This page continues
Choose a setup, then open the science or care guide when you need it.
Choose a complete setup
Six options compared by space, season, equipment, weekly care, and price.
Why shallow water works
Depth, altitude, summer heat, winter ice, and adult fish size in one practical guide.
Care through the seasons
Plants, fish, mosquitoes, mineral buildup, and the short fall routine.
Start here
Four of the five setups worth copying need no heater, no filter, no air pump and no carbon dioxide. The fifth needs only a pump.
Keep the water shallow and the surface wide, and the container refills its oxygen overnight. Plant heavily from day one so the plants take up ammonium1. Only the sunken tub pond, at a mean depth of 0.70 m, needs an air pump23.
What the cheapest one costs. The three liter jar totals about 12,000 AMD, at 10 to 15 minutes a week43. The balcony tub is about 24,000 AMD planted and unstocked43. The soil tank is built from free Armenian materials, and the same tank on bought aqua soil costs 24,000 to 48,000 AMD more43. Allow a little extra time while you get to know your setup3.
Compare the six setups, then open why shallow water works when you want the science. A jar of pond mud is the smallest place to begin.
Smaller than Model A
Before any of these, there is a two liter jar of pond mud, and it costs nothing at all.
Every setup above is a container you stock on purpose. There is one below all of them that you do not stock at all. Two spoons of sediment, a few dead leaves and water from a river or a pond, in a 1 to 2 liter jar, covered with cotton and an elastic band, on a windowsill. Buy nothing. Add nothing. Whatever appears was already in the mud, waiting as a resting egg. Ostracods and bdelloid rotifers have been hatched out of ditch sediment that was kept cold, rewarmed and flooded, and they were still emerging at the fourth weekly count5.
It sits next to this page for two reasons. It is the cheapest possible way to find out whether you actually enjoy watching a container of water, before you spend a dram on Model A. And that cotton cover answers the mosquito question below in the simplest way there is, because it works in both directions at once. Mosquitoes need very little water to lay in, they lay on the inner wall just above the waterline, and an egg becomes a flying adult in 7 to 10 days6.
The failure mode that has no equivalent indoors
Everything the physics says goes wrong, goes wrong out of sight.
You look at an indoor aquarium several times a day, from the side, through glass, with the light coming through it. Every fish gets counted without anybody deciding to count. You look at an outdoor container from above, through a surface that is busy reflecting the sky, often through a layer of plants, usually on your way somewhere else.
The oxygen low point is at dawn, and nobody is there at dawn. The temperature peak is mid-afternoon on the hottest day of the year, which is a day everybody spends indoors. The ammonia peak from a dead fish arrives on days two to five, from a fish that sank on day one2. Under ice the whole thing happens beneath an opaque lid. And when a cat, a heron, a jay or a neighbor's child takes a fish, the container looks exactly the same afterwards.
The answer is not vigilance, because vigilance is the precise thing this whole approach exists to avoid. The answer is design margin, plus one deliberate ritual.
Design so the worst case survives unobserved. That is the entire content of the stocking and summer sections. Half of the calculated limit exists for this reason and for no other2.
Count the fish out loud, once a week, at feeding time. Feeding is the one moment every animal comes to the surface. Thirty seconds, and it is the only reliable detector of a missing one. This is the single highest-value habit in the whole model.
Two alarms that need no attention. A light-coloured container, so a pale body on the bottom is visible from standing height, and a bare patch of substrate near the edge with no plants over it, as a viewing window.
A maximum and minimum thermometer, left in the water. It costs almost nothing and it tells you on Saturday what happened at three on Wednesday. Without it you are guessing about the most dangerous variable in an Armenian summer2.
A dead animal is a chemical event, not just a sad one. A 100 g fish is about 2.5 to 3 percent nitrogen by wet weight, so complete mineralisation in a 200 L container would raise total ammonia by 12 to 15 mg/L2, against a harmful threshold of 0.76 mg/L at pH 8 and 28 degrees27. A twenty-fold overdose, peaking on day three.
And four things that look like failures and are not. Green water is ugly, almost never lethal, normal outdoors in spring, and is a net oxygen producer during the day. String algae competes with plants and looks bad and is physically harmless. A little sediment blackening is mostly iron sulfide rather than a hydrogen sulfide hazard, because fresh water has roughly 250 to 350 times less sulfate than seawater and Yerevan tap water is unusually low in sulfate89. Mosquito larvae are a sign the container is understocked, not overstocked.
- Our own research on the Walstad method, which carries Walstad's own published statements and their citations. Used for: her filtration and aeration position and the reason for it; the two named maintenance verbs and the six-months-or-more water change figure; the first month described as a race against time; the soil and cap depths; soil sourcing and the Armenian exclusion list; the soak test protocol; the Ararat copper warning; and the finding that too poor a soil is also a failure. See the Walstad method.
- Our own physics research and our own arithmetic. Not a published source. Every figure on this page marked as calculated comes from one file, and each calculation states its model in that file so it can be argued with. Ours: the reaeration time constant table and the two-container comparison; the barometric correction to the oxygen solubility table; the night oxygen budget, the sensitivity analysis and the 0.5 g per litre stocking limit; the length to weight and minimum volume tables; the ammonia and alkalinity arithmetic; the accumulated freezing degree day reconstruction and its validation against published frost night and ice day counts; the ice thickness figures; the freeze-solid energy budget; the under-ice oxygen budget; the surface energy balance for equilibrium water temperature, the diurnal swing and the peak temperature tables; the soil frost depths; the hardness creep and evaporation arithmetic; the surface area and time constant recomputed for each model. The two inputs that move the stocking answer most, sediment oxygen demand and gas transfer velocity, are published ranges rather than measurements of an actual garden container, which is why the recommended figure is half the central calculated value. calculated, repeatable, unverified by a third party
- Our own build research: plants, containers and build recipes. The source of every container dimension, water volume, surface area, plant list, setup sequence, cost total and weekly minute figure attached to a model on this page, together with the plastics code rules, the food-safe container rules and the Armenian container sourcing. Weekly care times are approximate and depend on your setup and routine. Anything marked as an estimate in that file is an estimate here.
- Gazanik, Arshakunyats 54/49, Yerevan. Every price on this page marked as verified was read on 9 September 2026 from the shop's own product data, through its category pages and the price printed on each product page. Stock varies by item and by day and should be checked before travelling. gazanik.am read 9 September 2026
- Trujillo, J., Schwing, C.D., Muturi, E.J. and Cáceres, C.E. (2023). Heterocypris incongruens maintains an egg bank in stormwater habitats and influences the development of larval mosquito, Culex restuans. Ecology and Evolution 13(8): e10445. Open access. The ditch sediment hatching protocol and the ostracods and bdelloid rotifers recovered from it. DOI 10.1002/ece3.10445 open access
- US Centers for Disease Control and Prevention. Life Cycle of Aedes Mosquitoes, last updated 16 April 2024. Eggs laid on the inner walls of water-holding containers above the waterline, the small amount of water required, and the 7 to 10 day egg to adult development time. cdc.gov/mosquitoes/about/life-cycle-of-aedes-mosquitoes.html
- Emerson, K., Russo, R.C., Lund, R.E. and Thurston, R.V. (1975). Aqueous ammonia equilibrium calculations: effect of pH and temperature. Journal of the Fisheries Research Board of Canada 32(12): 2379 to 2383. The relation used for every un-ionised ammonia fraction on this page. This is the single most load-bearing formula behind the ammonia numbers. DOI 10.1139/f75-274 cited from standing knowledge, needs verification
- Our own research comparing nitrifying and heterotrophic bacteria. Used for: nitrifiers being 0.6 to 14 percent of the prokaryotes on real home biofilter media; heterotroph growth rates of 3 to 6 per day against well under 1 for nitrite oxidisers; the 30 to 50 percent loss of nitrification capacity after overfeeding in an inert-substrate tank; mineralisation of dead plant tissue straight back to ammonium; and the sulfate limitation on hydrogen sulfide in fresh water. See cycling and bacteria.
- Our own Armenia research. Used for: the Code of Administrative Offences articles and their amounts, the Law on Fauna, the Law on Hunting gear rule and the 2026 amateur fishing order, each cited in that file to the national legal register; Yerevan tap water by district, from Shahinyan (2005), Proceedings of the National Academy of Sciences of Armenia, Earth Sciences LVIII(1) pp. 41 to 45, sampling 2002 to 2004, giving 2.9 to 10.9 dGH by district, bicarbonate at 61 to 77 percent of anions, sulfate rarely above 11 or 12 percent, and metals at or below detection in almost all samples; the absence of published water data for Gyumri, Vanadzor, Dilijan and the Sevan towns; mosquitofish distribution in Armenia; and carbonate outcrops in the Hrazdan, Vedi and Argichi basins with the vinegar test. See Armenia.
Continue
Where to go next.
A jar from a pond
The version below the cheapest setup here: two litres of pond mud, and it costs nothing at all.
The windowsill tank
The long shallow tank the physics on this page keeps pointing at, built out in full.
Physics
Why volume cancels and only depth matters, worked properly, plus the altitude correction.
Livestock
What can actually live in an unheated tank, which is a different list from what can survive one.
Forgiving to demanding
Which species are forgiving and which are deferred problems, scored rather than asserted.