Gruta do Caldeirão (39° 30’ 11” N; 8° 24’ 32” W) is an archaeological site located about 140 km northeast of Lisbon (Portugal) at an altitude of 123 m a.s.l., 8 km north of the city of Tomar (Fig. 1A). The cave opens in calcareous dolomites of the Lower Jurassic, and its entrance, which faces south, is located in the northern slope of a valley in the right bank of River Nabão. The excavations carried out by one of us (J. Zilhão) between 1979 and 1988 (Zilhão, 1992, 1997) showed a 6.2 m stratigraphic sequence divided into 15 layers, containing Late Pleistocene (Middle and Upper Paleolithic) and Early Holocene (Neolithic) archaeological material (Fig. 1B). A zooarchaeological study of the large vertebrates showed that, during the Middle Paleolithic and the early Upper Paleolithic, the cave functioned in part as a large carnivore den; hyenas were the main bone accumulator, with contributions from leopards and the bearded vulture. In the later Upper Paleolithic, the bones were mainly accumulated by humans (Davis, 2002; Davis et al., 2007). Human remains have been identified in the Solutrean and Magdalenian layers (Trinkaus et al., 2001). Regarding the climatic reconstruction, and in agreement with the radiocarbon dates (ca. 22.5 -25.5 ka cal BP; Zilhão, 1997), the magnetic susceptibility (MS) of the Gruta do Caldeirão sedimentary sequence is lowest in Solutrean layers H to Fc, indicating a cold period related to the Last Glacial Maximum (LGM) (Ellwood et al., 1998), which includes, according to Rasmussen et al. (2014), Greenland Stadials (GS) GS 2.1, GS 2.2 and GS 3.

A preliminary study of the rodent assemblage (Póvoas et al., 1992; Brunet-Lecomte & Póvoas, 1993) identified 11 species (Table 1): Apodemus sylvaticus, Allocricetus bursae, Eliomys quercinus, Microtus arvalis, M. agrestis, M. (Iberomys) brecciensis, M. (Iberomys) cabrerae, M. (Terricola) lusitanicus, M. (Terricola) duodecimcostatus, Chionomys nivalis and Arvicola sapidus. Environmentally, the succession was interpreted as a landscape dominated by open and dry biotopes with forested areas from layers K to Fc (indicated by the presence of M. arvalis and A. bursae), turning into a more humid and forested biotope in layers Fb and Fa (represented by a high percentage of A. sylvaticus and M. (Terricola) spp.), and ending in a drier environment in layer Eb (shown by the major presence of M. arvalis, the low percentage of A. sylvaticus, and the high proportion of M. (T.) duodecimcostatus in relation to M. (T.) lusitanicus).

Against this background, our present objectives are threefold. Our first aim is to revise the Microtus (Iberomys) material, because current data suggest that the last occurrence of the extinct species M. (Iberomys) brecciensis in Iberia was at the end of Marine Isotope Stage (MIS) 6 or the beginning of MIS 5, e.g. at Maltravieso-Sala de los Huesos in Extremadura, dated to between 183–117 ka (Hanquet, 2011), or in layer I of Cova del Rinoceront in Barcelona, dated to ca. 84 ka (López-García et al., 2016). Our second aim is to reinterpret the assemblage, focusing on extinct taxa such as Allocricetus bursae and on species not currently found in the area, such as Microtus arvalis, Microtus agrestis and Chionomys nivalis, using the current distribution and habitat preference of these species (e.g. Paupério et al., 2017) as well as their first and last appearance data in Iberia (in the case of A. bursae). Our third aim is to apply the bioclimatic model (in accordance with Hernández-Fernández, 2001a, 2001b) in order to infer various climatic parameters and compare the Gruta do Caldeirão rodent assemblage with the climatic signals obtained by studying the magnetic susceptibility of the sequence (Ellwood et al., 1998), the current climatic parameters for the surrounding area, and the general dynamics of sea surface temperatures (SST) and pollen data from sea cores in the western Iberian margin (Naughton et al., 2007; Salgueiro et al., 2014; Turon et al., 2003).

Family Cricetidae Fischer, 1817

Subfamily Arvicolinae Gray, 1821

Genus Microtus Schrank, 1798

Subgenus Iberomys Chaline, 1972

Microtus (Iberomys) cabrerae Thomas, 1906

(Fig. 3: 1–15)

Material: nine left lower m1 (CAL/P11/F11oeste/Fb, CAL/P12/F13/Fc, CAL/P12/H5/I, CAL/P12/H2/I, CAL/P12/I4/I, CAL/P12/K1/K, CAL/P13/F11/Fc, CAL/P14/J5/Jb and CAL/N13/E2(2)/Eb), and 14 right lower m1 (CAL/P11/H3/Fc, CAL/P11/I8/I, CAL/P11/K2/K, CAL/P12/F15/Fc(1), CAL/P12/F15/Fc(2), CAL/P12/K1/K, CAL/P12/J5/Ja(1), CAL/P12/J5/Ja(2), CAL/P13/F8/Fc, CAL/P13/F10/Fc, CAL/O15/E2/Eb, CAL/L15/E1(1)/Eb, CAL/L15/E1(2)/Eb and CAL/N13/E2(2)/Eb).

Description: In terms of the description by Chaline (1972), modified by Ayarzagüena & López-Martínez (1976) and Cuenca-Bescós et al. (2014), the first lower molars (m1) recovered from Gruta do Caldeirão are characterized by clear labio-lingual asymmetry, more pronounced than in other microtines; on the labial side of the m1 there are only three re-entrants filled with cement; in some specimens there is a fourth, greatly reduced re-entrant (BRA4), allowing these specimens to be distinguished from other species of the genus Microtus and subgenus Terricola. Moreover, except for three juveniles (CAL/P11/K2/K, CAL/P12/F15/Fc(2) and CAL/P12/K1/K), all the identified teeth are large and feature the following: very marked labio-lingual asymmetry (mainly observed between triangles T4 and T5); a labial re-entrant angle 5 (LRA5); a scarcely to very pronounced angle between triangle T7 and the anterior cusp (AC); and a non-visible to well-developed double angle in triangle T6. The specimens from Caldeirão differs from M. (I.) brecciensis because in this fossil species the m1 are smaller and less asymmetrical than in M. (I.) cabrerae, and in general the LRA5 and the double angle of triangle T6 are non-existent or variably developed. All these features lead us to ascribe our material morphologically to the species M. (I.) cabrerae. This conclusion can be metrically corroborated by drawing a comparison between 17 measurable m1 (Table 4; excluding the juvenile or fragmented teeth CAL/P12/J5/Ja(2) and CAL/L15/E1(2)/Eb) and the extinct M. (I.) brecciensis from the Middle Pleistocene sites of Gruta da Aroeira (Torres Novas, Portugal) (López-García et al., 2018), Galeria, TD10 and TE 18–19 (all three from Sierra de Atapuerca, Burgos, Spain) (Cuenca-Bescós et al. 1999; López-García et al. 2008; 2011c; 2015) and the M. (I.) cabrerae from the Late Pleistocene sites of Gruta da Oliveira (Torres Novas, Portugal) (unpublished material), A. Romaní (Capellades, Barcelona, Spain), C. Gegant (Sitges, Barcelona, Spain) and Gorham’s cave (Gibraltar, UK) (López-García et al. 2008; 2011c; 2015) (Table 5; 6; Fig. 4).

Remarks: M. (I.) cabrerae (Cabrera’s vole) is currently endemic to the Iberian Peninsula, where it is widely distributed, with well-documented populations in the foothills of the Pyrenees, the southern Iberian System, the Baetic Sierras and the Central System, also extending the length of Portugal from SW to NE in a limited and patchy manner (Palomo et al., 2007; Paupério et al., 2017) (Fig. 5). It exclusively inhabits areas with a Mediterranean climate, a high water table, and all-year-round herbaceous cover (Pita et al., 2017). At present, the first appearance datum of M. (I.) cabrerae in the Iberian Peninsula occurs in the Marine Isotope Stage 5 (MIS 5) sites of Cueva de las Pinturas (Sesé & Ruiz-Bustós, 1992), Cova Bolomor (Guillem-Calatayud. 1995, 2001), Cueva del Camino (Laplana et al., 2013), Preresa (Sesé et al. 2011), Figueira Brava (Jeannet, 2000; Zilhão et al., 2020), and Gruta da Oliveira (unpublished material) (Fig. 5). During MIS 3 and MIS 2 M. (I.) cabrerae is represented outside its current range (Fig. 5), e.g. at Gorham’s cave (López-García et al., 2011a), Boquete de Zafarraya (Barroso Ruiz et al., 2006), El Portalón (López-García et al. 2010a), Cueva de la Zarzamora (Sala et al. 2011), Cueva de los Moros de Gabasa (Gil & Lanchares, 1988), Cueva de Aguilón-P7 (Cuenca-Bescós et al., 2010a), Cova dels Xaragalls (López-García et al., 2012a), Cova de Teixoneres (López-García et al., 2012b), Cova del Toll (Fernández-García & López-García, 2013), Abric Romaní (Fernández-García et al., 2018), Cova del Gegant (López-García et al., 2012b), Cova de Valdavara-1 (López-García et al., 2011b), El Salt (Fagoaga et al., 2018) and Cova Colomera (López-García et al., 2010b)

According to Póvoas et al. (1992), the long-tailed field mouse (Apodemus sylvaticus) represents more than 30% of the individuals in all the layers of the Gruta do Caldeirão sequence. The relative abundance of A. sylvaticus indicates that the landscape surrounding the cave featured good shrub cover and forest margins (Paupério et al., 2017). Moreover, the pine vole species M. (T.) duodecimcostatus and M. (T.) lusitanicus represent more than 30% of the individuals in all the layers, except K (13.8 %) and Jb (29.4 %). The relative abundance of these species indicates open landscapes and humid environmental conditions (Paupério et al., 2017). Worthy of note is also the presence of the extinct hamster Allocricetus bursae and of three vole taxa currently absent in the area: the European snow vole (Chionomys nivalis), the common vole (Microtus arvalis), and the field vole (Microtus agrestis).

Allocricetus bursae has been found only in layer K (Póvoas et al., 1992; initially assigned to the Mousterian, this unit has since been recognized as belonging in fact to an undiagnostic early Upper Paleolithic). The biotope preferences of this extinct hamster can be inferred from the present-day species that is phylogenetically closest to it, Cricetulus migratorius (the grey hamster). The latter’s present range extends from eastern Europe through Russia and central Asia to Mongolia and western China, where it inhabits dry grasslands, steppes and semi-deserts; arid areas with relatively sparse vegetation are preferred, and forests and damp habitats avoided (Kryštufek et al., 2016). The fossil record of A. bursae in the Iberian Peninsula goes back to the early Middle Pleistocene of Gran Dolina (Cuenca-Bescós et al. 2010b), with an age around 600 ka, and it is relatively abundant during the Middle Pleistocene and Late Pleistocene in central-south and eastern Iberia. At present, the most recent occurrence of the species is in the Late Pleistocene site of Cueva Ambrosio, with an age between 17.9–16.5 ka BP (Sesé & Soto, 1988).

Chionomys nivalis is found in layers Jb (early Upper Paleolithic), Fb (Solutrean) and Eb (Magdalenian) (Póvoas et al., 1992). It is a species mainly linked to the presence of stony soils with open meadows and herbaceous vegetation in mountainous regions above 1000 m (Paupério et al., 2017). In Portugal, it is currently only found in the northeastern Serra de Montesinho, at altitudes above 1340 m (Paupério et al., 2017). C. nivalis appears in the Iberian Peninsula during the Late Pleistocene (Sesé, 1994; Sesé & Sevilla. 1996), where it is well represented everywhere but the Levant (López-García, 2011).

Microtus arvalis and Microtus agrestis are represented in all layers (Póvoas et al., 1992). Both taxa inhabit grassland, but M. arvalis prefers open dry terrain with discontinuous herbaceous cover and M. agrestis prefers damp areas such as marshes, peat-bogs and river banks (Paupério et al., 2017). In Portugal, M. arvalis is currently only present in the extreme northeast, and M. agrestis lives exclusively in the north and north-central area (Paupério et al., 2017). Both species are also identified (Moreno-García & Pimenta, 2002) in Portugal outside its nowadays distribution in layers TP06 and TP09 of the Lagar Velho rockshelter (Lapedo valley, Leiria) with an age between 24–27 ka cal BP (Zilhão & Almeida, 2002). The first occurrence of these species in the Iberian Peninsula is in the Middle Pleistocene sites of Sierra de Atapuerca (ca. 400 ka) (Luzi & López-García, 2019; Luzi, 2018); both are well represented all over Iberia throughout the Late Pleistocene (López-García, 2011).

By comparison with current data (Table 7), the bioclimatic model characterizes the climate of the area around Gruta do Caldeirão as colder (ΔMAT = -3.3 °C to -1.6 °C) and relatively drier (ΔMAP = -95 mm to -49 mm) with the exception of layer Fc, where the precipitation would have been higher than nowadays (ΔMAP_Fc = +85 mm). Summers were similar to the present (ΔMTW = -0.9 °C to +0.2 °C), but winters were colder (ΔMTC = -5.2 °C and -3.0 °C). The study of the magnetic susceptibility of the sequence (Ellwood et al. 1998) also concluded that layers H to Fb corresponded to the coldest period of the sequence, in agreement with their radiocarbon dating to the Last Glacial Maximum (LGM).

The magnetic susceptibility and the MAT and MAP data derived from the rodent assemblage thus concur that the LGM in the area was characterized by relatively low temperatures and high precipitation (Fig. 6). This inference is consistent with the presence of species that according to López-García et al. (2010b) have mid-European requirements (e.g. the C. nivalis found in level Fb), as well as with the absence in layer Fc of species that have strict Mediterranean requirements (i.e. M. (I). cabrerae); it is also consistent with the suggestion made by Póvoas et al. (1992) to the effect that more humid conditions prevailed during the deposition of layers Fa and Fb.

These data coincide with the SST data from the western Iberian margin, which show differences between 1.5 °C and 4 °C for the LGM in relation to present-day temperatures (Salgueiro et al., 2014). Also, the LGM was characterized in western Iberia by a predominantly herbaceous environment, with Pinus starting to expand and an almost continuous presence of deciduous tree pollen (Naughton et al., 2007). In addition, the slight expansion of ericaceous communities detected in western Iberia suggests an increase in humidity near the continent at that time (Turon et al., 2003).

All lines of evidence therefore concur in contrasting the paleoclimatic conditions prevalent throughout the accumulation of the Pleistocene Gruta do Caldeirão sequence with those at present. Indeed, the area around Tomar nowadays falls within the temperate Mediterranean zone with warm summers (CSa) of the Köppen-Geiger classification (Beck et al., 2018).

Our revision of the rodent assemblage from Gruta do Caldeirão thus leads us to draw the following conclusions: