Snakes in a seasonally dry tropical forest in northeastern Brazil

in a seasonally dry tropical forest in northeastern Brazil Abstract: This study aimed to provide information on composition, abundance, and estimated snakes richness in an area of arboreal Caatinga (seasonally dry tropical forest) and analyze patterns of faunistic similarity between assemblages of snakes in the semiarid region of northeastern Brazil. The snakes found within the Fazenda Almas Natural Private Reserve (RPPN Fazenda Almas) were sampled during a 10-year period, with monthly monitoring, employing time constrained search, pitfall traps with drift fences, and donations from local collectors. Twenty-two snake species were recorded, with a predominance of terrestrial species with diurnal-nocturnal activity patterns. The species accumulation curves reached its asymptote, indicating that all possible species in the study area had been recorded. Our results indicated that the snake fauna in the study area is similar to other snake assemblages in localities with Caatinga vegetation in the Sertaneja Depression (“ Depressão Sertaneja ”) drylands, and that those assemblages constitute faunistic units typical of the interior of northeastern Brazil.


Introduction
The Caatinga biome extends throughout interior of northeastern region of Brazil and the northern part of the state of Minas Gerais and covers an area of approximately 852,261 km 2 (Ab'Sáber 1977, Prado 2003, Silva et al. 2017. The predominant vegetation type is the seasonally dry tropical forest and woodlands (SDTFW) (Pennington et al. 2009, Queiroz et al. 2017). The SDTFW comprises at least thirteen different phytophysiognomies with elevated densities of woody plants, known collectively as caatingas, whose production of leaves and flowers are dependent on seasonal rainfall (Andrade- Lima 1981, Prado 2003. The mean annual rainfall in the Caatinga biome varies between 200 and 700 mm, with mean annual temperatures varying from 25 o to 27 o C (Pennington et al. 2009). An important characteristic of the Caatinga is its high interannual rainfall variability with severe droughts that can last up to two years (Nimer 1972). The strong seasonal dryness makes even the rainy period sub-humid, and the dry season, lasts from five to seven months, almost totally arid (Ab'Sáber 1977).
The Caatinga has traditionally been described as a biome with low species richness and low numbers of endemic species, with its snake fauna is composed of elements that occupy the diagonal of open South American vegetation formations (Vanzolini 1974, Vanzolini & Williams 1981, Vanzolini 1988). However, the biome has been shown to be an important component of Brazilian biodiversity, with a significantly diversified fauna comprising endemic and threatened species (Rodrigues 2003, Leal et al. 2005, Albuquerque et al. 2012, Guedes et al. 2014a, Pereira-Filho et al. 2017.
Studies focusing on snake assemblages of northeastern Brazil have been quite limited and largely concentrated in coastal Atlantic Forest (e.g., Santana et al. 2008, França et al. 2012, Sampaio et al. 2018. Although inventories of the reptiles (Squamata) in Brazilian semiarid regions have been undertaken in recent years (e.g., Garda et al. 2013, Cavalcanti et al. 2014, Dal Vechio et al. 2016, Caldas et al. 2016, information concerning Caatinga snakes are generally limited to preliminary taxonomic lists derived from short term inventories. There have been few studies presenting detailed data concerning the natural history of the species found there (Vitt & Vangilder 1983, Mesquita et al. 2013a. Species richness is an essential measure to evaluate local and regional diversities. Studies of faunistic composition contribute greatly to our understanding on biodiversity and are, consequently, important components of ecological models and conservation strategies (Gotelli & Colwell 2001). Within that context, regional faunistic inventories are indispensable for providing a better comprehension of the structure, functioning, and natural variability of natural communities, and constitute an essential tool for conservation decision-making (Quintela et al. 2006). As such, the present study was designed to provide information concerning the composition, estimated species richness, and abundance of snakes in an area of Caatinga SDTFW, and to analyze the occurrence of faunistic similarity patterns between snake assemblages in the semiarid region of northeastern Brazil.

Study area
The present study was conducted in the Fazenda Almas Private Natural Reserve (RPPN Fazenda Almas), which covers a total of 5,500 hectares, of which 3,505 hectares (63.7%) correspond to protected area ( Figure 1). The average annual precipitation, temperature, and humidity are 560 mm (± 230 mm), 25°C, and 65% respectively (Paraíba 1985). Rainfall is concentrated from February to May (Paraíba 1985, CPTEC 2018.

Fieldwork/Snake Sampling
Snakes were sampled during monthly excursions (each one of eight consecutive days, followed by 20-day intervals) from Jan/2008 to Dec/2018 (totaling 132 months of monitoring). Snake sampling was carried out using combinations of three sampling methods: timeconstrained search, pitfall traps with drift fences, and donations from local collectors (sensu Sawaya et al. 2008, Martins & Nogueira 2012, Guyer & Donnelly 2012. Geographical coordinates of the sampled points (time constrained search and pitfall traps with drift fences lines) are in Table 1.
Time constrained search were performed by two experienced researchers walking slowly, searching for snakes in all visual accessible microhabitats for at least 15 hours per day (daytime corresponded to 10 h from 06:00 to 12:00 and from 13:00 to 18:00, and night time corresponded to 5 hours, from 18:30 to 00:00). The sampling effort and encounter rate of snakes were based on person-hours of search (Martins & Oliveira 1998, Guyer & Donnelly 2012). The total sample effort was 240 person-hours per month.
The pitfall traps with drift fences (Corn 1994, Cechin & Martins 2000 were placed along six lines containing ten 60 l buckets each (totaling 60 buckets) that were buried to ground level, approximately 10 m apart, and interconnected by 1 m high plastic tarps (Figure 3). A minimum distance of 200 m was maintained between trap lines to reduce the possibility of dependent sampling. The buckets remained open for eight consecutive days per month but were covered during the 20-day intervals between excursions. The total sample effort was 11520 bucket-hours per month.
Local collectors represented a complementary method of sampling. Snakes were captured by local residents, reserve employees, other searchers, or encountered dead along access roads leading to RPPN Fazenda Almas (Sawaya et al. 2008, Martins & Nogueira 2012. Local collectors were never encouraged to kill snakes, but they usually kill the snakes they encounter so we simply asked them to donated the snakes they killed and recorded some basic information about the location and time. The captured snakes were euthanized with an intracoelomic lidocaine injection (0.25 ml of a 100 mg/ml solution), preserved in 10% formalin, and stored in 70% ethanol (see Foster 2012), according to Portaria CFBio 148/2012-Anexo IV. The specimens were collected under an ICMBio collecting permit (SISBIO licenses 14105 and 65948-1). The specimens not collected were recorded, marked with cuts on their ventral scales, and subsequently liberated into their natural habitat. Voucher specimens were deposited in the herpetological collection of the Universidade Federal da Paraíba (CHUFPB; Appendix 1). The taxa nomenclature used in the present study followed Grazziotin et al. (2012) and Costa and Bérnils (2018).

Data analyses
To evaluate sampling efficiency, we constructed species accumulation curves for snakes sampled at the RPPN Fazenda Almas, constructed from 1000 randomizations without replacements, considering individual species abundance (Gotelli & Colwell 2001, Magurran 2004, Colwell 2011. Additionally, we used species richness estimators based on abundance data (Chao 1 and ACE) and based on nonparametric incidence (Bootstrap, Chao 2, ICE, Jacknife 1 and 2) to determine the expected richness of snakes in the area (Colwell & Coddington 1994, Magurran 2004, Colwell 2011. Species accumulation curves and richness estimates were constructed in EstimateS 9.1.0 software (Colwell 2011).

Leptotyphlopidae
Epictia borapeliotes (Vanzolini, 1996)  The use of pitfall traps with drift fences recorded 12 species and 73 individuals (Table 2), with no species recorded exclusively by that sampling method. That method recorded the largest numbers of individuals of Apostolepis cearensis Gomes, 1915 (n = 17) and Erythrolamprus viridis (n = 18) ( Table 2). Local collectors recorded 21 species and 164 individuals, although no species was recorded solely by that method ( Table 2). The species with the largest numbers of records made by local collectors were B. erythromelas (n = 27) and Thamnodynastes phoenix (n = 27). The species accumulation curves (constructed using the combined results of all three sampling methods) reached its asymptote at approximately 280 recorded individuals ( Figure 6A), corresponding to the 75th sampling month. Time constrained search reached its asymptote at approximately 141 recorded individuals ( Figure 6B), corresponding to the 80th sampling month. The species accumulation curves for pitfall traps and local collectors demonstrate a tendency to stabilize ( Figure 6C and D). All richness estimators indicated the existence of 22 species, except for Jacknife 2, which estimated a lower richness than the total number of species actually encountered (Table 3). At the end of the study period, one doubleton (Philodryas olfersii) and no singletons were recorded at RPPN Fazenda Almas.   Table 3. Species richness estimates for snake assemblage at RPPN Fazenda Almas, São José dos Cordeiros municipality, state of Paraíba, northeastern Brazil, using different estimators based on the numbers of individuals recorded. CSM-combination of all sampling methods, TCS-Time constrained search, PTDF-pitfall traps with drift fences, LC-local collectors. Sobs = total number of species observed in a sample, CI = 95% confidence interval. The Jacknife 1 and 2, Bootstrap, ACE, and ICE estimators did not generate confidence intervals.

Mean (± SD)/95% CI (Lower and UpperBound)
Observed richness (  Cluster analysis and NMDS demonstrated the same similarity patterns regarding the different snake assemblages (Figures 7 and 8, respectively). Cluster analysis (cophenetic correlation coefficient = 0.86) based on 70 species and 23 Caatinga localities generated two groups: a smaller group formed by three snake assemblages in different Caatinga Ecoregions, and a larger second group composed of the remaining assemblages (Figure 7). The larger second group also comprised two faunistic sets: a smaller set composed of snake assemblages from the São Francisco-Gurguéia and Complexo Ibiapaba-Araripe ecoregions, and another set composed of a group of snake assemblages from those same two ecoregions and of faunas from the Dunas do São Francisco and the Depressions Sertaneja Setentrional and Meridional (Figure 7). Non

Discussion
The snakes at RPPN Fazenda Almas were largely terrestrial and semi-arboreal, with wide distributions throughout the Caatinga biome as well as other Brazilian biomes. The predominance of xenodontines and species with wide geographic distributions in the study area makes it similar to other neotropical snake assemblages (e.g., Vitt & Vangilder 1983, Strüssman & Sazima 1993, Rodrigues 1996, , Sawaya et al. 2008, Mesquita et.al. 2013a, Pedrosa et al. 2014, Pereira-Filho et al. 2017, Sampaio et.al 2018 and follows a pattern of snake composition described for the Caatinga biome, including endemic species, such as Bothrops erythromelas, Thamnodynastes sertanejo, and T. phoenix (Guedes et al. 2014a(Guedes et al. , 2014b. In relation to the activity periods of those snakes at RPPN Fazenda Almas, greater numbers of diurnal-nocturnal and exclusively nocturnal species were recorded, being different from a study undertaken by Vitt & Vangilder (1983) and Mesquita et al. (2013a) who reported a dominance of strictly diurnal species in the Caatinga biome. The absence of aquatic/semi-aquatic species in the study area was probably related to the existence of only temporary bodies of water there (such as small streams, lakes, pools, and artificial waterholes) with very short periods of water retention and long periods without any standing water. Those temporary aquatic habitats do not favor the occurrence of snakes that are dependent on permanent bodies of water, and explains why aquatic/semi-aquatic species represent only a small component of the snake fauna of the Caatinga biome (Guedes et al. 2014a).
The snake assemblage at RPPN Fazenda Almas was rich in species as compared to other localities in the Caatinga biome (e.g., Miranda & Santos 2011, Muniz & Santos 2011, Garda et al. 2013, Cavalcante et al. 2014, Pedrosa et al. 2014, Benício et al. 2015aand b, Pereira et al. 2015, Pereira-Filho et al. 2017. The high species richness recorded at RPPN Fazenda Almas may reflect the long-duration study period (10 years of monthly monitoring) and the fact that the area represents the most well-preserved remnant of Caatinga vegetation in the state of Paraíba (see Lima & Barbosa 2014). Species records in faunistic studies can be influenced by the sampling efforts employed (duration of the study), the size of the area, as well as habitat heterogeneity and structure (Strüssman & Sazima 1993, Ugland 2003, Mesquita et.al. 2013b. Conserved environments are generally more heterogeneous and structured, providing adequate habitats for wide varieties of species as compared to homogeneous (and consequently less structured) environments (Lieberman 1986, Viana et al. 2001, Pardini & Umetsu 2006, Macedo et al. 2008.
Snake abundance in the study area is similar to that found in other Caatinga localities, with a dominance of the species Bothrops erythromelas, Thamnodynastes phoenix, Oxyrhopus trigeminus, and Erythrolamprus viridis. Those species are generalists and demonstrate significant plasticity in terms of their diets and habitat uses (Vanzolini et al. 1980, Vitt & Vangilder 1983. They occur in forested and shrub environments, as well as in perianthropic areas (Pereira-Filho et al. 2017) and are widely distributed throughout the Caatinga biome (Guedes et al. 2014a). Unfortunately, most studies of the snake fauna of the Caatinga have been short-term inventories that produced only preliminary species lists (Guedes et al. 2014a). Very few publications contain abundance data (e.g., Borges-Nojosa & Arzabe 2005, Borges-Nojosa & Santos 2005, Borges-Nojosa & Cascon 2005, Freire et. al. 2009, Garda et al. 2013, Cavalcante et al. 2014, Dal Vechio et al. 2016 and even then, usually significantly underestimate true snake abundances, which prevents any secure comparisons between different Caatinga sites. Time-constrained search and the use of local collectors are efficient methods for determining the richness and abundance of species in a given area, as opposed to pitfall traps with drift fences. Although snakes are often rather difficult to encounter (as compared to lizards and anurans) as many species demonstrate cryptic or secretive behaviors (Sazima & Haddad 1992, Sawaya et al. 2008, time-constrained search and local collectors have proven to be adequate methods for sampling snakes in many other herpetofaunistic inventories undertaken in the Caatinga and other Brazilian biomes (e.g., Sazima & Haddad 1992, Sawaya et al. 2008, Santana et al. 2008, Freire et al. 2009, Pontes et al. 2008, Mesquita et al. 2013b, Caldas et al. 2016, Pereira-Filho et al. 2017, Sampaio et al. 2018. While visual search represents a low-cost technique, their strength will depend on the sampling effort and experience of the researchers involved (Mesquita et al. 2013b). The same is true for the use of local collectors, who largely record only the most generalist species (in terms of habitat use), which have less propensity to flee (Sawaya 2003).
Pitfall traps with drift fences are widely used for sampling amphibians and reptiles (Corn 1994, Greenberg et al. 1994, Cechin & Martins 2000, Fisher & Rochester 2012), but were not found to be very efficient in the present study, recording only smaller numbers of species and individuals (with the exception of Apostolepis cearensis and Erythrolamprus viridis). Although the use of pitfalls traps is quite common, this method should only be considered as a complementary sampling technique associated with other sampling methods in long-duration studies of snake assemblages (Cunha & Nascimento 1993, Greenberg et al. 1994, Cechin & Martins 2000. As such, comparing the deficiencies of each sampling method and considering financial and time investments in relation to the numbers of species and individuals recorded, the use of pitfall traps in short-term studies of snakes in Caatinga areas with shallow rocky soils is not indicated (Cechin & Martins 2000, Mesquita et al. 2013b. The combination of three sampling techniques and extended fieldwork efforts (10 years of monthly monitoring), the absence of singletons, and species accumulation curves reaching their asymptotes, made it possible to affirm that probably all of the snake species occurring in RPPN Fazenda Almas were recorded in the present study. The Jacknife 2 estimator predicted a richness less than the actual total number of species recorded. Such unusual result may indicate the inefficiency of that estimator for snake inventories in Caatinga sites. Sampling is considered complete in biodiversity studies when each species is represented by at least two individuals, without any singletons, as there will only be an extremely low probability of any additional species being recorded even with additional sampling efforts (Chao et al. 2009). The performances of the richness estimators here, the absence of singletons, and the shape of the species accumulation curves allowed a secure interpretation of each sampling method, and indicated the extremely reduced probability of encountering more species in the snake assemblage with greater additional sampling efforts (Colwell & Coddington 1994, Chao et al. 2009).
The snake fauna in the study area demonstrated high similarity with typical assemblages found in Caatinga sites in crystalline terrains of the Sertaneja depressions. Localities with similar phytophysiognomies and at identical latitudes likewise demonstrated similarities in terms of their snake faunas. Faunistic similarities tend to appear as a function of geographic proximity between localities but decrease along latitudinal gradients (Nekola & White 1999, Willig & Bloch 2006. Stochastic processes and the evolutive and biogeographic histories of species lineages can also affect assemblage compositions ( The Caatinga snakes demonstrate species distribution patterns that suggest the occurrence of regionalization of assemblages in diverse habitats throughout that biome (Guedes et al. 2014b). Similarities among the assemblages analyzed in the present study were noted between most of the areas in the Sertanejas Meridional and Sententrional depressions, which form a large group composed of localities whose snake fauna is composed of very common and widely distributed species in arid landscapes dominated by xeromorphic vegetations. This snake fauna can be considered typical of interplateau depressions, corroborating that lowland semiarid Caatinga regions constitute a faunistic unit characteristic of the interior of northeastern Brazil (Guedes et al. 2014b). A study employing parsimony analysis of endemism evidenced that dryland Sertaneja depressions have large similarities in terms of their lizard faunas, and those faunas are largely composed of widely distributed species plus relictual elements generally related to less arid climates along the edges of those ecoregions (Mesquita et al. 2017). As such, the areas included within Sertaneja depressions have very similar Squamata reptile faunas that constitute important components of the Caatinga biota, in spite of the fact that Meridional and Sententrional Sertaneja depression ecoregions demonstrate distinct climatic, geological, and physiognomic parameters (Velloso et al. 2002).
The RPPN Fazenda Almas demonstrates an elevated snake species richness and constitutes an important conservation area for snakes of the Caatinga, reflecting its very well-preserved Caatinga vegetation (Lima & Barbosa 2014), in spite of the fact that is located in a region that has been greatly impacted by deforestation, mining, extensive cattle grazing, predatory hunting, and with susceptibility to desertification (Souza et al. 2004. It should be pointed out that RPPN Fazenda Almas is located in an area considered to have priority for conservation efforts and is of extreme biological importance to the semiarid region (sensu Veloso et al. 2002, MMA 2002, Tabarelli & Silva 2003. In spite of the growing number of herpetofaunistic inventories in the Caatinga biome during the last 10 years (Guedes et al. 2014a, Mesquita et al. 2017, there is still a need for long term studies in all Caatinga ecoregions. Further studies would complement and/or corroborate the species distribution patterns already published for the different Caatinga ecoregions, and will increase our understanding on the ecology, species richness, and biogeography of snake assemblages in the Caatinga seasonally dry tropical forests.

Supplementary Material
The following online material is available for this article: Appendix 1 -Voucher specimens of snakes collected at the RPPN Fazenda Almas, municipality of São José dos Cordeiros, state of Paraíba, northeastern Brazil.