Articles
Physics-informed AI control of irrigation: stable drainage EC management strategy for greenhouse tomato production
Article number
1468_51
Pages
393 – 400
Language
English
Abstract
High-wire tomato production in greenhouses accounts for 60% of fresh tomatoes in the US. Among various management practices, optimizing irrigation schedules is the most important for improving fruit quality and yield.
In commercial operation, irrigation scheduling is based on setpoint changes by experienced growers.
However, maintaining rootzone electrical conductivity (EC) within a narrow range, a key strategy to improve fruit quality, often becomes a challenge for growers.
To address this, we present and validate an artificial intelligence (AI) approach that couples a physics-informed evapotranspiration model driven by net radiation and canopy vapor-pressure deficit.
The AI-based irrigation control model with real-time feedback maintains drainage EC within the target range and meets crop water demand.
To validate the control model, cluster cherry tomato ‘Maxxiany XR’ (double-headed grafts; 4.4 heads m-2; transplanted 15 April 2025; harvesting from 9 June) was grown in a 484 m2 greenhouse compartment, while another identical setup was assigned for expert-grower-driven control.
A lysimeter system was used to measure the continuous drainage ratio and drainage EC. The AI autonomously scheduled irrigation intervals based on ET prediction.
The expert-driven control frequently monitored lysimeter measurements at least twice a day and changed setpoints that influenced irrigation intervals.
Over 80 days (10 May-28 July 2025), AI-driven control resulted in EC similar to the target EC and lower standard deviation (6±0.98 dS m-1) compared to the expert-control (7.95±1.83 dS m-1). The AI-driven control resulted in lower standard deviation in daily drainage ratio throughout the trial than the expert-driven control (35.5±12.1% vs. 37.5±17.5%). Eight-week cumulative yields with AI- and expert-driven control were comparable (9.43 vs. 9.47 kg m-2). These findings suggest that AI-based irrigation control maintained an ideal rootzone environment and stabilized drainage with minimal operator input, indicating a practical pathway for irrigation management under protected cultivation.
In commercial operation, irrigation scheduling is based on setpoint changes by experienced growers.
However, maintaining rootzone electrical conductivity (EC) within a narrow range, a key strategy to improve fruit quality, often becomes a challenge for growers.
To address this, we present and validate an artificial intelligence (AI) approach that couples a physics-informed evapotranspiration model driven by net radiation and canopy vapor-pressure deficit.
The AI-based irrigation control model with real-time feedback maintains drainage EC within the target range and meets crop water demand.
To validate the control model, cluster cherry tomato ‘Maxxiany XR’ (double-headed grafts; 4.4 heads m-2; transplanted 15 April 2025; harvesting from 9 June) was grown in a 484 m2 greenhouse compartment, while another identical setup was assigned for expert-grower-driven control.
A lysimeter system was used to measure the continuous drainage ratio and drainage EC. The AI autonomously scheduled irrigation intervals based on ET prediction.
The expert-driven control frequently monitored lysimeter measurements at least twice a day and changed setpoints that influenced irrigation intervals.
Over 80 days (10 May-28 July 2025), AI-driven control resulted in EC similar to the target EC and lower standard deviation (6±0.98 dS m-1) compared to the expert-control (7.95±1.83 dS m-1). The AI-driven control resulted in lower standard deviation in daily drainage ratio throughout the trial than the expert-driven control (35.5±12.1% vs. 37.5±17.5%). Eight-week cumulative yields with AI- and expert-driven control were comparable (9.43 vs. 9.47 kg m-2). These findings suggest that AI-based irrigation control maintained an ideal rootzone environment and stabilized drainage with minimal operator input, indicating a practical pathway for irrigation management under protected cultivation.
Authors
I. Dao, J. Hollick, C. Kim, C. Kubota, K. Tran
Keywords
human vs. AI, rootzone EC, greenhouse tomato production, controlled environment agriculture
Groups involved
Online Articles (70)