Late quaternary small mammals and paleotemperatures in Bulgaria and Italy

The aim of the study was to analyze the capacity of fossil remains of small mammals in the cave sediments for quantitative climate reconstructions. We have developed a number of small mammal-climate inference models using modern owl pellet samples from 144 sites. Canonical correspondence analysis was used to reveal the factors (climate parameters, agent of accumulation) that best reflect the main patterns of variation in the modern small mammal assemblages. Annual Mean Temperature, Maximal Temperature of Warmest Month, and Minimal Temperature of Coldest Month capture a large and statistically significant fraction of the overall variance in the recent small mammal data. Four inference models were tested. The results of cross-validation showed that Modern Analog Technique and Weighted Averaging Partial Least Squares (WAPLS) were the best in terms of performance statistics. Using these models it was possible to produce quantitative records of past temperatures based on small mammal stratigraphic cave sequences from Bulgaria (Cave 16 and Cave 15) and Italy (Castelcivita, Mura, and Continenza). For each locality both methods revealed similar trends in climate change, but differ somewhat in terms of specific values of the reconstructed temperatures. The differences between them were larger in non-analog situations. In this sense, WAPLS was considered more reliable in the reconstruction of climate features during Late Pleistocene. Reconstructions made on the basis of this method have important implications for understanding paleoenvironmental changes during Late Quaternary. The reconstructions support the previous qualitative interpretations and reconstructions based on other proxies.