A Preview of Berkeley Earth’s New High-Resolution Temperature Data Set

Since its founding, Berkeley Earth’s mission has been to provide the most accurate, independent, and reliable record of climate change for the global climate science and policymaker communities. 

With a beta release expected in early 2023, the next iteration of Berkeley Earth’s data set will represent a significant improvement in the ability to represent the long-term impacts of climate change at the local level.

Representing the Local Impacts of Climate Change

Berkeley Earth’s New High-Resolution Data Set (lower right), in comparison to the other three instrumental temperature data sets from NOAA, NASA, and HadCRU.

While the global target of keeping warming below 1.5°C has helped galvanize global action on decarbonizing and reducing emissions, the fact remains that no one actually lives at this global average. The impacts of global warming are undeniably experienced — and largely mitigated — at the local and regional levels. And policymakers, researchers, and business leaders need data that shows this. 

With 4x the resolution fo our existing data set, the Berkeley Earth High-Resolution Surface Temperature Data Set will represent historical weather and warming trends at a 0.25° x 0.25° spatial resolution. Designed to better capture small-scale spatial structure present in local weather and climate patterns, our new product will allow for visualization of local warming trends that until now have been unavailable. 

Capturing Spatial Variation in the World’s Highest-Resolution Temperature Time Series

Berkeley Earth’s High-Resolution Data Set (upper image), will help reveal how mountain ranges, coastlines, and other local variations interact with global climate changes. 

The new analysis uses advanced machine learning techniques to better incorporate patterns of weather variability and geographic variations. Berkeley Earth’s High-Resolution Data Set brings a number of key advantages compared to existing data products:

  • Improved visualization of local impacts — Crucial to key stakeholders worldwide, the enhanced representation of historical weather variations and trends at the local level will enable improved decision-making for mitigation and adaptation strategies; 
  • Improved input for climate and risk modeling — Improved representation of historical warming trends at the local level will better constrain, and therefore help improve, climate modeling, allowing for more accurate scenario modeling in both policy and risk management applications; 
  • Highest-resolution instrumental data product — Using surface observations, rather than satellites or reanalysis, Berkeley Earth’s High-Resolution Surface Temperature Data Set is by far the most detailed data set of its kind and offers less uncertainty than other temperature data sets at similar resolution; and
  • Historic time series from 1850-present — Consistent with our existing temperature data products, the enhanced high-resolution data set will continue to represent historical time series dating back to 1850, providing the most complete and accurate picture of historical warming trends of any temperature data product available. 

Of the other instrumental temperature data sets that feed the IPCC analysis of long-term climate change, two operate on 5°x5° grid scale and one operates on a 2°x2° grid scale.  In comparison to these products, Berkeley Earth’s shift to a 0.25° x 0.25° resolution will denote a quantum leap forward in the ability to characterize and represent long-term local changes. 

New Machine Learning Algorithms Improve Resolution

The new data product (upper image), will represent a 4x improvement in resolution compared to Berkeley Earth’s existing temperature data set.

This improvement in resolution has been made possible by two of Berkeley Earth’s innovations:

First, Berkeley Earth’s existing algorithms allow for the incorporation of significantly more instrumental observations than any other temperature product on the market, allowing for direct analysis on finer scales. 

Second, we have combined these observations with new algorithms that leverage machine learning technologies to recognize naturally-occurring weather patterns, and faithfully reproduce the associated small-scale variability.  In combination, these techniques provide insights into local variability that are unprecedented among global instrumental analyses.

Beta Release in Early 2023

We expect to make a beta version of the new data product publicly available early next year, with final publication of results expected in late 2023.  

For more information, licensing questions, or for priority access to the beta release, please contact admin@berkeleyearth.org.

We have updated our Privacy Policy to reflect the use of personalized advertising cookies placed on our website. By continuing to use our site, you acknowledge that you accept our Privacy Policy.

I accept