Satellites that measure microwave energy have been used since the 1970s to estimate how much of the ocean is covered by sea ice. These measurements rely on algorithms that convert satellite signals into sea ice concentration (SIC) based on their relationship to surface properties. These SIC products are widely used in climate studies and are even included in global weather and climate reanalyses. However, they are not direct observations and can contain significant errors, especially near the ice edge or during melting or freeze-up seasons.

In this study, we develop a method to check whether these satellite-derived SIC values are physically consistent. We do this by predicting what the satellite signal should look like, based on model and reanalysis data such as surface temperature, ice thickness, and wind. This process is called brightness temperature emulation. By comparing the emulated and observed satellite signals, we can detect when the satellite SIC retrievals are likely biased.

We apply this method to ERA5 reanalysis and CMIP6 climate model outputs, and analyze the errors using phase-space analysis—a way to understand how different environmental factors jointly affect the results. Our approach helps improve satellite-based sea ice estimates and their use in climate models.