Conference Agenda

The mismanagement of irrigation, fertilization, and cropping systems intensifies conflicts within the Water-Food-Ecology (WFE) system. Here, we developed a novel artificial intelligence (AI) Agent by integrating agro-hydrological model (AHC), machine learning algorithm, multi-criteria decision-making method, and large language model for WFE trade-offs, within in an AI development platform. The AI Agent can automatically optimize irrigation, fertilization, and crop planting structures through balancing irrigation water productivity (IWP), ecosystem service value (ESV), and crop yield. Applied to the Wuliangsuhai lake watershed of North China under 90 combined scenarios of irrigation and nitrogen application, and 8 planting structure scenarios. The AI Agent demonstrated significant improvements in sustainability of the WFE system by strategically reducing irrigation and nitrogen inputs, while optimizing wheat-to-maize ratio. Compared to conventional practices, the AI-driven solution increased IWP by 65%, reduced negative ESV by 20%, while limiting yield loss below 10%. Notably, the AI Agent shortened the time required for each decision cycle compared with traditional approaches, achieving a 5.7- to 11.5-fold increase in decision-making efficiency. This improvement was achieved by replacing extensive manual steps, including model execution, decision indicator calculation, and data transmission. This study highlighting the potential of AI Agent to enable intelligent management in complex agricultural systems.

Evaluating irrigation management performance requires reliable methods to benchmark applied water against crop water requirements under real production conditions. This study assessed irrigation strategies implemented by farmer teams participating in the Testing Ag Performance Solutions (TAPS) farm management competition hosted by the University of Nebraska-Lincoln using a process-based crop growth model. The DSSAT CERES-Maize crop model was calibrated and validated using multi-year field measurements, including site-specific soil properties, weather records, and observed maize phenology and yield from the TAPS research site. The validated model was then used to estimate irrigation requirements under the same environmental conditions experienced during the competition. Seasonal irrigation applied by participating teams was compared with model-estimated irrigation requirements to quantify deviations in water management. Considerable variation in seasonal irrigation amounts was observed among teams operating under identical field and weather conditions. Although many teams applied more irrigation than the model-estimated requirement, the overall differences were smaller than values reported in previous regional farm surveys. The results demonstrate the potential of combining crop simulation models with farmer-engagement programs to create a model-based benchmarking framework that supports improved irrigation decision-making and more efficient use of water resources in irrigated maize systems.

Effective use of soil moisture data for irrigation scheduling and field-scale water allocation is limited by spatial soil heterogeneity, measurement accuracy that depends on soil type, dynamic conditions at the soil-sensor interface, and sensor calibration constraints. Among sensor-based techniques, multi-depth FDR probes are widely used; however, their practical application is often hindered by soil-specific calibration requirements, temporal drift in sensor responses, and inter-probe variability. This study proposes an innovative Real-Time in situ Dynamic Post-Calibration (RTisDPC) procedure that enables the quantification of irrigation water requirements and actual evapotranspiration (ETa) using soil moisture data measured by uncalibrated sensors. The approach was tested in a drip-irrigated tomato field. RTisDPC uses known irrigation volumes as reference inputs to dynamically adjust root water uptake estimates, accounting for temporal changes in sensor performance. Daily ETa estimates showed strong agreement with independent sap flow measurements and reproduced seasonal trends consistently with FAO-56 calculations. The dynamic approach achieved water savings of 30% compared with standard on-farm practices and 22% relative to FAO-56-based scheduling, without reducing marketable yield. While fresh fruit water use efficiency remained stable, dry matter water use efficiency increased, indicating improved biomass conversion. These results highlight a cost-effective, scalable solution for climate-resilient sensor-based precision irrigation management.

Precision irrigation requires high-resolution, continuous soil moisture measurements to maximize crop growth and water use efficiency. Wireless soil moisture sensors are widely used nowadays; however, their installation often ignores in-field spatial-temporal variability. This study develops a method for designing wireless sensor network (WSN) distribution layouts for maximum field coverage with a minimum number of sensors while ensuring reliable communication. The proposed approach analyzes Sentinel-2 retrieved normalized difference moisture index as a proxy of soil water, followed by variogram modeling to estimate the effective range, which defines the maximum sensor-to-sensor distance. Sensor locations are then determined using a variogram-constrained clustering algorithm. In a case study over a 25 ha field in the Netherlands, this method distributes eight sensors, achieving approximately 90% spatial coverage. The effective range varies from 110 m to 423 m at mature and early potato growth stages, respectively, indicating the necessity for dynamic adjustment of sensor distribution layouts throughout the growing season. EU868 propagation simulations show that the WSN achieves packet delivery ratios above 90% under fully grown potato conditions. These results demonstrate that the proposed method is efficient and effective for designing WSN layouts that account for spatial heterogeneity while ensuring reliable data communication.

This study validates IrrigMonitor, an extension‑driven, vendor‑agnostic decision support platform designed to close the persistent technology–practice gap limiting precision‑irrigation adoption. The system standardizes heterogeneous commercial sensor data and integrates physics‑informed neural networks with Moving Horizon Estimation to produce 72‑hour irrigation forecasts. Co‑development with extension personnel shaped platform workflows, interface design, and communication output, resulting in text‑based, actionable recommendations preferred by producers over graphical dashboards. Preliminary on‑farm deployments across citrus and vegetable systems demonstrated 20–30% reductions in water use and >0.90 correlation between forecasted and measured soil‑water dynamics. The platform’s human‑in‑the‑loop learning mechanism captures real irrigation decisions, enabling localized model refinement that aligns AI-driven predictions with producer expertise. Targeted extension training programs further increased practitioner confidence, improved cross‑vendor data literacy, and strengthened statewide capacity for precision‑irrigation implementation. These results indicate that embedding adaptive machine learning within an extension‑centered framework significantly enhances usability, trust, and water‑use efficiency relative to threshold‑based commercial systems. The approach demonstrates a scalable pathway for integrating advanced analytics into routine irrigation management while maintaining the practical, interpretable outputs essential for widespread adoption.

Efficient water management remains a critical challenge in agriculture, particularly in water-limited and off-grid regions with constrained energy access. This study aims to develop and evaluate a solar-powered precision irrigation system integrating real-time sensing and climate-based decision support. The system employs a photovoltaic-powered deep-well pump (91 m depth) and an ESP32-based controller connected to soil moisture and temperature sensors. LoRa communication enables long-range data transmission up to 10 km for remote monitoring. Crop water requirements (CWR) are estimated using the FAO Penman–Monteith method based on observed climatic data. For tomato crops, the estimated seasonal CWR during the study period was approximately 365.4–405.4 mm season⁻¹ (equivalent to 4.06–4.50 mm day⁻¹), and irrigation scheduling was dynamically adjusted using both sensor feedback and evapotranspiration estimates. Field implementation was conducted in Banyu Urip Village, West Lombok Regency, during the March planting season. Results indicate that the system improved water-use efficiency by 18–37% and reduced irrigation water consumption by 22–40% compared to conventional practices. The photovoltaic systemmaintained energy autonomy of 85–100% under practical conditions. The system demonstrates a reliable and scalable solution for energy-efficient precision irrigation in water-limited environments.