Shallow wells along Kilifi County shorelines show rising conductivity during spring tides. We surveyed 210 households tracking source switching, boiling practices, and spending on vendor water. GIS overlays linked well depth and distance to mangrove buffers with taste complaint intensity. Households nearest eroded channels diversified sources fastest but faced higher costs. Community workshops co-produced a seasonal color-coded advisory using cheap handheld testers. Results inform subsidies for rainwater tanks paired with erosion control rather than tank programs alone.

Traditional disciplinary escalations can silence bullying reporting. Twelve schools tested structured restorative circles facilitated by trained teachers across one academic year. Anonymous termly surveys showed a decline in fear-based avoidance of corridors and cafeterias, with stronger trust in adult follow-through. Office referral data shifted toward earlier, lower-intensity incidents. Implementation fidelity depended on protected timetable slots and principal modeling. Limitations include self-report bias and uneven parent engagement in smaller towns.

Smartphone-based leaf image classifiers often assume reliable connectivity. We distilled a convolutional model to run on mid-range Android devices with batching disabled to save RAM. Field staff in two highland districts captured images of sweet potato foliage; on-device scores triggered SMS escalation only above confidence thresholds. Nighttime charging habits and shared family phones affected usage patterns more than model accuracy. The work underscores co-design with extension officers when rolling computer vision to remote agronomy workflows.

Post-stroke mobility programs are scarce outside Yerevan. We paired tablet-based exercise coaching with weekly physiotherapist video calls for forty-two participants in three provinces. Quantitative gait metrics improved modestly compared with a pamphlet control, but qualitative interviews highlighted electricity outages and family caregivers as decisive moderators. A simple offline video cache and SMS reminders raised adherence. Policy implications include bundled mobile data vouchers and clearer liability guidance for remote coaching.

Motorcycle-taxi and shared minibus routes shift weekly, leaving official maps outdated. We recruited drivers and students to collect anonymized GPS traces during peak and off-peak hours for eight weeks. Community validation sessions merged traces with hazard points such as flooded underpasses. The resulting vector layers improved a pilot trip planner and helped municipal planners discuss dedicated lanes without displacing vendors arbitrarily. Ethical protocols addressed phone battery drain and data ownership.

Erratic rainfall has disrupted coffee and maize incomes in western highland municipalities. Using anonymized loan ledgers and rainfall anomalies from 2016-2024, we trace how groups restructured debts after landslides versus drought. Borrowers with diversified non-farm income and rotating savings circles recovered fastest, while single-crop households faced longer arrears even when interest was frozen. Findings inform graduated relief templates that pair grace periods with soil conservation incentives rather than blanket forbearance.

Low-cost particulate sensors are spreading faster than governance frameworks can absorb them. We coordinated twelve neighborhood groups in two secondary cities to publish hourly PM2.5 and humidity readings under a minimal metadata schema compatible with regional open portals. Workshops focused on calibration drift, duplicate site naming, and privacy around GPS pins. After six months, merged feeds supported school recess advisories and helped municipal staff prioritize truck enforcement routes. The study argues that lightweight standards plus peer review between communities outperform top-heavy certification alone.