Research brief: climate aggregation and hidden VPD intervals
How does classifying hourly and daily mean VPD change the time flagged by the same rule on five-minute data?
Scope of this review
One public Reference greenhouse stream, 41 complete source-clock days, 11,808 intervals and 984 nominal hours. This extends our sensor study with a different question and complete-day cohort.
How we assessed the evidence
Checksum-pinned input; timestamp jitter normalized; missing pairs never interpolated. Calculate VPD per pair, then non-overlapping hourly and daily means. Reassign each mean to its source intervals and compare strict thresholds 0.3, 0.4, 0.5 and 0.6 kPa. Hold each value for five minutes; never extend beyond the final timestamp.
What the evidence supports
Mean preserved, classification changed
At 0.4 kPa, the original has 282.75 h below. Hourly means hide 24.08 h and introduce 18.33 h; daily means hide 255.75 h and introduce 21 h.
Threshold-sensitive effect
Daily hidden fractions range from 100.0% at 0.3 kPa to 14.2% at 0.6 kPa. The denominator is original below-threshold time at each threshold.
What this review does not establish
- One stream in one greenhouse; no causal crop, disease, savings or calibration inference.
- Two incomplete days excluded. Source timezone is unspecified; calendar bins use the source clock.
- Five-minute held values are a baseline, not continuous monitoring. The selected example is the day with the greatest hidden duration.
Hamfy’s proposed next step
Keep the raw export and show coverage, excursions and their timing alongside averages. Download the reproducible analysis or discuss an audit of your growing data.
Evidence to collect for a real project
- Repeat on other seasons, sensor positions and complete/incomplete cohorts.
- Compare with recorded alert delays, hysteresis and actual operator actions.
Source notes
- Public dataset · 2022–2023AGC 2022: first lettuce experiment, data dictionary and provenanceCEA Open Data / Koidra Tech Inc.; WUR AGC
Pinned source version; Reference compartment, Tair and Rhair. MIT permission and the unchanged input are included in our package.
- Primary method · 1998FAO-56, Chapter 3: Meteorological dataFAO
Saturation-vapour-pressure equation used for air VPD. It does not supply the 0.4 kPa review threshold used in this study.