This chapter is the most thesis-author-driven of the seven. The section headings below are scaffolding; the conclusions land after §4 results are finalised. Every bullet here is a placeholder.
§7.1 · What the numbers say
[High-level interpretation of §4 steady-state + failure-condition results. Hold one to two paragraphs per measurement variable. Refer back to the §3.2 hypotheses and state which were retained / rejected.]
DRAFT — populates after §4 windows complete.
§7.2 · What the numbers don’t say
The study is one greenhouse, one operator, one year, one EU-Central region, one set of sensors, one cloud provider. The numbers are informative; they are not generalisable without further work. Specific non-claims:
- The results do not establish that edge processing is universally cheaper / faster / more reliable. They establish the comparison for this specific configuration.
- The energy variable depends on a published AWS efficiency coefficient for eu-central-1; the coefficient has uncertainty bounds.
- The cost variable is sensitive to the AWS Free Tier; over-tier windows are excluded.
§7.3 · Threats to validity
- Single-site bias. No independent replication.
- Operator effect. The author wrote both the edge code and the cloud code; operator skill is a confounder.
- Sensor calibration drift. Capacitive moisture probes drift; not recalibrated within the measurement windows.
- Network locality. Hub is on residential fibre; results may differ on cellular or satellite uplink.
§7.4 · Future work
Open questions the next iteration should attempt:
- Cross-site replication. A second greenhouse + a second operator, under the same protocol.
- Different cloud regions. eu-central vs eu-west vs us-east — does the latency / cost story shift?
- Different cloud providers. AWS vs GCP vs Hetzner — same pipeline primitives, different bills.
- Longer measurement window. A full growing season, including winter shutdown periods.
§7.5 · Practical recommendation (single-operator, single-greenhouse)
[One paragraph. What would the author actually deploy if asked today? The thesis isn’t a sales pitch; this section is about engineering honesty.]
DRAFT — written last, after §4 numbers and §7.1 interpretation are locked.
Status: structural draft v0.1, 2026-05-07. Citation:
https://plantir.garden/thesis/2026/discussionis locked per ADR-011 — note that conclusions inside this chapter will change before defence; the URL stays put.