AI and IoT climate control: integrating climate and crop data
Climate computers and sensor-based fertigation are established commercial practice. What is still developing is the AI layer on top — models that combine environmental and crop data to stabilise climate, support irrigation and fertigation decisions, flag anomalies and make output more consistent. Most published results come from research systems or single trials.
- Established commercial practiceComputerised climate control and sensor-based fertigation (dosing from measured EC, pH and drain) are standard in commercial greenhouse production — see the climate-control and fertigation briefs below.
- Peer-reviewed reviewA May 2026 review describes AI use in four areas: environmental monitoring and control, crop-health monitoring with computer vision, AI-supported irrigation and fertigation, and robotics. It is a narrative review of published work, not a benchmark of commercial systems.Source: Kofi et al. (2026), “Use of Artificial Intelligence in Controlled Environment Agriculture: A Review”, J. Amer. Soc. Hort. Sci. 151(4):306–314 (published 1 May 2026)
- Single controlled study — not a performance guaranteeOne study cited in the review (Bonelli et al., 2024) found that sensor-based irrigation in hydroponic strawberry used 26% less water and 38% less fertilizer than timer-based irrigation. That result is for one crop and system and should not be read as an expected saving.Source: Kofi et al. (2026), “Use of Artificial Intelligence in Controlled Environment Agriculture: A Review”, J. Amer. Soc. Hort. Sci. 151(4):306–314 (published 1 May 2026)
- Research / experimental systemResearch examples in the review include machine-learning anomaly detection for indoor plant-wall climate systems (Liu et al., 2020) and an IoT, fuzzy-logic hydroponic controller that used 18% less energy than its own baseline (Khudoyberdiev et al., 2020). Both are experimental systems.Source: Kofi et al. (2026), “Use of Artificial Intelligence in Controlled Environment Agriculture: A Review”, J. Amer. Soc. Hort. Sci. 151(4):306–314 (published 1 May 2026)
- Peer-reviewed reviewLimitations the review lists: models need large, consistent datasets; predictions can fail when conditions move outside training data — a new pest or disease still needs human judgement; sensors and cloud infrastructure are costly; and growers need clear data-ownership and consent terms.Source: Kofi et al. (2026), “Use of Artificial Intelligence in Controlled Environment Agriculture: A Review”, J. Amer. Soc. Hort. Sci. 151(4):306–314 (published 1 May 2026)
Questions to put to suppliers
- Which decisions does the system make automatically, and which does it only recommend?
- What measured results exist for a comparable crop, climate and greenhouse type — and were they a trial or a commercial site?
- What happens on sensor failure, connectivity loss or power loss?
- Who owns the climate and crop data, and can it be exported if the supplier changes?
