piawaian perancangan negeri Selagor, Bab 3: Perindustrian’. Selangor, Malaysia: Jabatan Perancangan Bandar dan Desa, JPBD, p. 13. Jain, S. K. and Bhunya, P. K. (2008) ‘Reliability, resilience and vulnerability of a multipurpose storage reservoir’, Hydrological Sciences–Journal–des Sciences Hydrologiques, 53(2). Jayarathna, L., Rajapaksa, D., Managi, S., Athukorala, W., Torgler, B., Garcia-Valiñas, M. A., Gifford, R. and Wilson, C. (2017) ‘A GIS based spatial decision support system for analysing residential water demand: A case study in Australia’, Sustainable Cities and Society, 32, pp. 67–77. doi: 10.1016/J.SCS.2017.03.012. Jin, X., Li, E., Lu, S., Qiu, Z. and Sui, Q. (2013) ‘Coking wastewater treatment for industrial reuse purpose: Combining biological processes with ultrafiltration, nanofiltration and reverse osmosis’, Journal of Environmental Sciences, 25(8), pp. 1565–1574. doi: 10.1016/S1001-0742(12)60212-5. Judd, S. J. (2016) ‘The status of industrial and municipal effluent treatment with membrane bioreactor technology’, Chemical Engineering Journal, 305, pp. 37–45. doi: 10.1016/j.cej.2015.08.141. Judd, S. J. (2017) ‘Membrane technology costs and me’, Water Research, 122, pp. 1–9. doi: 10.1016/J.WATRES.2017.05.027. Jusoh, A., Su Shiung, L., Ali, aini and Noor, M. (2007) ‘A simulation study of the removal efficiency of granular activated carbon on cadmium and lead’, Desalination, 206, pp. 9–16. doi: 10.1016/j.desal.2006.04.048. Kampragou, E., Lekkas, D. F. and Assimacopoulos, D. (2011) ‘Water demand management: implementation principles and indicative case studies’, Water and Environment Journal, 25(4), pp. 466–476. doi: 10.1111/j.1747-6593.2010.00240.x. Kapoor, A., Tanjore, S. and Viraraghavan, T. (1995) ‘Removal of Selenium from Water and Wastewater’, International Journal of Environmental Studies, 49, pp. 137–147. Available at: https://www.tandfonline.com/doi/pdf/10.1080/00207239508711016 (Accessed: 28 March 2018). Kargari, A. and Mohammadi, S. (2014) ‘Evaluation of phenol removal from aqueous solutions by UV, RO, and UV/RO hybrid systems’, Desalination and water treatment, 54(6), pp. 1612–1620. doi: 10.1080/19443994.2014.891077. Karnib, M., Kabbani, A., Holail, H. and Olama, Z. (2014) ‘Heavy Metals Removal Using Activated Carbon, Silica and Silica Activated Carbon Composite’, Energy Procedia, 50, pp. 113–120. doi: 10.1016/J.EGYPRO.2014.06.014. Kasim, N., Mohammad, A. W. and Sheikh Abdullah, S. R. (2017) ‘Iron and manganese removal by nanofiltration and ultrafiltration membranes: influence of pH adjustment’, Malaysian Journal of Analytical Sciences, 21(1), pp. 149–158. doi: 10.17576/mjas-2017-2101-17. Katsou, E., Malamis, S. and Loizidou, M. (2011) ‘Performance of a membrane bioreactor used for the treatment of wastewater contaminated with heavy metals’, Bioresource Technology, 102, pp. 4325–4332. doi:

167

10.1016/j.biortech.2010.10.118. Kellis, M., Kalavrouziotis, I., Seferi, G. and Gikas, P. (2013) ‘Review of Wastewater Reuse in the Mediterranean Countries, Focusing On Regulations and Policies for Municipal and Industrial Applications’, Global NEST Journal, 15(3), pp. 333–350. Available at: https://journal.gnest.org/sites/default/files/Journal Papers/333-350_15-3_936_Kalavrouziotis.pdf (Accessed: 1 January 2018). Kettunen, R. and Keskitalo, P. (2000) ‘Combination of membrane technology and limestone filtration to control drinking water quality’, Desalination, 131, pp. 271–283. doi: 10.1016/S0011-9164(00)90025-0. Khazanah Nasional Berhad (2006) Section B: Planning and Implementation, Chapter 14 - Urban Infrastructure, Comprehensive Development Plan for SJER. 2006-2025. Johor Bahru, Malaysia. doi: 10.4135/9781412963930.n595. Kieger, J. C., Krentz, L. R. and Dziegielewski, B. (2015) Methodology for Evaluating Water Use in the Commercial, Institutional, and Industrial Sectors. Washington D.C. Available at: http://www.waterrf.org/PublicReportLibrary/4375.pdf (Accessed: 28 September 2017). Kisi, O. (2005) ‘Daily River Flow Forecasting Using Artificial Neural Networks and Auto-Regressive Model’, Turkish journal of engineering and environmental science, 29, pp. 9–20. Available at: http://journals.tubitak.gov.tr/engineering/issues/muh-05-29-1/muh-29-1-2-0307-1.pdf (Accessed: 11 July 2017). Kjeldsen, T. R. and Rosbjerg, D. A. N. (2005) ‘Choice of reliability, resilience and vulnerability estimator for risk assessments of water resources systems’, Hydrological Sciences–Journal, 49, pp. 755–768. Klein, R. J. T., Nicholls, R. J. and Thomalla, F. (2003) ‘Resilience to natural hazards: How useful is this concept?’, Environmental Hazards, 5, pp. 35–45. doi: 10.1016/j.hazards.2004.02.001. Koenig, L. (1973) ‘Optimal Fail-Safe Process Design’, Water Pollution Control Federation. Water Environment Federation, 45(4), pp. 647–654. Available at: http://www.jstor.org/stable/25037804. Koivunen, J. and Heinonen‐Tanski, H. (2008) ‘Dissolved air flotation (DAF) for primary and tertiary treatment of municipal wastewater’, Environmental Technology, 29(1), pp. 101–109. doi: 10.1080/09593330802009410. Koppol, A. P. R. R., Bagajewicz, M. J., Dericks, B. J. and Savelski, M. J. (2003) ‘On zero water discharge solutions in the process industry’, Advances in Environmental Research, 8(2), pp. 151–171. doi: 10.1016/S1093-0191(02)00130-2. Kowalski, M., Lynn, S., Waylen, C. and Bujnowicz, A. (2011) Freshwater use in the UK: manufacturing sector. United Kingdom. Available at: http://www.wrap.org.uk/sites/files/wrap/PAD101-201 - Manufacturing sector water report - FINAL APPROVED for publication - 12%2C03%2C12.pdf (Accessed: 3 October 2017). Krzeminski, P., van der Graaf, J. H. J. M. and van Lier, J. B. (2012) ‘Specific energy consumption of membrane bioreactor (MBR) for sewage treatment’, Water Science & Technology, 65(2), p. 380. doi: 10.2166/wst.2012.861. Ku, S.-J. and Yoo, S.-H. (2011) ‘Economic Value of Water in the Korean Manufacturing Industry’, Water Resources Management, 26(1), pp. 81–88. doi: 10.1007/s11269-011-9905-z. Kulkarni, S. J. (2017) ‘Recent Investigations for Phenol Removal from Effluent: An Insight into Research and Studies from 2013 to 2017’, International Journal of Science and Healthcare, 22(4), pp. 2455–7587. Available at: http://ijshr.com/IJSHR_Vol.2_Issue.4_Oct2017/IJSHR002.pdf (Accessed: 19 February 2018). Kundzewicz, Z. W. and Kindler, J. (1995) ‘Multiple criteria for evaluation of reliability aspects of water resource systems’, in Modelling and Management of Sustainable Basin-scale Water Resource Sytems, pp. 217– 224. Kurniawan, T. A., Chan, G. Y. S., Lo, W.-H. and Babel, S. (2006) ‘Physico–chemical treatment techniques for wastewater laden with heavy metals’, Chemical Engineering Journal, 118, pp. 83–98. doi: 10.1016/j.cej.2006.01.015. Kusin, F. M., Zahar, M. S. M., Muhammad, S. N., Zin, Z. M. and Sharif, S. M. (2016) ‘Assessing the potential use of abandoned mining pools as an alternative resource of raw water supply’, Water Science and Technology: Water Supply, 16(2), pp. 410–417. doi: 10.2166/ws.2015.150. Lahnsteiner, J., Van Rensburg, P. and Esterhuizen, J. (2017) ‘Direct potable reuse – a feasible water

168

management option’, Journal of Water Reuse and Desalination, 7(4), p. 15. doi: 10.2166/wrd.2017.172. Lambert, A. J. D. and Boons, F. A. (2002) ‘Eco-industrial parks: stimulating sustainable development in mixed industrial parks’, Technovation, 22, pp. 471–484. Available at: www.elsevier.com/locate/technovation (Accessed: 8 September 2016). Lan Anh, N. T., Willems, P., Boxall, J. B. and Saul, A. J. (2008) ‘An evaluation of three lumped conceptual rainfall-runoff models at catchment scale’, in Lan Anh, N. T., Willems, P., Boxall, J. B., and Saul, A. J. (eds) The 13th World Water Congress. France, p. 12. Available at: http://www.iwra.org/congress/resource/abs303_article.pdf (Accessed: 11 July 2017). Langenbach, K., Kuschk, P., Horn, H. and Kastner, M. (2009) ‘Slow Sand Filtration of Secondary Clarifier Effluent for Wastewater Reuse’, Environmental science and technology, 43, pp. 5896–5901. doi: 10.1021/es900527j. Lee, C. S., Robinson, J. and Chong, M. F. (2014) ‘A review on application of flocculants in wastewater treatment’, Process Safety and Environmental Protection, 92(2), pp. 489–508. doi: 10.1016/j.psep.2014.04.010. Lee, H.-B., Peart, T. E., Gris, G. and Chan, J. (2002) ‘Endocrine-disrupting chemicals in industrial wastewater samples in Toronto, Ontario’, Water quality research journal of Canada, 37(2), pp. 295–325. Available at: https://www.researchgate.net/profile/Thomas_King5/publication/242100130_Trends_in_the_Distribution_of_P CBs_Compared_to_PACs_in_Sediments_and_Mussels_of_Halifax_Harbour/links/00b7d52d166dac8cd400000 0.pdf#page=164 (Accessed: 11 January 2018). Lee, H. and Pin Tan, T. (2016) ‘Singapore’s experience with reclaimed water: NEWater’, InternatIonal Journal of Water resources Development, 32(4), pp. 611–621. doi: 10.1080/07900627.2015.1120188. Lemos, D., Dias, A. C., Gabarrell, X. and Arroja, L. (2013) ‘Environmental assessment of an urban water system’, Journal of Cleaner Production, 54, pp. 157–165. doi: 10.1016/J.JCLEPRO.2013.04.029. Leoi Leoi, S. (2015) Selangor water supply under threat - Nation | The Star Online, Star Online. Available at: https://www.thestar.com.my/news/nation/2015/05/11/selangor-water-supply-under-threat-syabas-pollution- caused-treatment-plants-to-be-shut-down-42-times/ (Accessed: 5 September 2018). Leverenz, H. L., Tchobanoglous, G. and Asano, T. (2011) ‘Direct potable reuse: a future imperative’, Journal of Water Reuse and Desalination, 1(1), pp. 2–10. doi: 10.2166/wrd.2011.000. Libralato, G., Volpi Ghirardini, A. and Avezzù, F. (2012) ‘To centralise or to decentralise: An overview of the most recent trends in wastewater treatment management’, Journal of Environmental Management, 94(1), pp. 61–68. doi: 10.1016/J.JENVMAN.2011.07.010. Lim, S.-R. and Park, J. M. (2010) ‘Interfactory and Intrafactory Water Network System To Remodel a Conventional Industrial Park to a Green Eco-industrial Park’, Industrial & Engineering Chemistry Research, 49, pp. 1351–1358. doi: 10.1021/ie9014233. Lin, H., Gao, W., Meng, F., Liao, B.-Q., Leung, K.-T., Zhao, L., Chen, J. and Hong, H. (2012) ‘Membrane Bioreactors for Industrial Wastewater Treatment: A Critical Review’, Critical Reviews in Environmental Science and Technology, 427(42). doi: 10.1080/10643389.2010.526494. Liu, C.-C. and Liu, J. C. (2016) ‘Coupled precipitation-ultrafiltration for treatment of high fluoride-content wastewater’, Journal of the Taiwan Institute of Chemical Engineers, 58, pp. 259–263. doi: 10.1016/J.JTICE.2015.05.038. Liu, Z., Kanjo, Y. and Mizutani, S. (2008) ‘Removal mechanisms for endocrine disrupting compounds (EDCs) in wastewater treatment — physical means, biodegradation, and chemical advanced oxidation: A review’, Science of the Total Environment, 407, pp. 731–748. doi: 10.1016/j.scitotenv.2008.08.039. Londhe, S. and Charhate, S. (2010) ‘Comparison of data-driven modelling techniques for river flow forecasting’, Hydrological Science Journal, 55(7), pp. 1163–1174. doi: 10.1080/02626667.2010.512867. Van Loon, A. F. (2015) ‘Hydrological drought explained’, Wiley Interdisciplinary Reviews: Water, 2(4), pp. 359–392. doi: 10.1002/wat2.1085. Lyu, S., Chen, W., Zhang, W., Fan, Y. and Jiao, W. (2016) ‘Wastewater reclamation and reuse in China: Opportunities and challenges’, Journal of Environmental Sciences, 39, pp. 86–96. doi:

169

10.1016/j.jes.2015.11.012. Madzin, Z., Fitri Shai-in, M. and Mohd Kusin, F. (2015) ‘Comparing Heavy Metal Mobility in Active and Abandoned Mining Sites at Bestari Jaya, Selangor’, Procedia Environmental Sciences, 30, pp. 232–237. doi: 10.1016/j.proenv.2015.10.042. Malamis, S., Katsou, E. and Haralambous, K. J. (2011) ‘Study of Ni(II), Cu(II), Pb(II), and Zn(II) removal using sludge and minerals followed by MF/UF’, Water, Air & Soil Pollution, 218, pp. 81–92. doi: 10.1007/s11270- 010-0625-4. Malamis, S., Katsou, E., Takopoulos, K., Demetriou, P. and Loizidou, M. (2012) ‘Assessment of metal removal, biomass activity and RO concentrate treatment in an MBR–RO system’, Journal of Hazardous Materials, 209– 210, pp. 1–8. doi: 10.1016/J.JHAZMAT.2011.10.085. Malaysian Administrative Modernisation and Management Planning Unit (2018) National Transformation Programme - Annual Report 2017. Kuala Lumpur, Malaysia. Available at: http://www.mampu.gov.my/images/pengumuman/NTP_AR2017_ENG.pdf (Accessed: 27 August 2018). Mamade, A., Loureiro, D., Covas, D., Coelho, S. T. and Amado, C. (2014) ‘Spatial and temporal forecasting of water consumption at the DMA level using extensive measurements’, Procedia Engineering, 70, pp. 1063–1073. doi: 10.1016/j.proeng.2014.02.118. Maria, A., Alves, B. and Norberta De Pinho, M. (2000) ‘Ultrafiltration for colour removal of tannery dyeing wastewaters’, Desalination, 130, pp. 147–154. Available at: www.elsevier.com/locate/desal (Accessed: 22 February 2018). Martínez-Villafañe, J. F., Montero-Ocampo, C. and García-Lara, A. M. (2009) ‘Energy and electrode consumption analysis of electrocoagulation for the removal of arsenic from underground water’, Journal of Hazardous Materials. Elsevier, 172(2–3), pp. 1617–1622. doi: 10.1016/J.JHAZMAT.2009.08.044. Mastrandrea, M. D., Heller, N. E., Root, T. L. and Schneider, S. H. (2010) ‘Bridging the gap: Linking climate- impacts research with adaptation planning and management’, Climatic Change, 100(1), pp. 87–101. doi: 10.1007/s10584-010-9827-4. Mayer, A. L. (2008) ‘Strengths and weaknesses of common sustainability indices for multidimensional systems’, Environmental International, 34, pp. 277–291. doi: 10.1016/j.envint.2007.09.004. Mehran, A., Mazdiyasni, O. and Aghakouchak, A. (2015) ‘A hybrid framework for assessing socioeconomic drought: Linking climate variability, local resilience, and demand’, Journal of Geophysical Research : Atmospheres, pp. 7520–7533. doi: 10.1002/2015JD023147.Received. Melin, T., Jefferson, B., Bixio, D., Thoeye, C., De Wilde, W., De Koning, J., Van Der Graaf, J. and Wintgens, T. (2006) ‘Membrane bioreactor technology for wastewater treatment and reuse’, Desalination, 187, pp. 271–282. doi: 10.1016/j.desal.2005.04.086. Mens, M. J. P., Gilroy, K. and Williams, D. (2015) ‘Developing system robustness analysis for drought risk management: an application on a water supply reservoir’, Nat. Hazards Earth Syst. Sci, 15, pp. 1933–1940. doi: 10.5194/nhess-15-1933-2015. Mert, B. K. and Kestioglu, K. (2014) ‘Recovery of Cr(III) from tanning process using membrane separation processes’, Clean Technology and Environmental Policy, 16, pp. 1615–1624. doi: 10.1007/s10098-014-0737-4. Mert, B. K., Sivrio, Ö., Glu, ˘, Yonar, T. and Özçiftçi, S. (2014) ‘Treatment of Jewelry Manufacturing Effluent Containing Cyanide Using Ozone-Based Photochemical Advanced Oxidation Processes’, Ozone: Science & Engineering, 36(2), pp. 196–205. doi: 10.1080/01919512.2013.860354. MIDA (2014) Impact to the Industries on water rationing extersise in Selangor. Kuala Lumpur, Malaysia. MIDA (2016) Guideline and Procedures for Waste Eco Park (WEP) Incentive and/or Expatriate Posts, Malaysian Investment Development Authority. Kuala Lumpur, Malaysia. Available at: http://www.mida.gov.my/home/tax-incentives-for-green-industry/posts/. Miller, D. J., Kasemset, S., Paul, D. R. and Freeman, B. D. (2014) ‘Comparison of membrane fouling at constant flux and constant transmembrane pressure conditions’, Journal of Membrane Science, 454, pp. 505– 515. doi: 10.1016/j.memsci.2013.12.027.

170

Ministry of Finance Malaysia (2015) Malaysia Economic Report 2015/2016, Chapter 3: Economic Performance and Prospect. Kuala Lumpur, Malaysia. Ministry of Health Malaysia (2010) Drinking Water Quality Standard. Available at: http://kmam.moh.gov.my/public-user/drinking-water-quality-standard.html (Accessed: 15 December 2015). Mishra, A. K. and Singh, V. P. (2010) ‘A review of drought concepts’, Journal of Hydrology. Elsevier B.V., 391(1–2), pp. 202–216. doi: 10.1016/j.jhydrol.2010.07.012. Mohamad, D. S. D. M. (1991) Malaysian: The way forward (Vision2020). Available at: http://unpan1.un.org/intradoc/groups/public/documents/apcity/unpan003223.pdf (Accessed: 16 January 2016). Mohammed, R. and Scholz, M. (2017) ‘Adaptation Strategy to Mitigate the Impact of Climate Change on Water Resources in Arid and Semi-Arid Regions: a Case Study’, Water resource management. doi: 10.1007/s11269- 017-1685-7. Mohapatra, M., Anand, S., Mishra, B. K., Giles, D. E. and Singh, P. (2009) ‘Review of fluoride removal from drinking water’, Journal of Environmental Management, 91, pp. 67–77. doi: 10.1016/j.jenvman.2009.08.015. Mohapatra, P. K., Siebel, M. A., Gijzen, H. J., Van der Hoek, J. P. and Groot, C. A. (2002) ‘Improving eco- efficiency of Amsterdam water supply: A LCA approach’, Journal of Water Supply: Research and Technology - AQUA, 51(4), pp. 217–227. Mohd Kusin, F., SYAKIRIN Md Zahar, M., Nurjaliah Muhammad, S. and Diyana Mohamad, N. (2016) ‘Hybrid off-river augmentation system as an alternative raw water resource: the hydrogeochemistry of abandoned mining ponds’, Environmental Earth Sciences, 75, pp. 230–245. doi: 10.1007/s12665-015-4981-7. Mohsen, M. S. and Jaber, J. O. (2003) ‘Potential of industrial wastewater reuse’, Desalination, 152(1–3), pp. 281–289. doi: 10.1016/S0011-9164(02)01075-5. Montgomery, M. A., Bartram, J. and Elimelech, M. (2009) ‘Increasing Functional Sustainability of Water and Sanitation Supplies in Rural Sub-Saharan Africa’, Environmental Engineering Science, 26(5), pp. 1017–1023. Morales, M. A. and Heaney, J. P. (2010) ‘Estimating Non-Residential Water Use with Publicly Available Databases’, in World Environmental and Water Resources Congress 2010: Challenges of Change, pp. 4283– 4292. Available at: http://ascelibrary.org/doi/pdf/10.1061/41114%28371%29435 (Accessed: 3 October 2017). Morales, M. A. and Heaney, J. P. (2014) ‘Classification, benchmarking, and hydroeconomic modeling of nonresidential water users’, American Water Works Association, 106(12), pp. 550–560. doi: http://dx.doi.org/10.5942/jawwa.2014.106.0150. Morales, M. A., Heaney, J. P., Friedman, K. R. and Martin, J. M. (2011) ‘Estimating commercial, industrial, and institutional water use on the basis of heated building area’, Journal - American Water Works Association, 103(6), pp. 84–96. Available at: file://homes.surrey.ac.uk/home/jaw_waternet.0073807.pdf (Accessed: 10 August 2017). Morales, M. A., Martin, J. M. and Heaney, J. P. (2009) ‘Methods for estimating commercial, industrial, and institutional water use’, in Florida Section of the American Water Works Association Water Conference. Orlando, p. 18. Available at: http://www.conservefloridawater.org/publications/10327351.pdf (Accessed: 10 August 2017). Moss, R. H. and Schneider, S. H. (2006) Climate scenarios and projections: The known, unknown, and the unknowable as applied to California. Aspen, CO. Available at: https://www.agci.org/sites/default/files/pdfs/lib/publications/2006_CSP.pdf (Accessed: 29 June 2017). Muhammad, N. (1998) Removal of heavy metals by slow sand filtration. Loughborough University. Available at: file://homes.surrey.ac.uk/home/Journals/2018/15427.pdf (Accessed: 19 February 2018). Muyibi, S. A., Raufu, A., Garoot, A. &, Eissa, S., Muyibi, S. A., Ambali, A. R. and Eissa, G. S. (2008) ‘The Impact of Economic Development on Water Pollution: Trends and Policy Actions in Malaysia’, Water Resource Manage, 22, pp. 485–508. doi: 10.1007/s11269-007-9174-z. National Water Service Commission (SPAN) (2015) ‘Buletin SPAN, Bil. 3’, Buletin SPAN, Bil. 3, 2015. National Water Service Commission (SPAN) (2016a) ‘Demand Side Management (DSM) strategy for Malaysia’. Kuala Lumpur: National Water Service Commission (SPAN).

171

National Water Service Commission (SPAN) (2016b) Sesi Taklimat Mingguan Isu Air YB Menteri (13 July 2016). Ndiaye, P. I., Moulln, P., Dominguez, L., Millet, J. C. and Charbit, F. (2005) ‘Removal of fluoride from electronic industrial effluent by RO membrane separation’, Desalination, 173, pp. 25–32. doi: 10.1016/j.desal.2004.07.042. Nguyen, C. M., Bang, S., Cho, J. and Kim, K.-W. (2009) ‘Performance and mechanism of arsenic removal from water by a nanofiltration membrane’, Desalination, 245, pp. 82–94. doi: 10.1016/j.desal.200. Nguyen, L. N., Hai, F. I., Kang, J., Price, W. E. and Nghiem, L. D. (2013) ‘Removal of emerging trace organic contaminants by MBR-based hybrid treatment processes’, International Biodeterioration & Biodegradation, 85, pp. 474–482. doi: 10.1016/j.ibiod.2013.03.014. Nicomel, N. R., Leus, K., Folens, K., Van Der Voort, P. and Du Laing, G. (2015) ‘Technologies for Arsenic Removal from Water: Current Status and Future Perspectives.’, International journal of environmental research and public health, 13(1). doi: 10.3390/ijerph13010062. Ning, R. Y. (2002) ‘Arsenic removal by reverse osmosis’, Desalination, 143, pp. 237–241. Available at: www.elsevieccom/locate/desal (Accessed: 14 January 2018). Nogueira, R., Brito, A. G., Machado, A. P., Janknecht, P., Salas, J. J., Vera, L. and Martel, G. (2009) ‘Economic and environmental assessment of small and decentralised wastewater treatment system’, Desalination and water treatment, 16(21), pp. 11–15. Available at: www.deswater.com (Accessed: 10 April 2018). O’Connor, M., Garnier, G. and Batchelor, W. (2014) ‘Life cycle assessment comparison of industrial effluent management strategies’, Journal of Cleaner Production, 79, pp. 168–181. doi: 10.1016/J.JCLEPRO.2014.05.066. Oh, K. S., Leong, J. Y. C., Poh, P. E., Chong, M. N. and Lau, E. Von (2018) ‘A review of greywater recycling related issues: Challenges and future prospects in Malaysia’, Journal of Cleaner Production, 171, pp. 17–29. doi: 10.1016/j.jclepro.2017.09.267. Olmstead, S. M. (2014) ‘Climate change adaptation and water resource management: A review of the literature’, Energy Economics, 46, pp. 500–509. doi: 10.1016/J.ENECO.2013.09.005. Olmstead, S. M. and Stavins, R. N. (2009) ‘Comparing price and nonprice approaches to urban water conservation’, Water Resources Research, 45(4). doi: 10.1029/2008WR007227. Ortiz, M., Raluy, R. G. and Serra, L. (2007) ‘Life cycle assessment of water treatment technologies: wastewater and water-reuse in a small town’, Desalination, 204(1–3), pp. 121–131. doi: 10.1016/J.DESAL.2006.04.026. Pal, P. and Kumar, R. (2012) ‘Treatment of coke wastewater: A critical review for developing sustainable management strategies’, Separation & Purification Reviews, 43(2), pp. 89–123. doi: 10.1080/15422119.2012.717161. Pasqualino, J. C., Meneses, M. and Castells, F. (2010) ‘Life Cycle Assessment of Urban Wastewater Reclamation and Reuse Alternatives’, Journal of Industrial Ecology, 15(1), pp. 49–63. doi: 10.1111/j.1530- 9290.2010.00293.x. Patil, D. S., Chavan, S. M. and Oubagaranadin, J. U. K. (2016) ‘A review of technologies for manganese removal from wastewaters’, Biochemical Pharmacology, 4, pp. 468–487. doi: 10.1016/j.jece.2015.11.028. Perrin, C., Michel, C. and Âassian, V. A. (2001) ‘Does a large number of parameters enhance model performance? Comparative assessment of common catchment model structures on 429 catchments’, Journal of hydrology, 242, pp. 275–301. Available at: http://ac.els-cdn.com/S0022169400003930/1-s2.0- S0022169400003930-main.pdf?_tid=581191d0-63f6-11e7-bd1c- 00000aab0f6c&acdnat=1499529715_5500830d2172eb2b6f9ed0c9e375ee64 (Accessed: 8 July 2017). Petala, M., Samaras, P., Kungolos, A., Zouboulis, A., Papadopoulos, A. and Sakellaropoulos, G. P. (2006) ‘The effect of coagulation on the toxicity and mutagenicity of reclaimed municipal effluents’, Chemosphere, 65(6), pp. 1007–1018. doi: 10.1016/J.CHEMOSPHERE.2006.03.035. Petrinic, I., Korenak, J., Povodnik, D. and Elix-Nielsen, C. H. (2015) ‘A feasibility study of ultrafiltration/reverse osmosis (UF/RO)-based wastewater treatment and reuse in the metal finishing industry’,

172

Journal of cleaner production, 101, pp. 292–300. doi: 10.1016/j.jclepro.2015.04.022. Piadeh, F., Alavi-moghaddam, M. R. and Mardan, S. (2018) ‘Assessment of sustainability of a hybrid of advanced treatment technologies for recycling industrial wastewater in developing countries: Case study of Iranian industrial parks’, Journal of Cleaner Production, 170, pp. 1136–1150. doi: 10.1016/J.JCLEPRO.2017.09.174. Piadeh, F., Alavi Moghaddam, M. R. and Mardan, S. (2014) ‘Present situation of wastewater treatment in the Iranian industrial estates: Recycle and reuse as a solution for achieving goals of eco-industrial parks’, Resources, Conservation and Recycling, 92, pp. 172–178. doi: 10.1016/J.RESCONREC.2014.06.004. Pillay, S. D., Friedrich, E. and Buckley, C. A. (2002) ‘Life cycle assessment of an industrial water recycling plant’, Water Science and Technology, 46(9), pp. 55–62. Available at: http://wst.iwaponline.com/content/46/9/55.abstract. Pintilie, L., Torres, C. M., Teodosiu, C. and Castells, F. (2016) ‘Urban wastewater reclamation for industrial reuse: An LCA case study’, Journal of Cleaner Production, 139, pp. 1–14. doi: 10.1016/J.JCLEPRO.2016.07.209. Pintor, A. M. A., Vilar, V. J. P., Botelho, C. M. S. and Boaventura, R. A. R. (2016) ‘Oil and grease removal from wastewaters: Sorption treatment as an alternative to state-of-the-art technologies. A critical review Oil & Grease’, Chemical Engineering Journal, 297, pp. 229–255. doi: 10.1016/j.cej.2016.03.121. du Pisani, P. L. (2006) ‘Direct reclamation of potable water at Windhoek’s Goreangab reclamation plant’, Desalination, 188(1–3), pp. 79–88. doi: 10.1016/J.DESAL.2005.04.104. Pizzol, M., Christensen, P., Schmidt, J. and Thomsen, M. (2011a) ‘Eco-toxicological impact of “metals” on the aquatic and terrestrial ecosystem: A comparison between eight different methodologies for Life Cycle Impact Assessment (LCIA)’, Journal of Cleaner Production, 19(6–7), pp. 687–698. doi: 10.1016/J.JCLEPRO.2010.12.008. Pizzol, M., Christensen, P., Schmidt, J. and Thomsen, M. (2011b) ‘Impacts of “metals” on human health: a comparison between nine different methodologies for Life Cycle Impact Assessment (LCIA)’, Journal of Cleaner Production, 19(6–7), pp. 646–656. doi: 10.1016/J.JCLEPRO.2010.05.007. Pryor, F. L. (1972) ‘The Size of Production Establishments in Manufacturing’, The Economic Journal, 82(326), pp. 547–566. Available at: http://www.jstor.org/stable/2229997 (Accessed: 26 September 2017). Purkayastha, D., Mishra, U. and Biswas, S. (2014) ‘A comprehensive review on Cd(II) removal from aqueous solution’, Journal of water process engineering, 2, pp. 105–128. doi: 10.1016/j.jwpe.2014.05.009. Qasim, S. R. (1999) Wastewater treatment plants: planning, design, and operation. 2nd edn. United States: CRC Press. Qdais, H. A. and Moussa, H. (2004) ‘Removal of heavy metals from wastewater by membrane processes: a comparative study’, Desalination, 164, pp. 105–110. Available at: https://ac.els- cdn.com/S0011916404001699/1-s2.0-S0011916404001699-main.pdf?_tid=eed8dd06-13cf-11e8-920e- 00000aacb362&acdnat=1518864622_2944409a069393af269bec75b971c255 (Accessed: 17 February 2018). Rajaram, T. and Das, A. (2008) ‘Water pollution by industrial effluents in India: Discharge scenarios and case for participatory ecosystem specific local regulation’, Futures, 40, pp. 56–69. doi: 10.1016/j.futures.2007.06.002. Raman, H. and Sunilkumar, N. (1995) ‘Multivariate modelling of water resources time series using artificial neural networks’, Hydrological Sciences -Journal-des Sciences Hydrologiques, 40(2), pp. 145–163. Available at: http://www.tandfonline.com/doi/pdf/10.1080/02626669509491401?needAccess=true (Accessed: 8 July 2017). Razavi, T., Coulibaly, P. and Asce, M. (2013) ‘Streamflow Prediction in Ungauged Basins: Review of Regionalization Methods’, Journal of hydrologic engineering, 18(8), pp. 958–975. doi: 10.1061/(ASCE). Renou, S., Thomas, J. S., Aoustin, E. and Pons, M. N. (2008) ‘Influence of impact assessment methods in wastewater treatment LCA’, Journal of Cleaner Production, 16(10), pp. 1098–1105. doi: 10.1016/J.JCLEPRO.2007.06.003. Renzetti, S. (1992) ‘Estimating the Structure of Industrial Water Demands: The Case of Canadian’, Land Economics, 68(4), pp. 396–404. Available at: http://www.jstor.org/stable/3146696 (Accessed: 22 September

173

2017). Renzetti, S. (2005) ‘Economic Instruments and Canadian Industrial Water Use’, Canadian Water Resources Journal, 30(1), pp. 21–30. Available at: http://www.tandfonline.com/doi/pdf/10.4296/cwrj300121 (Accessed: 2 October 2017). Renzetti, S. (2015) ‘Economic analysis of industrial water use’, in Dinar, A. and Schwabe, K. (Kurt A. . (eds) Handbook of water economics, pp. 87–102. Reynaud, A. (2003) ‘An Econometric Estimation of Industrial Water Demand in France’, Environmental and Resource Economics, 25, pp. 213–232. Available at: https://link.springer.com/content/pdf/10.1023%2FA%3A1023992322236.pdf (Accessed: 22 September 2017). Rodriguez-Garcia, G., Molinos-Senante, M., Hospido, A., Hernández-Sancho, F., Moreira, M. T. and Feijoo, G. (2011) ‘Environmental and economic profile of six typologies of wastewater treatment plants’, Water Research, 45(18), pp. 5997–6010. doi: 10.1016/J.WATRES.2011.08.053. Rodriguez, C., Van Buynder, P., Lugg, R., Blair, P., Devine, B., Cook, A. and Weinstein, P. (2009) ‘Indirect Potable Reuse: A Sustainable Water Supply Alternative’, International Journal of Environmental Research and Public Health, 6(3), pp. 1174–1203. doi: 10.3390/ijerph6031174. Rubio, J. and Tessele, F. (1997) ‘Removal of heavy metal ions by adsorptive particulate flotation’, Minerals Engineering, 10(7), pp. 671–679. Available at: https://ac.els-cdn.com/S0892687597000472/1-s2.0- S0892687597000472-main.pdf?_tid=8e99e7ba-13ce-11e8-9980- 00000aacb35e&acdnat=1518864031_0ae271127bfc54b7e90abc99141f8ce7 (Accessed: 17 February 2018). Sadr, S. M. K., Saroj, D. P., Kouchaki, S., Ilemobade, A. A. and Ouki, S. K. (2015) ‘A group decision-making tool for the application of membrane technologies in different water reuse scenarios’, Journal of Environmental Management, 156, pp. 97–108. doi: 10.1016/j.jenvman.2015.02.047. Safavi, H. R., Golmohammadi, M. H. and Sandoval-Solis, S. (2016) ‘Scenario analysis for integrated water resources planning and management under uncertainty in the Zayandehrud river basin’, Journal of hydrology, 539, pp. 625–639. doi: 10.1016/j.jhydrol.2016.05.073. Sailaja Kumari, B., Bhagawan, D., Himabindu, V. and Cherukuri, J. (2015) ‘Removal of fluoride from drinking water by adsorption onto Activated Alumina and activated carbon’, Journal of Engineering Research and Applications, 5(8), pp. 19–24. Available at: http://www.ijera.com/papers/Vol5_issue8/Part - 3/D58031924.pdf (Accessed: 20 February 2018). Sakr, D., Baas, L., El-Haggar, S. and Huisingh, D. (2011) ‘Critical success and limiting factors for eco- industrial parks: global trends and Egyptian context’, Journal of Cleaner Production, 19, pp. 1158–1169. doi: 10.1016/j.jclepro.2011.01.001. Sandoval-Solis, S., Mckinney, D. C., Asce, M. and Loucks, D. P. (2011) ‘Sustainability Index for Water Resources Planning and Management’, Journal of water resources planning and management, 137(5), pp. 381– 390. Available at: http://ascelibrary.org/doi/pdf/10.1061/(ASCE)WR.1943-5452.0000134 (Accessed: 29 June 2017). Sandy, T. and DiSante, C. (2010) Review of Available Technologies for the Removal of Selenium from Water. Charlotte, NC. Available at: http://www.namc.org/docs/00062756.PDF (Accessed: 20 February 2018). dos Santos Bazanella, G. C., da Silva, G. F., Vieira, A. M. S. and Bergamasco, R. (2012) ‘Fluoride Removal from Water Using Combined Moringa oleifera/Ultrafiltration Process’, Water, Air, & Soil Pollution, 223(9), pp. 6083–6093. doi: 10.1007/s11270-012-1342-y. dos Santos, J. D., Teresinha Veit, M., Moreno Palácio, S., da Cunha Gonçalves, G. and Regina Fagundes-Klen, M. (2017) ‘Evaluation of the Combined Process of Coagulation/ Flocculation and Microfiltration of Cassava Starch Wastewater: Removal Efficiency and Membrane Fouling’, Water, Air, and Soil Pollution, 228(238), p. 12. doi: 10.1007/s11270-017-3416-3. Santos, S., Ungureanu, G., Boaventura, R. and Botelho, C. (2015) ‘Selenium contaminated waters: An overview of analytical methods, treatment options and recent advances in sorption methods’, Science of The Total Environment, 521–522, pp. 246–260. doi: 10.1016/J.SCITOTENV.2015.03.107. Satterthwaite, D. (2007) The transition to a predominantly urban world and its underpinnings. Available at:

174

http://pubs.iied.org/pdfs/10550IIED.pdf (Accessed: 16 January 2016). Satterthwaite, D., McGranahan, G. and Tacoli, C. (2010) ‘Urbanization and its implications for food and farming’, Philosophical transactions of the Royal Society of London, 365(1554), pp. 2809–20. doi: 10.1098/rstb.2010.0136. Seah, H., Tan, T. P., Chong, M. L. and Leong, J. (2008) ‘NEWater-multi safety barrier approach for indirect potable use’, Water Science and Technology: Water Supply, 8(5), pp. 573–582. doi: 10.2166/ws.2008.130. Searcy, J. K., Hardison, C. H. and Langbein, W. B. (1960) Double-Mass Curves With a section Fitting Curves to Cyclic Data Manual of Hydrology: Part 1. General Surface-Water Techniques. Washington: United States Government Printing Office. Available at: https://pubs.usgs.gov/wsp/1541b/report.pdf (Accessed: 4 June 2017). Selangor State Assembly (2014) ‘Hansard, 7 April 2014 - Selangor State Assembly first meeting (Opening) of the second term’. Selangor: Selangor State Assembly, p. 14. Available at: http://dewan.selangor.gov.my/assets/pdf/Hansard/2014/Sesi 1/01.HANSARD.070414.pdf (Accessed: 5 September 2018). Selangor State Assembly (2015) ‘Statement from the selected committee on raw water resource management of the state of Selangor to the Selangor State Assembly regarding the implementation of the hybrid off river augmentation system (HORAS) 600 - 15/2015’. Selangor: Selangor State Assembly, p. 12. Available at: http://dewan.selangor.gov.my/assets/pdf/Penyata/2015/01/KERTAS BIL. 15 TAHUN 2015 - (JPK- SAM)_HORAS.pdf (Accessed: 5 September 2018). Selangor State Assembly (2017) ‘Pertanyaan-pertanyaan mulut daripada Y.B. Tuan Khasim Bin Abdul Aziz (N20 Lembah Jaya)’. Selangor: Selangor State Assembly, p. 43. Available at: http://dewan.selangor.gov.my/assets/pdf/sessions/2017/Sesi 3/bermulut/41-60.pdf (Accessed: 5 September 2018). Selcuk, H. (2005) ‘Decolorization and detoxification of textile wastewater by ozonation and coagulation processes’, Dyes and Pigments , 64, pp. 217–222. doi: 10.1016/j.dyepig.2004.03.020. Seth, D., Rehman, M. A. A. and Shrivastava, R. L. (2018) ‘Green manufacturing drivers and their relationships for small and medium(SME) and large industries’, Journal of Cleaner Production, 198, pp. 1381–1405. doi: 10.1016/j.jclepro.2018.07.106. Shaaban, A. J., Amin, M. Z. M., Chen, Z. Q., Kavvas, M. L. and Ohara, N. (2007) ‘Impact of climate change on the hydrology and water resources of Peninsular Malaysia’, in International Congress on River Basin Management. Shannon, M. A., Bohn, P. W., Elimelech, M., Georgiadis, J. G., Mariñas, B. J. and Mayes, A. M. (2008) ‘Science and technology for water purification in the coming decades’, Nature, 452(20), pp. 301–310. doi: 10.1038/nature06599. Shu, H.-Y. and Chang, M.-C. (2004) ‘Decolorization effects of six azo dyes by O3 , UV/O3 and UV/H2O2 processes’, Dyes and Pigments, 65, pp. 25–31. doi: 10.1016/j.dyepig.2004.06.014. Shuhaimen, M. S., Ahmad Ghazali, N. H., Wong, K. T. F., Loke, P. L. and Mohd Suhaimi, S. S. (2017) Rethinking Investment Incentives. Kuala Lumpur, Malaysia. Available at: www.bnm.gov.my (Accessed: 27 August 2018). Singapore Public Utility Board (2009) ‘Handbook on application for water supply’. Singapore: Singapore Public Utility Board, p. 40. Available at: https://www.pub.gov.sg/Documents/RPUB00003.pdf (Accessed: 29 December 2017). Singapore Public Utility Board (2011) Innovation in water. Singapore. Available at: http://www.pub.gov.sg/RESEARCH (Accessed: 30 December 2017). Singapore Public Utility Board (2016) ‘Singapore drinking water quality’. Singapore: Singapore Public Utility Board, p. 3. Available at: https://www.pub.gov.sg/Documents/Singapore_Drinking_Water_Quality.pdf (Accessed: 29 December 2017). Sipma, J., Osuna, B., Collado, N., Monclús, H., Ferrero, G., Comas, J. and Rodriguez-Roda, I. (2009) ‘Comparison of removal of pharmaceuticals in MBR and activated sludge systems’, Desalination, 250, pp. 653– 659. doi: 10.1016/j.desal.2009.06.073.

175

Sodamade, G. A. and Pearse, L. F. (2013) ‘Removal of iron and manganese using rapid sand filtration’, NSE Technical Transaction, 47(3), pp. 46–51. Available at: file://homes.surrey.ac.uk/home/Journals/2018/REMOVAL OF IRON AND MANGANESE USING RAPID SAND FILTRATION.pdf (Accessed: 27 February 2018). Solomatine, D. P. and Ostfeld, A. (2008) ‘Data-driven modelling: some past experiences and new approaches’, Journal of hydroinformatics, 10(1), pp. 3–22. doi: 10.2166/hydro.2008.015. Sombekke, H. D. M., Voorhoeve, D. K. and Hiemstra, P. (1997) ‘Environmental impact assessment of groundwater treatment with nanofiltration’, Desalination, 113(2–3), pp. 293–296. doi: 10.1016/S0011- 9164(97)00144-6. Song, P., Yang, Z., Zeng, G., Yang, X., Xu, H., Wang, L., Xu, R., Xiong, W. and Ahmad, K. (2017) ‘Electrocoagulation treatment of arsenic in wastewaters: A comprehensive review’, Chemical Engineering Journal. Elsevier, 317, pp. 707–725. doi: 10.1016/J.CEJ.2017.02.086. Soundharajan, B.-S., Adeloye, A. J. and Remesan, R. (2016) ‘Evaluating the variability in surface water reservoir planning characteristics during climate change impacts assessment’, Journal of Hydrology, 538, pp. 625–639. doi: 10.1016/j.jhydrol.2016.04.051. SPAN, N. W. S. C. (2015) ‘Buletin SPAN, Volume 3 (2015)’, Buletin SPAN, Bil. 3, 2015. Available at: http://www.span.gov.my/pdf/activity/2015/Buletin-SPAN-BIL-3-2015.pdf (Accessed: 13 November 2015). Srdjevic, B., Medeiros, P. and Faria, A. S. (2004) ‘An Objective Multi-Criteria Evaluation of Water Management Scenarios’, Water Resources Management, 18, pp. 35–54. Stavins, R. N. (2001) Experience with Market-Based Environmental Policy Instruments. Washington, D.C. Available at: http://www.rff.org (Accessed: 26 June 2017). Stokes, J. and Horvath, A. (2006) ‘Life Cycle Energy Assessment of Altrenative Water Supply System’, International Journal of Life Cycle Analysis, 11(115), pp. 335–343. doi: 10.1065/lca2005.06.214. Subramaniam, V. (2004) ‘Managing water supply in Selangor and Kuala Lumpur; Achieving world-class water utility company standard’, Bulrtin Ingenieur, pp. 12–20. Available at: http://www.bem.org.my/v3/body/pdf/einforeach/2004/BEM Jun04-Aug04 (Water Engineering).pdf (Accessed: 18 January 2016). Suhaila, J., Deni, S. M., Zawiah Zin, W. a N. and Jemain, A. A. (2010) ‘Trends in Peninsular Malaysia rainfall data during the southwest monsoon and northeast monsoon seasons: 1975-2004’, Sains Malaysiana, 39(4), pp. 533–542. Taddeo, R. (2016) ‘Local industrial systems towards the eco-industrial parks: the model of the ecologically equipped industrial areas’, Journal of Cleaner Production, (131), pp. 189–197. doi: 10.1016/j.jclepro.2016.05.051. Taheran, M., Brar, S. K., Verma, M., Surampalli, R. Y., Zhang, T. C. and Valero, J. R. (2016) ‘Membrane processes for removal of pharmaceutically active compounds (PhACs) from water and wastewaters’, Science of the Total Environment, 547, pp. 60–77. doi: 10.1016/j.scitotenv.2015.12.139. Talalaj, I. A. (2015) ‘Removal of organic and inorganic compounds from landfill leachate using reverse osmosis’, International Journal of Environmental Science and Technology, 12, pp. 2791–2800. doi: 10.1007/s13762-014-0661-5. Tan, J. (2012) ‘The Pitfalls of Water Privatization: Failure and Reform in Malaysia’, World Development, 40(12), pp. 2552–2563. doi: 10.1016/j.worlddev.2012.05.012. Tan, Y. L., Manan, Z. A. and Foo, D. C. Y. (2007) ‘Retrofit of Water Network with Regeneration Using Water Pinch Analysis’, Process Safety and Environmental Protection, 85(4), pp. 305–317. doi: 10.1205/psep06040. Tangsubkul, N., Beavis, P., Moore, S. J., Lundie, S. and Waite, T. D. (2005) ‘Life Cycle Assessment of Water Recycling Technology’, Water Resources Management, 19(5), pp. 521–537. doi: 10.1007/s11269-005-5602-0. Tassel, F., Rubio, J., Misra, M. and Jena, B. C. (1997) ‘Removal of mercury from gold cyanide solution by dissolved air floatation’, Minerals Engineering, 10(8), pp. 803–811. Available at: https://ac.els- cdn.com/S0892687597000587/1-s2.0-S0892687597000587-main.pdf?_tid=spdf-e1f5870e-2e19-4c2f-a306-

176

f4260d607014&acdnat=1519377914_6314600d614b0d65dffebb6684a85b08 (Accessed: 23 February 2018). Tchobanoglous, G., Butron, F. L. and Stensel, H. D. (2004) Watewater Engineering: Treatment and Reuse. Fourth. Singapore: McGraw Hill. Tchobanoglous, G., Leverenz, H., Nellor, M. and Crook, J. (2011) Direct Potable Reuse, A Path Forward, WaterReuse Research Foundation. Virginia. Available at: http://www.carollo.com/sites/default/files/tech- bulletins/DirectPotableReuse.pdf. Teresa, Z. P., Gunther, G. and Fernando, H. (2007) ‘Chemical oxygen demand reduction in coffee wastewater through chemical flocculation and advanced oxidation processes’, Journal of Environmental Sciences, 19, pp. 300–305. Available at: https://ac.els-cdn.com/S1001074207600497/1-s2.0-S1001074207600497- main.pdf?_tid=30476fb8-16ec-11e8-98cf- 00000aacb362&acdnat=1519206611_5ee6f803f73ff53071c55a945dbbfc37 (Accessed: 21 February 2018). The Cadmus Group (2006) 2003 Drinking Water Infrastructure Needs Survey: Modeling the Cost of Infrastructure. The Malaysian Water Association (MWA) (2015) Malaysia Water Industry Guide 2015. Kuala Lumpur, Malaysia. The Straits Times (2016) Toxic solvent in river, so no water for Klang Valley residents, SE Asia News & Top Stories - The Straits Times, The Straits Times. Available at: https://www.straitstimes.com/asia/se-asia/toxic- solvent-in-river-so-no-water-for-klang-valley-residents (Accessed: 5 September 2018). Tong, L., Liu, X., Liu, X., Yuan, Z. and Zhang, Q. (2013) ‘Life cycle assessment of water reuse systems in an industrial park’, Journal of Environmental Management, 129, pp. 471–478. doi: 10.1016/j.jenvman.2013.08.018. Tortajada, C. and Joshi, Y. K. (2014) ‘Water quality management in Singapore: the role of institutions, laws and regulations’, Hydrological Science Journal, 59(9), pp. 1763–1774. doi: 10.1080/02626667.2014.942664. Toze, S. (2006) ‘Reuse of effluent water—benefits and risks’, Agricultural Water Management, 80, pp. 147–159. doi: 10.1016/j.agwat.2005.07.010. Tsagarakis, K. P., Mara, D. D. and Angelakis, A. N. (2003) ‘Application of cost criteria for selection of municipal wastewater treatment system’, Water, air, and soil pollution, 142, pp. 197–210. Available at: https://link.springer.com/content/pdf/10.1023%2FA%3A1022032232487.pdf (Accessed: 14 April 2018). U.S. Environmental Protection Agency (1992) ‘Guidelines for Water Reuse’. Washington, D.C.: U.S. Environmental Protection Agency, p. 480. U.S. Environmental Protection Agency (2000) Detailed costing document for the final effluent limitations guidelines and standards for the centralized waste treatment industry. Washington, D.C. U.S. Environmental Protection Agency (2007) Wastewater Management Fact Sheet: Membrane Bioreactors. U.S.A. Uddin, M. T., Salatul, M., Mozumder, I., Figoli, A., Islam, A. and Drioli, E. (2007) ‘Arsenic removal by conventional and membrane technology: An overview’, Indian Journal of Chemical Technology, 14, pp. 441– 450. Available at: https://pdfs.semanticscholar.org/3753/97d5dabfb5574ab5585969f540fda63c768f.pdf (Accessed: 15 February 2018). UNEP (1995) Part B: Technological profiles, Source Book of Technologies for Freshwater Augmentation in Latin America and the Caribbean. Available at: http://www.unep.or.jp/ietc/Publications/techpublications/TechPub-8c/filter.asp (Accessed: 24 April 2018). UNEP (2011) Water Footprint and Corporate Water Accounting for Resource Efficiency. Available at: http://waterfootprint.org/media/downloads/UNEP-2011_1.pdf (Accessed: 1 September 2016). Urgun-Demirtas, M., Benda, P. L., Gillenwater, P. S., Negri, M. C., Xiong, H. and Snyder, S. W. (2012) ‘Achieving very low mercury levels in refinery wastewater by membrane filtration’, Journal of Hazardous Materials, 215–216, pp. 98–107. doi: 10.1016/j.jhazmat.2012.02.040. Vairavamoorthy, K., Gorantiwar, S. D. and Pathirana, A. (2008) ‘Managing urban water supplies in developing countries – Climate change and water scarcity scenarios’, Physics and chemistry of the Earth, 33, pp. 330–339. doi: 10.1016/j.pce.2008.02.008.

177

Varian, H. R. H. R. (2010) Intermediate Microeconomics: A Modern Approach. 8th edn. Edited by J. Repcheck. New York: W. W. Norton & Company. Vedwan, N., Ahmad, S., Miralles-Wilhelm, F., Broad, K., Letson, D. and Podesta, G. (2008) ‘Institutional Evolution in Lake Okeechobee Management in Florida: Characteristics, Impacts, and Limitations’, Water Resources Management, 22(6), pp. 699–718. doi: 10.1007/s11269-007-9187-7. Vincent, L., Michel, L., Catherine, C. and Pauline, R. (2014) ‘The energy cost of water independence: The case of Singapore’, Water Science and Technology, 70(5), pp. 787–794. doi: 10.2166/wst.2014.290. Vining, K. C. and Vecchia, A. V. (2007) Water-Balance Simulations of Runoff and Reservoir Storage for the Upper Helmand Watershed and Kajakai Reservoir, Central Afghanistan. United States. Available at: https://pubs.usgs.gov/sir/2007/5148/pdf/SIR2007-5148.pdf (Accessed: 22 June 2017). Vio De, F., Ribeiro, M. and Anderi Da Silva, G. (2010) ‘Life-cycle inventory for hydroelectric generation: a Brazilian case study’, Journal of Cleaner Production, 18, pp. 44–54. doi: 10.1016/j.jclepro.2009.09.006. Vlasopoulos, N., Memon, F. A., Butler, D. and Murphy, R. (2006) ‘Life cycle assessment of wastewater treatment technologies treating petroleum process waters’, Science of the Total Environment, 367(1), pp. 58–70. doi: 10.1016/j.scitotenv.2006.03.007. Vogel, R. M. and Bolognese, R. A. (1995) ‘Storage-reliability-resilience-yield relations for over-year water supply systems’, Water Resources Research, pp. 645–654. doi: 10.1029/94WR02972. Wang, W.-C., Chau, K.-W., Cheng, C.-T. and Qiu, L. (2009) ‘A comparison of performance of several artificial intelligence methods for forecasting monthly discharge time series’, Journal of Hydrology, 374, pp. 294–306. doi: 10.1016/j.jhydrol.2009.06.019. Wang, X., Zhang, J., Liu, J., Wang, G., He, R., Elmahdi, A. and Elsawah, S. (2011) ‘Water resources planning and management based on system dynamics: a case study of Yulin city’, Environment, Development and Sustainability, 13(2), pp. 331–351. doi: 10.1007/s10668-010-9264-6. Wasewar, K. L., Kumar, P., Chand, S., Padmini, B. N. and Teng, T. T. (2010) ‘Adsorption of Cadmium Ions from Aqueous Solution Using Granular Activated Carbon and Activated Clay’, Clean - Soil, Air, Water, 38(7), pp. 649–656. doi: 10.1002/clen.201000004. Wenzel, H., Dunn, R. F., Gottrup, L. and Kringelum, J. (2002) ‘Process integration design methods for water conservation and wastewater reduction in industry. Part 3: Experience of industrial application’, Clean Technologies and Environmental Policy, 4, pp. 16–25. doi: 10.1007/s10098-002-0146-y. Werner, A. T., Prowse, T. D. and Bonsal, B. R. (2015) ‘Characterizing the water balance of the Sooke Reservoir, British Columbia over the last century’, Climate, 3, pp. 241–263. doi: 10.3390/cli3010241. WHO (2006) ‘WHO Guidelines for the Safe Use of Wastewater, Excreta, and Greywater; Volume II: Wastewater Use in Agriculture’, World Health. France: World Health Organization (WHO), p. 204. doi: 10.1007/s13398-014-0173-7.2. Wilhite, D. A. and Glantz, M. H. (1985) ‘Understanding: the Drought Phenomenon: The Role of Definitions’, Water International, 10(3), pp. 111–120. doi: 10.1080/02508068508686328. WMO (2007) The Role of Climatological Normal in a Changing Climate. Geneva. Available at: https://www.wmo.int/datastat/documents/WCDMPNo61_1.pdf (Accessed: 4 June 2017). World Bank (2015) Purchasing Power Parities and the Real Size of World EconomiesL A Comprehensive Report of the 2011 International Comparison Program. Washington, DC. World Health Organization (2008) ‘Guidelines for drinking-water quality.’, WHO chronicle, 38(3), p. 564. doi: 10.1016/S1462-0758(00)00006-6. Worthington, A. C. (2010) ‘Commercial and Industrial Water Demand Estimation: Theoretical and Methodological Guidelines for Applied Economics Research’, Estudios de Economía Aplicada, 28(2). Available at: http://www.redalyc.org/html/301/30120357002/ (Accessed: 12 September 2017). Worthington, A. C. and Hoffman, M. (2008) ‘An Empirical Survey of Residential Water Demand Modelling’, Journal of Economic Surveys, 22(5), pp. 842–871. doi: 10.1111/j.1467-6419.2008.00551.x.

178

Wu, C. L., Chau, K. W. and Li, Y. S. (2009) ‘Predicting monthly streamflow using data-driven models coupled with data-preprocessing techniques’, Water Resources Research, 45(8), p. 23. doi: 10.1029/2007WR006737. Xiao, Y., Fang, L. and Hipel, K. W. (2018) ‘Agent-Based Modeling Approach to Investigating the Impact of Water Demand Management’, Journal of Water Resources Planning and Management, 144(3). doi: 10.1061/(ASCE)WR.1943-5452.0000907. Xu, C. (2002) Hydrological Models. Edited by C. Xu. Sweden: Uppsala University Department of Earth Sciences Hydrology. Available at: http://www.soil.tu- bs.de/lehre/Master.Unsicherheiten/2012/Lit/Hydrology_textbook.pdf (Accessed: 8 July 2017). Yadav, M., Wagener, T. and Gupta, H. (2007) ‘Regionalization of constraints on expected watershed response behavior for improved predictions in ungauged basins’, Advances in water resources, 30, pp. 1756–1774. doi: 10.1016/j.advwatres.2007.01.005. Zarfl, C., Lumsdon, A. and Berlekamp, J. (2014) ‘A global boom in hydropower dam construction’, Aquatic Sciences, 77(1), pp. 161–170. doi: 10.1007/s00027-014-0377-0. Zepon Tarpani, R. R. and Azapagic, A. (2018) ‘Life cycle environmental impacts of advanced wastewater treatment techniques for removal of pharmaceuticals and personal care products (PPCPs)’, Journal of Environmental Management, 215, pp. 258–272. doi: 10.1016/J.JENVMAN.2018.03.047. Zhang, N., Lin, L.-S. and Gang, D. (2008) ‘Adsorptive selenite removal from water using iron-coated GAC adsorbents’, Water Research, 42, pp. 3809–3816. doi: 10.1016/j.watres.2008.07.025. Zhang, Y., Ma, C., Ye, F., Kong, Y. and Li, H. (2009) ‘The treatment of wastewater of paper mill with integrated membrane process’, Desalination, 236(1–3), pp. 349–356. doi: 10.1016/j.desal.2007.10.086. Zouboulis, A. I., Sarasidis, V. and Moussas, P. A. (2010) ‘Removal of Copper From Synthetic Wastewaters by the Hybrid Coagulation–Microfiltration Process’, Separation Science and Technology, 45(11), pp. 1658–1666. doi: 10.1080/01496395.2010.487743.

179

Appendix A: Incidents of water rationing as reported in various tabloids Year State affected Reference 2005 Johor June, 2005: Seremban rations water Negeri Sembilan July, 2005: Water rationing imposed July, 2005: 24-hour water rationing July, 2005: Rain prayer answered for Mersing folk 2009 Sarawak August, 2009: Sarawak starts water rationing Sabah September, 2009: 85,000 people hit by water shortage in Labuan 2010 Johor February, 2010: Water rationing in Kluang to be affected Negeri Sembilan from today to March 13 March, 2010: Water rationing in Seremban to begin Sunday March, 2010: Sabah CM calls for water rationing March 2010: Possible water rationing for Penang if dry spell continues March, 2010: Dry spell may last until May March, 2010: Water rationing in Batu Pahat for one month April, 2010: JB in for water rationing 2012 Kuala Lumpur / July, 2012: Syabas wants to start water rationing Selangor13 July, 2012: Syabas wants to conduct water rationing immediately (Update) 2013 Kuala Lumpur / March, 2013: Syabas carries out water rationing in KL Selangor 2014 Kuala Lumpur / November, 2014: ‘Many industries suffered losses due to Selangor water rationing’ Perak August, 2014: Kossan posts higher profit, turnover hit by Johor month-long water cut in Selangor Negeri Sembilan June, 2014: Unscheduled water disruption because of Melaka over-consumption and dry spell, says Syabas May, 2014: Water rationing ends today

13 Water operator requested permission from SPAN to implement water rationing due to their inability to supply treated water to consumers

180

April, 2014: State FMM members seek alternative water supply sources April, 2014: Syabas queried over water sold to industries April, 2014: Water rationing in zone one, Taiping district to stop Thursday April, 2014: Factories: There is 'cost' for concern April, 2014: Rationing of water causing havoc for factories April, 2014: Kelantan looks to groundwater to keep the taps running March, 2014: Penang mulls need for water rationing March, 2014: Water rationing called off in five districts in Johor and Negri following rains March, 2014: Govt may be forced to declare water emergency 2015 Johor September, 2015: Folks in Tampin and Rembau face water Negeri Sembilan cuts Kuala Lumpur / September, 2015: Water warning irks Selangor govt Selangor14 Dec, 2015: Friday, 4 December 2015 2016 Johor March, 2016: It rained in Sabah but not enough to beat Sabah drought Perlis March, 2016: Wan Junaidi: Four dams at critical levels nationwide April, 2016: Span: Start water rationing April, 2016: Start water rationing now or be sorry, SPAN tells Perak govt April, 2016: Pahang spends RM60,000 daily to pump water into treatment plants April, 2016: Four million at risk of water cuts April, 2016: Beris Dam in Kedah at critical level April, 2016: Water rationing begins in northern Perlis April, 2016: Massive water crisis threatens nation

14 SPAN urged State to implement water rationing to prolong raw water availability

181

Appendix B: NAM model using MIKE Hydro Basin software B.1 Methodology In order to simulate the storage level in each reservoir, the rainfall-runoff (RR) model (NAM model using MIKE HYDRO Basin software) was used to estimate the inflow and was first calibrated using measured streamflow data in a gauged catchment. The NAM model was chosen due to its ability to simulate streamflow within a given catchment while incorporating precipitation, temperature, and evapotranspiration data (Agrawal and Desmukh, 2016). Furthermore, the software had also allow for the incorporation of reservoir operation rule into the estimation of inflow, which enables the simulation to be linked to a reservoir’s output and water level at a given time.

Based on the Department of Irrigation and Drainage’s (DID) record, all three reservoir catchments did not have any record of long-term streamflow measurements. In their study to assess the future trend of water availability in Klang Gates dam, Goh, Zainol and Mat Amin (2015) had calibrated the Klang Gates dam catchment against the downstream catchment (station ID: SF 3116430) which is an observed streamflow station operated and maintained by DID. However, upon further investigation, a gauged catchment between Langat and Semenyih dam presented a more ideal comparison for all three reservoirs as it was observed to be similar in size, topography, and land use. In addition, the availability of rainfall and evaporation data obtained from DID was extremely limited for the three reservoir catchments, which risks the model being inadequately calibrated. Thus, estimation of inflow was done using parameter transfer method. The NAM model was first calibrated and validated using measured streamflow data at a gauged catchment of similar physical characteristics and size, and the parameters was then used to estimate the streamflow in an ungauged reservoir catchments. Here, the catchment with the ID of SF3118445 was chosen to be used to calibrate all three reservoir catchments.

The NAM model is a conceptual, lumped hydrological model, which estimates the rainfall- runoff relationship by simulating the overland flow, interflow, and baseflow of a catchment (DHI, 2003). Simulations of each flows us done by continuously accounting for the water storage in different but mutually interrelated storages: snow storage (if any), surface storage, root zone storage (subsurface), and groundwater storage (Lan Anh et al., 2008). Basic data requirements for the NAM model include catchment area, initial conditions, and concurrent time series of precipitation, potential evapotranspiration, and streamflow discharge (for calibration and validation purposes) (DHI Water & Environment, 2008). It had required the

182

calibration of nine model parameters, which was first calibrated automatically using the auto- calibration function, then further adjusted manually to optimise each parameter. The nine parameters included were:

i. Maximum water content in surface storage (Umax) ii. Maximum water content in root zone storage (Lmax) iii. Overland flow runoff coefficient (CQOF) iv. Time constant for interflow (CKIF) v. Time constants for routing overland flow (CK1, 2) vi. Root zone threshold value for overland flow (TOF) vii. Root zone threshold value for interflow (TIF) viii. Time constant for routing baseflow (CKBF) ix. Root zone threshold value for ground water recharge (Tg)

The NAM model’s accuracy during the calibration period using the auto-calibration function was done based on a goodness-of-fit measures which was automatically optimised using the following numerical performance (objective functions):

i. Agreement between the average simulated and observed runoff (overall volume error): water balance ii. Overall agreement of the shape of the hydrograph: overall root mean square error (RMSE) iii. Agreement between the simulated and observed runoff (model performance): Pearson correlation coefficient (R) iv. Agreement between the simulated and observed runoff (model performance):

Nash-Sutcliffe efficiency (RNS)

To calculate the nine parameters above, data included in the model were primarily time series of precipitation, evapotranspiration, and streamflow. In order to ensure high accuracy during the calibration and validation of the NAM model, precipitation data used was checked for data completeness and consistency, using normal ratio method for the former, and double- mass curve for the latter. As the normal annual precipitation of the base station has a variation of more than 10%, the normal ratio method was used to fill in missing gaps using the following equation (Eq. B-1):

183

Eq. B-1

Where:

Px – missing precipitation at base station n – number of index stations

Pi – precipitation of the same period for the ‘ith’ index station

Nx – normal annual precipitation for base station

Ni – normal annual precipitation for ‘ith’ station

In order to determine the consistency of rainfall and streamflow data, a double-mass plot was plotted against precipitation and streamflow time series data of an adjacent station. Theoretically, concept behind the double-mass curve has been based on the fact that a graph of the accumulation of one parameter at one station against the accumulation of another parameter at a neighbouring station will plot a straight line, so long that the data are proportional (Searcy, Hardison and Langbein, 1960). A break in the curve suggest that there have been a change in the constant proportionality between the two variables, or that the proportionality is not a constant at all rates of accumulation. For precipitation, it is rarely affected by man-made changes; thus any breaks in the curve suggest a change in gage location, exposure, or observational methods. However, for streamflow data, the assumption of proportionality between two neighbouring stations might not be valid as conditions between the catchment or basin might be different (e.g.: difference in land-use, or differences in water users which influence the volume of water abstraction). However, if the variable ratio between the two stations are constant, minimised, or ignored, major breaks in the curve’s slope indicates the time which a change had occurred in the relation between the two quantities (Searcy, Hardison and Langbein, 1960).

B.2 Results and findings In order to calibrate the NAM model, daily observed streamflow, observed water level, and streamflow rating curve for station SF 3118445 was obtained from DID for the year 1986 to 2016, in addition to rainfall data for stations within close proximity of the catchment. SF3118445 catchment in relation with the three reservoir catchments.

184

The two parameters that were needed in order to calibrate the NAM model for SF 31185445’s catchment are rainfall and evaporation. Of the stations within DID’s database, only five stations were found to be within SF 3118445’s catchment. As for the evaporation station, there was only one station within the whole study site (EV 3117370); however, both the duration of the observed data and the percentage of missing data were less than the required minimum 30 years or 10% cut-off point, which is required to ensure that the data recorded by the station encompasses a complete description of the climate over a period of time (WMO, 2007). As there were no other data which is available to be used, the average monthly data were calculated based on available data instead. Details regarding the rainfall and evaporation stations used are summarised in Table B-1, while Table B-2 illustrate the calculated averaged observed evaporation data.

Table B-1: Rainfall and evaporation stations within SF 3118445 catchment

Station Station name Start year End year Total years Missing data number (%) RF Sek. Keb. Lg Sg. 01/01/1986 31/12/2016 31 5.23 3118102 Lui RF Sg. Gabai di Kg. 01/01/2008 31/12/2016 9 7.36 3118103 Lui RF Sawah Sg. Lui 01/01/1993 31/12/2016 24 17.95 3119001 RF Lalang Sg. Lui 01/01/2008 31/12/2016 9 0.91 3119002 RF Genting Peres 01/01/1993 31/12/2016 25 11.55 3119104 EV Pusat penyelidikkan 01/01/1985 31/05/2002 16 18.27 3117370 JPS Ampang

185

Table B-2: Calculation of average observed evaporation data for station EV 3117370

Jan Feb Mar Apr May June Jul Aug Sept Oct Nov Dec 1986 3.65 5.18 5.09 4.57 5.18 4.66 4.08 5.34 4.21 4.38 4.50 5.17 1987 - - 5.89 5.88 4.49 4.29 4.76 4.68 4.61 5.23 4.66 3.43 1988 4.59 4.92 4.81 4.78 4.82 4.22 4.23 3.83 4.57 4.73 4.28 4.21 1989 3.87 4.66 4.57 4.57 4.69 5.32 4.93 5.02 5.16 5.18 5.18 5.24 1990 4.53 5.52 4.69 5.00 - 5.12 4.30 5.38 4.67 4.88 3.93 4.73 1991 - - 5.28 4.48 5.08 4.97 5.23 4.62 5.47 4.88 5.30 5.06 1992 5.20 4.98 5.69 5.64 5.42 5.65 5.89 5.72 5.70 5.87 5.52 5.59 1993 5.48 - 5.85 - 5.96 - 5.98 4.36 4.46 3.90 - 3.64 1994 4.35 4.85 - 4.28 4.14 4.66 4.73 3.86 4.33 4.63 4.69 4.55 1995 4.69 4.60 - 4.88 4.75 4.84 - - 5.27 - 4.69 - 1996 4.84 - 5.42 - - - 4.81 - 5.20 4.77 4.13 3.82 1997 4.48 - 4.79 - 4.98 4.27 4.96 4.72 3.33 - - 5.05 1998 4.90 5.42 5.14 5.84 4.82 4.40 4.90 4.54 4.70 5.12 4.94 3.73 1999 3.79 4.96 4.54 5.31 4.60 4.62 5.19 4.96 4.73 4.12 5.22 3.96 2000 - 5.14 4.53 4.83 4.82 4.24 4.96 4.99 4.90 4.74 4.33 4.25 2001 4.30 4.26 5.65 5.31 4.88 5.17 5.42 4.56 4.83 4.38 4.15 4.44 Average 4.51 4.95 5.14 5.03 4.90 4.75 4.96 4.76 4.76 4.77 4.68 4.46

In order to fill in the rainfall data gaps for each station, two conditions need to be met; observed missing data must be less than 10%, and period of observation must be more than 30 years. Of the five stations, only RF 3118102 has met the specified criteria; as such only this station was fully treated for missing data using adjacent stations. Rainfall stations adjacent to station RF 3118102 were chosen based on its distance from the base station, total period of observed data, and percentage of missing data. Table B-3 summarised the stations which were used to fill in the gaps of the base station. As the variation of normal annual precipitation of the adjacent stations was more than 10% as compared to the base station, normal ratio method were used. Accuracy of normal ratio method was assessed using the correlation coefficient (R) as well as mean absolute error (MAE) (WMO, 2007), both of which were 0.52 and -0.06 respectively.

186

Table B-3: Stations used to fill in rainfall gaps for station RF 3118102

Station Station name Start year End year Total Missing Distance number years data (%) from base station (km) RF 3218101 TNB 01/01/1986 31/12/2016 31 8.67 4.00 Ponsoon RF 3217004 Kuala Seleh 01/01/1986 31/12/2016 31 9.86 14.89 RF 3217002 Empangan 01/01/1986 31/12/2016 31 1.93 14.97 Genting Klang RF 3018107 Ldg. 01/01/1986 31/12/2016 31 6.22 18.83 Dominion

Finally, each rainfall stations within SF 3118445’s catchment were checked for consistency of observed data using double-mass curve method against one another in order to determine if there were any inconsistencies in the dataset (Searcy, Hardison and Langbein, 1960). Results of the double-mass curve for each station showed no major breaks in rainfall accumulation in any of the stations, which indicates no major changes in proportionality between the observed data. Summary of the results of the double-mass curve is illustrated in Table B-4, while an example of the double-mass curve plot between station RF 3118102 and RF 3118103 is shown in Figure B-1.

Table B-4: Result of double-mass curve for each station

Consistency RF 3118102 RF 3118103 RF 3119001 RF 3119002 RF 3119104 RF 3118102 1 0.9993 0.9985 0.9992 0.9971 RF 3118103 0.9993 1 0.9985 0.9979 0.9994 RF 3119001 0.9985 0.9985 1 0.9957 0.9981 RF 3119002 0.9992 0.9979 0.9957 1 0.9992 RF 3119104 0.9971 0.9994 0.9981 0.9992 1

187

25000

R² = 0.9993

20000

15000

10000

5000 Accummulated Accummulated rainfall, 3118103 RF (mm)

0 0 5000 10000 15000 20000 25000 Accummulated rainfall, RF 3118102 (mm)

Figure B-0-1: Double-mass curve between station RF 3118102 and RF 3118103

To determine the period for NAM model calibration, completeness of rainfall and streamflow data from all the stations were cross referenced with one another, in addition to the availability of streamflow rating curve, which is summarised in Table B-5 below. Due to the distribution of the rainfall stations within the catchment, preference were given to the years which have observed rainfall data from all five stations, in addition to a high percentage of data completeness for the observed streamflow data, which rules out data from year 1986 to 2007. Furthermore, data for the year 2008 was also omitted as station RF 3118103 was well below the excepted limit of 90%. As such, period of 2009 to 2016 were initially considered to NAM model calibration and validation.

Table B--5: Rainfall and streamflow data completeness and streamflow rating curve check

Data completeness (%) Streamflow RF RF RF RF RF SF rating curve 3118102 3118103 3119001 3119002 3119104 3118445 1986 100 0 0 0 0 78.36 Yes 1987 100 0 0 0 0 100 Yes 1988 100 0 0 0 0 100 Yes

188

1989 97.81 0 0 0 0 89.59 Yes 1990 100 0 0 0 0 64.38 Yes 1991 100 0 0 0 0 98.36 Yes 1992 100 0 0 0 7.65 89.89 Yes 1993 100 0 54.79 0 82.19 96.44 Yes 1994 100 0 66.03 0 100 96.99 Yes 1995 100 0 36.44 0 57.26 74.25 Yes 1996 100 0 40.71 0 62.57 96.99 Yes 1997 100 0 82.47 0 100 95.07 No 1998 100 0 27.12 0 79.45 43.01 No 1999 100 0 48.22 0 95.89 65.75 Yes 2000 100 0 100 0 97.54 91.80 Yes 2001 100 0 63.29 0 44.66 94.25 Yes 2002 100 0 92.05 0 93.70 88.77 Yes 2003 100 0 75.89 0 100 92.05 Yes 2004 100 0 92.08 0 98.63 62.57 Yes 2005 100 0 92.05 0 92.05 70.96 No 2006 100 0 100 0 100 87.95 No 2007 100 0 100 0 100 96.99 Yes 2008 100 34.15 100 92.08 100 100 Yes 2009 100 100 98.36 100 100 96.44 Yes 2010 100 100 100 100 100 100 Yes 2011 100 100 100 100 100 93.42 No 2012 100 100 100 100 100 100 No 2013 100 100 100 100 100 100 No 2014 100 100 100 100 100 100 Yes 2015 100 100 100 100 100 100 No 2016 100 100 100 100 100 100 Yes

To check for consistency for the observed streamflow data for the year 2009 to 2016, double- mass curve was done for station SF 3118445 against SF 2917401, where SF 3118445 is a nested catchment within SF 2917401. It should be noted however that although the double- mass curve method is usually used to check the consistency of observed streamflow data, the

189

assumption of a constant ratio of monthly discharge between two stations may not be valid, as there might be other variables which might influence the annual discharge within one basin which does not affect another (Searcy, Hardison and Langbein, 1960). Nevertheless, this method has been known to provide reasonable estimation of the consistency of observed streamflow data; result of the double-mass curve is illustrated in Figure B-2 below.

The result of the plot indicated a major break in the double-mass curve from year 2009 to 2010, suggesting a possibility of an external factor causing the inconsistency to the data. This could be due to the regulating reservoir situated upstream of station SF 2917401 which does not influence the observed streamflow at station SF 3118554, or a change in measurement method in one or both stations. Post 2010 however, the relationship between the two stations were fairly consistent.

The streamflow rating curve which was available from 2010 to 2016 were then checked for major changes, as there were gaps of several years in which the curve was updated (Table B- 5). Based on the plotted curve (Figure B-3), it was observed that there were major changes in the relationship between the water level and streamflow sometime between 2010 and 2014; this could explain the increase in discharge rate while the water level remained fairly consistent, as illustrated in Figure B-0-4. In contrast, the rating curve for the year 2014 and 2016 remained similar. Due to this, the period of 2014 to 2016 were chosen for calibration of the NAM model.

14000

/s) 3 12000

10000

8000

6000

4000

2000 Accummulated Accummulated streamflow, 3118445 SF (m 0 0 5000 10000 15000 20000 25000 30000 35000 Accummulated streamflow, SF2917401 (m3/s)

Figure B-0-2: Double mass curve for SF 2917401 and SF3118445

190

300000

250000

200000

150000

Discharge, Discharge, m3/s 100000

50000

0 73000 74000 75000 76000 77000 78000 79000 80000 81000 Stage, m

2010 2014 2016

Figure B-0-3: Streamflow rating curve for SF 3118445

For period of 2014-2016, all five rainfall stations had complete data. As such, weightage for each rainfall stations were calculated using Thiessen polygon method to determine the mean precipitation for the catchment using ArcGIS 10.2 software; the weight for each stations are summarised in Table B-6. Furthermore, the average monthly evaporation from station EV 3117370 was used, where the evapotranspiration value were calculated by multiplying the rate of evaporation within the catchment by a pan coefficient of 0.75 (Jabatan Pengairan dan Saliran Malaysia, 1976). All treated rainfall and evaporation data were supplemented into the NAM model in MIKE HYDRO Basin software.

Table B-6: Weightage for rainfall stations in SF 31183445 catchment

Station Area (km2) Weight RF 3118102 36.54 0.08 RF 3118103 10.79 0.03 RF 3119001 127.82 0.30 RF 3119002 22.86 0.05 RF 3119104 232.00 0.54

Simulations were first done using the auto-calibration function in the MIKE HYDRO Basin software to optimise the parameters for the model, and were further adjusted manually by

191

trial and error to increase the accuracy of the simulated streamflow. The model was calibrated from 2013 to 2016 (Figure B-4), where results from 2013 were discarded to allow for the model to stabilize as the result of this period will depend more on the initial conditions rather than the parameter value.

In order to determine the accuracy of the results, they were assessed using four criteria: overall water balance error (WBer), root mean square error (RMSE), correlation coefficient

(R), and the Nash-Sutcliffe efficiency (RNS) (Goh, Zainol and Mat Amin, 2015). The closer the WBer and RSME is to zero, while R and RNS is to 1, the better the performance of the model. Result of the simulation and accuracy assessment for 2014-2016 are illustrated in Figure B-5 and Table B-7 respectively.

Table B-7: Accuracy assessment of model validation for year 2014 - 2016

Simulation Water Root mean square Correlation Nash-Sutcliffe

period balance error (RMSE) coefficient efficiency (RNS)

(WBer) (R) 2014-2016 0.30 7.70 0.63 -1.22

In an initial observation of the weighted rainfall pattern as compared to the observed streamflow, there were instances where the observed streamflow was higher than the volume of rainfall, specifically in the early period of the simulation (2014); this could be one of the reasons why the RMSE of the model was high as it could not replicate the peak flow. However, the increase in discharge in the events of relatively low rainfall also indicates that the streamflow rate is sensitive to the baseflow and lowers baseflow rate, as well as the time constant for routing the overflow, interflow, and baseflow.

192

30 79

25 78

76 15

75 Stage, m Stage, 10

Streamflow, Streamflow, m3/s 74

5 73

0 72

Qobs (m3/s) Wlobs (m)

Figure B-0-4: Observed water level and streamflow for SF 3118445 (1986-2016)

193

14 0

12 20

/s 10

8 60 80 6

100 Rainfall,mm

Streamflow, Streamflow, m 4 120 2 140 0

Rainfall Qobs Qsim

Figure B-0-5: Result of calibration for SF 3118445, period: 2014-2016

194

Unfortunately, even after rigorous attempts at calibrating the model, the result of the simulation had only managed to replicate the overall pattern of the streamflow discharge and not the volume – specifically the baseflow – for the catchment. The result of the simulated streamflow was consistently lower than observed streamflow, especially the low flows. The lower overall volume of the simulated streamflow could be due to inadequate representation of actual rainfall within the catchment due to limited rainfall stations, especially in the higher elevations which makes up approximately 20% of the whole catchment. As such, the model was not able to simulate the overall water balance within the catchment; this was apparent during calibrating the model. On the one hand, the model was able to achieve high accuracy for water balance, but the overall RMSE was high – this indicate that it was able to account for the water balance, but was not able to simulate the low flow and peak flow accurately. On the other hand, in the instances where the RMSE was relatively low, the water balance would be high. Additionally, as illustrated above, there were no other slice of time period in which the streamflow and precipitation data can be used to validate the model in order to ensure high validity. Couple with multiple sources of uncertainties which could further influence the accuracy of the result, another method was used to estimate the inflow into the reservoir.

195

Appendix C: List of parameters monitored and upper limit from Fifth and Seventh Schedule

Unit Standard A Standard B Temperature °C 40 40 PH - 6.0 – 9.0 5.5 – 9.0

BOD5 at 20°C mg/L 20 50 COD mg/L 80 200 Suspended solids mg/L 50 100 Mercury mg/L 0.005 0.05 Cadmium mg/L 0.01 0.02 Chromium, mg/L 0.05 0.05 Hexavalent Chromium, Trivalent mg/L 0.20 1.0 Arsenic mg/L 0.05 0.10 Cyanide mg/L 0.05 0.10 Lead mg/L 0.10 0.5 Copper mg/L 0.20 1.0 Manganese mg/L 0.20 1.0 Nickel mg/L 0.20 1.0 Tin mg/L 0.20 1.0 Zinc mg/L 2.0 2.0 Boron mg/L 1.0 4.0 Iron (Fe) mg/L 1.0 5.0 Silver mg/L 0.1 1.0 Aluminium mg/L 10 15 Selenium mg/L 0.02 0.5 Barium mg/L 1.0 2.0 Fluoride mg/L 2.0 5.0 Formaldehyde mg/L 1.0 2.0 Phenol mg/L 0.001 1.0 Free Chlorine mg/L 1.0 2.0 Sulphide mg/L 0.50 0.50

196

Oil and Grease mg/L 1.0 10 Ammoniacal mg/L 10 20 Nitrogen Colour ADMI15 100 200

15 American Dye Manufacturers Institute

197

Appendix D: Maximum allowable limit for non-potable and potable water reuse

Parameter Non-potable water standard, Potable water standard, Singapore and Iranian Malaysia drinking water industrial water reuse standard (Ministry of Health standard, and WHO Malaysia, 2010) Guidelines for the safe use of (unit in mg/l unless otherwise wastewater, excreta, and stated) greywater in agriculture (WHO1, 2006; Singapore Public Utility Board2, 2009; Piadeh, Alavi-moghaddam and Mardan3, 2018) (unit in mg/l unless otherwise stated) Total Coliform 0 in 100 ml3 0 in 100 ml E.coli - 0 in 100 ml Turbidity 0.5 – 2.0 NTU2 5 NTU Colour 10 – 20 HU2 15 TCU pH 6.8 – 7.32 6.5 – 9.0 Biological Oxygen 1.0 – 5.02 - Demand Chemical Oxygen 0 - 103 - Demand Suspended Solids 0 - 53 - Total Dissolved Solids 0 – 1003 1000 Total Solid 350 – 1,3002 - Hardness 100 - 2502 500 Chloride 100 – 5002 250 Free Residual Chlorine - 0.2 – 5.0 Combined Chlorine - Not less than 1.0 Conductivity 600 – 1,600 umhos/cm2 -

198

Ferum/Iron 5.01 0.3 Fluoride 1.01 0.4 – 0.6 Aluminium 5.01 0.2 Manganese 0.201 0.1 Cadmium 0.011 0.003 Arsenic 0.11 0.01 Plumbum/lead 5.01 0.01 Chromium 0.11 0.05 Cuprum/Copper 0.21 1.0 Zinc 2.01 3.0 Phosphate 1 – 42 - Selenium 0.021 0.01 Nickel 0.21 0.02 Clostridium perfringens - Absent (including spores) Coliform bacteria - - Colony count 22° - - Conductivity - - Enterococci - - Odour - - Taste - - Oxidisability - - Ammonia - 1.5 Nitrate - 10 Anionic Detergent - 1.0 MBAS Nitrite - - Total Organic Carbon - - (TOC) Mercury - 0.001 Cyanide - 0.07 Natrium/Sodium - 200

199

Sulphate - 250 Argentum - 0.05 Magnesium - 150 Mineral Oil - 0.3 Chloroform - 0.2 Bromoform - 0.1 Dibromochloromethane - 0.1 Bromodichlormethane - 0.06 Phenol - 0.002 Antimony - 0.005 Dibromoacetonitrile - 0.1 Dichloroacetic acid - 0.05 Dicholoacetonitrile - 0.09 Tricholoacetic acid - 0.1 Trichloroacetonitrile - 0.001 Trihalomathanes - total - 1.0 Aldrin / Dealdrin - 0.00003 DDT - 0.002 Heptachlor & - 0.00003 Heptachlor Epoxide Methoxychlor - 0.02 Lindane - 0.002 Chlordane - 0.0002 Endosulfan - 0.03 Hexachlorobenzene - 0.001 1,2-dichloroethane - 0.03 2,4,5- - 0.009 Trichlorophenoxyacetic acid 2,3,6-tricholophenol - 0.2 2,4- - 0.03 Dichlorophenoxyacetic acid

200

2,4-dichlorophenoxy - 0.09 2,4-dichlorophenol - 0.09 Acrylamide - 0.0005 Alachlor - 0.02 Aldicarb - 0.01 Benzene - 0.01 Carbofuran - 0.007 MCPA - 0.002 Pendimethalin - 0.02 Pentachlorophenol - 0.009 Permethrin - 0.02 Pesticides - - Pesticides - Total - - Polycyclic aromatic - - hydrocarbons Propanil - 0.02 Tetrachloroethene and - - Trichloroethene Vinyl chloride - 0.005 Gross alpha (α) - 0.1 Bq/l Gross beta (β) - 1.0 Bq/l Tritium - - Total indicative dose - -

201

Appendix E: Coefficient accuracy assessment

E.1 Overall profile for light, medium, and heavy industry Table E-1 summarised the number of customer points (CPs) that was used to estimate the water intensity coefficient. Of the 15,612 points, only 15,193 CPs were eventually used, as the 419 points had missing information regarding types of industrial classification.

Table E-1: Number of consumer points (CP) used for coefficient estimation

Service area (SA) Total CPs CP within SA CP outside SA Semenyih 479 349 130 Kuala Lumpur 223 137 86 Petaling Jaya 8,932 2,161 6,771 Hulu Langat 2,738 3,806 33 Kuala Langat 103 103 - Gombak 3,137 - 3,137 Total 15,612 5,455 (35%) 10,157 (65%) Table E-1 describes basic statistics for area and water consumption per premise, according to type/class of industry. From the table, it can be observed that the range for both area and water consumption for all three types of industry was large, with large standard deviation, indicating large variation of water consumption within each class. The high variability for both variables within each industry was due to the broad categories of industrial activity within each type of industry. According to the classification standard set by the Selangor Town and Country Planning Department as well as the Department of Environment, Malaysia, activities categorised within each industry type were classified based on wastewater discharge, schedule waste production, and noise emission, amongst others. Due to this, activities which was grouped within each class varies greatly; for example, both ‘manufacture of distilled water’ and ‘manufacturing of jewellery’ were classified as ‘light’ industries, while ‘waste treatment and disposal’, and ‘water collection, treatment, and supply’ were classified as ‘heavy’. However, in terms of water consumed during each manufacturing process, volume consumed by each industrial activities differs greatly, which contributed to the high variability of water consumption within each industry type. As such, additional information regarding the sub-classification of industrial activities within each industrial class could greatly reduce the variability of water consumption. However, arguments has also been made in several literature that although the high variability for industrial water consumption

202

as a whole is unavoidable (e.g. Billings and Jones, 2011), further disaggregation of the data could lead to unexplainable variation, which, in turn, would require additional information to explain (Dziedzic et al., 2015; Kieger, Krentz and Dziegielewski, 2015). For this research specifically, additional disaggregation for type of activities within each industrial class could further refine the calculation of the coefficient; however, the dataset was not available for use and thus, only the three industrial classification was used.

Table E-2: Descriptive statistics for light, medium, and heavy industry

Minimum Maximum Mean Std. Skewnes dev. s Light Area, m2 59.00 55,471.00 397.00 1,306.21 18.39 Water 0 66,894.44 92.98 1,237.84 36.07 consumption, m3 Medium Area, m2 115.00 79,626.00 1,401.00 3,273.95 10.40 Water 0 38,144.44 175.65 1,036.17 22.65 consumption, m3 Heavy Area, m2 210.00 18,851.00 2,296.00 2,825.90 2.92 Water 0 20,772.33 337.25 1,424.06 11.41 consumption, m3

As the calculation of coefficient is dependent on both the area and water consumption, correlation between the two variables were tested; results for the correlation are summarised in Table E-3 below. For light industry, a strong, liner correlation between both variables could be observed, while a medium monotonic relationship for medium and heavy type industry was observed instead. Results of the correlation test suggest that there is a liner relationship between area and volume of water consumption for light industry, indicating an increase in water consumption as size of premise increase. In comparison, for medium and heavy industries, results indicate that water consumption will also increase as the area of a premise increase, but to a lesser degree than light industry.

203

Table E-3: Correlation between area and water consumption for light, medium and heavy industry

N N’ Pearson Sig. Spearman's Sig. correlation rho Light 9,084 344 0.611 0.00 0.294 0.00 Medium 5,519 214 0.365 0.00 0.474 0.00 Heavy 590 12 0.316 0.00 0.494 0.00 N’ – Number of premises with annual water consumption = 0

Initial analysis of the water intensity coefficient for all three industry types had resulted in slightly varying profile, as illustration in Figure E-1, Figure E-2, and Figure E-3 for light, medium, and heavy industry respectively. Majority of the building areas for all three industries were below 10,000 m2, which is reflected the skewness in Table E-2. However, the pattern of water intensity coefficient for light and medium industries were more similar to one another as compared to heavy industry, recording a maximum of approximately 16 m3/m2. Furthermore, for both these industries, the highest variation for the coefficient was found in the premises with building area of less than 5,000 m2, indicating a wider range of possible activities conducted within these premises. On the other hand, this variation was markedly reduced as the premise area increase, indicating a more consistent water consumption/area ratio.

Conversely, the resulting coefficient for heavy industry was more constant (Maximum coefficient of 5 m3/m2), indicating a more homogenous water consumption relative to building area throughout the entire data population as compared to the other two industrial types. However, as compared to light and medium industries, although there is a high variation of water intensity coefficient for heavy industry in premises with area of below 2,000 m2, the highest calculated coefficient was found to be in larger premises of between 4,000 – 6,000 m2. Similar to the other two industry types, variation in water intensity coefficient had also significantly reduced as premise area increased. However, light and heavy industries had shown a slightly larger variation as compared to medium industry, where the calculated coefficient for medium industry was between 0-0.5 m3/m2, while light industry had a variation of 0 – 2.5 m3/m2.

204

Water intensity, m3/m2 - Light industry 18.00

16.00

2 14.00 /m 3 12.00 10.00 8.00 6.00

Waterintensity, m 4.00 2.00 0.00 0 10,000 20,000 30,000 40,000 50,000 60,000 Area (m2)

Figure E-0-1: Water intensity/area scatterplot - light industry

Water intensity, m3/m2 - Medium industry 16.00

14.00

2 12.00

/m 3 10.00

8.00

6.00

4.00 Waterintensity, m 2.00

0.00 0 10,000 20,000 30,000 40,000 50,000 60,000 70,000 80,000 90,000 Area (m2)

Figure E-0-2: Water intensity/area scatterplot - medium industry

205

Water intensity, m3/m2 - Heavy industry 5.00

4.50

2 4.00 /m

3 3.50 3.00 2.50 2.00 1.50

Waterintensity, m 1.00 0.50 0.00 0 2,000 4,000 6,000 8,000 10,000 12,000 14,000 16,000 18,000 20,000 Area (m2)

Figure E-0-3: Water intensity/area scatterplot - heavy industry

E.2 Coefficient accuracy assessment From the above findings, there is a possibility that the accuracy of using the coefficient to estimate water consumption could be increase if the entire population be divided into several ranges of building area, rather than using an overall average. In order to test this assumption, data grouped into set A were analysed using SPSS to generate the ranges of building area using hierarchical clustering (HC) method. Ranges of the building area resulted from the HC method, as well as the corresponding coefficient, are presented in Table E-4 below.

Table E-4: Industrial class range

Overall Hierarchical clustering (HC) Type Coefficient N Min area, Max area, Coefficient Std. N m2 m2 dev Light 0.15 4,548 0 2,200 0.15 0.54 4,427 2,200 6,000 0.20 0.44 90 6,000 9,000 0.16 0.34 14 9,000 12,800 0.67 1.29 8 12,800 18,000 0.32 0.43 4 18,000 > 18,000 0.50 0.76 5 Medium 0.14 3,542 0 1,600 0.15 0.45 2,781 1,600 7,300 0.11 0.30 659

206

7,300 14,000 0.15 0.30 73 14,000 20,200 0.17 0.32 18 20,200 > 20,200 0.09 0.07 11 Heavy 0.16 288 0 6,100 0.17 0.41 265 6,100 > 6,100 0.12 0.15 23

From the above table, two underlying patterns could be observed for the initial range for each industry type; it had the largest sample size, and the coefficient for the first range was similar to the coefficient calculated using the entire sample. The cause for the second pattern was interlinked with the first. As the method for hierarchical clustering groups values in sequential steps, the range of area for each cluster classes were formed based on similarities within classes while maximising dissimilarities between classes; however, this method does not allow for reclassification of dataset into different clusters. As most of the data samples had similar size for the smaller premises, this resulted in a larger initial range for all three industrial types. Conversely, this had also resulted in a high water intensity variability in the initial range; this characteristic was also reflected in the calculation of overall average coefficient. Thus, a similar coefficient value for the initial range and the calculated using the overall population was observed. However, the subsequent ranges for each industry type differs from the initial and overall coefficient, suggesting that there some dissimilarities between the calculated coefficient of larger premises and the premise size in the initial range in terms of both the premise size and corresponding coefficient. By segregating the larger premises into different ranges, the variation between the coefficients could be reduced, as represented by the decreasing value of standard deviation.

Accuracy of the predetermined area ranges were then tested using data in set B, and results of the accuracy assessment are presented in Table E-5, Table E-6, and Table E-7 for light, medium, and heavy industry respectively. Overall, neither methods could not produce a very high accuracy to predict industrial water consumption at premise level; regardless of the method used, on average, the water consumption per premise tend to be underestimated for light and medium industries, while overestimated for heavy industry. This, again, was principally due to the large differences in water consumption within each premise, resulting in high variance in the calculated water intensity for each industry type. In this sense, the use of HC method could be useful as it has the ability to group premises which has similar water intensity with one another, thus minimizing the variation within each class. For example,

207

using the HC method, the accuracy for water consumption estimation for large premises in the medium industry (area >20,200 m2) was significantly higher than any other ranges in all other types of industries (RNS = 0.93,

Table E-6), due to the extremely small variation in water intensity within this range (σ = ± 0.07). This, in turn, resulted in an average error of 307.57 m3/month, as compared to approximately 2,000 m3/month using the overall method.

Although the use of HC method had allowed for further reduction of variance within each range, it had resulted in a mixed result when applied to the dataset for this research specifically. For example, with the exception of premises within the range of 12,800 – 18,000 m2, the HC method outperformed the overall method for the light industry, where the error resulting from the HC method was consistently smaller than the overall method, resulting in the a lower RMSE value and higher RNS value. On the other hand, for both medium and heavy industries, there were only marginal differences for the overall RMSE and RNS values.

For heavy industry specifically, this could be due to the smaller range of calculated water intensities (maximum water intensity of approximately 4.5 m3/m2) as compared to light and medium industries. Furthermore, the calculated coefficient for heavy industry were more uniformed, where the majority of the samples had resulted in a coefficient of ≤ 1 m3/m2, with the exceptions of a few premises which had water intensities of above 1 m3/m2. This had translated to similar results RNS values for both the overall and the HC method, indicating that the use of HC method was not able to further reduce the variation within each area range compared to the use of the overall average coefficient. In this sense, the introduction of industrial sub-classification in addition to the HC method could aid in further reducing the standard deviation of coefficient within each area range. This pattern was also observed for medium industry, specifically for all other area ranges except for premises of area more than 20,200 m2; the relatively similar variation in water intensity in all other area ranges had resulted in similar accuracy between the two methods, indicating a need for additional information to reduce the dissimilarities. On the other hand, the introduction of additional information could possibly not explain the differences within each sub-classification, as emphasized by Kieger, Krentz and Dziegielewski (2015). This is because industrial water use for each firm/premise have been known to depend on other factors beyond production yield, such as technology use as well as water recirculation rate (Duport and Renzetti, 2001), leading to the possibility of large variation in water demand for similar industrial activities.

208

Based on the findings above, the use of coefficient using the HC method was used to estimate the water consumption for each premises. For medium industry specifically, the use of HC was chosen over the overall method primarily due to the higher accuracy for larger premises, while resulting in similar accuracy for the smaller premises. As for light industry, the use of HC method was preferred as it generated an overall higher accuracy as compared to the overall coefficient. Finally, for heavy industry, due to the little differences between the HC method and overall method, the use of coefficient generated from the HC method was also use to standardise the calculation method.

Subsequently, the accuracy of the method was further tested to see the aggregated error resulted from a cluster of premises, as the research requires the water consumption to be further aggregated to an IP level. The error had converged after a total of 20,000 randomly simulated configurations of IPs were done; parallel to the findings above, the simulations had resulted in a slightly lower average error using the HC method unlike the overall method, with an average overestimation of 19±41% and 22±42% respectively. Due to the lower average error of the HC method at an IP level, it further reinforced the applicability of this method to estimate the collective volume of industrial water consumption for an IP.

209

Table E-5: Accuracy comparison between hierarchical clustering and overall overage coefficient for light industry 3 3 3 Hierarchical clustering, HC Minimum error, m Maximum error, m Average error, m Std. dev. RMSE RNS Min area, Max N HC Overall HC Overall HC Overall HC Overall HC Overall HC Overall m2 area, m2 0 2,200 2,937 -292.89 -298.34 5,150.25 5,148.25 -1.43 -2.05 145.67 145.62 145.68 145.64 0.23 0.23 2,200 6,000 63 -987.31 -614.57 11,505.46 11,888.56 121.79 352.37 1,740.88 1,758.56 1,745.13 1,793.52 0.21 0.16 6,000 9,000 11 -2,724.36 -1,227.57 22,050.68 23,676.53 848.36 2,267.85 7,174.31 7,226.09 7,224.29 7,573.60 0.18 0.10 9,000 12,800 3 -3,449.18 -1,491.60 4,538.83 6,825.52 -757.00 1,304.02 3,744.90 3,904.39 3,820.64 4,116.40 0.39 0.29 12,800 18,000 2 1,081.56 -764.25 7,834.24 6,160.94 4,457.90 2,698.35 3,376.34 3,462.60 5,592.19 4,389.83 0.06 0.42 18,000 > 18,000 3 755.07 2,589.31 38,395.23 40,752.82 22,064.22 25,293.32 15,764.55 16,400.72 27,117.35 30,145.25 0.39 0.25 Average (Total) (3,019) -3,449.18 -1,491.60 38,395.23 40,752.82 28.37 41.84 1,019.56 1,107.98 1,019.95 1,108.77 0.35 0.23

Table E-6: Accuracy comparison between hierarchical clustering and overall overage coefficient for medium industry 3 3 3 Hierarchical clustering, HC Minimum error, m Maximum error, m Average error, m Std. dev. RMSE RNS Min area, Max N HC Overall HC Overall HC Overall HC Overall HC Overall HC Overall m2 area, m2 0 1,600 924 -226.93 -217.42 2,767.23 2,768.37 -8.16 -5.39 157.13 156.94 157.34 157.03 0.48 0.48 1,600 7,300 217 -937.32 -896.47 34,696.74 34,728.74 50.71 70.71 2,416.23 2,417.06 2,416.76 2,418.09 0.04 0.04 7,300 14,000 25 -1,288.64 -1,638.68 18,847.03 18,627.26 927.08 669.79 4,641.03 4,642.71 4,732.72 4,690.77 0.10 0.11 14,000 20,200 5 -2,679.92 -2,513.38 134.05 320.92 -1,589.50 -1,424.80 973.62 984.63 1,863.98 1,731.93 -0.66 -0.44 20,200 > 20,200 5 -1,961.94 -3,259.42 1,435.74 -973.80 -307.57 -2,040.75 1,096.15 923.43 1,138.48 2,239.95 0.93 0.59 Average (Total) (1,176) -2,679.92 -4,573.20 34,696.74 34,728.74 14.59 6.17 1,262.31 1,270.29 1,262.40 1,270.31 0.10 0.09

210

Table E-7: Accuracy comparison between hierarchical clustering and overall overage coefficient for heavy industry 3 3 3 Hierarchical clustering Minimum error, m Maximum error, m Average error, m Std. dev. RMSE RNS Min area, Max N HC Overall HC Overall HC Overall HC Overall HC Overall HC Overall m2 area, m2 0 6,100 177 -818.45 -799.51 3,385.62 3,403.66 -28.01 -22.17 511.21 511.19 511.98 511.67 0.17 0.17 6,100 > 6,100 17 -2,082.10 -2,878.97 11,153.57 10,356.70 122.22 -325.54 2,856.08 2,786.65 2,858.69 2,805.60 0.23 0.26 Average (Total) (194) -2,082.10 -2,878.97 11,153.57 10,356.70 -16.01 -50.47 984.85 969.83 977.38 963.65 0.22 0.24

211

Appendix F: Results for potable and non-potable reuse Table F-1: Estimated baseline and 15th year projected water saving from non-potable reuse by individual IP

Baseline, Adjusted (Non-potable) (m3/month) Projected Y15, Adjusted (Non-potable) (m3/month) WSS Service area IP Effluent, Effluent, Effluent, Effluent, Effluent, Effluent, Demand Demand 30% 60% 90% 30% 60% 90% CH_IP-1 954 272 544 816 1,284 366 732 1,098 CH_IP-2 5,614 1,600 3,200 4,800 7,555 2,153 4,306 6,460 CH_IP-3 45,509 12,970 25,940 38,910 61,249 17,456 34,912 52,368 Cheras CH_IP-4 73,082 20,828 41,657 62,485 98,359 28,032 56,064 84,097 CH_IP-5 15,577 4,439 8,879 13,318 20,964 5,975 11,950 17,924 Langat CH_IP-6 164,547 46,896 93,792 140,687 221,458 63,116 126,231 189,347 CH_IP-7 1,755 500 1,000 1,500 2,362 673 1,346 2,019 Kuala KL_IP-1 4,574 1,304 2,607 3,911 6,156 1,754 3,509 5,263 Lumpur Total, Langat (m3/month) 308,903 88,037 176,075 264,111 415,741 118,486 236,972 355,459 HL_IP-1 40,880 11,651 23,301 34,952 55,019 15,680 31,361 47,041 HL_IP-2 25,523 7,274 14,548 21,822 34,350 9,790 19,580 29,369 Hulu Langat HL_IP-3 48,499 13,822 27,644 41,466 65,273 18,603 37,206 55,808 HL_IP-4 25,074 7,146 14,292 21,438 33,746 9,618 19,235 28,853 HL_IP-5 66,516 18,957 37,914 56,871 89,522 25,514 51,027 76,541 PJ_IP-1 20,806 5,930 11,859 17,789 28,001 7,980 15,961 23,941 PJ_IP-2 23,471 6,689 13,379 20,068 31,589 9,003 18,006 27,009 Semenyih PJ_IP-3 1,066 304 608 912 1,435 409 818 1,227 PJ_IP-4 3,213 916 1,831 2,747 4,324 1,232 2,465 3,697 Petaling PJ_IP-5 7,478 2,131 4,263 6,394 10,065 2,868 5,737 8,605 PJ_IP-6 15,998 4,560 9,119 13,679 21,532 6,137 12,273 18,410 PJ_IP-7 35,144 10,016 20,032 30,048 47,299 13,480 26,960 40,440 PJ_IP-8 73,737 21,015 42,030 63,045 99,240 28,283 56,567 84,850 Sepang SE_IP-1 10,861 3,095 6,191 9,286 14,618 4,166 8,332 12,498

212

SE_IP-2 4,595 1,309 2,619 3,928 6,184 1,762 3,525 5,287 SE_IP-3 2,879 821 1,641 2,462 3,875 1,104 2,209 3,313 SE_IP-4 17,721 5,050 10,101 15,151 23,850 6,797 13,594 20,392 SE_IP-5 13,978 3,984 7,967 11,951 18,812 5,361 10,723 16,084 Total, Semenyih 430,281 122,629 245,259 367,888 579,100 165,042 330,087 495,128 (m3/month)

Table F-2: Estimated baseline saving from potable reuse from individual IP

Savings - Baseline, Adjusted (Potable) (m3/month)

Effluent, 30% Effluent, 60% Effluent, 90%

Reservoir Service area IP

Demand

lending

B 2/98 , Blending 15/85 , Blending 35/65 , Blending 2/98 , Blending 15/85 , Blending 35/65 , Blending 2/98 , Blending 15/85 , Blending 35/65 , CH_IP-1 954 8 41 95 16 82 190 24 122 285 CH_IP-2 5,614 48 240 560 96 480 1,120 144 720 1680 CH_IP-3 45,509 389 1,946 4,540 778 3,891 9,079 1,167 5,837 13,619 Cheras CH_IP-4 73,082 625 3,124 7,290 1,250 6,249 14,580 1,875 9,373 21,870 CH_IP-5 15,577 133 666 1,554 266 1,332 3,108 400 1998 4661 Langat CH_IP-6 164,547 1,407 7,034 16,414 2,814 14,069 32,827 4,221 21,103 49,241 CH_IP-7 1,755 15 75 175 30 150 350 45 225 525 Kuala KL_IP-1 4,574 39 196 456 78 391 913 117 587 1369 Lumpur Total, Langat (m3/month) 311,611 2,641 13,206 30,813 5,282 26,411 61,626 7,923 39,617 92,439 HL_IP-1 40,880 350 1,748 4,078 699 3,495 8,156 1,049 5,243 12,233 HL_IP-2 25,523 218 1,091 2,546 436 2,182 5,092 655 3,273 7,638 Semenyih Hulu Langat HL_IP-3 48,499 415 2,073 4,838 829 4,147 9,675 1,244 6,220 14,513 HL_IP-4 25,074 214 1,072 2,501 429 2,144 5,002 643 3,216 7,503

213

HL_IP-5 66,516 569 2,844 6,635 1,137 5,687 13,270 1,706 8,531 19,905 PJ_IP-1 20,806 178 889 2,075 356 1,779 4,151 534 2,668 6,226 PJ_IP-2 23,471 201 1,003 2,341 401 2,007 4,683 602 3,010 7,024 PJ_IP-3 1,066 9 46 106 18 91 213 27 137 319 PJ_IP-4 3,213 27 137 320 55 275 641 82 412 961 Petaling PJ_IP-5 7,478 64 320 746 128 639 1,492 192 959 2,238 PJ_IP-6 15,998 137 684 1,596 274 1,368 3,192 410 2,052 4,788 PJ_IP-7 35,144 300 1,502 3,506 601 3,005 7,011 901 4,507 10,517 PJ_IP-8 73,737 630 3,152 7,355 1,261 6,304 14,710 1,891 9,457 22,066 SE_IP-1 10,861 93 464 1,083 186 929 2,167 279 1,393 3,250 SE_IP-2 4,595 39 196 458 79 393 917 118 589 1,375 Sepang SE_IP-3 2,879 25 123 287 49 246 574 74 369 862 SE_IP-4 17,721 152 758 1,768 303 1,515 3,535 455 2,273 5,303 SE_IP-5 13,978 120 598 1,394 239 1,195 2,789 359 1,793 4,183 Total, Semenyih 437,437 3,679 18,394 42,920 7,358 36,789 85,841 11,037 55,183 128,761 (m3/month)

Table F-3: Estimated 15th year projected saving from potable water reuse from individual IP

Savings - Y15, Adjusted (Potable) (m3/month)

Effluent, 30% Effluent, 60% Effluent, 90%

Reservoir Service area IP

Demand

2/98 2/98 2/98

15/85 35/65 15/85 35/65 15/85 35/65

Blending, Blending, Blending, Blending, Blending, Blending, Blending, Blending, Blending, CH_IP-1 1,284 11 55 128 22 110 256 33 165 384 CH_IP-2 7,555 65 323 754 129 646 1,507 194 969 2,261 Langat Cheras CH_IP-3 61,249 524 2,618 6,110 1,047 5,237 12,219 1,571 7,855 18,329 CH_IP-4 98,359 841 4,205 9,811 1,682 8,410 19,623 2,523 12,614 29,434

214

CH_IP-5 20,964 179 896 2,091 358 1,792 4,182 538 2,689 6,274 CH_IP-6 221,458 1,893 9,467 22,090 3,787 18,935 44,181 5,680 28,402 66,271 CH_IP-7 2,362 20 101 236 40 202 471 61 303 707 Kuala Lumpur KL_IP-1 6,156 53 263 614 105 526 1,228 158 790 1,842 Total, Langat (m3/month) 419,387 3,555 17,773 41,470 7,109 35,546 82,940 10,664 53,319 124,411 HL_IP-1 55,019 470 2,352 5,488 941 4,704 10,976 1,411 7,056 16,464 HL_IP-2 34,350 294 1,468 3,426 587 2,937 6,853 881 4,405 10,279 Hulu Langat HL_IP-3 65,273 558 2,790 6,511 1,116 5,581 13,022 1,674 8,371 19,533 HL_IP-4 33,746 289 1,443 3,366 577 2,885 6,732 866 4,328 10,098 HL_IP-5 89,522 765 3,827 8,930 1,531 7,654 17,860 2,296 11,481 26,789 PJ_IP-1 28,001 239 1,197 2,793 479 2,394 5,586 718 3,591 8,379 PJ_IP-2 31,589 270 1,350 3,151 540 2,701 6,302 810 4,051 9,453 PJ_IP-3 1,435 12 61 143 25 123 286 37 184 429 PJ_IP-4 4,324 37 185 431 74 370 863 111 555 1,294 Petaling Semenyih PJ_IP-5 10,065 86 430 1,004 172 861 2,008 258 1,291 3,012 PJ_IP-6 21,532 184 920 2,148 368 1,841 4,296 552 2,761 6,443 PJ_IP-7 47,299 404 2,022 4,718 809 4,044 9,436 1,213 6,066 14,154 PJ_IP-8 99,240 848 4,242 9,899 1,697 8,485 19,798 2,545 12,727 29,697 SE_IP-1 14,618 125 625 1,458 250 1,250 2,916 375 1,875 4,374 SE_IP-2 6,184 53 264 617 106 529 1,234 159 793 1,851 Sepang SE_IP-3 3,875 33 166 387 66 331 773 99 497 1,160 SE_IP-4 23,850 204 1,020 2,379 408 2,039 4,758 612 3,059 7,137 SE_IP-5 18,812 161 804 1,877 322 1,608 3,753 483 2,413 5,630 Total, Semenyih (m3/month) 588,733 4,951 24,756 57,765 9,903 49,513 115,530 14,854 74,269 173,295

215

Table F-4: Estimated baseline saving at WSS from non-potable water reuse

Savings - Baseline, Adjusted (Non-potable) Dam (%) WTP (%)

total total non

- -

/month) /month) /month) /month) /month)

3 3 3 3 3

emand, IP IP emand,

Dam Dam demand (m Dam demand domestic (m treatment Water total plant, (m treatment Water non plant, domestic (m D (m 30% Effluent, 60% Effluent, 90% Effluent, 30% Effluent, 60% Effluent, 90% Effluent, Langat 4,813,924 2,002,592 12,026,571 5,003,054 311,611 1.84 3.69 5.53 0.74 1.48 2.22 Semenyih 3,061,132 1,273,431 19,676,143 8,185,275 437,437 3.69 7.38 11.08 0.57 1.15 1.72

Table F-5: Estimated 15th year projected saving at WSS from non-potable water reuse

Savings - Projected Y15, Adjusted (Non-potable) Dam (%) WTP (%)

total total n

- -

/month) /month) /month) /month) /month)

3 3 3 3 3

Scenario Dam Dam demand (m Dam demand domestic (m treatment Water total plant, (m treatment Water non plant, domestic (m IP Demand, (m 30% Effluent, 60% Effluent, 90% Effluent, 30% Effluent, 60% Effluent, 90% Effluent, Langat 6,258,101 2,603,370 12,026,571 5,003,054 419,387 1.91 3.82 5.73 0.99 1.99 2.98 S1 Semenyih 3,979,472 1,655,460 19,676,143 8,185,275 588,733 3.82 7.64 11.47 0.77 1.55 2.32 Langat 6,258,101 2,603,370 15,634,543 6,503,970 419,387 1.91 3.82 5.73 0.76 1.53 2.29 S2 Semenyih 3,979,472 1,655,460 25,578,986 10,640,858 588,733 3.82 7.64 11.47 0.59 1.19 1.78

216

Table F-6: Estimated baseline saving at WSS from potable water reuse (m3/month)

Savings - Baseline, Adjusted (Potable)

Effluent, 30% (m3/month) Effluent, 60% (m3/month) Effluent, 90% (m3/month)

/month)

domestic domestic

15/85

total demand demand total non

- -

/mont /month) /month) /month)

3 3 3 3

Dam Dam (m Dam (m demand treatment Water total plant, (m treatment Water non plant, (m IP Demand, (m

Blending,2/98 Blending, Blending,35/65 Blending,2/98 Blending,15/85 Blending,35/65 Blending,2/98 Blending,15/85 Blending,35/65 Langat 4,805, 1,999, 12,026, 5,003,0 311,611 2,664 13,321 31,083 5,329 26,64 62,166 7,993 39,964 93,249 530 100 571 54 3 Semenyih 3,061, 1,273, 19,676, 8,185,2 437,437 6,781 33,906 79,113 13,56 67,81 158,226 20,34 101,717 237,339 132 431 143 75 2 1 3

Table F-7: Estimated baseline saving at WSS from potable water reuse (m3/month)

Savings - Baseline, Adjusted (Potable) Dam (%) WTP (%)

Effluent, 30% Effluent, 60% Effluent, 90% Effluent, 30% Effluent, 60% Effluent, 90%

Dam

Blending, 2/98 Blending, 15/85 Blending, 35/65 Blending, 2/98 Blending, 15/85 Blending, 35/65 Blending, 2/98 Blending, 15/85 Blending, 35/65 Blending, 2/98 Blending, 15/85 Blending, 35/65 Blending, 2/98 Blending, 15/85 Blending, 35/65 Blending, 2/98 Blending, 15/85 Blending, 35/65 Langat 0.06 0.28 0.65 0.11 0.55 1.29 0.17 0.83 1.94 0.02 0.11 0.26 0.04 0.22 0.52 0.07 0.33 0.78 Semenyih 0.22 1.11 2.58 0.44 2.22 5.17 0.66 3.32 7.75 0.03 0.17 0.40 0.07 0.34 0.80 0.10 0.52 1.21

217

Table F-8: Estimated 15th year projected saving at WSS from potable water reuse (m3/month)

Savings – Projected Y15, Adjusted (Potable)

Effluent, 30% (m3/month) Effluent, 60% (m3/month) Effluent, 90% (m3/month)

/month)

domestic domestic

total demand demand total non

- -

/month) /month) /month) /month)

3 3 3 3

Scenario Dam Dam (m Dam (m demand treatment Water total plant, (m treatment Water non plant, (m IP Demand, (m Blending,35/65

Blending,2/98 Blending,15/85 Blending,35/65 Blending,2/98 Blending,15/85 Blending,35/65 Blending,2/98 Blending,15/85 6,247, 2,598, 12,026, 5,003,0 35,85 10,75 Langat 419,387 3,586 17,929 41,834 7,172 83,668 53,786 125,502 189 831 571 54 8 7 S1 3,979, 1,655, 19,676, 8,185,2 18,25 91,26 27,38 Semenyih 588,733 9,127 45,633 106,476 212,952 136,898 319,428 472 460 143 75 3 5 0 6,247, 2,598, 15,634, 6,503,9 35,85 10,75 Langat 419,387 3,586 17,929 41,834 7,172 83,668 53,786 125,502 189 831 543 70 8 7 S2 3,979, 1,655, 25,578, 10,640, 18,25 91,26 27,38 Semenyih 588,733 9,127 45,633 106,476 212,952 136,898 319,428 472 460 986 858 3 5 0

218

Table F-9: Estimated 15th year projected saving at WSS from non-potable water reuse (%)

Savings – Projected Y15, Adjusted (Potable) Dam (%) WTP (%)

Effluent, 30% Effluent, 60% Effluent, 90% Effluent, 30% Effluent, 60% Effluent, 90%

ending, ending,

Scenario Dam

Blending, 2/98 Blending, 15/85 Blending, 35/65 Blending, 2/98 Blending, 15/85 Blending, 35/65 Bl 2/98 Blending, 15/85 Blending, 35/65 Blending, 2/98 Blending, 15/85 Blending, 35/65 Blending, 2/98 Blending, 15/85 Blending, 35/65 Blending, 2/98 Blending, 15/85 Blending, 35/65 Langat 0.06 0.29 0.67 0.11 0.57 1.34 0.17 0.86 2.01 0.03 0.15 0.35 0.06 0.30 0.70 0.09 0.45 1.04 S1 Semenyih 0.23 1.15 2.68 0.46 2.29 5.35 0.69 3.44 8.03 0.05 0.23 0.54 0.09 0.46 1.08 0.14 0.70 1.62 Langat 0.06 0.29 0.67 0.11 0.57 1.34 0.17 0.86 2.01 0.02 0.11 0.27 0.05 0.23 0.54 0.07 0.34 0.80 S2 Semenyih 0.23 1.15 2.68 0.46 2.29 5.35 0.69 3.44 8.03 0.04 0.18 0.42 0.07 0.36 0.83 0.11 0.54 1.25

219

Appendix G: Profile of interviewed companies and summary of initial findings Table G-1: Profile of interviewed companies and summary of initial findings

Code Sub- Water use Water Final form Problems Strategies Challenges to Drivers to sector source of water encountered identified to increase increase use with water rectify the efficiency and efficiency and supply problem mitigate water mitigate water related risk related risk RI1 F&B Raw material Potable Filtered Water Transferred potable Not much they can Reputation - (consumptive water water rationing and water from one do in terms of they cannot , 30%), high Total factory to another process as potable have any stall in washing, Organic using water water is 30% of the production irrigation Carbon tankers. After raw material for as they (TOC) water rationing product currently hold occurrence, they production. They largest market invested attempted to use share in country approximately RM treated wastewater 1.5 million for for landscape installation of irrigation but was additional discouraged by permanent water local municipal tanks in all factory authority premises to

220

increase buffer time.

RI2 Paint Raw material Potable Deionised Water Install additional Identified Reputation - (60%), water water rationing water storage tanks possibilities of they need to washing using existing using treated fulfil customer storage containers wastewater demand. effluent as Decreasing substitute for potable water potable water. consumption to However, need to reduce cost of extensively study production microbial profile of wastewater generated before it can be considered safe to use RI3 E&E Cooling, Potable Deionised No issue No issue Limited funding Reputation - process (non- water water from company to they would be consumptive) implement any able to show a , maintenance additional proactive

221

technologies. Low attitude and will water tariff do not provide them a provide any competitive additional edge incentives. Getahindus Rubber Process (non- Potable Deionised Low water No change as it did No reason to Demand from (M) Sdn. product consumptive) water water and quality (high not affect final implement water customers - Bhd., RI5 and untreated salinity and product efficient most of their ground groundwate discoloration) technologies as customers is water r water tariff is low from international market. They will need to be able to comply to international requirement if and when it is requested by their customers ROHM- E&E Cooling, Potable Pure water No water for Engage MIDA and Internal Reputation - Waco product water expansion, local water challenges; higher need to fulfil

222

Electronics rinsing (non- water operators to discuss management might customer (M) Sdn. consumptive) rationing, and about possibility of not willing to demand (need Bhd., RI6 , maintenance low water building new water implement any to ensure quality treatment plant technologies continuous production)

223

Appendix H: Database for Chapter 4 Table H-1: Removal efficiency for unit processes

Reuse standard Potable Non-potable Treatment Train P-1 P-2 (+ NP-1) NP-1 NP-2

Unit process

ion

C/F/DAF RSF O B/GAC UF Chlorinat C/F/S SF Chlorination /O PST MBR RO UV SF MF UF RO BOD5 at % 53.5 70 60 83.8 85 83.8 70 60 43 97 83.52 70 85 83.52 20°C Source AS AB BA AR AR AR AB BA AF C L AB AR L COD % 47 38 60 82.5 77 57.5 38 60 93.5 97.03 3 38 45 77 97.03 Source AI AQ BA AR AR AM AQ BA C L BM AQ BM AR L Suspended % 77.2 77 59.4 62 98 77 65 99 99 77 90 62 99 solids Source AI Y AR AR BJ Y AF C BQ Y BO AR BQ Mercury, % 99 99.5 94.2 99.2 79.3 96.7 94.2 Hg Source AW S AN A AN AN AN Cadmium, % 29 96.6 84 99.73 96.6 99 98.5 96.6 98.5 Cd Source AZ X U R X A G X G Arsenic, As % 90 78 90 43 99 99 Source Q BH Q A J J Cyanide, Cn % 94.5 94.5 99.6 4 67 99.6 Source BM BM AO BM BM AO Lead, Pb % 29 99.8 99.98 99.9 71.5 99.8 98 99.3 99.8 98 99.9 Source AZ X S N V X D AY X BG N Copper, Cu % 98.6 99.6 96.7 90 99.8 46 99.6 96.7 80 97 99.6 90 99.8 97 Source AC X BM T N BF X BM D G X AX N G Manganese, % 99.2 79.5 83 79.05 99.95 79.5 83 84 95.3 79.5 95 99.2 95.3

224

Mn Source Z BI AA H R BI AA A AY BI Z BE AY Nickel, Ni % 98.6 89.3 90 99.6 63 89.3 65 99.3 95 99.6 99.3 Source AC BM I N BJ BM A B BG N B Zinc, Zn % 98.6 45 83.6 99.7 83 45 77 98.9 45 95 99.7 98.9 Source AC AB I N BF AB A B AB BG N B Iron, Fe % 98 94.3 96 99.02 99.9 98 45 96 94 97.62 45 99.9 97.62 Source BJ BI AA H BE BJ AB AA A L AB BE L Aluminium, % 67.6 74.7 99.2 67.6 74.7 98 78 74.7 78 Al Source BF BK H BF BK A BK BK BK Selenium, Se % 28 5 99.8 99.8 Source AG/ BB AG AG BC/ NE Fluoride, F % 46.5 47.9 3.39 94 97.2 3.39 97.2 Source BL BL W AH BL W BL Phenol % 96 95 96 96 95 99.8 87 17 87 Source T AD T T AD AK O AL O Colour % 60 6 94 99 70 80 60 6 94 85 95 92 70 95 Source AP AQ BA T AT AV AP AQ BA AU AQ BM AT AQ

225

Table H-2: Literature sources for removal efficiency

Code Source Effluent source Contaminant Unit process

A Bolzonella et al., 2010 Petrochemical Hg, Cd, As, Mn, Ni, Zn,Fe, Al, MBR Ba, As, B, Sn B Ipek, 2005 Synthetic wastewater Phenol RO C Melin et al., 2006 Industrial wastewater – unspecified TSS, COD, BOD MBR D Katsou, Malamis and Municipal wastewater Pb, Cu MBR Loizidou, 2011 F Dominguez-Tagle, J. Romero- Seawater B RO Ternero and Delgado-Torres, 2011 G Qdais and Moussa, 2004 Synthetic wastewater Cu RO, NF H Goher et al., 2015 Unspecified industrial wastewater Al, Fe, Mn GAC I Karnib et al., 2014 Synthetic wastewater Ni, Cd, Zn, Pb, Cr AC J Uddin et al., 2007 Industrial wastewater – unspecified As RO K Ndiaye et al., 2005; Colla et Hot tolling mill wastewater F, Sulphide RO al., 2016 L Talalaj, 2015 Leachate BOD, COD, Fe RO

M Pal and Kumar, 2012 Coke wastewater BOD, Colour O3 N Azimi et al., 2017 Industrial wastewater – unspecified Pb, Cu, Ni, Zn UF O Ipek, 2004 Synthetic wastewater Phenol RO P Pintor et al., 2016 Industrial wastewater – unspecified O&G Coagulation/flocculation, DAF, O3, UF, MF, RO Q Nicomel et al., 2015 River water As Coagulation/flocculation R Kurniawan et al., 2006 Industrial wastewater – unspecified Mn, Cd, Cu, Zn Coagulation/flocculation S Jusoh et al., 2007 Synthetic wastewater Pb GAC

226

T Cyr, Suri and Helmig, 2002 Pharmaceutical wastewater Phenol, Cu, Colour GAC

U Wasewar et al., 2010 Synthetic wastewater Cd GAC V Covaliu et al., 2016 Synthetic wastewater Pb, Zn PAC

W Petrinic et al., 2015 Metal finishing wastewater F UF X Muhammad, 1998 Synthetic wastewater Cu, Cr, Pb, Cd Sand filtration Y Langenbach et al., 2009 Synthetic wastewater SS Sand filtration Z Patil, Chavan and Synthetic wastewater Mg DAF, MF Oubagaranadin, 2016 AA Araby, Hawash and Diwani, Synthetic groundwater Fe, Mn O3 2009 AB Cheremisinoff, 2002 Industrial wastewater – unspecified BOD, Zn, Fe Sand filtration AC Rubio and Tessele, 1997 Synthetic wastewater Cu, Ni, Zn DAF

AD Kulkarni, 2017 Industrial wastewater – unspecified Phenol O3 AE Purkayastha, Mishra and Synthetic wastewater Cd Biswas, 2014 AF Tchobanoglous, Butron and Industrial wastewater – unspecified BOD, TSS Primary sedimentation tank Stensel, 2004 AG Sandy and DiSante, 2010 Industrial wastewater – unspecified Selenium GAC, UF, MF, RO, coagulation/flocculation, SF AH Aoudj et al., 2012 Synthetic wastewater F Coagulation/flocculation AI Koivunen and Heinonen‐ Municipal wastewater COD, SS DAF Tanski, 2008 AJ Sailaja Kumari et al., 2015 Synthetic wastewater F GAC AK Barrios-Martinez et al., 2006 Synthetic wastewater Phenol MBR AL Kargari and Mohammadi, Synthetic wastewater Phenol UV, RO 2014

227

AM Teresa, Gunther and Fernando, Coffee wastewater COD Coagulation/flocculation 2007; Lee, Robinson and Chong, 2014 AN Urgun-Demirtas et al., 2012 Oil refinery Hg MF, UF, RO

AO Jin et al., 2013 Coking wastewater Cyanide RO AP Selcuk, 2005 Synthetic wastewater Colour Coagulation/flocculation AQ Ciardelli, Corsi and Marcucci, Textile wastewater COD, colour UF, RO 2000 AR Bhattacharjee, Datta and Supernatant from clarified kraft TS, BOD, COD AC & UF Bhattacharjee, 2007 black liquor AS Abou-Elela et al., 2008 Board paper mill BOD, COD, TSS DAF

AT Maria, Alves and Norberta De Tannery wastewater Colour UF Pinho, 2000 AU Brik et al., 2006 Textile wastewater Colour MBR AV Brik et al., 2004 Textile wastewater Colour Chlorination AW Tassel et al., 1997 Synthetic wastewater Mercury DAF AX Zouboulis, Sarasidis and Synthetic wastewater Copper MF Moussas, 2010 AY Malamis et al., 2012 Municipal wastewater Mn, Pb RO AZ Al-Zoubi, Ibrahim and Abu- Synthetic wastewater Cd, Pb DAF Sbeih, 2015

BA Bernal-Martínez et al., 2010 Industrial wastewater – mixed, COD, colour O3 industrial park BB Zhang, Lin and Gang, 2008 Synthetic wastewater Selenium AC

228

BC Santos et al., 2015 Industrial wastewater – unspecified Selenium RO, AC, coagulation/flocculation BE Kasim, Mohammad and Synthetic groundwater Fe, Mn UF Sheikh Abdullah, 2017 BF Petala et al., 2006 Municipal wastewater Zn, Al, Fe, Cu Coagulation/Flocculation BG Malamis, Katsou and Synthetic wastewater Ni, Cu, Pb, Zn MF Haralambous, 2011 BH Ansari and Sadegh, 2007 Synthetic wastewater As GAC BI Sodamade and Pearse, 2013 Synthetic wastewater Mn, Fe SF BJ Abu Bakar and Abdul Halim, Automotive parts wastewater COD, TSS, Fe, Ni Coagulation/flocculation 2013 BK Kettunen and Keskitalo, 2000 Groundwater for drinking water Al RO, SF BL Chuang et al., 2006 Industrial wastewater – mixed, high- Cl, F DAF, GAC, RO tech industrial park

BM Cui et al., 2014; Mert et al., Electroplating wastewater/Jewellery CN, Cu, Ni O3, UV 2014 manufacturing effluent BN dos Santos et al., 2017 Food processing wastewater – Colour, Cyanide, COD MF cassava BO Ahmed, Tewfik and Talaat, Wastewater – unspecified SS MF 2002 BP Mert and Kestioglu, 2014 Tannery wastewater Cr UF, RO BQ Gupta et al., 2012 Wastewater – unspecified SS RO

229

Table H-3: Cost function for CAPEX and OPEX

CAPEX OPEX

Unit

process Cost function (USD) Cost function (USD/year)

Reuse standard Treatment Train Baseyear Baseyear Reference Reference 푙표푔(퐶) = 0.347 × 푙표푔(푥)1.448 C/F 푙표푔(퐶) = 0.222 × 푙표푔(푥)1.516 + 3.071 A 2014 A 2014 + 2.726 퐿푛 (퐶) = 14.5532 + 0.96495 퐿푛(푥) − 0.01219 퐿푛(푥)2 for x > 20 gpm 퐿푛 (퐶) = 13.9518 + 0.29445 퐿푛(푥) DAF E 1989 퐿푛 (퐶) = E 1989 − 0.12049 퐿푛(푥)2 21.2446 + 4.14823 퐿푛(푥) − 0.36585 퐿푛(푥)2 for x ≤ 20 gpm (14.472+ 0.5752⁄2) 0.716

RSF 퐶 = 푒 × 퐷 D 2003 0.25 X Capital Cost F 1995 1 - × 1.096833 P 0.8916 O3 Look-up table, O3 (Table H-5) D 2003 퐶 = 319 × 푄 C 2010 (12.634+ 0.9572⁄2) 0.832 0.559 푙표푔(퐶) = 1.669 × 푙표푔(푥) B/GAC 퐶 = 푒 × 퐷 D 2003 A 2014 × 1.096833 + 2.371 0.598 0.830 푙표푔(퐶) = 1.828 × 푙표푔(푥) Potable UF 푙표푔(퐶) = 1.003 × 푙표푔(푥) + 3.832 A 2014 A 2014 + 1.876 퐶 = (−0.000001 × 푄2) Chlorinatio 2 (10.4+ 1.07 ⁄2) 0.684 D 2003 + (2.36 × 푄2) B 2003 n 퐶 = 푒 × 퐷 × 1.096833 + 24813 (12.754 + 0.7502⁄2) 0.608 C/F/S 퐶 = 푒 × 퐷 D 2003 퐶 = 1.4 × 103 × (20 × 푄)0.5146 C 2010 × 1.096833 (14.472 + 0.5752⁄2) 0.716 퐶 = 푒 × 퐷

2 SF D 2003 C = 0.25 X Capital Cost F 1995 -

P × 1.096833 퐶 = (−0.000001 × 푄2) Chlorinatio 2 (10.4+ 1.07 ⁄2) 0.684 D 2003 + (2.36 × 푄2) B 2003 n 퐶 = 푒 × 퐷 × 1.096833 + 24813

230

0.8916 O3 Look-up table, O3 (Table H-5) D 2003 퐶 = 319 × 푄 C 2010 PST 퐶 = −0.000002푄2 + 19.29푄 + 220,389 H 1996 퐶 = 1.69푄 + 11,376 H 1996 푙표푔(퐶) = 0.639 × 푙표푔(푥)1.143 MBR 푙표푔(퐶) = 0.569 × 푙표푔(푥)1.135 + 4.605 A 2014 A 2014

+ 2.633 1

- 2 퐶 = 푒(14.472 + 0.797 ⁄2) × 퐷0.814 푙표푔(퐶) = 0.534 × 푙표푔(푥)1.253

NP RO D 2003 A 2014

× 1.096833 + 2.786

퐶 = (−3 × 10−6 × 푄2) UV Look-up table, UV (Table H-6) D 2003 B 2003 + (1.038 × 푄) + 4585

potable 2⁄ - (14.472 + 0.575 2) 0.716 SF 퐶 = 푒 × 퐷 D 2003 0.25 X Capital Cost F 1995

× 1.096833 Non 3 0.6

MF 퐶 = 5.36 × 10 × 푄 D 2010 C = 0.2 X Capital cost C 2010 2 - 0.598 0.830 푙표푔(퐶) = 1.828 × 푙표푔(푥)

NP UF 푙표푔(퐶) = 1.003 × 푙표푔(푥) + 3.832 A 2014 A 2014 + 1.876 (14.472 + 0.7972⁄2) 0.814 푙표푔(퐶) = 0.534 × 푙표푔(푥)1.253 RO 퐶 = 푒 × 퐷 2013 A 2014 × 1.096833 D + 2.786 Note: C – Cost of unit treatment; D – Design capacity in million gallons per day (MGD); Q – Design capacity in m3/d; gpm – gallon per minute

231

Table H-4: References for CAPEX and OPEX

Code Reference A Guo, Englehardt and Wu, 2014 B Economic and Social Commission for Western Asia, 2003 C Adewumi, Ilemobade and Van Zyl, 2010 D The Cadmus Group, 2006 E U.S. Environmental Protection Agency, 2000 F UNEP, 1995 G Qasim, 1999

Table H-5: Look-up-table (LUT) for Ozone

Design Capacity (m3/d) Cost, USD 90.85 524,942.03 329.33 637,154.99 378.54 655,117.58 1,022.06 711,831.35 1,703.43 866,473.55 2,460.52 1,020,895.85 3,104.04 1,315,222.30 3,785.41 1,499,007.33 6,813.74 1,667,528.62

Table H-6: Look-up-table (LUT) for UV

Design Capacity (m3/d) Cost, USD 90.85 21,425.92 329.33 29,235.60 1,022.06 41,219.82 2,460.52 66,274.00 6,813.74 244,264.48

Table H-7: ENR and BLS value for base year conversion

ENR1 BLS2 1989 1,988 50.3 1995 5,471 65.9 1996 5,620 66.7 2003 6,694 89.3 2010 8,802 113.3 2014 10,901 122.8 2016 11,500 128.0

232

Source: 1 Engineering News Records (ENR) at https://www.enr.com/economics 2 U.S. Bureau of Labour Statistics (BLS); all civilian workers 1982 - 2005 https://www.bls.gov/ncs/ect/sp/echistry.pdf 2001 - 2017 https://www.bls.gov/web/eci/echistrynaics.pdf

Table H-8: Inventory for coagulation/flocculation/sedimentation

Coagulation/Flocculation/Sedimentation Cashman, Lifespan Gaglione, A. Material Unit CH_IP-5 Process (years) Mosley, et al. (2014) Steel, low-alloyed {GLO}| market for | Steel 100 kg/m3 8.8x10-4 1.4x10-2 Alloc Def, U

Polyethylene, high density, granulate HDPE 100 kg/m3 5.8x10-6 9.2x10-5 {GLO}| market for | Alloc Def, U Concrete, normal Concrete 100 kg/m3 1.0x10-5 1.56x10-4 {RoW}| market for | Alloc Def, U Steel, low-alloyed Electrical 25 kg/m3 2.4x10-8 1.3x10-7 {GLO}| market for | steel Alloc Def, U Steel, low-alloyed Other steel 25 kg/m3 6.0x10-9 3.2x10-8 {GLO}| market for | Alloc Def, U Cast iron {GLO}| market Cast iron 25 kg/m3 1.8x10-7 9.6x10-7 for | Alloc Def, U Aluminium, wrought Aluminium 25 kg/m3 5.4x10-9 2.9x10-8 alloy {GLO}| market for | Alloc Def, U Copper {GLO}| market Copper 25 kg/m3 4.6x10-9 2.4x10-8 for | Alloc Def, U Cast iron {GLO}| market Cast iron 25 kg/m3 1.3x10-7 6.9x10-7 for | Alloc Def, U Steel, chromium steel Stainless 25 kg/m3 9.7x10-8 5.2x10-7 18/8, hot rolled {GLO}| steel market for | Alloc Def, U

233

Table H-9: Inventory for GAC and chlorination

Granulated activated carbon (GAC), and chlorination Cashman, Gaglione, J. CH_IP-5 Mosley, et al. Lifespan Material Unit (2014) Process (years) GAC Chlori GAC Chlori ne ne facility facilit

6 3 5

4 Reinforcing steel

- - - -

3 {GLO}| market

x10 x10 x10 Reinforcing steel 100 kg/m x10

0 for | Alloc Def,

4.1 1. 2.0

2.1 U

9 6 8

7 Concrete, normal

- - - -

6.5' concrete 3 {RoW}| market

x10 x10 x10 100 kg/m x10

piping 7 for | Alloc Def,

5.5 1.4 2.

2.9 U

8 5 7

6 Concrete, normal

- - - -

3 {RoW}| market

x10 x10 x10 Concrete 100 kg/m x10

for | Alloc Def,

4.8 1.2 2.3 2.5 U

Table H-10: Inventory for NF/RO (construction phase)

Nanofiltration (NF)/Reverse Osmosis (RO) Lifespan Bonton et Material Unit CH_IP-5 Process (years) al. (2012) Steel, low-alloyed {GLO}| Steel 60 kg/m3 5.0x10-5 3.0x10-5 market for | Alloc Def, U Concrete, normal {RoW}| Concrete 60 kg/m3 3.0x10-3 1.9x10-3 market for | Alloc Def, U Glass fiber {GLO}| market Fiberglass 60 kg/m3 7.0x10-5 4.3x10-5 for | Alloc Def, U

234

Table H-11: Inventory for MBR (construction phase)

Membrane bioreactor (MBR) Ioannou Lifespan -Ttofa et Material Unit CH_IP-5 Process (years) al. (2016) Steel, chromium steel 18/8, Stainless steel 20 kg/m3 4.6x10-4 1.2x10-4 hot rolled {GLO}| market for | Alloc Def, U Cast iron {GLO}| market for Cast iron 15 kg/m3 6.1x10-4 3.2x10-4 | Alloc Def, U Aluminium, wrought alloy Aluminium 30 kg/m3 3.0x10-4 1.6x10-4 {GLO}| market for | Alloc alloy Def, U Synthetic rubber {GLO}| EPDM 8 kg/m3 1.1x10-3 1.2x10-3 market for | Alloc Def, U Polyvinylchloride, bulk PVC 8 kg/m3 1.1x10-3 1.2x10-3 polymerised {GLO}| market for | Alloc Def, U Acrylonitrile-butadiene- ABS 5 kg/m3 1.8x10-3 4.8x10-4 styrene copolymer {GLO}| market for | Alloc Def, u Steel, chromium steel 18/8, Stainless steel 20 kg/m3 4.6x10-4 1.2x10-4 hot rolled {GLO}| market for | Alloc Def, U Polyvinylchloride, bulk UPVC 60 kg/m3 1.5x10-4 4.0x10-5 polymerised {GLO}| market for | Alloc Def, U

Table H-12: Inventory for DAF (construction phase)

Dissolved air flotation (DAF) O’Connor, Garnier Lifespan Material Unit and CH_IP-5 Process (years) Batchelor (2014) Concrete, normal {RoW}| Concrete 20 kg/m3 5.4x10-3 5.0x10-3 market for | Alloc Def, U Steel, low-alloyed {GLO}| Steel 20 kg/m3 1.9x10-4 1.8x10-4 market for | Alloc Def, U Glass fiber {GLO}| market Fiberglass 20 kg/m3 4.4x10-6 4.1x10-6 for | Alloc Def, U Aluminium, wrought alloy Aluminium 20 kg/m3 2.0x10-6 1.8x10-6 {GLO}| market for | Alloc Def, U

235

Table H-13: Inventory for SF, MF, UF, and UV (construction phase)

Sand filtration (SF), Microfiltration (MF), Ultrafiltration (UF), and UV Material Lifespan Unit Carré et al. (2017) CH_IP-5 Process

(years) SF MF UF UV SF MF UF UV

6 5 5 6 5 5

------

Aluminium, wrought alloy {GLO}| market for |

x10 x10 x10 x10 x10

Aluminium 60 kg/m x10

4 2 5

5 Alloc Def, U

9.0 4.4 5. 2. 1. 1.

7 6 8 6

- - - -

x10 x10 x10

Brass 60 kg/m x10 Brass {RoW}| market for brass | Alloc Def, U

2.3 8.8 6.2 2.3

4 4 4 5 4 4

------

x10 x10 x10 x10 x10

Cast iron 60 kg/m x10 Cast iron {GLO}| market for | Alloc Def, U

4 1 5

2.4 4.0 5.6 6. 1. 1.

4 4

- -

Chromium Steel, chromium steel 18/8 {GLO}| market for |

3 x10 60 kg/m x10

steel 3 Alloc Def, U

4.8 1.

5 6 5 5 6 6

------

Synthetic rubber {GLO}| market for | Alloc

x10 x10 x10 x10 x10

EPDM 60 kg/m x10 2

5 Def, U

7.3 8.2 2. 1.9 2. 6.6

5 6 -

Glass fibre - Glass fiber reinforced plastic, polymide

3 x10

reinforced 60 kg/m x10 injection moulded {GLO}| market for | Alloc

7 8 9. polyamide 3. Def, U

236

4 4 - Glass fibre -

Glass fiber reinforced plastic, polyester resin,

3 x10

reinforced 60 kg/m x10 3

9 hand lay-up {GLO}| market for | Alloc Def, U 1.

polyester 4.

3 2 4 3

- - - -

Polyethylene, high density, granulate {GLO}|

x10 x10

HDPE 60 kg/m x10

7 7

8x10 market for | Alloc Def, U

2.1 1. 5. 4.

6 7

- -

Polymethyl methacrylate, sheet {GLO}| market

3 x10

Plexiglas 60 kg/m x10 8

7 for | Alloc Def, U

3. 9.

6 7

- -

Polycarbonate {GLO}| market for | Alloc Def,

3 x10 Polycarbonate 60 kg/m x10

3 U

1.2 3.

6 5 4 6 5 5

------

Polyethylene, low density, granulate {GLO}|

x10 x10 x10 x10 x10

PE 60 kg/m x10

9 5

4 market for | Alloc Def, U

6.8 9. 1. 1.8 2. 5.0

5 6

- -

Polyphenylene Polyphenylene sulfide {GLO}| market for |

3 x10 60 kg/m x10

oxide Alloc Def, U

3.2 8.5

5 5 6 5

- - - -

Polypropylene, granulate {GLO}| market for |

x10 x10 x10 Polypropylene 60 kg/m x10

1 Alloc Def, U

2. 4.6 5.5 1.2

237

4 4 4 3 5 4 4 4

------

3 Polyvinylchloride, bulk polymerised {GLO}|

x10 x10 x10 x10 x10 x10 x10

PVC 60 kg/m x10 2

2 market for | Alloc Def, U

1.2 3.9 4.2 1.2 3. 1.0 1.1 3.

6 5 6 6

- - - -

Polyvinylidene Polyvinylfluoride {GLO}| market for | Alloc

x10 x10 x10 60 kg/m x10

fluoride Def, U

8.8 1.6 2.3 4.2

5 6

- -

Glass tube, borosilicate {GLO}| market for |

3 x10 Pyrex 60 kg/m x10

Alloc Def, U

3.7 9.8

3 4

- -

3 x10

Silica 60 kg/m x10 Silica sand {GLO}| market for | Alloc Def, U

1. 5.0

4 3 3 3 5 4 4 4

------

3 Steel, chromium steel 18/8, hot rolled {GLO}|

x10 x10 x10 x10 x10 x10 x10

Stainless steel 60 kg/m x10

3 1 7

7 market for | Alloc Def, U

2. 2.2 2.1 1. 7. 5.9 5. 3.3

5 6 5 3 5 6 5 4

------

3 Steel, low-alloyed {GLO}| market for | Alloc

x10 x10 x10 x10 x10 x10 x10

Steel 60 kg/m x10

2 2 6 8

7 Def, U

9.8 9. 9. 2. 2.6 2. 2.4 5.

238

Table H-14: Inventory for NP-1 and NP-2 treatment train (use phase)

NP-1 NP-2 Material Unit Process PTS MBR RO UV SF MF UF RO Input Material Sodium hydroxide, without Sodium 4.2x10- water, in 50% solution state kg/m3 0.002 hydroxide 3 {GLO}| market for | Alloc Def, U 1.3x10- Ultraviolet lamp {GLO}| UV lamp unit/m3 5 market for Z Alloc Def, U 1.2x10- Silica sand {GLO}| market for Sand kg/m3 2 | Alloc, Def, U Utilities x10lectricity, medium voltage Electricity kwh/m3 0.4 1.8 1.2 0.07 0.2 0.8 0.8 1.2 {MY}| market for | Alloc Def, U Output Emission to soil SS g/m3 65.0 34.7 3.5x10-1 77 20.7 1.4 0.9 Cd g/m3 2.0x10-2 2.0x10-4 1.9x10- 6.1x10- 2 4 As g/m3 4.3x10-2 5.6x10-2 9.0x10- 3 Cu g/m3 8.0x10-1 1.9x10-1 1.0 3.6x10- 4.0x10- 7.8x10- 3 4 7 Mn g/m3 8.4x10-1 1.5x10-1 8.0x10- 2.0x10- 1.0x10- 7.8x10- 1 1 2 5 Ni g/m3 6.5x10-1 3.4x10-1 9.5x10- 5.0x10- 2.0x10- 1 2 4 Al g/m3 14.7 2.3x10-1 11.2 3.0

239

Se g/m3 55.0x10-1 5.0x10- 1 F g/m3 4.9 1.7x10- 4.7 1

Emission to air

3 - kg

CH4 3 x10

CH4/m 2.5

Table H-15: Inventory for P-1 and P-2 treatment train (use phase)

P-1 P-2

ing

- 3

Material Unit 3 Process

Cl Cl

O O NP

SF SF

GAC

C/F/S

Blending

C/F/DAF 1/Blend Input

Material

- 2

- Aluminium sulfate, powder 3

Alum kg/m x10 {GLO}| market for| Alloc

3x10

4.5 Def, U

- -

Acrylic acid {GLO}| market

Polymer kg/m3

for| Alloc Def, U

7.5x10 7.3x10

2 -

Lime {GLO}| market for|

Lime kg/m3

Alloc Def, U 2.4x10

240

- -

Silica sand {GLO}| market

Sand kg/m3

for | Alloc, Def, U

1.2x10 1.2x10

3 -

Ozone, liquid {RoW}|

Ozone kg/m3

production| Alloc Def, U

5.9x10

- Activated carbon, granular Carbon kg/m3 {GLO}| market for activated

2.3x10 carbon, granular| Alloc Def, U

3 3 - - Sodium hydroxide, without Sodium water, in 50% solution state kg/m3

hydroxide {GLO}| market for | Alloc 4.2x10

4.2x10 Def, U

2 2 - - Sodium hypochlorite, without water, in 15% solution state Sodium kg/m3

hypochlorite {GLO}| market for| Alloc 1.4x10 1.4x10 Def, U Utilities Tap water {RoW}| tap water Potable m3/m3 0.8 0.7 production, conventional water treatment| Alloc Def, U x10lectricity, medium voltage Electricity kwh/m3 0.3 0.2 0.8 0.2 0.8 0.03 0.4 0.2 0.03 0.8 {MY}| market for | Alloc Def,

U Output Emission to soil

241

2 3 2 3

- - - - -

SS g/m3 21.1

7.5x10 1.3x10 5.6x10 7.3x10 1.1x10

7 3 6

- - - -

Hg g/m3

1.1x10

1.0x10 5.0x10 1.2x10 3.1x10

7 3

- - -

Cd g/m3

5.4x10

1.0x10 3.0x10 3.7x10

2 -

As g/m3

2.3x10

2.0x10

- -

Cyanide g/m3

2.8x10 1.1x10

2 -

Pb g/m3 0.2

0.2

8.8x10

2.2x10

7 8 2 5

- - - - -

Cu g/m3

6.0x10

4.6x10 5.4x10 6.0x10 4.1x10 1.6x10

242

8 3

- - -

Mn g/m3

2.5x10

8.0x10 2.8x10 5.3x10

- 5

- -

Ni g/m3 0.2

0.8

2.2x10

1.1x10 1.2x10

3 3 4 1 3

- - - -

- 1.6 Zn g/m3

1.7x10 1.8x10 3.4x10 1.2x10 1.1x10

2 5 7 2 2

- - - -

- 1.2 Fe g/m3

2.2x10 5.3x10 5.3x10 6.3x10 2.3x10

1 1 1

- -

- 2.2 Al g/m3

7.9x10 2.7x10 2.5x10

2 3 4

- - - Se g/m3

3.1x10 4.0x10 7.3x10

1 1 2 1

- - -

- 1.5 F g/m3

6.0x10 3.3x10 1.2x10 4.4x10

243

1 4 - - 3 Phenol g/m x10

2.0x10 4.

Emission to air

07 kg - CH4 3

CH4/m 2x10 4.

244