See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/345322824

Species distribution models predict the geographic expansion of an enzootic

amphibian pathogen

Article in Biotropica · October 2020

DOI: 10.1111/btp.12863

CITATIONS READS

8 170

3 authors:

Héctor Zumbado-Ulate Adrián García-Rodríguez

University of California, Riverside University of Vienna
25 PUBLICATIONS 364 CITATIONS 52 PUBLICATIONS 484 CITATIONS

SEE PROFILE SEE PROFILE

Catherine Laura Searle

Purdue University
66 PUBLICATIONS 1,576 CITATIONS

SEE PROFILE

All content following this page was uploaded by Héctor Zumbado-Ulate on 06 December 2020.

The user has requested enhancement of the downloaded file.

Received: 15 April 2020 | Revised: 15 July 2020 | Accepted: 22 July 2020

DOI: 10.1111/btp.12863

ORIGINAL ARTICLE

Species distribution models predict the geographic expansion

of an enzootic amphibian pathogen

Héctor Zumbado-Ulate1 | Adrián García-Rodríguez2,3 | Catherine L. Searle1

1
Department of Biological Sciences, Purdue
University, West Lafayette, IN, USA Abstract
2
Departamento de Zoología, Instituto de Globally, numerous amphibian species have declined due to the introduction of the
Biología, Universidad Nacional Autónoma de
chytrid fungus Batrachochytrium dendrobatidis (Bd). However, the understanding of
México, UNAM, Ciudad de México, México
3
Museo de Zoología, Escuela de Biología,
the spatiotemporal dynamics remains incomplete. Therefore, estimating the current
Universidad de Costa Rica, San José, Costa geographic distribution of Bd is urgently needed, especially in countries like Costa
Rica
Rica, where susceptible species are still recovering from Bd-driven declines. We
Correspondence conducted model tuning and spatial analysis to compare the habitat suitability for
Héctor Zumbado-Ulate, Department of
Biological Sciences, Purdue University, West
epizootic and enzootic Bd in Costa Rica and to identify data-deficient regions, oppor-
Lafayette, IN 47907, USA. tunistic sampling, and Bd hotspots. Our dataset combined two methods of detection
Email: hzumbado@purdue.edu
(histology and PCR methods) for a total of 451 Bd-positive records from 34 localities.
Funding information We found that the distribution of enzootic Bd in Costa Rica increased 60% since
Dirección General de Asuntos del Personal
Académico, Universidad Nacional Autónoma
previous estimates in the early 2000s and extended to highlands and dry lowlands
de México; Purdue University that were considered unsuitable for Bd. We also found that Bd is common across

Associate Editor: Ferry Slik

protected lands (80%) and within the herpetological provinces containing the high-
Handling Editor: Amanda Zellmer est amphibian richness and endemism in Costa Rica. Opportunistic sampling of Bd
has focused on sites where epizootics occurred with the strongest intensity, leading
to deficient or absent sampling across the Talamanca Range, the Nicoya Peninsula,
and the northern lowlands. Our results showed that PCR increased the power of
Bd detection in lowlands and favored the identification of Bd hotspots across the
Caribbean side of Costa Rica. Our results add to the understanding of disease spread
during enzootics and can be used to identify new hotspots for disease to mitigate
future outbreaks of this pathogen.
Abstract in Spanish is available with online material

KEYWORDS

Batrachochytrium dendrobatidis, Chytridiomycosis, disease, habitat suitability, histology, host,

kriging, qPCR

1 | I NTRO D U C TI O N a sixth mass extinction (Barnosky et al., 2011) driven by a variety of

anthropogenic actions such as habitat destruction, spread of patho-
Global biodiversity has steadily declined over the last 50 years, gens, introduction of non-native species, illegal harvest, and climate
and extinction risk at the species level has increased (Hooper change (Dirzo & Raven, 2003). Within vertebrates, amphibians are at
et al., 2012). Current extinction rates suggest we are approaching a higher risk of extinction than any other taxon (Monastersky, 2014;

© 2020 The Association for Tropical Biology and Conservation

Biotropica. 2020;00:1–11.  wileyonlinelibrary.com/journal/btp | 1

2 | ZUMBADO-ULATE et al.

Wake & Vredenburg, 2008). More than 40% of amphibian species of a highly virulent strain of Bd in Costa Rica (De León et al., 2019).
have declined or gone extinct (Stuart et al., 2004) due to multi- The rapid spread of this strain during the 1980s and 1990s through-
ple environmental threats (Collins, 2010; Daszak, Cunningham, & out the country caused deadly outbreaks of chytridiomycosis
Hyatt, 2003). Therefore, the identification and quantification of (i.e., epizootics, Briggs, Knapp, & Vredenburg, 2010), especially in
these threats are needed to reconstruct the history of species de- stream-dwelling amphibian species from mid- and high-elevation
clines and establish effective future management (Adams, Pessier, & ecosystems (1,000–2,500 m elevation; Bolaños, 2009; Lips, Green,
Briggs, 2017; Meredith, Van Buren, & Antwis, 2016). Green, & Papendick, 2003; Pounds et al., 2006). However, recent
With only 51,100 km2, Costa Rica is home of 215 species of studies have found low mortality and endemic Bd infection (i.e., en-
amphibians, which represents nearly 2.7% of globally described zootics, Briggs et al., 2010) across amphibians assemblages in all el-
amphibian species (Zumbado-Ulate, Nelson, et al., 2019). The accu- evations (Whitfield et al., 2017; Zumbado-Ulate, García-Rodríguez,
rate description of amphibian taxa in Costa Rica, as well as its ef- García-Rodríguez, Vredenburg, & Searle, 2019; Zumbado-Ulate,
fective monitoring and conservation has been facilitated with the Nelson, et al., 2019). Additionally, several declined species have
creation of conservation areas. Costa Rica possess a robust sys- been rediscovered since 2005 and may be recovering from epizo-
tem of protected areas (public and private), which safeguards 26% otics (Abarca, Chaves, García-Rodríguez, & Vargas, 2010; Chaves
of the country and favors the mitigation of anthropogenic threats et al., 2014; Gómez-Hoyos et al., 2018; Jiménez & Alvarado, 2017).
(Andam, Ferraro, Pfaff, Sanchez-Azofeifa, & Robalino, 2008; Lambin Therefore, information on the current distribution of Bd is urgently
& Meyfroidt, 2011; Langholz & Lassoie, 2001). Similarly, the estab- needed to prevent future outbreaks of disease, especially in regions
lishment of Savage's herpetological provinces in Costa Rica facili- where Bd is endemic and coexists with highly susceptible species
tated the study and conservation of amphibians (Sasa, Chaves, & (García-Rodríguez, Chaves, Benavides-Varela, & Puschendorf, 2012;
Porras, 2010; Savage, 2002). The Savage's herpetological provinces Zumbado-Ulate, Nelson, et al., 2019).
system divides Costa Rica into five units that represent a unique Central America has a combination of complex topography and
combination of life zones and herpetological taxa, therefore facili- climatic variability (Lieberman, Lieberman, Peralta, & Hartshorn,
tating specific conservation actions (Zumbado-Ulate, Nelson, et al., 1996; Powers, Becknell, Irving, & Pèrez-Aviles, 2009) causing varia-
2019). Despite the application of these conservation efforts, nu- tion in expected suitable localities for Bd occurrence even at similar
merous amphibian population declines (including enigmatic extinc- elevations (Puschendorf et al., 2009). Therefore, sensitive methods
tions) have occurred in pristine and protected sites of Costa Rica of detection are needed to accurately identify current hotspots
(e.g., Bolaños, 2009; Lips, Reeve, Reeve, & Witters, 2003; Pounds for disease and unsuitable regions for infection. Detection of Bd
& Crump, 1994). Thus, identifying environmental threats across has been conducted through histology, standard polymerase chain
protected lands and herpetological provinces and quantifying their reaction (PCR), and quantitative-PCR (qPCR; Annis, Dastoor, Ziel,
impacts on amphibian communities is crucial to propose specific Daszak, & Longcore, 2004; Berger, Speare, & Kent, 2000; Boyle,
conservation actions at the regional level (Nori et al., 2015). Boyle, Olsen, Morgan, & Hyatt, 2004). The choice of method of
A common threat to global amphibian populations is the in- detection has relied mainly on the type of study and sample. For
troduction of the chytrid fungus Batrachochytrium dendrobatidis example, histology is mostly used to assess the severity of damage
(hereafter Bd; Longcore, Pessier, & Nichols, 1999), which causes on amphibian skin caused by chytridiomycosis or to detect Bd on
chytridiomycosis, a deadly skin disease in some species (Berger samples where DNA degradation has occurred (Skerratt et al., 2011;
et al., 1998; Voyles et al., 2009). A recent assessment estimated Wandeler, Hoeck, & Keller, 2007). PCR detection is preferred when
that the global decline of at least 500 amphibian species is linked processing a large number of samples (Kriger, Hines, Hyatt, Boyle,
to Bd (Scheele et al., 2019). However, this number could be higher & Hero, 2006), or detecting Bd in regions where prevalence is low
because the status of many species remains unknown (e.g., species and destructive sampling of amphibian hosts is not possible or de-
that inhabit remote localities or exhibit a secretive biology). Similarly, sirable (Skerratt et al., 2011). Among PCR techniques, qPCR (Boyle
the distribution of Bd in several countries remains insufficiently de- et al., 2004) has become a popular sampling method to detect Bd
scribed due to the lack of continuous studies on regional dynamics from live animals in the field (e.g., Whitfield et al., 2017), museum
of Bd; therefore, the predicted distribution often relies in large- specimens (Cheng, Rovito, Wake, & Vredenburg, 2011), and the en-
scale continental models (James et al., 2015; Ron, 2005). Because vironment (Kirshtein, Anderson, Wood, Longcore, & Voytek, 2007).
every country uses different environmental policies and conserva- Therefore, building occurrence datasets based on the combination
tion strategies, updated distributions of Bd at the country level are of different methods of detection provides a solid framework for ac-
needed to accurately identify potential areas for future outbreaks curately estimating a pathogen's distribution across heterogeneous
and amphibian population recovery (Rohr, Halstead, & Raffel, 2011). environments (Chestnut et al., 2014; Skerratt et al., 2011).
Understanding of the spatiotemporal dynamics of Bd remains Species distribution models (SDMs) are powerful tools in con-
incomplete, even in well-studied countries like Costa Rica (De León servation. One common application of SDMs is to predict suitable
et al., 2019; Whitfield, Lips, & Donnelly, 2016). Although Bd has ex- areas for disease (Peterson, Bauer, & Mills, 2004; Pinkard, Kriticos,
isted in Costa Rica since at least the 1960s, optimal environmental Wardlaw, Carnegie, & Leriche, 2010). At the continental level, two
conditions during the middle 1980s may have favored the emergence suitability maps have been often used to predict the distribution of
ZUMBADO-ULATE et al. | 3

Bd across the Americas (James et al., 2015; Ron, 2005); however, 2.2 | Detection dataset
the large scale of these SDMs makes predictions of Bd distribution
difficult at the country level. To address this issue in Costa Rica, We build a detection-only dataset (hereafter “combined dataset”)
Puschendorf et al. (2009) built a Bd suitability map conditional on host based exclusively on detection on amphibian hosts using two meth-
presence with the maximum entropy (MaxEnt) algorithm (Phillips, ods of detection: histology and PCR. First, we reviewed literature
Anderson, & Schapire, 2006), a presence-background approach that (Goldberg, Hawley, & Waits, 2009; Picco & Collins, 2007; Saenz,
uses information from climatic conditions on detection-only data- Adams, Pierce, & Laurencio, 2009; Whitfield et al., 2017; Zumbado-
sets. This SDM predicted suitable areas for Bd during epizootics Ulate, Bolaños, Gutiérrez-Espeleta, & Puschendorf, 2014; Zumbado-
(e.g., mountain ranges between 1,000 and 2,500 m) and identified Ulate, García-Rodríguez, et al., 2019; Zumbado-Ulate, Nelson, et al.,
low suitability regions in several mid- to low-elevations. However, 2019; Table S2) and found 401 records of Bd in 13 localities of Costa
this model may not accurately reflect the enzootic distribution of Rica detected through standard PCR or qPCR. Then, we added these
Bd in Costa Rica because most detection trials were conducted on Bd-positive records to the histology-only dataset (50 Bd-positive
museum specimens collected during the 1980s and 1990s, when samples from 21 localities), which was used to produce the first
Bd was epizootic, and histology was the only method used for Bd suitability map for Bd in Costa Rica (Puschendorf et al., 2009) and
detection. Additionally, the suitability map lacked tuning methods, includes all the histology-detection data available for this country.
which generate more accurate modeling of geographic distributions Overall, the combined dataset consisted of 451 Bd-positive records
(Phillips & Dudík, 2008; Radosavljevic & Anderson, 2014). Therefore, from 34 localities (Figure S1) and evaluated 13 amphibian families
improving upon existing SDMs with the use of updated and robust and 90 species, plus five individuals that were identified at the genus
detection datasets along with increased rigor for model building and level. Given that PCR methods are more sensitive to detect Bd than
evaluation will improve the accuracy of predicting a pathogen's oc- histology (Kriger et al., 2006), we compared the abiotic range of de-
currence in a given region (Merow, Smith, & Silander, 2013; Rödder, tection of both methods in Costa Rica and found that histology could
Veith, & Lötters, 2008; Warren & Seifert, 2011). potentially lead to underestimation of the epizootic distribution, but
We generated an updated habitat suitability map for Bd in Costa only in the tropical dry forest (Figure S2).
Rica using a comprehensive dataset that combined two methods
of detection: histology and PCR. Our main goal was to describe
the current distribution of Bd and generate accurate predictions of 2.3 | Species distribution models
Bd occurrence in latitudinal and altitudinal regions. We also aimed
to estimate the occurrence of Bd across protected areas and her- We aimed to generated suitability maps to compare the endemic ge-
petological provinces in Costa Rica, and to identify data-deficient ographic distribution of Bd in Costa Rica (i.e., enzootic distribution,
regions, opportunistic sampling, and Bd hotspots. We hypothesize Zumbado-Ulate, Nelson, et al., 2019) with the distribution of the epi-
that (a) Bd increased its spatial distribution after epizootics, even zootic strain of Bd that emerged in the 1980s (De León et al., 2019).
reaching regions previously considered as unsuitable; (b) Bd is To generate the enzootic distribution, we used the combined dataset
highly common in protected areas and hotspots occur within the described above. To build the epizootic distribution, we used the his-
herpetological provinces with the highest amphibian richness; (c) tology dataset and rebuilt the first suitability maps for Bd in Costa
and opportunistic sampling is linked to regions where Bd caused Rica (Puschendorf et al., 2009) with tuned settings instead of default
declines during epizootics. settings (see below). Since these two suitability maps only differed in
the inputted Bd-presence data, we were able to directly determine
the effects of timing (epizootic vs. enzootic) and detection method
2 | M E TH O DS (histology only vs. histology + PCR) on Bd distribution.
Suitability maps were calibrated with the MaxEnt algorithm
2.1 | Climatic data and analyses (Phillips et al., 2006) in the R package “dismo” (Hijmans, Phillips,
Leathwick, & Elith, 2013). Since MaxEnt accounts for collinearity
To describe the abiotic conditions across Costa Rica, we down- and correlation between variables (Elith & Leathwick, 2009), we
loaded the 19 bioclimatic layers from WorldClim v1.4 (Table S1) at a included all the 19 bioclimatic layers from WorldClim v1.4 (see
spatial resolution of 30 arc-s (https://www.world​clim.org; Hijmans, Section 2.1) as predictors for model building. We conducted model
Cameron, Parra, Jones, & Jarvis, 2005). For data extraction and tuning (Phillips & Dudík, 2008; Radosavljevic & Anderson, 2014)
model construction, we calibrated all models to the extent of Costa with the R package “ENMeval” (Muscarella et al., 2014). For
Rica by cropping the 19 bioclimatic layers with a defined bounding data partition, we used the block method (Radosavljevic &
box (7.750–11.500 N, 82.150–86.250 W). Our maps and geographic Anderson, 2014). The block method traces latitude and longitude
®
analyses were created with ArcGIS 10.7 (ESRI ) and shape files from lines to split the occurrence localities into four bins of equal num-
the Atlas Digital Costa Rica 2014 (Instituto Tecnológico de Costa ber of occurrences and then iteratively uses k − 1 bins (k = 4) for
Rica, 2014). We used the World Geodetic System datum (WGS84) as model training and the withheld bin for testing, resulting in four
the coordinate reference system (CRS). runs for each combination of settings (Muscarella et al., 2014). In
4 | ZUMBADO-ULATE et al.

total, we generated 72 candidate models by varying regularization (Oliver & Webster, 1990) to identify Bd hotspots (i.e., regions where
multiplier values (ranging from 1 to 5 with increments of 0.5) and high prevalence of Bd is independent of opportunistic sampling). For
using different combinations of feature classes (L, Q, H, LQ, LH, this, we split the map of Costa Rica in nine equal regions using a fish-
LQH, LQHP, LQHPT, where L = linear, Q = quadratic, H = hinge, net map (cell size width = 122,500, cell size height = 117,500) and
P = product, and T = threshold). calculated nine regional values of Bd prevalence in Costa Rica. Then,
For model selection, we considered three metrics in the fol- we created a kriging predictive surface using a spherical semi-vario-
lowing priority order: (a) mean area under the curve (mean AUC), gram. Our predictions were tested with a zero-inflated linear model
(b) AUC difference, and (c) minimum training presence omission using region, sample size, and method of detection as predictors for
rate (mtpOR). High values of mean AUC reflect a better ability for a the Poisson regression component, and sample size and method
model to discriminate between conditions at withheld (testing) oc- of detection as predictors for the logistic regression component.
currence localities and those of background localities (Radosavljevic To evaluate the performance of our zero-inflated model, we ran a
& Anderson, 2014). If two models showed an identical mean AUC, Poisson regression using the same predictors and compared both
we reviewed the AUC difference between training and test set, models using a corrected Vuong test (Desmarais & Harden, 2013)
where best fitted models show lower values than over-fitted mod- with the R package “pscl” (Jackman et al., 2017).
els (Warren & Seifert, 2011). If that did not differentiate models, we
checked the mtpOR, that indicates the proportion of test localities
with suitability values lower than that associated with the low- 3 | R E S U LT S
est-ranking training locality. Values greater than the expectation of
zero typically indicate model overfitting (Tables S3 and S4). 3.1 | Geographic distribution of Bd

When comparing both SDMs, we found that habitat suitability for

2.4 | Geographic distribution of Bd enzootic Bd increased compared to the epizootic period in highlands
above 2,500 m and lowlands (<500 m), especially on the Caribbean
With the output of both suitability maps, we generated binary side (Figure 1a). The enzootic distribution also increased across the
maps (polygons representing the epizootic and enzootic predicted Pacific side of Costa Rica, even predicting the occurrence of Bd in
distribution of Bd) using the 10th percentile presence threshold the tropical dry and semi-dry forests that were predicted as unsuit-
(Radosavljevic & Anderson, 2014). This rigorous criterion discards able during epizootics (Figure 1b). We estimated the geographic dis-
the localities with the lowest 10% of suitability values and considers tribution of enzootic Bd in 28,390 km2 (55.6% of Costa Rica), which
the rest to be “suitable habitat.” To estimate the relative distribution represents an increase of 61% in the total area of occurrence from
of suitable and unsuitable conditions across Costa Rica's altitudinal epizootics (Figure 2a). We also found that Bd is predicted to occur
gradient, we extracted the value of elevation (altitude layer from the in 80% (10,500 km2; Figure 2b) of the geographic area classified as
2
WorldClim dataset at the same resolution of our models ~1 km ) and protected areas (26% of continental Costa Rica). When comparing
the presence or absence value from our binary maps for each cell of the occurrence of Bd among herpetological provinces (Figure 2c,d),
the country. Then, we grouped all elevation values into six altitudinal we found that Bd occurred in all five herpetological provinces, with
belts of 500 m (from 0 to 3,000 m elevation) and a seventh belt in- the highest overlap in the provinces with the highest amphibian
cluding all elevations above 3,000 m. Finally, we estimated the num- species richness: the Talamanca Range (89 species, 99.2% overlap),
ber of cells predicted as suitable or unsuitable for each belt based the Central Volcanic Range (159 species, 91.2% overlap), and the
on results from each model. We quantified the expansion of Bd by Caribbean lowlands (102 species, 58.3% overlap). Conversely, the
comparing the geographic area of both epizootic and enzootic binary occurrence of Bd in the North Pacific province was estimated in only
maps. We also estimated the predicted occurrence of Bd across pro- 5.6%, which is also the driest province and holds the lowest amphib-
tected areas and herpetological regions in Costa Rica by quantifying ian richness (66 species).
the overlap between the enzootic distribution of Bd and polygons Climatic variables showed different contributions to Bd suit-
representing the protected areas of Costa Rica and Savage's herpe- ability, ranging from 0.0% to 41.5% (Table S1). The climate pre-
tological provinces. dictors that most contributed to enzootic distribution were mean
precipitation of the driest month (BIO14 = 41.5%), minimum
temperature of coldest month (BIO6 = 13.9%), precipitation of
2.5 | Data-deficient regions, opportunistic warmest quarter (BIO18 = 13.8%), mean temperature of wettest
sampling, and Bd hotspots quarter (BIO8 = 12.8%), and precipitation of the coldest quarter
(BIO19 = 10.1%). The contribution of these predictors in our up-
We quantified and mapped the sampling of Bd in Costa Rica to iden- dated suitability map suggests that habitat suitability for enzootic Bd
tify data-deficient regions (i.e., regions where sampling is missing or in Costa Rica is strongly associated with seasonal climatic interac-
scarce) and opportunistic sampling. We used kriging interpolation tions across the year that favor or constrain Bd occurrence.
ZUMBADO-ULATE et al. | 5

F I G U R E 1 Species distribution models for Batrachochytrium dendrobatidis (Bd) in Costa Rica. (a) The enzootic distribution shows high
suitability across all elevations. (b) The epizootic distribution shows high suitability in mid-elevations throughout the country. The scale
at the right of each map shows the average probability of occurrence. The bars under each map show suitability by elevation with black
indicating the predicted percentage of climatic suitability for Bd. Altitudinal bars display belts of 500 m except the seventh bar that shows
all elevations above 3,000 m (1 = 0–500 m; 2 = 500–1,000 m; 3 = 1,000–1,500 m, 4 = 1,500–2,000 m; 5 = 2,000–2,500 m; 6 = 2,500–
3,000 m; 7 = 3,000–3,820 m)

3.2 | Data-deficient regions, opportunistic Rica. Our findings suggest that the distribution of Bd has shifted as
sampling, and Bd hotspots an enzootic, expanding its range to approximately 60% of the coun-
try and across all altitudinal belts of Costa Rica, from sea level to
Our results show that numerous localities in Costa Rica remain un- 3,800 m elevation, including the dry and semi-dry forest lowlands on
sampled for Bd, or with a very low sample size. The more extensive the Pacific side that have been historically considered unsuitable for
data-deficient regions include most part of the Talamanca Range, Bd (Puschendorf et al., 2009). Alternatively, the observed increase
the Nicoya Peninsula, and the Northwestern lowlands (Figure 3a). in Bd distribution from epizootic to enzootic dynamics could be re-
We found that most Bd sampling efforts have been conducted in lated to increased detection of Bd through time. Thus, the epizootic
midlands and highlands of the Central Volcanic Range, the north- distribution might be underestimated because histology is less sensi-
ern side of the Caribbean lowlands, Santa Elena Peninsula in the tive than PCR in detecting Bd in areas with low prevalence/intensity
Northwest Pacific, and Southern Pacific lowlands (Figure 3a). We (Kriger et al., 2006). One potential application that can be applied
found that enzootic Bd hotspots occur on the Caribbean side of to validate changes in the distribution during enzootics consists
Costa Rica (Figures 3b and S3), specifically on the foothills of the on using host–pathogen Integral Projection Models (IPMs; Ellner &
Central Volcanic Range and the Talamanca Range (Z = 5.15, p < .001) Rees, 2006). IPMs have already been used on the amphibian-Bd sys-
and the Caribbean lowlands (Z = 7.95, p < .001). We identified that tem and may be used to predict changes in the distribution of path-
Bd detection in hotspots increased with sample size (Z = 11.23, ogen loads over the course of enzootics based on environmental
p < .001) and PCR methods (Z = 14.46, p < .001). Our more robust predictors (Wilber, Langwig, Kilpatrick, McCallum, & Briggs, 2016).
zero-inflated linear model performed better than a Poisson regres- Complementarily, the use of spatial analyses to compare the abiotic
sion built with the same predictors (Z = −2.28, p = .01). range of detection of methods of detection could help identify areas
where an approach may fail to detect a pathogen, potentially leading
to underestimates of the geographic distribution (Figure S2).
4 | D I S CU S S I O N The increased suitability for Bd in lowlands of Costa Rica may
explain the enigmatic declines experienced by several amphibian
4.1 | Geographic distribution of Bd species in regions that were thought to be unsuitable for Bd (below
1,000 m elevation; Bolaños, 2009; Chaves et al., 2014; Zumbado-
In this study, our main goal was to describe the enzootic distribution Ulate, Bolaños, Willink, & Soley-Guardia, 2011). The suitable eleva-
of Bd and generate accurate predictions of Bd occurrence in Costa tional range predicted for enzootic Bd in Costa Rica matched with
6 | ZUMBADO-ULATE et al.

F I G U R E 2 (a) Predictive binary map showing the expansion of enzootic Batrachochytrium dendrobatidis (Bd) in Costa Rica. Predicted
distribution of enzootic Bd includes both blue and yellow polygons; (b) Predicted occurrence of enzootic Bd across protected lands in Costa
Rica; (c) Amphibian species richness within each herpetological province (from Zumbado-Ulate, Nelson, et al., 2019); and (d) predicted
occurrence of enzootic Bd within each herpetological province. Symbology: PN—Pacific Northwest, and PS—Pacific Southwest, CL—
Caribbean Lowlands, MSCC—Central Volcanic Range, CT—Talamanca Range

the elevational range reported for enzootic Bd in Panama, Mexico, endemism in Costa Rica: the Talamanca Range (99.2% overlap) and
Colombia, and Chile (Bacigalupe et al., 2019; Bolom-Huet, Pineda, the Central Volcanic Range (91.2% overlap). A recent assessment
Díaz-Fleischer, Muñoz-Alonso, & Galindo-González, 2019; Flechas (Zumbado-Ulate, Nelson, et al., 2019) found that 67 of 105 am-
et al., 2017; Kilburn et al., 2010; Woodhams et al., 2008). The cli- phibian species screened for Bd in Costa Rica tested positive and
matic variable contributions in our model (Table S1) also suggested confirmed that enzootic Bd is currently infecting species that ex-
that habitat suitability for enzootic Bd in Costa Rica is associated hibit multiple life-history traits across all elevations. Fortunately,
with seasonal climatic interactions across the year that favor or con- the lack of Bd-driven die-offs detected since epizootics (i.e., be-
strain Bd occurrence. For example, low precipitation of the driest fore 2,000) suggests that current amphibian communities in
month may reduce the ability of Bd to spread during the dry season. Central America might be exhibiting shifts in host responses to
On the other hand, predictors associated with suitable tempera- combat enzootic Bd and recover from epizootic declines (Christie
tures and precipitation for amphibian aggregation would favor Bd & Searle, 2018; Voyles et al., 2018).
transmission. We found that numerous localities in Costa Rica remain
Our results predicted that enzootic Bd occurs in 80% of the unsampled for Bd, especially across the remote highlands of the
area covered by protected areas and all five herpetological prov- Talamanca Range, the Nicoya Peninsula, and the Northwestern
inces (Figure S2b–d). Bd distribution was predicted to be high in lowlands (Figure 3a). In many other localities, the sample size ana-
the herpetological provinces with the highest rate of amphibian lyzed is well below the threshold of 60, the minimum recommended
ZUMBADO-ULATE et al. | 7

25% on local amphibian communities (Whitfield et al., 2012;

Zumbado-Ulate, García-Rodríguez, et al., 2019). We recommend
the use of interpolation methods (e.g., kriging interpolation) to re-
duce the effect of opportunistic sampling and accurately identify
high priority areas such as hotspots, high-risk areas, and refuges
(e.g., Hernandez-Stefanoni & Ponce-Hernandez, 2006).

4.2 | Implications for conservation

Our study demonstrates that the use of comprehensive datasets

that combine methods of detection generate robust predictions of
the geographic distribution of pathogens. In our study, the use of
the combined dataset increased the power to detect hotspots and
cold spots (e.g., Puschendorf et al., 2011). Our findings suggest
that other suitability maps for Bd in the Americas (at the conti-
nental and regional level) that were generated datasets derived
from only one method of detection (e.g., Bacigalupe et al., 2019;
Bolom-Huet et al., 2019; James et al., 2015; Ron, 2005) could be
reanalyzed in follow-up studies with combined datasets and inter-
polation methods to revalidate high-risk areas, Bd hotspots and
cold spots.
Our study also suggests that pathogens may expand their
ranges after epizootics. Similar studies have modeled the distri-
bution of pathogens during the enzootic stage, for example, West
Nile virus vector mosquitoes (Larson, Degroot, Bartholomay, &
F I G U R E 3 Historical sampling of Batrachochytrium dendrobatidis Sugumaran, 2010), tick vectors of Lyme disease (Illoldi-Rangel
(Bd) in Costa Rica. (a) Dots show the 172 localities where Bd et al., 2012), and sarcoptic mange in red foxes (Carricondo-Sanchez,
has been surveyed in Costa Rica. The dot size is proportional to
Odden, Linnell, & Odden, 2017). However, shifts in the geographic
the sample size per locality. Small dots and regions lacking dots
distribution from epizootics to enzootics are not yet documented.
represent data-deficient regions where Bd sampling is poor or
missing. (b) Kriging surface predicts the prevalence of enzootic For Bd, several studies have addressed the transition between epi-
Bd in Costa Rica based on 451 positive records detected through zootics to enzootics but mostly in terms of prevalence and suscepti-
histology or polymerase chain reaction (PCR). The map scale shows bility (Briggs et al., 2010; Catenazzi et al., 2017; Kilburn et al., 2010).
that the predicted percentage of infection has the highest values Given that enzootic Bd seems to maintain high levels of virulence
on the Caribbean lowlands and intermediate values on the Central
in Central America (Voyles et al., 2018), we recommend the use of
Volcanic Range, the Talamanca Range, and Southern Costa Rica.
The color-code dots show the number of Bd positives in each of backcast modeling to identify changes in the distribution of patho-
the 34 localities for each method of detection gens from epizootics to enzootics.
In a different context, our datasets can be used to predict
how climate change and other anthropogenic threats may influ-
size (Skerrat et al., 2008). Our results show that most Bd sampling ence suitable habitat for Bd in Costa Rica under current and fu-
efforts have focused on midlands and highlands of the Central ture scenarios (Rohr et al., 2011; Rohr & Raffel, 2010; Xie, Olson,
Volcanic Range, likely because this region was the most affected & Blaustein, 2016). A comprehensive assessment of a pathogen's
by epizootics (Puschendorf et al., 2009). We also found oppor- distribution can lead to more effective conservation strategies
tunistic sampling on the western side of the Caribbean lowlands, based on specific regions and species (Heard et al., 2018; Langwig
where the existence of experimental biological stations has fa- et al., 2015).
vored the application of long-term studies on Bd dynamics (e.g.,
Whitfield et al., 2013; Whitfield, Kerby, Gentry, & Donnelly, 2012). AC K N OW L E D G E M E N T S
We identified Bd hotspots in the areas of highest amphibian rich- We thank the Department of Biological Sciences at Purdue
ness in Costa Rica: the Caribbean side and the eastern foothills on University for funding this study through the Lindsay Fellowship,
the Central Volcanic Range and the Talamanca Range (Figures 3b the Andrews Environmental Travel Grant, and the Purdue Climate
and S3). Our findings are supported by recent studies that found Change Research Center (PCCRC) travel grant. A.G–R. is currently
that environmental conditions for Bd on the Caribbean side of supported by a postdoctoral fellowship from Dirección General de
Costa Rica favor an average endemic percent infection of about Asuntos del Personal Académico (DGAPA) at Instituto de Biología,
8 | ZUMBADO-ULATE et al.

Universidad Nacional Autónoma de México. We also thank two Bacigalupe, L. D., Vásquez, I. A., Estay, S. A., Valenzuela-Sánchez, A.,
Alvarado-Rybak, M., Peñafiel-Ricaurte, A., … Soto-Azat, C. (2019).
anonymous reviewers whose comments helped improve and clarify
The amphibian-killing fungus in a biodiversity hotspot: Identifying
this manuscript. and validating high-risk areas and refugia. Ecosphere, 10(5), e02724.
https://doi.org/10.1002/ecs2.2724
C O N FL I C T O F I N T E R E S T Barnosky, A. D., Matzke, N., Tomiya, S., Wogan, G. O. U., Swartz, B.,
Quental, T. B., … Ferrer, E. A. (2011). Has the Earth’s sixth mass
The corresponding author confirms on behalf of all authors that
extinction already arrived? Nature, 471(7336), 51–57. https://doi.
there have been no involvements that might raise the question of org/10.1038/natur​e 09678
bias in the work reported or in the conclusions, implications, or opin- Berger, L., Speare, R., Daszak, P., Green, D. E., Cunningham, A. A.,
ions stated. Goggin, C. L., … Parkes, H. (1998). Chytridiomycosis causes am-
phibian mortality associated with population declines in the
rain forests of Australia and Central America. Proceedings of the
AU T H O R C O N T R I B U T I O N S
National Academy of Sciences, 95(15), 9031–9036. https://doi.
Conceptualization: HZU. Data curation: HZU, AGR, CLS. Formal org/10.1073/pnas.95.15.9031
analysis: HZU and AGR. Methodology: HZU, AGR, CLS. Project Berger, L., Speare, R., & Kent, A. (2000). Diagnosis of chytridiomyco-
administration: HZU and CLS. Supervision: HZU and CLS. Original sis in amphibians by histologic examination. Zoos’ Print Journal, 15,
184–190.
draft: HZU. Writing—Review and editing: HZU, AGR, CLS.
Bolaños, F. (2009). Situación de los anfibios de Costa Rica. Biocenosis,
22(1–2), 95–108.
C O N S E N T FO R P U B L I C AT I O N Bolom-Huet, R., Pineda, E., Díaz-Fleischer, F., Muñoz-Alonso, A. L., &
The enclosed work is all original research carried out by the authors Galindo-González, J. (2019). Known and estimated distribution in
Mexico of Batrachochytrium dendrobatidis, a pathogenic fungus of
and is not under consideration for publication in another journal or
amphibians. Biotropica, 51(5), 731–746. https://doi.org/10.1111/
book. Any research in the paper not carried out by the authors, and btp.12697
all sources of funding are fully acknowledged in the manuscript. All Boyle, D. G., Boyle, D. B., Olsen, V., Morgan, J. A. T., & Hyatt, A. D. (2004).
authors approve the contents of the manuscript and its submission Rapid quantitative detection of chytridiomycosis (Batrachochytrium
dendrobatidis) in amphibian samples using real-time Taqman PCR
to the journal.
assay. Diseases of Aquatic Organisms, 60(2), 141–148. https://doi.
org/10.3354/dao06​0141
DATA AVA I L A B I L I T Y S TAT E M E N T Briggs, C. J., Knapp, R. A., & Vredenburg, V. T. (2010). Enzootic and
The data that support the findings of this study are openly available epizootic dynamics of the chytrid fungal pathogen of amphibians.
Proceedings of the National Academy of Sciences, 107(21), 9695–9700.
in The Purdue University Research Repository (PURR) at http://doi.
https://doi.org/10.1073/pnas.09128​86107
org/10.4231/439S-DN43 (reference https://purr.purdue.edu/publi​ Carricondo-Sanchez, D., Odden, M., Linnell, J. D. C., & Odden, J.
catio​ns/3424). We have included additional data as online support- (2017). The range of the mange: Spatiotemporal patterns of sar-
ing information. coptic mange in red foxes (Vulpes vulpes) as revealed by camera
trapping. PLoS One, 12(4), e0176200. https://doi.org/10.1371/
journ​a l.pone.0176200
ORCID
Catenazzi, A., Swei, A., Finkle, J., Foreyt, E., Wyman, L., & Vredenburg,
Héctor Zumbado-Ulate https://orcid.org/0000-0001-7039-9967 V. T. (2017). Epizootic to enzootic transition of a fungal disease
Adrián García-Rodríguez https://orcid. in tropical Andean frogs: Are surviving species still suscepti-
org/0000-0002-6556-9439 ble? PLoS One, 12(10), e0186478. https://doi.org/10.1371/journ​
al.pone.0186478
Catherine L. Searle https://orcid.org/0000-0002-6607-2299
Chaves, G., Zumbado-Ulate, H., García-Rodríguez, A., Gómez, E.,
Vredenburg, V. T., & Ryan, M. J. (2014). Rediscovery of the critically
REFERENCES endangered streamside frog, Craugastor taurus (Craugastoridae).
Abarca, J., Chaves, G., García-Rodríguez, A., & Vargas, R. (2010). Costa Rica. Tropical Conservation Science, 7(4), 628–638. https://doi.
Reconsidering extinction: Rediscovery of Incilius holdridgei (Anura: org/10.1177/19400​82914​0 0700404
Bufonidae) in Costa Rica after 25 years. Herpetological Review, 41(2), Cheng, T. L., Rovito, S. M., Wake, D. B., & Vredenburg, V. T. (2011).
150. Coincident mass extirpation of neotropical amphibians with the
Adams, A. J., Pessier, A. P., & Briggs, C. J. (2017). Rapid extirpation of a emergence of the infectious fungal pathogen Batrachochytrium den-
North American frog coincides with an increase in fungal pathogen drobatidis. Proceedings of the National Academy of Sciences, 108(23),
prevalence: Historical analysis and implications for reintroduction. 9502–9507. https://doi.org/10.1073/pnas.11055​38108
Ecology and Evolution, 7(23), 10216–10232. https://doi.org/10.1002/ Chestnut, T., Anderson, C., Popa, R., Blaustein, A. R., Voytek, M., Olson,
ece3.3468 D. H., & Kirshtein, J. (2014). Heterogeneous occupancy and density
Andam, K. S., Ferraro, P. J., Pfaff, A., Sanchez-Azofeifa, G. A., & Robalino, estimates of the pathogenic fungus Batrachochytrium dendrobatidis
J. A. (2008). Measuring the effectiveness of protected area net- in waters of North America. PLoS One, 9(9), e106790. https://doi.
works in reducing deforestation. Proceedings of the National Academy org/10.1371/journ​al.pone.0106790
of Sciences, 105(42), 16089–16094. https://doi.org/10.1073/ Christie, M. R., & Searle, C. L. (2018). Evolutionary rescue in a host–
pnas.08004​37105 pathogen system results in coexistence not clearance. Evolutionary
Annis, S. L., Dastoor, F. P., Ziel, H., Daszak, P., & Longcore, J. E. (2004). Applications, 11(5), 681–693. https://doi.org/10.1111/eva.12568
A DNA-based assay identifies Batrachochytrium dendrobatidis in am- Collins, J. P. (2010). Amphibian decline and extinction: What we know
phibians. Journal of Wildlife Diseases, 40(3), 420–428. https://doi. and what we need to learn. Diseases of Aquatic Organisms, 92(2–3),
org/10.7589/0090-3558-40.3.420 93–99. https://doi.org/10.3354/dao02307
ZUMBADO-ULATE et al. | 9

Daszak, P., Cunningham, A. A., & Hyatt, A. D. (2003). Infectious disease Instituto Tecnológico de Costa Rica. (2014). Atlas de Costa Rica 2014. In
and amphibian population declines. Diversity and Distributions, 9(2), RepositorioTEC. . https://repos​itori​otec.tec.ac.crhtt​ps://repos​itori​
141–150. https://doi.org/10.1046/j.1472-4642.2003.00016.x otec.tec.ac.cr/handl​e/2238/6749.
De León, M. E., Zumbado-Ulate, H., García-Rodríguez, A., Alvarado, Jackman, S., Tahk, A., Zeileis, A., Maimone, C., Fearon, J., & Meers, Z.
G., Sulaeman, H., Bolaños, F., & Vredenburg, V. T. (2019). (2017). The pscl package. R Package Version 1.17-4. http://Cran.r-
Batrachochytrium dendrobatidis infection in amphibians predates first Proje​c t.Org
known epizootic in Costa Rica. PLoS One, 14(12), e0208969. https:// James, T. Y., Toledo, L. F., Rödder, D., da Silva Leite, D., Belasen, A. M.,
doi.org/10.1371/journ​al.pone.0208969 Betancourt-Román, C. M., … Longcore, J. E. (2015). Disentangling
Desmarais, B. A., & Harden, J. J. (2013). Testing for zero inflation in count host, pathogen, and environmental determinants of a recently
models: Bias correction for the Vuong test. The Stata Journal, 13(4), emerged wildlife disease: Lessons from the first 15 years of amphib-
810–835. https://doi.org/10.1177/15368​67X13​01300408 ian chytridiomycosis research. Ecology and Evolution, 5(18), 4079–
Dirzo, R., & Raven, P. H. (2003). Global state of biodiversity and loss. 4097. https://doi.org/10.1002/ece3.1672
Annual Review of Environment and Resources, 28(1), 137–167. https:// Jiménez, R., & Alvarado, G. (2017). Craugastor escoces (Anura:
doi.org/10.1146/annur​ev.energy.28.050302.105532 Craugastoridae) reappears after 30 years: Rediscovery of an “ex-
Elith, J., & Leathwick, J. R. (2009). Species distribution models: Ecological tinct” Neotropical frog. Amphibia-Reptilia, 38(2), 257–259. https://
explanation and prediction across space and time. Annual Review doi.org/10.1163/15685​381-00003102
of Ecology, Evolution, and Systematics, 40(1), 677–697. https://doi. Kilburn, V. L., Ibáñez, R., Sanjur, O., Bermingham, E., Suraci, J. P., &
org/10.1146/annur​ev.ecols​ys.110308.120159 Green, D. M. (2010). Ubiquity of the pathogenic chytrid fun-
Ellner, S. P., & Rees, M. (2006). Integral projection models for species gus, Batrachochytrium dendrobatidis, in anuran communities in
with complex memography. The American Naturalist, 167(3), 410– Panamá. EcoHealth, 7(4), 537–548. https://doi.org/10.1007/s1039​
428. https://doi.org/10.1086/499438 3-010-0634-1
Flechas, S. V., Paz, A., Crawford, A. J., Sarmiento, C., Acevedo, A. A., Kirshtein, J., Anderson, C., Wood, J., Longcore, J., & Voytek, M. (2007).
Arboleda, A., … Amézquita, A. (2017). Current and predicted dis- Quantitative PCR detection of Batrachochytrium dendrobatidis DNA
tribution of the pathogenic fungus Batrachochytrium dendrobatidis from sediments and water. Diseases of Aquatic Organisms, 77, 11–15.
in Colombia, a hotspot of amphibian biodiversity. Biotropica, 49(5), https://doi.org/10.3354/dao01831
685–694. https://doi.org/10.1111/btp.12457 Kriger, K. M., Hines, H. B., Hyatt, A. D., Boyle, D. G., & Hero, J.-M. (2006).
García-Rodríguez, A., Chaves, G., Benavides-Varela, C., & Puschendorf, Techniques for detecting chytridiomycosis in wild frogs: Comparing
R. (2012). Where are the survivors? Tracking relictual populations histology with real-time Taqman PCR. Diseases of Aquatic Organisms,
of endangered frogs in Costa Rica. Diversity and Distributions, 71(2), 141–148. https://doi.org/10.3354/dao07​1141
18(2), 204–212. https://doi.org/10.1111/j.1472-4642.2011. Lambin, E. F., & Meyfroidt, P. (2011). Global land use change, eco-
00862.x nomic globalization, and the looming land scarcity. Proceedings of
Goldberg, C. S., Hawley, T. J., & Waits, L. P. (2009). Local and regional pat- the National Academy of Sciences, 108(9), 3465–3472. https://doi.
terns of amphibian chytrid prevalence on the Osa Peninsula, Costa org/10.1073/pnas.11004​8 0108
Rica. Herpetological Review, 40(3), 309–311. Langholz, J., & Lassoie, J. (2001). Combining conservation and devel-
Gómez-Hoyos, D., Méndez-Arrieta, R., Méndez-Arrieta, A., Seisdedos- opment on private lands: Lessons from Costa Rica. Environment,
de-Vergara, R., Abarca, J., Barrio-Amorós, C., & González-Maya, J. Development and Sustainability, 3(4), 309–322. https://doi.
(2018). Anuran inventory in a locality of the buffer area of La Amistad org/10.1023/A:10208​28610847
International Park, Costa Rica: Pilot study for Citizen Science appli- Langwig, K. E., Voyles, J., Wilber, M. Q., Frick, W. F., Murray, K. A., Bolker,
cation. Anales De Biología, 40, 57–64. https://doi.org/10.6018/anale​ B. M., … Kilpatrick, A. M. (2015). Context-dependent conservation
sbio.40.07 responses to emerging wildlife diseases. Frontiers in Ecology and the
Heard, G. W., Scroggie, M. P., Ramsey, D. S. L., Clemann, N., Hodgson, Environment, 13(4), 195–202. https://doi.org/10.1890/140241
J. A., & Thomas, C. D. (2018). Can habitat management mitigate dis- Larson, S. R., Degroot, J. P., Bartholomay, L. C., & Sugumaran, R. (2010).
ease impacts on threatened amphibians? Conservation Letters, 11(2), Ecological niche modeling of potential West Nile virus vector mos-
e12375. https://doi.org/10.1111/conl.12375 quito species in Iowa. Journal of Insect Science, 10(110), 1–17. https://
Hernandez-Stefanoni, J. L., & Ponce-Hernandez, R. (2006). Mapping the doi.org/10.1673/031.010.11001
spatial variability of plant diversity in a Tropical forest: Comparison of Lieberman, D., Lieberman, M., Peralta, R., & Hartshorn, G. S. (1996).
spatial interpolation methods. Environmental Monitoring and Assessment, Tropical forest structure and composition on a large-scale altitudinal
117(1), 307–334. https://doi.org/10.1007/s1066​1-006-0885-z gradient in Costa Rica. The Journal of Ecology, 84(2), 137–152. https://
Hijmans, R. J., Cameron, S. E., Parra, J. L., Jones, P. G., & Jarvis, A. (2005). doi.org/10.2307/2261350
Very high resolution interpolated climate surfaces for global land Lips, K. R., Green, D. E., & Papendick, R. (2003). Chytridiomycosis in
areas. International Journal of Climatology, 25(15), 1965–1978. https:// wild frogs from southern Costa Rica. Journal of Herpetology, 37(1),
doi.org/10.1002/joc.1276 215–218. https://doi.org/10.1670/0022-1511(2003)037[0215:CI-
Hijmans, R. J., Phillips, S., Leathwick, J., & Elith, J. (2013). Dismo: Species WFF​S]2.0.CO;2
distribution modeling. R Package Version 0.8-17. http://Cran.r-Project. Lips, K. R., Reeve, J. D., & Witters, L. R. (2003). Ecological
Org. https://cran.r-proje​c t.org/web/packa​ges/dismo/​index.html traits predicting amphibian population declines in Central
Hooper, D. U., Adair, E. C., Cardinale, B. J., Byrnes, J. E. K., Hungate, B. A., America. Conservation Biology, 17(4), 1078–1088. https://doi.
Matulich, K. L., … O’Connor, M. I. (2012). A global synthesis reveals org/10.1046/j.1523-1739.2003.01623.x
biodiversity loss as a major driver of ecosystem change. Nature, 486, Longcore, J. E., Pessier, A. P., & Nichols, D. K. (1999). Batrachochytrium
105–108. https://doi.org/10.1038/natur​e11118 dendrobatidis gen. Et sp. Nov., a chytrid pathogenic to amphibians.
Illoldi-Rangel, P., Rivaldi, C.-L., Sissel, B., Trout Fryxell, R., Gordillo-Pérez, Mycologia, 91, 219–227. https://doi.org/10.2307/3761366
G., Rodríguez-Moreno, A., … Sarkar, S. (2012). Species distribution Meredith, H., Van Buren, C., & Antwis, R. E. (2016). Making amphibian
models and ecological suitability analysis for potential tick vectors of conservation more effective. Conservation Evidence, 13, 1–6.
Lyme Disease in Mexico. Journal of Tropical Medicine, 2012, 959101. Merow, C., Smith, M. J., & Silander, J. A. Jr (2013). A practical guide to
https://doi.org/10.1155/2012/959101 MaxEnt for modeling species’ distributions: What it does, and why
10 | ZUMBADO-ULATE et al.

inputs and settings matter. Ecography, 36(10), 1058–1069. https:// chytrid fungus: The host’s perspective. Animal Conservation, 11(6),
doi.org/10.1111/j.1600-0587.2013.07872.x 513–517. https://doi.org/10.1111/j.1469-1795.2008.00210.x
Monastersky, R. (2014). Life-a status report. Nature, 516(7530), 158. Rohr, J. R., Halstead, N. T., & Raffel, T. R. (2011). Modelling the future
https://doi.org/10.1038/516158a distribution of the amphibian chytrid fungus: The influence of cli-
Muscarella, R., Galante, P. J., Soley-Guardia, M., Boria, R. A., Kass, mate and human-associated factors. Journal of Applied Ecology, 48(1),
J. M., Uriarte, M., & Anderson, R. P. (2014). ENMeval: An R pack- 174–176. https://doi.org/10.1111/j.1365-2664.2010.01891.x
age for conducting spatially independent evaluations and estimat- Rohr, J. R., & Raffel, T. R. (2010). Linking global climate and tempera-
ing optimal model complexity for Maxent ecological niche models. ture variability to widespread amphibian declines putatively caused
Methods in Ecology and Evolution, 5(11), 1198–1205. https://doi. by disease. Proceedings of the National Academy of Sciences, 107(18),
org/10.1111/2041-210X.12261 8269–8274. https://doi.org/10.1073/pnas.09128​83107
Nori, J., Lemes, P., Urbina-Cardona, N., Baldo, D., Lescano, J., & Loyola, Ron, S. R. (2005). Predicting the distribution of the amphibian pathogen
R. (2015). Amphibian conservation, land-use changes and protected Batrachochytrium dendrobatidis in the New World. Biotropica, 37(2),
areas: A global overview. Biological Conservation, 191, 367–374. 209–221. https://doi.org/10.1111/j.1744-7429.2005.00028.x
https://doi.org/10.1016/j.biocon.2015.07.028 Saenz, D., Adams, C. K., Pierce, J. B., & Laurencio, D. (2009). Occurrence
Oliver, M. A., & Webster, R. (1990). Kriging: A method of interpola- of Batrachochytrium dendrobatidis in an anuran community in the
tion for geographical information systems. International Journal southeastern Talamanca region of Costa Rica. Herpetological Review,
of Geographical Information Systems, 4(3), 313–332. https://doi. 40(3), 311–313.
org/10.1080/02693​79900​8941549 Sasa, M., Chaves, G., & Porras, L. W. (2010). The Costa Rican herpe-
Peterson, A. T., Bauer, J. T., & Mills, J. N. (2004). Ecologic and geographic tofauna: Conservation status and future perspectives. In J. H.
distribution of filovirus disease. Emerging Infectious Diseases, 10(1), Townsend, & J. D. Johnson (Eds.), Conservation of Mesoamerican
40–47. https://doi.org/10.3201/eid10​01.030125 amphibians and reptiles (pp. 510–603). Eagle Mountain, UT: Eagle
Phillips, S. J., Anderson, R. P., & Schapire, R. E. (2006). Maximum en- Mountain Press.
tropy modeling of species geographic distributions. Ecological Savage, J. M. (2002). The amphibians and reptiles of Costa Rica: A herpe-
Modelling, 190(3), 231–259. https://doi.org/10.1016/j.ecolm​ tofauna between two continents, between two seas, Chicago, IL: The
odel.2005.03.026 University of Chicago Press.
Phillips, S. J., & Dudík, M. (2008). Modeling of species distributions with Scheele, B. C., Pasmans, F., Skerratt, L. F., Berger, L., Martel, A. N.,
Maxent: New extensions and a comprehensive evaluation. Ecography, Beukema, W., … Canessa, S. (2019). Amphibian fungal panzo-
31(2), 161–175. https://doi.org/10.1111/j.0906-7590.2008. otic causes catastrophic and ongoing loss of biodiversity. Science,
5203.x 363(6434), 1459–1463. https://doi.org/10.1126/scien​ce.aav0379
Picco, A. M., & Collins, J. P. (2007). Fungal and viral pathogen occurrence Skerrat, L., Berger, L., Hines, H., McDonald, K., Mendez, D., & Speare,
in Costa Rican amphibians. Journal of Herpetology, 41(4), 746–749. R. (2008). Survey protocol for detecting chytridiomycosis in all
https://doi.org/10.1670/07-033.1 Australian frog populations. Diseases of Aquatic Organisms, 80(2),
Pinkard, E. A., Kriticos, D. J., Wardlaw, T. J., Carnegie, A. J., & Leriche, 85–94. https://doi.org/10.3354/dao01923
A. (2010). Estimating the spatio-temporal risk of disease epidemics Skerratt, L. F., Mendez, D., McDonald, K. R., Garland, S., Livingstone,
using a bioclimatic niche model. Ecological Modelling, 221(23), 2828– J., Berger, L., & Speare, R. (2011). Validation of diagnostic tests in
2838. https://doi.org/10.1016/j.ecolm​odel.2010.08.017 wildlife: The case of chytridiomycosis in wild amphibians. Journal of
Pounds, A. J., Bustamante, M. R., Coloma, L. A., Consuegra, J. A., Fogden, Herpetology, 45(4), 444–450. https://doi.org/10.1670/10-193.1
M. P. L., Foster, P. N., … Young, B. E. (2006). Widespread amphibian Stuart, S. N., Chanson, J., Cox, N. A., Young, B. E., Rodrigues, A. S. L.,
extinctions from epidemic disease driven by global warming. Nature, Fischman, D., & Waller, R. (2004). Status and trends of amphibian
439(7073), 161–167. https://doi.org/10.1038/natur​e 04246 declines and extinctions worldwide. Science, 306(5702), 1783–1786.
Pounds, J. A., & Crump, M. L. (1994). Amphibian declines and cli- https://doi.org/10.1126/scien​ce.1103538
mate disturbance: The case of the golden toad and the harlequin Voyles, J., Woodhams, D. C., Saenz, V., Byrne, A. Q., Perez, R., Rios-
frog. Conservation Biology, 8(1), 72–85. https://doi.org/10.104 Sotelo, G., … Richards-Zawacki, C. L. (2018). Shifts in disease dynam-
6/j.1523-1739.1994.08010​072.x ics in a tropical amphibian assemblage are not due to pathogen at-
Powers, J. S., Becknell, J. M., Irving, J., & Pèrez-Aviles, D. (2009). tenuation. Science, 359(6383), 1517–1519. https://doi.org/10.1126/
Diversity and structure of regenerating tropical dry forests in Costa scien​ce.aao4806
Rica: Geographic patterns and environmental drivers. Forest Ecology Voyles, J., Young, S., Berger, L., Campbell, C., Voyles, W. F., Dinudom,
and Management, 258(6), 959–970. https://doi.org/10.1016/j. A., … Speare, R. (2009). Pathogenesis of chytridiomycosis, a cause
foreco.2008.10.036 of catastrophic amphibian declines. Science, 326(5952), 582–585.
Puschendorf, R., Carnaval, A. C., VanDerWal, J., Zumbado-Ulate, https://doi.org/10.1126/scien​ce.1176765
H., Chaves, G., Bolaños, F., & Alford, R. A. (2009). Distribution Wake, D. B., & Vredenburg, V. T. (2008). Are we in the midst of the sixth
models for the amphibian chytrid Batrachochytrium dendro- mass extinction? A view from the world of amphibians. Proceedings of
batidis in Costa Rica: Proposing climatic refuges as a conserva- the National Academy of Sciences, 105(Supplement 1), 11466–11473.
tion tool. Diversity and Distributions, 15(3), 401–408. https://doi. https://doi.org/10.1073/pnas.08019​21105
org/10.1111/j.1472-4642.2008.00548.x Wandeler, P., Hoeck, P. E. A., & Keller, L. F. (2007). Back to the fu-
Puschendorf, R., Hoskin, C. J., Cashins, S. D., McDonald, K., Skerratt, L. ture: Museum specimens in population genetics. Trends in
F., Vanderwal, J., & Alford, R. A. (2011). Environmental refuge from Ecology & Evolution, 22(12), 634–642. https://doi.org/10.1016/j.
disease-driven amphibian extinction. Conservation Biology, 25(5), tree.2007.08.017
956–964. https://doi.org/10.1111/j.1523-1739.2011.01728.x Warren, D. L., & Seifert, S. N. (2011). Ecological niche modeling in
Radosavljevic, A., & Anderson, R. P. (2014). Making better MAXENT mod- Maxent: The importance of model complexity and the performance
els of species distributions: Complexity, overfitting and evaluation. of model selection criteria. Ecological Applications, 21(2), 335–342.
Journal of Biogeography, 41(4), 629–643. https://doi.org/10.1111/ https://doi.org/10.1890/10-1171.1
jbi.12227 Whitfield, S. M., Alvarado, G., Abarca, J., Zumbado-Ulate, H., Zuñiga,
Rödder, D., Veith, M., & Lötters, S. (2008). Environmental gradients ex- I., Wainwright, M., & Kerby, J. (2017). Differential patterns of
plaining the prevalence and intensity of infection with the amphibian Batrachochytrium dendrobatidis infection in relict amphibian
ZUMBADO-ULATE et al. | 11

populations following severe disease-associated declines. Diseases supports the existence of a climatic refuge from disease. EcoHealth,
of Aquatic Organisms, 126(1), 33–41. https://doi.org/10.3354/ 11(4), 593–602. https://doi.org/10.1007/s1039​3-014-0967-2
dao03154 Zumbado-Ulate, H., Bolaños, F., Willink, B., & Soley-Guardia, F. (2011).
Whitfield, S. M., Geerdes, E., Chacon, I., Ballestero Rodriguez, E., Population status and natural history notes on the critically endan-
Jimenez, R., Donnelly, M., & Kerby, J. (2013). Infection and co-in- gered stream dwelling frog Craugastor ranoides (Craugastoridae) in
fection by the amphibian chytrid fungus and ranavirus in wild Costa a Costa Rican tropical dry forest. Herpetological Conservation and
Rican frogs. Diseases of Aquatic Organisms, 104(2), 173–178. https:// Biology, 6(3), 455–464.
doi.org/10.3354/dao02598 Zumbado-Ulate, H., García-Rodríguez, A., Vredenburg, V. T., & Searle, C.
Whitfield, S. M., Kerby, J., Gentry, L. R., & Donnelly, M. A. (2012). L. (2019). Infection with Batrachochytrium dendrobatidis is common in
Temporal variation in infection prevalence by the amphibian chytrid tropical lowland habitats: Implications for amphibian conservation.
fungus in three species of frogs at La Selva, Costa Rica. Biotropica, Ecology and Evolution, 9(8), 4917–4930. https://doi.org/10.1002/
44(6), 779–784. https://doi.org/10.1111/j.1744-7429.2012.00872.x ece3.5098
Whitfield, S. M., Lips, K. R., & Donnelly, M. A. (2016). Amphibian de- Zumbado-Ulate, H., Nelson, K. N., García-Rodríguez, A., Chaves, G.,
cline and conservation in Central America. Copeia, 104(2), 351–379. Arias, E., Bolaños, F., … Searle, C. L. (2019). Endemic infection of
https://doi.org/10.1643/CH-15-300 Batrachochytrium dendrobatidis in Costa Rica: Implications for am-
Wilber, M. Q., Langwig, K. E., Kilpatrick, A. M., McCallum, H. I., & phibian conservation at regional and species level. Diversity, 11(8),
Briggs, C. J. (2016). Integral Projection Models for host–para- 129. https://doi.org/10.3390/d1108​0129
site systems with an application to amphibian chytrid fungus.
Methods in Ecology and Evolution, 7(10), 1182–1194. https://doi.
org/10.1111/2041-210X.12561 S U P P O R T I N G I N FO R M AT I O N
Woodhams, D. C., Kilburn, V. L., Reinert, L. K., Voyles, J., Medina, D., Additional supporting information may be found online in the
Ibáñez, R., … Rollins-Smith, L. A. (2008). Chytridiomycosis and Supporting Information section.
amphibian population declines continue to spread eastward in
Panama. EcoHealth, 5(3), 268–274. https://doi.org/10.1007/s1039​
3-008-0190-0
Xie, G. Y., Olson, D. H., & Blaustein, A. R. (2016). Projecting the
How to cite this article: Zumbado-Ulate H, García-Rodríguez
global distribution of the emerging amphibian fungal patho- A, Searle CL. Species distribution models predict the
gen, Batrachochytrium dendrobatidis, based on IPCC climate fu- geographic expansion of an enzootic amphibian pathogen.
tures. PLoS One, 11(8), e0160746. https://doi.org/10.1371/journ​ Biotropica. 2020;00:1–11. https://doi.org/10.1111/btp.12863
al.pone.0160746
Zumbado-Ulate, H., Bolaños, F., Gutiérrez-Espeleta, G., & Puschendorf,
R. (2014). Extremely low prevalence of Batrachochytrium dendro-
batidis in frog populations from Neotropical dry forest of Costa Rica

View publication stats