Abstract
Aim The emergence of Ecological Niche Modelling (ENM) in the 2000s advanced our ability to evaluate how biotic and abiotic factors constrain species distribution. Although highly employed in ecological investigations, the application of ENM in palaeontology (‘palaeoENM’) is less common, and the scarcity of the fossil record makes predictions uncertain with current ENM routines. Here we propose a methodological assembly especially designed for small datasets to enhance palaeoENM studies. Location Global. Time Period Permian/Triassic boundary (ca. 252 Ma). Major Taxa Studies Two conodont (Vertebrata, Conodonta † ) genera, Clarkina and Hindeodus . Methods We use the Non‐Parametric Probabilistic Ecological Niche (NPPEN) model improved with the generation of pseudo‐presences (PPs) to complete the dataset before applying the ENM, called NPPEN‐PP hereafter (i.e., NPPEN with PPs). NPPEN‐PP is performed using environmental parameters simulated with an Earth System Model constrained by proxy data. Then, we use Shapley values to investigate the individual contribution of each environmental variable in space and time in the model. Finally, we propose a simple procedure to quantify and compare the robustness of the simulations among studied taxa. Results The generation of PPs increases the performance of NPPEN but their robustness differs between the two taxa, probably due to a difference in sampling effort. The resulting palaeoENMs indicate that the two conodont taxa occupied distinct, albeit partially overlapping, ecological niches and that their distributions were influenced by different environmental variables. In particular, Hindeodus species remained in low latitudes and expanded their distribution to high latitude while Clarkina species moved from low to high latitude during the mass extinction event. The Shapley values allow us to explain this discrepancy by a higher adaptability of Hindeodus to anoxic environments compared to Clarkina . Main Conclusions We assess for the first time, through modelling, the ecological differentiation of two conodont genera. The methodological assembly proposed here can help palaeontologists but also ecologists to perform reliable ENM on small datasets and so extend our knowledge on drivers of past and present taxa distribution.