The phenology of many dragonflies have been shifting forward in time due to a warming climate (e.g., Hassall et al. 2007; Dingemanse & Kalkman 2008), and several geographic range expansions also have been reported (e.g., Beckemeyer 2009; McMurray & Simon 2011). Natural history surveys are key to informing the extent and rates of these trends; however, ongoing gaps in geographic information impede global assessments and predictions (Reece & McIntyre 2009). Florida is a key gap region due to its wide vertical gradient in climate that includes both the southernmost range of many north-temperate dragonflies and damselflies (Dunkle 1992), and the northernmost range of the largest group of endemic odonates found in any state or province (Dunkle & Westfall 1982). Further, despite a historically rich tradition in reporting and documenting Florida Odonata (e.g., Davis 1921; Byers 1930; Westfall 1941; Wright 1944; Needham 1946; Westfall 1953; Paulson 1966, 2001; Daigle 1978; Dunkle 1992), regional surveys and county checklists are now uncommon. This is concerning given that since the second half of the 20th century Florida has experienced rampant urbanization, eutrophication, freshwater modification and loss (Nagy et al. 2012), taxa introductions (Frank & McCoy 1992), and is poised for considerable climate change effects (Paulson 2001).

Here we report the first longitudinal surveys of adult dragonflies and damselflies in the Tampa Bay region (Hillsborough County, Florida, USA) since 1965 (Paulson 1999). The aims of our surveys are to: (1) determine the community composition in this county with a long-term goal to track shifts in phenology and range expansion among dragonflies (following Paulson 2001), and (2) report intraspecific measures of body size, fluctuating asymmetry, and parasitism. This intraspecific information is useful to provide insight on environmental stress (Daufresne et al. 2009; Gardner et al. 2011; McCauley et al. 2014), and how these stressors interact with a warming world to impact the phenology and ranges of dragonflies and damselflies (Hassall et al. 2007; Richter et al. 2008).

Materials and Methods

We conducted 2 surveys of adult dragonflies (Odonata: Anisoptera) and damselflies (Odonata: Zygoptera) at the University of South Florida Riverfront Park, Tampa, Florida, USA, located along a natural floodplain of the Hillsborough River (Hillsborough County, Florida, USA; 28.0705°N, 82.3786°W). This area is mostly fresh water wetland with hardwood hammocks, pine flatwoods, cypress domes, and open mowed recreational space for picnics, and a disc golf course. Lettuce Lake Conservation Park (97 ha) is located on the neighboring side of the river. The nearest previous survey of odonates was completed in Manatee County, Florida, USA, which is about 80 km south (i.e., Manatee River; Paulson 1966).

In our first survey, 1 individual sampled only dragonflies weekly for 1 yr (6 Oct 2013 to 5 Oct 2014) around midday (about 11:00 A.M. to 2:00 P.M.) for 1 h with a canvas sweep net. There are 2 sampling gaps in this survey: the park was closed on 20 Oct 2013, and again from 14 Dec 2013 to 6 Jan 2014. Our second survey occurred weekly from 22 Sep 2017 to 1 Dec 2017 in the late afternoon (about 3:00 P.M. to 5:00 P.M.). Here, 4 to 5 individuals sampled both dragonflies and damselflies for 25 min using canvas sweep nets, butterfly nets, and scoop nets. There are 4 sampling gaps in this second survey: 29 Sep 2017, 3 Nov 2017, 10 Nov 2017, and 24 Nov 2017. Finally, all captured odonates were placed into coin envelopes or jars marked with date and time, then then transported back to the University of South Florida for –10 °C freezer preservation.

Species and sex of each preserved specimen were identified following Garrison et al. (2006), Beckemeyer (2009), and Paulson (2011). As a surrogate of body size (Serrano-Menanes et al. 2008), the distance from nodus to pterostigma of left forewings was measured with digital calipers (± 0.02 mm; Pittsburgh item #93293, Carmarillo, California, USA). We estimated sexual size dimorphism in mean body size (mm) of each sex using linear regression models Ztests (α = 0.05), and when possible, we used linear mixed models to control for sampling effects between our 2 surveys (by treating survey identity as a random effect). Finally we compared the variance in body sizes with F-tests (assuming the ratio of the variances of males and females as one). Linear regression models analyses were performed in base R (R Core Team 2013) and linear mixed models with the lmer() function of the lme4 R package (Bates et al. 2015).

Following Lajeunesse (2007), wing-cell asymmetry was measured as the difference in wing-cell counts between the right and left forewings, where wing-cells were counted from nodus to pterostigma using a 10× loupe. To calculate and compare asymmetry frequency among individuals, a mixed-effect logistic regression model (with logit function) was implemented with the glmer() function of the lme4 package. This model included survey identify as a random-factor to control for sampling differences among the 2 surveys. A Z-test from a logistic model including a parity offset was used to test whether the ratio of asymmetric and symmetric individuals deviate from an equal frequency of observation between asymmetric and symmetric odonates (i.e., parity = 0.5). Finally, the ratio of asymmetric individuals and 95% confidence intervals were first estimated with the predicted values from the logistic model using the glht() function from the multcomp R package (Hothorn et al. 2014), and then using the inv.logit() function (generalized inverse logit function) from gtools R package (Warnes et al. 2013) to back-transform the 95% confidence intervals into a non-logged ratio scale.

Results

BODY SIZE VARIABILITY

Because of the low number of captures for most species, we were only able to assess sexual size dimorphism in 3 species (Table 1). Combining data from both surveys, there was female-biased sexual size dimorphism in both E. simplicicollis (linear regression models by sex: t = –7.04; P < 0.001; N = 264; female mean wing-length: 15.18; SD = 0.73; N = 184; male mean wing-length: 14.53; SD = 0.55; N = 80), and I. posita (linear regression models by sex: t = –7.29; P < 0.001; N = 133; female mean wing-length: 8.39; SD = 1.067; N = 79; male mean wing-length: 7.12; SD = 0.85; N = 54). Comparing body size variability between sexes, there were no differences in wing-length variances for E. simplicicollis (variance F-test = 1.178; df = [178, 79]; P = 0.411; variance ratio = 1.178; 95% confidence intervals = [0.797, 1.694]), or I. posita (variance F-test = 1.58; df = [78, 53]; P = 0.08; variance ratio = 1.575; 95% confidence intervals = [0.946, 2.561]).

Table 1.

Summary of two odonate surveys in Tampa Bay (Florida, USA). In rank order of capture abundance (TOTAL SAMPLED), odonates are grouped by infraclass (INF) as Zygoptera (Z) or Anisoptera (A), and sex. Wing-size and wing-cell fluctuating asymmetry is reported here as the proportion of asymmetric individuals and their lower (L) and upper (U) 95% confidence intervals (CI) were estimated using the Clopper & Pearson (1934) method. Sample sizes (N) of measurements differ from capture numbers due to wing damage during transport and preservation.

In contrast, M. marcella dragonflies had male-biased sexual size dimorphism (linear regression models by sex: t = –2.24; P = 0.029; N = 47; female mean wing-length: 13.68; SD = 1.297; N = 32; male mean wing-length: 14.48; SD = 0.887; N = 15). However, we caution that this overall male-biased sexual size dimorphism is inconsistent among our 2 surveys (Table 1; linear regression models by Sex: F = 6.21; df = 1; P = 0.0156; Survey: F = 6.51; df = 1; P = 0.0133; Sex by Survey: F = 10.03; df = 1; P = 0.002), and when accounting for this variability using a mixed-model, M. marcella dragonflies were sexually monomorphic (linear mixed models by Sex with Survey as random factor: t = 1.74; P = 0.088). Finally, there were no differences among the wing-length variances between females and males (variance F-test = 2.14; df = [46, 14]; P = 0.122; variance ratio = 2.14; 95% confidence intervals = [0.807, 4.634]).

Discussion

Similar to Paulson's (1966) observations, many of our estimated flight times indicate clear seasonality in Odonata activity in the Tampa Bay area, with either low abundance or absence during the coldest periods during the dry season (i.e., Nov–Mar). Tampa Bay is located at the southernmost frost point of the state (about 27.9506°N). Below this frost line, Odonata may have flight seasons yr round (Paulson 1999); however in Tampa Bay, there is still a high risk of at least 1 frost (0 °C) per dry season. In fact, our first survey experienced 0 °C on 7 Jan. This frost may explain the bivoltine flight time of the Neoptropical M. marcella (Fig. 1), because frost would have killed both adults and the invasive plant that nymphs use as habitat (the common water hyacinth, Eichhornia crassipes (Mart.) Solms (Commelinales: Pontederiaceae) (Beckemeyer 2009). The common water hyacinth is present at our site, but abundance has not been tracked for seasonal changes. Beginning in Mar and the start of the rainy season, the first dragonfly species to emerge were M. marcella and E. simplicicollis (Fig. 1), then later in the summer (about Jul), C. eponina, P. longipennis, and A. williamsoni emerged. The only species captured throughout the entire wet season until Oct was E. simplicicollis. Finally, the skimmer T. onusta emerged briefly in late fall (as also described by Paulson 1966), along with the second emergence of M. marcella (Fig. 1). Our second survey was too brief to estimate flight times of damselflies.

Fig. 1.

Capture dates of 12 odonate species captured between Oct 2013 to Oct 2014 (top panel), and again from Sep 2017 to Dec 2017 (bottom panel) in the University of South Florida's Riverfront Park in Tampa, Florida, USA. Dashed lines indicate the start and end of surveys, and shaded areas indicates the typical wet season. Blocks indicate the wk when individuals were captured; spaces without blocks indicate periods when no individuals were captured.

Odonates may indicate aquatic and terrestrial environmental health because of their semi-aquatic life cycles (Bustos-Baez & Frid 2003; Bybee et al. 2016), and because their body size and fluctuating asymmetry may be used as surrogates of environmental stress (Jentzsch et al. 2003). This includes sexual size dimorphism because environmental stress or change can differentially impact nymph growth between females and males (Baker 1986). There was some variability in these traits among the odonates we surveyed. We found that female E. simplicicollis were larger, but not more variable than males in forewing-length, which confirms other reports of female-biased sexual size dimorphism based on body-length (McVey & Smittle 1984) and wing-loading (mg per cm²; Locklin & Vodopich 2010). Female-biased sexual size dimorphism also was found among I. posita damselflies; however, size data for this species is limited, but Serrano-Meneses et al. (2008) reported this species as slightly male-biased. Likewise we could not find reports of sexual size dimorphism for M. marcella dragonflies; however, we found them to be monomorphic. We did not find differences in symmetry in either sex for E. simplicicollis, I. posita, or M. marcella; however, when combining data from each sex, I. posita damselflies were asymmetric. However, it is difficult to assess whether this asymmetry has functional importance at this site without clear stressors to link this asymmetry.

Following Lajeunesse (2007), the primary environmental stressor we aimed to measure was larval water mite parasitism; however, we did not capture any parasitized odonates at our site. Many of these species do have reports of water mite parasitism; for example, see Mitchell (1961) for E. simplicicollis; Botman et al. (2002) and Mlynarek et al. (2013) for I. posita; and Rodrigues et al. (2013) for M. marcella water mite parasitism within its native range (Brazil). Water mites are found in almost every type of freshwater habitat, but uncommonly in marine or brackish water (Wolcott 1905), and there are several permanent and temporary fresh water sources near our site. Our sampling site is located on the banks of the Hillsborough River which empties into Tampa Bay, and its salinity can fluctuate with the tides. However, this fluctuation is unlikely to affect water mite activity in our river section, which is over 32.2 km (20 miles) upriver from the mouth, and separated by the Hillsborough River Reservoir dam. Our sampling site also has several ephemeral ponds and swamp land that can dry up during the dry season (about 1 km); however, we sampled only near the river's edge since 23 Apr 2015, when 1 parasitized I. posita damselfly was captured (MJ Lajeunesse, personal observation). This female had 23 water mite larvae (7 thorax and 16 abdomen), which were dark orange with dark brownish-black spots, and 1 larvae was lodged between the ovipositor and the last abdominal (10th) segment. Given the lack of parasitized damselflies in our second survey, this parasitized individual may have migrated from a neighboring site and future surveys will target these ephemeral ponds to improve water mite sampling.

There are several ways we can improve the quality and scope of our surveys. Raebel et al. (2010) recommend collecting dragonfly exuviae for diversity surveys because it decreases capture biases (Lajeunesse et al. 2004), and in a similar vein, emergence traps also may provide better quality flight season data. These approaches avoid challenges with differentiating between site-residents and immigrants, which we could not separate in our surveys. Finally, Florida has a strong baseline of dragonfly and damselfly surveys in the early- to mid-20th century (Paulson 1966), but more recent surveys are lacking. We hope our surveys encourage more regional assessments throughout Florida. These would significantly improve long-term forecasts of changes in phenology and geographic ranges of odonates, especially if surveys span the broad latitudinal gradient of the Florida peninsula.