The alpine newt, Ichthyosaura alpestris, is very sensitive to habitat destruction and alteration which has led to declining populations across Europe. As this species is protected through the Bern Convention, it is essential to have a comprehensive understanding of its habitat requirements to ensure proper conservation measures. We trained, validated and optimized classification tree models based on data on local aquatic habitat conditions from 125 farmland ponds scattered over Belgium and Luxembourg where the alpine newt commonly occurs. To obtain user-friendly and representative models, data was preprocessed and stratified after which different degrees of pruning were applied for model optimization. In order to check the model's applicability for management, we predicted alpine newt occurrence with an independent dataset. The most robust and reliable model revealed that fish absence was the major driving factor followed by the thickness of the sludge layer. We found that fish presence established alpine newt absence and that fishless ponds with a sludge layer of 15 cm or more were predicted to host no alpine newts. The latter provides quantitative information for decision makers. Moreover, our results indicated that the amount of sludge could be associated with eutrophication and erosion. Regarding management practices, it is advised to assure the absence of fish and reduce sludge accumulation in ponds designated for the conservation of alpine newts, for example by designing temporary ponds not fed through fish-containing surface waters. Furthermore, we recommend to install buffer strips around a pond to reduce nutrient and soil run-off from the terrestrial environment.
