The DailyMeteo dataset
DailyMeteo publishes a daily gridded weather dataset covering Europe and the rest of the world's land area, from 1961 to 2026, at 1 km resolution. It has four variables — maximum (TMAX), minimum (TMIN), and mean (TMEAN) temperature, and total precipitation (PRCP) — each also available as monthly, annual, and long-term mean summaries. New days are added as our processing pipeline runs, with minimal lag.
- Coverage
- Global 1961–2026
- Resolution
- 1 km grid
- Accuracy
- R² up to 98%
- Updates
- Near-real-time
01 · Methodology
How the data is made
Observations come from public meteorological networks — GHCN-Daily, ECA&D, and GSOD. Duplicate stations and outliers are filtered out before anything is gridded.
Early, Late & Final products
The same variables are published in three streams, each trading lag against how thoroughly the underlying observations have been checked.
| Product | Early | Late | Final |
|---|---|---|---|
| How often is it made? | Every day | Once a month | Once a year |
| Lag | 4 - 8 days | 3 months | 1 year |
| Observations | GSOD | GHCN-Daily, ECA&D | GHCN-Daily, ECA&D |
| Gross error removal? | No | No | Yes |
A few auxiliary covariates go into the temperature model: a geometric temperature trend, a digital elevation model, and a topographic wetness index, combined through multiple linear regression to model the trend surface. Daily air temperature is then interpolated with space-time regression kriging, using that trend surface along with the elevation model and wetness index as covariates.
Precipitation — Europe & North America
- 1
Indicator kriging
Space-time indicator kriging predicts whether precipitation occurred at all.
- 2
Ordinary kriging
Space-time ordinary kriging then predicts the total amount, at the locations where it did.
Precipitation — everywhere else
We use the approach from NGCD-1: two surfaces are interpolated separately with Delaunay triangulation.
- 1
Precipitation amount
The raw precipitation surface, triangulated directly from station values.
- 2
Station elevation
A matching elevation surface, triangulated from the same station network.
The difference between that elevation surface and the 1×1 km terrain model is then used to correct the precipitation grid: a 10% adjustment per 100 m of elevation below 1,000 m above sea level, and 5% per 100 m above that.
02 · Format
Naming convention & units
Files follow var_day_yyyymmdd_proj.tif, e.g. tmax_day_20201231_3035.tif:
| Param | Meaning |
|---|---|
var | The variable — tmax, tmin, tmean, or prcp. |
proj | The projection — EQUI7 (EPSG:3035). |
Units — the API returns real values, not the scaled integers stored in the files:
| Param | Meaning |
|---|---|
tmean / tmax / tmin | °C ×10 |
prcp | mm ×10 |
03 · Services
Ways to get at the data
DailyMeteo offers a few ways to get at the data:
Point Query
One variable at a location, with a chart, plus CSV and plot download.
Polygon Query
One or more variables over a chosen time range and a drawn polygon (up to 3,000 km²). You get an email when it's ready.
Bulk Location Download
Every variable, the full time range, one location. Also delivered by email once it's done.
MeteoChat
Our AI assistant — ask for historical weather in plain language, described below.
04 · Validation
Data quality
We validate the grids with leave-one-station-out cross-validation for Europe and 10-fold cross-validation elsewhere. Coefficients of determination (R²) come out above 96% for the temperature variables and above 76% for precipitation. RMSE is 1.3°C for mean temperature, 1.6°C for max, 1.8°C for min, and 2.5 mm for precipitation.
R²
| R² | Max temperature | Min temperature | Mean temperature | Precipitation | ||||
|---|---|---|---|---|---|---|---|---|
| Continent | 1961–1990 | 1991–2020 | 1961–1990 | 1991–2020 | 1961–1990 | 1991–2020 | 1961–1990 | 1991–2020 |
| Europe (low-lying) | 0.975 | 0.979 | 0.962 | 0.961 | 0.98 | 0.981 | 0.776 | 0.763 |
| Europe | 0.976 | 0.977 | 0.963 | 0.959 | 0.981 | 0.979 | 0.784 | 0.772 |
| North America | 0.96 | 0.954 | 0.955 | 0.948 | 0.937 | 0.968 | 0.495 | 0.559 |
| South America | 0.818 | 0.886 | 0.895 | 0.934 | 0.883 | 0.936 | 0.391 | 0.374 |
| Africa | 0.882 | 0.909 | 0.886 | 0.923 | 0.913 | 0.941 | 0.324 | 0.134 |
| Asia | 0.977 | 0.98 | 0.973 | 0.976 | 0.981 | 0.985 | 0.453 | 0.3 |
| Oceania | 0.948 | 0.956 | 0.923 | 0.931 | 0.905 | 0.966 | 0.619 | 0.573 |
RMSE
| RMSE | Max temperature | Min temperature | Mean temperature | Precipitation | ||||
|---|---|---|---|---|---|---|---|---|
| Continent | 1961–1990 | 1991–2020 | 1961–1990 | 1991–2020 | 1961–1990 | 1991–2020 | 1961–1990 | 1991–2020 |
| Europe (low-lying) | 1.64 | 1.51 | 1.81 | 1.7 | 1.36 | 1.26 | 2.33 | 2.53 |
| Europe | 1.6 | 1.54 | 1.77 | 1.74 | 1.31 | 1.34 | 2.27 | 2.53 |
| North America | 2.54 | 2.67 | 2.44 | 2.54 | 3.74 | 2.07 | 5.4 | 5.6 |
| South America | 2.8 | 2.18 | 2.42 | 1.82 | 2.3 | 1.66 | 7.95 | 8.62 |
| Africa | 2.56 | 2.47 | 2.3 | 2.14 | 2.21 | 1.86 | 5.22 | 6.7 |
| Asia | 2.5 | 2.24 | 2.71 | 2.47 | 2.32 | 1.89 | 5.96 | 7.57 |
| Oceania | 1.78 | 1.64 | 2 | 2.03 | 1.8 | 1.26 | 4.4 | 4.9 |
05 · Assistant
MeteoChat
MeteoChat is GPT-3.5 Turbo, fine-tuned on top of the DailyMeteoAPI, so you can ask for historical weather in plain language instead of building a query by hand — something like “what was the average max temperature in Lisbon in July 2019.” It works out which endpoint that maps to, calls it, and hands back the data.
It keeps improving as people use it — questions asked through the assistant feed back into refining the model over time.