WATER POLLUTION AND HEALTH RISK CAUSED BY URBAN FLOODING IN CAN THO CITY: LESSONS LEARNT FROM THE FIELD CAMPAIGNS 2013

Proceedings of the 19 th IAHR-APD Congress 2014, Hanoi, Vietnam ISBN 978604821338-1 WATER POLLUTION AND HEALTH RISK CAUSED BY URBAN FLOODING IN CAN THO CITY: LESSONS LEARNT FROM THE FIELD CAMPAIGNS 2013 H. Q. NGUYEN (1), T. T. N. HUYNH (2), P. VAN DER STEEN (3), L. P. HO (4), A PATHIRANA (3), D. H. NGUYEN (1) & M BAINO- SALINGAY (3) (1) Institute for Environment and Resources (IER) Viet Nam National University Ho Chi Minh City (VNU HCM), Ho Chi Minh City, Vietnam, hongquanmt@yahoo.com, duynguyenhieu@gmail.com (2) University of Technology VNU HCM, Ho Chi Minh City, Vietnam, nguyenhuynh0608@gmail.com (3) UNESCO-IHE Institute for Water Education, Delft, the Netherlands, p.vandersteen@unesco-ihe.org, A.Pathirana@unesco-ihe.org, salin9@unesco-ihe.org (4) Center of Water Management and Climate Change (WACC) VNU HCM, Ho Chi Minh City, Vietnam, phi_hl@yahoo.com ABSTRACT Cities in developing countries, faced with rapid urbanization, encounter a number of problems that are connected to the urbanization process. Out of these urban flooding and water quality pollution are among the major ones. Increased imperviousness due to rapid urban densification, under-developed sewer systems, upstream flooding and tidal effects (in case of delta cities) are regular causes of urban flooding. Surface water pollution in urban areas come from both point and diffuses sources. While untreated wastewater from domestic, industrial activities are clearly defined pollution sources (i.e. point sources) discharging directly to local canals, other sources are diffuse, such as polluted runoff generated by the increase of impervious areas. This paper will present some initial results of flooding and water quality issues in Can Tho city based on an extensive monitoring campaign. The monitoring focused on different water bodies in Ninh Kieu District, Can Tho city, such as there are: (1): Open water (canals, rivers, lakes); (2): Flooded areas and (3): in Sewers. Flooded water quality and quantity were also considered. Sampling was before, during and after flooding events (average from 3 to 5 samples/site/event). Flooded water levels on the street, in the sewer and in the rivers were measured by automatic water level logging system. Public health risk due to flooding in Can Tho city were analysed by a Quantitative Microbial Risk Assessment (QMRA) approach. The results showed that both the open water and flooded water were highly polluted according to Vietnamese standards. The pathogen concentrations are highly variable during the flood event. Based on the QMRA method, the mean infection probabilities per urban flooding event for children is 0.2% o and 40% o ; for adults it is 0.1 % o and 14 % o for E. Coli and Salmonella, respectively. The results show that further investigations in the future on the flood and health related issues in Can Tho city are highly relevant. Keywords: Can Tho city, health risk assessment, monitoring, uncertainty, urban flooding 1. INTRODUCTION Cities in developing countries, faced with rapid urbanization, encounter a number of problems that are connected to the urbanization process. Out of these urban flooding and water quality pollution are among the major ones. Increased imperviousness due to rapid urban densification, under-developed sewer systems, upstream flooding and tidal effects (in case of delta cities) are regular causes of urban flooding. Surface water pollution in urban areas comes from both point and diffuse sources. While untreated wastewater from domestic, industrial activities are clear pollution sources (i.e. point sources) discharging directly to local canals, other sources are diffuse, such as polluted runoff generated by the increase of impervious areas. Polluted water in the rivers, drainage canals and sewer systems will come to the surface (streets, pavements etc) mixed with flood water. Thus flooding will likely impact on human health. For example, people are exposed to microbial contaminants while walking through flooded roads, playing with flooded water (normally kids in the cities). Different diseases are normally found related to flood events such as faecal oral diseases, vector-borne diseases, rodent-borne diseases, acute asthma and skin rashes (Tucci, 2001; ten Veldhuis et al., 2010; Taylor et al., 2011). Can Tho is the biggest city in the Mekong delta and has an area of 1,389 km 2. It is located on the southwest bank of the Hau River in the Mekong Delta. Can Tho is Vietnam s fourth largest city with approximately 1.25 million inhabitants. It was raised to the level of a first class city in June 2009, which is an administrative zone equivalent to a province and under the direct control of central government. Can Tho is considered the most important center for commerce, culture, education, and health services in the Mekong Delta (World Bank, 2012). Urbanization processes in Can Tho have rapidly increased in the last decades. The city is facing typical urban issues (e.g. flooding and water pollution) during its development processes as well as from future changes (Thy et al., 2010; Huong and Pathirana, 2013). For example, according to Thy et al. (2010), of 230 thousand houses in Can Tho city, 70 75% suffers the impacts of drainage problems, 10 15% are built over streams and canals. Huong and Pathirana (2013) showed significant impacts of flooding in 1

Can Tho city due to future predicted sea-level-rise and climate changes. This paper aims at providing an analysis of a typical urban flood event in the center of Can Tho city. The analysis comprised the dynamics of urban flood levels and water quality, especially contamination with pathogens. That information was subsequently used for a floodrelated health risk assessment. 2. METHODS 2.1 Study areas Ninh Kieu is the most urbanized and centralized district of Can Tho city. The urban drainage network in Ninh Kieu district consists mostly of combined closed sewer pipes. The network has been degraded seriously and has limited capacity. It is considered for upgrading to deal with urban pluvial flooding issues. In addition, the areas also suffer fluvial floods, especially during high tide. The high tides cause inundation in low lying areas through poor drainage systems. Most pipes are having direct outlets to the rivers, but often there is no valves, they are broken or of the wrong type. Both types of flooding are happening more frequent in the areas. Moreover, some other factors like local land subsidence (due to over groundwater extraction) and dynamic hydrological flow regimes in the main Mekong Rivers definitely impact to these urban areas. 2.2 Flood measurement and water samplings The monitoring focused on different water objects in Ninh Kieu District, Can Tho city,: (1): Open water (canals, rivers, lakes); (2): flooded roads and (3): sewers. Flooding levels and flood water quality were also considered. Hydrological measurements: The impacts of combined upstream and tidal floods were assessed by measuring the water level and flow discharge in the river and canals. In addition, water levels on the street and in the drainage canals/sewerage pipes (using automatic sensors) were recorded at the same time. The locations were the same as for some of the sampling locations. Water quality sampling: most of the water samples were taken at the same time as the hydrological measurements in the rivers, canal, lakes and on the street. Monitoring parameters included basic water quality like ph, COD, BOD 5, Nitrate (NO 3 --N), Amonium (NH 4 +- N), Phosphate (PO 4 3--P), Total P, Total N, Total suspended solids, and also microbiological ones such as E. coli, Total coliforms and Salmonella to serve the water quality and health risk assessments. It is very important to notice that the water quality sampling took place simultaneous with hydrological measurements. Water sampling frequencies are before, during and after flooding events (average from 3 to 5 samples per site). The Figure 1 shows the map of the monitoring campaign for a fluvial flooding event on 7 October 2013. Figure 1. The green line shows road networks in the Ninh Kieu District and the circles are showing the sampling sites where the red circle indicates (1): open water samples; the blue circle (2): flooded and/or sewer samples; the black circle (3): flooded samples; and the brown circle (4): sewer samples. 2.3 Quantitative Microbial Risk Assessment (QMRA) Quantitative Microbial Risk Assessment (QMRA) can be used to evaluate urban flood health risk, among other methods such as epidemiological population studies and comparison of the flood water quality with the EU water standard (Sterk et al., 2008). QMRA is a technique that has been developed for calculating the burden of disease from a particular pathogen (Haas et al., 1999; Howard et al., 2006). The calculation procedure consists of four successive steps: (1) Hazard identification; (2) Exposure assessment; (3) Dose-response relations; (4) Risk characterization. (Haas et al., 1999).Because of relatively large variation in pathogen concentrations and the nonlinearity of QMRA models, it is preferred to use average values with a probability distribution as inputs, rather than crisp values. To account for these variations and nonlinearity a Monte-Carlo simulation was applied, using the RiskAMP add-in for Excel software (Microsoft). In addition, to support the QMRA calculation, an interview campaign was conducted. A designed questionnaire was prepared consisting of 32 questions. The questions include general information (10); flooded situation in the areas, in their houses (8); flood related disease (7); adaptation measures (7). In total, initially 34 households were surveyed (see locations below). This information will provide estimated information on exposure pathways and the exposed population (Figure 2). Figure 2. Locations of interviewed citizens (yellow dots) 2

3. RESULTS 3.1 Flooding The campaign was executed during the highest tide period within a year (from September to November) 1. Flooded water comes to the city's streets either by overtopping of river banks or by excess water from sewer manhole surcharges. Both these cases were observed during the events (Figure 3). The flooding period on the road was ranging from 3 to 4 hours during the rising end receding tide, with flooded levels from 15 to 45 cm. There is only a short time lag between maximum river and sewer water levels (Figure 4). Figure 5. Total coliform variation at three surface water sampling locations (numbers 1 in Figure 1) during the flood event on 7 th Oct. 2013 (B1 and B2 are water criteria for irrigation and transportation, respectively) Figure 3. Flooding water on roads originates from rivers (left) and from sewer surcharges (right). Figure 6. E. coli variation at three surface water sampling locations (numbers 1 in Figure 1) during the flood event on 7 th Oct. 2013 (B1 and B2 are water criteria for irrigation and transportation, respectively) Figure 4. Recorded water levels in the river, sewer and roads during a flood event on the 7 th of October 2013. 3.2 Water quality dynamics Though more water quality parameters were measured (section 2.1), given the limitation of the paper content, only bacteriological quality results are presented here, namely E. coli and Total Coliform while the E. Coli and Salmonella are later used also for health risk assessment. Water quality in the river, in terms of coliforms are 5 to 50 times higher than the national standard for irrigation by surface water or transport on surface water (QCVN08:2008/BTNMT) (Figure 5) comparing to the B1, B2 as the lowest criteria for irrigation and transportation, respectively. The exceeding also observed with the E. coli parameter (Figure 6). However, the values are much smaller comparing to the Total coliform. It shows that the river water contains a lot of Total coliforms, but relatively few E.coli. The effect of flooding on pathogen concentrations in the sewer system was not evident from the measured pathogen concentrations in the sewer water (Figure 7). One may expect a dilution effect but this was not observed. At the Hoa Binh site, there is even an increase, but the reasons for that are not clear. At the Xang Thoi lake sewer, the concentration is a bit higher compared to the others. This was because it was observed at the site, the wastewater in the sewer seemed more concentrated. In general, the variations between samples from the same site were about one log unit. The water levels in the sewer system were still very high (full) during the sampling periods. Sewer sampling was limited and therefore could not capture a whole range of tide effects (rising and receding curves). Figure 7. Coliform variation at three sewer water sampling locations (numbers 2 in Figure 1) on the 7th of October 2013 1 According to observed water level at Can Tho station, the flood level on 7 th Oct. was the second highest water level in 2013. In the flooded water on the roads, water quality varied significantly during the flood event. There are certain 3

differences between flooded water caused by river overtopping (e.g. Hai Ba Trung, Tran Ngoc Que) and flooded water caused by sewer surcharge (e.g. Tran Van Kheo, Hoa Binh streets) (Figure 8, Figure 9). Figure 8 shows the large increase of E. coli after the flood event at the Tran Van Kheo and Hoa Binh streets, while the change is not so much as observed at the Hai Ba Trung and Tran Ngoc Que street. This observation clearly confirms that E. coli was certainly present in sewage water and thus the sewer surcharges cause an increase of E. coli. However, as observed in Figure 9, the variation of total coliforms is not so significant among the whole sites (except at Hoa Binh street which could be an outlier). As shown in Figure 5 and6, the river water contains high concentrations of total coliforms, but relatively few E.coli. In this regard, one could conclude that the E. coli in flooded water mostly is coming from the sewage system. Figure 8. E. coli variations of flooded water during the flooding event on the 7 th of October 2013 at some locations (numbers 2 in Figure 1) Figure 10. Coliform variation at Tran Van Kheo site during flood events 3.3 Health risk assessment According to the pre-defined 4 steps of the QMRA as described in section 2.3, the results are as follows: 3.3.1 Hazard identification The E. coli and Salmonella pathogens were used for the QMRA. In addition, data of 20 samples at 5 flooded sites were analyzed. The E. coli concentration varied from 1,600 MPN/100ml to 70,000 MPN/100ml. While the minimum Salmonella observed was 140 MPN/100ml and maximum was 9,100 MPN/100ml. The measured E.coli and Salmonella values have been used to define probability distributions of these pathogens. The Geometric fitting function (probability parameter of 0.000055 for E. coli and 0.00075 for Salmonella, respectively) was used. This fitting function (distribution) was obtained by input of all measurement data of E.coli and Salmonella in the Easyfit software 2. The distributions were used as input for Monte- Carlo simulations in @ RiskAMP (10,000 iterations), using the (Figure 11). Figure 9. Total Coliforms variations of flooded water during the flooding event on the 7 th of October 2013 at some locations (numbers 2 in Figure 1). There were some places that the water samples were taken both in the sewer and on the flooded streets e.g. at the Tran Van Kheo, Hoa Binh and Xang Thoi sites. At these sites, the flooded samples were taken after the sewer sample and the location was only 20 meter from the sewer manholes. Figure 10 shows that there is not much difference between the Total Coliforms in the sewer and in the flooded water. This could confirm that the coliform concentrations are not much different in the sewer and in the surface (river) water since a (large) part of the sewer and flooded water originated from the river water. Figure 11. Concentration probability distributions of E.coli (left) and Salmonella (right) of flooded water for 10,000 iterations 3.3.2 Exposure assessment According to ten Veldhuis (2010), the mean ingested volume for adults is 10 ml and children is 30 ml for a flooding event. These values are very close to those reported in Fewtrell and Smith (2007). However, it should be noticed the flood duration in Can Tho could be different from other places, thus the ingested volume can be different as well. Transportation custom (e.g. motorbike is popular in Vietnam) can also affect this estimation. In this study the above values were used. 2 www.mathwave.com/help/easyfit/index.html 4

3.3.3 Dose-response relations The β Poisson dose response model (Haas et al., 1999) is used to calculate the probability of E. coli and Salmonella infection (P inf ) for an event as follows: Where: µ : Dose of the pathogen α : Parameter that characterizes dose response function relationship N 50 : Median infection dose The dose was calculated as follows: Where: c : Concentration of a pathogen in water [nr of pathogens/l], from section 3.3.1 Hazard identification v : Ingested volume (section 3.3.2 Exposure assessment ) d : Dilution factor (as 1) The dose response parameters (α and N 50 ) are based on Haas et al (1999) where the values for α and N 50 are (0.1952; 3.01E+07), (0.3126; 2.36E+04) for E. coli and Salmonella respectively. 3.3.4 Risk characterization According to the dose response function, the mean infection probabilities per one urban flooding event to children is 0.00029 and 0.03960; to adults it is 0.0001 and 0.01396 for E. coli and Salmonella respectively. This result shows that Salmonella is apparently much more dangerous than E.coli. For example, from Salmonella, per flooding event about 4% of the children (and 1.4 % 0 of adult) will get a gastroenteritis infection. In addition, when comparing the Total coliform values to the Vietnamese Coastal Water Standard for bathing (QCVN 10:2008/BTNMT) and comparing the E. coli values to the Bathing water for inland waters of EU Directive 2006/7/EC, the measured concentration of flooded water are also much higher. Figure 12. Infection probability distributions for Salmonella of children and pedestrians. Figure 12 shows the infection probability distributions for Salmonella of children and adults when running the 10,000 Monte-Carlo simulations. This simulation shows that for a flood event, in 1,000 pedestrians, there is about 33% probability that 0 to 2 adults will get an infection. While in 1,000 children, it could be about 37% probability that 2 kids got infection. The above results are valid for one flooding event. This should be extrapolated and analysed for a whole year to get more solid conclusions on flood related health risk issues. Results from the interviewing campaign together with statistical information on flood frequency, population etc. can be used for this purpose. 4. DISCUSSIONS 4.1 Data uncertainty Uncertain flooding level data is because flooding issues in Can Tho city are caused by different reasons such as rainfall, tide, up-stream flow variations in time and space. Although tidal flooding is a major reason (SCE, 2013) and observed in this study, the flooding level data for the event on the 7 th Oct. 2013 may not be typical to represent the hydrological dynamics. Similarly, even more than that, water quality data, especially pathogen parameters often comprise a higher uncertainty since there are many processes involved (Novotny, 2002). In addition, given limited sample numbers at each site, these data are still highly uncertain especially in term of statistical assessment. 4.2 Lesson learnt In this study, an attempt was made to see the water quality (focusing on pathogen contaminants) variations in both temporal and spatial scales. A lot of efforts were made to co-ordinate a large group with totally about 30 people to work at the same time. Experiences in time allocation, human resources, budget arrangement, water sampling, preserving and analysis are up-taken for the next field campaign. 4.3 Next steps The first and most important next step is to collect more event data. This should include pluvial flood i.e. caused by rain. There are some other solutions that can be implemented in order to improve the QMRA results. For example, there should be an investigation about local people exposed to flooding under different conditions (e.g. tidal flood or rain flood or combined flood). Other parameters like Rotaviruses, Campylobacter, Cryptosporidium are often used to apply a QMRA, but were not used in this study. This could also be included in the next study either by using some available E. coli to pathogen ratio or by additional laboratory analysis(ten Veldhuis et al., 2010; Machdar et al., 2013). The risk characterization was assessed for a single event. This can be quantified annually based on e.g. the exposure pathway; flooding frequencies; exposed population. These information should be further elaborated using the results of social survey and local population statistics information. In addition, epidemiological data from local health services can be used to have flood related illness information. Especially, to get a better information on 5

exposed population, further social survey could be needed. The drainage system in Can Tho center has a large number of combined sewer overflows (CSOs) and the sewer water quality is highly dependent on the flooding water since these are open for inflows from the river into the sewer. Thus, the water quality measured during flooding event is essential important for sewage treatment system (e.g. waste water treatment plant capacity). Given the advances of mathematical model e.g. in water quality management (e.g. see in F. Blumensaat et al., 2012). These data can be severed for model calibration/validation. 5. CONCLUSIONS In this paper, we have presented some initial results of flooding and water quality issues in Can Tho city during flooding event based on an extensive monitoring campaign. The monitoring focused on different water objects over Ninh Kieu District, Can Tho city, as there are: (1): Open water (canals, rivers, lakes); (2): Flooded areas and (3): in Sewers. Flooded water quality and quantity were also considered. Samples were taken before, during and after flooding events (average from 3 to 5 samples per site/event). Flooded water levels on the street, in the sewer and in the rivers were also measured at the same time. The data showed that the water from different objects was highly polluted i.e. based on assessment of the pathogen contaminants. Based on the QMRA assessment we also find that, the mean infection probabilities per urban flooding event for children is 0.2% o and 40% o ; for adults it is 0.1 % o and 14 % o for E. Coli and Salmonella, respectively. Thus, the city is facing with a high health risk issue during flood event. Although the event data from this study can be a good reference, there are still sources of uncertainty in data collection e.g. to capture the dynamics. In addition, the results from the QMRA should be further elaborated in combination with other information. ACKNOWLEDGMENTS PRoACC (Post-doctoral Programme on Climate Change Adaptation in the Mekong River Basin) programme by the Netherlands Ministry of Development Cooperation (DGIS) through the UNESCO-IHE Partnership Research Fund. It was carried our jointly with UNESCO-IHE and Center of Water Management and Climate Change (WACC), Viet Nam National University Ho Chi Minh City (VNU HCM) REFERENCES F. Blumensaat, et al. (2012). "Sewer model development under minimum data requirements." Environ Earth Sci, 65, 1427-1437. Fewtrell and Smith (2007). Flooding and Infections. Urban Flood Management project report for the Flood Risk Management Research Consortium,, University of Wales, Aberystwyth: 37. Haas, C. N., et al. (1999). Quantiative microbial risk assessment. New York, Wiley. Howard, G., et al. (2006). "Quantitative microbial risk assessment to estimate health risks attributable to water supply: can the technique be applied in developing countries with limited data?" J. Water Health, 4(1), 49-65. Huong, H. T. L. and A. Pathirana (2013). "Urbanization and climate change impacts on future urban flooding in Can Tho city, Vietnam." Hydrology and Earth System Sciences, 17(1), 379-394. Machdar, E., et al. (2013). "Application of Quantitative Microbial Risk Assessment to analyze the public health risk from poor drinking water quality in a low income area in Accra, Ghana." Science of the Total Environment, 449, 134-142. Novotny, V. (2002). Water Quality: Diffuse Pollution and Watershed Management. Hoboken, New Jersey, John Wiley and Sons. SCE (2013). CAN THO (VIETNAM) - Comprehensive Resilience Planning for Integrated Flood Risk Management. Toward sustainable flood management in Can Tho and surrounding areas 2030, World Bank: 356. Sterk, G., et al. (2008). Microbial risk assessment for urban pluvial flooding. 11th International Conference on Urban Drainage. Edinburgh, Scotland, UK: 10. Taylor, J., et al. (2011). "Flood management: Prediction of microbial contamination in large-scale floods in urban environments." Environment International, 37(5), 1019-1029. ten Veldhuis, J., et al. (2010). "Microbial risks associated with exposure to pathogens in contaminated urban flood water." Water Res., 44(9), 2910-2918. Thy, P. M., et al. (2010). "Urban Expansion of Can Tho City, Vietnam: A Study based on Multi-Temporal Satellite Images." Geoinformatics, 21(3), 147-160. Tucci, C. E. M., Ed. (2001). Urban drainage in the humid tropics. Technical Documents in Hydrology No 40, Vol 1. Urban drainage in specific climates. Paris, UNESCO International Hydrological Programme (IHP-V). World Bank (2012). Tools for Building Urban Resilience: Integrating Risk Information into Investment Decisions. Pilot Cities Report Jakarta and Can Tho, Disaster Risk Management Team, East Asia and Pacific Infrastructure Unit (EASIN), The World Bank: 58. Copyrights Paper(s) submitted to the IAHR-APD2014 are interpreted as declaration that the authors obtained the necessary authorization for publication. 6