ShambaSense

How it works

From a weather station to advice a farmer can act on today.

  1. 1

    Data

    The JKUAT Conduit@Empathy station in Juja reports temperature, humidity, rain, wind, pressure and UV every few minutes.

  2. 2

    Insight

    Readings are cleaned and rolled up into daily summaries. Reference evaporation (ET₀) and a root-zone water balance estimate how dry the soil is.

  3. 3

    Action

    When the soil passes the crop's stress point, ShambaSense says to irrigate and by how much. Heat risk sets work and rest times.

  4. 4

    Impact

    Watering only when needed saves water compared with a fixed schedule and keeps the crop out of stress.

Limitations

  • All advice comes from one station in Juja. Rain can differ a lot a few kilometres away.
  • There is no soil moisture sensor. Soil water is modelled from rain and evaporation.
  • The model cannot see your own irrigation, so it assumes none has been applied.
  • The soil bucket starts full on the first day of data.
  • Flood, drought and crop thresholds are demo values, and impact figures are estimates.
  • Your farm profile and the reading history are stored only in this browser. Clearing site data removes them.
  • The 3-day forecast is a machine-learning model (scikit-learn) trained on this station's own history. It forecasts temperature and evaporation, not rain, and shows its measured accuracy.
  • The Assistant's answers are written by an AI model (Groq) from the same station data. It can word things imperfectly, so check the tabs for the exact numbers.
  • Every number comes from the ShambaSense backend. If it cannot be reached, the pages say so instead of showing made-up values.