Spatial microclimate · Cooling · 14 October 2026
The AC Cooled the Container Farm—but Not Uniformly
Data from 27 sensors exposed a trade-off that a single average can hide: when the AC was on, mean temperature fell by 0.58°C and VPD improved, but the sensor-to-sensor temperature range widened from 2.78°C to 5.03°C.
Mean temperature
−0.58°C
Lower during AC-ON observations.
Spatial range
+2.25°C
Wider during AC-ON observations.
Synchronized frames
544
406 AC-OFF and 138 AC-ON.
A good average can conceal a poor crop environment
Climate control is often evaluated with one average temperature. That is convenient, but it can hide simultaneous overcooling and overheating at different racks or positions.
In this facility, spatial standard deviation rose from 0.71°C to 1.25°C and the sensor range increased from 2.78°C to 5.03°C when the AC was on.
How the comparison was built
The observation covered 18–30 June 2026 in the 10-ft indoor farm. Temperature and humidity were recorded across 27 indoor nodes and aligned with AC status without fabricating intermediate values.
The resulting dataset contained 544 synchronized timestamps: 406 with the AC off and 138 with the AC on.
Lower temperature did not mean higher spatial compliance
The mean fell from 23.43°C to 22.85°C, but per-node temperature compliance decreased from 51.6% to 46.6%. Frames in which all nodes were simultaneously within 20–24°C fell from 20.9% to 1.4%.
Mann–Whitney comparisons indicated statistically significant differences in mean temperature, spatial standard deviation, temperature range, and compliance.
A benefit remained: humidity and VPD improved
Average relative humidity decreased from 83.97% to 65.98%, and air VPD rose from 0.463 to 0.946 kPa. Per-node VPD compliance increased from 21.1% to 57.5%.
The correct conclusion is therefore not that cooling was harmful. Cooling and dehumidification helped VPD, but the benefit came with a more uneven thermal field.
Engineering response
Control should not depend on a single sensor. Air distribution must be treated as a design variable, and dashboards should expose range, standard deviation, per-node compliance, and simultaneous compliance alongside the mean.
Conclusion
An average of 22.85°C looked acceptable, yet a 5.03°C spatial range and only 1.4% simultaneous compliance showed that plants did not experience the same environment. The next step is better circulation and spatially aware control, not simply more cooling power.
Interpretation boundary
These results describe one instrumented facility and the stated observation window. Operational associations should not be presented as universal causal effects. Airflow hypotheses require direct velocity measurements, smoke visualization, or validated CFD experiments.