CHAPTER 3 Impact of technical and human factors on water quality in ten small Quebec utilities

Table des matières

Overview. The spatial and temporal variation of drinking water quality in ten small Quebec municipal utilities examined in the second chapter of this study has brought a number of precious indications as for factors, i.e. parameters, potentially responsible for the difference observed in distributed water quality between nonproblematic and problematic utilities on a historical basis, while also enabling to document the water quality in studied utilities from the source to the consumer’s tap. This second stage of the present study was very important, since it allowed searching for potential causes of observed historical differences directly in water that currently reaches the local consumers. It allowed as well identifying factors upon which utility managers may act to improve the quality of distributed water in each of the two opposed utility groups (i.e., nonproblematic vs. problematic).

Despite those all-interesting results, the second chapter obviously left aside certain potential explaining factors. The latter are factors relating to the whole range of operational (i.e., disinfection-related), as well as infrastructure and maintenance characteristics, to go with human and organizational factors specific to each of the studied utilities. So, to make the study complete, it appeared essential to consider exploring all of the aforementioned characteristics, as potential factors explaining both current and historical water quality in the studied utilities. This is what is being done in the next and last chapter of the present study.

Abstract. Ten small Quebec municipal drinking water utilities have been studied as for their operational, infrastructure, and maintenance characteristics, along with human and organizational factors governing the utilities’ life. All of these utilities use surface water or groundwater under the direct influence of surface water and apply chlorination as the only treatment before distribution. The ten utilities were subsequently divided into two groups: four utilities that had never or rarely served water infringing upon the provincial drinking water microbiological standards (relating to fecal and/or total coliform bacteria), and six utilities that very often infringed upon said standards. The objective of this study was to investigate the impact of the utility operational, as well as infrastructure, and maintenance characteristics on current distributed water quality in small utilities and to explore the impact of human and organizational factors, which govern the principal utility manager's action, on historical water quality in the same utilities. The study includes three distinctive parts: the first one is a portrait of studied utilities’ operational, infrastructure, and maintenance characteristics; the second part is devoted to development of indicators of performance for the same utilities, whereas the last part deals with human and organisational factors. The portrait revealed interesting trends in terms of distinctive features between nonproblematic and problematic utilities. Utility performance indicators were systematically better for the nonproblematic group as compared to the problematic one, with a major input from disinfection-related performance sub-indicators, and those bearing on infrastructure and its maintenance. As for human and organizational factors, they allowed highlighting such issues like educational background, supplementary training, experience, awareness of and preparedness to take up new challenges, and support from local authorities.

Key words: drinking water, water quality, small utilities, performance indicators, human factor, Quebec

Résumé. Dix petits systèmes municipaux de distribution d’eau potable au Québec ont été étudiés en ce qui a trait aux caractéristiques d’opération, de même qu’à celles liées à l’infrastructure et à sa maintenance, auxquelles ont été joints les facteurs humains et organisationnels régissant la vie de ces systèmes. Tous ces systèmes utilisent de l’eau de surface ou de l’eau souterraine sous influence directe de l’eau de surface et pratiquent une simple chloration. Les dix systèmes furent par la suite répartis en deux groupes : quatre systèmes qui n’ont jamais ou ont rarement distribué de l’eau dérogeant aux normes microbiologiques provinciales relatives à l’eau potable (en ce qui a trait aux coliformes fécaux et/ou totaux) et six systèmes qui ont très souvent dérogé aux dites normes. L’objectif de cette étude était d’explorer l’impact des caractéristiques opérationnelles, de même que celles de l’infrastructure, et de la maintenance sur la qualité courante de l’eau distribuée par ces petits systèmes, et de sonder l’impact de facteurs humains et organisationnels liés à la personne du gestionnaire principal sur la qualité historique de l’eau desservie par les mêmes systèmes. L’étude inclut trois parties : la première est un portait des caractéristiques d’opération, de l’infrastructure et de la maintenance ; la deuxième est consacrée au développement d’indicateurs de performance pour les petits systèmes ; quant à la troisième, elle traite des facteurs humains et organisationnels. Le portrait a révélé des tendances intéressantes en terme de traits distinctifs entre systèmes non-problématiques et problématiques. Les indicateurs de performance des systèmes étaient systématiquement meilleurs dans le groupe des non-problématiques comparativement à celui des problématiques, avec un apport crucial des sous-indicateurs de performance de la désinfection et de ceux ayant trait à l’infrastructure et à sa maintenance. Pour ce qui est des facteurs humains et organisationnels, ils ont permis de mettre en exergue des aspects tels que la formation principale, la formation complémentaire, l’expérience, la conscience des nouveaux défis et du niveau de préparation requis pour y faire face, et enfin l’appui des autorités locales.

Mots-clés : eau potable, qualité de l’eau, petits systèmes, indicateurs de performance, facteur humain, Québec

There are about 1,000 small municipal drinking water utilities (i.e., serving 10,000 people or less) in Quebec (Gouvernement du Québec 1997). Utilities of that size are the most numerous in the province. In other respects, it is well known that small utilities often lack adequate technical, managerial, and financial capacity (USEPA 1999).

In Quebec, small municipal utilities that have chlorination as the only treatment applied to drinking water before its distribution to their customers had been found most frequently violating provincial drinking water standards regarding microbiological quality (Gouvernement du Québec 1997). This has been mentioned by the Quebec Ministry of Environment (QME) based on 1984 Quebec drinking water regulations (QDWR) follow-up information (Gouvernement du Québec 1984), before the publication of new QDWR in June 2001. These new QDWR have already affected or will affect infrastructure needs and human resources in practically all small Quebec drinking water utilities. This opinion is based on new, very stringent, microbial inactivation and (or) removal requirements, not to mention personnel training and many other new requirements (Gouvernement du Québec 2001).

The role of the distribution system infrastructure in serving drinking water with irreproachable quality is vital. For instance, storage tanks (Opferman et al. 1995) and distribution mains physical and chemical properties (LeChevallier et al. 1990) play a big role in the possibility for utility managers to maintain the quality unchanged from the point of treatment to the point of consumption. In recent years, numerous publications have focused on the impact of some distribution system infrastructure components (e.g., pipe material, storage tanks) on consumer’s tap water quality (Opferman et al. 1995; AWWA 1998). However, very few of those studies considered the impact of the supply system as a whole (including source characteristics, treatment plant, storage tanks, distribution pipes, and all other components). Similarly, very few studies considered the impact of the characteristics of utility management by the principal operator/manager (henceforward: manager), that is the organizational and human factors. In addition, interest in these aspects has been much higher in median and large utilities than in small ones.

The objectives of this study are: 1) to investigate the impact of the system infrastructure as well as the operational and maintenance characteristics on the distributed water quality in small utilities; and 2) to explore the impact of human and organizational factors tied to the utility manager.

This case study bears on ten small utilities of the province of Quebec (Canada). To determine their significance, utility and human factor characteristics are related to current and historical water quality in the distribution system. Utility characteristics are integrated through performance indicators, while human and organizational factors are analyzed on a qualitative basis.

Based on a Quebec Ministry of Environment (QME) database on regulatory follow-up, with data for 1997, 1998, and 1999, two types of small utilities were distinguished. The first type included utilities that had never registered coliform positive samples or had registered such samples only on extremely rare occasions. The second type encompassed utilities that often registered coliform positive samples. Based on these remarks, two concepts were defined: coliform episode and problematic utility . A coliform episode indicated one or a set of coliform positive samples occurring in a given distribution system during the three-year period (1997-1999), separated by at least 15 days from any other coliform positive sample in the same system. A problematic utility was defined as a utility that registered one or more coliform episodes in at least two of the three reference years. Consequently, utilities that registered no coliform episode, or had episodes in only one of the above-mentioned three years, were designated as nonproblematic utilities . It is important to note that the concerned QME database comprised data from 927 small Quebec utilities with results of about 65,000 water sample analyses for the above-mentioned three-year period. It has been noticed that about 25% of the 927 utilities (that is, 230 utilities) have been experiencing repetitive coliform episodes. It was precisely that fact that led to the differentiation into “nonproblematic” and “problematic”.

Among utilities appearing in the QME database, ten have been subsequently chosen for the present study. The selection of these ten utilities was based on the following criteria: 1) they used either surface water (lake or stream) or groundwater under direct influence of runoff (surface wells); 2) chlorination was the only treatment applied; 3) for logistic reasons, they had to be located relatively close (within a radius of about 150 km) to Quebec-City; 4) the 10 utilities encompassed a group of problematic utilities and a group of nonproblematic utilities; and 5) utility managers had to be in agreement with the proposed study and offer to co-operate by favouring easy access to all infrastructure components and archived water quality data, and being available for interviews. Under these criteria, four nonproblematic and six problematic utilities were selected.

First, a study of the spatial and temporal variation of drinking water quality was done in the ten utilities (see previous chapter; see also Coulibaly and Rodrigez, 2003b). Table 3.1 gives an overview of some specificities of the studied utilities, along with some data showing water quality variability all along the distribution systems.

Information about distribution system infrastructure bore on characteristics like chlorination plant and machinery, storage tanks, and distribution network (pipelines). To gather that kind of information, a questionnaire was built up in October 2001 and the manager of each of the ten utilities was asked to answer its content during a semi-directive interview with inquiries focused on distribution system components, operational practices (i.e., disinfection-related variables), and maintenance practices (see Appendix D). Questions about the distribution system components bore on the presence/absence, dimensions (or capacity) of some components (e.g., emergency chlorinator, storage tank, etc.), or the relative importance of some kind of material (e.g., percent of cast iron pipes, of PVC pipes, etc. in the total length of the distribution lines). Questions about operational practices were on details like the method of chlorine injection or the frequency of chlorine residual measurements. As for maintenance practices, they bore essentially on distribution network flushing and pipe break management. Personal observations made by the authors during an eight-month field work corresponding to sampling campaign in the ten concerned municipalities in 2001, as well as all information drawn from local archives or from usual ordinary talk with respective utility personnel will also be considered.

Average of 5 monthly values * Total organic carbon

Average of 5 monthly values at each location ** Heterotrophic plate count bacteria

Table 3.2 shows the main studied characteristics of distribution system operation, infrastructure, and maintenance.

It is important to recall that all ten studied utilities have chlorination as the only treatment applied and use surface water or groundwater under the direct influence of surface water. This fact makes much greater the role and potential impact of an efficient and competent manager to ensure that these systems constantly serve water of irreproachable quality. So, human and organizational factors are being considered herein for the ten small municipal drinking water utilities with regard to the main managerial personnel (see Appendix E).

The information (or data) collection method used was also a semi-directive interview of the manager of each of the concerned ten small utilities. This allowed asking all questions that appeared on a questionnaire at hand while enabling him to tackle his specific issues of interest. The questionnaire comprised two sections and thirteen clusters of questions (see Appendix D). The first section contained general information on the manager, whereas the second section was constituted of inquiries about the distribution system management. The questionnaire, in its entirety, permitted to inquire about the utility manager’s socioprofessional characteristics and the organizational factors influencing his work. The manager’s socioprofessional characteristics encompassed major professional indicators of competency such as his general education and (or) academic standard regarding the drinking water field, knowledge of regulatory texts or standards, being up-to-date on the drinking water industry burning questions, and so forth. As for organizational factors, they include the manager’s whole working universe (environment), with all of the latter’s influences and interactions like the manager’s networking capabilities (that is, relationships with peers, experts, consultancy; subscription to water quality/water resources management journals, membership of associations or other organizations working in the field of drinking water, etc.). Organizational factors also include the direct or indirect influences of local administration (i.e., municipal officials) and its policy in the field of drinking water.

The information collected on human and organizational factors was treated using the analytical techniques (or methods, processes) of the positivist/postpositivist stance. The positivist/postpositivist stance (see Denzin 1994; Huberman and Miles 1994; Guba and Lincoln 1989) may enable linking the analyzed socioprofessional features of each manager and the whole organizational structure surrounding the utility with its historical water quality, in terms of causes and effect.

The analysis was done in two steps: first, a comparative portrait of the different studied variables (i.e., characteristics) was drawn up between nonproblematic and problematic utilities; then, an integration of these variables is carried out through development of utility performance indicators. Such explanatory indicators may help revealing the potential contribution of the examined variables in explaining observed differences in the distributed water quality.

The presence or absence of an emergency chlorinating device (or chlorinator) in the distribution system was the first disinfection-related variable examined (see Appendix F). It appeared that such a device was mostly absent in both groups of utilities. However, it was interesting to note that measures existed (or were planned for near future) wherever the concerned mechanism was absent among nonproblematic utilities, whereas nothing existed, nor was planned, to compensate for the absence of emergency chlorinator among the problematic utilities.

As for the type of chlorinator, devices in the two groups were similar. No manually chlorinating utility was found among the nonproblematic (there was one among the problematic). Manual chlorination, obviously, is much less efficient than using a well calibrated chlorinator, since it can in no way ensure an equitable distribution of the applied chlorine dose according to water flowrate at any time of the day (or night) like the chlorinator does. And, this may mean total depletion of chlorine residuals, with subsequent microbial regrowth or recovery within distribution networks and the potential public health repercussions of such phenomena. So, the chlorination devices variable is of great importance, since it probably affects the disinfection effectiveness (i.e., CT value). And, indicators of disinfection efficacy to come will sum up potential impacts of this and other disinfection-related variables.

The efficacy of drinking water disinfection procedures is estimated using the CT concept. The CT is a concept that aims at ensuring sufficient contact time (T) and maintenance of adequate disinfectant residual concentration (C) to attain disinfection objectives set by the utility designer, guided by water quality standards promulgated by regulatory institutions (Gouvernement du Québec 2002).

Pathogens contained in source waters must be removed before water is served to consumers. Microbial cells can be eliminated either by physical removal (i.e., via diverse filter media) or by chemical inactivation (i.e., using disinfecting agents). According to Gouvernement du Québec (2002), the resulting “log” of cell reduction can be estimated as follows:

Log of reduction = Σ physical removals + Σ chemical inactivations

Since the ten small utilities of this study have no other treatment than disinfection (i.e., chlorination), only inactivation can be considered. Furthermore, that situation makes chlorination the only barrier between potential source water pathogens and the consumer’s tap. Therefore, it is essential to ensure that that barrier be as effective as it could be. The disinfection effectiveness is evaluated in terms of “log” of inactivation (Gouvernement du Québec 2002, USEPA 1999). This value is determined using the following formula:

Log of inactivation = CTavailable / CTrequired

As its name suggests, the CTavailable is the actual CT value measured at the utility by the designers. As for the CTrequired, it is a value the designer is provided with via tables compiled by the USEPA (1991 and 1999) that indicates the required CT value to inactivate 1 log of a given microorganism (virus or Giardia or Cryptosporidium ) in water with given characteristics (pH, temperature, etc.) (Gouvernement du Québec 2002).

CTavailable = Cresidual x T10 = Cresidual x Vu/QMAX x T10/T

Where :

Cresidual is the disinfectant concentration at the chlorination facility storage tank outlet;

QMAX is the peak flowrate at the storage tank outlet;

Vu is the useful volume in the storage tank (not the latter’s capacity); and

T10/T is the hydraulic efficiency factor

Based on these considerations, the following approximate CT values have been calculated for the utilities at study (see Table 3.3). These CT approximates, calculated using the relatively limited (primarily for T10/T and secondarily for QMAX) data available with the ten utility managers, enabled making relative comparisons between nonproblematic and problematic utilities as for disinfection efficacy.

Cresidual has been estimated by taking the mean of residual chlorine concentrations recorded at the facility outlet and the distribution system central part, since there was no sampling point available directly at storage tank outlets. QMAX was obtained directly with utility managers, which considered it as equalling the overall power of available distribution system feed pumps. Vu was considered as equalling 80 percent of the storage tank capacity. And, very conservatively, the T10/T factor (which varies between 0 and 1) was considered equalling 0.2 when chlorinated water is stored in a tank before its distribution, and equalling 0.6 when chlorination is done directly into the water main en route for the consumer’s tap.

Average CT values for nonproblematic utilities are significantly higher than those recorded for problematic utilities (229 mg·min/L vs 106 mg·min/L, respectively). This supports findings made in previous work (see Coulibaly and Rodriguez 2003b), where disinfection-related water quality parameters were constantly found better in nonproblematic utilities, and that, all along the distribution network. If disinfection parameters were always better in nonproblematic utilities as compared to problematic ones, so it may be logical to presume that water quality as a whole was better in the nonproblematic group, since there was no other treatment than chlorination. With chlorination alone, the maximum reasonable disinfection objective for such utilities appears the 4-log virus inactivation (required by the 2001 QDWR for surface water systems), which is achievable with a CT of 15 to 60 mg·min/L for most temperatures according to USEPA (1999). As for the 3-log Giardia cysts and, especially, the 2-log Cryptosporidium oocysts inactivation (two of the many other requirements brought in by the 2001 QDWR for surface water systems), they will necessitate supplementary disinfection, most probably ultraviolet (UV) radiation or ozone (O3). Note that while chlorine can achieve 3-log Giardia cyst inactivation, the CT requirement for 3-log inactivation of 100 to more than 300 mg·min/L will require high chlorine doses and (or) long contact times (USEPA, 1999). The performance indicators, to be addressed later on, will make clearer the differences foreseen.

The first studied infrastructure characteristic was the utility age (see Appendix F). Aging water mains, especially those made of iron-based material, can cause water quality deterioration within the distribution network, particularly through corrosion. In addition to favouring precipitation of metal ions, which can cause coloured water, pipe corrosion may favour the formation of tubercles within which a biological film can form or cause breaks in the main, both aspects being favourable conditions for deterioration of microbiological water quality (LeChevallier et al. 1990). A brief comparison of nonproblematic utilities to problematic ones according to their age permitted to find out that the average age is higher for nonproblematic utilities (42 years versus 37.7 years). However, 3 out of the 4 nonproblematic utilities are less than 30 years, whereas only 3 out of the 6 problematic utilities in that situation. In fact, withdrawing the “extremely aged” utility in each group (i.e., nonproblematic utility III and problematic utility I) would have made the nonproblematic group appear significantly younger than the problematic group: average age of 26 years vs. 33 years. For a better idea of what that represents, it may be helpful to mention that survey results for medium and large Quebec utilities (Villeneuve and Hamel 1998; Fougères et al. 1998) showed that 65% of them are 35 years old or less. In comparison, 3 out of the four nonproblematic utilities studied herein are in such a situation, whereas only half of problematic ones could claim being in that category. However, these are only general portrait considerations; categorizations and (or) conceptualizations to come in the indicators portion will be more appropriate for identifying the potential impact of the age factor on the groups’ historical water quality indicator.

Storage tanks may have different types of impact on distributed water quality according to their physical and (or) chemical properties (Opferman et al. 1995). According to their internal wall properties, they may improve chlorine contact with bulk water, thereby enhancing microbial inactivation. However, when tank capacities are big and water demand low (i.e., water travel time too long), storage tanks could also be locations were chlorine residuals undergo rapid decay even before water begins its travel through the distribution network, en route for the consumer’s tap. Based on these considerations, the storage tank variable was included in the CT variable at the indicator development stage. In other respects, possessing sufficient storage capacity may appear as a sign of clearsightedness from the utility designers, in case an emergency strikes (firefighting, draught, important main breaks, and so forth). Because sufficient data are available only for making comparisons between the two groups of utilities according to storage tank numbers, capacities, average storage durations, and storage tanks localization, comments will be restricted to general portrait aspects at this stage. Thus, all nonproblematic utilities but one have two storage tanks each, whereas only 2 out of 6 are in that situation among problematic utilities (see Appendix F). It has been noticed that the nonproblematic group average tank capacity (875.6 m3) is significantly bigger than the one for the problematic utility group (604.9 m3). Likewise, the average storage volume per nonproblematic utility (1532.2 m3) is much more important than the same average per problematic utility (806.6 m3). The average storage duration is the same in the two groups (that is, about 42 hours). It may be interesting to note that 2 out of the 4 nonproblematic utilities had a storage time of 48 hours or more in comparison to 4 out of the 6 for problematic utilities. As for information on storage tanks localization, it shows that, in both groups, only half of them are located at the chlorination facility. This situation may have different consequences according to the general configuration of each distribution system. Further, conceptualized, analysis will be done later for exploring the potential impact of storage tank characteristics on distributed water quality.

The type of pipes (i.e., cast iron pipe, PVC pipe, etc.) chosen by utility designers and managers is of great importance in terms of distributed water microbiological and physicochemical quality. For example, as mentioned earlier, iron-based pipe material may cause water quality deterioration within the distribution network through corrosion, with its corollary being coloured water, tubercles and biofilm formation, and even main breaks (LeChevallier et al. 1990).

Pipe material composition was as follows: 1.2% grey cast iron, 71.4% ductile cast iron, 26.2% PVC, and 1.2% other material—on average for nonproblematic utilities versus 22.8% grey cast iron, 46.7% ductile cast iron, 29.5% PVC, and 1% other material—on average for problematic utilities (percentage calculated from data shown in Appendix F). The most important distinction seems to be tied to the proportion of grey cast iron—22.8% for problematic utilities (that is, exactly 19 times as much as in the nonproblematic group). Such a proportion appears huge, since grey cast iron pipes, which are being progressively abandoned (Villeneuve and Hamel 1998; Fougères et al. 1998), are known to be very sensitive to corrosion, which can be detrimental to distributed water microbiological and physicochemical quality. Secondarily, the proportion of ductile cast iron is much bigger in the nonproblematic group than in the problematic one (nearly the double). Ductile cast iron, especially when coated, is considered a much more resistant material to corrosion than grey cast iron. As for PVC pipes, their proportion is approximately the same in the two groups. PVC pipes are not corrodible; however, they are much less resistant to pressure than cast iron pipes, and have been found to release their own substances in distributed water. On the whole, the proportion of PVC pipes appears normal in for both groups, and so is the case for the proportion of cast iron pipes (72.6% for nonproblematic group versus 69.5% for problematic one), judging by results of a 1999-2000 survey of small Quebec utilities (see Rodriguez et al. 2002; Coulibaly and Rodriguez 2003a): on average, 63% of the distribution pipes were made of cast iron, and 28% made of PVC. Further attempts to characterize the impact of pipe material on tap water quality will be made later through development of related indicators.

Periodical flushing may be an efficient way to ensure distribution system overall healthiness, since it makes possible taking out biofilm and corrosion tubercles, both of which favour drinking water microbiological quality deterioration within distribution lines (Antoun et al. 1999; Duranceau et al. 1999).

On average, nonproblematic utilities had 1.75 flushing events per year, whereas the problematic group average was 2.17. Moreover, 4 out of the 6 problematic utilities had at least 2 flushing events per year, with two utilities doing more; in comparison, none of the nonproblematic utilities had more than 2 flushing events in any one year. It is at first sight surprising to see that the higher average flushing number pertained to the group with the worse water quality record. However, the most important difference may not be in the number but in the way flushing is performed, the portions flushed (entire distribution network or only chosen parts of it), at what season(s) flushing is performed, and which are the reasons that made utility personnel flush the system.

The utility manager’s opinion on distribution network flushing frequency may be interesting to know, since it gives an indication of his propensity to maintain the status quo or make changes for the future. It also allows having a relatively good idea of why and how flushings have been executed in the past. Generally speaking, all managers are rather of the opinion that distribution system flushing events are not uncommon at their respective utilities. A proactive stance seemed however to dominate among nonproblematic utilities, which, on the whole, considered flushing exclusively as a preventive sanitary measure to consolidate their distribution system overall healthiness. As far as the problematic group of utilities was concerned, a clearly reactive stance could be foreseen: flushing seemed to be considered in that case as a curative measure to get rid of repetitive microbial invasions. That is probably the main reason why problematic utilities performed more flushing than nonproblematic ones. However historical water quality indicators rather tended to prove that flushing alone could not solve the problem. And new indicators bearing on that variable, among others, will allow for a more realistic estimation of the impact of that management practice on distributed water quality.

Main breaks are known to be a possible gate for micro-organism and (or) other contaminant entrance into distribution systems (McDonald et al. 1997; CMHC 1992). This means that good management of drinking water main breaks could only be beneficial to the ultimate consumer’s tap water quality. While main breakage rates appear much higher among problematic utilities as compared to nonproblematic ones (see Appendix F), it is surprising to note that in both groups, the highest breakage rate pertained to a relatively young utility (i.e., 28 for the nonproblematic, and 26 for the problematic). The overall group breakage rate for the nonproblematic utilities was only about 6/100 km/year, whereas the problematic group recorded more than twice as much (about 14/100 km/year). In other respects, it is interesting to note that only half of nonproblematic utilities declared having experienced main breaks in the previous year, whereas 4 out of the 6 problematic ones acknowledged having had them. Nevertheless, the main breakage rate in both groups might be considered as not giving serious cause for concern, since, according to McDonald et al. (1997), main break rate can be considered abnormally high when it exceeds 40/ 100 km/year (none of the ten utilities had this many). The group main break averages mentioned above (i.e., about 6/100 km/year for the nonproblematic, and about 14/100 km/year for the problematic) also appear rather acceptable, compared to the average for drinking water distribution systems of Ontario towns (25/100 km/year) (CMHC 1992), the average for U.S. towns’ distribution systems (about 13/100 km/year) (AWWA 1994), and the average for 114 small Quebec utilities (about 29/100 km/year) (Coulibaly and Rodriguez 2003a).

Main break frequency and distribution main leakage did not appear to cause any big concern among utility managers. This may be easily understandable, since, as seen previously, the overall nonproblematic and problematic group breakage rates are relatively low compared to those of other North American utilities. As for water loss through leaks, its impact seemed decidedly very little, which could mean that the overall portion of unaccounted for water was not bigger than 10 to 15%. All of the just said about breaks and leaks is in total agreement with the relatively young age of both utility groups. Indicators that will be identified below for main breakage will allow for more conclusive comparisons.

An abundant literature has been produced about water and environmental quality indices or indicators over the last three decades (Brown et al. 1970; Ott 1978; Yu and Fogel 1978; Dunette 1979; Ball et al. 1980; Porcella et al. 1980; Béron et al. 1982; Couillard et Lefebvre 1986; UNEP 1994; Zandbergen and Hall 1998; Cluis et al. 2001; Lence and Ruszczynski 2001).

Unlike usual water quality indices that are generally intended for characterizing a variable’s ‘‘state of being’’ in relation to a specified use (Laroux et Weber 1994), the performance indicators that are being developed herein will be oriented towards explaining a situation or demonstrating a phenomenon. As a matter of fact, these indicators will aim at explaining why the quality of the distributed water is better in nonproblematic utilities than in problematic ones (that is, the historical water quality), thereby demonstrating the impact of a number of crucial variables on the current (i.e., recent) water quality.

As indicated by Béron et al. (1982), for the identification of good indicators, it is better to stick to a relatively limited number of crucial variables, rather than trying to encompass all variables that may influence the phenomenon being characterized. Based on these considerations, a number of variables have been selected from those described in the last chapter in order to develop the indicators. The selected variables are shown in Table 3.4. Because of normally close relationships between some of the above-mentioned variables, a number of them have been considered as having their potential impact already expressed through connected variables that were retained for indicator development. As an example, the CT variable encompassed considerations for temperature, pH, free chlorine residual, and storage tank characteristics. The last named parameters or characteristics contributed either directly or indirectly to the CT value computation. In such cases, only the most “comprehensive” variable (e.g., the CT value) is retained. All individual variables have been conferred a weight, according to the relative importance of each of them based on pertinent literature indications (e.g., Béron et al. 1982; Couillard and Lefebvre 1986) and the concrete statistical levels of significance exhibited by the water quality, as well as operational and maintenance parameters, in the same small utilities in the two previous chapters (see also Coulibaly and Rodriguez 2003a,b). The parameters that exhibited the strongest significance in those chapters (e.g., disinfection-related ones) have been given the biggest weights.

As shown in Table 3.4, four kinds of parameters or variables have been retained for the development of indicators. First, an environmentally relevant variable was retained, which bears on the agricultural land use on the territory of the ten municipalities hosting the water utilities. This variable was taken from the first chapter. Second, five raw water quality variables were also retained from a recent sampling campaign in the same ten utilities (see second chapter). Third, three disinfection-related variables were chosen from the same sampling campaign. Fourth, four variables bearing on infrastructure and maintenance characteristics were selected (Table 3.4).

A number of major raw water characteristics have been included as variables in the determination of performance indicators. This is justified by the primary importance of source water quality for the studied utilities since they apply no other treatment than chlorination. The absence of sophisticated treatment (e.g., coagulation, flocculation, settling, filtration) makes the removal of natural organic matter and potential parasite cysts or oocysts quasi impossible. Thus, the capacity of such utilities to serve good quality water is much more impacted on by source water quality than it is for larger utilities.

For indicators development, the procedure used was the following. First, four explanatory “sub-indicators” have been identified. These sub-indicators corresponded to the four variable groups mentioned in Table 3.4. The four sub-indicators will be used to calculate a performance indicator for each of the ten utilities at study. Then, overall performance indicators are determined for both nonproblematic and problematic utilities. Finally, the two resulting overall indicators are compared to each other, and then put in relation with the recent distribution water quality (generated in 2001), which is represented by an indicator based on variables shown in Table 3.5. Weights of this table have been determined based on the same criteria than for Table 3.4.

For this study, the indicator computation method to be used is the weighted additive one. This method has been preferred to others (e.g., weighted multiplicative method) because it allows a linear transformation of performance points into primary indicators. Most importantly, the weighted additive method, which is based on arithmetic mean, will allow avoiding giving to much importance to low performance scores. So, as an example, this method is much less severe than the weighted multiplicative method (based on geometric mean) (Couillard and Lefebvre 1986). The weighted additive method proceeds as follows: the parameter (or variable) values are transformed into performance scores (see Appendix G for detailed explanations of how that procedure was carried out in this study), and the latter are weighted and added up to give a unique value (Yu and Fogel 1978; Ball et al. 1980; Béron et al. 1982; Couillard and Lefebvre 1986).

The general formula utilized for computations is the following:

n

Ip = Σ wiγi = w1γ1 + w2γ2+ ... + wnγn (Equation 1)

i=1

Where:

Ip is the utility performance indicator (weighted additive indicator);

wi is the weight for the ith variable;

γi is the performance score of the ith variable;

n is the number of variables.

As detailed in Appendix G, the performance levels vary from 0 to 100 in terms of performance points, which generally correspond to given percentile values. Table 3.6 shows the performance scores on all considered variables for utilities at study.

Using the Equation 1, all sub-indicators have been computed (Table 3.7). Adapting literature examples (e.g., Béron et al. 1982) to the specific nature of the variables and the objectives of the study, the following performance significance scale has been defined for sub-indicators and indicators: 0 through 20 ― E; >20 and ≤40 ― D; >40 and ≤60 ― C; >60 and ≤80 ― B; >80 and ≤100 ― A (Table 3.8 and Table 3.9). This scale has been made very conservative due to the empirical nature of most of variables (e.g., pipe age, main breaks). For utility performance sub-indicators determination, the amount of performance points on each variable was multiplied by this variable’s weight and added to the weighted performance points of the other variables pertaining to the same sub-indicator. Then, the resulting weighted sum was divided by the possible maximum weighted amount of points available on that sub-indicator and multiplied by 100. As for the utility performance indicator, it was computed by adding up the weighted values of the four sub-indicators by the corresponding variable group weight.

The agricultural land use sub-indicator demonstrated a relatively good impact on tap water quality indicator (Table 3.7 and Table 3.8). Of the four very good performances (i.e., score A) recorded for that sub-indicator, three resulted in acceptable current tap water quality indicator or better. On the other hand, none of the four utilities that had poor or very poor performances on that sub-indicator was found with high current tap water quality indicator (i.e., very good or good performance). Of the nine utilities that recorded the maximum level of performance on the raw water quality sub-indicator (i.e., 100 points), none had that much performance as current tap water quality indicator; instead, three of them exhibited poor or very poor performance in terms of current tap water quality indicators. For the disinfection-related sub-indicator (by far the most important one, since these utilities applied no other treatment), only one out of the four utilities that recorded a very good or good performance did not have at least an “acceptable” performance on current tap water quality. Of the six utilities that had ‘‘acceptable’’ performance or less on that sub-indicator, three exhibited poor or very poor performance on current tap water quality. As for the infrastructure and maintenance sub-indicator, it also showed a positive impact on the

current tap water quality indicator, as only two of the eight utilities that recorded either very good or good performance on that sub-indicator exhibited poor performance on current tap water quality. At the same time, all of the two utilities that did not have more than ‘‘acceptable’’ performance on that sub-indicator showed poor performance on current tap water quality. As for utility performance indicator and current tap water quality indicator, they will be commented below, using Figure 3.1, Figure 3.2, and Figure 3.3.

Using the earlier mentioned historical water quality indicator (i.e., nonproblematic vs. problematic), overall performance indicators have been identified for the two groups of utilities (Table 3.9). The overall performance indicators corresponding to the two stances of the historical water quality indicator were obtained by taking the non-weighted average of the four sub-indicator values (see Table 3.7) for each of the two groups of utilities, then weighting them by the corresponding variable group weight and adding them up. The current overall tap water quality indicators have been calculated using the same procedure. And, the same 0 to 100 scale, as for Table 3.8, was used to qualify utility group levels of performance.

A= Very good performance; B= Good performance; C= Acceptable performance; D= Poor performance; E= Very poor performance

There are many interesting comments to make about Table 3.9. First, all overall sub-indicators but one favour the nonproblematic group of utilities. The only one in favour of the problematic group is the raw water quality overall sub-indicator. Although the nonproblematic group also performs well on that sub-indicator, this important remark

* Historical water quality indicator

A = Very good performance; B = Good performance; C = Acceptable performance; D= Poor performance

furnishes a big support to comments made in the last the chapter about the fact that differences seen in current and historical tap water quality between the two utility groups probably have their main causes inside the distribution system, not in the source water. In other respects, it is interesting to notice that the problematic group of utilities performs relatively well on the infrastructure and maintenance overall sub-indicator, only slightly less than the nonproblematic group. That indicates that infrastructure and maintenance are in good condition in the nonproblematic and problematic group alike. However, when it comes down to the agricultural land use overall sub-indicator and, especially, to the disinfection-related overall sub-indicator, the situation is unequivocally in favour of the nonproblematic group of utilities. It appears more and more probable that, for current tap water quality, the disinfection-related overall sub-indicator is the central explaining factor of the overall much better situation of the nonproblematic group of utilities as compared to the problematic group. As for the overall performance indicator and current overall tap water quality indicator, they are commented beneath, along with utility performance indicator and current tap water quality indicator.

A graphical representation of utility- and overall performance indicators, as well as current tap water quality- and current overall tap water quality indicators is given in Figure 3.1. This figure shows that in every aspect of utility performance and tap water quality indicators, the situation in the nonproblematic group of utilities is better than the one in the problematic group. The significant differences observed between real values of the overall performance indicators for the nonproblematic and problematic group (75 and 59, respectively) and, particularly, between the current overall tap water indicator values (68 and 36, respectively) come in support of the last assertion. Figure 3.2 confirms the hypothesis that better performance corresponds to better consumer’s tap water quality. Indeed, in Figure 3.2, the current (i.e., 2001) microbiological tap water quality varies in direct proportion to the utility performance indicator. This finds also good support in Figure 3.3, although the Pearson Determination Coefficient (adjusted R2 = 0.27) does not, at first sight, seem to confirm it. Indeed, it is easily understandable that the R2 be not high, since the number of observations (i.e., statistical cases; n = 10) is very limited. A careful examination of Figure 3.3 permits to notice that, on the whole, the current tap water quality indicator is better with higher utility performance indicator.

The determination of variable weights (e.i., wi) showed in Table 3.4 and Table 3.5 was based on two approaches. The first one consisted in taking into consideration of all variables that exhibited a relatively good level of significance (at least at the 10% level, P<0.1 ) in previous stages of the study, that is, in Chapters 1 and 2. The more significant the variable proved to be, the bigger was its weight. The second approach entailed consideration of all potential explanatory factors that have not yet been considered in the ten utilities. These factors have been conferred weights based on literature indications (that guided the author’s judgment). The fact of considering certain variables for indicator

Figure 3.1. Relationships between utility performance indicators and current tap water quality indicators in nonproblematic (NP) utilities with those in problematic (P) utilities

Figure 3.2. Relationship between utility performance indicator and current (2001) microbiological tap water quality

Figure 3.3. Graphical representation of the relationship between the utility performance indicator (upi) and the current tap water quality indicator (twi)

development because they turned out to be statistically significant in previous stages of the study involved an a priori stance. That is the reason why variables have been fixed conferred weights before indicators were subjected to a sensibility analysis.

Two approaches of sensitivity analyses are being proposed for the utility performance indicator: 1) making sub-indicator weights to vary; 2) excluding (i.e., withdrawing) sub-indicators.

Varying utility performance sub-indicator weights (through doubling or halving of their constitutive individual variable original weights) yielded eight scenarios (see Appendix H, Table H.1). In fact, that operation represented much more than simply doubling or halving original variable weights: it often implied simultaneous adjustment of some or all other variable weights to maintain the sum of all weights equal to 1. When one sub-indicator weight is doubled, the weight of at least one of the remaining three sub-indicators is reduced. This weight reduction is mainly executed at the expense of the most weighted sub-indicator among the other three, which often fell on the disinfection-related sub-indicator. This process narrowed the gap between sub-indicator weights. Note that doubling the latter sub-indicator’s weight resulted in cancelling all others’ weight since that sub-indicator represented more than half of the overall weight. Likewise, when one sub-indicator weight is halved, the weight of at least one of the remaining three is raised. The rise fell mainly on the least weighted sub-indicators, which were the agricultural land use sub-indicator and the raw water quality sub-indicator. This process also tended to diminish the gap between sub-indicator weights. So, eventually, these weight changes had the effect of giving more impact to sub-indicators (or variables) that did not have much of it in the original scenario.

The impact of sub-indicator weight variations is visible (see Appendix H, Table H.2). However, in all of the eight scenarios, the nonproblematic group of utilities showed a higher overall performance indicator. Moreover, in most cases, the gap between the overall performance indicator values of the nonproblematic and problematic utility groups remained very comparable to the one obtained in the original scenario (that is 75 (B) vs 59 (C), respectively; so the original gap is about 15 performance points). In fact, in seven of the eight concerned scenarios, the gap varies between 10 and 20 performance points, with the only one remaining being about 8 points in favour of the nonproblematic group. Overall, the nonproblematic group of utilities had exclusively good performances, whereas the problematic group reached such level of performance only three times out of eight.

One-at-a-time cancellation of utility performance sub-indicators yielded four scenarios (see Appendix H, Table H.1). Except for one case (when the disinfection-related sub-indicator was cancelled, resulting in ten individual variables with equal weights), the same approach of raising the least sub-indicator weights while reducing the biggest ones (as described above) was applied, and with the same tendency of narrowing the gap between the remaining sub-indicator weights.

The impact of sub-indicator cancellations is obvious. As an example, conferring an identical weight (that is 0.1) to all other individual variables except the ones composing the disinfection-related sub-indicator resulted in a very comparable overall performance indicator (the closest of all) between the nonproblematic and the problematic groups of utilities (78 and 75, respectively) (Appendix H, Table H.2). However, in all other three scenarios, the gap between the two utility groups (in terms of performance points) varies between 10 and 20 as was the case above. Anew, on an overall basis, the nonproblematic group of utilities exhibited exclusively good performances in these last four scenarios, whereas the problematic group scored as much only on two occasions out of four.

Because of the particular nature of this information, human and organizational aspects were treated as a case study (see Appendix E). According to Huberman and Miles (1994), a case is a phenomenon of some sort occurring in a bounded context—in fact, the unit of analysis. Two cases are being analyzed in a comparative style. The first case is represented by a group of four nonproblematic utilities; the second—by a group of six problematic utilities (see Figure 3.4). But these two cases may not be monolithic blocks; within each case, there may be certain differences between member utilities. And these differences may appear interesting enough to necessitate a brief analysis herein. Hence, the analytic strategies that will be followed are within-case comparisons (between utilities in each group as for managers’ socioprofessional characteristics and organizational factors) and across-case comparisons (between the two groups of utilities (i.e., cases) according to the same variables or distinctive features). To allow for within-case analyses, clusters have been identified whenever possible. In the first case (i.e., nonproblematic or case-1), only cluster A could be identified, whereas two clusters ( B and C ) were identified in the second case (i.e., problematic or case-2) (see Figure 3.4). The criteria for identifying the clusters are as follows. For case-1, cluster A is constituted of utilities that recorded no more than one coliform episode during the considered three-year period. In case-2, cluster B is formed of utilities that recorded from two to four episodes (in fact, utilities I and VIII recorded exactly two episodes each), and cluster C —of utilities that recorded five or more episodes

during the studied three-year period.

Figure 3.4. Clustering of the studied utilities according to the level of their being nonproblematic or problematic

The first step of qualitative comparative analyses will be done between clusters identified within case-2. Hopefully, these analyses will allow for identifying certain interesting distinctive features between case-2 member utilities.

Since only one cluster has been identified in case-1 (i.e., cluster A ), no cluster comparison could be made for that case. As for the two case-2 clusters (i.e., B and C ), three important differences have been noticed between them: the two cluster B managers were the only ones to have had some educational background dealing with the water issue. They have got the lesser problematic utilities among the problematic group. The same cluster B managers were those who most unequivocally welcomed supplementary training to come with 2001 QDWR implementation. Surprisingly, cluster C managers, with the most problematic utilities of all, appeared to enjoy better support from local municipal authorities than cluster B managers.

For the first point mentioned as a difference (i.e., educational background), there is no doubt that utility managers who have got certificates (or diplomas) in civil engineering or water sanitation were better prepared for the job, and were much more likely to be effective and get good results (i.e., distributed water quality records) than those who have come to learn directly on the job (see Appendix E). The level of being problematic for the two compared clusters supports this. The second difference is rather evocative of the managers’ mental predisposition, with those supposedly best prepared for the job being also the ones that were most willing to get better. The third point may have something to do with the accuracy of cluster C managers’ responses, since, usually (as it will be demonstrated by across-case analyses), the more problematic a small municipal utility is, the weaker is the support its manager gets from local authorities.

As mentioned earlier, the goal of across-case analyses is to identify and interpret significant differences between case-1 and case-2 utilities, i.e., between nonproblematic and problematic utilities. First of all, case-1 utility managers were older: 3 out of the 4 case-1 managers were aged (i.e., more than 50 years old), with mean age equalling 42 years, whereas 4 out of the 6 case-2 managers were of mature years (i.e., 30 to 50 years old), with mean age equalling about 38 years. It is important to note that none of the ten municipal utility managers was very young (i.e., of age less than 30 years). Because, generally speaking, experience comes with age, it is understandable to presume that case-1 managers were also more experienced in the field of drinking water. Indeed, 3 out of the 4 case-1 managers were experienced or better, whereas 4 out of the 6 case-2 managers were little experienced or lesser. This of course is an interesting indication, considering the historical water quality indicators of the two groups of utilities. So, it appeared that the utilities with the best historical water quality record were also those with the most experienced managers. That is logical, but certainly not necessarily compulsory.

A relatively surprising finding was that none of the case-1 managers indicated to participate in conferences or seminars, whereas one third of case-2 managers mentioned to take part in such events. But this fact might not be so decisive: first, only 2 out of the 6 case-2 managers claimed to do so and, second, the number of those conferences and/or seminars and their participants’ academic or training level might also mean a lot. Another somewhat surprising finding is that only half of case-1 managers clearly claimed to have a good knowledge of the 2001 QDWR about six months after their publication, whereas 5 out of the 6 case-2 managers claimed good knowledge of new DWR at the same period. A possible explanation of this seemingly laxness from apparently good managers might be simply that “publication” did not mean immediate implementation, since the mentioned standards had to come into effect only one year after their publication date.

And, again, half of case-1 managers were in favour of new DWR training requirements, whereas 5 out of the 6 case-2 utilities were favourably disposed towards them. A possible explanation of that is that case-1 managers did not see the necessity of such requirements because of their utilities’ good historical water quality record. Moreover, half of case-1 managers considered their level of training already adequate (with regard to new DWR implementation) versus only one third (2 out of 6) of case-2 managers; this might also be an explaining factor.

Judging by the answers given by utility managers, case-2 utilities might appear much closer to being ready for full compliance with new DWR than case-1 ones. However, taking into account the whole situation of the concerned utilities, case-2 managers are rather suspected of underestimating the immensity of challenges their respective utilities faced with regard to new provincial standards. Conversely, case-1 managers’ relative reserve could indicate their being really aware of the difficulty of tasks that fall on them due to new DWR, as well as their concern about being able to take up such challenges. This would probably explain why only 1 out of the 4 case-1 managers had clearly claimed an overall good appreciation of 2001 QDWR, whereas 4 out of the 6 case-2 managers claimed satisfaction or better.

The higher degree of case-1 managers’ awareness of challenges facing them was confirmed by the great relevance of issues that they mentioned as positive: half of them mentioned big issues like strengthened bacteriological control and training needs for managers. None of case-2 managers mentioned these factors. Instead, case-2 managers rather complained about a supposedly excessive number of samples required by virtue of new DWR, and the all-known financial needs. So, overall, case-1 managers’ preoccupations were much closer to a better consumer’s tap water quality than were case-2 ones’.

The municipal authorities’ support was certainly not the least factor. All case-1 managers stated receiving satisfactory or better support from local officials. As for case-2 utilities, only half could claim satisfactory support. This could be a tremendous difference, especially considering that these utilities had no other funding possibility than the one coming through local authorities, whether that be municipal, provincial or federal money.

Finally, it is regretful to notice that across-case analyses could not be applied to the fact of the utility manager's having educational background dealing with the water issue. In fact, none of case-1 managers had such a background; as a result, no comparison could be made between the two cases as for that distinctive feature. It could be reasonably presumed however that if the concerned information were available in case-1, the comparison would have confirmed observations made in within-case analyses (i.e., between case-2 utilities) concerning that feature.

To conclude, across-case analyses led to the following clear distinctive features: 1) case-1 managers were older, but much more experienced; 2) case-1 managers appeared to be more aware of challenges brought in by new 2001 QDWR and, hence, better prepared to face them; and, 3) case-1 utilities appeared to receive significantly more support from their local authorities than case-2 ones when it came down to the drinking water utility needs.

Distribution system operational, infrastructure, and maintenance variables analyzed herein showed some interesting trends in terms of distinctive features between the nonproblematic and problematic groups of utilities in relation to their distributed water quality. The trends noticed in the general portrait features have been almost systematically confirmed by relating indicators.

Almost all indicators point towards better performances in nonproblematic utilities, which are also those having the best current water quality in the distribution system. While, on the whole all indicators are better in the nonproblematic group, a specific focus comes on disinfection-related performance sub-indicators, and those for infrastructure and its maintenance. It appears that these factors are really those that have the biggest impact on distributed water quality in small utilities at study.

The sensitivity analyses applied to the utility performance indicator showed that the methodology employed stands the test of individual variable and sub-indicator (or variable group) weight changes. As a matter of fact, in the twelve scenarios tested, the nonproblematic group of utilities exhibited exclusively good performances, whereas the problematic group matched that only on some occasions (in 5 out of 12 scenarios), with overall performance numeric values systematically lower than those of the nonproblematic group of utilities.

The developed small utility performance indicators suggest that it is very difficult to make good tap water from bad source water; however, it is very feasible to improve water quality between the source and the consumer’s tap when adequate operational, infrastructure, and maintenance, as well as human and organizational resources are brought together.

Qualitative studies in the field of drinking water are rare. This study gives indications that human and organizational factors probably play a much more important role in the quality of the consumer’s tap water than most stakeholders notice. It is obvious that even the most sophisticated and complete equipment will not bring satisfaction in terms of distributed water quality over a long span if not handled by a sufficiently qualified staff, supported by an adequate organizational structure.

The comparative analyses accomplished in this study as for their managers’ socioprofessional characteristics and their organizational factors allowed identifying a number of distinctive features, some of which appeared really worth attention. Within-case analyses permitted to point out distinctive features between clusters of utilities pertaining to case-2 (nonproblematic group). Three interesting distinctive features emerged within case-2 when clusters B and C managers were compared: 1) educational background dealing with the water issue; 2) supplementary training issues relating to new QDWR; and 3) support from local authorities. As for across-case analyses, they allowed highlighting such important distinctive features as experience, awareness of and preparedness to face new challenges brought in by 2001 QDWR, and all-around support from local authorities, all of which heavily favoured case-1 utilities.

The findings of this study may be helpful for small utility managers, by allowing more perceptiveness in their daily operational practices and favouring a better understanding and awareness of their role and place in protecting public health through drinking water supply. The findings may also be helpful for municipal officials and government bodies in terms of personnel recruitment and (or) training policy making, and also in terms of better understanding and assessing of the small utilities’ specific infrastructure needs and subsequent allocation of appropriate resources.

To end, it has been identified a number of bias sources that could have affected this study’s findings as well as their interpretations. Among the potential biases, the most important are probably tied to little size of study sample (only ten utilities), and to manager interviews: what they say is not necessarily what they really know, which is a common problem in qualitative studies involving attitudes, behaviour and opinions. Nonetheless, it is reasonable to think that the argumentations developed in this paper could be useful for those interested in a better understanding of small utilities’ specificities and ways to make them serve, on a constant basis, drinking water of irreproachable quality and in sufficient quantity.

Antoun, E.N., Dyksen, J.E., and Hiltebrand, D.J. 1999. Unidirectional flushing – a powerful tool . J. Am. Water Works Assoc. 91: 62-71.

AWWA. 1994. An assessment of water distribution systems and associated research needs. American Water Works Association, Denver, Colorado.

AWWA. 1998. Water:\stats : the water utility database. American Water Works Association, Denver, Colorado.

Ball, R.O., Asce, M., and Church, R.L. 1980. Water quality indexing and scoring. Proceedings of the American Society of Civil Engineers (ASCE) 106: 757-771.

Béron, P., Valiquette, L., Patry, G., and Brière, F. 1982. Indices de qualité des eaux. Trib. Cebedeau 467: 385-391. Cebedoc Éditeur, Liège, Belgique.

Brown, R.M., McClelland, N.I., Deininger, R.A., and Tozer, R.G. 1970. A water quality index ― do we dare? Water Sewage Works 177: 339-343.

Cluis, D., and Laberge, C. 2001. Climate change and trend detection in selected rivers within the Asia-Pacific region. Water Inter. 26: 411-24.

CMHC. 1992. Urban infrastructure in Canada. Canada Mortgage and Housing Corporation, Ottawa.

Couillard, D., et Lefebvre, Y. 1986. Indice de qualité de l’eau pour détecter l’impact de la pollution diffuse urbaine. Can. J. Civ. Eng. 13: 631-638.

Coulibaly, H.D., and Rodriguez, M.J. 2003a. Portrait of Drinking Water Quality in Small Quebec Municipal Utilities. Water Qual. Res. J. Canada 38: 49-76.

Coulibaly, H.D., and Rodriguez, M.J. 2003b. Spatial and Temporal Variation of Drinking Water Quality in Ten Small Quebec Utilities. J. Environ. Eng. Sci. 2: 47-61.

Denzin, N.K. 1994. The art and politics of interpretation. In: Handbook of qualitative research. N. K. Denzin and Y. S. Lincoln (eds.), pp. 500-515. Thousand Oaks, CA: Sage.

Dunette, D.A. 1979. A geographically variable water quality index used in Oregon. J. Water Pollut. Control Fed. 51: 53-61.

Duranceau, S.J., Poole, J., and Foster, J.V. 1999. Wet-pipe fire sprinklers and water quality . J. Am. Water Works Assoc. 91: 78-90.

Fougères, D., Gaudreau, M., Hamel, P.J., Poitras, C., Sénécal, G., Trépanier, M., Vachon, N., et Veillette, R. 1998. Évaluation des besoins des municipalités québécoises en réfection et construction d’infrastructures d’eaux. INRS-Urbanisation, Montréal, 266 p.

Gouvernement du Québec. 1984. Règlement sur l’eau potable. Éditeur officiel du Québec, Québec. 7 p.

Gouvernement du Québec. 1997. L’eau potable au Québec. Un second bilan de sa qualité : 1989–1994. Ministère de l’Environnement et de la Faune, Québec. 36 p.

Gouvernement du Québec. 2001. Règlement sur la qualité de l’eau potable. Ministère de l’Environnement, Québec. 19 p.

Gouvernement du Québec 2002. Guide de conception des installations de production d’eau potable. Volumes I et II. Ministère des Affaires municipales et de la métropole, Ministère de l’environnement.

Guba, E.G., and Lincoln, Y.S. 1989. Fourth generation evaluation. Newbury Park, CA: Sage.

Huberman, A.M., and Miles, M.B. 1994. Data management and analysis methods. In: Handbook of qualitative research. N. K. Denzin and Y. S. Lincoln (eds.), pp. 428-444. Thousand Oaks, CA: Sage.

Laroux, T., et Weber, J.-L. 1994. Réflexions sur les critères de définition et de choix des indicateurs d’environnement. IFEN.

LeChevallier, M.W., Schulz, W.H., and Lee, R.G. 1990. Bacterial nutrients in drinking water . In: Assessing and controlling bacterial regrowth in distribution systems. AWWARF (ed.), pp. 143–201. American Water Works Association Research Foundation, Denver, Colorado.

Lence, B.J., and Ruszczynski, A. 2001. Managing water quality under uncertainty : application of a new stochastic branch and bound mothod. In: Risk, Reliability, Uncertainty and Robustness of Water Resources Systems. J.J. Bogardi and Z.W. Kundzewicz (eds.), pp. 143-152. International Hydrologic Series, Cambrigde University Press, Cambridge, UK.

McDonald S, Daigle L, and Félio G. 1997. Water distribution and sewage collection in Canada – assessing the condition of municipal infrastructure, results from questionnaires to Canadian municipalities. Client Report A-7016.1, Institute for Research in Construction, National Research Council Canada.

Opferman, D.J., Buchberger, S.G., and Arduini, D.J. 1995. Complying with the STWR : Ohio’s Experience. J. Am. Water Works Assoc. 87: 59–67.

Ott, W.R. 1978. Environmental indices ― theory and practice. Ann Arbor Science Publishers Inc., Ann Arbor Michigan, pp. 369.

Porcella, D.B., Petersen, S.A., and Larsen, D.P. 1980. Index to evaluate lake restoration. Proc. Am. Soc. Civil Eng., J. Env. Eng. Div. 106: 1151-1169.

Rodriguez, M.J., Coulibaly, H.D., and Banville, J. 2002. Strategies for ensuring a safe drinking water in small utilities of Quebec (Canada). Proceedings of the American Water Works Association Annual Conference and Exposition (ACE), New Orleans, La. 16-20 June.

UNEP. 1994. Overview of Environmental Indicators: State of the Art and Perspectives. United Nations: United Nations Environment Program, Nairobi.

USEPA. 1999. Disinfection profiling and benchmarking guidance manual. United States Environmental Protection Agency, EPA-815-R-99-013.

USEPA. 1999. Handbook for capacity development: developing water system capacity under the Safe Drinking Water Act as amended in 1996. United States Environmental Protection Agency, Office of Water (4606), EPA 816-R-99-012.

Villeneuve, J.P., and Hamel, P.J. 1998. Synthèse des rapports INRS-Urbanisation et INRS-Eau sur les besoins des municipalités québécoises en réfection et construction d’infrastructures d’eaux. INRS-Urbanisation, Montréal, 50 p.

Yu, J.K., and Martin, M.F. 1978. Development of a combined water quality index. Water Resources Bulletin 14: 1239-1250.

Zandbergen, P.A., and Hall, K.J. 1998. Analysis of the British Columbia water quality index for watershed managers: a case study of two small watersheds. Water Qual. Res. J. Canada 33: 519-549.