Model hub

Every model states its method, its data source and its status. Experimental models are never presented as operational.

Agronomy models

Leaching requirement

Available

Method: FAO-29: LR = ECw / (5·ECe,t − ECw), with crop salt-tolerance thresholds (Maas–Hoffman).
Inputs: Calibrated profile ECe, irrigation water EC, crop.
Outputs: Leaching fraction and required leaching depth (mm).

Version 1.0 · Physics-based

Soil salinity forecast

Experimental

Method: LightGBM baseline, then LSTM once field data accumulates.
Inputs: EC, moisture and water-table series, ET0, rain, irrigation events.
Outputs: Forecast ECe per depth over 7 days.

No release yet

Water table forecast & drainage trigger

Experimental

Method: LightGBM with an operating rule at the crop's critical depth.
Inputs: Water-table depth, irrigation, rain, ET0.
Outputs: Forecast depth and a drainage run recommendation.

Version 0.1 · Physics-based

Irrigation scheduling

Available

Method: FAO-56 water balance with a salinity stress factor as the baseline, plus a DDPG agent trained in simulation. Agent actions pass a rule-based safety envelope.
Inputs: Soil moisture, crop and stage, ET0, salinity.
Outputs: Next irrigation date and depth (mm).

Version 1.0 · Physics-based

Salt-stress detection from images

Planned

Method: CNN with transfer learning.
Inputs: Field or drone images.
Outputs: Salt-stress map.

No release yet

Field salt balance

Available

Method: Salt mass in with irrigation versus out with drainage, and the change in profile storage.
Inputs: Irrigation and drainage flow and EC, profile ECe.
Outputs: Net salt accumulation or removal (t/feddan).

Version 1.0 · Physics-based

Aquaculture models

Un-ionized ammonia

Available

Method: Emerson et al. (1975): NH3 fraction of total ammonia as a function of pH and temperature.
Inputs: Total ammonia nitrogen, pH, water temperature.
Outputs: NH3 concentration and risk level.

Version 1.0 · Physics-based

Dissolved oxygen & dawn risk

Available

Method: DO saturation (APHA 4500-O) and projection of the overnight decline to dawn.
Inputs: DO, temperature and salinity readings.
Outputs: Percent saturation, projected dawn minimum and aeration advice.

Version 1.0 · Physics-based

Daily feed ration

Available

Method: Tilapia feeding tables by size and temperature, reduced when oxygen is low or ammonia is high.
Inputs: Fish count, average weight, water temperature, DO, ammonia.
Outputs: Daily feed, number of meals and feed conversion ratio.

Version 1.0 · Physics-based

Company management models

Subscription demand forecast

Planned

Method: Time-series forecasting (ETS or Prophet).
Inputs: Past subscriptions, season, areas.
Outputs: Expected subscriptions next season.

No release yet

Farmer churn risk

Planned

Method: Gradient-boosting classifier.
Inputs: Usage, response to advice, payments.
Outputs: Probability each farmer does not renew.

No release yet

Company KPIs

Planned

Method: Indicator dashboard (no machine learning).
Inputs: Revenue, cost per feddan, water saved.
Outputs: Monthly and seasonal indicators.

No release yet