PROPOSAL OF A WATER TREATMENT PLANT QUALITY INDEX This paper focuses on developing a quality index for measuring the

results of a conventional water treatment plant. It aims to provide a useful tool that allows the effective plants comparison by means of a methodology beyond compliance with drinking water regulations. The research procedure used in formulating this WTPQI was based on the methodology used in beginning of 1970s for developing the Water Quality Index (WQI) and like this one it attempted to incorporate many aspects of the Delphi method. Afterwards, the WTPQI was applied to ten different Brazilian conventional water treatment plants which average flow rate range from 100 to 4300 L/s and all of them utilize rectangular basin as settling unit. The results pointed out the WTPQI usefulness as a plant evaluation tool. It was verified a clear tendency that plants achieving more elevated WTPQI is the same that achieving good performance in terms of filtered water turbidity. In such way the WTPQI can arise as a reliable tool to manage water supply systems in near future. Suggested keywords: Water treatment plant quality index, water treatment plant evaluation, water treatment. INTRODUCTION AND RELEVANCE Factors limiting water treatment plant performance was usually related to (i) the suitability between the raw water characteristics and the treatment process train, (ii) the ratio between the influent flow rate and the water treatment plant capability and, probably the most important, (iii) the operation accuracy. Also, the global evaluation of a plant have to join the finished water quality which is related with the dosage and the type of coagulant, the run filters, the possibility of short circuits and other factors. This multiplicity of factors has been raising many difficulties to the professionals to set up the reliable hierarchy among them. This hierarchy would define more accurately the activities of the water supply system managers in terms of operation and/or enlargement of the water

treatment plants. Consequently, the water treatment plants evaluation has been frequently done just based on the relevant compliance to the drinking water regulations. The global evaluation of the water treatment plants, and consequently the operation quality, shows some important questions: i) For plants with same treatment process and producing finished water in compliance with drinking water regulations, how is possible to evaluate the plant performance or, in the similar way, to compare the operation costs when the raw water quality is distinct? ii) The overflow index, in terms of some hydraulic parameters such as the flocculation time, the settling velocity and the rate filtration, is or not more relevant than operation accuracy to get the finished water quality? iii) Is it correct to compare plants with same treatment process, and similar raw and finished water quality if the operation costs are different? iv) Which processes must be prioritized in terms of the enlargement of water treatment plants? v) Is there a remarkable distinction between the operation quality among plants operated by the same company in regions with different social and economical levels? It is still possible to insert in these questions the environmental impact evaluation on the water sources by the discharge of the sludge from sedimentation basins and the backwash water from the filters. In such way, there is a necessity of a methodology to permit the managers of water supply systems realize a truthful analysis of the water treatment plants to define in a best way the wisest application of the available financial support and human resources. The perspective to develop a methodology for evaluation of water treatment plants is not a new idea. Renner et al. (1993) presented a discussion concerning the main results of the implementation of an optimization program of treatment processes, named CPE (Comprehensive Evaluation Program), in 36 plants in USA. This program identified some limitations in the performance of 29 plants, in function of operational problems, and the real possibility to optimize these units

without large investments also. The option of a quality index was justified by the use of easy available data related to the operational routine of the plant. It will permit a good comprehension by the population (as WQI does), because in many situations the responsible for the application of financial resources does not have a clear knowledge about the processes concerning the water treatment. The mentioned comprehension by the public will help the use of the WTPQI as a tool for the population consciousness by the relevance of a good performance of the water treatment plants, minimizing in a second instance the outbreak risks thru the drinking water. OBJECTIVE The paper proposes a Water Treatment Plant Quality Index (WTPQI) as an evaluation tool for the water supply system administrations become the comparisons among different plants more precise. Additionally, the paper proposes: i) to list the intervenient parameters on the performance of water treatment plants; ii) to define an hierarchy for these parameters according to their role in performance of the plants; iii) to validate the WTPQI basing on the daily operational data of ten conventional water treatment plants with different sizes.

LITERATURE REVIEW Optimization and Evaluation of Water Treatment Plants The water treatment for human consumption, as one of most important features of sanitary engineering, has been facing a clear dichotomy last years. The successive drinking water regulations have been presenting more restrictive in terms of the number of parameters and their maximum levels. On the other side, there is a progressive deterioration of the natural water quality by means concentrated and diffuse pollution mainly as a consequence of human activities. In a first moment more restrict levels to the filtered water turbidity were focused on the higher efficiency of the chlorination in inactivation of pathogenic microorganism, and in a second phase toward the perspective to increase the protozoa removal. In this last context, many researches have emphasized a higher removal associated to finished water with turbidity lower than 0,1 NTU. As an example, a research was carried out with some filters in pilot and actual scale monitored during two years. It was demonstrated the more consistent Giardia and Crypto removal was reached with low filtered water turbidity (0,1 to 0,2 NTU), despite the determination coefficient was not high (r2 = 0,64). Furthermore, when the performance of water treatment plant varied with the fluctuations of raw water quality a high variability in cyst concentration was observed in the effluent (NIEMINSKI & ONGERTH, 1995, apud LECHEVALIER & AI, 2004). The development of a methodology to optimize the water treatment plants began in USA and Canada in the end of 1980s with the objective to increase the protection against some pathogenic microorganisms. Named Composite Correction Program (CCP), some objectives of this methodology were to define the best performance of sedimentation, filtration and disinfection processes. There was established the highest settled water and filtered water turbidity values of 2 NTU and 0,1 NTU, respectively, with a permissible peak after backwashing of 0,3 NTU by less than 15 minutes. The CCP optimization concepts were expanded to many other activities. The program Partnership for safe water employed CCP such as

a basis to the development of its Phase III with the objective to improve the treatment for a better water quality. This program was developed by the association of six American entities, and in May 1988 217 water treatment plants supplying more than 90 million people were taking part of it (GUIDELINES FOR PHASE IV, 2003) In 1984 the DEP (Department of Environmental Protection) of Pennsylvania, with the objective to assure the distributed water quality, made a start the implementation of the FPPE program (Filter Plant Performance Evaluation) aiming to determine the plant effectiveness in terms of the particle removal at same range size of the cysts and oocysts of protozoa. Until 1996 290 plants were evaluated and in 1988 more than 60 % were producing effluent with turbidity higher than 0,2 NTU. In 1996 this percentage was reduced to 4 %. Afterwards some CCP concepts were inserted in the FPPE program. There was done the capacity evaluation of each water treatment plant with a current use of standard sheets to obtain temporal series of raw, settled and filtered waters. By means the comparison of these graphs there was possible to assess the plant capacity to produce better quality water despite the changes of raw water (CONSONERY et al., 1997). The Delphi Methodology The Delphi method concept can be understood as a product of a Rand Corporation project in 1950s, concerning the application of the opinion of specialists. It can be developed in two different techniques. The more common is the pen and paper version. In this situation, a monitor elaborates a questionnaire that is sent to a group of respondents. When these questionnaires return, the monitor summarizes the results and basing on them develops a new one. The group of respondents has at least one chance to change his opinion. This technique is known conventional Delphi. The other one, named Delphi conference, the monitor is substituted by a computer program, which makes the printout simultaneously and returns the responses to the respondents. After the last response, the software makes a report and the new questionnaire.

This method has an advantage to carry out the process in real time. In both ways, some characteristics define the method: (i) the anonymity; (ii) the interaction; (iii) the feedback; (iv) the statistical representation of the results (LINSTONE & TUROFF, 1975). Taylor & Ryder (2003) utilized the Delphi methodology to define a management plan of 25 multiple uses reservoirs. This information concerned basically the necessary water levels to guarantee the survival of fishes. Questionnaires were elaborated for each reservoir and were sent to 26 specialists and the number of respondents by reservoir varied from 2 to 8. It was possible to the same specialist answered questionnaire related to more than one reservoir. The first questionnaire asked them about a list of the more vulnerable species and the period in which each species was particularly sensible to the variations of the reservoir level. In the second questionnaire, the specialists reevaluated their responses in function of the opinion of the entire group. The research has gotten a return of 85 % and a high convergence of opinions for all reservoirs. The research showed the applicability of Delphi methodology to deal with several information to the management of complex environmental questions. The index development The transmission in an intelligible way to the population the data and the parameters of water treatment plants is not an easy task. However, it is not a question restricted to this specific area. There were many efforts trying to reproduce in only value the meaning of a data set. Brown et al. (1970) employed the Delphi methodology to develop the Water Quality Index (WQI) based on the opinion of a group of 142 water quality specialists. This research was composed by three questionnaires. In the first a list with 35 parameters, randomly selected, was sent to the group. For each parameter, the respondents had to choose among three options: include, no include and undecided. There was possible to include other parameters that were absent in this first list. The respondent had to assign values for each parameter selected as include from 1 to 5.

The results of this first round were sent to the respondents with the second questionnaire opening the possibility to the respondents compare their responses with the other ones, and, occasionally, reevaluate them. There was a request of a list including the 15 most important parameters. In the third questionnaire, for each one of the 9 selected parameters, the respondent had to draw a curve as, in his judgment, the best way to represent the influence of this parameter in water quality. The nine average curves employed to define the WQI were a combination of the responses of all respondents. Among 142 specialist invited in the first round, 94 (66 %) returned the first questionnaire in time to take part in the second round, and from this group 82 % completed and returned the second questionnaire. The WQI value was defined upon a sum represented by the Equation 1:
WQI = wi qi
i =1 n

(1)

in which: WQI: the water quality index, a number between 0 and 100; wi: the unit weight of ith parameter, a number between 0 and 1; qi: the quality of the ith parameter, a number between 1 and 100, extracted from the respective curve; n: number of parameters. Based on the same methodology employed in the development of the WQI, Nages et al. (2001) proposed an index system to assess the recreation water quality in New Zealand. They used the Delphi methodology to resume the judgment of 18 specialists from consulting engineering, environmental management companies, research institutes and universities. The remarkable new in this research was the final definition of the index. Distinctly of the WQI, there were not established weights for each parameter and the index of a specific water source will be the lowest value extracted from these curves. The justification was an aggregation of many individual scores could hide a low value of a specific parameter. METHODOLOGY
7

The present work can be defined such as an applied research with the objective to provide a quantity tool to help the water supply system managers, and a quality one to classify the performance of the water treatment plants by means numerical methods. The universe of this research was limited by the conventional water treatment plants, with horizontal sedimentation basins, treating typical raw water to produce filtered water turbidity lower than 0,5 NTU and absence of total coliform in compliance to the Brazilian Drinking Water Standards. The methodology to formulation the WTPQI was based on the same utilized to the development of the WQI. In such way, after the establishment of all parameters, and respective weights and grading criteria, there were defined two different formulations in terms of a summation and a multiplicative forms:
WTPQI
N n = Wj Qj 1 i= j = 1 i N n WTPQI = QWj 1 j i =1 j =

(2)

(3)

where: Wj: weight rated to each parameter established by the judgment of the specialists; Qj: value rated to the water treatment plant for each parameter selected according to the developed criterion; j: each parameter included in the index; i: each group of parameters to comprise the index such as rapid mix, flocculation, factors; n: number of parameters included in the index; N: total number of groups of parameters that will constitute the index. The methodology to the development of the WTPQI was divided in three phases (opinion research, definition of the grading criteria, and the index validation) as follow. Opinion research There was carried out an opinion research to select the intervenient parameters to be included in the WTPQI, and the respective weights, with 18 professionals with expertise in water treatment. Of the total panel, 16 completed and returned both questionnaires. The group was selected
8

sedimentation,

filtration,

disinfection,

and

operational

focusing different professional formations and distinct geographical areas of Brazil. This research was constituted in two different phases according to the Delphi characteristics. After the literature review, there was elaborated a first list of the intervenient parameters in water treatment as shown in Table 1.

TABLE 1- Parameters included in the first questionnaire

GRM TRM Jtest Gf Tf Gp Nc Vc Gps Gin Vs Rapid Mix Velocity Gradient Rapid Mix Detention Time Routine Jar Test Realization Flocculation Velocity Gradient Flocculation Time Velocity gradient thru the ports of flocculator Number of compartments Average velocity in the flocculated water channel Velocity gradient thru the ports of sedimentation basin Velocity gradient thru the inlet baffle of sedimentation basin Sedimentation Surface Loading Rate or Terminal Settling Velocity VF QL Tfilt Drf Exp Vupf Lair Lwate
r

Flow-through Velocity Weir Loading Rate Filtration Rate Run Filter Filter bed expansion Upflow water wash velocity Washing with auxiliary air scour Washing with auxiliary surface water system Detention time in the clearwell Number of compartments of the clearwell Instruction level of the operational staff

Tc Ncl IL

This list was utilized in the elaboration of the first questionnaire sent to 18 professionals selected. The panel was composed by graduate professionals responsible by researches in water treatment, designs and operation of water treatment plants, regarding universities, sanitation companies and consulting engineering of six Brazilian states in the two most developed and populous regions like showed in Table 2. TABLE 2: Professional fields of participants in the panel
Plant operator Designer Researcher Researcher/Desig ner Designer/Operato r Researcher/Opera tor 2 4 7 1 1 1

The first questionnaire was divided in three parts. The first one has presented an introduction explaining all phases of research and showing to the participant his role in it. The second part has explicated all instructions for a correct filling of the questionnaire. Finally, the third part was constituted by the initial list (Table 1) of the parameters which the respondent would have to evaluate by one the categories include, no include and undecided, and he could suggest additional parameters absent in the first list. After his judgment, the respondent would have to rate (up to 100) only those parameters marked include according to their relevance to water treatment.

10

After the finish of the first phase, there was elaborated a report with numerical summary of the responses of all participants. It was incorporated in this report the inclusion percentage of each parameter, average, median, mode, quartiles, and an abstract of the commentaries, the participants responses and a column to review their initial responses. After an evaluation of the opinion of the entire group, the respondents were asked to review their responses, keeping or modifying them. The parameters included in the index were divided into six groups according to the step of the conventional treatment such as, per example, Rapid mix, Flocculation, Sedimentation, Filtration, Disinfection, and Operational quality. Based on the weights rated to the parameters by the panel it was determined the weight of each group in function of the treatment effectiveness. The main reason for the separation in groups was the possibility to comprise a complete index, formed in function of the indexes of each step of the water treatment. In such way, it will be possible to identify which group is responsible by an eventual low grade of the water treatment plant. Development of the grading criteria After the definition of the parameters included in the index and their respective weights, the following phase of the research was begun. In this phase it was established the grading criteria based on the premises set up by the Brazilian Technical Standards Association (1990) and the literature. Validation of the WTPQI The final phase of the research was composed by a comparative study between the final grade provided by the index to a specific water treatment plant and the monitoring data in terms of filtered water turbidity. The scope of this last phase was to choose the final formulation of the index (summation or multiplicative), and to verify the validation of the grade provided by the index to the treatment. In other words, whether the water treatment plant evaluated with a high WTPQI had presented a good performance concerning the filtered water turbidity. With this objective, the developed index, in summation and

11

multiplicative forms, was applied to ten conventional water treatment plants of two most important Brazilian states with average flow rate range from 100 to 4300 L/s. These plants were selected according to the easiness of access and the reliability of the monitoring data provided by the respective directions. The comparison between the WTPQI and the filtered water turbidity was done with tables in which were presented the day, the WTPQI (summation and multiplicative), and the mean values of filtered water turbidity for six months data of 2003 and 2004 (three months related to the drought season and three months to the rainy season). There were calculated the following values for each season according to the Brazilian and American Drinking Water Standards: time percentage of the operation plant with filtered water turbidty 0,5 NTU; time percentage of the operation plant with filtered water turbidty 0,3 NTU; time percentage of the operation plant with filtered water turbidty 0,1 NTU; value lower than 95% of the filtered water turbidity values. For verifying whether the WTPQI was correlated to the filtered water turbidity values, there were calculated linear and non-linear correlation coefficients. This analysis focused to assess whether an occasional reduction of the WTPQI was followed by higher filtered water turbidity levels. The last analysis was based on the premise about a plant with good performance probably will produce a high water quality even when a variable raw water quality as influent, or the finished water turbidity will not change with occasional alterations of raw water characteristics. For this, there were calculated the correlation coefficients (r) among raw, settled and finished waters for each plant. RESULTS AND DISCUSSION Participants responses and the definition of the weights for each

12

parameter The first round of the research was carried out from January to April 2004 with 89 % of returning among 18 questionnaires. The justifications and comments from the first questionnaire were sent to the respondents in the second round with the objective to show the opinions of the other participants. No parameter listed in the first questionnaire (Table 1) was dismissed, and among the parameters suggested by the respondents none was inserted in the second round because the necessary information about them was not easily accessible or the parameter was very subjective. Per example, there were some suggestions in terms of general situation of the laboratory, plant versatility, and others. Other kinds of suggestions concerning the raw water quality were not accepted because this index focused to evaluate the treatment despite the raw water characteristics. Beyond this fact, none suggested by more than three respondents. As previously mentioned, the respondents were instructed to assign values from 0 to 100 for only the selected parameters marked include. This rating system was chosen to become easier the filling of the questionnaires because. However it was relevant the relative importance of each parameter and the weight assigned by the respondent in a proportion with the total points distributed by him. Therefore, the final score of each parameter was divided from the total points distributed by the respondent and the sum of all distributed points have totalized 100. It is shown in Figure 1 the relative importance of each parameter in terms of global performance of the water treatment plants. According to the panel, the rate filtration is the most relevant parameter, answering by approximate 9 % of the performance, agreeing with the tendency of the national and international water drinking standards to reduce the finished water turbidity. The settling velocity and the flocculation velocity gradient, with the filtration rate, were responsible by 23 % of the plant efficiency. The first parameter represents the assurance of the settling of flocs and the second the suitable formation of them. new parameters were

13

Jtest

Vupf

IL

GMR

Gps

Exp

Gin

Gp

QL

Vs

TMR

Vc

LAIR

Tc

Nc

FIGURE 1: Accumulated median of the weights of each parameter By the observation of the vertical bars shown in Figure 1, it is possible to verify the highest ranges among the responses to the most significant parameters, emphasizing the agreement of the respondents concerning the relevance of them, but a clear disagreement in terms of the weight to be assigned. Based on these results, the weights for the WTPQI determination would have to be defined. In this context, what would be the best way for this scatter of results? Maybe other rounds could help to reach a higher convergence, but this option was not suitable in function of the time expended in each round. Moreover, it should be noted there was a little change expressed in the second questionnaire and several respondents did not modify their scores. In such way, two important decisions were taken to the definition of the final weights: (i) avoiding the influence of extreme points, the median was chosen as the best measurement of the group opinion; (ii) all parameter were included and the weights were multiplied by the inclusion rate of each parameter, so the parameters with 100 % inclusion rate had their weights were kept, and the others had theirs reduced. For the development of grading criteria some parameters were unified. As an example, the Tf and Gf parameters were rated as a couple and the weight is the sum of each one. The Lair and Lwater parameters were transformed in only parameter named Laux (auxiliary wash), the weight of it was defined as a median of all weights assigned to them. In the same
14

LWATER

Drf

VF

Gf

Tf

TFILT

Ncl

1 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0

,0 ,0 ,9 ,9 ,8 ,8 ,7 ,7 ,6 ,6 ,5 ,5 ,4 ,4 ,3 ,3 ,2 ,2 ,1 ,1 ,0 ,0

5 0 5 0 5 0 5 0 5 0 5 0 5 0 5 0 5 0 5 0 5 0

context, it was defined the group of Exp and Vupf parameters because the inclusion of both would be overrating the same aspect related to the bed filter wash. Finally, the last transformation was carried out dividing each weight from total score, for all weights sum was 1. The final weight of each parameter is shown in Table 3.

15

TABLE 3: Final weights of all parameters

Group Rapid Mix Flocculati on Parame ter GMR TMR Gf-Tf Gp Nc Vc Gps Gin Vs VF QL Wei ght 0,06 0,03 0,14 0,04 0,03 0,03 0,04 0,05 0,08 0,04 0,06 Group Filtratio n Disinfect ion Operatio n Param eter Tfilt Drp Exp Laux Tc Nl Jtest IL Weig ht 0,09 0,04 0,04 0,03 0,05 0,02 0,07 0,06

Sediment ation

Development of grading criteria After the definition of weights for each parameter there was necessity to establish the grading criteria. In reality, this definition is substituting the mentioned curves drawn by the respondents for the WQI. In this phase, as explicated in the methodology, the parameters were divided into six groups: Rapid mix, Flocculation, Sedimentation, Filtration, Disinfection and Operation. Due to the limit to the size of the paper, there will be detailed, among 19 parameters, only the settling velocity (surface loading rate) integrating of the group Sedimentation. The function of this step of treatment is to remove by gravity the flocs to lower the solids concentration on filters. Among the intervenient factors to the sedimentation effectiveness the more important are the settling velocity, the inlet and outlet arrangements, and sludge removal. The high variability of size, density, and particle shape has been presenting difficulty to develop a mathematical model for flocculant settling. In such way, the ideal horizontal-flow sedimentation basin, even its simplicity, was applied to estimate the particle behavior. Some simple suppositions characterize this model: (i) in sedimentation zone the particles settle in analogous way as in rest tank with the same depth; (ii) the flow and particle concentration are uniform all over the transversal section; (iii) there is not scouring when the particles reach the sludge zone. The grading criterion was defined based on the rate Vs/Vs,

16

considering Vs the surface loading rate with a progressive increasing of 5 % from the design rate on and Vs was considered the highest surface loading rate established by The Brazilian Technical Standards Association (40 m3/m2.day). Of course, all plants which sedimentation basins were operating with surface loading rates lower than that received the maximum grade. In a real sedimentation basin the terminal settling velocity of the particles tends to increase in function of the differential settling. In this way the sedimentation effectiveness will be higher than that estimated by the ideal model. In contrast with differential settling, the wind effects, the different temperatures, the currents as a result of distinct densities, and other factors caused short-circuits, floc rupture and scouring of settled sludge reducing the sedimentation efficiency. Due the difficulty to synthesize the influence of these factors and considering the positive effects of differential settling can compensate these negative effects, for the Vs grading criterion, shown in Figure 2, it was utilized the ideal model already described.

Vs
100 90 80 70

Score

60 50 40 30 20 10 0 0 10 20 30 40 50 60 70 80 90 100 110 120

Vs (m3/m2.d) 1000m3/day 1000 < Cap < 10000 m3/day >10000m3/day

FIGURE 2: Grading criterion established for Vs WTPQI Application Finally, after the definition of the weights and the grading criteria the final phase of the research was to apply the WTPQI, in multiplicative

17

and summation forms (equations 2 and 3), to ten water treatment plants. The plants were not identified because the focus of this research was to evaluate the applicability of the WTPQI, and there was no intention to assess the plant performance which managers permit the free access to the plant data. As previously mentioned, there were utilized six months data, in terms of drought and rainy seasons for three months each. As an example, the WTPQI was calculated and presented in Table 4.

18

TABLE 4: Determination of the WTPQI for one of the plants of the sample
Paramete r GMR (s-1) TMR (s) GF (s-1) TF (s) Flocculato r type Gp (s-1) Nc Vc (m/s) Gpss-1) Gin (s-1) Vs (cm/min) VF (cm/s) QL (L/s.m) Filter type Tfilt (m/dia) Drf Filtered volume (m3/m2/run ) Washwate r volume (m3/m2/filt er) Exp (%) Laux Residual Cl (mg/L) T (min) pH Disinfecti on Ncl Daily data median for each parameter in six months 1340,65 0,40 35,18 835,87 Hydraulic 55,11 5 0,29 19,29 8,19 3,02 0,26 1,64 Downflow Dual media 279,66 1 60 100 100 100 80 100 100 100 0,04 0,03 0,03 0,04 0,05 0,08 0,04 0,06 0,09 0,04 1,80 3,00 4,00 5,00 6,40 4,00 6,00 9,00 1,00 1,13 1,15 1,20 1,26 1,42 1,20 1,32 1,51 25,24 3,40 Sco re 100 100 30 Weig hts 0,06 0,03 0,14 Weigh ts x Score 6,00 3,00 4,20 Weig hts ^ Score 1,32 1,15 1,61 9,04 2,09 Scor e by grou p (Sum ) 9,00 Scor e by grou p (Mul) 1,52

Group

Rapid Mix

Flocculati on

Sediment ation

475,30 100 6,06 32 Surface wash (scrapping) 0,87 1,47 7 The water was introduced in the bottom of the tank and, after the chlorine dosage, flowed over a chicane Superior According to the change of raw water turbidity 10 0,05 0,5 1,12 100 1 0,04 0,03 4,00 0,03 1,20 1 0,04 4,00 1,20 17,03 2,17

Filtration

1,5 50 0,02 1 1,08

1,21

IL Operatio n WTPQI Jar test

100 100

0,07 0,06

7,00 6,00

1,38 1,32 13 74,8 1 1,82 51,6 2

19

After the WTPQI determination, the main point was the final definition by multiplicative or summation forms for the WTPQI equation. The distinction between both equations was focused in the possibility of a plant with a low grade, in a specific parameter, has its final score more significantly affected when the multiplicative form is utilized. The plants with a more uniform grading among the parameters keep the final score approximately constant by both equations. The question, which answer intends to reach in the following analysis, is: only parameter with low score will have a significant impact in the global performance of the plant is able to justify the option by the multiplicative form of the WTPQI equation? For evaluating the applicability of the index, its values were compared to filtered water turbidity. This comparison was done by means scatter graphs determining the correlation between The WTPQI (in both forms), and the percentage of daily mean values of filtered water lower than 0,5 and 0,3 NTU. Also, there were made graphs in terms of the WTPQI and the turbidity value higher than 95 % filtered water turbidity values. For the last validation of the WTPQI, there was accomplished another scatter graph concerning the index and the daily filtered water turbidiy. With the exception this last graph, all ones were divided in drought and rainy season. The correlation was exploited to evaluate the association degree between the index and the filtered water turbidity percentage lower than a previous established value. The filtered water turbidity was selected as the mark to assess the WTPQI applicability because, besides to be a parameter of treatment effectiveness, all plants have been monitoring it. For a better comparison among the obtained values, they were organized in Table 5, outstanding the more significant results (R2), and also including the correlation with the percentage below 0,7 NTU, as an intermediary between the maximum level (1,0 NTU) and the recommended 0,5 NTU. For the rainy season, it is possible to observe that the multiplicative form (Equation 3) presented more significant results when compared with the filtered water turbidity percentage lower than

20

0,1 and 0,3 NTU, and with the value higher than 95 % of turbidity data. On the other hand, the summation form (Equation 2) has showed more significance for the percentage lower than 0,5 and 0,7 NTU. However, in the drought season, despite the better results with summation form, the WTPQI values were correlated with high p values. This fact indicates a no correlation among these variables. Table 5 Linear correlation coefficients (R)
<= 0,1 NTU WTPQI Summation WTPQI Multiplicative R R2 p R R2 p 0,37 0,14 0,37 0,56 0,31 0,15 Percentage of filtered water turbidity values below Rainy Season <=0, <= 0,5 <=0,7 95% <= 0,1 3 NTU NTU NTU NTU 0,55 0,80 0,85 -0,62 -0,05 0,30 0,63 0,73 0,38 0,00 0,16 0,02 0,01 0,10 0,90 0,85 0,73 0,69 -0,86 0,38 0,73 0,53 0,48 0,74 0,14 0,01 0,04 0,06 0,01 0,31 the established limit Drought Season <=0, <= 0,5 <=0, 3 NTU 7 NTU NTU 0,33 0,47 0,48 0,11 0,22 0,23 0,39 0,21 0,20 0,21 0,25 0,24 0,04 0,06 0,06 0,60 0,51 0,53 95% -0,50 0,25 0,17 -0,23 0,05 0,55

As the R coefficient just measures the linear correlation among the variables, another coefficient was also calculated named Kendall Tau () presented in Table 6. This coefficient assesses too the no-linear correlation between two variables, besides it is less susceptible to the outliers. The values presented in Table 6 demonstrated better correlation for the WTPQI multiplicative for both seasons, and in the drought season with better significance levels. In the drought season, despite higher r value for summation form, it was observed high p values for both WTPQI forms recommending the rejection of the linear correlation among the tested variables. Despite this fact, the WTPQI multiplicative was more significant for demonstrating the existence of a no-linear correlation among those variables. Nevertheless, the best results for the WTPQI summation in the drought season does not invalid the first conclusion, that is, the WTPQI multiplicative demonstrated more suitability in the rainy season, in which there are more difficulties to keep the filtered water quality. Also, retaking the same question elaborated in the beginning of this discussion: in terms of removal turbidity, is only parameter with a low grade able to confer a noteworthy impact in the global performance of the plant to justify the option by the WTPQI multiplicative? In principle,
21

the answer is yes. Based on this analysis it is supposed that only parameter, such as Vs or Gcor, can affect the plant performance as much as to decide by a final low grading even all other parameters are suitable. This decision was confirmed by the highest correlation values for the WTPQI multiplicative when more restrict filtered water turbidity standards were utilized. Afterwards, it was evaluated whether establishing of goals to the filtered water turbidity was inducing to a correlation between then and the WTPQI. In this context, there were elaborated scatter graphs between the daily WTPQI values and the daily average of filtered water turbidity, for ten plants with six months data. It was characterized a correlation between then, a little higher to the summation form for the linear correlation (r) and the same to the multiplicative form in terms of the nolinear correlation (Table 7). Table 6: Correlation between filtered water turbidity and the WTPQI for all data
WTPQI - S WTPQI - M R -0,47 -0,39 r2 0,22 0,15 -0,37 -0,42

With the finality to evaluate the possible correlation between the raw water and settled water turbidity, and between the settled and filtered water turbidity, the linear correlation coefficients were calculated for all plants which data were available, as shown in Table 8.

22

Table 7: Correlation between turbidity values of Raw/Settled Water and Settled/Filtered Water
Raw Water/Settled Raw Water/Filtered Water Water R r2 R r2 WTP 1 0,557 0,310 0,130 0,017 WTP 3 0,571 0,326 WTP 4 0,778 0,605 WTP 5 0,527 0,278 0,636 0,405 WTP 6 0,561 0,314 0,475 0,226 WTP 7 0,451 0,204 0,399 0,159 WTP 8 0,754 0,569 0,642 0,412 WTP 9.A 0,714 0,510 0,652 0,425 WTP 9.B 0,532 0,283 0,783 0,613 WTP 9 B and A Before and after enlargement Plants

The plants with good performance have to be able to produce constant quality filtered water independently of the raw water quality. In function of this premise and the Table 8 data, the WTP 7 presented the best results for sedimentation and the WTP 2 the same for filtration. On the other hand, the WTP 4 and the WTP 9B presented the worst performance for sedimentation and filtration, respectively. In this context, another question arose. Do the water treatment plants producing filtered water with a regular quality have the higher WTPQI values? Trying to solve this question, there were elaborated the graphs shown in Figure 3, with the WTPQI (summation and multiplicative forms) on the horizontal axis and the r2 values (obtained for settled and filtered water) on the vertical axis.
r 2 ( s e t t le d t u r b id i ty / f il te r e d t u r b id i ty ) x W T P Q I 0 ,7

E T A IX .B

r 2 ( s e t tl e d tu r b i d it y /f il te r e d t u b i d it ) x W T P Q I 0 ,5 0 0 ,4 5

E T A IX .B
0 ,6

E T A IX .A
0 ,5

E T A V III
0 ,4 0 0 ,3 5

E T A V

E T A IX .A
0 ,4 0 ,3

E T A VE T A V II I

r2

E T A V I
0 ,2 0 ,1

r2 0 ,3 0

E T A V II
0 ,2 5 0 ,2 0

E T A V I

E T A II
0 ,0

E T A V II
0 ,1 5

-0 ,1

76

78

80

82

84

86

88

90

92

94

64

65

66

67

68

69

70

71

72

73

74

75

76

77

78

W T P Q I S u m m a t io n W T P Q I S u m m a t io n : r 2 : r 2 = 0 , 0 2 8 7 ; r = 0 , 1 6 9 3 , p = 0 , 7 1 6 7

W T P Q I M u lt ip li c a t i v e fo r m W T P Q I M u l ti p l ic a t iv e f o r m : r 2 : r 2 = 0 , 1 9 7 1 ; r = - 0 ,4 4 3 9 , p = 0 , 3 1 8 4

FIGURE 3: Scatter plots r x WTPQI for summation and multiplicative forms (rainy season)
2

E T A II

It was observed in these graphs that the affirmative answer for the previous question was completely rejected for the WTPQI summation.
23

Despite its low magnitude, for the multiplicative form, the results were significantly better than those obtained by the summation form. The supremacy of the WTPQI multiplicative can be confirmed comparing it with the correlation between settled and filtered water turbidity. In other words, the plants with the lowest WTPQI multiplicative values presented too the lowest correlation between settled and filtered water turbidity. Also, the WTPQI usefulness was verified by the comparison of the results obtained by the WTP B and WTP A, demonstrating the index sensibility to the plant improvements followed by the enhancing of filtered water quality. CONCLUSIONS It was verified that two rounds were not enough to reach a higher consensus among the respondents for a definition of the weights mainly to the more relevant parameters. For some parameters the dispersion increased after the second round. Concerning the parameters hierarchy, the questionnaires demonstrated an evident consensus about the more relevant to the treatment effectiveness. According to a tendency of the national and international standards, which have been emphasizing progressively the reduction of filtered water turbidity, the rate filtration was chosen the most relevant parameter by the panel. Despite some limitation in function of the sample size of ten water treatment plants, the significant correlations pointed out a tendency of the plants producing good filtered water quality usually present high WTPQI values. The higher correlation values were presented to the WTPQI multiplicative when more restrict standards were established. However, for plants with high scores in all parameters the final WTPQI is approximately the same for both formulations. Based on the premise of plants with good performance must be able to produce regularly high filtered water quality, despite the changes of raw water, the WTPQI multiplicative was more efficient than the WTPQI summation. It was evident when the comparison was made in terms of r2 values between settled and filtered water turbidity, showing more sensible and able to classify the plants, conferring better scores to those

24

with a more accurate performance. Finally, the correlations confirmed the principles in which the grading criteria for 19 parameters were based on, and the WTPQI multiplicative as a good indicator to make the plants hierarchy. In this way, the WTPQI multiplicative may be an interesting tool to the water supply system administrations. RECOMMENDATIONS Evidently, a more complete analysis has to involve, besides the WTPQI and the finished water quality, an index for the raw water in terms of its higher or lower treatment feasibility. New researches may improve the WTPQI increasing its accuracy, basing on the disagreements arose in these two rounds. Also, a research opinion about the developed grading criteria may contribute significantly to the better index accurateness. Finally, the WTPQI application to a higher number of plants would be useful to confirm, or not, this tendency. REFERENCES BRAZILIAN TECHNICAL STANDARDS ASSOCIATION (ABNT) Water treatment plants design, NBR 12216, Rio de Janeiro, 1990. (in Portuguese) BROWN, R. M.; MCCLELLAND, N. I.; DEINIGER, R. A. & TOZER, R. G. - A water quality index do we dare?, Water & Sewage Works, Chicago, v 117, n.10, p.339-43, October 1970. CONSONERY, P. J.; GREENFIELD, D N. & LEE, J. J. - Pennsylvanias filtration evaluation program, JAWWA, v.89, n.8, p. 67-77, August 1997. GUIDELINES FOR PHASE IV: Partnership for safe water, AWWA et al, 2003. HELSEL, D. R. & HIRSCH, R. M. - Statistical methods in water resources, U.S. Geological survey, 503 p., 2002. LINSTONE, H. A. & TUROFF, M. - The Delphi Method: techniques and applications, Addison-Wesley Publishing Company: Massachusetts, USA, 620 p., 1975. HEALTH MINISTRIO DA SADE MS. Portaria 518: Normas e padro de potabilidade maro.2004. da gua destinada ao consumo humano. Braslia,

25

NIEMINSKI, E. C. & ONGERTH, J. E. Removing Giardia and Cryptosporidium by conventional treatment and direct filtration. JAWWA, New York, v. 87, n. 9, p. 96-106, September1995 apud LeCHEVALIER, M. W. & AI, K. Water Treatment and Pathogen Control, WHO, London, 112 p., 2004. RENNER, R. C. et al. - Composite Correction Program Optimizes Performance at Water Plants. JAWWA, v.85. , n. 6, p.67-74, June 1993. TAYLOR, J. G. & RYDER, S. D. - Use of the Delphi method in resolving complex water resources issues. Journal of the Americam Water Resources Association, New York, v. 39, n.1, p. 183-189, February 2003. USEPA. Optimizing water treatment plant performance using composite correction program.Cincinnatti, Ohio : USEPA CERI.1998.168p.( EPA/625/387/013)

26