1. Tracking insect migration at continental scales is intractable using exogenous markers because of tiny body size and high improbability of recapture. Naturally occurring endogenous isotopic markers, such as tissue δ2H and δ18O, are a means of assigning origins to both vertebrate and invertebrate populations, but the success depends upon derivation of a robust algorithm linking measured tissue isotope values with large-scale geospatial isotopic patterns (isoscapes) in the terrestrial hydrosphere.
2. We derived a North American dragonfly wing δ2H and δ18O isoscape from known-origin dragonflies of three species (Aeshna interrupta, Aeshna umbrosa and Pachydiplax longipennis) obtained across North America. A strong relationship (r2 = 0·75) was found between wing δ2H and hydrologic geospatial δ2H patterns, and between wing δ2H and δ18O (r2 = 0·92). The strong coupling between emergent insect tissue and hydrologic spatial patterning suggested that this dragonfly isoscape may be applicable to other aquatic emergent migratory insects in North America and elsewhere.
3. As a proof of concept, we used the wing isoscape algorithm to map the probability of natal origin of Common Green Darners (Anax junius) migrating through southern Texas. Results showed that these Texan dragonflies were a mix of local and far-distant migrant (e.g. northern United States) individuals. We suggest that this isoscape algorithm opens new opportunities to quantify the migration and natal origins of dragonflies and other aquatic emergent insects where conventional methods have failed.
A compelling aspect of the evolution and life history of many organisms is their annual or seasonal movements, often over vast geographic scales (Hobson & Norris 2008). Such directed movements or migrations define the interactions migrants have with their environment and the range of biotic and abiotic forces they encounter. Migration also results in profound challenges for the conservation of animals as they move across jurisdictional boundaries and require suitable habitats to sustain them at distinct breeding and wintering sites and during stopover (Faaborg et al. 2009). By establishing migratory connectivity among populations throughout the annual cycle, conservation efforts can presumably be made to be more effective (Norris et al. 2006; Kelly et al. 2008). It is also clear that events occurring during one period and location in the annual cycle can influence organisms subsequently, and this has led to renewed interest in defining migratory connectivity, especially for migrant birds (Webster & Marra 2005).
Although less recognized or poorly studied, several species of large insects are migratory, for example, moths and butterflies (e.g. Red Admiral, Vanessa cardui), locusts (e.g. desert locust, Schistocerca gregaria) and dragonflies (e.g. Common Green Darner, Anax junius; Russell et al. 1998; Holland, Wikelski, & Wilcove 2006; Stefanescu, Alarcón, & Àvila 2007). In terms of biomass, movements of some insect species compete favourably with other well-studied migrant vertebrates (Holland, Wikelski, & Wilcove 2006). Some dragonflies (e.g. Anax junius) apparently show migration patterns similar to those of songbirds (Wikelski et al. 2006), but the main difference between insect and bird migration is the involvement of more than one generation of insects throughout the annual cycle. Individual insects that carry out a complete return migration have not yet been found. Perhaps, the most prominent example is the annual movement of Monarch Butterflies (Danaus plexippus) from eastern North America to their overwintering sites in central Mexico (Malcolm & Zalucki 1993). Several dragonfly species also perform spectacular migratory movements throughout the world (Russell et al. 1998; Corbet 1999; Freeland et al. 2003; Artiss 2004; Anderson 2009).
The major challenge limiting our understanding of migratory insect ecology is the ability to infer natal origins of individual animals, or to follow them throughout their travels. In many cases, extrinsic markers such as active transmitters (satellite, VHF, radiotransmitters) or passive recorders of light cycles (geolocators) have led to major advances (Hobson & Norris 2008; Bächler et al. 2010). However, for animals and insects that are too small to carry such devices, mark–recapture or the use of naturally occurring intrinsic markers such as molecular or stable isotope measurements have proven to be a useful option (but see Wikelski et al. 2006). Such intrinsic marker techniques do not rely on instrumenting animals or on successful recapture. Rather, they provide a forensic tool to infer prior location based on our understanding of spatial patterns of markers in nature and how they are transferred and recorded in animal tissues.
A major breakthrough in the use of naturally occurring stable isotopes to track migrant animals at continental scales was the discovery of the direct linkage between well-known spatial hydrologic isotopic patterns (isoscapes) in deuterium (2H) and oxygen (18O) to animal tissues (Hobson & Wassenaar 2008; West et al. 2010). Ultimately, precipitation isotopic composition drives the H isotopes in plant-based food webs, which, in turn, is transferred to metabolically inert tissues of top consumers, and thus the well-known hydrologic isoscape patterns can be used to infer the origin of where tissues were grown. This breakthrough has led to numerous successful applications of defining origins of several taxa (reviewed in Hobson 2008), including birds and butterflies (Wassenaar & Hobson 1998; Hobson, Wassenaar, & Taylor 1999; Brattström et al. 2010).
The isotopic tracing of natal origins of Monarch Butterflies wintering in Mexico was based on a calibrated wing δ2H isoscape for the breeding range in eastern North America and was based on wing samples from known natal sites across the breeding range. While this approach provided a high-spatial-resolution wing isoscape, this intensive approach is rarely practical in the broad application of isotopic tracking of insect origin in general. Instead, a more practical approach to tracing insect migrations will be through the derivation of generalized predictive isoscapes based on the species of interest and long-term average hydrologic isotope patterns (Bowen, Wassenaar, & Hobson 2005), an approach used extensively in isotope studies of avian migrants (Hobson et al. 2009a,b). This approach fundamentally presumes that a robust relationship between the isotope values of the tissue of interest and the hydrologic isoscape can be established (i.e. the precipitation-tissue isotopic discrimination factor). By measuring a sufficiently large sample of known-origin individuals from across their range, the algorithm resulting from such analyses can then be used to create a spatially explicit tissue isoscape, which can be used to derive an understanding of origins of migratory individuals and populations using Bayesian or likelihood-based methods of assignment (Hobson et al. 2009a,b; Van Wilgenburg & Hobson 2011).
Our objectives were to establish the dual isotopic (δ2H, δ18O) relationship between dragonfly wing chitin and precipitation in North America and to use this algorithm to create a spatially explicit wing-chitin isoscape to be applied to the study of migrant dragonflies and potentially other aquatic emergent migratory insects. As proof of concept, we used the H isotopic composition of a population of Common Green Darners moving through southern Texas and a statistical model to depict the probable natal origins of this potential mix of migrating and local dragonflies.
Materials and methods
Dragonfly tissue isotopic calibration
To derive a robust algorithm relating long-term geospatial hydrologic isotope patterns to dragonfly wing-chitin isotope values, dragonfly wings were provided from the collection of DRP (collected from 1947 to 2008). These dragonfly samples included (assumed) non-migratory Aeshna interrupta (n = 36), Aeshna umbrosa (23) and Pachydiplax longipennis (127). These three species provided the appropriate spatial coverage across North America to explore the full range of predicted precipitation δ2H values (Fig. 1). Because many of the dragonfly samples had been historically treated with acetone to preserve their coloration, we also evaluated the effect of colour preservation on δ2H and δ18O values.
Stable isotope measurements
All dragonfly wings were cleaned of surface oils in a 2 : 1 chloroform/methanol solvent rinse and prepared for stable-hydrogen and oxygen isotope analysis at the Stable Isotope Hydrology and Ecology Laboratory of Environment Canada in Saskatoon, Canada. Stable-hydrogen isotope analyses of dragonfly wings were conducted using the comparative equilibration method described by Wassenaar & Hobson (2003) and through the use of calibrated keratinous protein H/O isotope reference materials (Qi, Coplen, & Wassenaar 2011). Stable-hydrogen and oxygen isotope measurements were performed on H2/CO derived from high-temperature (1400 °C) flash pyrolysis of 350/700 μg wing subsamples using continuous-flow isotope-ratio mass spectrometry. All results for non-exchangeable δ2H and for δ18O are expressed in the typical delta notation, in units of per mil (‰), and normalized on the Vienna Standard Mean Ocean Water – Standard Light Antarctic Precipitation (VSMOW-SLAP) standard scale. Two keratin laboratory reference materials (CBS, KHS) were used to normalize the results, using the previously assigned values of −197·0 and −54·1‰ for δ2H, and +2·3 and +21·3‰ for δ18O, respectively.
Evaluation of acetone treatment
Treatment of dragonfly wings for preservation purposes typically involves the use of acetone. We first tested for possible deleterious effects of acetone colour preservation treatment on wing δ2H using freshly collected wings from Aeshna interrupta and Aeshna eremita from two Canadian locations (Waskesiu, Saskatchewan, and Delta, MB, Canada). This test was important because many current and future studies may rely on archived or extant specimens from museum and natural history collections, and thus, establishing the effect of preservation is crucial. For each individual, wing material (forewing and hindwing) from one half was treated with reagent grade acetone for 24 h, while those from the other half were stored without acetone treatment. Following drying, all samples were then cleaned by soaking and rinsing in a solution of 2:1 (v/v) chloroform/methanol and processed in the manner described earlier.
Assigning natal origins to potential migrants
A geographic information system (GIS) and geostatistical assignment method were used to link dragonfly wing stable-hydrogen isotope ratios (δ2Hw) from 42 individuals of Common Green Darner captured during fall migration in southern Texas to established geospatial hydrologic isotope patterns (δ2Hp) of eastern North America using the methods described previously (Hobson et al. 2009a,b). A rastered GIS hydrologic isoscape (δ2Hp; http://www.waterisotopes.org) was used as a basis for predicting natal location by using the calibration curve determined for known-origin dragonflies across North America (see Results). ArcGIS® 9 Spatial Analyst tool was used to transform the Bowen, Wassenaar, & Hobson (2005) hydrologic raster to a δ2Hwing map of the same spatial scale by applying our derived wing-precipitation calibration, where each raster cell represented a local δ2Hwing value rather than a δ2Hp value. This δ2Hwing isoscape was then imported into the R Version 2·10·1 (R Development Core Team 2010) statistical program where the rgdal, sp and raster packages were used to import and export raster data into the R interface; plot 2-D spatial data in map form, and perform spatial statistics on the raster surfaces. We assessed the likelihood that each cell within the transformed isoscape represented a potential origin for each dragonfly sampled using:
where f(y*|μc, σc) is the probability that any given cell (pixel) on the map represented a potential origin for an individual origin y*, given an expected mean (μc) of δ2Hwing based on the predicted value within the calibrated isoscape and the expected standard deviation (σ) of δ2Hwing between individuals growing their wings at the same locality. A value of σ = 14·7‰ was estimated using the standard deviation of the residuals from the regression equation linking δ2Hwing with δ2Hp. To estimate ‘probability of origin’, we used the normalized likelihoods calculated using eqn 1 as follows:
Applying eqn 2, we obtained a set of spatially explicit probability densities for each individual dragonfly. To assign all individuals to the basemap, we reclassified the spatially explicit probability densities into likely versus unlikely origins, by specifying an odds ratio. Based on 3 : 1 odds that a given assigned dragonfly had really originated from a cell within that range, we identified the set of cells that defined the upper 75% of estimated ‘probabilities of origin’ (using eqn 2) and coded those as one, and all others as zero. Each dragonfly was assigned to multiple potential origins within the isoscape at the same time. The results of the assignment for each individual were summed and mapped on the δ2Hwing isoscape to obtain the likely natal origin of the population.
Historical acetone coloration treatment had little influence on the δ2H and δ18O values of preserved dragonfly wings (Fig. 2), showing no significant differences between treated and untreated samples (paired t-test, δ2H: P = 0·09; δ18O: P = 0·55). The strength of the regression was weaker for δ18O values (r2 = 0·66) compared with δ2H values (r2 = 0·99). Therefore, we did not correct our historical acetone-treated sample δ2H values for any effects of acetone treatment (to transform them to non-acetone values). Wing δ2H values were regressed against the predicted hydrogen isotope hydrologic isoscape. We found a strong correlation between the hydrologic isoscape and wing δ2H, explaining 75% of the variance in dragonfly wing δ2H (Fig. 3). The subset of (untreated) samples also examined for wing δ18O values showed a strong relationship with wing δ2H values (Fig. 4), primarily reflecting the expected meteoric co-relationship. The effect of species on the δ2Hwing–δ2Hp relationship was tested by an analysis of covariance (ancova). No effect of species was found for the slopes and the Y-intercepts of the regression lines (F < 0·7, P > 0·10). Similarly, the effect of collection year on the δ2Hwing–δ2Hp relationship was tested using a GLM and ancova approach. We found no year effects in the GLM (GLM, year, F = 0·002, P > 0·10) or when using periods of 20 years as a factor in an ancova (period, F < 2·4, P > 0·10).
Using the newly derived algorithm depicted in Fig. 3 and the hydrologic H isoscape basemap, we created a continent-wide model dragonfly wing δ2H isoscape (Fig. 5). Our assignment of a sample of 42 Anax junius captured during fall in southern Texas revealed some exceptionally long-distance migrants (north-eastern USA, Fig. 6a) as well as individuals from the southern Gulf states (Fig. 6b). At the population level, all dragonfly individuals were well distributed in origin across east-central USA (Fig. 6c).
We found a strong relationship between δ2H isoscapes based on precipitation and those measured in dragonfly wing δ2H for the North American continent. This relationship provided a means of creating a dragonfly wing isoscape that can be used to infer natal origins of migratory individuals, as revealed by our test case of presumed migratory Anax junius intercepted in southern Texas. These findings open up considerable new possibilities for tracking migration and natal origins of dragonflies and other aquatic emergent insects using the same isotopic approach demonstrated for terrestrial and aquatic birds and Monarch Butterflies (Hobson & Wassenaar 2008). Importantly, while the phenomenon has been suspected in several species, ours is the first demonstration of actual long-distance migration of individuals of any aquatic emergent insect.
A striking aspect of this investigation was the exceptionally tight relationship we determined between dragonfly wing δ2H and δ18O values. This finding was important because it is currently not well understood how the well-established H vs. O meteoric co-relationship is maintained in food webs although promising results have been shown recently by Chesson et al. (2011) for a number of agricultural animals in the USA. Because O in animal tissues is derived from diet, drinking water and air, we might expect the expected isotopic coupling of water isotopes to break down at higher trophic levels, and to be influenced by the relative role of hydration in overall nutrition (Pietsch et al. 2011). Dragonflies are typically high-trophic-level consumers in the larval stage, but their wing chitin is derived entirely from their aquatic environments at formation (Tillyard 1917). This close isotopic coupling with their ambient water in the larval stage undoubtedly contributes to the strong relationship we determined between wing δ2H and δ18O values. An advantage of this coupling is that additional information on origins might be available through the measurement of both isotopes in dragonfly wings related to deuterium excess (Clark & Fritz 1997; Bowen et al. 2009) that can reveal origins of emergent insects from more highly evaporative or dynamic systems (e.g. wetlands, ponds).
The only published information on migratory insects comparable to ours involves terrestrial groups such as butterflies and beetles (Table 1). Hobson, Wassenaar, & Taylor (1999) derived a strong relationship between Monarch wing chitin and mean growing season precipitation δ2H across their eastern breeding range. Similarly, Brattström et al. (2010) reported as strong a regression for three species of European butterflies. However, using a controlled experiment with Monarchs raised on known (milkweed) plant water, Hobson, Wassenaar, & Taylor (1999) derived the best theoretical relationship between environmental water and monarch wings as δ2Hwing = −53 + 0·5*δ2Hwater (r2 = 0·99). These regression approaches underline the unavoidable variance associated with using predicted hydrologic isoscapes based on the long-term growing season average precipitation δ2H, and factors associated with individual variance in H isotopic discrimination taking place at the plant–water interface and subsequent plant-larval stage. Such isotopic variance is clearly implicit in models predicting dragonfly wing δ2H. Other H isotope data available include relationships between beetle exoskeleton chitin and meteoric waters as a means of deriving a palaeoecological proxy for temperature (Miller, Fritz, & Morgan 1988; Gröcke et al. 2006). These palaeo-studies are more difficult to compare with our data because authors analysed the nitrated chitin component and, in the case of Gröcke et al. (2006), only examined precipitation averaged over a narrow period of the annual cycle, nor are these species migratory.
Table 1. Results of regression between mean amount-weighted precipitation δ2H values and wing or exoskeleton chitin δ2H for various insect taxa
The migration of aquatic emergent insects like those of the Odonata in North America is a complex phenomenon involving multi-generational linked movements and life-history strategies that can respond flexibly to weather, climatic and other environmental forces (Trottier 1971; Corbet 1999; Matthews 2004; May & Matthews 2008). This complexity presents a challenge to investigations of insect movements globally. The stable isotope method advocated here, and particularly the use of both δ2H and δ18O analyses of wing chitin, thereby presents a unique means of inferring origins of insects without the burden and shortcomings of initial marking and unlikely recapture. We encourage future refinements of this approach to examine potential use of isotopes to improve our knowledge and to help better quantify migratory connectivity for the host of small-bodied species for which conventional mark–recapture approach has long remained intractable. Stable isotopes thereby are an important new tool for informing conservation efforts for those overlooked migrant species for which little is known about their migratory ecology.
This study was supported by operating grants to KAH and LIW from Environment Canada. We thank G.J. Bowen and two anonymous reviewers for critical comments on an earlier draft of the manuscript.