Integrated descriptive, predictive, and prescriptive models for hydroponic crop optimization

Overview

This research program develops an integrated modeling framework for hydroponic crop production that unites crop physiology, artificial intelligence, and computational optimization. Unlike traditional agricultural crop models designed for soil-based systems, these models are purpose-built for the unique dynamics of hydroponic environments. The framework operates in three layers: descriptive models that capture physiological processes in biologically meaningful terms, predictive models that forecast crop performance under unseen environmental conditions, and prescriptive models that recommend optimal growing conditions to maximize yield, nutritional value, and energy efficiency.

The work addresses a critical gap in controlled-environment agriculture, where growers often rely on either black-box machine learning tools that lack transparency or legacy soil-based models that fail to capture hydroponic-specific dynamics. By producing models that are both accurate and explainable, the program offers a practical decision-support foundation for commercial hydroponics operators, vertical farming companies, and space agriculture applications.

Technical specifications

Modeling approach:

• Descriptive models represent major physiological processes of crop growth with parameters that are biologically meaningful and interpretable
• Data-driven calibration using historical and experimental data produces predictive models capable of temporal and spatial extrapolation
• Prescriptive models apply computational optimization to identify growing conditions that maximize yield, nutritional value, and energy efficiency
• Iterative experimental feedback loops continuously refine model accuracy

Key features:

• Validated on corn and soybean using USDA historical data, achieving predictive performance comparable to or exceeding state-of-the-art black-box machine learning models
• Designed to handle multiple environmental variables including light, temperature, CO2, pH, growing media, and humidity
• Explainable outputs that allow growers and researchers to understand the physiological basis of recommendations
• Architecture extensible to additional crops and, longer term, to microgravity conditions for space agriculture

Technology readiness level

The descriptive and predictive modeling components have been validated on corn and soybean using USDA historical datasets, demonstrating strong fit and accurate extrapolation performance. The next phase focuses on hydroponic lettuce, beginning with descriptive model design, experimental data generation under varied growing conditions, model calibration, and prescriptive model development. Prescriptive recommendations will be tested experimentally, with results fed back into the dataset to iteratively improve accuracy. The long-term vision extends the framework to space agriculture by incorporating gravity as a model variable.