Gary Logsdon, Director, Water Process Research

Black & Veatch, Cincinnati

Alan Hess, Senior Project Manager

Black & Veatch, Philadelphia

Patrick Moorman, Instrumentation Specialist

Black & Veatch, Kansas City

David Koch

Black & Veatch, Detroit

Michael Chipps, Principal Research Scientist

Thames Water Utilities, Reading, U.K.

The U.S. EPA promulgated the Interim Enhanced Surface Water Treatment Rule (IESWTR) on December 16, 1998 in the Federal Register, Vol. 63, No. 241, pp. 69478-69521. This rule will take effect on December 17, 2001. Essential features of the IESWTR were reviewed in the Journal AWWA by Pontius (1999). Water systems treating surface water or groundwater under the influence of surface water and serving 10,000 or more persons will have to comply with the IESWTR. The principal aspects of this rule that will affect filter operation and maintenance are the tightening of the filtered water turbidity performance standard and the requirement for monitoring of turbidity of individual filters.

When the IESWTR takes effect, conventional filtration plants and direct filtration plants (filtration plants in which chemical coagulation is employed) must produce filtered water that is 0.3 ntu or lower in 95 percent of samples taken in finished water sent to the distribution system. In addition, turbidity of the water produced by each filter must be measured with an on-line turbidimeter and the results recorded every 15 minutes. Turbidity excursions must be documented and reported, and depending on the severity of the excursion, the water utility would have to undertake different levels of action. Turbidity events and consequences are:

• > 1.0 ntu in 2 measurements at 15 minute intervals; produce a filter turbidity profile
• > 0.5 ntu in 2 measurements at 15 minute intervals after 4 hours of operation; produce a filter turbidity profile
• > 1.0 ntu in 2 measurements at 15 minute intervals in each of 3 consecutive months; prepare a filter profile and prepare a filter self-assessment report
• > 2.0 ntu in 2 measurements at 15 minute intervals in each of 2 consecutive months; arrange for a comprehensive performance evaluation.

The IESWTR includes other provisions, but these are the ones that will have the greatest impact on the filtration process, monitoring, and associated record keeping.

More and more reliance is being placed on instrumentation and electronics in the monitoring and operation of water filtration plants. Decisions on water treatment are being made on the basis of instrument read-outs or electrical signals transmitted from a remote sensor to a computer screen or data panel in the control room or a similar place in the treatment plant. Carrying out an effective quality control program is the key to being able to trust the instruments and the information they are providing.

On-line turbidimeters must be installed in a manner that enables them to measure water samples that are representative of the full flow of the water being sampled. The best sources of guidance for filtered water turbidimeter installation are the manufacturers’ manuals and A Practical Guide to Particle Counting (Hargesheimer and Lewis, 1995). The latter is an important source of information for two reasons. First, particle counters are more sensitive than turbidimeters so some of their requirements are more stringent. Second, the number of utilities installing particle counters is increasing, and if turbidity data and particle count data are to be compared, common water samples should be analyzed. To do this, a flow stream from one sample tap would need to be split and routed to a turbidimeter and to a particle counter, or the two sample streams would need to be extracted in close proximity, using similar sample taps.

Properly locating the sample tap for an on-line turbidimeter is important. The tap in the header coming from the filter box should be placed as far upstream from bends or flow control valves as practical. Avoid placing taps near unusual turbulent flow patterns. Hargesheimer and Lewis did not detect any difference in particle counts for sample tubes mounted on the top, side, or bottom of the pipe, but turbidimeter manufacturers do make specific recommendations. Hach manuals (Hach, 1997 and Hach, 1999) recommend installation of the sample tap in the side of the pipe, at the midpoint between top and bottom, or in the top, with the sample line protruding to the middle of the pipe. The latter is indicated in the manuals as the preferred location. GLI International (1999) recommends a sample tap installed at mid-depth in the side of the pipe and flush with the pipe wall as a good location, but the best location is to have a tap at the same place but extend the sample tube into the middle of the pipe. Hach (1999) notes that sample taps that draw water from the top or bottom of a pipe may cause interference from air bubbles or bottom sediments.

After a tentative location has been selected for a sample tap, the practicality of this location needs to be considered before tapping into the header. Flow from the tap should travel to the turbidimeter under the forces of water pressure within the pipe, or gravity, or a combination of those. The available head to provide flow to the turbidimeter may vary depending on head loss in the filter bed, and this has to be considered. Pumping filter effluent samples will break up particles and may change the turbidity, and particle count sample streams must not be pumped ahead of the particle sensor. (Finished water sample flows, analyzed for regulatory purposes, are likely to be pumped in most filtration plants. They represent water sent to the distribution system, not water from individual filters.)

Filtered water turbidimeters are placed in the pipe gallery, close to the sample tap. The installation is just the initial effort involved with an on-line turbidimeter, so the instrument has to be accessible for future maintenance and calibration. Select a location where these tasks can be done easily, so plant staff is not discouraged from doing these necessary tasks at the needed frequency. Climbing over or crawling under large pipes does not encourage frequent checks of turbidimeters.

For accurate turbidity measurement, the sample line from the source to the turbidimeter must be a material that will not corrode or otherwise contribute to the turbidity of the water being transported to the turbidimeter. Hargesheimer and Lewis found that stainless steel, copper, and PTFE tubing were satisfactory for use as sample lines for particle counters, and these materials are suitable for turbidimeters. Hach (1997) recommends use of ¼ inch rigid or semi-rigid tubing, routed as directly as possible from the sample point to the turbidimeter. Use of longer tubing causes a longer lag time between the extraction of the sample from the plant’s piping and the analysis of that sample. Hargesheimer and Lewis (1995) found no influence on particle count for tubing lengths of up to 10 feet, but recommended that the length of sample tubing be as short as possible. The flow through an on-line turbidimeter is small compared to flow from a filter so it can be wasted after analysis. The design of the waste line ought to facilitate easily checking the rate of flow through the turbidimeter, because this is one of the operating and quality control checks that operators must perform periodically.

Standard Methods for the Examination of Water and Wastewater (APHA, 1995) contains an extensive discussion of statistics, quality assurance and data quality in Part 1000, INTRODUCTION and is the recommended source for information on those topics. Specific procedures to be followed for inspection, cleaning, maintenance, and calibration of instruments and measuring devices are described in manuals provided by the manufacturers.

Inspection and performance checks should be performed on a frequent basis for comparison of on-line and bench turbidimeter results, rate of flow, and condition of tubing. On-line turbidimeters need to be checked about once per week to once per month to verify that the rate of flow through the instrument is within the range specified by the manufacturer. At many plants the head at the sample tap will vary according to the head loss occurring within the filter at that time. Noting the head loss on the filter when the flow is checked can aid in interpretation of the results.

If transparent sample tubing is used for the sample line, the condition of the tubing requires regular checking. Some types of tubing can permit growth of biofilms on the tubing wall, and this could lead to erroneous readings. Flush sample tubing thoroughly after it is cleaned, before reconnecting it to the turbidimeter.

On a daily basis or weekly basis, the reading of an on-line turbidimeter should be compared to the turbidity of a sample of water as determined by a bench turbidimeter. This can be done by collecting a filtered water grab sample at the on-line turbidimeter and recording the reading of that instrument. The on-line reading is compared to the turbidity of the grab sample as determined on a bench model that has first been subjected to a secondary standard check. If the two readings fall outside the expected range of differences usually noted for the on-line turbidimeter versus the bench turbidimeter, recheck the results. If the difference continues to be observed, the on-line turbidimeter should be recalibrated. Do not merely adjust the reading of the on-line turbidimeter to coincide with the reading of the bench turbidimeter. The comparison of bench and on-line turbidity readings is done to learn if the on-line turbidimeter is still reading within an accepted range as compared to the bench model, not to serve as the actual calibration of the on-line turbidimeter. Record the results after the comparison is made.

Calibration of on-line turbidimeters is addressed in the IESWTR. The Rule says, "In addition to monitoring required by § 141.74, a public water system subject to the requirements of this subpart that provides conventional filtration treatment or direct filtration must conduct continuous monitoring of turbidity for each individual filter using an approved method in § 141.74(a) and must calibrate turbidimeters using the procedure specified by the manufacturer." The requirements for calibration will differ from manufacturer to manufacturer. The manufacturer’s instrument manual for on-line turbidimeters is the appropriate guide to calibration requirements and procedures.

The availability of easily used testing equipment and a plentiful supply of low-turbidity water or "particle-free" water will make the task of calibrating turbidimeters much easier when these activities are performed by treatment plant staff. Information on water preparation is presented in Standard Methods (APHA, 1995) in Method 2310, Turbidity, and Method 2560, Particle Counting and Size Distribution. Hargesheimer and Lewis (1995) prepared quantities of up to about 1 liter at a time by pumping high quality water through a cartridge filter. They used a peristaltic pump with a Masterflex Model 7019 pump head, 9 mm ID silicone tubing, and a 0.22 m m pore size cartridge filter (Sterivex-type from Millipore). The water was pumped directly into the clean bottle or container in which it was to be analyzed. The quantities of water needed for quality assurance procedures involving turbidimeters will depend on the number of these devices in use at a particular plant. The person responsible for water quality should review the situation at the plant and determine the most appropriate means of producing the needed quantities of water for dilution or testing.

Dedicated test equipment may be needed when multiple instruments are to be inspected, maintained, or calibrated in an efficient and cost-effective manner. Use of a stainless steel laboratory cart equipped with all of the supplies, expendable items, and apparatus needed for cleaning and/or calibrating turbidimeters would save staff time during the performance of quality control (QC) procedures. At treatment plants that are not large enough to have an elevator servicing the pipe gallery, a laboratory cart could be purchased and left in the pipe gallery for the QC work. Standard Methods and the operating manuals of the instrument manufacturers are good sources for specifics on how to properly equip a turbidimeter QC cart.

Whenever an on-line turbidimeter is calibrated the turbidimeter body must be cleaned and flushed thoroughly. Also, flush the sample line supplying water to the turbidimeter. Always document all procedures carried out for cleaning or calibrating the turbidimeter.

The need for quality control is not diminished because a water system is classified as a small. Data must be accurate, regardless of system size. Small systems may find it appropriate to use a contract service for quarterly calibration of turbidimeters, especially if water treatment plant staff do not have the apparatus or the time to conduct the quality control procedures.

The IESWTR requires plants to gather and monitor a lot of turbidity data. The rule requires that turbidity values of individual filters must be monitored at 15-minute intervals. This in itself may not seem too difficult but the rule also specifies turbidity values, which if exceeded will trigger a requirement for additional filter monitoring and reporting. These values vary depending on how long the filter has been in service.

For example:

Small plants, with only a few filters, may be able to meet these monitoring requirements by manually monitoring and alarming the data from a strip chart recorder or data logger. Most larger plants will require the use of some type of computerized data management system such as a plant SCADA (Supervisory Control and Data Acquisition) system to accomplish this.

A common approach to providing computerized data management is through the use of programmable logic controllers (PLCs) and personal computers (PCs). The PLCs gather the data from online turbidity analyzers and then transmit the data to a database or spreadsheet program running on the PC.

Recordkeeping will be an important task for compliance with the IESWTR, as reporting and recordkeeping requirements stipulate that systems must maintain the results of individual filter monitoring for three years. For a single filter, data must be recorded each 15 minutes. This requires 4 turbidity values per hour or 96 per day for a filter operated continuously for 24 hours. The number of quarter-hour intervals in a year is 35,040. Filter backwashing would reduce the number of turbidity data entries by a few percentage points, but with a filter operating all day, every day (except for backwash), over 30,000 turbidity data values could be recorded and kept in a year’s time. In three years, this total could amount to 100,000 for a filter in nearly constant use. If time were kept on a 24-hour clock, data would be needed for the hour (e.g. 0915, 1445, etc.), day, month, and year of each turbidity value. This could necessitate recording and keeping as many as 500,000 data entries over a 3-year period, for just one filter. Clearly data acquisition and storage will have to involve computers, as this would be a burdensome task if done manually.

Some desirable features for a monitoring and reporting system are summarized below:

a. The system should continuously monitor and alarm the turbidity values received from online turbidity analyzers. The system should alert the plant operator any time that a real-time turbidity value is above a preset desired limit (such as 0.20 or 0.25 NTU) or any time that the value is changing a rate faster than some preset limit (such as greater than 0.2 NTU per hour increase). The turbidity level selected as the alarm limit will depend on the water quality goals that have been adopted by the utility, but the limit would definitely need to be lower than a turbidity level that could cause regulatory reporting for an individual filter.

b. The system should log the actual turbidity values to a report spreadsheet or database at least once every 15 minutes. The system should only monitor and log values if the associated filter effluent valve is at least partially open (not closed). Typically an online turbidity analyzer will continue to send values to the PLC even if the filter is not in service. This could result in false or erroneous alarms.

c. The system should alarm and flag or mark any turbidity values that exceed the limits set by the IESWTR or the limits set by the applicable regulatory agency. The system software should be smart enough to automatically change the alarm limits after the filter has been in service for four or more hours as described in the rule. One way to automatically change the alarm limits is to monitor the effluent valve position and automatically reduce the alarm limit set points when the filter effluent valve has been open (not closed) for four or more hours after a backwash has occurred.

And finally, one important step in the data monitoring program is to educate that plant operators on how to use this data to avoid operational situations which could lead to "exceedance alarms". By alerting operators to alarm conditions before an actual exceedance occurs it may be possible to avoid the exceedance condition. Some plants are even considering an automatic shutdown of a filter if turbidity approaches an alarm/exceedance condition.

Generally the quality of filtered water produced by a filter is poorer just after a new filter run starts, and quality then improves as the run progresses. This has been referred to as the initial improvement period, or filter ripening. At filtration plants where operators have coagulation under control, the long period of time when the filter is producing excellent water should not present problems. Late in a filter run, if turbidity breakthrough begins to occur, an alert operator can stop the filter and backwash it before a turbidity limit is exceeded. The portion of a filter run in which turbidity control may be the most challenging occurs at the start of the run, during the initial improvement period.

The initial improvement period was studied and reported upon in the 1960's (Cleasby and Baumann, 1962; and Robeck, Dostal, and Woodward, 1964) and was studied in detail by Amirtharajah and Wetstein (1980). Because of the increased risk of passage of pathogens during the initial improvement period, the turbidity spike that occurs at the start of a filter run should be minimized.

Several techniques for minimizing the time needed to attain routine or normal low-turbidity water after starting a filter run are described in this paper, including:

• filter to waste,

• the delayed start,

• the slow start,

• adding polymer or coagulant to backwash water, and

• adding coagulant chemical or cationic polymer to the settled water as it fills the filter box after backwash is terminated.

During the course of AWWARF Project 2511 (Filter Maintenance and Operations Guidance Manual, referred to in this paper as the AWWARF Manual Project), authors of this paper asked water utilities to report on techniques they used to minimize the degradation of filtered water quality at the start of a filter run. Utilities provided information on operations at 37 surface water filtration plants where the initial turbidity spike nearly always was held to 0.3 ntu or lower, a goal of the Partnership for Safe Water. Information was presented for 7 surface water filtration plants that did not consistently hold down the initial turbidity spike to 0.3 ntu. Techniques used for controlling initial filtered water quality included conditioning the backwash water or the influent settled water after backwash with a polymer or with a coagulant chemical, holding the filter out of service for a period of time after backwashing it, increasing the filtration rate gradually when the filter is put in service, and using filter to waste. Some plants used none of the above techniques, and some used more than one.

Data provided by the utilities are summarized in Table 1. This table indicates that the majority of plants that successfully controlled the initial turbidity spike employed more than one method of doing this. When multiple techniques are given for controlling the initial turbidity, they are listed in the order in which they are performed by the utility or utilities. Only three plants placed all of their reliance on filter to waste, while at 15 plants, filter to waste was used in combination with one or more other methods for lowering turbidity when the filter run began. Using other approaches to controlling the initial turbidity spike in addition to filter to waste may enable filtration plant operators to reduce the duration of the spike and shorten the time needed for filter to waste. A single approach successfully controlled the initial spike at 9 plants, whereas multiple approaches were used with success in 26 plants. Among the seven plants that did not consistently control the initial turbidity spike to 0.3 ntu, four used no special techniques to control the spike, and only one used a combination of two approaches. Three of the seven plants not consistently controlling the initial turbidity spike were lime softening plants. Two lime softening plants were able to meet the Partnership for Safe Water goal of keeping the initial turbidity spike to 0.3 ntu or lower when filter runs are started. Prevention passage of calcium carbonate crystals may be more difficult than controlling coagulation floc at the start of a filter run.

Filter to waste, sometimes called rewash, involves wasting the water from the first portion of a filter run rather than putting it into the clearwell. Water from filter to waste can be recycled back to the raw water. Generally this would improve the quality of the raw water. Some plants carry out the filter to waste process for a fixed period of time, but this practice may not achieve the goals for filter to waste. Filter to waste is done to prevent the discharge into the clearwell of the initial high-turbidity water that is produced by the filter.

Bucklin, Amirtharajah, and Cranston (1988) studied filter to waste and stated, "In summary, a filter to waste procedure is only of value if (1) the post backwash characteristics for a particular system are understood, and (2) there are clear objectives for the use of the filter to waste period in terms of what exactly is to be accomplished." They did, however, conclude that because alternatives to filter to waste are available, other approaches to controlling the initial filtered water quality may be attractive to water utilities.

One basis for managing filter to waste is to monitor the filtered water turbidity, and waste it until the filtrate meets the turbidity goal of the utility. At that time, the filtered water quality is acceptable for discharge into the clearwell. At filtration plants where particle counting is done routinely, particle counts in filtered water could be used as the quality criterion for termination of filter to waste. Whether particle counting or turbidity measurement is used, the important concept is that filter to waste is managed on the basis of the filtered water quality (the real reason for using filter to waste) instead of being managed only on the basis of a time period for wasting filtered water. The length of time for filter to waste is site-specific and may vary with source water quality or pretreatment approach. The question of how long to operate filter to waste needs to be resolved on a plant-by-plant basis.

At some filtration plants the filter to waste piping may not be adequate to carry away the wasted filtrate when the filter is operated at the full filtration rate. In this circumstance, filter to waste would have to be conducted with the filter operating at a reduced rate, and after filter to waste ended the filtration rate would need to be increased. Increasing the filtration rate after filter to waste is ended could cause a spike or increase in turbidity, so careful management of the rate change is required. Unless turbidity of the filter to waste stream is measured continuously, and the turbidity of the filtered water is measured continuously, operators will not know if increasing the filtration rate at the end of filter to waste has had an adverse effect of water quality.

At the end of filter to waste, a brief deterioration of filtered water quality may occur even if filter to waste is done at the full filtration rate, because of the need to change from diversion of filtered water to waste to directing the filtered water to the clearwell. Closing the filter to waste valve and opening the valve to the clearwell piping might cause a filter disturbance and temporarily impair water quality. Again the existence of a turbidity spike, its duration, and its magnitude would be very difficult to detect without having continuous monitoring of the turbidity of filtered water, both in the wasting mode and the production mode. Bucklin, Amirtharajah, and Cranston (1988) observed that at the Bozeman Water Treatment Plant, where filter to waste was practiced, the turbidity spike was related to the closing of the filter to waste valve, the opening of the clearwell valve, and the filtration rate in effect during the change-over. They explained that closing and opening the valves produced sudden rate changes in the filter.

A well-designed system should ensure that the opening of the filter outlet valve and the closure of the filter to waste valve can be achieved smoothly and with no overall rate change to water coming out of the filter. One way to do this is to set up the piping with a three-valve scheme where the normal rate control valve is followed by two diverter valves in parallel; one to the clearwell and the other to waste. This allows diverting the controlled flow from waste to the clearwell with a minimal disturbance to the flow rate, thus helping to minimize the turbidity upset. Sometimes questions are asked about the filter to waste rate. In seeking to minimize water loss a slow filter to waste rate may be proposed. This simply delays ripening and leads to the risk of a hydraulic shock when the filter comes back into service, potentially causing turbidity or particle breakthrough.

Major reconstruction often is needed to retrofit a plant with filter to waste piping and valves. This can be an expensive option for existing filtration plants, and may limit use filter to waste for controlling turbidity spikes to new plants or to plants where piping and valves are already in place for this procedure.

Allowing filter media to "settle in" after backwashing, for a time of 1/4 hour to 24 or 48 hours has been done at some plants, and some operators believe this helps minimize the initial turbidity spike when the filter is started. In this strategy for controlling the initial turbidity spike, after backwashing is completed settled water is introduced to the filter box and held for a period of time before filtration begins. A delayed start could improve initial filtered water quality by allowing the filter media to consolidate slightly after backwashing, tightening the pore structure. Also during the delay time, floc particles that were not washed out of the filter box during backwash would have the opportunity to settle to the top of the filter media, and floc not removed from the expanded filter bed during backwash could attach to the filter media.

Using the delayed start method does not require special equipment or filter to waste piping. This procedure is applicable, however, only when plants have sufficient filtration capacity so it is not necessary to have all filters in service at the same time. During times of maximum water production, employing a delayed start much longer than 1/2 hour may be impractical.

Delayed filter starts have been evaluated at three full scale plants and found to improve initial filtered water quality (Baird and Hillis, 1998). To evaluate the degree of quality improvement attained, they depended on particle counting. The total particles in the 2 to 5 µm range were compared for the ripening period for control filters (no delayed start) and test filters operated with a delayed start. In the three plants, delayed starts reduced the particles during the ripening period from 35 percent to slightly over 42 percent. Baird and Hillis presented figures depicting particle counts in filtered water versus time showing reductions in the peak particle count of over 50 percent for the delayed start filters as compared to the control filters with regular start.

A similar positive benefit of filter resting was reported by Pizzi (1996), who presented data from a study of two filters at a plant in Ohio. One filter was started immediately, whereas the other was rested for 4 hours. The turbidity peak from the rested filter was about half that of the peak for the filter placed into service immediately after backwashing, and the ripening period for the rested filter was 45 percent shorter (34 minutes versus 62 minutes).

Resting filters before starting a new run is not a cure-all, as four plants participating in the AWWARF Manual Project reported using delayed starts but did not consistently control their initial turbidity spike. Nor can filters be rested for an unlimited length of time. This also applies to water filtration plants that have excess capacity, where the practice of keeping backwashed filters on "standby" for future service might be common.

Microbiological problems can develop if filters are out of service for an excessive time and then returned to service without being backwashed again. Wierenga (1985) observed that bacteria grew in out-of-service filters after the chlorine residual was depleted. Filters need to be backwashed after being held out of service for a long period of time. In general two days may be an upper limit for leaving a filter out of service without a second backwash when it is returned to service. In the summer, when water temperatures are in the range of 25^o to 30^o C, two days might be an excessive time to leave a filter out of service without a backwash before it is used again because bacteria can grow much more rapidly than when water is warm. If questions exist about the status of a filter, a quick check of the dissolved oxygen concentration in the filter could be made. If biological action has brought dissolved oxygen down near zero, the filter needs to be backwashed before it is placed in service.

No construction is needed to employ the delayed start procedure if extra water production capacity exists. Care is required to ensure plant output is not affected, either in terms of volume or quality with extra flow going through remaining filters.

At some filtration plants, the initial turbidity spike is managed by starting the filter at a low filtration rate and gradually increasing the rate over a period of time, such as 15 to 30 minutes. For this procedure to be used, the filters must have rate control valves that can be gradually increased. This approach would not be applicable at declining rate filtration plants unless the filters were equipped with rate control valves that were used at the beginning of the filter run to limit the filtration rate. Use of slow start is not new. Hudson mentioned British experience with "slow start modules" in his discussion of a paper on filter rate changes (Cleasby, Williamson, and Baumann, 1963). From this we can infer that the concept of starting a filter slowly has been in existence for at least four decades.

Results for using a slow or gradual start to attain better quality during filter startup are mixed. Some investigators have not found this to be beneficial, but others have seen a benefit. On the basis of particle counting at full-scale filters, Borrill and McKean (1993) concluded that the slow start procedure was delaying the passage of particles but not decreasing the total number passed during ripening.

In the studies of Hillis and Colton (1995) nearly 50% of the 2-5 µm particles penetrating the filter during a 48 hour run did so during the ripening period. The effect of the slow start was to prolong the ripening period from about 25 minutes to 45 minutes. The slow start showed a two- peak ripening period in contrast to the single peak when no slow start was used. Measuring 2-5 µm diameter particles, the single peak reached 30,000 particles/mL, whereas the two peaks were each around 22,000 particles/mL. After ripening, the filtrate contained around 50 particles/mL. The cumulative number of particles passed in a given filtered water volume showed a clear benefit of the slow start, as it reduced treated water particle counts by about 30%. However the authors were aware that the counts during ripening were still high. Hall and Pressdee (1995) reported a similar reduction (average 30%) in particle counts comparing a slow start with a full rate start. Once again, however, the filter with the slow start did produce a particle count spike during the ripening period. It must be concluded that although the slow start makes some impact, removing 30% of a large number of particles still leaves a large number of particles. However trimming the height of a peak may enable a utility to meet its goals, albeit not necessarily the spirit of those goals.

In their AWWARF-sponsored study of the characteristics of initial filter effluent quality, Bucklin, Amirtharajah, and Cranston (1988) reported that at the Helena Water Treatment Plant a direct relationship was found between the magnitude of the initial turbidity peak and the rate of flow when a full-scale filter was placed into service. They wrote, "The flowrate at which a filter is started immediately following backwash is proportional to the turbidity peak. This is a good argument for the incremental startup of a filter after a backwash." They concluded that the incremental, or gradual, filter starting procedure reduced the length of time needed for the initial improvement period. The standard procedure for placing a filter in service at Helena during the AWWARF study was "…. to slowly put the filter back on line in incremental flow rate steps …" This procedure was compared to an instantaneous startup mode during the study. Table 2 in this paper summarizes the data for filter runs listed in the AWWARF report (Bucklin, Amirtharajah, and Cranston; 1988) for which the ultimate rate attained was either 2.6 or 2.8 gpm/sf, based on Tables 11 and 12 of the Bucklin, Amirtharajah, and Cranston report. Using an incremental startup did not lower the peak turbidity observed on startup, but this procedure substantially decreased the duration of the turbidity peak (1 hour average for incremental startup versus 3 hours average for instantaneous.)

An example of the schedule for an incremental startup is presented in Table 3 of this paper. This table was based on Figure 15 of the AWWARF report of Bucklin, Amirtharajah, and Cranston, who did not specify the rate of change for moving the valve (the number of seconds required for the valve to change its position) during the incremental changes. During the filter run for which startup rate data are given in Table 3, the peak turbidity reached during startup was approximately 0.27 ntu, according to Figure 15 in Bucklin, Amirtharajah, and Cranston’s AWWARF report.

The benefit to be gained from using the slow start approach for limiting turbidity spikes at the start of a filter run may be a function of the valve control mechanisms at the filter plant. Bearing

in mind that Cleasby, Williamson, and Baumann (1963) reported that slow and gradual rate changes were less disruptive than abrupt changes, we can infer that the very slow and gradual opening of a valve would be preferable to intermittent but jerky operations. Thus if the six valve position changes illustrated in Table 3 were each made gradually over 30-second intervals rather than abruptly in 1 or 2 seconds, the effect of changing the valve position about every two minutes should be minimized.

Operating the rate control valve slowly, in steps; instead of quickly or abruptly, in steps, is recommended. Furthermore, because Cleasby, Williamson, and Baumann found that the magnitude of the rate change influenced water quality (smaller changes caused less deterioration) operating the valve in multiple, smaller steps is preferred over using only a few, larger rate changes to go from a filter at rest to operation at the full rate.

At one plant participating in the AWWARF Manual Project currently in progress, filter flow is ramped up by the SCADA system, using flow increases of about 0.2 mgd every 2 minutes beginning at a rate of 2 mgd per filter. This sequence of step increases and time intervals equates to a rate increase of 1 mgd in 10 minutes. The time needed to attain the total increase of 3 to 3.8 mgd (from the initial 2 mgd up to maximum set points ranging from 5 to 5.8 mgd per filter) therefore would be 30 to 38 minutes. Often the initial turbidity does not exceed 0.1 ntu, and the worst case reported may be 0.16 ntu for a peak, with an hour needed to go below 0.1 ntu in filtered water. This plant also adds polymer to backwash water.

Three filter plants participating in the AWWARF Manual Project reported that the initial turbidity spike was not controlled consistently to 0.3 ntu or lower by use of gradual rate increases at the start of filter runs. Furthermore one plant tried this approach and it did not work at that installation. This approach does not come with a guarantee of success. As described above, a filter rate start up that is slower, and more gradual, done with a larger number of smaller rate steps would have a better likelihood of success.

No construction is needed if operators can manipulate filter rate control valves slowly, gradually, and easily. If valves have electric operators, changes to programs controlling valve operation may enable operators to use the gradual or slow start technique if they are not doing so presently.

Addition of coagulant chemical or polymer to backwash during a portion of the filter backwash has been practiced for about three decades, beginning with Harris (1970) at the Contra Costa County Water District in California. Harris reported that filter media was conditioned to improve particle adsorption by adding nonionic polymer in the filter backwash water at a dosage of about 0.10 mg/L. He noted that conditioning the filter media by adding the polymer during backwash was not a substitute for coagulating with an adequate dosage of alum. Harris stated that at his plant, filters could be started at a rate of 10 gpm/sf and immediately produce effluent turbidity of 0.10 or lower. The plant was equipped with filters having anthracite and sand media, and air scour was used, rather than surface wash, to provide auxiliary scour.

Pilot plant studies (Yapijakis, 1982) showed benefits of adding polymer to backwash water for direct filtration of surface waters in the New York-New Jersey area. In one case, using a nonionic polymer dosage of 0.10 mg/L to 0.15 mg/L reduced the duration of the initial improvement period by about 50 percent but did not decrease the turbidity peak. In the other water studied, using nonionic polymer dosages of 0.05 mg/L to 0.15 mg/L reduced the turbidity peak and decreased the filter operation time required to attain a filtered water turbidity of 0.3 ntu.

Cranston and Amirtharajah (1987) presented turbidity data for filter starts in which addition of alum to backwash water was tested at the Bozeman, Montana water filtration plant. According to Figure 21 of their paper, the highest turbidity peak occurred when no alum was added to the backwash water. The most effective strategy was to add alum for the final 2.5 minutes or for the full 5 minutes of backwash, at a concentration of 17 to 18 mg/L in the backwash water.

Cleasby et al (1992) evaluated a direct filtration plant in Bozeman, Montana as part of an AWWARF project. They reported that the turbidity spike during the ripening period was completely eliminated by use of nonionic polymer to condition filter media during backwashing, a practice similar to that of Harris (1970). At Bozeman, a nonionic polymer dosage of 1.3 mg/L was fed into the backwash water during the last 3 minutes of the filter backwash.

Addition of coagulant or polymer to backwash water helps to condition the remaining suspended solids that were not washed out of the filter media during backwashing and the suspended solids that remained in the water over the filter media before the end of the backwashing.

At the end of a filter backwash, low turbidity, unused backwash water (typically, finished water) remains in the filter underdrain, under the media. If coagulant chemical or polymer addition is not stopped for a short time interval before the end of backwashing, some chemically conditioned backwash water will remain in the piping and underdrain cavity below the filter support material and filter media. At the start of the next filter run, the small volume of finished water containing polymer or coagulant will be discharged to the clearwell. The length of time between stopping the coagulant or polymer feed and ending the backwash will depend on the rate of flow of backwash water, and the volume of water that would be in the piping and underdrain. Plant staff should be able to calculate this from plans of the filters and data on the backwash flow rate. If backwash water is treated by adding polymer or coagulant, stop the addition of the chemical for a short time before the backwash water ends, as described above.

No clear consensus on ways to determine coagulant or polymer type and dosage and the duration of dosing was found in the literature or in the survey of utilities done for the AWWARF Manual Project by the authors. Pilot plant testing and evaluation of a single full-scale filter equipped with filter to waste are two ways to evaluate addition of coagulant or polymer. In either of these situations, the filtered water could be wasted instead of flowing into the clearwell. If a full-scale evaluation is performed at a filter plant lacking filter to waste, operators will need to use caution and keep state regulatory engineers fully informed of plans, procedures, and testing results.

Adding coagulant chemical or polymer to backwash water does not require as much construction and equipment as adding filter to waste, but the equipment needs are significant. To condition backwash water with chemicals, it is necessary to provide a chemical supply tank, a chemical pump, piping, valves, and controls.

Some water utilities where alum is the primary coagulant, including Grand Rapids and Greenville, SC, control or mitigate the initial turbidity spike at the start of a filter run by adding alum to the settled water entering the filter box as it is refilled following the termination of backwashing. At Greenville, a comparison of filter starts alternating with added alum and without added alum demonstrated that adding alum gave statistically lower turbidity peaks at startup. Success has been attained with an alum slug dose that is the equivalent of about 4 mg/L in all of the water above the filter media after the filter box has refilled. Mixing of the slug dose takes place because it is added during the early phase of refilling the filter, when turbulence is higher in the water above the washwater trough. At Milwaukee, cationic polymer is added in a slug dose to the settled water entering the filter box as it is refilled following the termination of backwashing, and then the polymer is fed into the filter influent for one hour.

Janssens et al. (1982) showed that an overdose of coagulant at the beginning of a filter run could improve initial filter effluent quality. In addition to the change in porosity that might be brought about by the addition of coagulant, adding coagulant to settled water flowing into the filter box could help to destabilize floc fragments left in the filter at the end of the backwash.

Operators who try the slug dosing method or the short-term continuous feed method are advised to start at low dosages of coagulant or polymer and gradually increase the dosage if no positive effect is seen in the filter effluent quality. Furthermore, to develop solid evidence that the procedure is effective, they should perform alternate filter runs with added coagulant or polymer and filter runs having no added coagulant or polymer. In this way it can be established that the lowered turbidity spike at the start of the run was a result of adding the chemical to the filter influent water and not a consequence of source water quality and treatability changes that occurred over time as the testing program was under way. A common mistake that has to be avoided is the assumption that "if a little bit is good, a lot is much better." Overdosing either an inorganic coagulant or a polymer when settled water flows into the filter box after backwash could have serious consequences. If excessive alum is added to the influent settled water, mudballs might develop in the filter. The ability of excess polymer dosages to cause problems of short filter runs and mudball formation is well known.

Some water utilities treat source water for which alum is somewhat ineffective in very cold water. If this condition exists, the period of time between addition of alum to the filter box and the start of the run may be too short for the alum to react. This could cause dissolved alum to pass through the filter bed and into finished water. Where alum reaction time has been a problem in winter, this technique for controlling the initial quality of water produced at the start of a filter run may not be helpful. If the procedure is tried, monitoring the filter effluent for dissolved aluminum for the first hour is strongly recommended, with samples collected every 10 minutes, or more frequently. For comparison purposes, two or three dissolved aluminum samples would be required during the stable portion of the filter run, after the end of the initial improvement period.

If a full scale evaluation is performed at a filter plant lacking filter to waste, operators will need to use caution and keep state regulatory engineers fully informed of plans, procedures, and testing results

Equipment needs for adding coagulant chemical or polymer to filter influent water range from very simple to something similar to that required for adding chemicals to backwash water. No construction is needed if the chemical is added manually in a slug dose. To provide for a continuous chemical feed into the filter box for a sustained period of time such as one hour, a chemical supply tank, chemical pump, piping, and controls would be needed.

Implementation of the Interim Enhanced Surface Water Treatment Rule will bring about significant changes in turbidity monitoring and reporting at plants that are not already using on-line turbidimeters to monitor the quality of water from each filter. Large amounts of data will need to be collected and stored. Utilities will have to be capable of doing this by December 2001. A variety of follow-up actions will be required if defined turbidity limits are exceeded. When on-line turbidity monitoring data are used for regulatory purposes and not just for providing plant operators with real-time information on filter performance (and especially changes in performance) quality control and instrument calibration issues will assume greater importance.

At filtration plants where coagulation is being done correctly, operators should be able to manage their filters so that low-turbidity filtered water is produced after the initial improvement period. With a turbidimeter installed at each filter, operators ought to be able to terminate a filter run before turbidity breakthrough causes regulatory problems. The remaining aspect of filter operation needing attention is the initial improvement period that occurs when a filter is returned to service after backwashing. Several techniques for lowering filtered water turbidity during the initial improvement period are reviewed. Plant operators are encouraged to evaluate and use multiple approaches for controlling the initial filtered water turbidity and shortening the duration of the initial improvement period as an approach to producing better filtered water quality and avoiding compliance problems related to turbidity excursions.

Some of the information presented in this paper was developed in work done for the AWWA Research Foundation on Project 2511, Filter Maintenance and Operations Guidance Manual. The Project Manager for this ongoing project is Traci Case. The support of both the AWWA Research Foundation and the Project Manager are gratefully acknowledged. The cooperation of over three dozen participating water utilities is sincerely appreciated.

Amirtharajah, A. and Wetstein, D.P. 1980. "Initial Degradation of Effluent Quality During Filtration," Journal AWWA, 72:9:518-524.

APHA, AWWA, and WEF (American Public Health Association, American Water Works Association, and Water Environment Federation). 1995. Standard Methods for the Examination of Water and Wastewater. 19^th ed. Washington, D.C.: APHA.

Baird, G. and Hillis, P. 1998. "Full Scale Evaluation of Filter Start-up Strategies to Reduce Particle Passage into Drinking Water Supply," In Proceedings AWWA Water Quality Technology Conference, November 1998, San Diego, California. CD ROM from AWWA, Denver, Colorado.

Borrill, R.J. and McKean, J. 1993. "Improving the Operation of Drinking Water Filters Using Particle Size Analysis," In Proceedings AWWA Water Quality Technology Conference, Miami, Florida, November, 1993, 1611-1630.

Bucklin, K., Amirtharajah, A., and Cranston, K.O. 1988. "The Characteristics of Initial Effluent Quality and Its Implications for the Filter to Waste Procedure," AWWA Research Foundation Report, Denver, Colorado.

Cleasby, J.L. and Baumann, E.R. 1962. "Selection of Sand Filtration Rates," Journal AWWA, 54:5:579-602.

Cleasby, J.L., Williamson, J.M. and Baumann, E.R. 1963. "Effect of Filtration Rate Changes on Quality," Journal AWWA, 55:7:869-880. (This article contains a discussion by H.E. Hudson, Jr., and closure by the authors.)

Cleasby, J.L., Sindt, G.L., Watson, D.A., and Baumann, E.R. 1992. "Design and Operation Guidelines for Optimization of the High-Rate Filtration Process: Plant Demonstration Studies," AWWA Research Foundation, Denver, Colorado.

Cranston, K.O. and A. Amirtharajah. 1987. "Improving the Initial Quality of a Dual-Media Filter by Coagulants in the Backwash." Journal AWWA. 79:12:50-63.

GLI International. 1999. Operating Instruction Manual No. T53/8320, Revision 1-499. Milwaukee, Wisconsin: GLI International.

Hach Company. 1999. 1720D Low Range Process Turbidimeter. Loveland, Colo.: Hach Company.

Hach Company. 1997. 1720C Low Range Process Turbidimeter Instruction Manual. Loveland, Colo.: Hach Company.

Hall, T. and Pressdee, J.R. 1995. "Removal of Cryptosporidium during Water Treatment." In Proceedings of Workshop on Treatment Optimization for Cryptosporidium Removal from Water Supplies. West, P.A. and Smith M.S. (eds.) HMSO, London.

Hargesheimer, E.E. and C.M. Lewis. 1995. A Practical Guide to On-Line Particle Counting. Denver, Colo.: AWWARF and AWWA

Harris, W.L. 1970. "High-Rate Filter Efficiency," Journal AWWA, 62:8:515-519.

Hillis, P. and Colton, J.F. 1995. "Use of Particle Monitors to Evaluate Filter Backwash and Start-up Strategies," In: Advancing the Science of Water: An International Technology Transfer Conference, UKIWR, AWWARF, IWO; Warwick, U.K. 28-29 September, 1995, 225-229.

Janssens, J.G, Adam, C. and Buekens, A. 1982. "Statistical Analysis of Variables Affecting Direct Filtration," In: Water Filtration, Weiler, R. and Janssens, J.G. (Eds.) KVIV, Antwerp, 4.65-4.80.

Pizzi, N. 1996. "Optimizing Your Plant's Filter Performance," Opflow, 22:5:1, 4, 5.

Pontius, F.W. 1999. "Complying with the Interim Enhanced Surface Water Treatment Rule." Journal AWWA, 91:4:28, 30, 32, 187-188.

Robeck, G.G., K. A. Dostal and R.L. Woodward. 1964. "Studies of Modifications in Water Filtration." Journal AWWA, 56:2:198-213.

Wierenga, J.T. 1985. "Recovery of Coliforms in the Presence of a Free Chlorine Residual," Journal AWWA, 77:11:83-88

Yapijakis, C. 1982. "Direct Filtration: Polymer in Backwash Serves Dual Purpose," Journal AWWA, 74:8:426-428.