A long-term perspective is essential for understanding environmental change. To be able to access the past, environmental archives such as marine and lake sediments that store information in the form of diverse proxy records are used. Whilst many analytical techniques exist to extract the information stored in these proxy records, the critical assessment and refinement of current methods in addition to developing new methods is crucial to improving our understanding. This study aims to improve our knowledge on diatom species used for reconstructing ocean surface conditions, especially temperature and sea ice variability over time. We define the distribution and the relationship to sea surface temperature (SST) and sea ice concentrations (SIC) of the species Fragilariopsis oceanica, Fragilariopsis reginae-jahniae and Fossula arctica using diatom training sets from the northern North Atlantic. We further assess the effect of separating these species compared to grouping them under F. oceanica, as has been done in the past. Our results suggest that while these three species share similarities such as the preference for stratified waters induced by sea ice or glacier meltwater, they also exhibit heterogeneous distributions across the northern North Atlantic, with individual optima for SST and SIC. This also affects quantitative reconstructions based on our data, resulting in lower SST and higher SIC estimates when the species are separated in the surface sediment and down-core diatom assemblages.