Temperature mapping: how many sensors, where, and for how long

Two questions get confused constantly, and confusing them is expensive.

Mapping asks: where in this space are the warmest and coldest points, under realistic conditions? It is a time-bounded study with many sensors, run once at qualification and repeated when something changes.

Monitoring asks: are the points that matter staying in range, right now and continuously? It is a permanent installation with far fewer sensors.

Mapping produces the evidence that justifies where monitoring sensors go. Skip it and you are placing permanent sensors wherever the cable reached — which is the answer an auditor will extract from you in about ninety seconds when they ask "why is this sensor here?"

Why regulators ask for it

EU GDP requires an initial temperature mapping exercise on the storage area before use, carried out under representative conditions, and repeated according to a risk assessment and whenever significant modifications are made. WHO's guidance on storage and distribution takes the same position, and its technical supplement on temperature mapping of storage areas is the most useful free document on the subject. Revised Schedule M brings Indian manufacturers into the same expectation through its qualification and quality-risk-management requirements.

The common thread across all of them is justification. Nobody prescribes a universal sensor count, because a 1.5 m³ pharmacy fridge and a 4,000 m³ high-bay warehouse are not the same problem. What they prescribe is that your decisions come from documented evidence.

How many sensors

Small chambers — under about 2 m³ (refrigerators, freezers, small incubators, stability cabinets): a minimum of nine sensors is the widely applied baseline — eight distributed toward the corners of the usable volume plus one in the geometric centre. Add one more at any position you suspect, such as directly in front of the evaporator or immediately inside the door.

Walk-in cold rooms and chambers: start from the same logic in three dimensions — the eight corners of the usable storage volume plus the centre — and then add sensors at every position the room's physics makes suspicious:

• immediately in the evaporator air stream (the coldest point, usually)
• at the air return (often the warmest)
• just inside the door, at product height
• against any external wall, especially a sun-facing one
• near any heat-generating equipment, including lighting and forklift charging points
• at the top of the highest racking position that will hold product

Large warehouses: grid the floor plan and work in vertical layers, low, mid and high, because vertical stratification is the dominant effect. A high-bay pharmaceutical warehouse routinely shows 3–5 °C between floor level and the top rack. Spacing is a judgement call documented in the protocol; what matters is that no significant volume is unrepresented and that every identified risk position has a sensor on it.

For any of these, the honest summary is: the number comes out of your risk assessment, and the risk assessment is part of the deliverable. A protocol that says "nine sensors because the supplement says nine" for a 200 m³ cold room will not survive review.

Where the count usually goes wrong

The door. Loading bays and cold-room doors are the single biggest thermal load on most storage areas, and they are chronically under-instrumented because the doorway is inconvenient. This is precisely why the doorway needs a sensor: you are trying to characterise the worst case, not the comfortable case.

The top of the racking. Sensors get mounted at eye level because that is where a person can reach. Product sits up to the top beam. Map where the product is.

Void volume that is not really void. "We do not store anything in that corner" is a statement about today's stock plan, not about the qualified volume. If the racking exists, it will eventually hold product.

How long, and when

Duration: at least 72 hours of continuous recording, and preferably seven days. The argument for a week is simple — it captures a full operational cycle including a weekend, when HVAC schedules change, traffic stops, and a surprising number of facilities behave differently.

Sample interval: one to five minutes. Fast enough to resolve a defrost cycle and a door opening. Long intervals smooth away exactly the events you are looking for.

Season matters. A study run in January in Gujarat and a study run in May describe different buildings. Best practice is to map in the most challenging season, or to map in both and document the comparison. If you map in a mild season, say so in the report and justify it — an undocumented seasonal blind spot is a finding.

Empty first, then loaded. Empty mapping characterises the space itself and belongs to operational qualification. Loaded mapping characterises the space as you actually use it, with the thermal mass and the blocked airflow that real pallets create, and belongs to performance qualification. Both are expected; loaded is the one that determines your permanent sensor positions.

The challenge tests nobody enjoys

Two deliberate failures belong in the study, because they are the two failures that will happen anyway:

Door-open challenge. Open the door for a defined period under a defined load and record the recovery. This produces your realistic expectation for how long a normal operation pushes the space out of band — which is what makes it possible to set alarm delays that are meaningful instead of arbitrary.

Power-failure challenge. Cut power for a defined period and record the excursion and recovery. This gives you the answer to the question you will be asked after every real outage: how long do we have? Without this test, the answer is a guess.

Both need to be scheduled, risk-assessed and agreed with operations in advance. Both are also the tests most often quietly omitted, and an experienced auditor knows to look for them.

Instrument requirements

Calibration before and after. Pre-study calibration traceable to a national standard — in India, NABL-accredited under ISO/IEC 17025 — and a post-study verification. The post-study check is what lets you defend the data if a sensor drifted mid-study; without it, a single suspect channel can cast doubt on the whole report.

Accuracy appropriate to the band. For controlled room temperature, ±0.5 °C is generally adequate. For 2–8 °C work, where the band is only six degrees wide, ±0.3 °C or better is the sensible specification, and a Class-A PT100 at ±0.15 °C near 0 °C removes the argument entirely.

One timebase if you can get it. Nine single-channel loggers means nine clocks, and even small drift makes correlating a single event across positions awkward. Multi-channel instruments — ABH-300 carries four Class-A PT100 channels plus on-board temperature, humidity and barometric pressure — reduce the count of clocks, calibration certificates and batteries you have to manage, which is a real reduction in study overhead.

Enough memory for the whole study at your chosen interval. At a one-minute interval, seven days is about 10,000 readings per channel. The 16 MB parts in the ABH line hold up to 409,600 readings, so buffer wrap is not a constraint at study scale — but check the arithmetic against whatever instrument you use, because a wrapped buffer silently discards the beginning of your study.

What the report has to contain

A mapping report that survives review includes:

1. The protocol, pre-approved, with acceptance criteria defined before the data was collected
2. A drawing or plan with every sensor position numbered and located, in all three dimensions
3. The rationale for the sensor count and positions, traced to a risk assessment
4. Calibration certificates for every instrument, pre- and post-study
5. The raw data, retained, not just the summary charts
6. Identified hot and cold spots, explicitly named
7. Challenge test results — door open, power failure — with recovery times
8. Deviations encountered during execution and their assessment
9. A conclusion that specifies permanent monitoring positions, derived from the hot and cold spots
10. The re-mapping interval and the triggers that force an early re-map

Item 9 is the one that connects the study to the rest of your life. A mapping report that identifies extremes and then does not say where the permanent sensors go has left the most valuable part of the work undone.

Re-mapping triggers

Periodic re-mapping on a risk basis — commonly every two to three years for a stable facility — plus an immediate re-map after any of:

• changes to the refrigeration or HVAC plant, including a replaced compressor or a rebalanced system
• changes to racking layout, aisle configuration or storage height
• a change in the maximum stored load
• new construction adjacent to the space, or a new heat source inside it
• a change in the door configuration or traffic pattern
• any excursion investigation that concludes the thermal behaviour of the space is not what the previous study described

That last trigger is the one people resist and the one that matters most. If the space surprised you, your model of it is wrong.

Closing the loop

The output of mapping is a decision: these positions, monitored continuously, represent this space. The value of the whole exercise depends on that decision being carried through faithfully — which means the permanent monitoring system needs to sit at those exact positions, with equivalent or better accuracy, on a calibration programme, producing records under the same integrity controls as the study.

The same ABH-300 instrument is designed to serve both the mapping study and the routine monitoring that follows, deliberately, so the study data and the ongoing record share a data model, a reporting format and a calibration programme. If you have a mapping report and want to turn its conclusions into a monitoring design, send it over — that is a conversation we have often.

---

Related: pharma warehouse monitoring · cold room monitoring · stability chamber monitoring · validation support