The large mammal analysis is based on a total of 4904 specimens: 838 from S.U. 13, 120 from S.U. 103 and 3946 from S.U. 14 (Table 1). Forty percent of the specimens was too fragmented for identification. Species determinations were made using the comparative collection from the Large Mammals Laboratory at the Department of Human Studies, University of Ferrara. For each taxon, we calculated the Number of Identified Specimens (NISP) and the Minimum Number of Individuals (MNI). Because the remains of Ursus spelaeus had a different mode of accumulation (see par.3.1), this species is considered separately.

The small mammal remains consist of disarticulated bone fragments collected by water-screening using sieves of 0.5 mm mesh during the excavation campaigns from 2009 to 2013. Part of the material has been previously published in Arnaud et al. (2014) and Arzarello et al. (2012).

The small mammal assemblage comprises a total of 1187 remains, corresponding to a minimum number of 373 individuals (Table 2). The specific attribution of this material was mainly based on the best diagnostic elements: maxilla and isolated teeth for rodents, mandible and maxilla for shrews, mandible, maxilla, isolated teeth and humeri for bats.

The taxonomic classification follows Wilson and Reeder (2005), except Clethrionomys glareolus (for the priority over Myodes, see Tesakov et al., 2010). Data on the distribution and habitat of the species were taken from Amori et al. (2008), Boitani et al. (2003) and Mitchell-Jones et al. (1999).

We calculated the palaeodiversity using the Simpson index of Evenness = 1-∑(p_i ²), where p_i is the proportion of individuals in the i^th species (Harper, 2005 and Magurran, 2004). The evenness index is constrained between 0 and 1. The index has been calculated using PAST 3.04 avoiding redundant determinations (i.e., for Arvicola amphibius, the individuals determined as Arvicola cf. amphibius and Arvicola sp. were not included in the Simpson index calculation) (Hammer et al., 2001).

The SDQ index (Heinrich, 1978) quantifies the difference in enamel thickness between the anterior and the posterior wall of each triangle of arvicolid molars. This index was calculated for m1 of Arvicola amphibius of Ciota Ciara. SDQ was measured for the three main triangles T1, T2 and T3 of first lower molars (SDQ3, according to Kosciow and Nadachowski, 2002, Lippi et al., 1998 and Markova, 2005) in order to include the maximum number of specimens. Higher values for these indices generally correspond to early evolutionary stage of Arvicola. However, remarkable differences exist between modern populations, and can be correlated to altitude (Kratochvil, 1981) and/or latitude (Maul et al., 1998 and Röttger, 1987). Furthermore, critics have been moved to use SDQ as a strong tool for biochronological reconstruction (Escudé et al., 2008). Although this coefficient cannot provide conclusive evidence of the relative age of a fauna, combined with other data – such as the whole composition of the mammal assemblage – SDQ can give additional information about the chronological position of a site (Maul et al., 2000).

In order to reconstruct the palaeoenvironment at Ciota Ciara cave, we used the method of habitat weighting (Andrews, 2006 and Evans et al., 1981), assigning each small mammal taxon to the habitat(s) where it can be found at present in Europe. For this purpose, habitats were divided into six types (Cuenca-Bescós et al., 2009, López-García et al., 2014 and López-García et al., 2010): open land with either dry and wet meadows (OD and OH, respectively); woodland environments, divided into open woodland, woodland margins and forest patches (OW) and woodland and mature forest habitat (W); water, areas along streams, lakes and ponds (Wa); and habitats with a suitable rocky or stony substratum (R) (Table 3).

In order to infer the palaeoclimatic data from Ciota Ciara cave assemblage, we applied the Mutual Climatic Range method (Elias, 1997 and Pross, 2000 among others, see also Blain et al., 2009 and López-García et al., 2014 and López-García et al., 2010). This method assesses potential palaeoclimatic conditions by evaluating the current distribution of all taxa occurring in each level.

Two climatic factors were calculated: the mean annual temperature (MAT) and the mean annual precipitation (MAP), using climatic maps of Italy (Attorre et al., 2008) and data provided by the network of Italian meteorological research stations over a period of 30 years. Then, we could compare them with the modern climatic data from the weather station of Borgosesia (359 m a.s.l.). For the area surrounding Ciota Ciara cave, the current data show an MAT = 10.7 °C and MAP = 1009 mm. Because of the scarce interpolations (only two), we also used the linear regression method (Hernández Fernández, 2004, Hernández Fernández, 2006, Hernández Fernández et al., 2007 and Hernández Fernández and Peláez-Campomanes, 2005).

The Italian Peninsula is characterized by different climates, due to its peculiar position in the centre of the Mediterranean Sea and its geography. In Italy, two macro-bioclimates are present: the temperate bioclimate, which is found in the north-eastern Peninsula up to the Marche region, the Po Valley and part of the Appenines as opposed to the Mediterranean bioclimate found in the Tyrrhenian coast up to the Ligurian region and the southern Adriatic coast. These macro-bioclimates are subdivided into nine different climates, and climatic conditions may change abruptly over a few tens of kilometres, from the mildness of the seashore to the harshness of coastal mountain summits (Stoch, 2009).

The faunal assemblage is composed of Ursus spelaeus, Ursus arctos, Canis lupus, Vulpes vulpes, Meles meles, Martes martes, Lynx lynx, Panthera leo, Panthera pardus, Rupicapra rupicapra, Cervus elaphus, Bos sp., Bos vel Bison, Sus scrofa, Stephanorhinus sp. and Marmota marmota. A right upper incisor previously determined as Hystrix sp. from S.U. 13 (Arnaud et al., 2014, Arzarello et al., 2012, Arzarello et al., 2013 and Daffara et al., 2014) has been revised and assigned to Marmota marmota (Fig. 2, no. 5).

Ungulates appear to be underrepresented in S.U. 13 and 103 (3,8% of identifiable remains) but in S.U. 14 there is a clear increase (13,4% of determinable remains).

Ursus is the best represented carnivore in the Ciota Ciara cave, and Ursus spelaeus is the most copious species in all the stratigraphic units (S.U. 103: 27.50%; S.U. 13: 40,77%; S.U. 14: 59.04%). Ursus arctos is progressively rare from the top to the bottom of the sequence: (S.U. 103: 8.33%; S.U. 13: 4.22%; S.U. 14: 1.01%). Ursus spelaeus used caves during dormancy and as a nest for the cubs (Pacher and Stuart, 2009), which probably accounts for this species being overrepresented in the assemblage.

The other taxa are represented by only a few remains, but, especially in S.U. 14, the biodiversity is high. The ungulates are more abundant in the lower part of the stratigraphic sequence: in S.U. 103 and 13, the only herbivores present are Rupicapra rupicapra, Cervus elaphus and a bovid, while in S.U. 14 Rupicapra rupicapra, Cervus elaphus, cf. Dama, Bos primigenius, Bos vel Bison, Stephanorhinus sp., and Sus scrofa are present. Fedele (1968) also reported the occurrence of Capra ibex and Castor fiber but they have not been recovered in modern excavations. In a breach of the site of Ara, near Ciota Ciara cave, a mandible of Stephanorhinus kirchbergensis has been found (Giacobini and Strobino, 1978), attesting to the presence of this species in the area of Monte Fenera.

Considering the MNI in S.U. 14, the most abundant taxon is Ursus spelaeus (18 specimens), followed by Cervus elaphus (five specimens) and Marmota marmota (three specimens). All the other taxa are represented by only one or two specimens. The deciduous teeth of bear are abundant (71 specimens); unfortunately, cubs of U. spelaeus are not distinguishable from U. arctos. Considering the different ecology between U. spelaeus and U. arctos, the deciduous teeth probably could be referred to U. spelaeus, U. arctos preferring small dens for dormancy, while U. spelaeus commonly used large caves (Kurtén, 1976).

The abundance of carnivores in the cave could be the result of selective use of the cave by these taxa, but error sampling due to the limited number of the remains or anthropic selection cannot be excluded until more accurate taphonomic analysis are available.

Ciota Ciara cave is unique in the Italian Peninsula context because of the relatively high abundance of Pliomys coronensis, and the high biodiversity of bats (Fig. 3). The Simpson index of Evenness indicates that in both Units, the biodiversity is high and relatively equally distributed (S.U. 14: 0.88 and S.U. 13: 0.86) (Fig. 4B).

The presence of one sample identified as Erinaceus sp. is remarkable. It is a right P4 and, based on the samples present in our collection, falls more into the size range of present Erinaceus roumanicus than that of E. europaeus. Unfortunately, the lack of specific works on the dental differences (especially on P4 measurements and morphology) between these two species does not allow a specific determination.

In Fedele (1966), only few remains of Sorex gr. araneus, Chionomys nivalis, Microtus arvalis, Arvicola sp. and Glis glis have been signalled in a trench (Trench II) excavated in the interior of the cave, approximately 20 m far from the 2009-2014 investigated area.

In S.U. 103, only two specimens of Pliomys coronensis have been found. S.U. 13 and 14 are both dominated by Clethrionomys glareolus which, together with Apodemus (Sylvaemus), Glis glis, Eliomys quercinus and the bats, contributes to the forest component (W). Thus, the environment surrounding the cave was dominated by woodland, woodland margins and forest patches (Fig. 4). These considerations are confirmed by the presence of typical woodland small mammals such as Glis glis and Sciurus vulgaris, the latter rather rare in Late Pleistocene sequences because of its diurnal habits (Berto, 2012). Also, the chiropters found in Ciota Ciara cave confirm the presence of these kind of habitats. Species as Rinolophus hipposideros, R. ferrumequinum, Barbastella barbastellus and Plecotus sp. normally hunt in wooded areas, while Myotis myotis prefers grassland environments. Although the chiropters have different habits than the other small mammals, these species are considered sedentary, their areal varies from 15 to 30 km from the nest, even though sporadic migration up to 50 km are registered.

Nevertheless, a climatic change is visible along these two Stratigraphic Units. The environment was more open during the deposition of S.U. 14 and the cave was surrounded by an open woodland environment with exposed rocks. In this Unit percentages of Microtus arvalis and M. (Terricola) gr. multiplex-subterraneus are high (16,11% and 13,37% respectively) and cold climate indicators, such as Cricetus cricetus, Microtus cf. gregalis and Chionomys nivalis, are present, although rare. The change to a more temperate climate is confirmed also by temperatures (MAT) and precipitations (MAP) (Fig. 5), even though this result must be considered with prudence, given the few interpolations found with the Mutual Climatic Range. It is possible to observe an improvement in climatic conditions, but the temperature for S.U. 14 indicates a “pleniglacial” condition that does not entirely reflect the situation visible in the faunal assemblage. Clethrionomys glareolus always stands as the most represented arvicolid. This is a strong indicator of a forested environment and the species can be underrepresented in the thanatocoenoses because it lives in bushes and it is considered difficult to pray. In addition, bats related to wooden environments are constantly present. The linear regression method tested on our sample shows that The MAT oscillation is less severe (S.U. 14, MAT = 9.4 °C; S.U. 13, MAT = 11 °C) while the MAP is much higher (S.U. 14, MAP = 1756 mm; S.U. 13, MAP = 1933 mm) but it must be noted that the standard deviation of this method is high (3.367 °C and 470.615 mm respectively, data inferred from Hernández Fernández, 2004).