Why We Need Sensors
Our first season crop loss made the case for data infrastructure on this farm. We will rely on rainfed farming for one or two more seasons while we build capital, but when we are ready to invest in irrigation, we will need to measure rainfall, soil moisture, temperature, humidity, wind, and solar radiation so we can make decisions based on numbers instead of guesses.
What We Compared
We evaluated weather stations across three price tiers and soil sensors across two.
Weather stations we looked at:
- Budget tier (under $200): Ecowitt WS-2902, Ambient Weather WS-2000. WiFi-only, mechanical anemometers, tipping-bucket rain gauges. Adequate sensors but every moving part is a failure point in salt air.
- Mid-range ($300-400): WeatherFlow Tempest. No moving parts, solar-powered sensor, lightning detection, IP67 rated. Genuinely impressive for $349. This was our second choice.
- Professional ($1,000+): Davis Vantage Pro2 line. The standard in agricultural weather monitoring across Kenya. Expandable, well-documented API, 40-year track record.
Soil sensors we looked at:
- DIY tier ($30-60 per node): DFRobot capacitive sensors on ESP32 boards. Cheap but requires programming, manual calibration, custom firmware, and ongoing maintenance. We do not have the time for that right now.
- Plug-and-play LoRaWAN ($100-200 per node): Dragino LSE01, SenseCAP S2105, Milesight EM500-SMTC. Factory-calibrated, 5-10 year battery life (manufacturer rates 10 years; hot climates reduce this to 5-7 years), IP66/IP67 rated. No programming needed.
- Alibaba 7-in-1 RS485 sensors ($35-77): Measure pH, NPK, EC, moisture, temperature. The NPK readings are not laboratory-grade — they are useful for trend monitoring only, not for making nutrient decisions.
What We Chose
Weather Station: Davis Vantage Pro2 Plus with Sonic Anemometer
| Component | SKU | Price (USD) |
|---|---|---|
| Wireless Vantage Pro2 Plus ISS (standard shield) | 6327M | $1,310 |
| WeatherLink Live (headless, WiFi + Ethernet) | 6100 | $425 |
| Sonic Anemometer (replaces mechanical) | 6415 | $725 |
| Subtotal | $2,460 |
The Pro2 Plus includes temperature, humidity, barometric pressure, rainfall, wind speed and direction, solar radiation, and UV index. Solar radiation is critical — it feeds the evapotranspiration calculation that tells us how much water the crops are actually losing.
We chose the standard (passive) radiation shield over the fan-aspirated version deliberately. The fan adds $420 and introduces a motor that corrodes in salt air. In Gongoni’s 24-30 degree climate with consistent coastal breeze, the passive shield is adequate.
The WeatherLink Live was chosen over the Console because it has an Ethernet port for a more reliable connection to a 4G router, and we do not need a touchscreen display sitting in a dusty field. We monitor from a phone.
Why the Sonic Anemometer
This is the single most important upgrade. The standard Davis anemometer uses mechanical wind cups with bearings. In coastal salt air, those bearings corrode and seize within 1-2 years. Replacement costs $50 per year for the bearing cartridge, plus labour.
The sonic anemometer has no moving parts. No cups, no vane, no bearings. It measures wind using ultrasonic pulses. At $725 it is expensive, but over ten years it saves roughly $500 in parts and eliminates the most common failure mode for weather stations near the coast. For a station that we need to trust for irrigation and planting decisions, that reliability is worth paying for.
Soil Sensors: Dragino LSE01
| Component | Qty | Unit Price (USD) | Total (USD) |
|---|---|---|---|
| Dragino LSE01 (moisture, temperature, EC) | 4 | $147 | $588 |
Each sensor measures soil moisture, temperature, and electrical conductivity using Frequency Domain Reflectometry. They communicate via LoRaWAN, run on a lithium battery rated for 10 years at 15-minute reporting intervals (realistically 5-7 years in Kilifi’s heat), and are IP66 rated (dust-tight and resistant to heavy rain).
The Dragino LSE01 was chosen over the SenseCAP S2105 ($144) for one specific reason: factory calibration for saline-alkali and loamy soils. Our Gongoni site has black cotton soil with coastal salinity, so calibration accuracy matters. The SenseCAP is a strong sensor — if we were inland, we might have chosen it instead.
We are starting with four sensors: two in the maize block and two in the green gram block, placed at the one-third and two-third points along each block’s longest axis. That gives one sensor per 1.5 acres.
LoRaWAN Gateway: Dragino LPS8N
| Component | Qty | Unit Price (USD) | Total (USD) |
|---|---|---|---|
| Dragino LPS8N gateway (EU868) | 1 | $157 | $157 |
| Gateway solar power kit (40W panel + 12V 15Ah battery + MPPT controller) | 1 | $100 | $100 |
The gateway receives data from the soil sensors over LoRaWAN at 868 MHz (licence-free in Kenya) and forwards it to the cloud over the internet. One gateway covers the entire six-acre farm easily — LoRaWAN range is 2-10 km in rural areas, and our farm is roughly 250 by 100 metres.
The gateway needs continuous power, so we are running it on a dedicated solar kit. Grid power in rural Magarini is unreliable.
Full Cost Breakdown
| Item | USD | KES (approx.) |
|---|---|---|
| Davis Vantage Pro2 Plus ISS | $1,310 | 170,300 |
| WeatherLink Live | $425 | 55,250 |
| Sonic Anemometer | $725 | 94,250 |
| 4x Dragino LSE01 soil sensors | $588 | 76,440 |
| Dragino LPS8N gateway | $157 | 20,410 |
| Gateway solar kit | $100 | 13,000 |
| Mounting tripod (Davis 7716) | $175 | 22,750 |
| IoT SIM card (1NCE: 500 MB starter) | $11 | 1,430 |
| Hardware subtotal | $3,491 | ~KES 453,830 |
| Shipping and import duties (~21.5%) | ~$750 | ~97,500 |
| Total landed cost | ~$4,241 | ~KES 551,330 |
Recurring costs:
| Item | Annual Cost |
|---|---|
| WeatherLink Pro subscription (5-minute data) | ~$48/year |
| Safaricom 4G M2M data for gateway backhaul | ~KES 18,000-30,000/year |
| Sensor maintenance (temp/humidity board, connectors, corrosion treatment) | ~$60/year |
| Dragino battery replacement (amortised over 5-7 years, 4 units) | ~$15/year |
| Gateway solar battery replacement (amortised over 2-3 years) | ~$12/year |
| CropNuts sample shipping (Kilifi to Nairobi, 2x/year) | ~KES 2,000/year |
| Total annual | ~KES 25,000-35,000 (~USD 190-270) |
What We Ruled Out
WeatherFlow Tempest ($349): Our second choice. No moving parts, IP67, solar powered, lightning detection, free API with no subscription. For the price, it is remarkable. We ruled it out because it cannot expand — you cannot add soil moisture stations, leaf wetness sensors, or additional temperature probes to a Tempest system. The Davis system has seven remaining channels for future sensors. If budget were the only constraint, the Tempest would be the right call.
Budget weather stations (Ecowitt, Ambient Weather): At $150-200, the price is attractive. But they use mechanical anemometers and tipping-bucket rain gauges — both of which corrode in coastal salt air within 1-2 years. Replacement parts are cheap ($30-50) but the maintenance frequency is the problem. On a remote farm, an unreliable weather station is worse than no weather station.
Fan-aspirated radiation shield ($420 upgrade): Improves temperature accuracy from plus or minus 1.0 degrees to plus or minus 0.5 degrees in direct sun. The improvement is real but modest in our climate (consistent breeze, temperatures rarely above 32 degrees). The fan motor is another salt-air corrosion point, needing replacement every 2-3 years at $125 each. We chose zero moving parts over marginal accuracy.
Davis soil moisture sensors ($95 each plus $525 transmitter station): Davis makes its own soil moisture sensors. They work, but $905 for a 4-sensor soil station — on top of the $2,460 weather station — pushes the total past $3,400 before import. The Dragino sensors at $147 each are cheaper, measure EC in addition to moisture and temperature, and have 5-7 year effective battery life with no external power required.
Handheld NDVI sensors and drones: On six acres, a handheld NDVI sensor or a drone with a multispectral camera is not cost-justified. The DJI Agras T50 drone costs $17,999, and KCAA commercial drone licensing takes 6-12 months. A handheld GreenSeeker costs $500-700 and gives you one data point per walk-through. At our scale, walking the field with your eyes is faster and free. We will revisit drones if the farm grows past 20 acres.
Connectivity Plan
The sensors talk to the gateway over LoRaWAN. The gateway talks to the cloud over Safaricom 4G.
LoRaWAN operates at 868 MHz, which is licence-free in Kenya under the ISM band allocation. However, the Communications Authority of Kenya (CAK) technically requires Type Approval for imported radio equipment. The process costs KES 5,000-10,000 and takes 2-4 weeks. We will file for approval when we import the gateway.
For initial cloud backhaul, we have budgeted a 1NCE IoT SIM (10 euros, 500 MB). At 50-100 MB per month of gateway traffic (sensor data plus firmware updates, NTP sync, and ChirpStack overhead), this SIM will last approximately 6-12 months — it is a starter option, not a long-term solution. We plan to switch to a Safaricom M2M SIM (~KES 1,500-2,500/month) once the 1NCE allowance is exhausted.
There is no public LoRaWAN network in Kilifi County. The nearest community gateways are in Nairobi. We are running our own private gateway with ChirpStack as the network server.
Lightning Protection
Coastal Kenya’s long rains (April-June) bring significant thunderstorm activity. A weather station mast at 3-5 metres is the tallest object on flat farmland, making it a natural lightning strike target. Losing a $2,460 weather station and $745 worth of soil sensors to a single surge would set the project back months, so lightning protection is part of our deployment plan from day one.
The protection has three layers:
- Anemometer cable surge protector. The Davis 7854 surge protector (~$25) installs inline on the cable between the sonic anemometer and the ISS. This is the most exposed cable run on the mast and the most likely path for a surge to enter the electronics.
- Copper earthing rod. A 2-metre minimum copper earthing rod driven into the soil at the mast base, connected to the mast with 10mm² copper bonding cable. This gives lightning current a low-resistance path to ground rather than through the sensor electronics.
- Surge-protected power strips. The LoRaWAN gateway and WeatherLink Live are connected through surge-protected power strips at the indoor end. This protects against surges entering through the power or data cables.
The total cost for lightning protection is approximately KES 5,000-8,000 ($38-62) — a small addition to a KES 551,000 sensor deployment that protects against a total-loss scenario.
What We Still Do Not Know
Soil type. The Dragino LSE01 is factory-calibrated for saline-alkali and loamy soils, and our Gongoni site has black cotton soil with coastal salinity. But “should match” is not the same as “verified.” Before we deploy the sensors, we need a laboratory soil test from CropNuts (KES 8,000-15,000 per full test) to confirm soil composition and establish a calibration baseline. Without that test, the moisture readings will be directionally correct but not precise.
Gateway placement. LoRaWAN range is 2-10 km in rural flat terrain, and our farm is well within that. But the actual range depends on antenna height, obstructions, and terrain. We plan to mount the gateway on a 3-5 metre pole at the highest point on the farm, but we will not know the exact coverage until we test it on-site.
Salt air impact timeline. The Davis Pro2 Plus housing is UV-stabilized ASA plastic, which should handle coastal conditions. The sonic anemometer has no moving parts to corrode. But we do not have first-hand data on how fast the humidity sensor drifts in our specific environment. The manufacturer suggests replacing the temperature/humidity sensor board every 12-18 months in coastal sites ($60 per replacement). We will track actual drift against laboratory readings to determine our own replacement schedule.
Status
Researched and costed. Not yet funded. The total landed cost of approximately KES 551,000 is significant for an early-stage farm. This is not a next-month purchase — it is a goal we are working toward, and we will deploy it before the equipment rather than after the next crop loss.
When we fund and deploy the system, we will publish the installation process, the first data, and whether the numbers match what we expected.