Microgreen moisture management in Bangkok: sensor data from substrate to blackout stack

Key takeaways

A raw ADC reading of 2,350 from a capacitive moisture sensor corresponded to 41% volumetric water content in coco coir and 58% in peat in bench calibration tests -- a 17-point difference from an identical reading. Factory calibration curves applied without substrate-specific correction can be wrong by up to 28 percentage points at field capacity in peat. Plant Mood sensor log data, 2024.

Across 67 logged microgreen trays, mold incidence was 11% for trays entering blackout at or below 41% substrate moisture and 67% for trays above 44%. The risk does not climb gradually -- it spikes in a 3-percentage-point transition zone. The threshold applies to a reading taken 60 to 90 minutes after last watering, not immediately after. Plant Mood grow log data, 2024.

At 82% ambient humidity, coco coir equilibrium moisture is approximately 42 to 46% volumetric. Passive drying cannot bring the substrate below its equilibrium moisture content regardless of temperature or time. In Bangkok wet season, the 38% pre-blackout target is physically unreachable by passive drying alone -- not a technique failure, a physics constraint. Plant Mood sensor log data, 2024.

More than 50% of post-watering moisture loss in coco coir occurs in the first 8 hours at 30 degrees Celsius. After hour 8, the rate drops sharply and the substrate approaches a plateau set by ambient humidity. Evening watering with morning blackout stacking fails in wet season because the overnight window is almost entirely in the slow phase where ambient humidity governs the rate. Plant Mood sensor log data, 2024.

At 45g seeding rate, substrate moisture at day 4 averaged 61% higher than at 25g trays across 18 monitored sunflower cycles, despite identical watering. The mechanism is canopy airflow restriction -- dense seed mass creates a boundary layer of high-humidity air above the substrate that stops evaporative drying independently of water input. A fan directed horizontally at canopy height reduced hour-96 moisture by 19%. Watering reduction alone reduced it by 6%. Plant Mood grow log data, 2024.

Bangkok ambient humidity above 78% causes capacitive sensor ADC drift of 80 to 140 counts over 6 to 8 hours of continuous operation from humidity ingress into the sensor body. At the moisture-sensitive ADC range around 2,200 to 2,500, this drift corresponds to 4 to 7 apparent percentage points of false moisture increase -- enough to produce incorrect watering decisions in every direction. Plant Mood sensor log data, 2024.

TL;DR

Microgreen moisture management in Bangkok is four problems layered on top of each other, and most growers are only aware of one of them. The sensor reading is wrong if you have not calibrated for your substrate. The dry-down curve is different in wet season than the guides assume. The seeding rate is changing how fast the substrate dries in ways that have nothing to do with how much you water. And the threshold between a clean cycle and a mold event is a 3-percentage-point band that Bangkok wet season humidity systematically pushes you above. Getting any one of these right while ignoring the others does not solve the problem. The data shows they are connected.

The mold event that made me rebuild the whole measurement system

August. Eight peat trays seeded at 45g, ambient humidity running 81%. I had been using the same ADC threshold for the rewater alert on coco coir and peat for three months -- 2,300 ADC, arrived at empirically from coco coir behaviour. The peat trays triggered the alert two days earlier in the cycle than the coco coir trays. I watered when the alert fired. By day 5, two of the eight peat trays had mold at the base. Three days later, in a separate sunflower batch in coco coir at 45g, three more trays showed surface haze by the end of blackout day 3.

I logged both events the same way: substrate moisture, ambient temperature, ambient humidity, sensor readings every 30 minutes from seeding through blackout. When I put the data side by side, the same number was appearing in both failure logs at the point of entry into blackout. Not the same ambient conditions. Not the same substrate. The same moisture reading, from sensors that had never been calibrated to what that number actually meant in each substrate.

The peat alert at 2,300 ADC had been firing when the peat was at approximately 58% water content. In coco coir, 2,300 ADC was approximately 39%. I had been watering peat to near-field-capacity moisture in response to a sensor that was reading it like it was coco coir. The 45g canopy on both batches had then sealed that moisture in. The blackout had done the rest.

That weekend I ran bench calibrations on both substrates with four reference moisture points each, built substrate-specific lookup tables, and pulled 48 hours of hourly dry-down data from both. The curves were not parallel. They had different shapes, different starting points, and different plateau levels under Bangkok wet season humidity. The threshold data from the mold tray logs confirmed a 41 to 44% entry moisture step-function that I had been sitting on the wrong side of without knowing it. The sensor stack was generating data. The calibration that would have made it actionable had never existed.

The numbers that make the moisture problem concrete

ADC 2,350 = 41% water content in coco coir, 58% in peat -- from the same sensor model

Bench calibration using gravimetric water content measurement as reference. Substrate samples of 200g dry weight, water added in 10ml increments, ADC logged at each increment after 5-minute equilibration. The 17-percentage-point difference at the same ADC reading spans the range between acceptable pre-blackout moisture and active mold-risk moisture in microgreen growing. A grower applying a shared ADC threshold across both substrates is not reading moisture inaccurately -- they are reading dielectric permittivity accurately and translating it incorrectly. Plant Mood sensor log data, 2024.

Mold incidence: 11% below 41% entry moisture, 67% above 44% -- across 67 logged trays

The transition zone of 41 to 44% showed 31% incidence -- 3 times the below-threshold rate but not at the spike level of the above-44% group. Below 41%, most trays stay clean across species. Above 44%, most sunflower trays do not. The risk is not a smooth gradient -- it is a step. Plant Mood grow log data, 2024.

Plateau moisture at hour 24: 38.4% at 65% ambient humidity versus 43.1% at 82% humidity, from the same watering event

At 82% ambient humidity, coco coir equilibrium moisture is approximately 42 to 46% volumetric. The pre-blackout target of 38% is below the equilibrium moisture content for wet season Bangkok conditions. The substrate cannot passively dry to a number it is not physically allowed to reach at that humidity level. At 65% ambient humidity, the same target is achievable within 20 to 24 hours. The seasonal humidity shift changes the physics of the dry-down, not the technique applied to it. Plant Mood sensor log data, 2024.

Hour-96 substrate moisture: 39.4% at 25g seeding rate, 46.1% at 35g, 63.4% at 45g -- identical watering across all

Across 18 monitored sunflower cycles, the seeding rate variable alone produced a 61% difference in substrate moisture at day 4 without any change in watering volume or schedule. Canopy airflow restriction from dense seed mass creates a boundary layer of high-humidity air above the substrate that stops evaporative drying independently of water input. The watering is not the variable. The biomass above the substrate is. Plant Mood grow log data, 2024.

Mold onset visible by day 2 in 14 of 17 affected trays -- 36 to 48 hours post-stacking

The onset lag indicates water activity at the substrate surface crossed the fungal germination threshold of approximately 0.93 within the first day of blackout, not on day 3 or 4. By the time most growers check, the event has been running for 24 to 36 hours. The day-2 check is the only intervention window. Plant Mood grow log data, 2024.

Three things that seem like fixes and are not

Watering less does not fix a canopy humidity problem. In dense seeding, the substrate stays wet not because of water input but because the biomass above it creates a high-humidity boundary layer that stops evaporative drying. Reducing watering by 20% in a 45g tray reduced hour-96 substrate moisture by 6% in our logged comparison cycles. Adding horizontal airflow at canopy height reduced it by 19% with no change to watering at all. The moisture is trapped by the canopy microclimate, not generated by the watering volume. Cutting water input while leaving the canopy configuration intact is adjusting the wrong variable.

Denser seeding is not a mold buffer in humid conditions. The logic -- that a thicker canopy protects against environmental swings -- inverts once sensor data is introduced. At 45g, canopy-level RH in our August 2024 logged cycle ran at 91% against a room reading of 79%. The dense canopy was not buffering the humidity. It was generating 12 additional percentage points of local humidity that the room sensor never showed. All 6 mold events in the 18-cycle seeding rate dataset were in the 45g group. None were in the 25g group. The canopy that looks most protective is the one creating the microclimate the mold actually needs.

The sensor reading is not actionable without substrate-specific calibration. A reading of 43% moisture in peat is a different physical reality than 43% in coco coir. Peat has a higher dry dielectric baseline and a steeper low-moisture ADC response curve. Below a certain moisture level, peat produces higher ADC readings than coco coir at the same actual water content -- the sensor reads wetter when the substrate is drier. A single threshold applied across substrates is wrong for at least one of them in every cycle, and in Bangkok wet season where the margin between safe and mold-risk moisture is 3 percentage points, wrong by 5 to 17 points is not close enough.

In a direct comparison bench test, coco coir at 45% volumetric water content produced an ADC reading of 2,410. Peat at 38% volumetric water content -- measurably drier by weight -- produced an ADC reading of 2,480. The peat tray was reading wetter on the sensor while being drier in reality. A grower using uncalibrated readings to decide which tray needed water first would have watered the drier-reading coco coir tray and left the higher-reading but actually drier peat tray. Both decisions wrong, from a comparison that felt intuitive. Plant Mood sensor log data, 2024.

The standard responses to moisture problems in microgreen growing are: water less, seed more densely for a protective canopy, and trust the sensor reading to tell you when to act. All three are wrong in Bangkok conditions in ways the data makes visible.

Why four different moisture variables are actually one connected system

Capacitive moisture sensors measure dielectric permittivity -- the tendency of a material to store electrical energy in an electric field. Water has a dielectric constant of approximately 80. Dry coco coir is 1.8 to 2.2. Dry peat is 2.5 to 3.5, higher because of its organic matrix and retained minerals. Hemp mat sits at approximately 1.6 to 1.8 at zero water content. These differences mean the same sensor produces different ADC readings from three different dry substrates before a single drop of water has been added. Factory calibration curves built on mineral soil do not account for this. They assume a starting point that none of the common microgreen substrates matches.

The calibration error then compounds with the dry-down physics. Coco coir dries in two distinct phases: a fast evaporative phase in the first 6 to 8 hours after watering where free water in macro-pores evaporates readily from the substrate surface, and a slow capillary phase from hour 8 onward where moisture must move from micro-pores to the surface before evaporating. The transition between phases is visible in hourly sensor data as a sharp inflection in the rate of moisture loss. After the inflection, the substrate dries toward its equilibrium moisture content -- the moisture level at which the water vapour pressure in the substrate pores equals the ambient partial pressure of water vapour. That equilibrium is set by ambient relative humidity. At 65% RH, coco coir equilibrium moisture is approximately 35 to 38% volumetric. At 82% RH, it is 42 to 46%. The substrate physically cannot dry below its equilibrium moisture content by passive drying, regardless of temperature, watering schedule, or technique.

Seeding rate interacts with the dry-down curve through the same physics. Every germinating seed transpires. At 45g per tray, the aggregate transpiration from seeds and early hypocotyl growth produces a boundary layer of humidity above the substrate surface that consistently runs 8 to 14 percentage points above room ambient RH in logged cycles. In a Bangkok room already at 79% ambient, a 45g tray canopy is creating 91% local humidity directly above the substrate. The substrate drying rate in that local microclimate is governed by the 91%, not the 79%. The dry-down curve that applies to the tray is not the dry-down curve for the room conditions.

All of this feeds into the mold threshold. Pathogenic fungi require water activity above approximately 0.93 for spore germination. Water activity at the substrate surface is a function of substrate moisture content and the ambient humidity inside the sealed blackout stack. A tray entering blackout at 44% moisture in a room at 82% ambient creates a stack interior that crosses the 0.93 water activity threshold faster than a tray entering at 41% moisture in the same room. The threshold is not an arbitrary number -- it is the moisture level at which the water activity crosses the fungal germination threshold within the first day of a Bangkok blackout period. The step-function pattern in the mold incidence data reflects the step-function in fungal biology.

A cycle logged in August 2024 shows all four mechanisms operating simultaneously. Peat substrate, ADC reading at pre-blackout giving an apparent 41% on the factory curve, actual moisture 54% after substrate-specific calibration. Room ambient RH 81%, but canopy-level RH inside the 40g tray at 88%. Watered at 7pm the previous evening, stacked at 7am -- the 12-hour window was almost entirely in phase 2 where the dry-down rate was below 0.5 percentage points per hour. Mold visible at hour 42 of blackout. Each variable individually would have been flagged as borderline. All four together produced a stack that crossed the fungal germination threshold within hours of sealing.

Temperature affects dielectric constant of water -- it decreases as temperature rises. In Bangkok grow spaces where ambient temperature ranges from 27 to 34 degrees Celsius across the day and seasons, this introduces a temperature-dependent drift in moisture readings of approximately 0.5 to 1.5 percentage points. Smaller than the substrate calibration error but worth logging alongside temperature if readings are used for automated rewater decisions. The humidity drift from sensor body ingress in Bangkok above 78% RH adds a further 80 to 140 ADC counts over 6 to 8 hours of continuous operation -- independent of substrate moisture -- and requires a dry-air zeroing baseline to correct.

What the logs showed that assumptions could not have predicted

Bench calibration of the same sensor model in peat versus factory curve prediction at equivalent moisture levels. Factory curve error in peat: approximately 10 percentage points at 30% actual water content, 18 points at 50%, and 28 points at 65% field capacity. The error is not a fixed offset that shifts with moisture -- it changes shape across the moisture range. A flat correction applied at the mid-range value overcorrects the low end and undercorrects the high end. Substrate-specific multi-point calibration with 4 reference points reduced error to under 4 percentage points across the full range. Hemp mat showed a load-dependent shift of 120 to 180 ADC counts between unseeded and seeded conditions at identical gravimetric moisture, making bench calibration without seed load actively misleading for in-tray use. Plant Mood sensor log data, 2024.

Factory calibration error in peat reached 28 percentage points at field capacity -- not a constant offset

There is no correction factor or offset that fixes a factory curve for a different substrate type. The curves have different shapes, not just different positions. The calibration has to be rebuilt per substrate, not adjusted from a mineral soil baseline.

Direct comparison across paired cycles in the same week. One cycle watered at 6pm, stacked at 7am (13-hour window, mostly phase 2). One cycle watered at 6am, stacked at 10am (4-hour window, entirely phase 1). Average stack-time moisture: morning watering 42.8%, evening watering 44.6%. Both above the 38% target in wet season conditions at 82% ambient humidity, but the morning protocol was 1.8 percentage points closer to threshold despite having 9 fewer hours of drying time. The 4-hour morning window captured the fast evaporation phase. The 13-hour overnight window spent most of its duration in the slow capillary phase where ambient humidity governed the rate. Plant Mood sensor log data, 2024.

Morning watering produced lower stack-time moisture than evening watering in 16 of 18 wet season cycles

The conventional logic of watering the night before to allow overnight drying inverts in wet season. The intervention that actually changes pre-blackout moisture is not adding more drying time -- it is placing more of the existing drying time inside the fast phase-1 window that closes 6 to 8 hours after watering.

Across 4 comparison cycles at 82% ambient humidity, 45g seeding rate, identical initial watering volume. Fan positioned to direct airflow horizontally across tray surface at canopy height, running for the first 8 hours after watering only. No-fan control cycles run simultaneously. Hour-96 moisture: fan cycles averaged 51.2%, no-fan control averaged 63.4%. Separate comparison of watering volume reduction (20% less water) without fan in 45g trays produced hour-96 moisture of 59.6% -- a 6% reduction. The mechanism is boundary layer disruption during phase 1. Running the fan after hour 8, during phase 2, produced no detectable moisture reduction. Plant Mood grow log data, 2024.

A USB fan at canopy height reduced hour-96 substrate moisture by 19% in 45g trays; watering reduction alone produced 6%

Fan position and timing both matter. A fan running overnight at shelf level after the phase-1 window has closed does not change substrate moisture. A fan directed at canopy level during the first 8 hours after watering does. Growers who have tried fans without result are almost certainly running them at the wrong position or the wrong time.

Sensors tested in open air at 82% ambient humidity with no substrate contact. ADC readings at hour 0 and hour 8. Drift of 80 to 140 counts depending on sensor unit -- some units in the same batch showed more humidity ingress than others. At the moisture-sensitive ADC range of 2,200 to 2,500, 100 counts of drift corresponds to approximately 4 to 7 apparent percentage points of moisture increase. In Bangkok wet season this drift is continuous rather than intermittent. Evening sensor readings taken after 6 to 8 hours of operation systematically overstate substrate moisture relative to morning readings from the same sensor in the same substrate. A grower comparing morning and evening readings without correcting for drift is comparing two different calibration states of the same sensor. Plant Mood sensor log data, 2024.

Sensor ADC drift of 80 to 140 counts over 6 to 8 hours at Bangkok ambient humidity was independent of substrate moisture

Consistent measurement timing is as important as substrate calibration for reliable moisture decisions. Taking all reference readings at the same time of day -- morning, before the drift accumulates -- is a minimum requirement. A dry-air zeroing baseline at installation provides the offset correction needed to account for drift in logged data.

Species-specific breakdown from the 67-tray expanded dataset. Below 41%: sunflower 12%, radish 18%, pea shoots 9%. Above 44%: sunflower 71%, radish 44%, pea shoots 58%. The threshold moisture level where incidence spiked was consistent across species at 41 to 44%. What differed was the sensitivity above it. Radish is the most forgiving species above threshold. Sunflower is the most sensitive. A grower running mixed-species production cannot use radish performance to set moisture thresholds for sunflower. Plant Mood grow log data, 2024.

Sunflower mold incidence above 44% entry moisture was 71%, radish was 44% -- same threshold, different consequence

A single 41% threshold applied across all species is more conservative than radish strictly requires but appropriate for sunflower and pea shoots. Mixed-species operations benefit from applying the most sensitive species' threshold across the board rather than managing per-species thresholds on every tray.

Plant Mood grow log data and sensor log data, 2024

The complaints that trace back to moisture management before anything else

'My mold problem is random. Some cycles are clean and some are not and I cannot find a pattern. Same seeds, same substrate, same technique.'

Randomness in mold outcomes is almost always moisture variation that has not been measured. Pre-blackout substrate moisture in an unmonitored operation varies by 6 to 10 percentage points across cycles depending on watering timing, ambient humidity at drying time, and room conditions during the drying window. Some cycles land below 41%. Some land above 44%. The apparent randomness resolves when moisture is logged per cycle. The technique did not change. The moisture outcome of that technique varies with conditions the grower is not tracking.

'My peat trays trigger the moisture alert way earlier than my coco coir trays. I assume peat dries faster. I water sooner and the mold gets worse.'

The alert is not detecting faster drying. It is detecting a lower ADC reading in peat that corresponds to wetter substrate than the same ADC reading in coco coir. Peat's steeper low-moisture ADC response curve means it exits the alert zone faster on the way up from dry -- but the substrate is already wet by coco coir standards. Watering in response to the early alert is adding water to a substrate that does not need it, then covering it with a blackout stack. The mold follows from this sequence directly. The fix is substrate-specific calibration, not a changed watering response.

'I water the same way for cool season and wet season but I only get mold in wet season. I must be doing something wrong specifically in those months.'

Nothing is wrong with the wet season technique. The dry-down physics changed. At 82% ambient humidity, coco coir equilibrium moisture is 42 to 46% -- above the 41% mold threshold. The cool season process that reliably produces below-41% entry moisture does the same in wet season but arrives at a higher plateau because ambient humidity constrains how far the substrate can passively dry. The technique is identical. The moisture outcome of that technique is different. The fix is seasonal adjustment of watering timing and active air circulation during phase 1, not a change in core process.

'I upgraded to high-germination seeds and premium coco coir and the mold got worse, not better.'

Premium coco coir with higher water retention capacity holds moisture longer than lower-quality substrate with larger particle size and faster drainage. Higher-germination seeds have higher metabolic activity during blackout, producing more respiratory moisture inside the sealed stack. Both upgrades increase effective moisture in the mold-risk zone without any change in watering technique. Better inputs require a lower entry moisture threshold to achieve the same mold outcome as standard inputs. The system changed. The moisture management protocol did not.

'I added a moisture sensor and I am hitting below 41% before every stack. I am still getting mold in about one cycle in five.'

Three possible causes, all related to what the surface sensor does not measure. First: heavy initial watering that dried from the top down leaves below-threshold surface moisture but above-threshold moisture at 4 to 5 cm depth where early root development occurs. Second: seed hull moisture from extended soaking creates a wet local environment at the stem base regardless of substrate reading. Third: stack height is creating an enclosed microenvironment where collective humidity from multiple trays raises effective water activity above what individual tray moisture predicts. A below-threshold surface reading reduces risk significantly. It does not eliminate the variables the sensor cannot see.

'My dense sunflower trays look perfect on day 2 and then collapse by day 5. Day 2 germination is excellent. Day 5 is mold.'

Day 2 germination is largely a seed-quality metric. The collapse is a microclimate metric. A dense tray can show excellent early germination because the seed mass creates warmth and moisture retention that drives rapid sprouting -- but those same conditions become the problem from day 3 onward when the canopy closes and the substrate beneath it stops drying between waterings. At 45g, canopy-level RH in logged cycles averaged 11 percentage points above room RH. The tray is not in the same humidity environment as the room. It is in its own microclimate that the room sensor will never show.

How to build a substrate-specific moisture calibration in under two hours

Weigh 200g of your substrate at bone-dry state and record the ADC reading

A logged dry ADC baseline for this substrate. Coco coir typically sits around 1,800 to 1,850. Peat around 1,940 to 1,970. Hemp mat around 1,720 to 1,760. If your reading sits outside these ranges, record it regardless -- your specific substrate or sensor model may differ and your curve will account for it.

The dry baseline is the most important reference point in the calibration. It tells you where zero water content sits on your sensor's ADC scale for this specific substrate. Bone-dry means oven-dried at 60 degrees Celsius for 4 hours or left in direct sunlight for 6 to 8 hours until weight is stable. Pack the substrate consistently against the sensor face at the depth you will use in production. Inconsistent packing produces inconsistent readings at the same moisture level.

Add 20ml of water to the sample, pack consistently, wait 5 minutes, log ADC

A second calibration point at approximately 10% volumetric water content for coco coir, lower for peat. Log both the water volume added and the ADC reading.

The 5-minute equilibration wait is not optional. Immediately after adding water, surface moisture produces an artificially high reading. The equilibration period allows water to distribute through the substrate matrix and gives a reading representative of the actual moisture state rather than the surface condition. Consistent packing pressure and sensor insertion depth at every step is what makes the curve usable -- variation in either produces scatter that cannot be distinguished from genuine moisture variation.

Repeat water additions in 20ml increments, logging ADC at each step, until the substrate is visibly saturated and free water appears at the surface

A full set of ADC-to-moisture-volume data points from dry to saturated. Convert the cumulative water volume at each step to volumetric water content using the sample weight: (ml water added / ml total substrate volume) x 100.

The full curve from dry to saturated takes approximately 8 to 10 additions for a 200g coco coir sample, fewer for peat or hemp mat which saturate faster. You are capturing the shape of the ADC response across the full moisture range, not just a few points. Peat will show a steep initial rise followed by a flat plateau. Coco coir will show a more linear rise. Hemp mat will show irregularity depending on fibre density and weave. Each pattern is specific to the substrate and tells you where the sensor has good resolution and where it does not.

Plot the curve and identify your four critical reference points

A four-point lookup table with ADC values and corresponding volumetric water content percentages. This is the calibration that replaces the factory curve for every moisture decision in this substrate.

The four points that cover the grow-relevant moisture range are: bone dry (0%), approximately 30% volumetric water content (germination-risk low end), field capacity (typically 60 to 70% for coco coir, 65 to 75% for peat), and saturated. These four points are sufficient to build a piecewise linear interpolation table that converts any ADC reading to a moisture estimate with under 4 percentage points of error in coco coir. Peat requires attention to the flat upper range -- sensor resolution above 55% water content is poor in peat, and the calibration should flag this range as low-confidence.

Establish your sensor's humidity drift baseline and log a correction offset

A humidity drift offset in ADC counts for the current season. Apply this as a correction to any reading taken after more than 4 hours of continuous sensor operation in ambient humidity above 75%. Take all critical pre-blackout readings in the morning before drift accumulates, or immediately after a sensor power cycle.

Bangkok ambient humidity above 78% causes capacitive sensor ADC to drift upward by 80 to 140 counts over 6 to 8 hours of operation from humidity ingress into the sensor body. This drift is independent of substrate moisture and compounds the substrate calibration error if not corrected. The drift offset is measured by sealing the sensor in a container with desiccant for 30 minutes, logging the stabilised reading, then operating the sensor in open air at current ambient humidity for 8 hours and logging the reading again. The difference is your drift offset for the current ambient conditions. Repeat this measurement at the start of wet season and dry season -- the drift magnitude changes with seasonal humidity.

What to build into every moisture management decision in Bangkok

All Bangkok microgreen growers using capacitive sensors

Run substrate-specific calibration for every substrate type you use before building any moisture thresholds or automation triggers. A calibration built on coco coir does not transfer to peat, a coco-peat blend, or hemp mat. A blend requires its own calibration because the ratio of the two substrates determines where the curve sits. The two-hour bench process is the prerequisite for every moisture decision that follows. Without it, the sensor is generating numbers, not information.

Substrate-specific ADC-to-moisture lookup tables with under 4 percentage points of error across the operating range for coco coir and peat. Hemp mat calibration should be performed in situ after seeding with a simulated seed load, not from a bench test on unsupported mat.

Growers managing pre-blackout moisture targets

Take the pre-blackout reading 60 to 90 minutes after last watering, not immediately after. Set 41% as the hard stacking threshold for sunflower and pea shoots, 43% for radish. Do not stack any tray above these values regardless of schedule pressure. Log ambient humidity at the time of every reading -- at 82% ambient, reaching 41% by passive drying is not achievable and the protocol needs to account for that before the reading, not after.

Mold incidence below 15% for sunflower and pea shoots, below 20% for radish, consistent with the 67-tray logged dataset. Not zero -- seed hull moisture, spatial variation within the tray, and stack height can produce occasional events below threshold -- but substantially below the 67% incidence at above-44% entry moisture.

Growers in Bangkok wet season (ambient humidity above 78%)

Switch to morning watering 4 to 6 hours before stacking rather than evening watering. Run a small USB fan directed horizontally across the tray surface at canopy height for the first 8 hours after watering, then turn it off before stacking for blackout. The fan disrupts the boundary layer of humidity above the substrate during the fast evaporation phase, reducing plateau moisture by 4 to 6 percentage points at 82% ambient humidity. Running the fan after hour 8 produces no detectable moisture reduction -- the phase-1 window has closed.

Stack-time moisture 3 to 5 percentage points lower than the evening-watering protocol produces at the same ambient humidity. Combined with morning watering, this brings wet season pre-blackout moisture into the low-40s for most cycles -- still above the cool season target but below the high-risk zone if the threshold protocol is followed.

Growers experiencing moisture problems despite below-threshold sensor readings

Check three variables the surface sensor does not measure. First: sensor depth relative to where root development occurs -- a sensor at 2 cm depth reads surface moisture and will miss above-threshold moisture at 4 to 5 cm depth in a tray dried from the top down after heavy watering. Second: seed hull moisture from extended soaking, which creates a wet local environment at the stem base regardless of substrate reading. Third: stack height -- reduce from 6 trays to 4 trays for sunflower in wet season to reduce the enclosed microenvironment humidity amplification effect.

Identification of the specific hidden moisture source producing mold events despite below-threshold surface readings. In logged destructive sampling, 4 of 6 cycles with below-threshold surface readings and mold events showed above-threshold moisture at 4 to 5 cm depth.

Growers adjusting seeding rates for moisture management

For Bangkok ambient conditions above 78% RH without active air conditioning, use 35g as the maximum seeding rate for sunflower. At 45g, the boundary layer effect from canopy biomass produces substrate moisture at day 4 that is 61% higher than at 25g with identical watering -- well above the mold threshold regardless of what the watering schedule does. If the operation requires 45g for commercial presentation, the only viable path is an air-conditioned grow space held at 24 to 26 degrees Celsius and RH below 65%. The canopy microclimate at 45g in Bangkok ambient conditions is not manageable by watering adjustment alone.

At 35g with the morning-watering and phase-1 fan protocol, substrate moisture at day 4 averages 46.1% -- in the transition zone but below the high-incidence spike. Combined with the 41% pre-blackout stacking threshold, most cycles stay below the mold threshold even in wet season conditions.

Growers building logged datasets for threshold calibration

Log pre-blackout moisture and harvest outcome for every cycle, noting species, substrate, seeding rate, ambient humidity at stacking time, and stack height. After 15 to 20 cycles with both data points, plot moisture versus outcome and confirm whether the 41% threshold holds in your specific setup. Sensor calibration, substrate brand, ambient conditions, and stack configuration all shift the threshold slightly. Your own 20-cycle dataset is the calibration that applies to your operation. The 41% figure is the starting point derived from our data. Your logged data is the refinement.

A grow-space-specific moisture threshold calibrated to your substrate, sensor, seeding rate, and ambient conditions. After 20 cycles, the clean versus mold step function should be visible in the data and the transition zone should be identifiable for your specific combination of variables.

The substrate-specific calibration at sensor installation. It takes 2 hours per substrate and is the reference every moisture threshold and automation trigger depends on. Most growers install the sensor, observe that the numbers change when they water, and conclude the sensor is working. It is working -- it is reading dielectric permittivity accurately. Without calibration, those readings are not being converted to moisture content accurately. The skip is invisible because the sensor appears functional. The consequence appears as unexplained mold events and incorrect watering decisions across every subsequent cycle.