Climate Change API

High resolution IPCC climate models from the 6th assessment report

Are you interested in exploring climate data and modelling climate indicators? Are you curious about how climate change is affecting your region? The newly released Open-Meteo Climate API will be interesting for you.

The Climate API provides access to climate data from the Intergovernmental Panel on Climate Change (IPCC), that is the foundation of the 6th assessment report, highlighting ongoing temperature increases, more extreme temperatures, increased rainfall, and melting of Arctic sea ice and glaciers.

Emissions have continued to rise – albeit at a slowing rate – and it will be “impossible” to stay below 1.5C with “no or limited overshoot” without stronger climate action this decade.

Carbon Brief Q&A for the IPCC 6th assessment report

Climate change is not a phenomenon that affects all regions equally. While some areas might experience more frequent and intense rainfall, others are confronted with extended periods of drought, which can have a severe impact on agriculture. The change in temperature patterns is also unevenly distributed, with cities being particularly vulnerable to heatwaves. The graphic below illustrates how areas are effected by changes in temperature, soil moisture and precipitation with ongoing rising temperature.

Source: IPCC 6th assessment report

High-Resolution Climate Models

To better estimate climate change on a regional scale rather than a continental scale, climate models evolved significantly over the past few decades, from large-scale models with resolutions of up to 1000 km to high-resolution models with resolutions as high as 10 km. This increase in resolution allows for more accurate representation of regional and local weather patterns and is critical for studying the impacts of climate change on a smaller scale.

The IPCC "HighResMip" working group aimed at enhancing comprehension of regional climate change through global climate simulations at resolutions ranging from 10 to 50 km. Data from renowned institutes worldwide were made available as Open-Data via the CMIP6 project. Nevertheless, managing the massive data volume of over 15 Petabytes presents a challenge.

This is where Open-Meteo steps in, by downloading the most accurate and highest resolution climate models for common weather variables such as temperature, humidity, wind, clouds, solar radiation, precipitation, and soil moisture. By converting this data into an API, it has become incredibly convenient to access all available climate data for a single location within seconds.

The initial release of the Open-Meteo Climate API provides access to seven climate models that offer data from 1950 to 2050. Unlike other simulations, the data is not calculated for different emission scenarios and is only available until 2050. While emission scenarios are crucial for estimating climate change until 2100, they do not immediately affect the years until 2050. The graph below illustrates how various scenarios of increasing emissions result in higher temperature increases after 2060.

Effect of different emission scenarios on temperature (IPCC 6th assessment report)

Daily Data from 1950 until 2050

The Climate API provides daily climate data rather than climate indices such as the number of heat days or tropical nights. This level of detail enables users to calculate climate indices themselves with greater flexibility.

Daily data can be used to run models to predict for example crop yield, pest infestation, water balance and photovoltaic energy production. Using multiple models allows for better expression of uncertainty compared to relying on a single climate model. The chart below show the increase in temperature in Berlin, Germany from 1950 until 2050 as predicted by multiple climate models.

Temperatures from 1950 until 2050 in Berlin from seven climate models (Source: Open-Meteo)

Although the Climate API provides data from past and recent years, it should not be confused with actual weather measurements or reanalysis. The control period in climate data from 1950 until 2015 is primarily used for model validation, rather than representing the past weather.

Comparing data from a specific day to measurements or weather model data bears no relation to the past weather, as weather patterns resulting from climate models do not match in place and time with past weather. Furthermore, a single year in a climate model may be hotter or colder than averages, but it provides no indication of the weather in the upcoming year. Climate models simulate changes over extended periods and do not forecast day-to-day weather.

Climate Model Comparison and Downscaling

Climate models are based on different numerical methods, run at varying resolutions, and incorporate different assumptions regarding how the Earth's climate system responds to emissions. The period from 1950 to 2015 is commonly used to assess the accuracy and reliability of climate models before using them to make future predictions with increased emissions. Despite this, direct comparisons between multiple models can still yield different outcomes.

Let's consider an example with the city of Lucerne in Switzerland, which is located near the Swiss Alps. When comparing multiple climate models, there can be temperature differences of up to 6 Kelvin as illustrated below.

Uncorrected (Raw) climate forecast for Lucerne, Switzerland

These differences arise due to varying model resolutions and assumptions about mountainous terrain versus urban areas.

To address this issue, Open-Meteo uses ground truth data, such as measurements, to adjust the climate forecast. This is achieved by downscaling all climate models to a 10 km resolution using the ERA5-Land reanalysis as reference data, which is available through the Historical Weather API.

Statistical errors are corrected by calculating average long-term averages from 1950 to 2015, and climate normals are calculated for different months to capture seasonal effects. The resulting error-corrected data ensures that climate models are more comparable and suitable for different applications, although it does not guarantee that the data accurately represents future conditions and may hide systematic error in climate models.

Once the climate forecast for Lucerne is adjusted, all climate models demonstrate a similar trend in relation to each other (see below).

Error corrected climate forecast for Lucerne, Switzerland

The default setting in the Open-Meteo Climate API provides error-corrected data, but raw data is available as well. However, it is still recommended to analyze the entire 1950 to 2050 time range and compare it with actual past weather data from the Historical Weather API.

Climate API Documentation

The Open-Meteo Climate API documentation provides detailed information and an overview of all available climate models and data. The data can be visualized worldwide instantly. Although the API is primarily designed to support further data analysis rather than illustrating climate change to the general public, it certainly can be a good foundation for other open-source projects aiming at visualizing climate change at a local level.

As research on climate change progresses, additional climate models may be added to the API in the future to improve reliability. To keep up with the latest developments, you can subscribe to this blog for free and will be immediately notified via email.

If you have any questions or suggestions, feel free to leave a comment below.