Abstract. The carbonate chemistry of sea ice is known to play a role in global carbon cycles, but its importance is uncertain in part due to disparities in reported results. Variability in physical and biological drivers is usually invoked to explain differences between studies. In the Canadian Arctic Archipelago, “invisible polynyas” – areas of strong currents, thin ice, and potentially high biological productivity – are examples of extreme spatial variability. We used an invisible polynya as a natural laboratory to study the effects of inferred initial ice formation conditions, ice growth rate, and algal biomass on the distribution of carbonate species by collecting enough cores to perform a statistical comparison between sites located within, and just outside of, a polynya near Iqaluktuttiaq (Cambridge Bay, Nunavut, Canada). At both sites, the uppermost 10-cm ice horizon showed evidence of CO2 offgassing, while carbonate distributions in the middle and bottommost 10-cm horizons largely followed the salinity distribution. In the polynya, the upper-ice horizon had significantly higher bulk total inorganic carbon (TIC), total alkalinity (TA), and salinity, potentially due to freeze-up conditions that favoured frazil ice production. The middle-ice horizons were statistically indistinguishable between sites, suggesting that ice growth rate is not an important factor for the carbonate distribution under mid-winter conditions. The thicker (non-polynya) site experienced higher algal biomass, TIC, and TA in the bottom horizon. Carbonate chemistry in the bottom horizon could be explained by the salinity distribution, with the strong currents at the polynya site potentially playing a role in desalinisation; biology did not have a noticeable impact. We did see evidence of calcium carbonate precipitation, but with little impact on the TIC : TA ratio, and little difference between sites. Because differences were constrained to relatively thin layers at the top and bottom, vertically averaged values of TIC, TA, and especially the TIC : TA ratio were not meaningfully different between sites. This provides some justification for using a single bulk value for each parameter when modeling sea ice effects on ocean chemistry at coarse resolution. Exactly what value to use (particularly for the TIC : TA ratio) likely varies by region but could potentially be approximated from knowledge of the source seawater and sea ice salinity. Further insights await a rigorous intercomparison of existing data.