Smartphone-based IoT Sensor Network for Real-Time Soil Health Monitoring

Thesis Abstract

The rapid adoption of smartphone-enabled sensing and the rising demand for timely soil health information motivate the development of a real-time, field-deployable monitoring system that integrates smartphone-based sensors with low-cost IoT devices to assess soil health indicators and guide precision management. The study addresses the problem of spatial-temporal disconnect between soil health data and on-farm decision-making due to limited access to timely, location-specific soil metrics. The aim is to develop and validate a smartphone-based IoT sensor network capable of real-time monitoring of key soil health parameters (moisture, temperature, pH, electrical conductivity, and micronutrient proxy indicators) across diverse agroecologies, and to evaluate its predictive utility for crop-precursor conditions. Specific objectives include (1) to design a modular sensing architecture combining smartphone sensors, wireless IoT nodes, and cloud analytics; (2) to calibrate and validate sensor readings against standard laboratory analyses for soil moisture, pH, and electrical conductivity (n=150 soil samples across five farms); (3) to develop machine learning models to predict laboratory soil fertility metrics (macronutrients and organic matter) from in-field sensor data (n=3000 sensor-day records); (4) to assess user acceptability and operational reliability among farmers (n=60 participants) through a mixed-methods approach; and (5) to propose a deployment framework aligned with farmers’ decision-support needs. The methodology adopts a convergent parallel mixed-methods design underpinned by the Technology Acceptance Model (TAM) and the Diffusion of Innovations theory to explore user uptake and adoption barriers. The population comprises arable farms across three distinct soil types within a temperate region. A stratified random sampling strategy yields 60 farms for field testing and 15 farmer focus groups for qualitative insights. Data collection instruments include calibrated IoT sensor nodes for soil moisture, temperature, pH, and electrical conductivity; smartphone-based data capture with a dedicated mobile application; laboratory analyses at accredited soil testing facilities following standard methods (e.g., oven-dried moisture, pH meter, EC meter, and Kjeldahl/NPK for nutrient proxies); and semi-structured interviews and focus groups. Validity and reliability are established through cross-validation of sensor data with laboratory results (n=150 samples), test-retest reliability for the mobile app (intraclass correlation coefficient >0.85), and calibration curves for each sensor parameter. Data analysis employs descriptive statistics and time-series analyses to characterize sensor performance, followed by regression and random forest algorithms to predict lab-based soil properties from in-field measurements; model evaluation uses RMSE, R-squared, and cross-validation errors. A theoretical model linking sensing fidelity, data latency, and decision-support accuracy guides analysis. The expected findings include (i) high correlation between smartphone-IoT sensor readings and standard laboratory measurements for moisture (R2 > 0.90), pH (R2 > 0.85), and EC (R2 > 0.88); (ii) robust predictive models predicting soil organic matter and available phosphorus with RMSE within acceptable agronomic error ranges; (iii) demonstration of timely alerts and field-scale fertility mapping enabling targeted irrigation and nutrient management; and (iv) evidence on user acceptance and adoption potential, indicating favorable attitudes toward use among the majority of farmers, with identified barriers such as data privacy concerns and device maintenance needs. The study contributes to knowledge by integrating mobile computing, IoT sensing, and machine learning to deliver real-time soil health intelligence at farm scale, extending existing precision agriculture frameworks with a smartphone-centric, cost-effective solution suitable for resource-limited contexts. The main conclusion is that smartphone-based IoT sensor networks can provide accurate, timely, and actionable soil health information, substantially improving decision-making for irrigation and nutrient management. Recommendations include scaling the architecture to larger geographic and soil diversity, creating open datasets for model generalization, implementing robust battery-saving and data-security features, and developing policy-oriented guidelines to support adoption, training, and interoperability with existing farm management information systems.