Aerosol mixing state drives a supersaturation-dependent bias in cloud condensation nuclei hygroscopicity retrievals

Abstract. Cloud droplet number depends on aerosol hygroscopicity, the readiness of aerosol particles to take up water. Operational products retrieve this property, denoted κ, from cloud condensation nuclei (CCN) measurements and are widely used to constrain and evaluate models. The retrieval infers an activation threshold by matching cumulative particle number to measured CCN concentration, assuming that every particle above this diameter activates and none below it. It is shown here that this assumption fails at a remote marine site, producing a systematic bias in retrieved κ. At the ARM Eastern North Atlantic site, retrieved κ falls from 0.28 to 0.06 across the measured supersaturation range, diverging from the value predicted by aerosol composition. A coincident hygroscopicity tandem differential mobility analyser record shows that genuine size-dependent hygroscopicity accounts for only about one-fifth of this decline and measures substantial hygroscopic heterogeneity. Forward-simulating the retrieval using the measured κ distributions reproduces the observations at every setpoint, accounting for 92–118 % of the discrepancy. Insufficiently hygroscopic particles fail to activate despite their size, causing the inferred activation diameter to be too large and biasing κ low through its inverse-cubic dependence on diameter. The observed super-saturation dependence is therefore largely a mixing-state-dependent retrieval artefact rather than an aerosol property. When propagated through an aerosol activation calculation, use of retrieved rather than predicted κ produces an approximately 9–20 % underestimate in predicted cloud droplet number. More generally, integration-based CCN retrievals can convert aerosol mixing-state heterogeneity into apparent supersaturation dependence, affecting observational constraints on aerosol activation and aerosol–cloud interactions.