Data & methodology

How when.garden knows when to plant

Every figure is precomputed from public climate data, then labeled with its source and a confidence score. We never present a modeled grid estimate as a local station reading.

The method, briefly

  1. Derive last/first frost probabilities and frost-free days from the nearest dense station history (NOAA GHCN-Daily), else model them from the NASA POWER grid.
  2. Classify each location's Köppen zone. Frost zones plant relative to frost dates; tropical zones plant relative to wet/dry seasons from POWER precipitation.
  3. Anchor each crop's sow / transplant / harvest windows to those dates using a curated offset table, then attach a confidence derived from data source and season fit.

Sources & licenses

NOAA GHCN-Daily

Public domain (CC0)

Daily station observations — the basis for last/first-frost probabilities (50% & 10%), frost-free days and growing degree days where stations are dense.

NASA POWER Agroclimatology

US-gov open, 0.5° grid

Fills gaps in sparse regions. Anything derived from it is tagged source = modeled and shown with lower confidence.

FAO ECOCROP

FAO open data

Crop temperature, precipitation and photoperiod ranges — the breadth behind our crop list.

Köppen-Geiger 1km (Beck et al. 2023)

CC BY 4.0

Climate-zone classification and the frost ⇄ tropical regime switch.

OpenStreetMap / Nominatim

ODbL

Place names and coordinates.

Building an app on this data? See the developer API.

Provenance & reproducibility

Every published figure resolves to the source snapshot, the parameters, and the pipeline stage that produced it. Frost dates are derived over the WMO reference window with a minimum station history before we publish them.

Derivation parameters
  • Reference window: 1991-2020
  • Minimum years: 20
  • Frost threshold: 0°C
  • Gap-fill: >50 km or >300 m → modeled
  • Method: curated
Source snapshots