Performance Characteristics of Open and Closed Bead Filters in A

Performance Characteristics of Open and Closed Bead Filters in A

Performance Characteristics of Open and Closed Bead Filters in a

Closed Recirculating Tilapia Production System

Brent A. Skeen
Environmental Research Laboratory
University of Arizona
Tucson, Arizona / J. Jed Brown
Environmental Research Laboratory
University of Arizona
Tucson, Arizona
Kevin Fitzsimmons
Environmental Research Laboratory
University of Arizona
Tucson, Arizona / Gary Dickenson
Environmental Research Laboratory
University of Arizona
Tucson, Arizona


The relative efficiencies and operational costs of two replicated enclosed floating bead and one open floating bead filter, with continuous and intermittent cleaning, respectively, were determined over a ten month period. In addition, one round of intensive diurnal sampling was conducted to compare the nutrient removal rates of the two filters types over a twenty four hour period. In phase one of the study a closed recirculating system was constructed requiring each bead filter to equal amounts of filter influent from a common source. Analyses of water quality parameters were conducted on influent and effluent as a means of evaluating removal rates and relative efficiencies of the filters. Over the course of the ten month period the enclosed bead filters and the open filter were not significantly different for the removal of total suspended solids (TSS) or in the reduction of total ammonia nitrogen (TAN). The enclosed filters were found to significantly lower dissolved oxygen (DO). The open bead filter added significant amounts of DO, a result of its continual recirculation of water within the system. The open bead filter was also found to remove a significantly greater amount of nitrite.

In phase two of the study, system plumbing was adjusted to significantly increase the load on the filters, with the flowrate of the open filter considerably greater than those of the enclosed filters. This change was conducted in order to operate the open bead filter within the suggested design parameters, which accommodate greater volumes of water than the enclosed filters. Sampling during this period revealed increased system levels of TSS and TAN after a morning feeding. Concomitantly, removal rates of the filters increased and peaked in the late evening, then tapered off into the early morning hours. Removal of TSS and TAN by the open filter was, for a daily average, greater than removal rates of the enclosed filters. The TSS removal rate for both enclosed units was an average of 18.0 g hr-1, while the open filter averaged 39.8 g hr-1, over the twenty four hour sampling period. Similarly, TAN removal rates averaged 0.4 g hr-1 and 1.9 g hr-1 for the closed and open filters, respectively. As in phase one of the study, the open filter was found to add significant amounts of DO to the filtrate. The open filter is capable of filtering solids and nitrifying ammonia, at greater rates, and can accommodate higher feeding levels and subsequent production of solids and ammonia in the production system. The experimental data have elucidated the importance of determining when to feed in relation to sample collection for proper filter systems evaluation.


In comparing filtration systems for application in aquaculture there has been little research carried out to establish standard practices and procedures for their analysis and evaluation (Piedrahita et al., 1996). This is true of most types of filtration systems, but it is particularly important to obtain such information for newly developed bead filters (Malone 1992) which are becoming increasingly more important filtration systems in aquaculture.

Bead filters are a type of granular media filter that utilize low density plastic beads. The beads are used as media in a variation of a fluidized bed reactor (Owsley et al. 1989). The beads perform two functions; they can capture solids and they provide a substrate for bacteria to colonize and perform biofiltration (Malone et al. 1993). While many different species of bacteria may colonize the beads, the most important function is performed by nitrifying bacteria that oxidize ammonia to less toxic nitrate.

The purpose of this study was to determine the performance of two different types of bead filters under similar conditions. Filtering efficiencies of two closed bead filters and one open bead filter were compared in a closed, recirculating system producing tilapia. One of the primary objectives of this study was to help establish a standard protocol to evaluate and compare the performance of granular media filters. This study was part of a larger study to examine the efficiency of solids removal of a wide variety of filters under different conditions.

Methods and Materials

Two identical Armant Aquaculture Inc. (Vacherie, Louisiana) model CBF-4 enclosed upflow expandable granular biofilters (C1 and C2) and an A1 Aquaculture (Bush, Louisiana) CCMB expandable media self cleaning hydraulically driven open biofilter (O) were tested. While both systems were designed as biofiltration units to reduce ammonia and facilitate the removal of solids, fundamental differences exist in their design, operation and maintenance requirements.

Among the first generation bead filters designed for aquaculture applications, the Armant models employ a two-stage upflow media filtration unit design. Influent is drawn into the lowest chamber of the 225 cm tall filter. Once in the unit, process water is filtered through the first of two reaction chambers, a fluidized sand bed. The sand volume contained therein is 0.0283 m3, with a specific surface area of 69.68 m2. Process water is then channeled upward through the floating bead media contained in the uppermost chamber of the unit. The bead media has a volume of 0.113 m3 and a specific surface area of 129.7 m2.

The A1 CCMB unit is described as a continuous cleaning biofilter, and utilizes a unique design to alleviate some of the fouling problems associated with older bead filters. This is accomplished through the continual motion of beads along the outer perimeter of the top portion of the filter whereupon the beads reach a critical mass and collapse toward the center of the filter. Bead media are then forced downward by water motion. Upon reaching the bottom, beads are then pushed upward by water motion and positive buoyancy until they reach the top again, where the cycle repeats. This filter has a bead volume of 0.433 m3. The A1 filter was placed in a 6.01 m2 holding tank, into which sump water was pumped. In addition, treated effluent is continually recirculated from the filter to the holding tank.

Phase I

In phase one of the study (monthly sampling), the system was plumbed to allow all three filters to receive and filter 24 l/min influent from a common sump (Fig. 1). Filtered effluent from the two enclosed units was then fed to 16 rearing tanks, with the open unit continuously cycling effluent directly back to the sump. Discharged water from the rearing tanks was gravity drained to the sump. Several varieties of tilapia were used for the study, with a starting total fish biomass of 323 kg. The experiment was fed a floating semi-extruded feed twice daily, with an average total input of 1,165 g day-1. The two enclosed floating bead filter were backflushed twice weekly to remove trapped solids within the bead media. Although the open filter continuously recirculated its beads we found backflushing was also necessary, albeit at relatively infrequent intervals, to improve bead turn-over and movement within the system.

A determination of the relative operational costs of the Armant and A1 filters tested was an important objective during this phase of our study. Labor, backflush water and energy consumption were all recorded or calculated for the ten month study period. Frequency and duration of backflushing and other procedures necessary for the efficient operation of the filters was determined, and a regular schedule of said maintenance adhered to. The amount of water used in maintaining the filters was also an important consideration. Lastly, the daily power consumed by each filter was calculated based on measured amperage and volt values. Sampling for phase I of the experiment was conducted once per month for a period of ten months. Samples were collected in acid washed plastic containers in the early morning before feeding. Influent to the filters as well as effluent from each of the three filters was collected.

Phase II

For the 24 hour sampling trial plumbing was adjusted to increase flow to both C filters from 24 l min-1 to an average flow rate of 68.7 l min-1 each, close to the maximum allowable flow rates for these specific units. The flow rate to filter O was increased to 161.5 l min-1, a significant increase from 24 l min-1 in phase one of the study (Fig. 4), but still considerably less than the maximum of 322 l min-1 possible for this unit. A large rearing tank was also added to the system, again containing mixed varieties of tilapia, to raise the total biomass to 337 kg. Daily input of feed also increased slightly, which in turn increased the production of solids and nutrients in the system. Type of feed used and feeding practices were maintained.

During the phase II experimental run, the fish were fed at 08:00 hour and the first sample collected at noontime of day one. Additional samples were collected and analyzed every six hours thereafter, for a total of five sampling rounds over a twenty-four hour period. All laboratory analysis of nutrients and general water quality parameters was conducted immediately, for both Phase I and II of the study, using analytical techniques as described in Standard Methods for the Examination of Water and Wastewater (1995).

Figure 1. Phase I System Plumbing (Monthly sampling).

Figure 2. Phase II System Plumbing (Diurnal Sampling).

Results and Discussion

Phase I: Monthly sampling

After several months of analysis it was apparent that the enclosed bead filters were at least as effective, if not more effective, than the open bead filter in the removal of solids and in the nitrification of ammonia from system water. Average influent (sump) values for TAN and TSS were, respectively, 1.2 mg l-1 and 12.3 mg l-1 over the ten month study period. At the relatively low flow rates during this phase of the study, C1 and C2 were able to reduce TAN levels at a rate of 8.5 g day-1 (Fig. 5) and TSS at 124.0 g day-1 (Fig. 6), on average, over the test period. The O filter was also able to reduce TAN and TSS levels, respectively, by an

Figure 5. Monthly TAN Removal Rates

average of 6.2 g day-1 and 102.5 g day-1 for the tested period. Although these removal values appear to be similar they become divergent when data are normalized by their respective filter bead volumes. Normalized TAN removal rates for the enclosed systems averaged 72.0 g day-1 m-3 of bead material. By contrast, normalized TAN removal rates for filter O averaged 14.4 g day-1 m-3 for the test period. Similarly, TSS removal rates for the C filters averaged 995.5 g

Figure 6. Monthly TSS Removal Rates

day-1 m-3, while values for filter O averaged only 213.1 g day-1 m-3 for the test period. These differences in normalized removal rates between the enclosed filters and the open filter may be explained by the fluidized sand beds contained in the lower chambers of the enclosed filters. Based on our results, the additional filtration process contained within the enclosed filters significantly reduced TAN and suspended solids concentrations.

Figure 7. Monthly DO Removal Rates

The relative performance of filter O was improved however, in comparing removal rates of DO in the system (Fig. 7). Filter O added DO at a rate of 19.0 g day-1 to the system, in comparison to filters C1 and C2, each of which removed DO at an average rate of 104.7 g day-1 over the ten month test period. These results are due to the fundamental differences in operation between the enclosed and open filters. Bacteria in both systems consume oxygen. In addition, oxygen is consumed within the bead media by a host of other biological and chemical transformations. Hence, the observed oxygen reduction in the enclosed systems was expected. However, the open recirculating filter oxygenated the effluent to 71.0 % air saturation through three different mechanisms. First, sump water was continuously pumped into the filter’s surrounding holding tank at an extremely high flow rate. Second, influent was continuously recirculated by a second pump up through the center of the filter and then back down into the surrounding holding tank. Third, filtered water from the bead holding bin, located at the top of the filter, continually overflowed into the holding tank. These three combined actions, and their resultant turbulent flows into the holding tank, served to negate the effects of DO removal caused through microbial action in the filter.

Despite requiring more labor and water for backflushing and maintenance needs over the ten month period, the enclosed filters were found to be the more economical of the two tested filter types to operate (Table 1). Filters C1 and C2 required an average of 86.6 l day-1 of non-system water for backflushing, a significantly greater volume than the 35.1 l day-1 required to backflush and maintain filter O. In addition, time required for maintenance of filter O was only half that of the enclosed units, an average of 0.13 hr day-1 for filters C1 and C2 and only 0.07 hr day-1 for filter O. Despite these disadvantages, the enclosed units were still slightly cheaper to operate due to a much lower energy consumption, an average of 17.6 kw day-1, than filter O which required 78.6 kw day-1. However, the relative cost-effectiveness in operation of these units is spatially variable, and a function of local economics for the aquaculturist (i.e., local utility and labor costs). Based on economics in the Tucson, Arizona area and on our specific experimental design, C1 and C2 were an average of 64 % cheaper to operate than the open unit. Filter O, however, is capable of operating with a much lower powered pump than the 1.5 hp unit used in this study. Recent discussions with the manufacturer of the open filter have revealed that these systems may function properly with as little as a 0.3 hp pump. Switching to a lower horsepower pump may reduce the cost of operating the open filter and increase its monetary efficiency in relation to that of the enclosed units.

Labor / BF Water / Energy / Total Cost
($/d) / ($/d) / ($/d) / per day (U.S.$)
C1 / 1.44 / 0.02 / 1.34 / 2.80
C2 / 1.68 / 0.02 / 1.46 / 3.16
O / 0.84 / 0.01 / 3.82 / 4.67

Table 1. Operational Costs

Phase II: 24 hour sampling

The performance of filter O was significantly improved, relative to the enclosed units due to the increased flow rates and loads on the filters, for the diurnal sampling trial. Mean influent concentrations of TSS were 44.7 mg l-1 and mean removal rates over the 24 hour sampling period were 18.0 g hr-1 for C1 and C2, and 39.8 g hr-1 for filter O. Normalized for bead volume, C1 and C2 removed 159.3 g hr-1 m-3, while filter O removed 91.9 g hr-1 m-3. Initial TSS removal rates for C1 and C2 averaged 22.9 g hr-1 four hours after an early morning feeding (Fig. 8). Removal rates for these units then rose to an average of 34.5 g hr-1 near midnight, from which removal rates continually declined for the remainder of the test period. The initial TSS removal rate of filter O was higher than rates for the enclosed units, at 59.0 g hr-1. After a slight early evening drop-off in removal rate the filter peaked during the late night hours, at a removal rate of 139.0 g hr-1. Although normalized removal rates were, again, slightly lower for the open unit than the two enclosed units, it must be noted that these filters are available to the consumer as a complete unit. Hence, absolute removal rates will be a more important consideration for the aquaculturist than normalized values.

TAN removal rates also showed a gradual increase up to the midnight sampling round, after which levels steadily decreased (Fig 9). With average influent concentrations of 0.5 mg l-1 over the 24 hour sampling period, filter O proved superior in the removal of TAN over C1 and C2. The average TAN removal rate for O was 1.9 g hr-1 for the 24 hour sampling period, or 4.4 g hr-1 m-3 of bead material, while the mean removal rate was 0.4 g hr-1 for filter C1 and C2, or 3.9 g hr-1 m-3. Again, filter O’s removal performance peaked during the late evening to midnight hours, where filtration rates were as high as 3.4 g hr-1. Removal rates for C1 and C2 peaked during the same hours but at much lower levels, averaging only 1.7 g hr-1.

DO removal rate values for the enclosed and open filters followed a different trend over the diurnal sampling period (Fig. 10), from TSS and TAN values. They remained fairly uniform for all filters during all of the sampling events. C1 and C2 removed DO from system water at an average rate of 3.7 g hr-1. Conversely, filter O again increased DO levels in system water at an average rate of 6.9 g hr-1 over the twenty four hour period, with its final effluent having a mean oxygen saturation of 84.7%.