AI-powered Pheno-Farm Server: Making adaptive farming decisions

Context Agricultural productivity faces growing challenges due to climate variability, resource constraints, and the demand for sustainable practices. Precision agriculture, powered by artificial intelligence (AI) integrate high-throughput phenotyping, offers practical advancements for monitoring crop traits, growth patterns, and responses to environmental factors. The Pheno-Farm Server (PFS) leverages AI-driven approaches to integrate real-time data, analyse it, and enable adaptive decision making for efficient crop management in diverse farming environments such as glasshouses, rain-out shelters (ROS), experimental farms, and open fields. This empowers stakeholders to make informed, data-driven decisions. Objective This study evaluates the potential of the PFS in advancing precision agriculture through adaptive AI models. It focuses on the system's ability to collect, process, and analyse data by developing a robust pipeline for integrating phenotyping datasets and optimizing nitrogen application. Additionally, it assesses the effectiveness of AI/ML-based adaptive decision-making in improving crop management and resource utilization. Methods The PFS was implemented using a structured framework integrating hardware, software, and data processing components. Data from phenotyping platforms, including plant area (PA) measurements and digital biomass (DBM) estimates from four cereal cultivars under varying nitrogen levels (NL) and drought conditions (DC), were aggregated into the PFS. Three regression models such as Ridge Regression, Support Vector Regression (SVR), and Random Forest were trained and evaluated using a structured pre-processing pipeline. Hyperparameters were optimized through grid search with 5-fold cross-validation. Model performance was assessed using the coefficient of determination (R 2 ) and normalized root mean squared error (NRMSE). Tests were conducted to validate the system's reliability and applicability. Results and conclusions The experimental implementation of the PFS demonstrated its capability to collect and analyse data from various sources effectively. The SVR model had the highest accuracy with an R 2 of 0.992 and an NRMSE significantly lower than other models followed by Random Forest and Ridge Regression. SVR was good at understanding complex relationships and worked well with unseen nitrogen levels. Random Forest showed limitations in generalization due to data dependency. The integration of AI-powered servers like PFS improve precision agriculture by making real-time data analysis possible, allowing smart decisions, and using resources more efficiently. Significance The integration of AI-powered systems like PFS represents a significant advancement in sustainable farming. These systems help drive innovation and tackle important agricultural problems, leading to farming that is productive, profitable and environmentally friendly. The study shows how AI can change traditional farming and set new standards for data-driven farming practices.