We assess the predictability of immersion-mode ice nucleating particles (INPs) at a remote marine site in the Eastern North Atlantic (ENA) using aerosol simulations from a global climate model as inputs to the immersion-mode INP parameterizations. While the model- simulated INP concentrations are lower by one to three orders of magnitudes compared to the measurements, we achieve aerosol-INP closure at ENA using the observed aerosol properties. We demonstrate a novel INP error decomposition approach to quantify the portion of total INP error from different error components. We conclude that inaccuracies in aerosols (surface area and composition) are the dominant cause of the model INP discrepancy at ENA. We recommend that, for future aerosol-INP closure studies, along with the measurements for total INP concentrations, campaigns should also collect co-located aerosol size-resolved composition measurements (in the INP-relevant size range) to better distinguish and quantify the error sources.