What a mold event looks like in sensor data before you can see it: 8 documented incidents

Key takeaways

In 7 of 8 documented mold incidents, relative humidity at tray level exceeded 91% for more than 4 consecutive hours before any visible mold appeared — Plant Mood grow log data, 2023-2024.

Substrate moisture readings above 72% on day 3 after blackout removal correlated with mold onset in 6 of 8 incidents, regardless of variety.

Visible mold appeared on average 22 hours after the sensor threshold breach, meaning intervention during the data window would have saved the tray in most cases.

Airflow absence — not humidity alone — was the shared condition in every single one of the 8 incidents. Rooms with a fan running did not produce a mold event even when humidity peaked above 88%.

Pea shoots and sunflower were involved in 6 of the 8 incidents. Both hold moisture in the seed hull longer than fine-seed varieties, extending the window of risk.

The smell threshold — that faint sourness before mold is visible — arrived an average of 8 hours after the sensor breach and 14 hours before visible growth. It is a mid-stage signal, not an early one.

TL;DR

Mold does not arrive without warning. Across 8 incidents we logged in Bangkok, the sensor data shifted before the tray did — every time. Humidity at tray level, substrate moisture on day 3, and the absence of moving air were the three variables that predicted every event. If you are waiting to see the mold, you are already 22 hours too late. The signals are readable. You just have to be measuring the right things at the right height.

The tray that looked fine at 9pm

Day 5 of a sunflower grow. The blackout dome came off on schedule. The seedlings were standing upright, hulls mostly shed, cotyledons beginning to open. Nothing looked wrong. I watered lightly, the way I always do after blackout, and left the tray on the rack. By 7am the next morning there was white fuzz across the entire left half of the tray. Full loss. That was incident number three. After it happened, I went back to the sensor log and looked at what was happening at 9pm the night before when I thought the tray was fine. Relative humidity at tray level: 93%. Substrate moisture: 74%. Fan: off. The fan had tripped a switch at some point that afternoon. I had not noticed.

That was the incident that made me start treating the sensor log as a predictive document rather than a record of what had already happened. The data was showing me a mold event in progress. I was looking at the tray instead of the screen.

Mold is not a watering mistake

The standard advice is that mold in microgreens comes from overwatering. Water less, mold less. Most guides stop there.

Across our 8 incidents, watering volume was not the distinguishing variable. In 5 of the cases I had watered the same amount I always water for that variety at that stage. The distinguishing variable was whether air was moving across the tray surface after watering. A tray watered at 80% of normal volume with no airflow produced mold. A tray watered at 120% of normal volume with a fan running at low speed did not. Moisture that evaporates does not become mold. Moisture that sits at the surface in stagnant humid air does.

Plant Mood grow log data, 2023-2024. 8 incidents reviewed. Fan state logged at time of mold onset for each incident: 8 of 8 had fan off or obstructed. Watering volume alone could not predict outcome.

The numbers that changed how I read a tray

The average gap between the first sensor threshold breach and visible mold across 8 documented incidents. That is the intervention window most growers do not know they have — Plant Mood grow log data, 2023-2024.

91% RH for 4+ hours

The humidity threshold at tray level that preceded mold onset in 7 of 8 incidents. Not ambient room humidity — tray-level humidity, which regularly runs 6 to 11 percentage points higher in a dense canopy.

72% substrate moisture on day 3

Substrate moisture reading after blackout removal that correlated with mold in 6 of 8 incidents. Below 65% on day 3: zero mold events in the log regardless of ambient humidity.

Average time between sensor breach and detectable smell — the sourness that experienced growers recognize before they see anything. The smell is a mid-stage signal, not an early warning.

What the 8 incidents actually showed

In every incident, we logged both ambient room humidity and a sensor positioned at canopy height inside the grow rack. The average divergence was 8.4 percentage points. A room reading of 80% RH — which feels manageable — corresponded to 88.4% at tray level. At 82% ambient, the tray was already over 90%. Growers who monitor only room humidity are reading the wrong number.

Tray-level humidity diverges significantly from room humidity

Room humidity targets published in grow guides are calibrated for temperate climates and ambient sensors, not for Bangkok apartments with dense tray canopies. A 'safe' ambient reading here is not safe at tray level.

Of the 8 incidents, 6 had their sensor threshold breach on day 3 after dome removal — not day 1 or 2. On days 1 and 2, the canopy is shorter and airflow reaches the substrate more easily. By day 3, the canopy is dense enough to trap humidity at the surface. Watering on day 3 into a stagnant room is the highest-risk action in the entire grow cycle.

Day 3 post-blackout is the highest-risk window

Day 3 post-blackout needs active airflow management, not passive. This is the day to run the fan before watering, not after.

Six of 8 incidents involved pea shoots or sunflower. Both varieties retain moisture in the seed hull significantly longer than fine-seed crops like radish or broccoli. In a humid Bangkok environment, that hull moisture does not fully dry between waterings if air is not moving. The hull becomes the initial mold site, not the substrate.

Pea and sunflower carry risk specific to seed hull moisture retention

For pea and sunflower in Bangkok conditions, hull dryness matters as much as substrate moisture. Visual inspection needs to include the seed layer, not just the canopy.

Temperature, watering volume, seed density, and substrate type all varied across the 8 incidents. Fan state did not. Every single incident occurred with the fan off or physically obstructed. This held even in the two incidents where ambient humidity was relatively moderate — 78% and 81% respectively.

Fan state was the only variable present in all 8 incidents

Airflow is not one factor among many. It is the factor that makes the other variables manageable or dangerous. A grow space without guaranteed airflow is structurally exposed regardless of how carefully you water.

Plant Mood grow log data, 2023-2024

The tray that smelled fine went first

Smell is a lagging indicator, not a leading one. In our log data, detectable odour appeared an average of 8 hours after the sensor threshold breach and 14 hours before visible mold. That means the biological process that produces mold was already running for 8 hours before your nose registered anything. In a Bangkok apartment where you are not sleeping in the grow space, those 8 hours are probably the hours between your last check and your morning visit. By the time it smells wrong, half the intervention window is already gone.

Incident six was the clearest example. I checked the tray at 10pm, smelled nothing, went to bed. Sensor log shows the humidity breach happened at 11:30pm. I checked again at 6:30am. Visible mold across the stem base of a full pea shoot tray. The smell I would have caught had I checked at 7am was gone — overwhelmed by the stronger smell of active mold. The sensor showed the problem. I was not reading the sensor.

Most growers treat smell as their primary early-warning system. If it smells off, check the tray. If it smells clean, the tray is fine. This feels reliable because smell is immediate and sensor-free.

Why the 22-hour window exists and what closes it

Mold spores are already in your grow space. They are on the seeds, in the air, on every surface. What determines whether they colonise a tray is whether conditions cross the threshold for germination and hyphal growth simultaneously. For the species most common in warm-humid indoor environments — primarily Botrytis, Fusarium, and Rhizopus in Bangkok conditions — that threshold is roughly 90% relative humidity sustained for long enough to allow spore germination, combined with a substrate that is not drying between waterings. The 22-hour window exists because spore germination and early hyphal growth are not visible. The mycelium is building before there is anything to see. Visible mold is late-stage colonisation, not early-stage.

In Bangkok, the mechanism accelerates compared to temperate-climate grow guides because ambient temperature sits between 28 and 34 degrees Celsius without air conditioning — which is the optimal growth range for most warm-climate mold species. A mold event that might take 48 hours to become visible in a 20-degree Celsius European grow room can reach the visible stage in 18 to 22 hours here. The grow guides written for temperate climates are describing a slower process.

The substrate moisture reading matters because it reflects the water activity at the medium surface — the number that determines whether the spore has the free water it needs to germinate. A tray reading of 72% moisture in our coco coir substrate corresponds to water activity conditions that most common mold species can exploit. Below 65%, germination slows significantly. This is why the day-3 moisture threshold matters: it is not about aesthetics or root health, it is about the biological threshold that separates dormant spores from active colonisation.

What growers say went wrong and what actually went wrong

'I barely watered it. I cannot figure out why it molded.'

Watering volume is rarely the proximate cause. The question is not how much water went in but how much evaporated before the next watering cycle. In a Bangkok room with no airflow and 80%+ ambient humidity, even a light watering on a dense canopy does not dry between cycles. The substrate holds the moisture because the air above it is already saturated. Reducing watering volume without fixing airflow just slows the process slightly.

'It happened so fast. The tray was fine yesterday.'

It was not fine yesterday. The sensor data shows it was not fine. The visual inspection showed nothing because mold in the 8 to 20 hour range is not visible to the naked eye. 'Fine yesterday' usually means 'not visibly colonised yesterday.' The conditions for the event had already been present for hours. Bangkok temperatures compress the mold development timeline relative to temperate-climate advice, which creates the impression of sudden onset when the process was running on schedule.

'I have a fan but it still molded.'

Fan position is as important as fan presence. In three of our incidents where a fan was technically running, the airflow was directed at the room wall, not across the tray canopy. In a grow rack with multiple shelves, air from a floor fan does not reach upper shelves without deliberate positioning. 'Fan running' and 'airflow across the tray surface' are not the same condition. The sensor on the upper shelf showed stagnant humidity even while the lower shelves were adequately ventilated.

'My humidity is fine — I checked.'

Where the sensor is sitting matters more than the reading itself. A wall-mounted hygrometer or a sensor sitting on a shelf above the trays reads ambient room air, not canopy-level air. In a dense tray canopy, relative humidity 5 centimetres above the substrate surface can be 8 to 12 percentage points higher than what a wall sensor shows. The reading is accurate for the air where the sensor lives. It tells you nothing about the air where the mold is deciding whether to grow.

Intervention at different points in the 22-hour window

Intervention at hour 0 to 6 (sensor breach, no smell, no visible sign)

Fan on immediately, reduce ambient humidity if possible, do not water again until substrate drops below 60% moisture. Spore germination is in early stages. The hyphal network has not established. Drying the surface environment at this stage disrupts germination before colonisation takes hold.

Tray is recoverable in the majority of cases. This is the window that sensor monitoring opens and visual inspection cannot. In 5 of our 8 incidents, retrospective analysis showed that intervention at this stage would have prevented loss.

Intervention at hour 8 to 14 (sensor breach ongoing, faint smell, no visible mold)

The smell means early hyphal growth is producing volatile compounds. The mycelium is established but not yet surfacing. Fan on, reduce humidity, consider removing the tray from the rack to isolate it from adjacent trays. Misting with diluted hydrogen peroxide at 0.5% can interrupt surface colonisation at this stage in our experience — though results varied across incidents.

Partial tray recovery is possible in some varieties. Radish and broccoli responded better to intervention at this stage than pea or sunflower, which had mold rooted in the seed hull rather than the substrate surface.

Intervention at hour 18 to 22 (visible mold, full colonisation underway)

Visible mold means the mycelium has broken the surface. At Bangkok temperatures, sporulation can begin within a few hours of visible growth. The primary concern at this stage is not saving the tray — it is preventing spore dispersal to adjacent trays. Remove the affected tray from the rack immediately. Do not water adjacent trays until airflow has run for at least two hours.

Tray is a loss in all 8 of our documented cases at this stage. The variable that matters now is containment. A single mold event that spreads to two adjacent trays has tripled the cost of the failure.

What we changed after incident eight

Any grower running pea, sunflower, or other large-seed varieties

Position a humidity sensor at canopy height inside the grow rack, not at room level. If this is not possible, add 8 percentage points to whatever your wall sensor reads and treat that as the tray-level figure.

You stop making decisions based on air that is not in contact with your tray. Ambient room humidity in Bangkok is a misleading number for mold risk assessment.

Any grower in a Bangkok apartment without dedicated climate control

Run airflow across tray surfaces for a minimum of 30 minutes after every watering, before domes or covers go back on. Not at the room wall. Across the canopy surface. The 30 minutes is the period of peak moisture vulnerability.

Surface moisture dries before it has time to create the sustained high-humidity microenvironment that spore germination requires. This single change eliminated mold events in our operation for 14 consecutive cycles after incident eight.

Any grower on day 3 post-blackout with pea or sunflower

Check substrate moisture before deciding whether to water. If the reading is above 65%, skip the watering entirely and increase airflow instead. Day 3 is not the day to water on schedule — it is the day to water based on the actual moisture state of the substrate.

You break the pattern where day-3 watering into a saturated dense canopy creates the exact conditions the sensor data identified as the primary risk window.

Any grower who monitors the grow space only by visual inspection

Add a smell check as a second-stage signal — but understand it is mid-stage, not early. If you detect the sourness before visible mold, you have roughly 14 hours before the tray is a visible loss. Use them: fan on, isolate the tray, reduce humidity.

You recover some trays you would otherwise have found fully colonised in the morning. The smell check does not replace sensor monitoring but it is better than visual-only.

Checking fan position after moving trays. Every time a tray gets repositioned on the rack, the airflow geometry changes. A fan that was effectively ventilating one configuration may be directing air past the tray entirely after a reshuffle. We confirmed this with spot humidity readings in three incidents where the fan was running but the upper shelf was still showing 91% at canopy level.

The signal was always there

Every one of these 8 incidents was documented after the loss. But the data that would have prevented the loss was available before it. That is the uncomfortable part. The sensor logs showed the breach. The problem was that nobody was treating the sensor as a decision-making tool in real time. This week, if you are growing pea or sunflower in a Bangkok apartment: position something that measures humidity at canopy height, not room height. It does not need to be sophisticated. A 200 baht digital hygrometer sitting on the rack shelf is measuring the right air. Your wall unit is not. That one change is what shifted the mold event from 'I cannot figure out what went wrong' to 'I can see exactly when it started and what I should have done.' Start there.