Potential effect of habitat disturbance on reproduction of the critically endangered harlequin frog Atelopus varius in Las Tablas, Costa Rica

Potential effect of habitat disturbance on reproduction of the critically endangered harlequin frog Atelopus varius in Las Tablas, Costa Rica . We studied a population of Atelopus varius in Las Tablas Protected Zone in southwest Costa Rica, where we estimated occupancy rates of tadpoles along the Cotón River. In addition, we report the first tadpoles observed in the wild in 20 years. Tadpole rate of occupancy was greater in habitat containing native forest than in disturbed areas bordering cattle pasture. This same pattern was also reflected in adult hotspots, where encounter rates were higher for adults in habitat surrounded by forest versus pasture. We present evidence for the potential effect of habitat modification on the presence and reproduction of A. varius and suspect that over time this modification impacts the species' demography. However, further study is necessary before we can confirm that habitat change alone was the key factor involved in patterns of decline for the species.


Introduction
The variable harlequin frog Atelopus varius is categorized as Critically Endangered on the IUCN Red List of Threatened Species following a drastic population decline since the 1980s (Pounds et al., 2010). Despite this decline, Atelopus varius is one of the few harlequin frog species remaining in Central America. Reproductive populations have recently been recorded in Costa Rica and Panama (González-Maya et al., 2013;Perez et al., 2014;Barrio-Amorós and Abarca, 2016), providing the opportunity to carry out research and conservation efforts to better understand the factors leading to the decline and how to address them.
Although A. varius is the best-known species within the genus (Lötters, 1996), information about current threats and their effect is scarce. Ongoing factors impacting the species include chytrid fungus, invasive species, and habitat loss and degradation (Richards-Zawacki, 2009;Pounds et al., 2010;Perez et al., 2014;González-Maya et al., 2018). Chytrid fungus is the most well-known and persistent threat affecting the species (Brem and Lips, 2008), and the Atelopus genus, while other described threats both directly and indirectly impact the species (Pounds et al., 2010).
Habitat disturbance is the main driver of amphibian population decline and extinction worldwide (Blaustein et al., 2011). However, the implications of habitat disturbance on population are unknown for many amphibians (Cushman, 2006), including the genus Atelopus. Understanding the effects of this threat factor is essential to develop and prioritize management actions to avoid the extinction of endangered species.
Because the species is restricted to a very limited area along riverbanks, riparian habitat disturbance may have a negative effect on many aspects of the ecology of A. varius (Richards-Zawacki, 2009). Unfortunately, the effect at the population level is unknown. Here, we show evidence of the impact of habitat disturbance on the reproduction of A. varius along the Cotón River, Las Tablas Protected Zone, Costa Rica. We also present the first records of harlequin frog tadpoles in the wild in at least 20 years.

Material and methods
Our study site is located at Las Tablas Protected Zone (LTPZ), Puntarenas, Costa Rica. LTPZ is in southeastern Costa Rica, specifically along the Pacific flanks of the Talamanca mountain range (8.93º N and 82.82º W). LTPZ spans for over 19,000 ha and is part of La Amistad Biosphere Reserve. Our specific study site is situated in and along the Cotón River ( fig. 1), at 1,300 m a.s.l., with a mean annual precipitation of 3,500 mm and a mean annual temperature of 19 ºC (González-Maya and Mata-Lorenzen, 2008).
We conducted surveys along a 2.2 km section of the Cotón River, in LTPZ ( fig. 1), along both river banks. Two experienced researchers (each with at least 1.5 years surveying the population) performed the surveys. One researcher focused on searching for adults and the other focused the search on tadpoles.
The same observers performed all the surveys during the study to minimize detection bias. Mean days surveyed per month were 7.4 (± 1.52) with a mean effort of 4.24 hours/day/person (± 1.61). The surveys were carried out monthly from October 2016 to April 2017, corresponding to the reproductive season of A. varius. The only break in surveys occurred following abnormally high water levels during November 2016 which prevented access to the study area due to hazardous conditions. All surveys consisted of visual encounter surveys (VES) along the river banks; for adults this entailed surveying at least 3 m from the water-line into the forest and for tadpoles we searched over and under rocks and in pools along the water's edge. Tadpoles of A. varius were identified by their exclusive ventral sucking disc and distinguishing chromatophores, which differ from larvae of other sympatric anuran species, as well as the depressed-flattened body and massive proximal caudal musculature (Lötters, 1996). Tadpole records include one or more individuals which were assigned to Gosner's stages (1960) per month through subaquatic photography. Each adult individual detected was captured and measured. We photographed each adult's ventral and dorsal surfaces to 1) identify individuals and to create a capture-recapture histories database, 2) define movement of individuals and 3) prevent pseudoreplication. Only adult detections between October and December 2016 were used in the analysis, because this is time period when aggregation for reproduction occurs. Uncertainty about the identification between males and females did not allow discrimination by sex categories in all cases; only females with snout to vent length (SVL) > 33 mm were identified.
Geographic location was recorded for each individual or tadpole and was georeferenced using QGIS software (QGIS Development Team, 2016). Each location was classified into one of two vegetation categories based on satellite imagery and verified using field observations, according to the composition and dominant vegetation elements: disturbed forest (perturbed) zone and late successional forest zone ( fig. 1). The perturbed zone is dominated by pastures and scrublands with a few dispersed natural vegetation patches and crops. The late successional forest zone is dominated by natural vegetation that suffered selective wood extraction over 15 years ago and which is currently recovering.
Capture-mark-recapture data were used to estimate the mean distance that A. varius individuals moved. We used the Points2One application (available at: https://plugins.qgis.org/plugins/points2one/) to calculate distance between capture and recapture locations for each individual. Mean distance covered by A. varius individuals was 76 m, so we established sub-transects of 100 m to divide the 2.2 km stretch we surveyed into 22 segments. The 100 m segments were used to identify aggregations of individuals of greater magnitude than that expected for random events. These aggregations were identified with the Getis-Ord Gi* statistic (Ord and Getis, 1995) return the z-score (GiZ) of each segment and 5 % significance, a GiZ greater than 1.96 indicates a hotspot. We used the fixed distance option of 200 m (two segments) to identify hotspots. This method works by evaluating each feature within the context of neighboring features, thus identifying the degree of spatial clustering. We used the Getis-Ord Gi* statistic because we assumed that the spatial distribution of harlequin frog (adult or tadpole) hotspots are spatially associated during the reproductive season since adults aggregate during these months (Lötters, 1996;Savage, 2002).
We also estimated the occupancy rate of tadpoles along the river. The occupancy was modelled through Unmarked package (Fiske and Chandler, 2011) for the R language (R Core Team, 2016). We tested the null model (constant detection and occupancy) against detection (p) and occupancy (Psi) depending on vegetation cover zone (perturbed or late successional forest areas). The best-fitting model was selected based on Akaike's Information Criterion (AICc), which was adjusted for small sample size and derived measures (Burnham and Anderson, 2002). Model selection was performed with AICcmodavg package (Mazerolle, 2016).

Results
We identified 87 adult individuals during the reproductive season (October-December 2016), where at least nine were females and three of them were in amplexus. Furthermore, 33 records of tadpoles with 195 individuals in total were taken ( fig. 2). We detected between 1 and 55 individuals per record (mean: 5.9 ± 10.1). The tadpoles were between stages 26 in January 2016 and 39 in April 2016, Gosner's stages (1960).
Despite equal survey effort, tadpoles were recorded almost exclusively in the late successional forest zone. We identified two hotspots for tadpole records using the defined segments of the Cotón River. The lengths of these hotspot segments were 200 and 400 m with GiZ value > 2.19 ( fig. 3A). On the other hand, only 22 % of the adults were recorded in the perturbed zone. One adult hotspot was identified, and it was exclusively within the late successional forest zone. This hotspot was 400 m long with GiZ value > 2.16 ( fig. 3B). Adults and tadpoles hotspots hadan overlap of 293 m (50.56 % and 75.85 % of tadpole and adults hotspots, respectively).
The occupancy model supports the hypothesis that habitat disturbance has a negative effect on A. varius

Discussion
Despite A. varius being the most studied species within the genus Atelopus (Lötters, 1996), ecological aspects are still almost entirely unknown. After the dramatic population decline and local extinctions in Costa Rica and Panama in the 1980s, remnant populations such as those studied here are an important opportunity for research, especially on demography and threat impacts (Muths et al., 2011). Herein, we present the first records of A. varius tadpoles in the wild in at least 20 years (Lötters, 1996;La Marca et al., 2005;Lötters pers. comm.). It is important to consider that almost all tadpoles were restricted to the forest zone of the Cotón River, where sun exposure is less due to continuous forest canopy cover. Hotspots of adult individuals were also restricted to the forest zone, and tadpole occupancy was higher in the forest than in the perturbed zone. This spatial pattern is the first evidence indicating that habitat disturbance affects reproduction and could have a strong impact on demographic parameters in A. varius. However, to avoid premature generalizations we recommend a multi-year study in order to confirm that the spatial pattern is explained by habitat disturbance as a driver of reproduction site selection in A. varius and not by some other factor.
To date, habitat disturbance (specifically habitat loss) has not been a factor considered in the decline  and extinction of the Atelopus genus (La Marca et al., 2005). On the other hand, threat factors such as chytrid fungus have been well documented and have been associated with amphibian extinction and population decline processes (Ryan et al., 2008;Crawford et al., 2010), and more specifically for Atelopus species (Brem and Lips, 2008;La Marca et al., 2005). Although there is no doubt that habitat disturbance has been a factor involved in amphibian decline and extinction globally (Blaustein et al., 2011), including other Atelopus species (La Marca et al., 2005), we demonstrate a potential effect of habitat disturbance over A. varius reproduction and we suspect this disturbance also impacts population demography. However, we cannot confirm that habitat disturbance alone was the key factor involved in decline patterns for the species, although we suspect that it is a contributing factor.
Species' disturbance tolerance is widely used as an indicator of ecosystem conservation status (Segurado et al., 2011) and it is also used during routine assessment of species' conservation status and categorization on the IUCN Red List of Threatened Species (IUCN, 2016). However, disturbance tolerance of A. varius is based on expert inference, with little empirical evidence (Segurado et al., 2011). Expert opinion is generally based on anecdotic and limited records, although in the absence of information it is an important resource when inferring impacts of known threats. However, it can be easily misapplied from one species to the next in the absence of specific data on occupancy during different life stages, specifically on breeding populations.
This new information stands in contrast to inference sometimes made unknowingly by some experts. Therefore, we feel it is prudent and necessary to provide evidence that the presence of Atelopus varius individuals (especially adults) in a perturbed zone is not proof of a population' tolerance to disturbance, nor does it indicate that a population is breeding or thriving in disturbed habitat. Evidence from this study suggests that successful reproduction occurs almost exclusively in the forest zone (i.e. native habitat). We urge biology researchers and conservation scientists to avoid speculation about the effects of habitat disturbance on species, especially those on the brink of extinction. Instead we need to increase ecological research efforts for remnant populations of threatened species, including A. varius, with the goal of conducting appropriate conservation and management actions to avoid their extinction.