Are tropical reptiles really declining? A six-year survey of snakes in a tropical coastal rainforest: role of prey and environment

Introduction: Even though snake declines seem to be a reality in many parts of the world, some reports are based on anecdotal evidence and there is a need of prolonged and intensive studies, especially in the tropics, for corroboration. Objective: To investigate if snake populations in Drake Bay are decreasing, and if there is a relationship with prey, time, temperature, rain and moonlight. Methods: We counted snakes seen per hour when walking along a single trail in the coastal forest of Drake Bay, Costa Rica. We walked the trail at night for a total of 842 nights (over 4 000 hours of observations), from 2012 through 2017 and recorded all the individual snakes we could see with head flashlights. We used ANOVA tests to check correlations among counts per hour with moonlight and rain; and graphic analysis for associations with diet, temperature, month and year. Results: We recorded 25 species (five families); which feed mostly on terrestrial vertebrates. Counts per hour have fallen over the years, especially for species that prey on amphibians and reptiles; Mastigodryas melanolomus has remained in similar numbers; and Siphlophis compressus has not been seen since May 2016. Temperature is relatively constant along the year in Drake, but month strongly affected the counts, which increased from August to September. Most species were seen more often in rainy nights (0.11 per hour versus only 0.03 per hour in nights without rain, ANOVA P < 0.05); we saw less Leptodeira septentrionalis on bright nights (0.12 per hour, versus 0.21 per hour in dark nights; ANOVA, P = 0.01), but all other species were unaffected by moonlight (ANOVA, P > 0.05). Conclusion: Night field counts of snakes in Drake Bay, Costa Rica, are not strongly affected by light or temperature, but are lower when there is no rain and show a strong decline from 2012 through 2017, particularly for species that feed on amphibians and reptiles. We have no reason to believe that the decline is an erroneous interpretation or that the snakes moved elsewhere, the decline of snakes in Drake seems to be real and needs attention from the conservation authorities.

Here we compare nighttime snake sightings with prey, rain, moonlight and temperature; along a six year period, in an intensively sampled tropical coastal rainforest trail.

MATERIALS AND METHODS
We counted snakes seen per hour while walking at night along a trail (our "transect"), repeatedly over a six year period (2012-2017), in lowland tropical rainforest, 12-38 m above sea level, near the Agujas River (Location and trail map in Fig. 1, Digital Appendix 1); the site is in Drake Bay, South Pacific of Costa Rica (N 08.69420 -08.69490 & W 083.67421 -083.67495). All counts and data recording were made by Stice and Gómez as part of their work as field guides, while accompanied by small groups of tourists, who did not participate directly in the study. Except when necessary to corroborate identification, we did not touch the animals.

Snake counts:
We walked the trail at night for a total of 842 nights (over 4 000 hours of field observations), using head flashlights to count all the snakes that we could see along the transect without disturbing the vegetation to look for them; identified species in situ using the guides by Savage (2002) and Solórzano (2004); and photographed them for taxonomic corroboration (representative photographs for all species seen are presented in Digital Appendix 2). We worked mostly from 1 930 to 2 200 hours, but a few counts started at 1 730 and ended at 2 245 hours (full details of sampling hours and dates are presented in Digital Appendix 2).
Snake diet: Prey type preference was taken from Savage (2002) and Solórzano (2004); categories are not mutually exclusive, and full details are presented in Table 1.
Annual and monthly data for rain and temperature: Rain and temperature data were kindly provided by the Instituto Meteorológico Nacional de Costa Rica from the nearest meteorological station at Rancho Quemado, partially covering the period 2012-2017. The institute reported that no rain records were available from January 2012 to July 2012, December 2016 to March 2017 and September 2017 to December 2017. No temperature records were available from January 2012 to July 2012 and from September 2017 to December 2017, so these periods were not included when calculating the mean monthly values that appear in Fig.  5 (Digital Appendix 1).
Note: Rancho Quemado, where the meteorological station is located, has an altitude of 240 m above sea level while the survey site in Drake Bay has an altitude of 12 to 38 m, so we expected a difference in temperature between the study site and the meteorological station. To assess this difference, we made temperature measurements directly in Drake Bay for seven months, from February 2017 to August 2017, and compared them with Rancho Quemado for that same time period. We found that average temperatures in Drake Bay are 2 to 3 °C higher than temperatures in Rancho Quemado, but that the trends along the timeline are the same. We did not have the funds to do this comparison in other years, but have no reason to doubt that the difference existed along the six years because lower sites are warmer that higher sites in a predictable value (Salter, Hobbs, Wheeler, Kostbade, & Trenton, 2005).

Rain:
We used both official rain records by Instituto Meteorológico Nacional de Costa Rica (i.e. government data), and our own classification of rain condition during fieldwork as "No Rain", "Light Rain" and "Heavy Rain", to make a double assessment of possible correlations with snake counts. We applied ANOVA tests for our classification and visual inspection for the government data (Full details in Fig. 5, and in Table 1 and Table 2, Digital Appendix 1).
Moonlight: Instead of using published moonlight data, which might not apply under the forest cover (cloudy skies produce dark nights even when the moon is full), we recorded moonlight conditions on every trip directly in the trail; we classified nights as "Dark" (practically no moonlight); "Bright" (good moonlight allowing us to follow the trail easily) and "Semibright" (intermediate between Dark and Bright).

Statistical analyses:
In the case of rain and moonlight, we analyzed "Snakes seen per hour" independently for each of the most common species, and pooled data for the rare species into a single category called "Others" so that counts were enough for ANOVA tests (Table 1 and Table 2 in Digital Appendix 1). We used visual inspection for comparisons among months and among years (full details and graphs in Digital Appendix 1, Fig. 2, Fig.  3, Fig. 4, Fig. 5, Fig. 6 and Fig. 7). Raw data sets are available on Zenodo (DOI: 10.5281/ zenodo.3531953).
Effect of prey: Most species at Drake feed on terrestrial vertebrates (Table 1) and, over the years, the number of species with each prey type remained relatively constant (Fig. 2 in Digital Appendix 1). However, the numbers of individuals seen per hour fell continuously, from 2012 through 2017, for snake species that feed on amphibians and reptiles (Fig. 3 in Digital Appendix 1).
Annual patterns: Our analysis ( Fig. 4 in Digital Appendix 1) indicates that, despite wide oscillations in the number of animals seen per hour, in recent years, encounters with Leptodeira septentrionalis, Imantodes cenchoa and Enuliophis sclateri are less common, M. melanolomus has remained in similar numbers; and Siphlophis compressus has not been seen after April 2016.
Monthly pattern: Visual inspection of Fig. 5 (Digital Appendix 1) shows a monthly pattern marked by a stable temperature (H = 59.80, P = 0.5), rain falling mostly from February through October (H = 70.07, P = 0.5), and snake counts increasing from August to September (H = 2.14, P = 0.5).

Effect of rain:
Statistical data for rain appear in detail in Table 1 and Fig. 6 (Digital Appendix 1). Encounters per hour with I. cenchoa were not affected by rain (ANOVA P > 0.05), but L. septentrionalis was seen more often in nights without rain (0.20 per hour, versus only 0.12 per hour in rainy nights; ANOVA P < 0.05). All the other species (pooled) had the opposite pattern: they were found more often in rainy nights (0,11 per hour versus only 0,03 per hour in nights without rain, ANOVA P < 0.05). The same applies to the abundant E. sclateri: 0.04 per hour in rainy nights versus only 0.01 with no rain (ANOVA P < 0.0003).

Effect of moonlight:
We saw less L. septentrionalis on bright nights (0.12 per hour, versus 0.21 per hour in dark nights; ANOVA, P = 0.01), but all other species were unaffected (ANOVA, P > 0.05; Table 2 and Fig. 7 in Digital Appendix 1).

DISCUSSION
We found that most of the studied snakes, feed primarily on terrestrial vertebrates, and were less seen as the years passed, this was especially true for amphibians and reptiles predators. This could be related with the reduction in the availability of some, but not of all, of their prey species, because some frogs, for example, remained abundant throughout the study period (Gianfranco Gómez, personal observation). Specialist snakes are highly vulnerable to prey decline because they are less likely to exploit alternative resources in response to shifting environmental conditions (Terborgh & Winter, 1980;Gaston, 1994), and, according to Webb and Shine (1998), snakes that ambush prey are particularly susceptible because of their low rate of food acquisition.
The stable temperature of this rainforest along the year, experiences most of its climatic variation due to rain falling, mainly from February through October. Snake counts increasing at the end of the rainy season, matches previous studies with tropical species. Basically, the rainy season allows the reproduction of amphibians, and their increased activity attracts the snakes that prey on them (Duellman, 1958;Shine & Madsen, 1996;Daltry, Ross, Thorpe, & Wüster, 1998;Luiselli & Akani, 2002;Oliveira & Martins, 2001;Morrison & Bolger, 2002). There are, however, reports of a different pattern, i.e., of no correlation between counts of Neotropical snakes and rainfall (Henderson & Hoevers, 1977;Martins, 1994;Bernarde & Abe, 2006), which only match our findings for I. cenchoa, and future researchers could evaluate if its arboreal behavior can explain its "independence" from rain. In our study, the fewer sightings of L. septentrionalis on nights with strong moonlight could be a behavior to avoid predation, like that of Crotalus viridis in Colorado (see Clarke, Chopko, & Mackessy, 1996), but it was the exception, we saw with the same frequency all the other species at Drake both in dark and illuminated nights. This is unusual, as many snakes increase their activity in full moon, when prey are more visible (Lillywhite & Brischoux, 2012;Connolly & Orrock, 2018), like the tropical tree snake Boiga irregularis that searches for forest spots where moonlight is stronger (Campbell, Mackessy, & Clarke, 2008). Like any study, this study has limitations, and the main one is the impossibility of detecting all the snakes in the forest. Another one is the fact that a reduction in the number of species seen along the trail does not mean that the snakes are dying, it has to be taken into account that maybe they just moved from the area for a variety of reasons that could include microclimatic variation and the search of prey elsewhere. We do not think the presence of humans could cause this, as the number of people walking the transect every night is low and it is limited to a brief period of the night. Despite changes in the leaf litter can have a large impact on invertebrates, and in turn this affects amphibians and reptiles that are part of the litter food web (Henderson, Dixon, & Soini, 1978;Martins, 1994), we did not notice differences in leaf litter during the study period (see Fig. 3, Digital Appendix 2).
Small environmental fluctuations can also affect the visibility of snakes (Seigel, Collins, & Novak, 1987;Zamora-Camacho, Moreno-Rueda, & Pleguezuelos, 2010;Rugiero, Milana, Petrozzi, Capula, & Luiselli, 2013), and of course there is a possibility that snakes hide, or leave the trail altogether. However, we have no reason to believe that our ability to see the snakes has decreased over the years, and a few species remained constant in our counts along the study. We also do not think they moved to less disturbed habitat because of human alteration of habitats around the reserve as some species have reappeared in subsequent years. We believe that the decline of snakes in Drake is real and needs attention from the conservation authorities.
Ethical statement: authors declare that they all agree with this publication and made significant contributions; that there is no conflict of interest of any kind; and that we followed all pertinent ethical and legal procedures and requirements. All financial sources are fully and clearly stated in the acknowledgements section. A signed document has been filed in the journal archives.