As the subject of proper sampling of GAC (granular activated carbon) treatment units has arisen, allow me to explain the principles and suggest some guidelines.
Contaminants are trapped in microscopic crevices within the carbon. The crevices start out empty. The higher the "affinity" or "partitioning coefficient" of the contaminant, the quicker it get stuck and the slower it gets released from the crevice. At design flows, the large amount of empty crevices will remove all of the contaminant quickly. As time progresses, some contaminant will be released (see partitioning) and move deeper into the AC bed. As the crevices at the beginning of the bed become completely filled, contaminant will move deeper into the bed. Eventually, the contaminant will have completely filled the bed and will "break through".
There are very few operational variables. During system design the flow rate per area and the depth of the filter can be chosen. During operation the flow rate can be metered and the media can be changed. Although the pressure drop across the filter can be measured, because of the removal mechanism, pressure drop does not reveal anything about the removal of contaminant. Because a GAC filter also behaves as a depth filter, pressure drop may indicate other filtration and/or contamination problems.
Lacking any physical indications from the equipment, the only way to tell if the filter is working is to test the water before and after treatment. Furthermore, because of the removal process, the removal efficiency will stay at 100% for quite a while, then fairly quickly decrease to zero. So a treated water sample showing no contaminant gives no indication of how "full" the GAC is, nor is a treated water sample with 2% of the raw contaminant likely to be repeatable at a later date.
Given the above constraints, in order to implement GAC to provide reliable removal, a two tank system has been implemented. The two tanks are operated in series. The first tank removes all the contaminant until its "break through". The second tank ensures that contaminant breaking through the first tank does not reach the distribution system. After break through, the first tank, being full of contaminant is removed. The second tank is put in the first position and a new (virgin) second tank is installed.
The effectiveness of the filter can only be determined by sampling. Because no accurate field tests currently exist for organic contaminant samples these samples must be sent away to be tested. A significant delay (weeks to months) between sampling the water and receiving the results must be factored into the design of the removal system. The time to break through (of the second tank) must be significantly more than the time delay in sampling, or sample could break through the second filter and into the distribution system before the operator knew the first filter had broken through.
To determine break through, the water exiting the first tank must be sampled on a regular basis. To determine whether the second tank is working (and what the contamination level entering the distribution system is) the water leaving the second tank must be sampled. To determine whether the system is being operated in accordance with its design, the raw water must be sampled.
All three should be sampled whenever the media is changed, in order to determine whether the new media is working. Intermediate samples should be taken when break through is expected. Raw water and treated water samples should be taken as appropriate to check for variability in the raw water concentration and confirm the safety of the treated water.
Most small systems sample all three sites every 3 months. As long as the time to breakthrough is longer than six months this is sufficient. The 3 months time is loosely based on table 9B. As a sample history and a break through history is developed, the sampling schedule could be optimised.
Given a constant tank size, higher flow and/or higher contaminant concentration mean quicker break through. That is bad, since break through means getting new tanks, disposing of old tanks, and sampling. That's just money. More importantly, quicker break through can mean that the system no longer guarantees clean treated water (see above). When the standard size tank does not last long enough, many designers chose to add more tanks.
More tanks are generally bad, also. Putting tanks in parallel means more sampling and difficult flow balancing. Putting tanks in series means more head loss.
It has been proposed to use a large first tank followed by a small second tank. The second tank would have to be large enough to guarantee a time to break through longer than the sample and change out delay. Both tanks would have to be changed at the same time, so the second tank would never be "full" when changed. The actual time to breakthrough for the second tank would never be measured and would have to estimated from the performance of the first tank. Such an estimate would be "rough", at best, since the characteristics of the two tanks would be different.
GAC does not remove contaminants independently. GAC affinity of contaminants depends on the total chemistry of the water.
If there is a history of sample and GAC change out, sample reductions may be posible. Provided the raw water concentration is constant and the time to breakthrough is significantly longer than the reporting and change out delay, there is no need to sample until shortly before the next break through is expected. Should the intermediate sample be "clean", a follow up sample should be taken soon (monthly).
For most small systems, quarterly sampling is best because
Part 5-1.52 Table 9B requires sampling of raw water quarterly for contaminated sources. Part 5 is quite silent on sampling of treated water. Ideally, a sampling schedule and authority to modify and enforce the sampling schedule should be written into the design approval. 5-1.22 requires design approval. 5-1.71 requires operation in accordance with design.
Grade C Manual section on Granular Activated Carbon.