What is the potential for extirpating spectacled caiman from Comprehensive Everglades Restoration Plan projects in South Florida?

Spectacled caimans ( Caiman crocodilus Linnaeus, 1758) are established invaders in the United States, Cuba, and San Andres Island, Colombia. They have been established in South Florida since the 1970s and are found primarily within Comprehensive Everglades Restoration Plan (CERP) projects. These projects provide suitable habitat and may provide dispersal pathways via water management activities. Caiman presence in these areas directly conflicts with the CERP’s goals, and as a generalist species with a broad diet, they can impact biological resources. Past removal efforts failed to extirpate caimans, but their efficacy has not been well evaluated. We addressed caimans via removal surveys during 2012–2021 with the goal of evaluating extirpation or maximum species control within South Florida’s CERP projects. Documented opportunistic removals for this study began in December 2012, and systematic efforts have been ongoing since October 2017. We evaluated efficacy of caiman removals by analyzing 10 years of opportunistic and systematic survey data, plus associated removal data, along 11 survey routes during 2012–2021. We also conducted necropsies to collect biological information which could be used to improve removal efforts. We removed 251 caimans during 2012 to 2021, and the rate of caiman removals per year increased from 5 animals during 2012 to a peak of 47 animals during 2020. Necropsies revealed reproductive information (nesting/ hatching timelines) that we applied to improve our removal rates. Caiman encounter rates declined from a peak of 1.55 ± 0.66 caiman/hr and 0.72 ± 0.38 caiman/km during 2013 to a low of 0.18 ± 0.09 caiman/hr and 0.03 ± 0.02 caiman/km during 2020 and slightly increased during 2020–2021, likely because of a change in search effort. We evaluated and discussed the potential for extirpating caiman from CERP projects and provide a data-driven prescription for removal efforts.


Introduction
Spectacled caimans are established invaders in Florida and Puerto Rico, United States (US), Isla de la Juventud, Cuba, and San Andres Island, Colombia (Balaguera-Reina and Velasco 2019). Caimans were first reported as introduced in South Florida by King and Krakauer (1966) and as established by Ellis (1980), who reported a breeding population at what is now the Homestead Air Reserve Base (HARB). King and Krakauer (1966), Ellis (1980), and Moore (1954) attributed presence of caimans to an increase in stocking and selling by pet dealers in response to trade laws passed during the 1950s to protect declining American alligator (Alligator mississippiensis Daudin, 1802) populations. These caimans were identified as being sourced from South America (Moore 1954), probably from Colombia (Ellis 1980), and Roberto et al. (2021) molecularly identified Colombia as the most likely origin of Florida caiman using museum samples.
Although HARB has been the principal area, caiman sightings and captures have been reported throughout the Greater Everglades (Figure 1). King and Krakauer (1966) reported caimans as far north as Palm Beach County whereas Wilson and Porras (1983) reported sightings from the Miami canal system and a female with hatchlings along Tamiami Trail. Sightings have also occurred within and adjacent to important conservation areas such as Everglades National Park (ENP; Butterfield et al. 1997;Meshaka et al. 2000), Biscayne National Park, Everglades and Francis S. Taylor Wildlife Management Area (WMA), Frog Pond WMA, and Big Cypress National Preserve as well as Comprehensive Everglades Restoration Plan (CERP) projects such as the Biscayne Bay Coastal Wetlands project (BBCW) and C-111 Spreader Canal Western Project , increasing concerns about effects this species can have on these threatened ecosystems ( Figure 1).
CERP projects were established to restore the Everglades ecosystem via freshwater deliveries designed to recreate historical wetland conditions. Caimans occupy most habitat types present in the CERP projects, including man-made and disturbed wetland habitats, and they appear to utilize new habitats within these projects as they expand. The presence of invasive species in these projects directly conflicts with the ecological goals and objectives of the CERP to improve native plant and animal abundance and diversity (National Research Council 2015). Although invasive wildlife is recognized as a threat to the CERP (National Research Council 2015), they are not as actively managed as part of each CERP project as invasive plants.
As an aquatic habitat generalist with a broad diet (Ellis 1980;Alvarez del Toro 1982;Thorbjarnarson 1993;Aguire and Poss 2000;Bontemps et al. 2016), spectacled caimans have the potential to impact South Florida's biological resources. Caimans in South Florida may prey upon protected species such as young American crocodiles (Crocodylus acutus Cuvier, 1807). Although it is less likely, caimans may also prey upon eastern indigo snakes (Drymarchon couperi Holbrook, 1842), which have been known to occur near or within both the BBCW and the C-111 Project (Steiner et al. 1983;Enge et al. 2013;Chandler et al. 2022). Indigo snake observations in southern Florida demonstrate that they may spend substantial time in both freshwater and saline wetlands (Steiner et al. 1983;Metcalf et al. 2021), putting them at risk in areas where caimans overlap. Caimans may also compete for food and space with native crocodylians. An expert panel collaboration to assign research and management priorities for Florida's invasive reptiles identified caiman as having a "highest impact concern" (Engeman and Avery 2016).
Caiman extirpation efforts in South Florida started in 1977 at HARB when 44 caimans were removed during a one-year effort, which could have conceivably represented all the caimans on the airbase at the time (Ellis 1980). However, very few efforts have been reported since then, and consistent documentation of caiman removals at HARB was not available until November 2004 (Brandt and Mazzotti 1990;AMEC 2012). In 2012, a caiman removal feasibility study was performed at HARB to evaluate the effect of four courses of action: no action, opportunistic removals, periodic removal surveys, and eradication (AMEC 2012). This study concluded that periodic surveys would be the most desirable management option since opportunistic removals had failed to achieve the intended results and the eradication plan was not deemed desirable due to high up-front costs and potential negative impacts to native crocodylians (AMEC 2012). Since then, the US Department of Agriculture personnel on the airbase periodically removed 44 caimans between 2009-2019 that interfered with base operations (J. Edwards and J. Friers pers. comm.).
We evaluated trends in caiman observations and removals over time and discuss them in the context of potential extirpation or maximum control within or adjacent to South Florida CERP projects (BBCW, C-111 Project) and natural protected areas (ENP, Biscayne National Park). The goal of our effort was to reduce or eliminate the direct conflict that this invasive species poses to the goals of the CERP and reduce potential impacts to biological resources. Our specific objectives were to 1) improve caiman removal rates based on ecological (effect of environmental variables on encounter rates) and biological (reproductive information gained via necropsies) knowledge empirically collected across South Florida, 2) define a data driven methodological approach to manage this invasion, and 3) to test the hypothesis that caiman encounter rates will decrease and stay low (at or near zero) in response to caiman removals after systematic efforts. We defined "maximum control" and "extirpation" as sustained zero encounter rates from a described area through intermittent periods of at least 5 consecutive surveys and 5 years, respectively, verified via monitoring. For this to happen, we assumed that there will be no deliberate releases of new caimans which would facilitate reinvasion.

Field and laboratory work
We conducted spotlight surveys during 2012-2021 along 11 routes ( Figure 2) by truck, foot, and/or boat using Fenix HP15 900-lumen headlamps and a Fenix TK41C 1,000-lumen tactical flashlight (Fenix Lighting, Littleton, CO, USA). We selected routes based on previously reported caiman sightings, layout of navigable roads and levees, and location within or along the footprints of CERP projects and protected natural areas. Habitats varied within and among routes, so we used truck and/or foot travel to survey wetlands, canals, and ditches and boats to survey ponds, lakes, and other areas of open water. We recorded air temperature (Celsius), relative humidity (%), and average wind speed (mph; one minute average) with a Kestrel 3000 weather meter (Kestrel Instruments, Boothwyn, PA, USA) at the start and end of each survey. We also recorded general weather conditions and lunar phase during each survey.
We began opportunistically removing caimans in BBCW in December 2012 (only one survey was conducted during 2012), and systematic surveys  Figure 3). We also conducted systematic removal surveys at HARB, in collaboration with airbase personnel, during January 2019-February 2020. However, due to the COVID-19 pandemic, access to the airbase was restricted from March 2020 through most of 2021. Airbase biologists continued periodic surveys during this time, but no caimans were detected (J. Edwards pers. comm.).
We identified spectacled caimans, estimated total length to the lower bound of 0.25 m increments, and collected geospatial coordinates (Universal Transverse Mercator zone 17N, map datum WGS 1984) with a Garmin handheld global positioning system unit (Garmin International, USA). We used the tactical flashlight and 10 × 42 mm binoculars to verify species identification when necessary. Since American alligators and American crocodiles were also present in the study area, if we were unable to identify the species then we recorded animals as "unknown." Observation time, cloud cover, habitat type, and number of animals observed were recorded at each observation.
Size estimates indicate the lower bound of the size increment; for example, a 0.5 m estimate ranges from 0.5-0.74 m, and so on. Total length estimates were grouped into size aggregates loosely following the size class recommendations of Gorzula and Seijas (1989) which were based on snout-vent length. We defined the size aggregates as hatchling (≤ 0.25 m during September-October), juvenile (≥ 0.25 m during November-August and ≤ 1.25 m), or adult (> 1.25 m). If we could not estimate total length to the nearest 0.25 m increment, then the animal was either assigned to a size aggregate or no size estimate was recorded. We validated caiman size estimates by calculating the percentage of correct estimates compared to the measured total lengths of subsequently captured animals (n = 216). Size estimates were 91.2% correct in the 0.25 m increments and 97.2% correct in the aggregates, so we are confident that our size estimates were relatively accurate for uncaptured animals.
We attempted to remove all caiman detected. Capture methods were varied to maximize probability of capture and included hand captures, snake tongs, wire snares, and toggle darts. Starting in 2017, we recorded air temperature (Celsius), water temperature (Celsius), salinity (psu), and water depth (cm) at our caiman detections. Air and water temperatures were collected with a stainless-steel, instant-read cooking thermometer (Taylor Precision Products, Oak Brook, IL, USA), salinity was collected with a handheld temperature-corrected refractometer (Agriculture Solutions, Kingfield, ME, USA), and water depth was collected via a folding ruler (U.S. Tape Company, Pennsburg, PA, USA). We humanely euthanized all captured caiman in accordance with the American Veterinary Medical Association guidelines (American Veterinary Medical Association 2020).
Euthanized animals were subsequently necropsied following standard necropsy procedure of Farris et al. (2013), and all specimens and samples were stored at −20 °C. We collected morphometric data (head length (HL), snout-vent length (SVL), total length (TL), tail girth (TG), and mass), assessed the general condition of major organ systems (i.e., normal or abnormal), and attempted to determine reproductive status. We assigned caimans to size classes based on SVL so that we could examine trends by size class. We considered caiman ≥ 60 cm snout-vent length (SVL) as adults, caiman ≥ 14 cm and < 60 cm SVL as juveniles, and caiman < 14 cm SVL as hatchlings. Adult and juvenile classifications follow Gorzula and Seijas (1989) recommendations. However, we added a lower bound for juveniles based on the creation of the hatchling size class. The hatchling size class was delineated by using capture and morphometric data collected from individuals in South Florida.

Data Analysis
We used a variety of travel methods during surveys, and we did not always survey the same transect lengths during each survey of a given route. So, we calculated encounter rates per hour and kilometer by year and route to account for sample heterogeneity (sampling by foot versus sampling by truck, area covered versus time spent) and assess spatial and temporal variation. We assessed effects of environmental variables collected at the beginning of each survey (air temperature, humidity, wind speed, weather condition, and lunar phase), on-the-spot environmental variables (air and water temperature, salinity, water depth), and habitat (wetlands, canals, ditches, and ponds/lakes/open water) on both encounter rates and number of caiman observed across routes via generalized linear models using the "glm" function in R version 4.2.1 (R Core Team 2022). Survey routes without any caiman encounters were excluded from the encounter rate calculations and generalized linear models to avoid bias from all-zero encounter rate values. We also assessed effects of survey effort, survey distance, and number of observers on the number of caimans removed per survey using the "glm" function in R. Variables were selected based on critical values (p-value < 0.05) and the best model was defined based on the lowest Akaike Information Criteria (AIC) score, the closest value to 1 when tested for under/overdispersion based on model residuals, and the highest pseudo R 2 (Dobson 2002). If these parameters were too close to choose the best candidate model, then we ran a likelihood-ratio test via the "lrtest" function in R as a final criterion to choose the best model. We evaluated trends in number of animals captured per year, season (wet or dry), and month as well as by sex and size class. We defined the South Florida wet season as May to October and the dry season as November to April each year. We calculated annual mean caiman capture probability (number of caimans captured/total number of caimans observed) from each survey where at least one caiman was detected to assess the proportion of caimans captured versus escaped over time. We analyzed differences in morphometrics (HL, SVL, TL, TG, and mass) of captured animals by habitat type to assess whether we could target removals of specific size classes by searching in specific habitat types. To do so, we tested data for normality via Shapiro-Wilks tests with the "shapiro.test" function and then ran Kruskal-Wallis tests with the 'kruskal.test' function in R to analyze differences in morphometric data by habitat type.

Results
We conducted a total of 330 spotlight and removal surveys during 2012-2021 totaling 958.6 hr of survey effort and 6,935.6 km of survey distance (Table 1). Mean survey effort was 2.9 ± 1.2 hr, and mean survey distance was 21.0 ± 10.0 km. Most of the surveys were done by vehicle (n = 240), followed by a combination of vehicle/foot (n = 58), or vehicle/boat (n = 19). We obtained a total of 317 caiman observations ranging from 0 to 29 observations per survey. Based on total length size estimates, most caiman observations were juveniles, followed by hatchlings, then adults ( Table 2). The remaining 30 caimans were not assigned a size estimate or size aggregate. Caimans were observed throughout the year with peaks in October (n = 107), March (n = 56), and January (n = 50). Caimans were most observed in either shallow-water habitats such as marsh (n = 104) and ditches (n = 45) or in deep-water habitats such as canals (n = 72) and ponds/lakes (n = 36 Combined annual mean caiman encounter rates exhibited a noticeable decrease from a peak of 1.55 ± 0.66 caimans/hr and 0.72 ± 0.38 caimans/km during 2013 to a low of 0.18 ± 0.09 caimans/hr and 0.03 ± 0.02 caimans/km during 2020, then they slightly increased during 2020-2021 (Figure 4). Mean encounter rates per survey route ranged from 0.0 to 1.82 ± 0.75 caimans/hr and 0.0 to 0.86 ± 0.44 caimans/km depending on the year and route, decreasing across all routes by the end of the study period (Figure 4). The decrease in both combined and per survey route encounter rates is more accentuated when analyzed by area covered than by time spent (Figure 4). The L-31N Canal, Aerojet Canal, C-110 Canal, and C-111 Canal routes had zero caiman observations during the last three years of the study. The HARB, Rocky Glades North, and Frog Pond routes had zero caiman observations during the last two years of the study (Figure 4). The Rocky Glades South route had zero caiman observations during 2019 and 2021 ( Figure 4). We also had zero caiman observations from our surveys on this route during 2020, but python removal program participants reported one verified sighting that year. The Lake Chekika route exhibited a steady reduction in encounter rates, which were near zero by the end of the study period (Figure 4). The BBCW and L-357 Canal routes had variable annual encounter rates, but both routes exhibited decreases over the course of the study period (Figure 4). Despite these overall decreases, encounter rates increased in the BBCW route and were stable in the L-357 Canal route during the last two years of the study (Figure 4).
Relationships between environmental variables collected at the beginning of surveys and encounter rates were non-linear and thus not amenable to generalized linear modeling. Model-fitting analyses of environmental variables collected on spot and the number of caimans observed showed that the global Negative Binomial (NB) model outperformed both the Poisson (P) and Gaussian (G) distribution structure models (AIC = 297.3,327.1,and 414.8,respectively). Residual tests also corroborated the NB distribution as the best model for our data due to no evidence of overdispersion (0.91) compared with P and G distributions (1.62 and 4.53, respectively). Finally, likelihood-ratio tests provided strong evidence that the best model was the global NB (χ 2 = 31.76, p-value < 0.000). The NB model showed significance in the intercept (z = −2.77, p-value = 0.006) and in only one environmental variable (water temperature z = 1.98, p = 0.05) explaining around 27.4% of the variation found in the number of caimans observed. Relationships between survey effort and number of caimans removed were also non-linear. Model-fitting analyses showed that the NB model again outperformed both the P and G distribution structure models (AIC = 591.4, 965.49, and 1469.0, respectively), though there was evidence of overdispersion (2.25). The NB model showed significance in the intercept (z = −2.85, p-value = 0.004) and in the survey effort (z = 5.96, p-value = < 0.000) and survey distance (z = −4.05, p-value = < 0.000) variables explaining around 22.5% of the variation found in the number of caimans removed.
We captured and removed a total of 251 caimans during 2012-2021 (79.2 percent of the total observed) which were comprised of 90 males, Hatchling animals were defined as snout-vent length < 14cm, juveniles as ≥ 14cm and < 60cm snout-vent length, and adults as ≥ 60cm snout-vent length. Capture probability was calculated as the number of caimans captured divided by the total number of caimans observed. 22 females, and 139 unsexed individuals (36 adults, 131 juveniles, and 81 hatchlings; Table 2, Figure 5). We did not collect the required morphometric data to assign the three remaining animals to a size class. The rate of caiman removals per year increased from five animals during 2012 to a peak of 47 animals during 2020 ( Figure 5). Annual mean caiman capture probability decreased from a maximum of 1.0 during 2012 to a minimum of 0.58 ± 0.08 during 2018 and then increased to 0.88 ± 0.07 during 2020 ( Figure 5). We have mostly removed males and individuals of unknown sex each year, and the number of unsexed individuals remained relatively stable -these numbers are largely comprised of hatchlings ( Figure 5). Regarding size class, we have mostly removed juveniles and hatchlings each year, and the number of adults steadily decreased during the last five years of the study period ( Figure 5).
Necropsies yielded useful reproductive information for an adult female captured at BBCW on May 7 th , 2018, which contained 43 mature follicles with a mean length and diameter of 30.68 ± 6.53 cm and 31.35 ± 6.60 cm, respectively, and a combined mass of 900 g. Another female caiman with similarly sized follicles was captured at HARB during late April 2018 (J. Edwards pers. comm.). Given the state of follicle development in these females, it seems likely that caiman nesting in South Florida occurs from mid-June to mid-July. Retroactively calculating the range of reported incubation periods (Allsteadt 1994;Gorzula 1978;Staton and Dixon 1977;Alvarez del Toro 1974 in Staton andDixon 1977) from the first appearance of hatchlings in mid-late September results in an oviposition date range from mid-June to mid-July. Hatchling capture dates and associated morphological features, such as the presence of umbilical scars, further corroborate this timeline. We have removed four pods of hatchlings (55 out of 81 total hatchlings removed) and one adult female since gaining this information in mid-2018.

Discussion
Our results suggest that despite increasing caiman removal rates (i.e., more individuals removed per year and almost all detected individuals removed) we observed an overall decrease in caiman encounter rates across all routes. These observations suggest there may have been a reduction in the number of caimans in the effective survey area of our routes. We have achieved our definition of maximum control within the effective survey area of all our survey routes. Additionally, the sustained zero encounter rates for up to three years in several routes Aerojet Canal, suggests we may be approaching our definition of extirpation on these routes (Table 1). However, there were no reported caiman sightings from these routes prior to our efforts, so it is possible that caimans were originally absent from these routes. The lack of prior sightings in these areas could simply be due to a lack of prior reporting or search effort (i.e., past caimans may have been missed and then emigrated or became locally extinct).
It is likely that caimans were present but unobservable because their territory or home range did not overlap with the effective survey area of our routes. There also could have been an increase in wariness and associated behavioral adjustments by the caimans. This latter issue has been documented for crocodylians with increasing size and for surveys done in areas with human disturbance or hunting (Webb and Messel 1979;Pacheco 1996;Ron et al. 1998;Naveda-Rodríguez et al. 2020;Pereira et al. 2022). We attempted to address these issues by adjusting our search efforts. We analyzed satellite imagery and selected any wetlands which were in the vicinity of our survey routes but may not have been thoroughly covered or readily visible from the route for targeted surveys. Our targeted surveys yielded additional sightings and captures in the BBCW and L-357 Canal routes, notably during 2020 and 2021, but they did not yield any additional caimans on any of the other survey routes. Despite the targeted surveys and increased survey effort we detected and removed noticeably fewer caimans during 2021 than in previous years. Surveys during 2022 detected even fewer caimans (n = 10), the lowest number since 2017 -when survey effort and the number of survey routes substantially increased (Figure 3). We are catching all size classes as they present themselves at both CERP projects, and we are allowing fewer animals to escape over time. We are cautiously optimistic that the removal efforts are generating an impact, but we cannot rule out that the decrease in detections may be due to previously mentioned issues. Continued monitoring is essential to verify our observed trends.
We had fluctuations in survey effort caused by funding and logistical constraints which affected the number, rate, and location of surveys conducted (Table 1, Figure 3). These effort fluctuations could have influenced the observed variation (or lack thereof) in removals and caiman encounter rates. However, it should be noted that years exhibiting the most removals did not align with the years exhibiting the highest survey effort (Figures 3, 5). Additionally, our encounter rate calculations do not account for imperfect detection probability (probability of a caiman being detected during a survey given it is present) and assumes all animals present during any survey are counted or all sites occupied are accounted for, which biased our calculations (MacKenzie et al. 2006;Royle and Dorazio 2006). Further analyses using hierarchical models that account for imperfect detection and unequal survey effort over time (e.g., Kéry and Royle 2016;Vajas et al. 2021) could help to reduce this uncertainty and better evaluate any potential patterns. Using Kéry's (2002) approach for calculating the minimum number of times that a site must be visited to infer absence could further reduce uncertainty.
Most successful invasive species eradication and control projects have been on islands and have focused on mammals, birds, insects, and plants (Simberloff et al. 2018;Robertson et al. 2019). Invasive herpetofauna research and management efforts have lagged, and control or eradication efforts have seldom been attempted or successfully achieved (Kraus 2009).
Yet, the number of research and management efforts for these species have been increasing (Kraus 2015). Successful invasive herpetofauna eradications so far have mostly occurred on islands (Kraus 2009). Despite our study site's location on the mainland of Florida, South Florida functions as a habitat island bounded on three sides by water and on the fourth by frost (Ewel and Myers 1990). This frost barrier could limit the northward expansion of caimans (Brandt and Mazzotti 1990) and improve our prospects for maximum control or extirpation.
Based on our results, we recommend that the basic prescription for maximum species control or extirpation should be to conduct consistent and thorough removal surveys along standardized routes for a period of at least five years. Surveys should be conducted two to three times per week for the initial two to three years, given there is adequate funding, so that surveys are conducted during several nesting seasons and water fluctuations which may increase caiman detectability and accessibility. Thereafter surveys could segue into a less frequent (bi-monthly, monthly, per season) schedule which could be implemented as part of a multispecies monitoring effort. Nest searches should be conducted during mid-July to late-August to optimize removal of unhatched nests and attendant females. To enhance the probability of extirpation, rather than maximum species control, we recommend extending the survey areas to include private lands where possible. This will allow surveyors to seek out potential nesting areas and females in properties that could serve as refugia near public lands. This is particularly relevant to BBCW, which is surrounded by a matrix of private lands. Several invasive herpetofauna control or eradication efforts in Florida appear to have been initiated too late to adequately control or eradicate the target species (Kraus 2009). However, the results of our removal efforts to date suggest that we have achieved maximum control in the effective area of our survey routes and extirpation may be possible. Our efforts should only continue to improve as we gain new information and tools for increasing efficiency. Mitigating this invasive species and its impacts within CERP project footprints will be essential to attaining the goals and objectives of the CERP. The relatively rapid decline of caiman encounter rates in both CERP projects, over 60 years after their introduction, gives us cautious optimism that our efforts may be successful. . The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.