Small mammal distributional patterns in Northwestern Argentina

Quantitative evaluations of species distributional congruence allow evaluating previously proposed biogeographic regionalization and even identify undetected areas of endemism. The geographic scenery of Northwestern Argentina offers ideal conditions for the study of distributional patterns of species since the boundaries of a diverse group of biomes converge in a relatively small region, which also includes a diverse fauna of mammals. In this paper we applied a grid-based explicit method in order to recognize Patterns of Distributional Congruence (PDCs) and Areas of Endemism (AEs), and the species (native but non-endemic and endemic, respectively) that determine them. Also, we relate these distributional patterns to traditional biogeographic divisions of the study region and with a very recent phytogeographic study and we reconsider what previously rejected as 'spurious' areas. Finally, we assessed the generality of the patterns found. The analysis resulted in 165 consensus areas, characterized by seven species of marsupials, 28 species of bats, and 63 species of rodents, which represents a large percentage of the total species (10, 41, and 73, respectively). Twenty-five percent of the species that characterize consensus areas are endemic to the study region and define six AEs in strict sense while 12 PDCs are mainly defined by widely distributed species. While detailed quantitative analyses of plant species distribution data made by other authors does not result in units that correspond to Cabrera's phytogeographic divisions at this spatial scale, analyses of animal species distribution data does. We were able to identify previously unknown meaningful faunal patterns and more accurately define those already identified. We identify PDCs and AEs that conform Eastern Andean Slopes Patterns, Western High Andes Patterns, and Merged Eastern and Western Andean Slopes Patterns, some of which are re-interpreted at the light of known patterns of the endemic vascular flora. Endemism do not declines towards the south, but do declines towards the west of the study region. Peaks of endemism are found in the eastern Andean slopes in Jujuy and Tucuman/Catamarca, and in the western Andean biomes in Tucuman/Catamarca. The principal habitat types for endemic small mammal species are the eastern humid Andean slopes. Notwithstanding, arid/semi-arid biomes and humid landscapes are represented by the same number of AEs. Rodent species define 15 of the 18 General Patterns, and only in one they have no participation at all. Clearly, at this spatial scale, non-flying mammals, particularly rodents, are biogeographically more valuable species than flying mammals (bat species).

records) of the species (Murguía & Llorente-Bousquets, 2003). Several methods for identifying AEs proposed in recent years are based on Platnick (1991)'s definition. However, most methods traditionally applied to recognize AEs are non-spatially explicit. Conversely, Szumik et al. (2002) and Szumik & Goloboff (2004) formalized a method that explicitly considers the spatial aspects of species distribution to identify AEs, evaluating the degree of overlap between the geographical distributions of species through the application of an optimality criterion (Szumik & Roig-Juñent 2005).
Northern Argentina is a transition zone between the tropics and subtropics: the Tropic of Capricorn passes through Jujuy, Salta, and Formosa provinces in the north. Due to its tropical-subtropical location, Northwestern Argentina (NWA) provides a good opportunity to investigate some of the factors that can determine the distribution of species. The boundaries of a diverse group of biomes converge in a relatively small region, which also includes a diverse fauna of mammals. Moreover, NWA includes the largest portion of the southern central Andes, whose endemic vascular plant species distribution has been recently studied . In this sense, the geographic scenery of NWA offers ideal conditions for the study of distributional patterns of animal species and for comparisons between floral and faunal patterns. Also, specifically dealing with small mammals, the large phylogenetic divergence between the three orders that inhabit NWA allows testing the generality of the patterns found (in the present and in previous studies, particularly those of Sandoval et al., 2010 andFerro, 2014) and their transcendence to a small group of certain forms of life (i.e. a particular genus, family, order).
Although some studies have characterized the major South American and Argentinean biomes, including those represented in NWA, as distinctive biogeographic provinces (Cabrera & Willink, 1973;Cabrera, 1976;Vervoorst, 1979; among others), these characterizations were traditionally essentially qualitative, and most were based almost exclusively on its floristic components. For example, regarding the study of Cabrera (1976), which is almost invariably used by researchers working in Argentina when they need to assign the study area or flora to a phytogeographic unit, it proposes a hierarchical partition of the regions. However, the partition criteria Cabrera effectively applied is not explicit, and, although the author argues that he categorizes successive levels of the phytogeographic hierarchy by endemism and taxa predominance, in fact, the criteria are not exclusively taxonomic and are not always hierarchically applied (Ribichich, 2002). To characterize phytogeographical areas, the author refers to the extension of the types of vegetation, and to "what taxa of different ranks are alternately dominant, conspicuous, unique, important, frequent, endemic, dominant, abundant, distinctive, prominent, present, common, constant, typical, represented, accompanying, associated or co-dominant in the represented area or ecological communities" (Ribichich, 2002), without presenting any formal definition of the meaning of such terms. In particular, Cabrera does not clarify what concept or what synonyms of endemic he used (Ribichich, 2002). For other than floristic taxa, with very few exceptions (e.g., Díaz Gómez, 2007;Lizarralde & Szumik, 2007;Aagesen et al., 2009;Navarro et al., 2009;Sandoval et al., 2010;Szumik et al., 2012;Sandoval & Ferro, 2014), these regions have been assumed as natural, and its congruence with the distribution of species was not tested. In fact, quantitative evaluations of species distributional congruence are very recent compared to narrative biogeographic regionalizations that have been made for decades by different naturalists. Quantitative approaches allow assessing previously proposed biogeographic regionalization and even identify previously undetected areas of endemism.
It has been previously assessed the classic phytogeographic regionalization of NWA by using small mammal distributional records (Sandoval et al., 2010;Sandoval & Ferro, 2014) and the authors have been able to relate the obtained patterns to the traditional phytogeographic units within the region and to conclude that obtained distribution patterns in general concur with those units. However, in both cases they miss the opportunity of: (1) Analyzing the results obtained from mammal distributional data with those obtained by Aagesen et al. (2012) in order to outline a more general biogeographical scheme for the region, i.e., an scheme that take into account both faunal and floristic elements.
(2) To deeply discuss what previous works discarded as 'spurious' areas (i.e., patterns that merge highly divergent units, e.g., merging eastern and western Andean slopes, or humid and arid habitats; Sandoval & Ferro, 2014), considering that some of these 'spurious' areas may even include a higher portion of endemic species than the areas of endemism (AEs) that they did recognize (Sandoval & Ferro, 2014) and that some of these areas may also survive changes in grid size (one of the identified criterions for areas support; Aagesen et al., 2009;Casagranda et al., 2009;Szumik et al., 2012;DaSilva et al., 2015). Since AEs may or may not include a single biogeographic unit (e.g., the endemic flora of the region do not; Aagesen et al., 2012), we want to reassess the issue but seriously analyzing the AEs that were previously dismissed as spurious (Sandoval & Ferro, 2014). And (3) To address all small mammal distributional records in an integrated and comparative analysis to evaluate the distributional patterns of non-flying (i.e. marsupials and rodents) versus the patterns of flying small mammals (i.e. bats) (assessment outlined in Sandoval et al., 2010). Bats and terrestrial small mammals (marsupials and rodents) exhibit disparate attributes, making them ideal for comparative purposes. Many bats have extensive ranges (Lyons & Willig, 1997). Actually, it has been mentioned that there is an apparently common pattern of low rates of endemism in several birds and bats that may be due to the high vagility of flying versus non-flying vertebrates (Myers & Wetzel, 1983). Contrary, most marsupials and rodents have relatively restricted ranges, probably due to their lack of vagility and poor dispersal abilities.
In this paper, we applied a grid-based explicit method to analyze distributional records of Northwestern Argentinean small mammal species in order to recognize patterns of species co-occurrence and the species that determine them. Also, we relate these distributional patterns to traditional biogeographic divisions of the study region and with a very recent phytogeographic study , and we reconsider what was previously rejected as 'spurious' areas (Sandoval & Ferro, 2014). Finally, we assessed the generality of the patterns found by evaluating the distributional patterns of non-flying versus the patterns of flying small mammals.

Study region.
The study region is located in the southern part of the central Andes. The tropical eastern slopes of the Andes are recurrently recognized as a continental and global hotspot of diversity and endemism for several taxa (e.g., Rahbek & Graves, 2001;Barthlott et al., 2005;Kier et al., 2005;Orme et al., 2005;Sklenář et al., 2011;Patterson et al., 2012). The study region is located where the high diversity and endemism of the eastern Andes slopes gradually fade southward and comprises the northwestern portion of Argentina (NWA), including the political provinces of Jujuy, Salta, Tucumán, Catamarca, and Santiago del Estero (Fig. 1), between parallels ~22°S and ~30°S, and meridians ~62°W and ~69°W (Fig. 1). The entire region comprises 470,184 km 2 , which represents approximately 17% of the total Argentinean territory (2,791,810 km 2 ).
Due its location between the tropical and subtropical latitudes and the complex relief, with the Andean orography on the western half of the region, which varies between 500 and 6000 m altitude, this region is characterized by high environmental heterogeneity, great climatic contrasts and a mosaic of different biomes. All these biomes come together and some of them reach their limits of distribution in the region (Cabrera & Willink, 1973;Cabrera, 1976;Burkart et al., 1999;Morrone, 2001a).
The biogeographic description outlined below follows Cabrera (1976), Burkart et al. (1999) and Morrone (2001a). The mountain ranges are located to the west and they are oriented north to south. The Sierras Subandinas, the Sierras Pampeanas, the Cordillera Oriental, and the Cordillera de los Andes are the substrate on which are developed the biomes of High Andes, Puna and Prepuna; while on its slopes and intermountain valleys are developed the xeric scrublands of the Monte, the montane rainforest of the Yungas, and the semiarid woodlands of the Chaco. The Sierras Subandinas and Pampeanas show an insular form and are surrounded by forests of Yungas and Chaco. The other mountain systems are relatively continuous, reach higher altitudes and separate the forest environments on the eastern slope of the open environments of Monte, Prepuna, Puna, and High Andes on the tops and the western slope. The eastern plains correspond to the Chaco biome.
Data set. The taxonomic treatment used herein basically follows that outlined in Wilson & Reeder (2005) and Barquez et al. (2006). We used distributional information of 124 small mammal species (10 species of Didelphimorphia, 41 of Chiroptera, and 73 species of Rodentia), i.e., those species whose adults range in weight from four grams to two kilograms (Wilson et al. 1996) and inhabits North-western Argentina (Appendix and Supplementary Material).
The main source of data was the specimens reviewed in all the major systematic collections of Argentina (see Acknowledgements). We also completed the data including published localities of occurrence for small marsupials and bats, and for species of the rodent families Chinchillidae, Caviidae, Octodontidae, and Abrocomidae (Díaz & Barquez, 2007;Díaz et al., 2000;Jayat et al., 2009;Jayat et al., 2013;Jayat & Ortiz, 2010;Mares et al., 1997), which have very few specimens based records and have no major problems with their alpha taxonomy. The complexity of the taxonomic history of most other species of Rodentia and the variety of approaches taken by different authors as to the limits of each taxa make very difficult to compare the entities mentioned in the literature with those considered in this paper.
Localities were taken from museum tags or from collector catalogues. All localities were checked and georeferenced using gazetteers, maps, or satellite images and field trips. We agree with Tabeni et al. (2004) in the use of georeferenced records to optimize the quality of the analysis, contrary to the use of distributional maps for the species, which tend to overestimate species diversity in the regions under study (but see Soberón et al., 2000).
Finally, we divided the species into two classes according to its distributional range, restricted to the study region and widely distributed. Distributional analyses. The method used to identify the AEs and Patterns of Distributional Congruence (PDCs) of small mammals in Northwestern Argentina was proposed by Szumik et al. (2002) and Szumik & Goloboff (2004). This method implements an optimality criterion that explicitly considers the spatial location of the species in the study region. The study region is divided into cells, and the sets of cells (=candidate areas; Aagesen et al., 2013) are evaluated through an index of endemicity. Individual values of endemicity between 0 and 1 are calculated for each species depending on the fit of the species geographical distribution to the candidate area being evaluated. Individual values are combined for all the species that contribute to a set to obtain the value of endemicity (score) of the area. The analysis results in a number of areas with maximum scores. This optimality criterion is implemented in the computer program NDM and its viewer VNDM (Goloboff, 2005; available at http://www.lillo.org.ar/phylogeny/). As Sandoval & Ferro (2014), we will consider basically two areas categories: areas of endemism (AEs, defined only or mostly by sympatric endemic species) and local patterns of distributional congruence (PDCs, defined only or mostly by sympatric species not endemic to the study region). It should be noted that the optimality criterion that was used is based on the classic concept of areas of endemism. Therefore, ideally, the taxa to be used in such an analysis of endemism should be those that are maximally endemic, i.e., those whose geographic ranges are smaller than the study region. Even so, the method can be extended to be applied in the search for patterns of distributional congruence determined both by species of restricted ranges (i.e., AEs) as by widely distributed species whose ranges exceed the study region (Aagesen et al., 2009), which will constitute local patterns of distributional congruence (PDCs) but not AEs in strict sense. While AEs in strict sense provide strong basis for biogeographic regionalization, PDCs provide first-step testable hypotheses of AEs for future analyses of neighbouring regions or analyses at more inclusive scales (Szumik et al., 2012). The distributional congruence of non-endemic species may be representing parts of larger AEs, local co-occurrence by particular combinations of ecological or climatic factors, or an artifact due to unevenness of sampling localities. It is beyond the purposes of this study to identify the causal factors of the distributional concordances identified. Whether the factors restricting the species ranges are a combination of present day ecological and physical phenomena or consequence of a history of vicariance and speciation, it is not a prerequisite for the detection of the pattern, which is, in turn, the first step in the elucidation of the process generating them (Sandoval & Ferro, 2014). As Aagesen et al. (2009), we use the optimality criterion to analyze species distribution in a regional context while relaxing the criterion of endemism. Accordingly, we will treat the scoring species as characterizing and not always as endemic species: the scoring species may be involved in the characterization of an area being or not being endemic or exclusive to that area.
We performed analyses with five different databases: first, one with a dataset that comprises all the small mammal species (SM database, 124 species); then, with a dataset that comprise terrestrial small mammals (orders Didelphimorphia and Rodentia; TSM, 83 species); and finally, three analyses, each comprising the species of Didelphimorphia, Chiroptera, and Rodentia separately (Di, 10 species, Ch, 41 species, and Ro, 73 species). Because there is no formal argument to use only one grid cell size (Aagesen et al., 2009;Casagranda et al., 2009), we analyzed the matrix of georeferenced data using grid cells of four different sizes (1, 0.75, 0.5, and 0.25° latitudelongitude). The use of several grid sizes increases the chance of finding different areas given the topographical complexity of the study region; moreover, using several grid sizes provides some kind of measure of support for a particular area of endemism: those areas which survive changes in grid size can be considered more strongly and clearly supported by the data (Aagesen et al., 2009;Casagranda et al., 2009;Szumik et al., 2012;DaSilva et al., 2015; although some very restricted areas can be strongly supported by the data, but would not be found by larger grids; Augusto Ferrari, com. pers.). Additionally, grid sizes used in this analysis were already used for the study region facilitating comparisons with previous studies (e.g., Aagesen et al., 2009;Díaz Gómez, 2007;Navarro et al., 2009;Sandoval et al., 2010;Sandoval & Ferro, 2014;Szumik et al., 2012). If the cells are too small the number of artificially empty cells increases and reduce the number of sympatric species, preventing correct detection of AEs.
The practical problem with fine grids is that available data are usually incomplete including many gaps. To counteract this, we established filling values, which minimize the number of empty cells (Arias et al., 2011;Casagranda et al., 2009;Szumik & Goloboff, 2004). Therefore, we analyzed our matrix considering different filling values for assumed and inferred presences for smaller cells (square cells of 0.5° and 0.25° per side; Tab I). An assumed presence implies that, even if a species has not been recorded in a cell, its presence, although unconfirmed, is suspected or assumed by the user due to its proximity to records in neighbouring cells (Szumik & Goloboff, 2004). An inferred presence is implemented by the method itself when a species is absent from a cell but present in surrounding cells (Szumik & Goloboff, 2004). Therefore, as Szumik & Goloboff (2004) highlighted, the assumed presences are given prior to the analysis, whereas the inferred presences are postulated as part of the analysis. Not setting filling values for the analysis of data when using grids with larger cell sizes is based on the fact that density of points of occurrence in relation to the proportion of the study region that is covered by a cell of large size makes it unnecessary. For a brief explanation on how the algorithm works see Aagesen et al. (2009).
Grid origin was arbitrarily defined and located at 70° long W and 22° lat S. We carried out the endemicity analysis Tab. I. Results of the endemicity analysis on data set from different taxa of small mammals (SM, small mammals; TSM, terrestrial small mammals; Ch, Chiroptera; Di, Didelphimorphia; Ro, Rodentia). The total number of consensus areas obtained by using different search (analytical) parameters (i.e., cell sizes and filling values; T) and the number of biogeographically informative areas (BI, or areas not almost equivalent to the study area, see main text) as a subset of the first are indicated. Cell sizes (in degrees) and filling values (first value: assumed presences, second value: inferred presences) are also indicated. Taxa  SM  TSM  Di  Ch  Ro  Consensus areas  T  BI  T  BI  T  BI  T  BI  T  BI  1°x1° 0-0  8  7  6  6  1  1  4  3  7  through a heuristic search and default NDM parameters: searching groups of cells by adding/eliminating one cell at a time, and saving groups defined by two or more endemic species and with scores higher or equal to 2.0. Groups of cells with more than 50% of species in common were ruled out retaining those with highest scores, and groups of cells with a score up to 1% inferior were stored in memory. We obtained consensus areas using the loose consensus rule in VNDM (see Aagesen et al., 2013). The distributional patterns were evaluated in the context of traditional biogeographic divisions by plotting the distribution of the defining species upon the terrestrial eco-regions as defined by Burkart et al. (1999) and Olson et al. (2001). The results were mapped using the program Global Mapper v11.02. Finally, all consensus areas were grouped in General Patterns (GPs) employing a qualitative criterion (see DaSilva et al., 2015), according to their fit to the traditional biogeographic divisions considered.

RESULTS
We reviewed a total of 8,673 specimens, and we identified 8,400 of them to the species level. From these, 8,117 had precise punctual localities and represented 119 of the 124 NWA species included in this work (Appendix). Altogether, the principal database contains 11,596 georeferenced records of 124 of the 128 known NWA small mammal species [exceptions are Necromys amoenus (Thomas, 1900), Oxymycterus akodontius Thomas, 1921, Graomys edithae Thomas, 1919, and Ctenomys fochi Thomas, 1919.
The results we obtained by analyzing data from the different taxa of small mammals are summarized in Tables I to IV. The analysis resulted in 165 consensus areas. Four large areas almost equivalent to the study region were not considered in the following characterization (see Discussion). The remaining 161 areas represent PDCs or AEs in strict sense characterized by 7 species of marsupials, 28 species of bats, and 63 species of rodents. Thirty-two species (25% of all study region small mammal fauna that characterize consensus areas) are endemic to the study region and define six GPs that are AEs in strict sense (Tab. III). Other 12 GPs are mainly defined by widely distributed species (whose distributions exceed the study region) and therefore we consider these GPs as PDCs. Of all the 18 GPs, 6 (3 of the AEs in strict sense and 3 PDCs) run transversal to the Andes, in an east-west direction, and group together eastern and western Andean biomes, joining up lowland and moist or dry highland species (Figs 14-19). On the other hand, the other 9 PDCs run parallel to the Andes, in a north-south direction, including eastern or western Andean biomes (Figs 2-10). The other 3 AEs are located in the south of the study region, either on the eastern moist Andean slopes and lowlands or on the western dry valleys (Figs 11-13). Of the 18 GPs, 6 were previously identified (Sandoval & Ferro, 2014). The other 12 are novel distributional congruence patterns or are more detailed or precise definitions of some patterns previously outlined. Since the characterization of the GPs already reported by Sandoval & Ferro (2014) is basically the same, the new characterization is not included in the following statement. The number of areas equivalent to each of such GPs, the databases and cell sizes of the grids that allowed their identification, and the number, identity and distributional fit of the characterizing species are provided in Figs. 2-7, and in Tables II and III.

A patch of the Northern Argentinean Yungas: Finca Las
Capillas (Fig. 4). Two consensus areas were equivalent to this pattern, which is nested within the North-eastern Andean Slopes pattern. This PDC were only obtained analyzing the SM database with the smallest grid sizes (Tab. II). Three species characterize this PDC. There is one endemic species characterizing these areas (Tab. III).

Eastern Andean Slopes and Lowlands ("Argentinean
Yungas and Chaco") ( Fig. 5). Two consensus areas were equivalent to this pattern. This PDC were obtained only analyzing the Ch database (Tab. II). Fifteen species characterize this PDC. There are none endemic species characterizing these areas (Tab. III).

North-eastern Andean Slopes and Lowlands ("Northern
Argentinean Yungas and Chaco") ( Fig. 6). Six consensus areas were equivalent to this pattern. This PDC were obtained analyzing the SM, TSM, and Ro databases with big grid sizes (Tab. II). Twenty-five species characterize this PDC. There are nine endemic species characterizing these areas (Tab. III).

Mountain Tops and Western Andean Slopes ("Northern
Argentinean Highlands") ( Fig. 7). Three consensus areas were equivalent to this pattern. This PDC were only obtained analyzing the TSM and Ro databases with large grid sizes (  (Fig. 16). Ten small consensus areas were obtained in the north of the study region (Tab. II). These areas were characterized by nine species ( (Fig. 19). Three consensus areas were equivalent to this GP. This AE was obtained analyzing the SM, TSM, and Ro databases with a moderate grid size. Three endemic species characterize this AE (Tab. III), of which one correspond to the western arid slopes (P. aureus Mares, Braun,  and the other two inhabit both the western arid slopes and the eastern lowlands. Tab. II. Results of the endemicity analysis on data from different data set of small mammals. The number of consensus areas obtained by analyzing data from each of the taxa considered (S, small mammals; T, terrestrial small mammals; D, Didelphimorphia; C, Chiroptera, R, Rodentia) with grids with different cell sizes (a: 1°; b: 0.75°; c: 0.5°; d: 0.25°) is shown. Filling values are not shown. The General Pattern numbers (GP) correspond to those in the text.

DISCUSSION
Floristic vs faunal patterns. As highlighted by Godoy-Bürki et al. (2013), since endemism is a fundamental criterion to identify conservation areas, it is urgent that botanists and zoologists to conduct local distributional analyses, "in order to obtain an integrated solution for protecting both the flora and fauna in the region" (Godoy-Bürki et al., 2013). In fact, in NWA there are hotspots for vascular plants and several animal taxa, including mammals (Szumik et al., 2012). However, knowledge on distributional patterns and endemism for this region is very uneven among different taxa.
Surprisingly, it seems that, while detailed quantitative analyses of plant species distribution data do not result in units that correspond to Cabrera's phytogeographic divisions at this spatial scale , analyses of animal (here, small mammal) species distribution data generally do. So, despite the system Cabrera used was not based on quantitative studies and did not apply consistent criteria for defining the individual phytogeographic units , what he observed in local and regional flora composition may be determining animal distributions (e.g. Sandoval et al., 2010;Szumik et al., 2012;Sandoval & Ferro, 2014), although plant endemic species distribution quantitative analyses result in areas of endemism that do not correspond to his divisions . Although many of the areas we obtained, both AEs and PDCs, are constituted by cells that include a wide array of habitats, the natural history and habitat preferences of the characterizing species allow us to certainly relate them to phytogeographic units previously defined by Cabrera (Cabrera & Willink, 1973;Cabrera, 1976). All the obtained AEs are located in the south of the study region, either in the eastern Andean slopes and the western Andean biomes, and even in many cases those landscapes are merged in one AE. Similarly to what happens with vascular plants , small mammal endemic species are far from evenly distributed in the study region. Of the 33 endemic species (Table 3), only five are more or less latitudinally widely distributed [T. cinderella (Thomas, 1902), T. sponsorius (Thomas, 1902), B. lactens (Thomas, 1918), G. domorum (Thomas, 1902), and M. shiptoni (Thomas, 1925]. Interestingly, endemism, unlike what happens in vascular plants , does not declines towards the south, but does declines towards the west of the study region. Ten endemic species have their distributional ranges located in the north of the study region meanwhile 18 are in the south of the study region. On the other hand, 23 species inhabit in environments of eastern Andean slopes and lowlands meanwhile only ten inhabit western Andean biomes. The geographical location of our AEs and PDCs are mostly congruent with the most important hotspots for endemic vascular plant species. Our largest GPs include almost all the endemic species hotspots defined by Godoy-Bürki et al. (2013). And there are even coincidences at a more local scale: our GP16 Figs 6-9. Fig. 6, a case of GP5, North-eastern Andean Slopes and Lowlands PDC recovered through the analysis of TSM database, with a grid of cells of 0.75°x0.75° and without filling values. Fig. 7, a case of GP6, Mountains Tops and Western Andean Slopes PDC recovered through the analysis of TSM and Ro databases, with a grid of cells of 1°x1° and without filling values. Fig. 8, a case of GP7, Eastern Lowlands PDC recovered through the analysis of TSM database, with a grid of cells of 1°x1° and without filling values. Fig. 9, a case of GP8, Widespread Western High Andes PDC recovered through the analysis of SM, TSM, and Ro databases, with a grid of cells of 0.75°x0.75° and without filling values. 6 7 8 9 ( Fig. 17) includes the Sierra del Aconquija, between the provinces of Tucuman, Salta and Catamarca, which has the highest concentration of endemic plant species in the study region (Godoy-Bürki et al., 2013); and our GP15 (Fig.  13), which includes Andalgalá, Santa Maria, and Ambato in Catamarca, was recognized as another important AE for plant endemism (Godoy-Bürki et al., 2013). Endemism and co-occurrence patterns. Highest numbers of endemic species (but not necessarily recognized AEs) are found in the eastern Andean slopes in Jujuy and Tucumán/Catamarca, with nine and eight endemic species respectively, and in the western Andean biomes in Tucumán/ Catamarca, with eight endemic species. Of these three high endemism sites that we a priori and qualitatively identified, our analysis was able to recover two patterns as AEs in strict sense, while the other was recovered as PDCs (which are characterized both by endemic and non-endemic species). However, both endemism patterns are impoverished in term of defining species: in any case the recovered areas were characterized by the complete assemblage of endemic species but only by a subset of such species (Tab. III). The South-eastern Andean Slopes pattern is an AE and corresponds to an endemism peak. Sandoval & Ferro (2014) recovered this pattern, characterized by 14 rodent species, 12 been endemic to the study region, been our GP also characterized only by rodent species (but far fewer).
Figs. 10-13. Fig. 10, a case of GP9, North-western High Andes PDC recovered through the analysis of SM, TSM, and Ro databases, with a grid of cells of 0.25°x0.25° and filling values of 80 for assumed presences and 100 for inferred presences. Fig. 11, a case of GP13, South-eastern Andean Slopes AE recovered through the analysis of SM, TSM, and Ro databases, with a grid of cells of 0.25°x0.25° and filling values of 20 for assumed presences and 40 for inferred presences. Fig. 12, a case of GP14, South-eastern Andean Slopes and Lowlands AE recovered through the analysis of SM, Ch, and Ro databases, with a grid of cells of 0.25°x0.25° and filling values of 60 for assumed presences and 80 for inferred presences. Fig. 13, a case of GP15, South-western High Andes AE recovered through the analysis of SM, TSM, and Ro databases, with a grid of cells of 0.5°x0.5° and filling values of 0 for assumed presences and 20 for inferred presences.

11 12 13
Despite the low number of our characterizing species, this AE has a high endemicity score and repeated occurrence (Tab. IV). This pattern corresponds to the Southern sector of the Yungas forest of NWA, which has been differentiated by phytogeographers as an impoverished version of the Northern sector rather than on exclusive floral elements (Cabrera, 1976;Morales, 1996;Brown et al., 2001). Moreover, Aagesen et al. (2012) reported that vascular flora endemism declines gradually towards the south of the study region. However, this area was recovered for insects (Navarro et al., 2009), for a taxonomically diverse data set when analysed using quantitative approaches (Szumik et al., 2012), and for rodents (Sandoval & Ferro, 2014; our study), revealing the Southern sector of the Argentinean Yungas as a distinctive faunal AE, and, at least in the case of endemic small mammals, not a impoverished one. On the other hand, regarding the case of the other obtained AE that corresponds to an endemism peak, the Monte biogeographic province in NWA was recover as and named as the South-western High Andes pattern. This western Andean biome occurs on the arid western (rain shadow) slopes. As the Yungas forest reach its southern tip on the humid eastern Andean slopes in NWA, the Monte province reaches its northernmost extension on the west-faced slopes of the same mountain ranges. Also, one of the highest concentrations of endemic plant species were found mostly Unlike to what happens with endemic vascular plants Godoy-Bürki et al., 2013), whose hotspots at a regional scale are concentrated in the arid ecoregions of NOA (with 80% of the endemic flora; Aagesen et al., 2012), the principal habitat types for endemic small mammal species are not the western or eastern arid or semi-arid biomes, but the eastern humid Andean slopes, with 20 endemic species (versus 13 in arid and semi-arid environments). Notwithstanding, arid/semiarid biomes and humid landscapes are represented by the same number of AEs: one for arid/semi-arid biomes (GP15), one for humid landscapes (GP13), and four mixed GPs, which include both habitat types (GP14,16,17,and 18). Contrary to our results, Aagesen et al. (2012) reported that very few vascular plant species are restricted to subhumid or humid environments in this region. They highlighted this surprising poverty in endemic species of the otherwise diverse Tucumano-Bolivian Yungas forest at a local scale. It should be noted that Aagesen et al. (2012) stressed that the mountain grasslands above the tree line were included in the Puna (following Ibisch et al., 2003), meanwhile in the phytogeographical scheme outlined by Cabrera (1976), which we follow, these grasslands are considered the highest strata of the Yungas. Notoriously, supporting Ibisch et al. (2003) scheme, we obtained the Mountain Tops and Western Andean Slopes pattern (a PDC), which, even though it represents an association between areas, apparently not correspond to a transition zone between biogeographic units but a natural unit characterized by species mostly inhabiting all environments represented in the area. Obtaining this GP allows formally grouping all those highlands environments (High Andes, Puna, and mountain grasslands) that harbour common assemblages of rodent species. This issue will be presented in more detail in another manuscript. On the other hand, the Widespread Eastern Andean Slopes allows us to identify the entire latitudinal extension of the Yungas forest as a discrete biogeographic unit regarding the NWA geographic context. This is a stable pattern since it was recovered with different databases with different grid sizes (Tab. II). However, only 5 of the 36 species that characterize the Yungas forest are exclusive or almost exclusive of the NW Argentinean Yungas. This may well indicate that the Yungas of NWA may be part of a larger AE extending beyond the limits of our study region: the Yungas forest in NWA is the tail of a broader area that extends towards the north through Bolivian Yungas and probably to southern Peru. Alternatively, the recognition of distributional congruence in the species distributions may be due to a similar response to a unique combination of ecological and historical factors restricting species ranges that not necessarily affect in the same way these species north or southwards. The recognition of this pattern was also reported for bats and marsupials (Sandoval et al., 2010) and for rodents (Sandoval & Ferro, 2014). Meanwhile, the North-eastern Andean Slopes, near the Bolivian border, is just traversed by the Tropic of Capricorn. Certainly, the northern part is the richest sector of the NWA Yungas forest which gradually impoverished southward (Ojeda & Mares, 1989;Ojeda et al., 2008), althought this statement do not applies to endemic small mammal species (see above). For bats and marsupials (Sandoval et al., 2010) and for rodents (Sandoval & Ferro, 2014) the Northern sector of the Argentinean Yungas was also recovered. Contrary to the low number of endemic species found by Sandoval & Ferro (2014) for this pattern, according to our

19
results, almost half of the defining species (9 of the 23) are exclusive or almost exclusive to this PDC. Notably, even defined as a PDC, the fact of containing several endemic species would not allow a delimitation of an AE in the context the analyses we performed. Probably, this PDC would be obtained as an AE if distributional records of only endemic species are included in a study that do not include widespread species . In that case, perhaps we could define a PDC and an overlapping AE, and they would belong to the same GP. However, given that our analyses include both endemic and widely distributed species in the datasets, we can only speculate. We identify Argentinean High Andes (North-western High Andes, GP9) and Puna (Widespread Western High Andes) provinces as discrete biogeographic units in the geographic context of NWA. Maybe, the High Andes and the Puna of NWA are part of larger AEs extending southward in Argentina and northward into Bolivia. For rodents (Sandoval & Ferro, 2014) the Argentinean High Andes and Puna were also recovered. However, as for vascular plants, the relative low number of endemic highland species is unexpected and may be underestimated as collection localities in those landscapes are scarce . As Sandoval & Ferro (2014) mentioned, large areas of these highlands remain unexplored and so further exploration should clarify the existence of different AEs associated with mountaintops on the highlands of NWA. Despite that, one of the highest concentrations of endemic plant species were found mostly in the Central Andean Puna environments, especially in transition zones with the Southern Andean Yungas (Godoy-Bürki et al., 2013). Szumik et al. (2012) were able to recover the Chaco biogeographic province but later Sandoval & Ferro (2014) failed to recognize it neither as an AE nor as a PDC. We recovered the Chacoan biome both as hybrid areas, mixed with the Yungas environments (the Eastern Andean Slopes and Lowlands and the nested North-eastern Andean Slopes and Lowlands and South-eastern Andean Slopes and Lowlands) and as independent areas (Eastern Lowlands, GP7). Although the scattered sampling localities may prevent or difficult its identification as a sympatric distributional pattern, the Chacoan small mammals, and the species inhabiting both biomes (Chaco and Yungas), allow the recognition of this biogeographic province as a PDC (see also Sandoval & Barquez (2013) for a detailed interpretation of the Chacoan bat fauna assemblages).
Finally, despite the identification of four large areas almost equivalent to the study region, we were not able to obtain a pattern equivalent to the southern part of the central Andes, which was recognized as a single AE, defined by at least 53 species of vascular plants that are widely distributed within the study region but are endemic to it . We considered our four wider consensus areas biogeographically not informative at this spatial scale because they have a too large surface (as mentioned, almost equivalent to the study region) and are characterized by many widespread species, in some cases presenting high endemicity values, all these making impossible their interpretation at this analysis scale. In our spatial scale of analysis, these areas could be interpreted as areas that integrate transition zones, because they present varied and complex characterizing faunas. So, it is likely that these areas are part of a more inclusive pattern evident to a larger scale of analysis. However, this interpretation is purely speculative. For these reasons, those areas are not considered in this discussion.
'Spurious' areas? Finally, we obtained three sets of 'merged-unit' areas: latitudinally extensive (GP10) and northern (GP11 and 12) and southern (GP16, 17, and 18) groupings of eastern and western Andean biomes. All these areas combine consensus areas corresponding to well defined classic main patterns. Sandoval & Ferro (2014) interpreted these areas as 'spurious' areas and, as Aagesen et al. (2013) stated, as a likely outcome of using the 'loose' consensus rule. The loose rule merges areas if they share a user defined percentage of their defining species with at least one other area in the consensus . The 'loose' consensus rule is a useful tool for detecting gradual overlapping distribution patterns and replacement among areas. Therefore, Sandoval & Ferro (2014) considered that the spurious areas may well be indicating a transition zone between the main patterns involved. Some of our 'merged-unit' areas actually seem to be what Sandoval & Ferro (2014) called 'spurious' areas, since they more likely represent turnover gradients rather than congruence patterns (Augusto Ferrari, com. pers.). Interestingly, Aagesen et al. (2012) do not share our view. In their analyses all areas were characterized by species from very different altitudes that do not necessarily strictly coexist but are founded in the same cell or cell set because in the study region several phytogeographic units are separated from each other by very short distances. They consider those AEs as valid and their high endemic areas include several environments and are characterized by vascular plant species that are not sympatric but inhabit highly altitude and aridity variable landscapes. This is the case of GP11, 12 (both PDC), and 16 (an AE), because in these areas species that are not sympatric at all, which inhabits very environmentally different landscapes, are grouped.
However, GP10 (a PDC), 17, and 18 (both AEs) include some species that are in fact sympatric. Moreover, these areas are among the high endemic areas found when analyzing the endemic vascular flora  and present high concentrations of endemic plant species (especially the transition zones from the Southern Andean Yungas to the High Monte, Godoy-Bürki et al., 2013). So, interestingly, small mammals and vascular plants appear to define the same AEs in NWA (Lone Aagesen, com. pers.). This finding is very important, especially considering that areas with high levels of endemism are priorities for conservation because they reflect past and potentially future speciation events (Rosenzweig, 1995;Fa & Funk, 2007). Therefore, areas identified by their high number of endemic taxa represent essential conservation targets particularly if occurring in rather small geographic areas (Vane-Wright et al., 1991;Linder, 1995;Peterson & Watson, 1998).
Taxa. Rodent species define 15 of the 18 GPs, while bats define four and marsupials only define one. The GP defined by bats and marsupials and two defined by bats are also defined by rodents. So, only three GPs are not recovered through the analysis of rodent species distributional data, and even these GPs have rodent characterizing species when analized together with all small mammal species (GP3 and GP10). Only in GP4 rodents have no participation at all. Clearly, at this spatial scale, non-flying mammals, particularly rodents, are bigeographically more valuable species than flying mammals (bat species). The other terrestrial small mammals considered (marsupials) include very few (10) and not always distributionally overlapping species in NWA, and this is probably why they are so little involved in the characterization of areas, compared to rodents. Therefore, as suggested by Sandoval & Ferro (2014), rodents are very useful to analyse patterns of endemism and biographical regionalization in small geographical areas. This is probably due to the fact that a large proportion of mammals with restricted distributions are rodents, independently of their taxonomic level (genera/ species) or spatial scale (Arita et al., 1997;Danell, 1999;Danell & Aava-Olsson, 2002), that they are present in all type of biomes from tropical moist forest to gelid highland deserts, and that they constitute quite conspicuous assemblages as a consequence of adaptation to environments (Mena & Vázquez-Domínguez, 2005;Patterson et al., 1998). Moreover, rodents are the most diverse order of mammals containing 42% of known species (Musser & Carleton, 2005), thus frequently being the most important part of mammal assemblages concerning number of species.
Using different search (analytical) parameters. Numerous studies support the hypothesis that the results obtained at a given spatial or temporal scale can not necessarily be obtained or be significant at larger or smaller scales (Lyons & Willig, 1999, and references cited therein), so it is interesting and valuable to document scale dependence of different patterns. Casagranda et al. (2009) analyzed the implications of using different scales of analysis with the optimality method, recovering some patterns at all spatial scales, but also recovering certain patterns only in some particular scale. According to Laffan & Crisp (2003), we are closer to finding a complete representation of the true pattern of distributional congruence and endemicity varying the spatial scale of analysis than without doing it. In our analysis, the dependence of local patterns of distributional congruence and endemicity on the spatial scale does not seem to be decisive: most patterns remain constant during the analysis at different scales (i.e., grid sizes). Our GP1 and 2 are obtained using four different grid sizes, GP16 and 17 are obtained using three different grid sizes, and GP4, 5, 6, 8, 12, and 14 are obtained using two different grid sizes (Tab. II). Distributional congruence and endemicity patterns that do not support changes in the cell size can be a simple artifact of a specific grid, while resistance to changes in cell size is an important factor in assessing a particular distributional pattern (Aagesen et al., 2009). According to this framework, our GP3,7,9,10,11,13,15, and 18 could be artifacts of the technique without a real biogeographical significance. Considering this, we think that in fact recognition of GP3 (A patch of the Northern Argentinean Yungas: Finca Las Capillas), which constitutes a nested area within the Northern sector of the Argentinean Yungas, may be due to the substantial sampling effort invested in this area by Dr. Barquez and its working group: the area was visited on numerous occasions, in different seasons, the last almost 20 years. Thus, this small area of the Argentinean Yungas might be seen as a sampling artifact rather than isolated populations inhabiting only there in Argentina. Further exploration around this area should clarify whether the disjunction of the distribution of the characterizing species is real, or if this area is part a continuous area extending northward. However, as highlighted by Aagesen et al. (2012), in some instances, dependence on obtaining a pattern to a particular grid cell size, may relate to the characteristics of the available data and not necessarily with the validity of the pattern, since small cell sizes requires collection sites are very close to each other. In the cases that collection sites are not so closely, the pattern will emerge only when using larger cell sizes which allow wider distances between the collection sites . We think this is the case of GP7, 9, 13, and 15. General Patterns 13 (South-eastern Andean Slopes) and 15 (South-western High Andes), which constitute AEs in strict sense, was already discussed (see above).

CONCLUSIONS
While detailed quantitative analysis of plant species distribution data does not result in units that correspond to Cabrera's phytogeographic divisions at this spatial scale, analysis of animal species distribution data does. Even in a region whose biogeographical patterns have been studied based on several of its floral and faunal species, we were able to identify previously unknown meaningful faunal patterns (Eastern Andean Slopes and Lowlands, North-eastern Andean Slopes and Lowlands, South-eastern Andean Slopes and Lowlands, Eastern Lowlands, and Mountain Tops and Western Andean Slopes) and more accurately define some of those already identified. We identify PDCs and AEs that conform Eastern Andean Slopes Patterns, Western High Andes Patterns, and Merged Eastern and Western Andean Slopes Patterns, some of which are re-interpreted at the light of known patterns of the endemic vascular flora. Endemism do not declines towards the south, but do declines towards the west of the study region. Highest numbers of endemic species (but not necessarily recognized AEs) are found in the eastern Andean slopes in Jujuy and Tucumán/Catamarca, and in the western Andean biomes in Tucumán/Catamarca. The principal habitat types for endemic small mammal species are the eastern humid Andean slopes. Notwithstanding, arid/semi-arid biomes and humid landscapes are represented by the same number of AEs. Rodent species define 15 of the 18 GPs, and only in one they have no participation at all. Clearly, at this spatial scale, non-flying mammals, particularly rodents, are bigeographically more valuable species than flying mammals (bat species).