Case Study: A Sealed Bottle Garden at Eight Years

The original build

A ten-litre glass demijohn with a four-centimetre neck, planted through a paper funnel and a set of extension tools made from bamboo canes. Drainage layer of three centimetres of expanded clay, a fleece barrier, then eight centimetres of a coir, bark, pumice and charcoal mix, sloped from three centimetres at the front to eleven at the back.

Original planting: Hypnum cupressiforme carpet, two Leucobryum cushions, one Fittonia albivenis, one Pilea depressa, a small Asplenium fern and a scatter of oak leaf litter. Hardscape was three pieces of the same dark slate, inserted through the neck one at a time and positioned with bamboo rods, which took an entire evening.

It was left with the neck loosely covered for a year to allow the moisture balance to settle, then corked and sealed with a wax collar in late 2019. Since then it has been opened twice: once in 2021 to remove a dead fern, and once in 2024 to trim growth that had reached the glass across the whole front.

Year one to three: the shake-out

The first year saw the losses. The Asplenium fern declined slowly over eighteen months, browning frond by frond, almost certainly from insufficient light at the bottom of a tall vessel where the moss above it had grown to shade the floor. It was removed in 2021, the only significant intervention in the build's history.

One Leucobryum cushion yellowed at the margin during the second winter and slowly shrank to about a third of its planted size. It stabilised there and has persisted, pale and compact, ever since. The other cushion has done the opposite and roughly doubled.

The Fittonia, which most guidance treats as a short-lived terrarium filler, turned out to be the most durable plant in the vessel. It has died back to the stem twice and regrown both times, and currently occupies a quarter of the floor area. The Pilea grew vigorously for two years, reached the glass, and has since existed in a permanent cycle of growing to the wall, browning back and regrowing.

Year four to eight: equilibrium and self-generated litter

The most striking development is the litter layer. Eight years of shed Fittonia leaves, dead Pilea stems and senesced moss have built a genuine detritus layer roughly a centimetre deep across the floor of the bottle. It is being processed by a springtail population that arrived accidentally in the original substrate and has been visible ever since — a fine white dusting across the glass at the substrate line on humid mornings.

That litter layer is doing real work. It is the nutrient cycle of the system, returning what the plants take up, and its presence explains why a sealed vessel with no fertiliser input has not become progressively nutrient-poor over nearly a decade.

The water cycle has proved extraordinarily stable. The bottle fogs on its western side each afternoon and clears by morning, exactly as it did in 2020. The drainage reservoir has a visible water level that has moved perhaps a centimetre across eight years. There is no evidence of the seal leaking, and no water has been added since it was corked.

What it looks like now

Honestly assessed: it is not a beautiful composition any more. The careful slope is largely hidden under growth, two of the three slate pieces are invisible, and the Fittonia and Pilea have overrun the deliberate negative space that made the original build read as a landscape. What was designed as a miniature woodland has become a dense, tangled green mass.

It is, however, fascinating. The glass is clean, the growth is vigorous, the springtails are active, and the whole thing has needed roughly one hour of attention in eight years. As a demonstration of a functioning closed system it is completely convincing.

The tension between these two assessments is the real finding. A sealed bottle garden is a biology experiment that happens to be attractive for its first two or three years. If you want a composition that stays legible, you need access for trimming — which means a wide aperture and a lid you can remove, which means accepting a slightly less romantic seal.

What would be done differently

First, a wider neck. Every regret about this build traces back to the four-centimetre aperture. A twelve-centimetre opening with a fitted glass lid would have allowed twice-yearly trimming and the composition would still be readable.

Second, no fast-growing vascular plants. The Fittonia and Pilea are why the design disappeared. A moss-only planting, or moss with a single slow fern, would have held its shape for far longer. In a sealed vessel, vigour is a liability rather than a virtue.

Third, more litter from the start. The self-generated detritus layer turned out to be the engine of the system's stability, and seeding a deeper leaf litter at build time would likely have accelerated the establishment of the springtail population and the nutrient cycle by a year or more. That is the one thing this build proved that most short-term terrarium advice never mentions: in a closed system, the decomposition pathway is not a side effect. It is the design.

Extended notes

• David Latimer's sealed spiderwort, planted in 1960 in a ten-gallon carboy and watered only once, remains the standard reference for the longevity of closed systems.
• Oxygen in a sealed vessel cycles daily — produced by photosynthesis in the light period, consumed by respiration and decomposition at night. The system stays viable because the two processes balance over time, not because oxygen accumulates.
• Wax-sealing a cork is reversible with gentle heat from a hairdryer. Do not use adhesive; you will eventually want to open the vessel and a glued cork usually means breaking the glass.

Frequently asked