Back to gallery
Presentation
PhD Dissertation Defense: Low-Cost Soil Moisture Sensing for Smallholder Irrigation
A PhD defense in English: motivation, research questions, related work, sensor and node design, calibration, the field study, accuracy and irrigation results, contributions, limitations and future work.

1 / 12
All examples are AI-generated with FlyPPT.
Source inputExpand
This is the complete input used to generate this example — nothing was edited by hand.
PhD Dissertation Defense Low-Cost Soil Moisture Sensing for Smallholder Irrigation: Design and Field Evaluation Candidate: Alex Morgan — Department of Agricultural and Biosystems Engineering Advisor: Prof. L. Chen 1. Motivation Irrigation accounts for roughly 70% of freshwater withdrawals worldwide, and smallholder farms (under 2 ha) grow a large share of the food in Sub-Saharan Africa and South Asia. Yet most smallholders irrigate by calendar or by feel. Over-irrigation wastes water and pumping fuel and leaches nitrogen below the root zone; under-irrigation cuts yield at critical growth stages. Commercial soil moisture probes cost $150–$800 per node and usually need lab calibration and a data subscription, which puts them out of reach for farmers earning a few dollars a day. Central question: can we deliver "good enough" soil moisture information for under $15 per node — and does it change irrigation decisions on real farms? 2. Research questions RQ1. Can a capacitive sensor built from standard PCB material and a low-power microcontroller reach volumetric water content (VWC) accuracy within ±3% after a simple field calibration? RQ2. How do sensor drift, soil texture, salinity and temperature affect accuracy over a full growing season? RQ3. Does sensor-guided irrigation reduce water use without reducing yield on smallholder plots? RQ4. What do farmers need from the advisory interface to actually act on the readings? 3. Related work - Reference methods: gravimetric sampling is the gold standard but labor-intensive and destructive; time-domain reflectometry (TDR) is accurate but costs $1,000+ per installation. - Commercial capacitance / FDR probes: good accuracy, but high cost and proprietary loggers. - Low-cost sensors in the literature: resistive probes corrode within weeks; hobby-grade capacitive boards show large unit-to-unit variation (±8–12% VWC) and strong temperature sensitivity. - Decision support for smallholders: SMS advisory services and satellite products exist, but satellite soil moisture is kilometer-scale — far too coarse for a 0.5 ha plot. - Gap: no low-cost design has been validated over full seasons, across soil types, with both yield outcomes and farmer adherence data. 4. Sensor and node design - Capacitive fringe-field sensor etched on a 1.6 mm FR-4 board, conformal coating plus epoxy-sealed electronics; 70 MHz oscillator; onboard thermistor for temperature compensation. - Three sensing depths (15, 30 and 45 cm) on a single stake. - Node: low-power microcontroller, long-range sub-GHz radio, 2 × AA batteries; 38 µA average sleep current; estimated battery life 14 months at 30-minute sampling. - One solar-powered gateway per village relays readings to a phone app and SMS alerts. - Bill of materials: $11.40 per node at 500 units. 5. Calibration method - Two-point field calibration: one reading in air-dry soil and one at saturation, taken by the farmer or extension worker at installation. - Soil-texture correction table built from lab measurements on 6 soils, from sand to clay. - Temperature compensation model fitted on 4,800 lab readings between 5 °C and 45 °C. 6. Field study design - 3 sites in semi-arid regions, 2 growing seasons (2023 and 2024). - 48 smallholder plots (0.2–0.8 ha) growing tomato, onion and maize. - Randomized design: 24 sensor-guided plots vs. 24 farmer-practice control plots. - Ground truth: 2,976 weekly gravimetric samples and 6 TDR reference stations. - Advisory: color-coded thresholds (green = no irrigation needed, amber = irrigate within 2 days, red = irrigate today) delivered by SMS and app. 7. Results — sensor accuracy - Lab: RMSE 1.9% VWC after calibration, versus 7.8% uncalibrated. - Field: RMSE 2.6% VWC across all soils — 2.1% in sandy loam, 3.4% in clay with higher salinity. - Drift: median 0.4% VWC per month; 93% of nodes stayed within ±3% after a full season. - Temperature compensation reduced diurnal error from 2.2% to 0.6% VWC. - Reliability: 91% of nodes still operating at season end; most failures were cable damage from rodents and hand tools. 8. Results — irrigation outcomes - Water use: sensor-guided plots applied 31% less irrigation water (412 mm vs. 597 mm per season on average). - Yield: no significant difference (tomato +4%, onion −2%, maize +1%; all p > 0.3). - Water productivity: +38% (kg of crop per m³ of irrigation water). - Pumping cost: down 27%, an average saving of $46 per plot per season — the node pays for itself within one season. - Adherence: farmers followed the advisory on 72% of days; adherence was higher for SMS alerts than for the app. 9. Contributions 1. An open hardware design for a sub-$12 capacitive soil moisture node with built-in temperature compensation. 2. A two-point field calibration procedure that brings low-cost sensors within ±3% VWC without lab equipment. 3. A two-season randomized field evaluation linking low-cost sensing to water use, yield and cost on smallholder plots. 4. Design guidelines for farmer-facing irrigation advisories: color-coded thresholds, SMS first, one action per message. 10. Limitations - Heavy clay and saline soils (EC above 4 dS/m) remain outside the ±3% target. - Two seasons only; durability beyond 18 months is untested. - 48 plots at three sites; results may not generalize to other crops or to rain-fed systems. - Advisory thresholds were crop-generic and did not model evapotranspiration. 11. Future work - Salinity-aware calibration using a second measurement frequency. - Combining node data with weather forecasts for predictive irrigation scheduling. - Shared-node models: one sensor serving neighboring plots through spatial interpolation. - A cooperative-scale pilot with 500 nodes and a pay-per-season service model. 12. Publications and acknowledgments - 3 journal articles (2 published, 1 under review) and 2 conference papers from this work; hardware files and calibration code released under an open license. - Thank you to my advisor and committee, the field teams, and the farmers who hosted the trials on their land. Questions?