Assessment of three removal methods for an invasive lizard, Peters’ rock agama (Agama picticauda)

Peters’ rock agamas (Agama picticauda) are petrophilic lizards native to Western sub-Saharan Africa and are established in multiple areas in Florida. Where established, this species is conspicuous and numerous and has been the top reported non-native reptile to Florida’s state wildlife agency, the Florida Fish and Wildlife Conservation Commission, since 2021. There is a need to identify effective removal methods that can be applied in different habitats and by homeowners and professionals in both public and private lands. We evaluated three removal methods for A. picticauda in Martin County, Florida, USA. Using a before-after-control-impact design, we evaluated targeted removals by lizard-fishing (baited line on the end of an extendable pole), air rifle, and fabric minnow traps baited with crickets at three sites per treatment across two seasons (fall 2024 and spring 2025) compared to control sites with no removal treatment applied. Each method was applied to each site for two 1-h treatments per season. We observed population reductions at all sites after removal treatments during the warm season. Air rifle was the most effective method in terms of the number of lizards removed, effort, and cost efficiency, but the limited temporal application of 2-h removal effort per site per season was not enough to distinguish population reduction potential among methods. Our calculated prescribed take levels of approximately 30% of the average population estimates are reasonable to achieve based on current routine removal efforts. This information may be applicable to other management efforts on small invasive lizards and may be informative for non-lethal capture efforts of similar lizard species.

Peters’ rock agamas (Agama picticauda) are an invasive lizard established in multiple populations in Florida (Nuñez et al., 2016). Where established, this species is numerous and conspicuous (N. Claunch, pers. obs.). Agama picticauda is an omnivorous lizard with insects as the primary component of its diet (Yeboah, 1982; Enge et al., 2004; Henigan et al., 2019; Ofori et al., 2023). In its native range of southern Ghana, ants, beetles, and grasshoppers form the bulk of its diet (Ofori et al., 2023). This species has been documented cannibalizing smaller individuals (Henigan et al., 2019) and will also forage on flowers (Yeboah, 1982) and discarded human food such as bread (Ofori et al., 2018) and strawberries (Claunch, pers. obs.). Agama picticauda thus has the potential to impact Florida’s native lizard fauna (of which most native species are smaller than A. picticauda), horticulture and gardening efforts, and imperiled insects such as the Miami tiger beetle (Cicindelidia floridana). The ecological impacts of agamas in Florida are not well studied, likely due to a combination of factors including the challenges of assigning impacts to agamas versus other established non-native lizards (e.g., Leicephalus carinatus; Connor et al., 2013), the lag time for impacts to become obvious (Lockwood et al., 2013), or the paucity of data on native insect diversity and abundance (Suanders, 2019) to quantify their impacts, mainly, insectivorous lizards. This species has been documented to cause damage to infrastructure through burrowing behavior and to be a nuisance to private property due to climbing and defecation (Florida Fish and Wildlife Conservation Commission Nuisance Wildlife Database, 2024). Agama picticauda has documented potential for long-distance dispersal (Broxton et al., 2019) and has been reported from 45 out of 67 counties in Florida (EDDMapS, 2025), making the establishment of new populations throughout Florida likely. Due to the wide range of potential and demonstrated impacts, it is important to develop and compare removal efforts for A. picticauda.

Peters’ rock agamas (agamas hereafter) are present in public lands and private residential areas (Florida Fish and Wildlife Conservation Commission Nuisance Wildlife Database, 2024). Thus, there is a need to identify effective removal methods that can be applied in different habitats by both homeowners and professionals. There are multiple potential removal methods for agamas (Table 1), but the effectiveness of these methods has not been formally evaluated. Agamas are swift lizards with a high propensity to flee (Mitchell et al., 2021), making the pursuit of individuals difficult. It is likely that multiple methods in concert are needed to achieve effective removal. This is due to the limitations of each removal method and individual variation in response to removal efforts (e.g., trap-shy individuals).

As a key example, glue boards are effective at catching agamas as an active trapping method (Henigan et al., 2019; Dzikowski et al., 2019). Their use requires placement in anticipated escape routes and being monitored closely to prevent bycatch and ensure that both glue boards and any captured lizards are removed immediately. When used as a passive trapping method (as advertised by the manufacturer as a “set and forget” method), glue boards are inhumane and have high potential for bycatch (Reiss et al., 2022; De Ruyver et al., 2023). Several other methods exist for lizard removal applications. Air rifle use is an effective method that eliminates the potential for non-target bycatch when performed by wildlife professionals (Lovich and Bauer, 2012). The use of air rifles requires trained shooters able to hit a small and fast-moving target. However, air rifles cannot be applied in all areas due to safety concerns and legal restrictions in certain municipalities. In addition, air rifles may cause additional damage when infrastructure serves as a backstop. Lizard-fishing is a capture method where a bait item (usually an insect) is secured to the end of a line attached to a pole to attract the lizard (Durden et al., 1995). The lizard then bites down on the insect, and the user swings the lizard into a bucket or similar receptacle for capture (Durden et al., 1995). There are variations on the method: use of extendable crappie rods with insects secured near the end versus traditional fishing rods with reeled lines and use of baited hooks when the captured animals are intended for lethal removals. Lassoing is an active method that involves pursuing a lizard with a lasso secured at the end of a stick or fishing rod tied from a string, floss, or fishing line (Lovich and Bauer, 2012). While some commercially available traps are marketed for lizards (e.g., IggyTrap of Florida, LLC, Boca Raton, FL, USA), these traps are often limited in availability and may be limited in applicability for smaller species such as agamas. Collapsible fabric minnow traps have some utility for catching fast-moving lizard species such as ameivas and agamas (C. Goodman, University of South Florida, pers. comm.). The fabric nature makes these traps less likely to overheat compared to metal traps, and lizards may be more likely to enter them as shelter in open areas. Here, our main aim is to test the best and most effective methods to remove these lizards.

For this study, we selected three removal methods: trapping, lizard-fishing, and air rifle removals. We selected these methods because the U.S. Department of Agriculture Wildlife Services (USDA WS) and the United States Army Corps of Engineers (USACE) have successfully employed air rifle removals; lizard-fishing may be possible to achieve with little training; and trapping may scale to larger removal efforts. Both fishing and trapping may be perceived more positively by the public and may be feasible as homeowner control measures with minimal collateral damage/bycatch potential. We evaluated the effectiveness of methods in four ways: by estimating the numbers removed and the proportion of population removed, the catch per unit effort (CPUE), and the costs associated with each method. We extrapolated the costs to approximate the costs of conducting a larger removal effort with each method. Finally, we estimated the prescribed take level (PTL) for agamas for each method using the best available data. PTL is a simulation method grounded in fundamental aspects of harvest theory derived from the standard and generalized population logistic model (Wade, 1998; Runge et al., 2009). The method uses maximum population growth rate (r max), population size (N), and a recovery factor (F) that, in the case of a non-native invasive species, can help determine harvest rates required to reduce the population and maintain abundance at an acceptable level when eradication is not a feasible management option.

Sites and study design

We conducted our study in collaboration with the USACE on the Saint Lucie Lock and Dam and Herbert Hoover Dike, areas with large numbers of agamas (Figure 1). We selected 12 sites from existing structural features inhabited by agamas: riprap (n = 4), rock piles (n = 4), or water control structure rocks (n = 4). We demarcated a 600-m 2 area for each site, as agamas are highly territorial, and this represents nearly double the published territory size (Yeboah, 1982). To prevent the influence of treatment application methods on nearby sites, each site was also separated by a minimum of 600 m 2 that served as a buffer zone where no counts or control measures were applied.

We applied fundamental aspects of a before-after-control-impact (BACI) design (Smokorowski and Randall, 2017), conducting a multiple time series study with and without interventions to evaluate different removal methods. To estimate abundance, we conducted repeated counts of agamas and used N-mixture models (Royle, 2004a, b). We counted agamas by conducting searches at a fixed distance around the perimeter of the 600-m 2 area footprint of each site for 5 min per site. Perimeter counts were conducted from a vehicle to prevent agamas from fleeing, and binoculars were used to confirm species. Each site was surveyed three times to estimate detection probability and abundance. All counts were conducted alone by the same observer between 09:00 and 10:00. Because WS operations have ongoing invasive reptile removal contracts in the area, we employed a non-intervention buffer period of at least 2 months, where no agama removal methods were applied prior to beginning site counts for each site and associated buffer zone.

Sites were assigned to one of the three removal treatments or control (no removal), with three replicates of each assignment. To avoid bias in site characteristics influencing interpretation of results, site types (i.e., riprap, rock pile, water control structure rocks) were assigned strategically to ensure even distribution of site types across treatments. We applied only one treatment per site, per day, for a total application period of 1 h, from 09:00 to 10:00, such that each site received one treatment application during a total 2-week period. Treatments were then repeated in a second 2-week period. Treatment order was assigned randomly across all the sites within each 2-week period. Following the completion of all removal treatments, 5-min fixed-distance perimeter counts were conducted at each site following treatments as described above. We conducted these efforts in two seasons: warm season (September–November 2024) and cold season (January–April 2025). We did not conduct counts or removals during rain events, as agamas retreat and are not observed or available for removal during rain (Claunch, pers. obs.). Instead, schedules were delayed until the next fair day without rain.

We collected environmental covariate data (cloud cover, ambient temperature, windspeed, humidity, barometric pressure) using a handheld Kestrel 3500 Weather Meter (Nielsen-Kellerman Company, Boothwyn, PA, USA) immediately before conducting counts or removal treatments. We collected morphometric data (sex, snout-vent-length, weight, reproductive characteristics) on removed agamas that were retrievable. Individuals were dissected and examined for testis size and status of ductus deferens (presence of sperm), ovarian follicle sizes, status of oviduct (distended), and presence of oviductal eggs. For agamas that were not retrievable (e.g., agamas hit with an air rifle pellet that fell into a deep crevice), sex and age class were estimated based on the size and coloration of the lizard.

Removal methods

Our three removal methods were air rifle, lizard-fishing, and trapping. Due to the risk of improper application of glue boards by the public (e.g., as a passive trapping method), we opted to not evaluate the use of glue boards in this study. We used an air rifle (Marauder, Benjamin Airguns, St. Louis, MO, USA) with.22-caliber Terminator pellets (Haendler & Natermann Sport GmbH, Hannoversch Münden, Germany) to lethally shoot agamas via pellet to the head. Air rifle use was limited to USDA Wildlife Services personnel certified in the safe and lawful use of rifles. Death was confirmed by pithing with an awl. For lizard-fishing, we used a 16.5-foot extendable crappie pole (Black Widow, B’n’M Pole Company, West Point, MS, USA) affixed with a length of approximately 120 cm of 1-mm waxed whipping twine (Robline, Teufelberger Fiber Rope GmbH, Wels, Austria) followed by 30 cm of a 14-lb (6.3 kg) test monofilament fishing line (Sufix Promix, Rapala VMC Corporation, Minnetonka, MN, USA) attached to a kirbed-point bait hook (size 10 Carlisle, O. Mustad & Son Americas Inc, Miramar, FL, USA), baited with a live adult cricket. Lizards were identified and approached slowly as the cricket was slowly “cast” and allowed to crawl on a surface near the agama until it bit on the bait. Once bitten, the pole was pulled to attempt to hook the lizard. If captured, the lizard was immediately removed from the hook and euthanized by pithing. For trapping, we deployed four collapsible fabric minnow traps: two of each umbrella-style (PuDong, Yiwu PuDong Trade Co., Ltd., Yiwu, Zhejiang, China) and hexagon-style traps (12 holes “The Magic Foldable Fishing Trap,” Fauppt, Shenzhen Fanhang E-commerce Co., Ltd., Shenzhen, Guangdong, China). Traps were weighted with a rock inside and baited with approximately 25 crickets dispersed from a bait container (Cricket Tube, Danielson Company, Inc., Auburn, WA, USA) with the narrow-ended cap removed, allowing the crickets to move around within the trap. After 1 h of deployment, the observer returned to the traps to remove any captured lizards. Captured lizards were euthanized by pithing following removal from the traps. This included euthanasia of any non-native non-target lizards (e.g., curly tailed lizards, basilisks) to remain in compliance with state laws that prohibit their release and ongoing contractual obligations by the Wildlife Services Operations to remove these species from the site.

Data analysis

We compared the estimated abundance of Agama at sites with different treatments applied using a BACI design. To estimate the population of agamas at each site, we used the program PRESENCE (version 2.15.15; Hines, 2006) to evaluate our count data with the Royle Poisson binomial N-mixture models (Royle, 2004b). We performed a separate model for each season (warm pre-removals, warm post-removals, cold pre-removals, cold post-removals), as our data were too sparse for multiseason models. All sites were confirmed to be occupied with agamas prior to starting counts and assigning treatments. We assumed no immigration or emigration during each count period. To keep consistent with our sampling design, we used the same model for each season, where abundance (λ) was allowed to vary by treatment and detection probability (r = observer detectability of an available agama × availability for detection) was allowed to vary by site visit. Given our small number of sites and counts, we only considered the most basic Poisson binomial N-mixture model fitting our experimental design (without covariates). We extracted population estimates for each removal treatment, calculated the density per treatment group by dividing the abundance estimate by the combined area of sites per treatment group (1,800 m 2), and assessed differences in density among treatments and time periods using two-tailed unpaired Z -tests.

We calculated CPUE per season by dividing the total number of agamas removed by the total removal effort for the season (6 h per season per method). We opted not to correct the CPUE estimates for individual traps deployed (we deployed four traps per application), because many traps can be deployed and checked by an individual observer in the same hour, as that observer would have focused effort using other active methods (lizard-fishing and air rifle). We compared CPUE among removal methods using the Kruskal–Wallis test, followed by post hoc comparisons with the Dunn test.

We calculated total and disposable costs for each method. Total costs included personnel salary, equipment costs, and disposable costs. We estimated personnel costs for the salary of one biological technician (U.S. General Schedule pay rate GS-5-5) performing each removal application method for 2 h per treatment per season. The upfront equipment cost for each treatment is approximately $610 for the air rifle, $75 for traps, and $30 for the extendable crappie pole, line, and hooks for lizard-fishing. We did not include travel or vehicle maintenance costs in our estimate of total cost. Disposable costs included pellets ($0.06 per pellet) and crickets ($0.05 per cricket). Air rifle pellets and crickets baited on hooks were only expended when a target lizard was identified, whereas traps were always baited with approximately 25 crickets per trap.

We used the total cost and total cost-per-lizard data to calculate potential costs of steady removal applications with similar removal rates over a calendar year. Because our complete removal data already account for some variability across seasons, we assumed a constant removal rate per method and 12 h of effort a week over a calendar year to calculate the cumulative number of lizards removed. We calculated the cumulative total cost at each timestep (each week) as:

Where t = the timestep (in this case, week), fixed cost reflects the cost of equipment at the start of the project for each method, disposable costs reflect varying disposable costs associated with each method (e.g., crickets, pellets), and salary reflects the cost of the employee’s time associated with each timestep (in this case weekly salary). Then, we calculated the cumulative cost per lizard at each timestep (each week) as:

Where we input the number of lizards removed (n removed) associated with each method and cumulative costs at the relevant timestep as above. We calculated the PTL of the population following the equations of Runge et al. (2009):

Where the F o represents the management objective, set F o = 2 to reflect the goal of reducing the population size of agamas by maximizing take to above maximum sustainable yield; r max is the maximum growth rate expected at low population densities (Runge et al., 2004); and N is the population size. We assumed linear density dependence under the standard logistic model (Runge et al., 2009) because we do not have information about density dependence in this species. The tildes (~) over N and r max denote that we used estimates generated from statistical distributions for these variables. We estimated r max using two methods. First, we used a simplified Euler equation, estimating the finite growth rate of a discrete-aged breeding population in a stable age distribution in Slade et al. (1998):

Where l is the survival to maximum longevity, p is the yearly probability of adult survival, λ is the finite growth rate, l α is the pre-reproductive survival, b is a fixed fecundity value, α is age at first reproduction, and ω is the maximum longevity of reproductive adults. We used the parameter assumptions of the λPM model from Slade et al. (1998), where pre-reproductive survival (l α) was taken from an observed value, adult survival (p) estimated from the mean survival of adults, and survival to maximum longevity (l) calculated as:

Second, we used the demographic invariant method (DIM; Niel and Lebreton, 2005), which is often used for species where life history data are not well described and require only two input variables—age at first reproduction (α) and adult survival (s):

We estimated life history parameters based on the literature on invasive populations of A. picticauda in Florida and congeners in the native range (A. agama), which may, in some cases, represent A. picticauda based on genetic evidence (Leaché et al., 2016). To account for uncertainty associated with applying congeneric data to a different species, applying native range data in an invasive range, and little knowledge of life history parameters in the invasive range, we used available data to generate distributions for each input variable using the Monte Carlo (MC) simulation method [denoted by a tilde (~) over each variable]. Using the lower and upper estimates available for each variable, we generated estimates assuming a normal distribution using 10,000 iterations in the program R v. 4.5.0 (R Core Team, 2025) using package fitdistrplus v. 1.2-4 (Delignette-Muller and Dutang, 2015). We set adult survival probability (p in the Euler equation and s in the DIM) between 0.759 and 0.899 based on a mark recapture study of A. agama that examined 4-month survival probability of adults of both sexes (Luiselli et al., 2011). Survival to reproductive age l α was calculated between 0.125 and 0.65 based on a common garden study of A. picticauda hatchlings incubated at different temperatures raised to near-adult sizes (Steele and Warner, 2020). We estimated fecundity (b) by multiplying MC generated estimates of clutch size and number of clutches by a fixed sex ratio of offspring. We estimated clutch size between 4 and 13 based on clutch sizes reported for A. picticauda in Florida in Blunden and Krysko (2007) and necropsy data collected by NC. Although there is no concrete evidence, the number of clutches per year for A. picticauda in Florida is posited to be up to 3 (Blunden and Krysko, 2007); therefore, we estimated values between 1 and 3. This species has temperature–sex-determination (Charnier, 1966; Steele et al., 2018); to simplify the model, we assumed an average sex ratio of 0.5, as information about A. picticauda nesting ecology is lacking in either range. Age at first reproduction (α) was estimated between 0.6 and 0.87 years based on a minimum of 219 days for females to reach a reproductive size (74 mm, Steele and Warner, 2020) confirmed by distended oviducts in a 74-mm specimen collected in Martin County, Florida (Claunch, unpubl. data) and a minimum of 244 days for males to reach a subadult size of 75–85 mm (Steele and Warner, 2020). The smallest reported male in Florida observed in adult breeding colors, captured in this study, is 100 mm SVL. This specimen had a testis size of 7 mm, within the range of reproductively active A. agama adults reported in Daniel (1960) in the native range. The age at last reproduction (ω) was estimated between 1.2 and 8 years based on longevity data from a congeneric species in the wild, Agama hartmanni (Flower, 1925 in Carey and Judge 2000), and posts in a reptile forum discussing captive longevity of A. agama (ReptileForums.co.uk, 2010).

For each method, we converted from λ using r max = ln(λ max). We calculated r max from each method using 10,000 iterations from the estimates generated for each input parameter. We then generated a combined estimate for r max from 10,000 iterations of the average of Euler and DIM distributions to calculate PTL. To account for uncertainty in our population estimates (N), we similarly generated estimates from the lower and upper agama population estimates from this study using a log-normal distribution (Dillingham and Fletcher, 2008) as described above.

We removed 37 agamas over 36 h of effort using three removal methods (n = 28 during the warm season, n = 9 during the cold season; Figure 2). Considering total lizards removed, the air rifle method was most effective (n = 27), followed by traps (n = 6), and lizard-fishing (n = 4). Fewer lizards were captured in the cold season for all methods, with eight lizards removed by air rifle, one lizard removed by lizard-fishing, and no lizards removed using traps. Ambient temperature during counts and removals ranged 21.5–34.3°C, mean = 26.9°C during the warm season and 13.0–29.9°C, mean = 21.9°C during the cold season.

There were no significant differences in the density of agamas among removal sites versus control sites prior to removal efforts (p > 0.7), indicating similar starting densities of agamas among treatment assignments at the start of the project (Figure 3). The density of agamas was reduced by nearly half after removal methods were applied in the warm season (z = 2.31, p = 0.04). Within the warm season, a reduction in abundance was observed in all groups after treatments were applied: control (z = 2.16, p = 0.03), air rifle (z = 2.00, p = 0.05), trap (z = 2.39, p = 0.02), and fishing (z = 1.86, p = 0.06), although fishing was marginally non-significant at alpha = 0.05. Within each season, densities were not significantly different among the removal treatment groups versus the control group. We did not detect a difference in agama density between the warm post-removal counts and the cold pre-removal counts, during which no removal treatments were applied (z = −0.66, p = 0.51). We also did not detect a difference in density between cold pre-removal counts and cold post-removal counts (z = 0.13, p = 0.90), and this was true within each treatment group.

The CPUE estimates per method per season are presented in Table 2. While CPUE did not differ significantly by season (χ 2(1) = 2.15, p = 0.14) or site type (χ 2(1) = 0.49, p = 0.78), we detected a difference in CPUE by method (χ 2(1) = 6.06, p = 0.048), with air rifle exhibiting greater CPUE than both fishing (p = 0.028) and trap methods (p = 0.04).

Although having the highest total cost, air rifle was the cheapest method when considering disposable costs per lizard and total costs per lizard removed (Table 2). Air rifle was also the cheapest method for total disposable costs per lizard within a season. Lizard-fishing and traps had comparable total costs, with a lower total cost-per lizard removed with traps, but a lower disposable cost per-lizard removed with lizard-fishing (Table 2).

When extrapolating the cost of removal across a longer time span, the total cost per lizard is expected to decrease from our initial estimates for each method. Within 17 weeks, the cost per lizard reduces to approximately $14–15 with the air rifle method, while lizard-fishing stalls approximately $93 per lizard within 8 weeks, and traps approximately $72 within 9 weeks (Figure 4).

Based on our input parameters, we estimated r max to be 0.25 ± 0.06 using the Euler equation and 0.41 ± 0.04 using the DIM, resulting in a combined estimate of 0.33 ± 0.09. The yearly PTL per 600-m 2 site was estimated at 30.94 ± 12.52 or approximately 30% of the average estimated population per site. Within a treatment group (three sites), this equates to PTL of 92.82 ± 37.56 (Figure 4) and a PTL of 378.28 ± 150.24 lizards per year across all study sites combined (a total area of 7,234.6 m 2).

In all groups, we observed a reduction in agama population density during the warm season after the removal treatment period, and this reduced population density was unchanged through the end of the cold season. This is corroborated by the fact that more agamas were removed during the warm season compared with the cold season. When considering the results of the abundance models alone, similar reductions were observed regardless of the treatment group. It is important to note that our study design is limited, as only 12 h of effort was applied within a season per treatment group (2 h per site per season). This limited temporal application of our removals may obscure our ability to understand the population reduction potential of each method. In practice, removal efforts are often longer in application with more regular site visits (several hours across several weeks at each site). Thus, we cannot make a definitive recommendation on the population reduction potential of each method using abundance modeling alone.

Our abundance data highlight some important aspects of A. picticauda ecology. Our observations of stable population estimates from the warm post-removal period through the cold pre- and post-removal period indicate that there was no substantial recruitment or deaths of agamas during this period. Although reproductive life history data are generally lacking in this species in Florida, hatchling season is estimated from May to August based on previous investigations of this species (Enge et al., 2004). Our necropsy data of removed agamas corroborate a lack of recruitment during winter (of 22 individuals necropsied, oviductal eggs were only detected in a single individual on October 3). If substantial deaths occurred due to cold snaps (e.g., Mazzotti et al., 2016), we would expect an observed reduction in density between the warm post-removal and cold pre-removal periods. In January, nighttime temperatures occurred as low as 3.9°C at our sites. Our data suggest that, at a population level, agamas appear resistant to these acute changes in temperature at the sites evaluated. The measured cold tolerance of agama in the Miami area averages 9.77°C, after which they are able to recover (Mothes et al., 2019). Agamas may be most vulnerable to population reduction during the end of the warm season (October–November). An expanded application of our study methodology (e.g., removal treatments applied for a longer treatment duration) may reveal differences in population reduction potential.

When considering the number of agamas removed, more agamas were removed during the warm season than the cold season. Based on our abundance models, fewer agamas were present in the population during the cold season. The reduced number of removed lizards in the cold season may also reflect a behavioral change at lower temperatures, such as reduced willingness of lizards to pursue bait in lower temperatures (Cabanac, 1985). This difference may also reflect a lower availability for removal. For example, if lizards are more likely to shelter in burrows or other microsites at low temperatures (Cecchetto et al., 2019), there would be fewer available for detection or subsequent removal by any of our methods.

Regardless of season, the greatest number of agamas was removed using the air rifle method. Air rifle is the current method preferred for removals by the WS Operations personnel. While fabric minnow traps are used by other groups attempting to capture agamas (Suarez, University of Florida, pers. comm.; Goodman, University of South Florida, pers. comm.), fishing methods do not seem widely applied in a removal context. We did not capture enough agamas with our traps to distinguish efficacy between the umbrella or hexagon configurations. There were some difficulties noted with the fishing and trap methods. During fishing removal treatments, agamas were sometimes observed but not captured due to wind complicating the casting and aim of the baited hook, lizards being unwilling to approach the baited hook, or the lizard removing the cricket without being hooked (i.e., eating half of the cricket). This occurred with both adult and juvenile agamas. During trapping removal treatments, agamas were sometimes found sitting on top of the trap or facing the trap from the outside, attempting to get to the crickets. This suggests that though agamas may be interested in the traps due to the bait within, they are not able to easily locate the entrances. This warrants future investigation into trap modifications that may improve efficacy.

Based on our removal results, it is unsurprising that the CPUE reflects a similar trend, with air rifle having the highest CPUE regardless of season. Traps had a higher value for CPUE than lizard-fishing, but only in the warm season, as no lizards were trapped during the cold season. It is possible that CPUE could be increased with a greater number of traps deployed, as more than four traps could be managed by a single observer within a 1-h time span. However, increasing trap density would also increase costs. While the per-item disposable costs were nearly equal for pellets and crickets, actual disposable costs were the highest for traps because they were always baited with 25 crickets per trap. Ultimately, the air rifle method was the cheapest when considering the disposable and total costs per lizard removed. When considering potential removal applications across a longer timescale at 12 h of effort per week, the costs per lizard are reduced, and air rifle remains the most cost-effective method. Our estimates of potential cost are coarse, as these estimates do not account for population change associated with continuous removal. They may better reflect the potential costs of removal methods at a larger scale, for example, across multiple sites rather than revisiting the same site every week.

Our estimate for PTL of agamas is not unrealistic to achieve. During routine removal of agamas outside of this study, our observer (ML) has removed 67 agamas across several sites within a single day and removed 34 agamas across 3 similarly sized sites within an hour using the air rifle method. Based on our CPUE estimates, achieving a three-site PTL of 92.82 ± 37.56 would take approximately 41.25 ± 16.69 h of effort with air rifles, 278.49 ± 112.69 h with lizard-fishing, and 185.64 ± 75.12 h with traps at the removal rates measured in this study. In practice, the effort would likely be reduced. It is important to note that our estimates of PTL are based on published data that include estimates from congeners in the native range of A. picticauda. The r max estimates from the DIM, which only rely on two life history metrics, are 1.67 times higher than the r max estimates using the Euler equation. Our PTL estimate, calculated from an average of these two methods, may thus be refined with a comprehensive study of reproductive and population ecology of A. picticauda in its invaded Florida range, to include mark-recapture studies, observational study, and necropsies of lizards across the year to determine survival rates, clutch sizes, number of clutches per year, sex ratios, and sizes and ages at first reproduction. Comprehensive demographic life history data would be useful to parameterize integrated population models for more accurate estimates of r max and PTL.

Overall, we may observe differences in removal success rates and, subsequently, costs, when applied at a larger scale or in different areas. Likewise, a larger population monitoring study with a gradient of habitats with an increased number of counts per season could also alter abundance estimates and, in turn, PTLs. For example, we may improve estimates with a study conducted within the warm season that leverages 15 urban and 15 non-urban sites with 12 h of removal effort applied per site during the removal periods and 5–10 counts at each site before and after treatments. We used a single observer to apply all three techniques. At scale, different observers may have different results based on individual technique (Hubert and Fabrizio, 2007) such as differences in aiming for air rifle or posture/“casting” for lizard-fishing. Our sites in this study have limited access by people and have little to no structural change over time. It is difficult to extend our results to other areas of agama establishment because we may observe different abundances or removal rates in more populated areas. As one example, in an urban area, agamas may have more frequent exposure to people and face a more structurally dynamic environment, which may influence their habituation toward people or neophobia toward new objects (Carter et al., 2012). Agamas generally have high flight-initiation distances (Mitchell et al., 2021). If agamas are more habituated toward people, they may be easier to approach using the lizard-fishing method (Carter et al., 2012). If they are less fearful of new objects, they may be more willing to explore and enter traps (Carter et al., 2012). Furthermore, the removal data informing our CPUE and cost estimates are limited from our efforts during a standardized time period, from 9 to 10 a.m., regardless of season. It is possible that the efficacy of methods changes throughout the day based on thermoregulation behavior, influencing the availability of lizards for removal (Rangel-Patiño et al., 2020). Thus, it would be prudent to test methods across a gradient of habitat types with different observers at longer temporal scales to generate more general recommendations for management of Peters’ rock agama across its established range in Florida.

We conducted a study evaluating the effect of three removal methods (air rifle, lizard-fishing, and traps) on A. picticauda populations compared to non-removal sites and evaluated the efficacy of these methods in terms of the numbers of lizards removed, catch per unit effort, and cost. We calculated the prescribed take limits for agamas based on our findings combined with existing data on reproductive parameters for this and related species. Our results demonstrated that air rifle was the most effective method in terms of the number of lizards removed, effort, and cost efficiency, but the limited temporal application of 2 h of removal effort per site per season was not enough to distinguish population reduction potential for each method. Our data show promising results for removal methods of agamas but are limited in scope due to the size and location of the study. To improve management recommendations for this species, a study similar to ours could be conducted at a larger geographic scale, with an increased number of counts per season to include more habitats with increased effort for each removal method. Additionally, estimates for PTLs would be improved by a concentrated study of agama ecology in Florida to refine survival and reproduction estimates to inform model parameters. Finally, based on our experiences, we recommend evaluating alternate trap types and modification of existing traps such as the addition of ramps or alternate lures for efficacy.