Smartphone-based IoT Soil Health Monitoring and Decision Support System

Thesis Abstract

Soil health is a critical determinant of agricultural productivity and sustainability, yet farmers globally face challenges in acquiring timely, accurate soil information to guide management decisions. The study addresses the gap between conventional soil testing, which is expensive and infrequent, and the demand for continuous, field-scale monitoring that informs adaptive practices. The aim is to develop, validate, and evaluate a smartphone-based Internet of Things (IoT) soil health monitoring and decision support system (DSS) that integrates affordable sensor data, citizen-science–style field observations, and cloud-based analytics to deliver real-time soil health indices and management recommendations. Specific objectives include (i) designing an integrated mobile sensing platform comprising low-cost in situ sensors for soil moisture, temperature, pH, and electrical conductivity, coupled with a portable spectrophotometric module for organic matter proxy measurement; (ii) implementing edge- and cloud-based data processing pipelines, including data quality checks, imputation, and feature extraction; (iii) developing a soil health index (SHI) framework and crop-specific nutrient and moisture management rules embedded within a rule-based and machine learning–driven DSS; (iv) evaluating system usability, adoption potential, and impact on management decisions among 120 smallholder farmers across three agrarian districts; and (v) validating the SHI against laboratory soil analyses and crop yield outcomes over two growing seasons. Methodologically, the study adopts a mixed-methods research design anchored in a pragmatic paradigm to maximize practical applicability. A purposive sample of 60 farmers will participate in a pilot, with 60 additional farmers allocated to a comparative conventional-management group, yielding a total population of approximately 120 farmers. Data collection instruments include calibrated IoT sensors deployed in farmers’ fields (n=180 sensor deployments across sites), a smartphone app with an integrated user interface, structured questionnaires, semi-structured interviews, and field-level crop yield records. Laboratory validation will involve standard soil property analyses (pH, EC, texture, organic carbon, total nitrogen) on 360 composite soil samples collected seasonally. Reliability and validity will be ensured through instrument calibration, test-retest procedures, and triangulation across sensor data, laboratory results, and farmer-reported outcomes. Data analysis will employ descriptive statistics and inferential methods, including multiple linear regression and ANOVA to assess the relationship between SHI components and yield, time-series analysis for sensor data, and machine learning models (random forest, gradient boosting) to predict soil health outcomes and optimize DSS recommendations. The theoretical framework will be guided by the Technology Acceptance Model (TAM) to evaluate user adoption and the Resource-Based View (RBV) to interpret organizational capability gains from deploying the system. A conceptual model illustrating the data-flow from sensing to decision support and farmer action will be developed and tested. Expected findings include (i) the smartphone-based IoT platform achieving high data capture continuity (>90%) with robust imputation for missing values and low sensor drift over two growing seasons; (ii) the SHI demonstrating strong correlations with laboratory soil properties (r > 0.75 for organic matter proxies and nutrient indicators) and with observed crop yields (r > 0.6); (iii) the DSS producing actionable, agronomically sound recommendations with a decision accuracy rate exceeding 80% for fertilizer and irrigation scheduling under varied climatic conditions; (iv) positive shifts in farmer decision-making behavior and perceived usefulness, supported by TAM-based indicators showing gradual increases in perceived ease of use and perceived usefulness over time; and (v) demonstrable yield and input-use efficiency improvements in the treatment group relative to the control group. The study contributes to knowledge by advancing an integrative, scalable model for real-time soil health monitoring that bridges on-farm sensing with decision support, thereby enabling precision management in resource-constrained farming systems. It also advances methodological discourse on combining IoT-enabled data streams with machine learning-based DSS in agronomy, and offers a practical framework for validating soil health indices against empirical crop outcomes. The conclusions will emphasize the viability of smartphone-based IoT solutions for democratizing soil health information, with policy and extension implications to promote widespread adoption, capacity building, and continued innovation in low-cost, participatory soil monitoring technologies. Recommendations include refining sensor calibration protocols, expanding the SHI framework to accommodate additional soil constraints (e.g., salinity, micronutrient status), enhancing offline functionality for remote areas, and conducting longitudinal studies to assess long-term agronomic and environmental impacts.