Smart Orchard IoT for Precision Pest Monitoring and Intervention

Thesis Abstract

The study addresses the escalating pest pressures in commercial apple and citrus orchards and the limitations of conventional calendar-based spraying, which often results in excessive chemical use, environmental contamination, and suboptimal pest control. The aim is to develop and evaluate a Smart Orchard Internet of Things (IoT) system that enables precision pest monitoring and targeted interventions, thereby reducing chemical inputs while maintaining or enhancing yield and fruit quality. Specific objectives are to (1) design an integrated sensor network that continuously monitors microclimate, ambient pest activity indicators, and tree physiological stress; (2) develop a data fusion framework that detects pest risk levels in real time and triggers site-specific interventions; (3) implement machine learning models to forecast pest outbreaks and optimize spray schedules; (4) assess the agronomic, economic, and environmental performance of the system in multi-location field trials; and (5) evaluate user acceptance and operational feasibility among farm managers and extension officers. Methodologically, the study adopts a mixed-methods research design comprising a quantitative field experiment and qualitative user assessments. The population includes two commercial orchards in temperate regions, with a total cultivated area of 150 hectares and diverse cultivar composition. A stratified random sampling approach selects 40 hectares for IoT deployment and 10 hectares as control. Data collection instruments comprise (i) a sensor suite installed at 120 tree canopies, including microclimate sensors (temperature, humidity, leaf wetness), pest activity proxies (pheromone traps, acoustic sensors, canopy infrared thermography), and soil moisture probes; (ii) handheld and automated phenotyping devices for fruit size and quality metrics; (iii) farm management records for input costs, labor, and yield; and (iv) semi-structured interviews and surveys to capture stakeholder perceptions. Data will be collected over two full growing seasons to capture temporal pest dynamics. Validity and reliability are ensured through calibration protocols, pilot testing of sensors, and triangulation between sensor data and manual scouting. Instrument validity will be enhanced using content validity checks with entomology and horticultural experts, while reliability will be demonstrated through test-retest measures and inter-sensor calibration. Analytical techniques include time-series analysis and anomaly detection to identify pest risk signals, with regression models (multilevel linear and logistic) to quantify relationships between environmental variables, sensor-derived indices, and pest incidence. A random forest and gradient boosting framework will be employed for pest outbreak forecasting, while causal inference methods (difference-in-differences) will assess the effect of site-specific interventions on yield and fruit quality. Economic evaluation will use partial budgeting and net present value (NPV) calculations, complemented by sensitivity analyses on input costs and pest pressure scenarios. The theoretical basis leverages the Technology Acceptance Model (TAM) to interpret user adoption, and the Diffusion of Innovations theory to explain uptake patterns among growers. The study also draws on ecological pest management theory to justify reduced agrochemical reliance through targeted interventions. Expected findings include (i) a significant reduction in total pesticide applications (up to 40–60%) without compromising yield or fruit quality; (ii) improved pest suppression during peak risk periods through real-time intervention triggers; (iii) robust predictive models with area under the receiver operating characteristic (AUC) scores above 0.85; and (iv) positive net economic benefits under realistic cost scenarios. Environmental benefits are anticipated from lower chemical loads and reduced non-target impacts, assessed via life-cycle considerations. The study contributes to knowledge by integrating IoT-enabled sensing, data fusion, and machine learning to operationalize precision pest management in perennial fruit systems, offering a scalable blueprint for adoption in diverse orchard contexts. It provides methodological advancement in calibrating sensor networks for pest risk detection, as well as practical guidelines for deployment, data governance, and decision-support interfaces. The main conclusion is that Smart Orchard IoT systems can deliver sustainable pest control with measurable agronomic and economic gains when combined with farmer-centric implementation strategies. Recommendations include expanding to multi-species pest complexes, integrating drone-based scouting for canopy-level validation, and developing open-access datasets to foster cross-regional model transferability.