Testate amoeba transfer function performance along localised hydrological gradients

Testate amoeba transfer functions are widely used for reconstruction of palaeo-hydrological regime in peatlands. However, the limitations of this approach have become apparent with increasing attention to validation and assessing sources of uncertainty. This paper investigates effects of peatland type and sampling depth on the performance of a transfer function using an independent test-set from four Sphagnum-dominated sites in European Russia (Penza Region). We focus on transfer function performance along localised hydrological gradients, which is a useful analogue for predictive ability through time. The performance of the transfer function with the independent test-set was generally weaker than for the leave-one-out or bootstrap cross-validations. However, the transfer function was robust for the reconstruction of relative changes in water-table depth, provided the presence of good modern analogues and overlap in water-table depth ranges. When applied to subsurface samples, the performance of the transfer function was reduced due to selective decomposition, the presence of deep-dwelling taxa or vertical transfer of shells. Our results stress the importance of thorough testing of transfer functions, and highlight the role of taphonomic processes in determining results. Further studies of stratification, taxonomy and taphonomy of testate amoebae will be needed to improve the robustness of transfer function output.