Apollo Terra
Terra × Mitra
Not connected. Only the four-zone scenario summary is shared; coordinates and imported measurement rows are excluded. Mitra advises inspection, never physical actuation.
Root-zone trajectory
Four irrigation zonesZone 1
Select any crop or soil section to inspect its current state.
Observation overlay
Import timestamped soil measurements to compare modelled and measured volumetric water content.
Columns: date, zone, theta1. Date YYYY-MM-DD, zone 1–4, theta1 in m³/m³.
Weather outlook
Load the current provider forecast for the selected location.
| Date | Temperature | Rainfall | ET₀ | Evidence |
|---|
Provider forecasts can be refreshed and compared with later observations. Seasonal and annual operational products remain withheld until locally verified skill is available. API use is for local research; provider commercial terms require separate review.
Historical skill
May 1 issue context, June–September lower-tercile rainfall event. Every prediction trains only on earlier seasons. The analysis uses final reanalysis vintages, which differ from data available in real time.
Out-of-sample event probabilities
Reliability and abstention
| Season | Forecast risk | Observed rain | Training threshold | Outcome | Analog years |
|---|
Experiments
Experiments preserve forcing, parameter values, provenance and results. Hosted records belong to the signed-in researcher; the local edition uses SQLite.
Methods and sources
A field in four dimensions
The three spatial dimensions show crop rows, three soil horizons, roots and irrigation zones. Time is the fourth dimension: every day in the replay uses an actual dated weather record to advance a conserved water ledger. Plant height and root depth are research scenarios; weather forcing alone does not calibrate a crop model.
Water and energy
Rain and scheduled irrigation enter the top layer. Infiltration capacity and layer saturation bound storage. Excess becomes runoff or deep drainage. Evaporation and root uptake can consume only water above the assumed wilting point. A finite hydrogel reservoir exchanges water with soil; its uptake is removed from soil and release returns to soil. The displayed residual is recomputed from the complete ledger.
NASA forcing uses the FAO Hargreaves approximation with extraterrestrial radiation to estimate reference evapotranspiration. MET Norway supplies the hosted short-range weather outlook under CC BY 4.0. Forecast ET₀ is estimated here with Hargreaves, not supplied as a measured value. Daily aggregates include at least 18 hours of forecast coverage; their temperature mean is an approximation. This is a bucket model; layer dimensions, rooting, crop coefficient and treatment effects require independent field calibration.
Historical skill
The first actual benchmark uses 1991–2025 Solapur representative-grid data. A five-neighbor model uses January–April rain and temperature to estimate the probability of below-tercile monsoon rainfall. It is compared with climatology and persistence in rolling-origin seasons after 20 training years. Three-year block bootstrap intervals and reliability accompany the result. Failure to meet any gate produces abstention.