The reconstruction of past environments by means of macrofossil and pollen analysis is commonly based on the modern ecological preferences of the taxa that may have produced these fossils. Here we present a modelling approach, in which we use modern vegetation–surface height relationships to quantify past surface heights in an Arctic ice-wedge polygon mire. Vegetation composition and ground surface height (GSH) were assessed in a polygon mire near Kytalyk (Northeastern Siberia). Cluster analysis revealed five plant communities, which are clearly separated with respect to ground surface height, frost surface height and coverages of open water and vegetation. Based on the composition of modern vegetation we constructed two sets of potential fossil types (plant macrofossils and pollen), an extensive one and a more restricted one to reflect different conditions of preservation and recognisability. We applied Canonical Correspondence Analysis to model the relationships between potential fossil types and measured GSH. Both models show a strong relationship between modelled and measured GSH values and a high accuracy in prediction. Finally, we used the models to predict GSH values for Holocene peat samples and found a fair correspondence with expert-based multi-proxy reconstruction of wetness conditions, even though only a minor part of the encountered fossils were represented in the GSH models, illustrating the robustness of the approach. Our approach can be used to reconstruct palaeoenvironmental conditions in a more objective way and can serve as a template for further palaeoecological studies.