Build a permanent slow sand filter

London drank slow sand filtered water from 1829. It is the oldest municipal water treatment still in everyday use anywhere, it needs no power, no chemicals and no moving parts, and a well-built one outlives the person who built it.

The biosand filter is this technology compressed into a bucket and run in batches. This is the full-size version: continuously fed, far higher output, and better performing, because it is doing what the process was actually designed to do.

Build this if you expect to be on your own water for a season or longer. Build the bucket version if you need something working this week.

How this differs from the bucket version

The critical difference is the constant head of water above the sand. In the bucket version the 5 cm standing layer only keeps the biolayer alive between batches. Here, 70 cm to a metre of water sits permanently on top, and its weight is what drives the flow. That steady pressure and continuous feed is why a slow sand filter outperforms the batch version: the biological layer is never disturbed, never starved and never allowed to dry.

Sizing it for your household

Slow sand filtration is governed by one number: surface loading rate, the speed water descends through the bed, measured in metres per hour.

Work it the other way round for a household. A family of four needs perhaps 100 litres a day for drinking, cooking and washing, which at the conservative 0.1 m/h loading rate is met by one square metre of filter running for one hour.

That is the surprise in the numbers: the area needed is small. A single 200 litre drum has a surface area of about 0.25 m², giving 25 to 60 litres an hour, which comfortably covers a household running continuously. An IBC tank cut down, or a lined blockwork box a metre square, gives you far more than you will use.

Do not be tempted to run it faster. The loading rate is not a throughput suggestion, it is what makes the biology work. Push water through faster than about 0.4 m/h and you convert a biological filter into a mechanical strainer: it will still take out silt, and it will stop removing pathogens. The slowness is the treatment.

The build

The vessel. A food-grade 200 litre drum, an IBC with the cage retained, or a rendered blockwork or concrete box. It must be opaque or shaded, because light in the water column grows algae in the supernatant and that will blind the sand surface within weeks.

Total depth needed: roughly 5 to 10 cm drainage gravel, 5 cm fine gravel, 70 to 100 cm sand, then 70 to 100 cm of water above that, plus freeboard. That is a vessel about 1.6 to 2.2 metres deep. An IBC on its end or a tall drum works; a shallow water butt does not, and this is the most common reason a home build underperforms.

1. Fit the outlet before anything else. It leaves at the very base, then rises outside the vessel to discharge at a level just above the top of the sand bed. This is the same trick as the bucket version and it is what guarantees the bed can never drain dry, whatever else goes wrong. Getting this wrong is the one build error that kills the filter rather than merely slowing it.
2. Lay graded gravel over the outlet: 5 to 10 cm of coarse (10 to 20 mm) first, then about 5 cm of fine (2 to 5 mm). The grading stops sand migrating down into the coarse layer and out.
3. Wash the sand properly. This is hours of work and it is not optional. Rinse in batches until the water runs clear, then rinse twice more. Unwashed sand releases fines for months and every one of them ends up in your drinking water.
4. Place the sand wet, not dry. Fill the vessel with clean water to above the gravel first, then add sand into the water a bucketful at a time. Sand poured dry traps air pockets that create channels, and a channelled bed passes unfiltered water straight through for the rest of its life. This step is the difference between a filter and an expensive bucket of sand.
5. Level the surface and let it settle for a day.
6. Fit a diffuser above the sand: a perforated plate, a layer of larger stones, or a splash plate, so incoming water never lands directly on the surface and gouges it.
7. Fit an overflow near the top so the water level cannot exceed the design head, and site it so overflow goes somewhere harmless.
8. Fill slowly from the bottom if you can, by back-filling through the outlet, which pushes air out ahead of the water. Filling from the top traps air in the bed.

Ripening

A new filter is sand. It becomes a filter when the biological layer establishes, and that takes time you cannot shortcut.

Allow 2 to 6 weeks. During that period the output is clearer than the input but not biologically treated. Boil or chlorinate everything you draw off, and keep doing so until the filter has been running steadily for at least a month.

Ripening is faster if:

• The feed water carries some organic content. Very clean feed water starves the biolayer and it establishes slowly.
• The temperature is above about 10°C. A filter commissioned in a British January will take the long end of that range, or longer.
• The flow is steady. Start it, set the rate, and leave it alone.

The signs it has ripened: output turbidity drops noticeably, a visible brown or green film develops on the sand surface, and the flow rate declines slightly as the layer thickens.

Scraping, and the half-bed trick

Over weeks to months, the schmutzdecke thickens and the flow falls. Eventually you have to remove it.

The standard method:

1. Stop the inflow and let the water level drop to just above the sand.
2. Scrape off the top 1 to 3 cm of sand, no more. You are removing the biolayer and the trapped material, not resurfacing the bed.
3. Level what remains.
4. Refill slowly, from the bottom if possible.
5. Restart, and treat the output as unripened for 2 to 6 weeks while the layer redevelops.

That ripening penalty after every scrape is the real cost of this filter, and it is why the next technique is worth knowing.

This is standard practice in municipal slow sand plants and almost never mentioned in household guidance. On a small filter, divide the surface with a temporary batten or simply work carefully to a marked line.

When to scrape: when flow has dropped to roughly half its ripened rate, or when the water level above the sand climbs because the bed can no longer take the inflow. Not on a calendar.

Resanding: after perhaps 10 to 20 scrapes the bed has lost enough depth to approach the 0.7 m minimum. Top it back up with washed sand. The traditional method places the new sand underneath the retained top layer, which preserves the mature biology, though for a household filter simply adding clean sand on top and accepting one ripening period is far easier and perfectly workable.

Winter, and other ways it fails

Cold. Biological activity slows markedly below about 10°C and largely stops near freezing, so a British winter reduces treatment performance even though the filter keeps flowing. Insulate the vessel, site it in a shed or partly below ground if you can, and lean harder on boiling in winter. A filter that freezes solid is likely finished: ice expansion disturbs the bed structure and creates channels you cannot see.

Drying out. The one truly fatal error. If the bed dries, the biolayer dies and the sand cracks, and cracks become permanent channels. The high outlet described in the build exists to make this impossible; check it has not been modified or bypassed.

Channelling. Water finding a fast path through rather than percolating evenly. Caused by dry-placed sand, air pockets, a disturbed surface, or scraping too deep. The symptom is output that stays turbid however long the filter runs. The cure is excavating and re-laying the bed wet, which is a day's work.

Algae in the supernatant. Light on the standing water grows algae which then blinds the sand surface, forcing scrapes far more often than necessary. Cover it or shade it.

Overloading. Feeding heavily silted water clogs the surface in days rather than months. Let silt settle before the filter, or run a settling tank ahead of it. A slow sand filter is a polishing step, not a mud trap.