Historical Weather Data From 1940 until now
More than 80 years hourly weather data using Copernicus ERA5
3 weeks ago, the Copernicus ERA5 initiative made available additional weather reanalysis information dating back to 1940. After non-stop downloading, the Open-Meteo Historical Weather API now provides access to more than 80 years of past weather data.
ERA5 Overview
ERA5 is a highly regarded weather reanalysis dataset that utilizes a variety of observations from weather stations, aircraft, buoys, radar, and satellites to construct a comprehensive record of past weather conditions. By using mathematical models to estimate missing data, reanalysis datasets are able to fill in gaps and provide detailed historical weather information for locations that may not have had weather stations nearby, such as rural areas or the open ocean.
ERA5 encompasses an extensive range of weather variables, not only at the surface but also at atmospheric levels, and is publicly available through the Copernicus Climate Data Store (CDS). However, navigating the vast amount of data can be daunting, and obtaining a continuous time-series for individual locations can be challenging.
Fortunately, Open-Meteo simplifies the process of utilizing ERA5. Instead of downloading enormous amounts of data from the CDS, users can obtain continuous time-series for key weather variables, including temperature, humidity, precipitation, wind, and solar radiation in hourly resolution since 1940.
Although not all weather variables from ERA5 are available on Open-Meteo due to the enormous amount of storage required to store this much data. Important variables can be added upon request.
Combined with ERA5-Land
ERA5 employs a spatial grid-spacing of 25 km, which is insufficient to represent local effects such as urban heat islands, thunderstorms, or local wind systems. This is particularly relevant in coastal and mountainous regions, where caution should be exercised.
Although the resolution is adequate for the upper atmosphere, processes near the surface would benefit from a higher resolution, which would enable better resolution of local effects, such as near-surface temperature, humidity, soil moisture, and temperature.
To address this issue, the ERA5-Land reanalysis dataset, which has a resolution of 10 km, is available. It incorporates improved land and soil models, enabling it to better represent near-surface processes.
At Open-Meteo, the Historical Weather API seamlessly integrates ERA5-Land, and when data is retrieved for a single location, ERA5-Land with 10 km resolution is automatically selected.
It is worth noting, however, that ERA5-Land is only available starting from 1950, rather than 1940. Hopefully, the next ERA5-Land release will include the remaining 10 years.
Consistency
Ensuring consistency is a crucial aspect of reanalysis data. When dealing with 80 years of continuous data, it can be challenging to prevent statistical errors resulting from different measurement sensors or variations in the way data is assembled.
Changes in sensors used by weather stations or the addition of more satellite observations can slightly alter measured values or significantly increase the amount of observed data available, respectively, which can accidentally introduce change signals that may be misinterpreted. Although there are only a limited number of weather stations and no satellite data available between 1940 and 1970, there are numerous weather stations worldwide with pristine past weather records that span back well beyond 1940.
Despite these challenges, ERA5 endeavors to maintain consistency as much as possible. Its proven track record of consistency is evidenced in the literature, and ongoing studies are being conducted to assess data quality.
Next steps
To access historical weather data, you can utilize the Open-Meteo Historical Weather API, which allows you to download data in JSON, CSV, or XLSX formats.
Moreover, the regular weather forecast API offers access to weather data from previous weeks and months as well. By adding the parameter "&past_days=90", for instance, you can retrieve 90 days of past data in conjunction with 7 days of forecast. The forecast API utilizes weather models with resolutions of up to 1 km, but consistency cannot be guaranteed. Weather models receive updates every few months that slightly enhance accuracy but may introduce varying behaviors. Nonetheless, this data is suitable for training machine learning models to enhance weather forecasts.
To stay informed about the latest advancements in weather-related developments, be sure to subscribe to our free newsletter.