Data Sources

Seven public datasets integrated into the AgriTwin database

All data is downloaded and processed by the agritwin-etl pipeline. Raw files are stored in agriTwin-etl/data/; processed Parquet files in data/processed/. The loader bulk-copies Parquet into PostgreSQL via COPY FROM STDIN.

Weather

ERA5-Land

ECMWF / Copernicus Climate Data Store

Monthly reanalysis of land-surface climate variables at ~9 km resolution, covering the full ERA5 archive. Stored at H3 res-6 cells only (~1,007 cells over Konya). Higher-resolution cells look up their res-6 parent at query time.

Spatial resolution~9 km (0.1°)
H3 level storedres-6
Temporal stepMonthly
Coverage2018 – 2023
Feature nameUnitNotes
temperature_2m°CMonthly mean of 2 m air temperature
precipitationmm/monthMonthly total (daily mean × days in month)
solar_radiationMJ/m²/monthSurface solar radiation downwards, monthly total
dewpoint_2m°CMonthly mean 2 m dewpoint temperature
wind_u_10mm/sEastward wind component at 10 m
wind_v_10mm/sNorthward wind component at 10 m
Weather

ERA5-Land — Minimum Temperature

ECMWF / Copernicus Climate Data Store

Separate ERA5-Land download for daily minimum temperature, resampled to monthly minima. Used in crop suitability scoring to assess frost risk during critical growth stages.

Spatial resolution~9 km (0.1°)
H3 level storedres-6
Temporal stepMonthly
Coverage2018 – 2023
Feature nameUnitNotes
temperature_2m_min°CMonthly minimum of daily 2 m minimum temperature
Vegetation

MODIS MOD13A3

NASA Land Processes DAAC

Monthly vegetation index product derived from Terra MODIS imagery. Provides NDVI (Normalized Difference Vegetation Index) and land cover classification at 500 m resolution, mapped to H3 res-9 cells.

Spatial resolution500 m
H3 level storedres-9
Temporal stepMonthly
Coverage2018 – 2023
Feature nameUnitNotes
ndviNDVI index (−1 to 1); higher values = denser green vegetation
land_cover_typeclassMODIS MCD12Q1 IGBP land cover class (integer); aggregated by mode
Water

MODIS MOD16A2 — Evapotranspiration

NASA Land Processes DAAC

8-day global terrestrial evapotranspiration product derived from MODIS land cover, albedo, and LAI inputs. Actual ET reflects real vegetation water use and is a direct input to irrigation planning.

Spatial resolution500 m
H3 level storedres-9
Temporal step8-day
Coverage2018 – 2023
Feature nameUnitNotes
actual_etmm/8-dayActual evapotranspiration per 8-day composite period
Soil

SoilGrids v2.0

ISRIC — World Soil Information

Global gridded soil property predictions at 250 m resolution from the ISRIC World Soil Information service, derived from machine learning on soil profiles, remote sensing, and terrain data. Four depth layers (0–5, 5–15, 15–30, 30–60 cm) are averaged to a single mean value per cell. These are static — a single observation per cell, no temporal dimension.

Spatial resolution250 m
H3 level storedres-9
Temporal stepStatic (2021)
Depth averaged0–60 cm (4 layers)
Feature nameUnitProperty
soil_phpHSoil pH in water (phh2o)
soil_organic_carbong/kgSoil organic carbon content (soc)
soil_clayg/kgClay fraction mass fraction
soil_sandg/kgSand fraction mass fraction
soil_siltg/kgSilt fraction mass fraction
soil_bulk_densitykg/dm³Fine earth bulk density (bdod)
soil_cecmmol/kgCation exchange capacity at pH 7
soil_nitrogeng/kgTotal nitrogen content
Terrain

SRTM Digital Elevation Model

NASA / USGS

Shuttle Radar Topography Mission 1 arc-second (~30 m) global DEM. Elevation, slope, and aspect are derived from the DEM tiles and stored as columns directly on the spatial_cell table (not in the observation table). These are static.

Spatial resolution~30 m (1 arc-sec)
H3 level storedres-7/8/9 cells
Temporal stepStatic (2000)
AggregationMean elev/slope; circular mean aspect
ColumnUnitDescription
elevationmMean elevation of H3 cell pixels
slope°Mean slope angle of H3 cell pixels
aspect°Circular mean aspect (0°=N, 90°=E, 180°=S)
Economics

FAOSTAT Producer Prices

Food and Agriculture Organization of the United Nations

Annual producer price data for Turkey by crop commodity, sourced from the FAOSTAT API. The latest available year is used for each crop. Prices are in USD/tonne and feed directly into the profit projection formula on the Yield & Profit page.

Spatial resolutionNational (Turkey)
Temporal stepAnnual
Tablecommodity_price
CropPrice (USD/tonne, approx.)
Wheat~240
Barley~210
Sugar Beet~45
Sunflower~530
Maize~220
Chickpea~680
Lentil~620
Cotton~1,450
Economics

TAGEM Production Costs

Turkish Ministry of Agriculture and Forestry — TAGEM

Itemised agricultural production cost surveys published by the Turkish Ministry of Agriculture and Forestry (TAGEM) for the 2022 season. Costs cover seed, fertilizer, pesticide, fuel, machinery, labor, and overhead. A CPI adjustment factor of 1.40 is applied to convert 2022 USD costs to approximate 2024 values, reflecting Turkish agricultural input inflation.

Spatial resolutionNational (Turkey)
Reference year2022 → adjusted to 2024
CPI factor1.40
Tableproduction_cost

Cost categories stored per crop: seed, fertilizer, pesticide, fuel, machinery, labor, overhead. All are surfaced in the cost breakdown card on the Yield & Profit sidebar.

Agronomy

Crop Requirements

FAO AgroEcological Zones + literature

Monthly crop climate and soil requirements hand-compiled from FAO AgroEcological Zone guidelines and peer-reviewed agronomy literature. Stored in the crop_requirement table. Used by the ETL scoring engine to compute suitability via the trapezoidal fuzzy membership function.

Crops8
Features per croptemperature, precipitation, solar radiation, min temperature, soil pH
Temporal granularityMonthly (growing season only)
Membership paramsreq_min, req_optimal, req_max per month per feature