One Air, Two Appetites

The general claim holds

That a technology needs to be kept inside a moisture envelope is not a metaphor borrowed from horticulture. It is written into the standard. ASHRAE’s recommended envelope for data equipment is a dew point of 5.5–15 °C with relative humidity at or under 60%, and the allowable range widens only under conditions and classes.

What makes it a genuine envelope rather than a ceiling is that the two edges fail differently. Too dry and you get electrostatic discharge, which kills components outright. Too damp and you get condensation, short circuits, and corrosion — and where copper and silver corrosion is a concern the ceiling comes down further still, to 60% or below 50%. Neither failure is a lesser version of the other. They are separate mechanisms that happen to bound the same variable.

The same quantity is necessary and destructive at once. What separates the two is dose — which is this site’s window principle, arriving in a machine room.

And the plants really do cool

Not marginally. Measured green roofs run surface temperatures far below conventional ones — reductions on the order of 20–30 °C on the roof surface in hot conditions, with nearby air several degrees cooler, and cooling-load reductions reported from about 15% up to 70% depending on climate, plant type and construction. The mechanism is not mysterious: shade, plus evapotranspiration converting incident radiation into latent heat instead of surface temperature.

So the thermal half of the proposal is well supported. A living layer is a real cooling technology, and for a system whose hardware is trying to shed heat, that is a straightforward benefit.

The correction: VPD is the plant’s variable

Vapor pressure deficit measures how hard the air pulls water out of a wet surface. That is exactly the right quantity for a leaf, for moss, and for skin — which is why the knot already computes it three ways, anchoring plant VPD at air temperature and the human gradient at skin, 35 °C, the ISO 7933 quantity. It is the right variable when the thing you care about is losing water.

Dry hardware is not losing water. Its failure modes are condensation, which is governed by dew point and how close a surface sits to it, and static and corrosion, which track relative humidity. Those are not restatements of VPD; they answer different questions about the same air. A machine does not care how hard the air could pull moisture out of it. It cares whether moisture will arrive on it, and whether there is enough moisture around to bleed a charge.

So the claim survives with its variable swapped. Every technology has an operational moisture envelope — true, general, and load-bearing. That the envelope is best expressed as VPD — true for anything evaporating, and a category error for a circuit board. Same physics, different natural variable, because the failure is different. This site made the same mistake once in the other direction and wrote it down: a threshold travels with a quantity, not with a name.

The two appetites

Now the part that makes stacking them a design problem rather than a synergy.

The previous note established that moss is bound by water and not by light. This one closes the loop badly: the transpiration that keeps the living layer alive is a humidity source pushing the machine toward its upper failure edge. Put them in one air and the plant’s health and the hardware’s health are in direct competition, with the plant winning locally and the corrosion arriving slowly enough that nobody attributes it correctly.

They do agree about temperature. Both want cooler. The disagreement is entirely about water.

Which gives the actual design rule

Couple the heat, decouple the water. The plants belong on the outside of a vapour barrier, doing latent cooling where their humidity vents to the sky rather than into the equipment volume. Take the temperature benefit, refuse the moisture. Almost every working green-roof-over-occupied-space is already built this way, and for exactly this reason.

What that forfeits is worth naming: the moss then has to get its water from rain and irrigation rather than from a shared humid volume, which is the expensive part and the part the original idea was implicitly solving. There is no free version where one air serves both.

Plants are a buffer, not a controller

One more limit, because “regulate” is doing heavy lifting in the proposal. A controller has a setpoint and acts against error. Evapotranspiration has neither: it runs hardest when it is hot and dry, which is when the reservoir is emptying, and it stops when the water is gone — which is precisely when the buffering was most needed. It is driven by the same weather it is meant to damp.

That makes a planted layer a genuine thermal mass and latent buffer, and not a regulator. It narrows the swings; it does not hold a value. Any envelope that must stay inside ASHRAE’s band still needs mechanical control, and the plants reduce the load on it rather than replacing it. Claiming otherwise would be the same over-reach as calling resonance a second energy source: real effect, wrong job title.

Rests on: ASHRAE TC9.9 thermal guidelines for data equipment (recommended 5.5–15 °C dew point, ≤60% RH; ESD at the dry edge, condensation and copper/silver corrosion at the damp edge); measured green-roof thermal studies reporting surface-temperature reductions of tens of degrees and cooling-load reductions from ~15% to ~70% across climates; and the standard poikilohydric account of bryophyte water relations. Offered, not proven: that the couple-heat / decouple-water rule is the right architecture for this particular stack — it is the conventional answer, and nothing here has been built or measured.

Kin to The Green Window (what the fibre keeps), Insensible (anything warm and wet in unsaturated air is losing water), The Dose Window (the same stressor, either side of a line), and the knot, which already reads one air three ways.