Edge Early Warning of Threshold-Defined Heat-Stress Episodes in Tomato: A Proof-of-Concept Low-Cost ESP32 Multi-Sensor Node with On-Device Machine Learning

Heat stress during flowering reduces tomato fruit set, yet low-cost monitoring nodes typically report a threshold breach only after it occurs. This proof-of-concept study developed CropWhisperAI, an ESP32 node with air-temperature/humidity, soil-temperature, soil-moisture, light and reservoir-level sensors and tested whether threshold-defined heat-stress episodes (air > 35 °C or soil > 32 °C for at least 15 min) can be forecast 15–60 min ahead on the microcontroller. Over 107 days, the node logged 30,816 five-minute records beside a single field-grown tomato plant at one site, containing 42 episodes on 42 days, 29 of them in June; the models therefore forecast threshold crossings, not physiological damage. Two rules, a time-of-day baseline and four learned models were compared using day-blocked cross-validation and event-level metrics. At 30 min, LightGBM reached an area under the precision–recall curve of 0.919 (95% CI 0.862–0.961), warned before 40 episodes and raised 0.07 false alarms per day, against 0.28 for a trend-extrapolation rule. A nine-parameter logistic model for the ESP32 warned before 41 episodes with a median lead of 30 min and 0.10 false alarms per day, and its C implementation reproduced every Python alarm decision. Air and soil temperature carried nearly all predictive information. Under seasonal and leave-one-month-out hold-outs, accuracy fell and no method was consistently best. The results establish engineering feasibility only; validation on multiple plants, sites and seasons, linked to plant responses, is required. By warning growers before heat stress sets in, such low-cost nodes could help farms adapt to rising temperatures, contributing to SDG 2 (Zero Hunger), SDG 13 (Climate Action) and SDG 9 (Industry, Innovation and Infrastructure).

Authors

Institutions

Publication Details

Journal
AgriEngineering
Published
2026-11-17
DOI
https://doi.org/10.3390/agriengineering8110493
Primary Topic
Smart Agriculture and AI
Type
article
Field-Weighted Citation Impact
0.00
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
article

Edge Early Warning of Threshold-Defined Heat-Stress Episodes in Tomato: A Proof-of-Concept Low-Cost ESP32 Multi-Sensor Node with On-Device Machine Learning

S G Rahul, Rithikkaa Suresh Jyothi Bai, Amirthalakshmi Thirumalai Maadapoosi, Parinitha Mohanavelu
1 links
AgriEngineering
Smart Agriculture and AI
article

Edge Early Warning of Threshold-Defined Heat-Stress Episodes in Tomato: A Proof-of-Concept Low-Cost ESP32 Multi-Sensor Node with On-Device Machine Learning

S G Rahul, Rithikkaa Suresh Jyothi Bai, Amirthalakshmi Thirumalai Maadapoosi, Parinitha Mohanavelu
article en

Abstract

Heat stress during flowering reduces tomato fruit set, yet low-cost monitoring nodes typically report a threshold breach only after it occurs. This proof-of-concept study developed CropWhisperAI, an ESP32 node with air-temperature/humidity, soil-temperature, soil-moisture, light and reservoir-level sensors and tested whether threshold-defined heat-stress episodes (air > 35 °C or soil > 32 °C for at least 15 min) can be forecast 15–60 min ahead on the microcontroller. Over 107 days, the node logged 30,816 five-minute records beside a single field-grown tomato plant at one site, containing 42 episodes on 42 days, 29 of them in June; the models therefore forecast threshold crossings, not physiological damage. Two rules, a time-of-day baseline and four learned models were compared using day-blocked cross-validation and event-level metrics. At 30 min, LightGBM reached an area under the precision–recall curve of 0.919 (95% CI 0.862–0.961), warned before 40 episodes and raised 0.07 false alarms per day, against 0.28 for a trend-extrapolation rule. A nine-parameter logistic model for the ESP32 warned before 41 episodes with a median lead of 30 min and 0.10 false alarms per day, and its C implementation reproduced every Python alarm decision. Air and soil temperature carried nearly all predictive information. Under seasonal and leave-one-month-out hold-outs, accuracy fell and no method was consistently best. The results establish engineering feasibility only; validation on multiple plants, sites and seasons, linked to plant responses, is required. By warning growers before heat stress sets in, such low-cost nodes could help farms adapt to rising temperatures, contributing to SDG 2 (Zero Hunger), SDG 13 (Climate Action) and SDG 9 (Industry, Innovation and Infrastructure).

AgriEngineeringVol. 8(11)
Amrita Vishwa Vidyapeetham (IN)
Zero hunger
Smart Agriculture and AI
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon