Back to BlogClimate AI Guide
XLinkedInFacebook

Open Climate Datasets for ML Projects in 2027

Suman SharmaSuman Sharma

Finding the right dataset is half of any climate ML project. These are the ones I keep coming back to, grouped by what you're trying to do.

Weather and climate

DatasetWhat it isGood for
ERA5 (ECMWF / Copernicus)Global hourly reanalysis, 1940 onwardWeather features, historical baselines
ERA5-LandHigher-resolution land variablesSoil moisture, land temperature
CMIP6Climate model projectionsFuture scenarios (risk modelling)
CHIRPSRainfall estimates (stations + satellite)Drought, agriculture, tropics
GPM IMERG (NASA)Global satellite precipitationRainfall-triggered hazards

Satellite imagery

DatasetGood for
Sentinel-210 m optical: land cover, crops, forests
Sentinel-1Radar, sees through clouds: floods, surface change
LandsatLong history, thermal bands
Harmonized Landsat Sentinel-2 (HLS)Dense, consistent time series
Sentinel-5PAtmospheric gases (NO₂, CO, etc.)

Comparison: Sentinel-2 vs Landsat.

Land cover and forests

DatasetGood for
ESA WorldCoverGlobal 10 m land cover
Dynamic World (Google / WRI)Near-real-time land cover
Global Forest Change (Hansen et al.)Annual tree cover loss since 2000
JRC Tropical Moist ForestsDeforestation and degradation in the tropics
JRC Global Forest Cover 2020Forest baseline relevant to EUDR

Used in practice in Detecting Deforestation with Satellite Imagery.

Hazards

DatasetGood for
GloFAS (Copernicus)Global river discharge forecasts
Global Flood DatabaseHistorical flood extents
NASA Global Landslide CatalogLandslide events
EM-DATDisaster events and impacts (registration required)

Related: flood early warning, landslide early warning.

Air quality

DatasetGood for
OpenAQAggregated ground measurements worldwide
CAMS (Copernicus Atmosphere)Global atmospheric composition forecasts and reanalysis

Emissions

DatasetGood for
EDGAR (JRC)Global emissions by sector and country
Climate TRACEFacility-level emissions estimates

Tips

  • Start with the cloud versions (Planetary Computer, Earth Engine, AWS) to avoid downloading terabytes.
  • Learn xarray. Most gridded climate data is NetCDF or Zarr.
  • Cite properly. Most datasets require specific attribution.

Frequently asked questions

What is ERA5?

ERA5 is the fifth-generation global atmospheric reanalysis from ECMWF, distributed through the Copernicus Climate Change Service. It provides hourly estimates of many weather and climate variables from 1940 onward.

Where can I access these datasets?

Many are available through the Copernicus Climate Data Store, Copernicus Data Space Ecosystem, NASA Earthdata, USGS, Google Earth Engine, and cloud platforms like Microsoft Planetary Computer and AWS Open Data.

Can I use these datasets commercially?

Many are free for commercial use under open licences, but not all. Check each dataset's licence and attribution requirements before building a product on it.

Contact

Get in Touch

Want to call me? book a call and I'll respond in the available time slot. I will ignore all soliciting.