Influence of Observational Temperature Data Sets on ECS and TCR Estimates

Plain Language Summary

When studying how the climate responds to increasing levels of greenhouse gases, scientists often use two key metrics: the Equilibrium Climate Sensitivity (ECS) and the Transient Climate Response (TCR). However, the estimates for these metrics can vary depending on the historical temperature data product used. Different climate data-centers create their own temperature records, and these can differ in many ways. Even within a single data-center's product, there are variations. In our study, we examine these differences, particularly focusing on how each individual data product (or various groupings of them) influence the likelihood estimates of ECS and TCR. Our analysis suggests that the estimates of ECS and TCR are highly sensitive to the choice of observational data sets. Even when we group temperature data from sources using similar sea temperature records, it skews the results. Additionally, we find that variations within one data set can be as impactful as differences between multiple ones for the estimates of ECS and TCR.

Abstract

Uncertainties in estimates of Equilibrium Climate Sensitivity (ECS) and Transient Climate Response (TCR) are influenced by observational temperature data sets. Variability exists not just among the data products, but also within the creation of each one. This includes significant variations among ensemble members within a single data product. 

Using the optimal fingerprint approach combined with Bayesian updating, we quantify the uncertainties in ECS and TCR estimates arising from both individual data sets and their various groupings. Our methodology, utilizing both spatial and temporal data, shows impacts on the estimates of ECS and TCR. As we assess different groupings of observational data products, we observe that using products sharing identical Sea Surface Temperatures (SST) introduce discernible biases. 

These results highlight that variations among ensemble members within a single data product are as influential as the disparities across multiple data products.