21.11.2024

Graphic above: Illustration of the reconstructions of the global mean temperature only from temperature measurements over land (white curve) and from temperature measurements on the ocean surface (blue curve). The new study shows the long-term consistency of both time series and the deviations in the period between 1900 and 1930. Graphic: University of Leipzig

A team led by lead author and junior professor Dr. Sebastian Sippel from the University of Leipzig shows that the measured values ​​of the global ocean surface temperatures are not consistent with the temperatures measured on coasts and over land and other paleoclimatic data. The reason for this assumed too cold ocean cold period is likely to be insufficient information about changes in the measurement methods during this time.

Parallel to the main method used, the DKRZ led by department head Dr. Christopher Kadow investigated the same issue using artificial intelligence methods. For this purpose, the incomplete observation data or those collected using different measurement methods were reconstructed using AI. For the reconstruction, the model was trained with large amounts of data from the Coupled Model Intercomparison Project CMIP6. The calculations were carried out on the GPU part of the DKRZ's HPC system "Levante".

Despite its different approach, the AI ​​method reaches the same conclusions as the main study and thus provides additional security for its results.