Distribution and invasive potential of the black-tufted marmoset Callithrix penicillata in the Brazilian territory

Biological invasions are one of the greatest existing threats to biodiversity. Invasive species can cause economic and environmental damage. Callithrix penicillata is naturally found in the Brazilian savanna and Caatinga. Its introduced populations have become a conservation problem due to its high occupancy potential, native fauna predation, competition with native primates, congeners and hybridization. We used Species Distribution Modeling (SDM) through the Maxent software in this study in order to identify areas with a higher probability of C. penicillata occurrence. The AUC value was close to 1 (AUC=0.966), with a curve value close to 1. Through the Jackknife test we observed that temperature seasonality was the variable most related to distribution (AUC=0.86), which agrees with other studies that show climatic variables influencing primate distribution. The Atlantic Forest in the Southeast and South regions of Brazil was indicated as susceptible to invasion by C. penicillata. The marmoset C. penicillata has become a successful invader of Atlantic Forest areas, causing depreciation in many native species and other problems. However, biological invasions might be mitigated or even extinguished through successful interventions and management.


INTRODUCTION
Biological invasions are responsible for significant environmental alterations and are one of the greatest existing threats to biodiversity [1]. Once settled in a new habitat, the invasive species threatens the native biodiversity, being able to cause potentially irreparable economic and environmental losses [1]. In order to mitigate this global problem, tools have developed that enable us to predict invasion events [2]. Among these, Species Distribution Modeling (SDM) has become increasingly important for predict biological invasions [3,4]. Species distribution models have been used in biogeography, conservation, ecological and paleontological studies [5].
Species Distribution Modeling can be designed promptly and with a low budget, helping to identify areas in which a species has a higher probability of occurring [6]. Precisely identifying areas that may be successfully occupied by invasive species is one of the greatest challenges when studying biological invasions [1]. Data used to determine the distribution of a species in a given geographical area is usually scarce and incomplete, which hinders conservation and management projects [7]. These projects are only made possible by knowing which areas have already been

MATERIAL AND METHODS
We used the Maxent software (www.cs.princeton.edu/~schapire/maxent/) in order to estimate the potential species distributions. This algorithm requires the entry of a set of layers or environmental variables (e.g., precipitation rates, altitude, etc.) and a set of georeferenced occurrence locations in order to generate a SDM of a given species [7,26], as used bellow.
The marmoset occurrence locations (302 occurrence points) ( Figure 1) were attained through an extensive literature review in the Web of Science (appswebofknowledge.ez25.periodicos.capes.gov.br), Scielo -Scientific Electronic Library Online (www.scielo.org/php/index.php), and Academic Google (scholar.google.com.br) databases, as well as consultations to biological collections (Appendix I) and in the Global Biodiversity Information Facility (gbif.sibbr.gov.br), Mammal Networked Information System (https://ecologicaldata.org/wiki/mammal-networked-information-system), SpeciesLink (splink.cria.org.br) and Táxeus (taxeus.com.br) databases. Unreferenced data was georeferenced through the Geoloc tool (splink.cria.org.br/geoloc) and Google Earth. Records with inaccurate information about the locality were discarded. The environmental variables used in this study were the 19 listed by Hijmans et al. (2005) [27], attained from consulting the WordClim database (www.worldclim.org). Additionally, data on the altitude and vegetation of biomes were attained from the Instituto Nacional de Pesquisas Espaciais -INPE (www.dpi.inpe.br/Ambdata/index.php). In order to reduce overfitting, which tends to be larger with larger number of dimensions, through principal component analysis (PCA), we used six variables that together explained 99% of the data variation, BIO 4 (temperature seasonality), BIO 5 (Max Temperature of Warmest Month), BIO 10 (Mean Temperature of Warmest Quarter), BIO 11 (temperature mean of the coldest quarter), BIO 12 (Annual Precipitation) and BIO 15 (Precipitation Seasonality).
An independent dataset was then built and divided into training data and testing data in order to assess the quality and reliability of the model. The testing dataset was created by using a 25% randomization of the presence points (totaling 227 training points and 75 test points). The adjustment measure in the model was a random prediction with an AUC value = 0.5.
Two other statistical parameters were taken into account, the omission rate and the binomial proportion [28]. These parameters help us understand how much the model failed to predict the occurrence of test points and how statistically significant it is. Complementary analysis of data overlay in the marmoset occurrence in Conservation Units, phytogeographic domains and priority conservation areas (www.mapas.mma.gov.br/i3geo/datadownload.htm) were carried out using the DIVA-GIS software (www.diva-gis.org/download).

RESULTS AND DISCUSSION
Our results show that the sites which are more susceptible to C. penicillata invasion outside their likely occurrence area are in the Southeast of the Atlantic Forest ( Figure 2).
The calculated values regarding the model's reliability were AUC=0.923. Pearce & Ferrier (2000) [29] consider that values over 0.75 are indicators of good model performance, and therefore the closer the area under the curve is to 1, the smaller the probability of the model being a result from a random prediction. The model presented low values for both the omission rate (0.000) and the binomial test (0), indicating that the generated models are significantly different from those generated at random. Regarding the environmental variables that most influenced the model prediction, the Jackknife test showed that the species distribution is closely related with the variables: Max Temperature of Warmest Month (BIO 5) AUC=0.84, and Temperature Seasonality (BIO4) AUC=0.83, followed by Annual Precipitation (BIO 12) AUC= 0.82. The influence of these variables on the species distribution is due to its natural habitat being the Cerrado, which has differences in the temperature seasonality throughout the year. Two distinct seasons are markedly present: the hot and rainy season (from October to April), in which 75% of the precipitation takes place and temperatures range from 20°C to 28°C, and the cold and dry season (from May to September), with temperatures going as low as 16°C and relative air humidity getting close to 20% during droughts [30]. The Atlantic forest presents the highest potential for invasion, where the predominant climate is the humid tropical climate, which is marked by medium to high temperatures and high air humidity throughout the year and regular and well distributed rainfall [31]. The characteristics of this biome favor the occupation of C. penicillata in these areas. Climatic and environmental factors may interfere in the marmoset distribution in the Brazilian territory, thus creating areas with higher or lower invasion probability. However, it is necessary to consider that results are limited to the data currently available on locations of occurrence of the species who are mainly from the southeastern region of Brazil.
Other primate studies have shown the existence of an influence between environmental variables and the distribution limits, as well as the use of space [see 25,32,33]. For instance, the distribution limits for Brachyteles arachnoides (É. Geoffroy, 1806) are influenced by climatic factors (AUC=0.994) such as temperature and precipitation [34]. The environmental variables that most influenced the B. arachnoides distribution were temperature seasonality (AUC=0.96), followed by annual temperature mean (AUC=0.93) and maximum temperature of the hottest month (AUC=0.93). As found in our study, temperature seasonality was the variable most correlated to distribution. For Callithrix flaviceps (Thomas 1903), study with model attained through logistic regression (with a 95.6% concordance value) showed that climatic factors seem to limit its distribution, suggesting that there are areas with higher probabilities (> 40%) of species occurrence [25]. The occurrence of C. flaviceps was positively related to relative humidity (0.8057, sd ± 0.0229), and it seemed to show a preference for Ombrophilous forest areas (more than 50% of the occurrence was in Ombrophilous Forests).
In order to understand the impact of C. penicillata invasion, we analyze its distribution area throughout the country along with the Protected Areas (PA, Conservation Units -CU in Brazil) ( Figure 3), Priority Conservation Areas and the Atlantic Forest domain itself ( Figure 4). When C. penicillata occurrence is overlaid with Brazilian Conservation Units, it is evident that the species has invaded Conservation Units belonging to the Atlantic Forest (Appendix I).  The black-tufted marmoset (C. penicillata) is a generalist invader [14], with a plastic diet, a high habitat occupation potential, and a capability of occurring in a widely variety of phytophysiognomies such as disturbed areas or secondary vegetation [36]. Additionally, Callithrix genus is also one of the most frequent in illegal trade, commonly commercialized as pet [37]. It has great potential on the predation of the native fauna (bird eggs and birds, amphibian and serpent hatchlings) [14,20,38]. Causing many problems in the Atlantic Forest areas where they were introduced [13], as direct competition with native primate species for habitat and resources [39] and hybridization (H) with the endemic Callithrix ssp. which may result in the loss of unique genotypes, endemism suppression and population depreciation [40], as well as the transmission of diseases to both native primates and human beings [41].
Primates of the Callithrix genus are reservoir for diseases that afflict primates, including humans, and are often potential transmitters. Marmosets are classified by the National Health Foundation (Fundação Nacional de Saúde -FUNASA) as a host species and/or a possible biological risk parasite reservoir, and therefore are monitored by the Surveillance and Control Coordination of Biological Risk Factors (Coordenação de Vigilância e Controle dos Fatores de Risco Biológico) in order to prevent and avoid any changes in key and conditional environmental factors related to human health [42]. Records dating back to 1930 describe diseases being transmitted from primates down to humans, with encephalomyelitis (from Herpesvirus simiae) being one of the first [43]. Still, diseases from fungal, viral, bacterial and helminthic origins are currently described as cycling between humans and other primates, such as rabies, herpes B, monkeypox, common cold, poliomyelitis, measles, yellow fever, dengue and others [42,43].
Dozens of marmosets have been diagnosed with rabies, and human deaths have been caused by marmoset-transmitted rabies [44]. The Callithrix genus was the second most stricken by yellow fever among non-human primates [45] in the recent outbreaks of wild yellow fever that hit the southeast of the country between 2017-2018, the area most susceptible to invasion, caused high mortality of marmosets. The death of these animals mainly in urban areas with high population and occurrence of the Aedes aegypti mosquito increased the concern of health agencies about the risk of reurbanization of the disease in the country [46,47]. Moreover, intestinal parasites may be transmitted by marmosets down to humans in urban areas frequented by marmosets and humans (such as parks or squares) [48].
Callithrix ssp. generally has the ability to survive in fragmented areas [5,10]. Disturbed environments, especially when close to urban areas, are susceptible to colonization by generalist primates as the Callithrix genus [5,9]. In southeast of the country, especially in the invaded areas, primates are living in environments near human settlements, they opportunistically interact with people aiming to supplement their diet [50]. This greater proximity then increases the risks of transmitting diseases for humans, and also increase the exposure of marmosets to parasites, risk attack by domestic animals, hunting for pets and susceptible to roadkill and electrocution when using power lines [19,30,31].

CONCLUSION
The Atlantic Forest is vulnerable to biological invasions since it is already under of pressure coming from urbanization, and degradation and environmental fragmentation. The marmoset Callithrix penicillata has become a successful invader of Atlantic Forest areas, causing depreciation in many native species, which already deal with various other pressures. Factors linked with urbanization and industrialization are harder to control since the human expansion process is not likely to be contained. However, biological invasions might be mitigated or even extinguished through successful interventions and management strategies when well applied can bring excellent results [see: 51, 52].
The generated model reached desirable reliability rates and may be used to help plan the control of C. penicillata invasions. Further studies are needed in order to design better control measurements. Our results provide data that may contribute to the conservation of the Atlantic Forest by helping and clarifying the potential biological invasion process by the C. penicillata, outlining the current invasion profile, based on environmental characteristics. Showing, therefore, which regions are more likely to be invaded, as well as which environmental conditions may contribute to or limit the invasion; data which may be used in future conservation and management projects.