Modeling the Effects of Commercial Harvest on Population Growth of River Turtles
Abstract
Commercial turtle harvest is considered one of the major contributing factors to declines in turtle populations. Few long-term studies have evaluated turtle population response to harvest and little is known about demographic rates for many turtle species. We gathered demographic rates from the literature for snapping turtles (Chelydra serpentina), smooth softshells (Apalone mutica), and spiny softshells (Apalone spinifera), which are harvested in Missouri, and developed deterministic, density-independent, stage-based matrix models to assess turtle population response to plausible harvest rates we estimated from field sampling. Further, we used population modeling to determine annual harvest rates that would result in λ = 1 for each demographic scenario. We developed one model for snapping turtles and another for both softshell species combined due to the lack of available species-specific demographic data for either softshell species. Using mean demographic rates for survival and fecundity, snapping turtle populations had a growth rate (λ) of 1.023; at minimum demographic rates λ = 0.891 and at maximum demographic rates λ = 1.199. When we applied plausible, field-estimated annual harvest rates under mean demographic rates, populations decreased in all instances except when harvesting only juveniles at the minimum harvest rate. At mean demographic rates, annual harvest of both adults and juveniles should be ≤ 2.3% to maintain a stationary population (λ = 1). For softshell turtles, λ was 0.952 at mean demographic rates, 0.838 at minimum demographic rates, and 1.163 at maximum demographic rates. Under mean and minimum demographic rates, no field-estimated harvest could be sustained, as any annual harvest rate resulted in λ < 1. For both snapping turtles and softshells, harvest was sustainable when demographic rates were at the maximum values, which are highly improbable to occur frequently: annual harvest of 16.3% of both adult and juvenile softshells and 18.6% of adult and juvenile snapping turtles resulted in λ ≥ 1. In both species, elasticity analyses demonstrated that adults, which are the most vulnerable to commercial harvest, were the most important segment of the population demographically. These results corroborate the findings of other studies which indicate that even low annual harvest rates may have detrimental effects on the long-term sustainability of turtle populations at localized scales.
Commercial turtle harvest is an important cause of turtle population declines (Ceballos and Fitzgerald 2004; Moll and Moll 2004; Schlaepfer et al. 2005), though little is known about the long-term demographic effects of harvest (Congdon et al. 1994). Because of their life-history characteristics and demographic rates, such as high adult survivorship, low nest and hatchling survivorship, and late age at first reproduction, turtle populations are not highly amenable to sustainable commercial harvest (Crouse et al. 1987; Congdon et al. 1993, 1994; Crouse and Frazer 1995; Galbraith et al. 1997; Heppell 1998; Zhou and Jiang 2008). For the snapping turtle (Chelydra serpentina), annual survivorships of 0.88 to 0.97 have been reported (Galbraith and Brooks 1987; Congdon et al. 1993, 1994), and maintaining this high annual adult survivorship is among the most important factors contributing to long-term population sustainability (Congdon et al. 1994). However, adults are also the most desirable from a harvest standpoint because they are often sold by weight on the food market (Brown et al. 2011).
In North America, for the few turtle populations with available long-term life-history and demographic data, both field data and population modeling suggest that small increases in annual mortality can be detrimental to population sustainability (e.g., Brooks et al. 1991; Congdon et al. 1994). Painted turtle (Chrysemys picta) populations in Minnesota are susceptible to overharvest when 4%–5% of females are removed (Gamble and Simons 2003). Similarly, a 10-yr study of an ornate box turtle (Terrapene ornata) population in Wisconsin indicated that population declines may occur if total annual adult mortality exceeds 5% (Doroff and Keith 1990). Further, there is no evidence that turtle populations exhibit density-dependent reproductive responses when removals occur (Brooks et al. 1991). Thus, these turtle populations are unable to adequately compensate for losses owed to commercial harvest through increased reproduction (Congdon et al. 1994).
In Missouri, the snapping turtle (Chelydra serpentina), the smooth softshell (Apalone mutica), and the spiny softshell (Apalone spinifera) are the 3 turtle species subject to commercial harvest. Commercial turtle harvest may take place year round with no size or bag limits; the only restrictions are the type of net used and the waterways in which turtles may be captured (e.g., the Missouri, Mississippi, and St. Francis rivers). Currently in Missouri, there are no data available on population size or structure for these species and no indication of the effects of commercial harvest on population sustainability.
Given the potential sensitivity of turtle populations to harvest (e.g., Congdon et al. 1994; Gamble and Simons 2003), it is important to address how harvest could affect population growth rates for snapping turtles and softshells in Missouri. Our objective was to assess the effects of commercial turtle harvest using plausible harvest rates estimated from our field research (Zimmer 2013). Using demographic rates in the literature for the 3 target species, as well as harvest rates estimated during our 2011 and 2012 field seasons (Zimmer 2013), we modeled population growth and the effects of commercial harvest on turtle populations under mean, minimum, and maximum survival and fecundity rates using female-only stage-structured matrix models (Lefkovitch 1965). Additionally, we determined the rate of annual harvest that did not result in population declines (i.e., harvest rates that resulted in λ ≥ 1) for populations under mean, minimum, and maximum demographic rates. We also conducted sensitivity and elasticity analyses to determine which demographic parameters were most important to population growth rates (λ) and examined the implications of these analyses for conservation of harvested turtle populations.
METHODS
Literature Review
We conducted a literature review to obtain demographic rates for use in matrix population models for the following parameters: nest survival, hatchling survival, juvenile survival, adult survival, duration (years) of the juvenile stage class, clutch size (number of eggs in a single nest), breeding frequency (the proportion of females reproducing annually), and number of clutches produced annually. We used several filters to select data for this analysis. We used data from the literature when sample sizes were greater than 10 turtles or nests. Additionally, we excluded demographic rates that were estimated under unusual circumstances (e.g., decreased adult survivorship resulting from an abnormal increase in predation; Brooks et al. 1991). Because the models we constructed were female-only, we only included adult survivorship rates that were either female-specific (e.g., Congdon et al. 1994) or reported as equal between the sexes (e.g., Paisley et al. 2009). Where authors reported a mean value for a demographic rate, we used the mean value rather than a range (e.g., Congdon et al. 1994 reports nest survivorship over 17 yrs to range from 0% to 64%, mean 23%). Because rates of nest survivorship were sparsely reported in the literature, we also included the reported proportions of known nests not destroyed by predators, a commonly cited source of mortality for nests (Plummer 1976; Robinson and Bider 1988).
Data Summarization
We divided snapping turtle data (Table 1) from the softshell data (Table 2) but combined softshell species due to the paucity of demographic data for both smooth and spiny softshells. Additionally, in 2 instances we used rates reported for the snapping turtle where softshell data were unavailable (i.e., hatchling survivorship and breeding frequency). We included our estimates of abundance and harvest rates (Zimmer 2013) in our data summary.
Using these data we calculated mean, minimum, and maximum values for snapping turtles and for the softshell turtles combined (Table 3) to represent demographic rates under mean, minimum, and maximum demographic conditions. To estimate mean, minimum, and maximum fecundity (F) for each species, we used the following variables: nest survival, clutch size, breeding frequency, adult survival, and number of clutches produced annually:
where number of clutches = number of clutches laid annually; clutch size = number of eggs laid in a single nest; breeding frequency = proportion of adult females reproducing annually; adult survival = annual survival of the adult female stage; and ŜN = survival of nests (i.e., the proportion of eggs surviving and transitioning to the hatchling stage). The value 0.5 assumes a 1:1 sex ratio at birth for all 3 species (Graham and Graham 1997).
In this stage-structured model with annual increments, the model treats turtles within a stage as having a probability of surviving and transitioning to the nest stage each year and a probability of surviving but remaining at the current stage. To calculate the probability that a juvenile survives and remains in the current stage, we used the following equation (Crouse et al. 1987):
where Pi = the proportion of individuals surviving and remaining within stage class i; pi = the survival of stage class i; di = the duration stage class i, i.e., the number of years spent in that stage class. Similarly, including the above variables pi and di to calculate the juvenile survive and transition value, we used the following equation (Crouse et al. 1987):
where Gi = the proportion of individuals surviving stage class i and transitioning to the next stage class. For snapping turtles, the mean duration of the juvenile stage was estimated to be 8.5 yrs based on averaging juvenile duration values from the literature (Christiansen and Burken 1979; Congdon et al. 1987). For softshells, the mean juvenile duration was 6.5 yrs (Johnson 2000; Ernst and Lovich 2009).
Matrix Modeling
To examine population growth of snapping turtles and softshells in the absence of harvest, we developed female-only, density independent, deterministic 3 × 3 stage-based matrix models (Lefkovitch 1965; Skalski et al. 2005) with stages defined as hatchling (age 0), juvenile (age 1 to adulthood; Tables 1 and 2 indicate specific values for each species), and adult (reproductive). The general model used had the following form:
where Fi is the fecundity of the stage class i (calculation described above). We developed 6 separate stage matrix models for estimating population growth without harvest, 3 for snapping turtles and 3 for softshells using the mean, minimum, or maximum demographic values (Table 3; Figs. 1 and 2).
Citation: Chelonian Conservation and Biology 13, 2; 10.2744/CCB-1109.1
Citation: Chelonian Conservation and Biology 13, 2; 10.2744/CCB-1109.1
We determined the stable stage distribution (i.e., the proportion of hatchlings, juveniles, and adults in the population when population growth, λ, stabilized) for each population and applied it to the mean value of our abundance estimates per 2 km for unharvested populations of snapping turtles and softshells (Zimmer 2013) to generate initial vectors of abundance. For snapping turtles, mean abundance was 90 turtles per 2 km. For softshells, mean abundance was 62 turtles per 2 km. These abundance values were used with the proportion of turtles in each stage at the stable age distribution to calculate the initial abundance vector (Figs. 1 and 2). Thus, we assumed our models were applicable to the 2-km scale, which represents localized harvest as observed in Missouri rivers.
Harvest Matrices
We applied harvest matrices using harvest rates estimated from mock harvests conducted in the 2011 and 2012 field seasons (Tables 1 and 2; Zimmer 2013). These harvest rates were based on our estimates of abundance (per 2 km) and capture data from mark–recapture studies conducted in the same years and indicated plausible proportions of the population that we were able to “remove” at a 2-km scale while mimicking commercial harvest procedures (Zimmer 2013). From these plausible harvest proportions, we calculated a mean and found the minimum and maximum harvest rates for snapping turtles and for softshells (Table 3). We applied these mean, minimum, and maximum harvest rates to each of the 6 stage matrices described above to examine the effects of varying annual harvest rates (hereafter referred to as λHmean, λHmin, or λHmax) when applied to populations exhibiting mean, minimum, or maximum demographic rates. Harvest matrices were constructed in the following form (Skalski et al. 2005):
where h is the rate of escapement or harvest survival for each stage, i. We examined the effects of mean, minimum, and maximum annual harvest rates for 3 harvest scenarios: 1) harvest of adults only, 2) harvest of juveniles only, and 3) harvest of both adults and juveniles, and then we applied these to populations exhibiting mean, minimum, or maximum demographic rates.
Ultimately, we examined 30 population growth situations for both snapping turtles (Table 4) and softshells (Table 5): unharvested population growth at (mean/minimum/maximum) demographic rates, and then population growth under (mean/minimum/maximum) demographic rates for populations harvested at (mean/minimum/maximum) harvest rates, for each of the 3 harvest scenarios (adult-only/juvenile-only/adult-and-juvenile) using our plausible field-estimated harvest rates (Table 3; Zimmer 2013). Finally, we determined the annual harvest rate that could be sustained for each unharvested population growth scenario in order to maintain λ = 1 (i.e., the sustainable annual harvest rate).
Sensitivity and Elasticity Analysis
We conducted sensitivity and elasticity analyses to determine which demographic parameter was most important to population growth. We calculated sensitivity (s) following methods described by Caswell (2001):
where vi is the ith element of the reproductive value vector (v) or the reproductive value specific to each class (the left eigenvector of the matrix), and wj is the jth element of the stable stage vector (w) or the proportion of individuals in a given stage at stable stage distribution (the right eigenvector of the matrix). Similarly we calculated elasticity, e, or the proportional sensitivity of λ to a change in a specific demographic parameter following methods described by Crouse et al. (1987):
where aij is the matrix element for which we are examining how λ is affected upon changing this value and sij is the sensitivity of λ to a change in that value (calculated above).
RESULTS
Snapping Turtles
Using mean demographic rates for populations in the absence of harvest, snapping turtle populations grew at 2.3% (λ = 1.023) annually. Under maximum demographic rates, populations grew at about 19.9% each year (λ = 1.199) but declined by 10.9% per year at minimum demographic rates (λ = 0.891; Table 4, Fig. 3).
Citation: Chelonian Conservation and Biology 13, 2; 10.2744/CCB-1109.1
When field-estimated annual harvest rates (Table 3) were applied to population growth models, mean annual harvest (21%) resulted in population decline (λ < 1) for all harvest scenarios except when either only adults or only juveniles were harvested from populations under maximum demographic rates (Table 4). When the maximum field-estimated annual harvest rate was applied (43%), harvest was not sustained for any scenario except when only juveniles were harvested from populations under maximum demographic rates (λ = 1.045). When the minimum field-estimated annual harvest rate was applied (7%), harvest was sustainable for all maximum demographic rate scenarios as well as for when only juveniles were harvested under mean demographic rates.
Under mean demographic rates, 2.3% of both adults and juveniles could be harvested annually without causing population decline. Under maximum demographic rates, 18.6% of adults and juveniles could be harvested annually while maintaining stable or growing populations. Under minimum demographic rates, no annual harvest rate could be sustained by populations (Table 4).
Softshell Turtles
The baseline population growth rate for softshells was unsustainable and decreased slightly under both mean and minimum demographic rates (λ = 0.952 and λ = 0.893, respectively; Fig. 4). Under maximum demographic rates, baseline softshell annual population growth was 16.3% (λ = 1.163; Table 5, Fig. 4). Thus, because all scenarios at minimum and mean demographic values would continue to show declines, we focused only on harvest under maximum demographic rates.
Citation: Chelonian Conservation and Biology 13, 2; 10.2744/CCB-1109.1
When field-estimated annual harvest rates (mean = 24%, minimum = 6%, maximum = 79%; Table 3) were applied to softshell population growth under maximum demographic rates, population growth was unsustainable when harvesting both adults and juveniles at these rates but was sustainable when either adults or juveniles were the only group harvested. When the minimum field-estimated harvest rate was applied annually, population growth maintained sustainable levels of λ > 1 for all harvest scenarios. Under maximum demographic rates, an annual harvest rate of 16.3% could be sustained when both adults and juveniles were harvested (Table 5).
Sensitivity and Elasticity Analysis
Sensitivity and elasticity analysis indicated that survival of the adult stage has the greatest influence on population growth. For snapping turtles a 1% increase in adult survival annually would result in a 0.63% increase in annual population growth (Fig. 5) while, for softshells, a 1% increase in adult survival would result in an annual population growth increase of 0.51% (Fig. 6).
Citation: Chelonian Conservation and Biology 13, 2; 10.2744/CCB-1109.1
Citation: Chelonian Conservation and Biology 13, 2; 10.2744/CCB-1109.1
DISCUSSION
Our results indicate that under mean demographic rates even modest field-estimated harvest rates (Zimmer 2013) would not be sustainable at a local scale. These results mirror those of other studies that have suggested that low harvest rates of snapping turtles can be detrimental to their long-term viability. For example, in Michigan continued harvest of 10% of adults annually would reduce the number of adults by 50% within 20 yrs (Congdon et al. 1994). The effects of annual harvest were even more evident for softshells, for which populations were not sustainable at any combination of harvest parameters under mean and minimum demographic conditions. However, these results need to be placed in an appropriate context, including scale of harvests and our use of mean, minimum, and maximum demographic values. For example, both softshells and snapping turtles had some harvest scenarios which were sustainable. Although demographic rates would not be maintained consistently for long periods (i.e., 25 yrs represented in our models), these results illustrate that harvest may be sustained during some years, at a local scale, when population growth is maximized. However, it should be noted that maximum rates are not expected to be the normal demographic pattern for river turtles, given that there are numerous natural life-history parameters (e.g., low hatchling survivorship, late reproduction; Congdon et al. 1993; Galbraith et al. 1997) and interannually variable environmental factors (e.g., flooding, predation; e.g., Wilbur 1975; Plummer 1976; Congdon et al. 1987) that limit growth rates.
As our sensitivity and elasticity results indicated, survival of the adult stage has the greatest influence on population growth, a finding supported by other studies (e.g., Crouse et al. 1987; Congdon et al. 1994). This result is also evident when examining the effects of adult-only harvest versus juvenile-only harvest, where growth rates were consistently lower when only adults were harvested. Minimizing harvest of this stage would aid in reducing the potential for additional harvest pressure on these populations, and it might be possible to manage harvest of adults through use of size limits. The implementation of slot limits has been suggested for sustainable management of other harvested turtle populations (Gamble and Simons 2004; Brown et al. 2011). Additionally, softshell turtles exhibit sexual dimorphism, with adult females reaching a far greater body size than males. Female smooth softshells may reach up to 35.6 cm in carapace length, and female spiny softshells up to 54 cm, twice the length or more of an adult male (Johnson 2000; Ernst and Lovich 2009). Because turtles are often sold by weight (Brown et al. 2011), female softshells are likely targeted because of their greater weight. Placing size limits on softshell turtles could reduce the harvest pressure that may currently be placed on reproductive females, thus reducing impacts on turtle populations. Such restrictions have been suggested in order to reduce harvest pressure on adult female painted turtles in Minnesota (Gamble and Simons 2004). Further, in Texas the enforcement of bag and size limits for adult female turtles has been suggested for the same reasons (Brown et al. 2011). Although reducing the harvest pressure placed on the adult stage could increase the pressure placed on juveniles, the juvenile stage is much-less important to the overall demographics of the population.
In Missouri, commercial harvesters may collect turtles year round. Limiting harvest to specific times of year could also reduce harvest pressure placed on adults. For example, softshells often use sandbars and banks to lay nests (Johnson 2000). Limiting turtle harvest to areas away from sandbars and banks, which are used during breeding and nesting season (May through August; Barko and Briggler 2006), may reduce the harvest mortality of egg-laying females. This approach would reduce localized effects of harvest on female turtles, which may be more vulnerable to harvest when searching for nest sites if harvesters trap within areas containing sandbars. Brown et al. (2011) have suggested preventing harvest during breeding and nesting season for turtle populations in Texas. Additional modeling of various harvest scenarios would be a fruitful line of investigation to determine the ultimate impact of these management strategies.
Our results illustrate the inability of both snapping turtles and softshells to sustain minimum annual harvest pressure at a local scale when populations exhibit mean demographic rates. Assuming such consistency over a long period (i.e., 25 yrs) is not plausible, but there is insufficient information to adequately parameterize a stochastic model. Deterministic models result in overall higher population growth rates compared with stochastic models; therefore, our population growth models may be optimistic considering the limitations of the models we used. Additionally, the variability in demographic rates of turtles as evidenced in our literature review indicates that variability exists in the demographic rates of populations over time. For example, reproductive rates for turtles can fluctuate greatly from year to year (Wilbur 1975; Plummer 1976) and, while turtles may exhibit poor years where mortality rates are high and fecundity is low, these poor demographic years could be buffered by subsequent years when conditions are more favorable.
We conducted these simulations at a local scale using literature-derived values, but it is unlikely that commercial harvesters in Missouri are removing turtles at the annual harvest rates we have reported throughout all harvestable areas. Although we are not suggesting river-wide declines in river turtles due to commercial harvest at the current commercial harvest rates, the potential for local-scale effects does exist, particularly when commercial harvest regulations do not restrict the number of turtles that may be removed. Thus, high amounts of harvest activity may occur at a local scale, and local populations could be reduced if the same areas were trapped year after year. It follows that harvesters would then be likely to move to new trapping locations (i.e., according to the law of diminishing returns) in order to increase their catch. Still, because harvesters are not required to report specific locations where turtles were collected (aside from river of capture), it can be difficult to assess the scale of harvest activity. Local-scale effects of harvest on turtle populations have been observed for other turtle populations: harvesters have reported declining numbers of turtles in harvested areas for snapping turtles on the upper Mississippi River (Paisley et al. 2009) and for alligator snapping turtles (Macrochelys temminckii) in Louisiana (Boundy and Kennedy 2006). Further, harvesters have indicated that for the spiny softshell, reduced population size resulting from harvest activity may last for periods up to 3 or 4 yrs (Breckenridge 1955).
In conclusion, our models suggest that river turtles have only a modest capacity to withstand harvest, at least during some years, but the overall impact depends on the scale of harvest and the repeatability of harvest in the same locales. Further, the high variability of turtle demographics indicates that low annual harvest rates may be sustainable by populations in years when populations exhibit high demographic rates. However, in average or poor years, which we expect to be the normal scenario, even low annual harvest rates cannot be sustained. To reduce the risk of localized effects of harvest activity on turtle populations, harvest regulations could be modified to restrict harvest to specific regions of the river. Additionally, placement of size and bag limits, with an emphasis on juvenile take, can aid in decreasing the harvest pressure placed on the important adult segment of the population. Finally, upon examination of these observations, closing commercial turtle harvest could be a justifiable management action. For regions where harvest is relatively unlimited this may be an abrupt option, albeit supported by the results of this research, past studies, and by the evidence concerning turtle life-history traits and the known negative effects of sustained increases in mortality (Congdon et al. 1994; Galbraith et al. 1997). Additional modeling of these management actions and the resulting effects on population sustainability would aid in determining the effectiveness of the approaches described here.
Population growth matrices, including initial population vectors and stable stage distribution, for snapping turtle (C. serpentina) populations exhibiting mean (top), minimum (middle), and maximum (bottom) demographic rates.
Population growth matrices, including initial population vectors and stable stage distribution, for softshell turtle (A. mutica, A. spinifera) populations exhibiting mean (top), minimum (middle), and maximum (bottom) demographic rates.
Baseline population growth for snapping turtles (C. serpentina) at stable stage distribution under mean, minimum, and maximum demographic rates.
Baseline population growth for softshell turtles (A. mutica, A. spinifera) at stable stage distribution under mean, minimum, and maximum demographic rates.
Snapping turtle (C. serpentina) elasticity values under mean survival conditions at stable stage distribution. Bars represent the elasticity of transitioning to the next stage class (stippled bars), the elasticity of remaining in a class (striped bars), and the elasticity of fecundity (solid bar).
Softshell turtle (A. mutica, A. spinifera) elasticity values under mean survival conditions at stable stage distribution. Bars represent the elasticity of transitioning to the next stage class (stippled bars), the elasticity of remaining in a class (striped bars), and the elasticity of fecundity (solid bar).
Contributor Notes
Handling Editor: Peter V. Lindeman
