Dragonfly and Damselfly (Insecta, Odonata) Distributions in Ontario, Canada: Investigating the Influence of Climate Change

We analysed temperature data and odonate distribution data collected in the province of Ontario, Canada, over approximately sixty years. Analysis of temperature data from 31 weather stations collected in the years 1945–2000 showed an overall signifi cant increase in the minimum, maximum and mean monthly temperatures; these trends were not adjusted for changes in urbanisation. Comparison of county level presence/absence data for odonates from the 1950 s and 2002 found a slight decrease in the northernmost distributions of some species, although no signifi cant patterns were evident. Lower sampling coverage in the larger, more northerly counties in Ontario, as well as the assessment of distributions based on county records may limit the sensitivity of our approach in detecting changes in odonate species distributions over time. Future work should focus on increasing the coverage, uniformity and geographic detail of available datasets, as well as evaluating range change through testing predictions based on the ecology and biogeography of odonate species.


Climate Change, Ecological Responses & Odonates
Compelling evidence for a global trend in environmental warming continues to accumulate (IPCC 2007, Karl andTrenberth 2003). Likewise, researchers have found ecological and evolutionary responses to the eff ects of climate change in a number of plant and animal groups (Huntley 1991, Hughes 2000, with a range of data specifically from arthropods (Parmesan et al. 1999, Warren et al. 2001. Further study of the ecological and evolutionary responses of a number of plant and animal groups is necessary to better understand and predict future climate change responses, and to potentially mitigate the detrimental eff ects of climate change. Organisms can respond to climate change in a number of ways: by going extinct, by adaptation in situ, by range change or expansion (Coope 1995), and/or through plastic changes in life-history patterns (Butterfl ied and Coulson 1997, Hassall et al. 2007, Paremesan 2007. Th e type of response observed will likely be a function of both the type of organism involved (based on its ecology and life-history), as well as the home-range conditions of that organism. Dragonfl ies and damselfl ies (Insecta: Odonata), off er great potential as an indicator group for anthropogenic disturbances, such as habitat alteration, and climate change (Corbet 1999 Chapter 12 and sources therein). Th e majority of dragonfl y species have an obligate aquatic larval stage, linking their ecological success to the distribution and quality of a range of aquatic habitat types throughout the landscape. As an aerial adult, odonates demonstrate a potential for dispersal and recolonization (Conrad et al. 1999), with some species migrating great distances in response to seasonal changes in weather and climate (Wikelski et al. 2006). As such, the possibility of range change and expansion is high in this group, if other ecological factors (such as the distribution of appropriate habitat for specialist species) allow for such responses.
Recent work has investigated the change in range size of dragonfl ies and damselfl ies to detect the eff ects of climate change (De Knijf et al. 2001, Ott 2001, Ott 2007. Work published by Hickling et al. (2005) detected a signifi cant expansion of the total range, as well as the northernmost range extent, of non-migratory odonates in Britain. Hassall et al. (2007) have also found signifi cant changes in the phenology of British odonates, with advances of 1.51±0.060 days per decade (or 3.08±1.16 days per degree rise in temperature) in the leading edge (fi rst quartile date) of the fl ight period between 1960 and 2004. As odonates are a ubiquitous group of organisms, found on many continents and in many biomes, investigation of the response of this group could prove a useful indicator for the eff ects of climate change in regions throughout the world.
In this current study, we have assessed data on the distribution of dragonfl ies and damselfl ies in the province of Ontario, Canada, using surveys at diff erent time periods (1950´s and 2002) to evaluate potential changes in species ranges. We compare these preliminary results to recorded changes in climatic conditions in Ontario that have arisen over this same time period. As with any survey work, one of the important interpretive constraints is knowing whether a recorded absence actually corresponds to a genuine absence. We discuss our results and their implications in the light of these constraints, and we suggest potential directions for future research.

Habitat and Climate in Ontario
Ontario is a large province, covering an area of 1,076,395 km² (approximately 15% of which is covered with water) divided among three geographic regions: the Precambrian Shield (a rocky region with many lakes and wetlands that comprises over half of the province) in the northwestern and central regions, the Hudson Bay lowlands in the extreme north and northeast and the Great Lake/St. Lawrence Valley region in the south. Th e northernmost point of the province lies at 56°51'N, extending south to 41°54'N at the end of Point Pelee in Lake Erie, a distance of over 1600 km (the westeast extent of Ontario is from 95°10'W on the border with the province of Manitoba to 74°19' where it adjoins the province of Quebec, along the St. Lawrence River, just over 1500 km). Th ere are a total of 47 administrative counties in the province (see Figures 1 and 2).
Th e climate of the province ranges from a moderate humid continental climate in the southernmost regions of southern Ontario (Peel et al. 2007) to a more severe humid continental climate in the central portions of the province. Th ese regions have relatively hot summers and cold winters, though summers are shorter in the north. Northern Ontario (especially above 50°N latitude) has a much longer and more severe winter.

Th e Odonate Fauna of Ontario
Ontario has a rich odonate fauna, attributable at least in part to the diversity of habitat types distributed on a north-south gradient within the province. Many species that are associated with more southerly habitats in the United States, such as the Carolinean forests, are found in southern Ontario, while more northerly-distributed species, such as those found within the boreal forests, are common to the northern regions of the province (Dunkle 2000, Catling et al. 2004). As such, species ranges within this region may be more signifi cantly aff ected by climate change than in other geographies.
Th e data used for these analyses (see below) indicate the presence of 134 species of Odonate, distributed over a total of 9 families: Aeshnidae, Cordulegastridae, Corduliidae, Gomphidae, Libellulidae and Macromiidae in the suborder Anisoptera (dragonfl ies); Calopterygidae, Lestidae and Coenagrionidae in the suborder Zygoptera (damselfl ies) (see Table 1). Th e species within these groups utilize a large range of habitat types, from large lakes, ponds and marshes to small streams and large rivers.
Many species are habitat generalists, while some, such as Nehalennia gracilis, specialize in specifi c habitats (in this case, sphagnum bogs, similar to N. speciosa in Europe (Lam 2004, Dijkstra 2006).
Due to their range of habitat preferences, some species are found in only a single region of the province; for example, many species, such as Sympetrum vicinum and Celithemis elisa, are found only in the southermost regions of the province, while species such as Aeshna juncea and Somatochlora whitehousei are found only in the north. Still other species, such as Leucorrhinia hudsonica, are found throughout the province. Some species, while found over a large area, have a patchy distribution, due to the limited presence of their preferred habitat type. Finally, some species, such as Anax junius, are migratory: in this case, individuals will fl y south in the fall, and the off spring they produce will return to Ontario in the spring. In the case of A. junius, only some populations are migratory; others, often found in the same habitats with individuals that migrate, remain in Ontario throughout the year, completing their entire life cycle in these lakes and ponds.

Unadjusted estimates of climate change in Ontario
We assembled data on maximum and minimum monthly temperatures collected at 31 weather stations throughout Ontario (Canadian Centre for Climate Modelling and Analysis, http://www.cccma.ec.gc.ca/data/data.shtml), starting in January 1945 and continuing through 2000, (Figure 1, Table 2). Th ese data approximately coincide with the time period of our odonate distribution data (see below). Fluctuations in the mean monthly temperatures of three winter months and three summer months over these decades are shown in Figure 3a and 3b. . Odonate presence/absence was recorded and analysed on a county by county basis. Northern range extensions were calculated by assessing the mean and northernmost extent of the counties in which each species was found. Table 1. List of Ontario odonate species found in datasets used in this analysis. Species are grouped by suborder and family. Th e change in northernmost latitude by county between the 1950´s and 2002 is listed as increased (+), decreased (-) or unchanged (U).  To test whether overall temperatures in Ontario have changed over this period, and estimate the size of these changes we fi tted a General Linear Model (GLM) to the climatological data with year as a covariate and site as a random factor. Th ese analyses indicated a signifi cant increase in minimum, mean and maximum temperature over the period 1945 to 2000 (see Table 3 for results), as well as a signifi cant site eff ect. Th e rates of temperature increase were relatively consistent throughout the year such that temperatures in Ontario have increased at the rate of about 0.02°C per year (1°C in 50 years ) in both summer and winter months (Figure 3a and 3b, Figure 4). Th ese results are in line with larger-scale assessments of climate change, which have found increases in mean annual temperatures throughout southern Canada ranging from 0.5°C to 1.5°C over the 20th Century (Bonsal et al. 2000, Zhang et al. 2001). Our analysis does not account for important eff ects such as urbanisation (Prokoph and Patterson 2004), which could artifi cially increase estimates of rates of change of temperature at many of our locations where weather stations are situated. Th e signifi cant diff erences in temperatures among sites are not unexpected, since sites in the northern regions of Ontario are consistently colder throughout the year.

Changes in Odonate Distributions in Ontario
To assess potential responses to the changing climate of Ontario, we assembled county distribution data for dragonfl ies and damselfl ies from two sources. First, historical distributions were gleaned from the volumes of The Odonata of Canada and Alaska by Walker (vol. 1, 1953Walker (vol. 1, , vol. 2, 1958 and Walker and Corbet (vol. 3, 1978). Data for suborder Zygoptera ( For current distributions we have synthesized data from volume 4 of Ontario Odonata by Catling, Jones and Pratt (2004). Th ese recent data represent known distributions as of 2002 and comprise 6,700 recorded sightings by 43 diff erent contributors. Records were compiled into databases at the regional level and then passed to a provincial complier to create a single database. Unusual records were discussed with contributors to ensure a level of certainty. Voucher specimens were required for the most   signifi cant records. Included in these diff erent surveys over several decades are data on 134 species, with 39 damselfl ies (suborder Zygoptera) and 95 dragonfl ies (suborder Anisoptera) (see Table 2 for species list).
Since the records of occurrence were only at the county level, some analysis was required to match historic and current distributions of Ontario odonates to geographical latitudes. For each county, the geographic boundaries were used to determine the northern extent of the county in terms of latitude, which allowed for an estimate of the maximum northerly latitude of a species based on the northern-most county observation. Th e median latitude of each county was also determined given the geographical boundaries of the county. Th e median latitude of the county enabled an estimation of the average latitude of the range of the species within Ontario given its presence in a number of counties. Some counties that had been sampled by Walker were not sampled in the 2002 data; when historic and current data were matched, data from a total of 41 counties could be assembled, though inevitable variation exists as to the extent of sampling within diff erent counties.
Of the 134 species of odonate included in the study, 29 showed an increase in northern range extension between the two datasets (based on the northernmost latitude of the northernmost county in which the species was detected), while 40 species actually showed a decrease. Sixty-fi ve species showed no change in their To determine whether diff erences could be found among taxonomic groups, we compared family-level distributions between datasets. We found similar patterns within family to those for the overall order ( Figure 6) -thus, even at the family level, no signifi cant overall increase (or decrease) was detected for any group.
It has been suggested that odonate species which inhabit lentic habitats (non-moving water, such as lakes and ponds) have larger and more northern distributions than those that inhabit lotic (river and stream) habitats (Hof et al. 2006). Hof and colleagues found this to be the case for odonates throughout Europe and North America, suggesting that the lower stability of lakes and ponds through time (in comparison to streams and rivers) require lentic species to have higher dispersal rates. As such, lentic species might be quicker to respond to increasing temperatures, and thus might expand their ranges northward more rapidly than lotic species. We compared lentic to lotic species (92 species versus 42 species, respectively), but found no diff erence in the pattern of latitudinal extent between these two groups (Figure 7).

Preliminary conclusions
Our fi rst-step analysis of evaluating the responses of Odonates to changes in climate has concentrated on temperature, although we recognise that rainfall may also have the potential to infl uence the breeding ecology (and hence distribution) of Odonates (Cannings and Cannings 1998). We have found that, while increases in the recorded minimum, maximum and mean monthly temperatures in Ontario have been observed over the time interval between 1945 and 2000, no consistent pattern of change in the northern extensions of our studied odonate species is detectable. Th is is surprising, as changes in range size and northward extent, as well as changes in life history, have been observed in odonates in a warming European climate (Hickling et al. 2005, Hassall et al. 2007, Ott 2007. A number of diffi culties often arise in the collection of time-series data such as those used for these analyses. First, uneven sampling eff ort may lead to diff erences in levels of detection at diff erent times. For example, if a lower amount or intensity of sample eff ort (fewer sample events, or fewer collectors) takes place during a specifi c phase of the sampling, species may be underrepresented, and the full distribution of some species-especially rare species-may not be determined (Southwood & Henderson 2000). As is often the case with survey data, later samples in our data (from the 2002 source) were done more consistently-by a large number of people-than our earlier data, which was collected by Walker and his colleagues. Th us, the detection of an expanded range for a species-particularly a species with low densities or a patchy distribution-may be an artefact of sampling.
Secondly, uneven sampling within a sample period (for example, our 2002 dataset) may lead to biased assessments of species distributions. Sampling may favour urban areas, with large numbers of collectors and with greater access to habitats. Large areas with low human population densities will likely be sampled less eff ectively and could result in uneven recordings of species distributions. Th is could be problematic in our dataset, in that the northernmost counties of Ontario are the largest, least populated and least accessible; these are also the most crucial to sample in the detection of northward range changes.
Th e use of county-level presence data is also problematic, in that there is a great range of sizes in county areas in Ontario (see Figure 2); thus, recording presence at the county level represents an uneven sample eff ort. Many modern surveys record latitude and longitude of the actual sample location -this is made all the more easy due to the availability of hand-held global positioning system (GPS) receivers. However, older data, such as those used here, often are not recorded with such accuracy. A potential problem with county record data when analyzing changes is distribution patterns comes if county sizes are clustered -that larger counties are grouped in an area together. In our data, the fact that larger counties in Ontario tend to be in the northern part of the province could lead to an infl ated increase in northern range extensions in species, as any record within a large county could represent a considerable northward latitude increase for that species.

Future Work
We observe that, while we did not fi nd a signifi cant trend in the change of species distributions in our dataset, this may be due not to absence of a change, but an inability to detect such change with the current data. Ontario is a large province, with many remote areas that are diffi cult to sample. Still, increased interest has been seen in recent years in dragonfl ies and damselfl ies, refl ected by the development of regional groups of odonate enthusiasts throughout North America, and the increased availability of fi eld guides. We welcome this increase in public interest and awareness, and hope that it continues to contribute to the growing availability of data species distributions. We encourage our readers, in Ontario and elsewhere, to become involved in the collection and synthesis of dragonfl y distribution data, which will improve our understanding of these organisms. Th ose specifi cally interested in Ontario odonata can become involved by contacting the Natural Heritage Information Centre with the Ontario Ministry of Natural Resources, the Toronto Entomological Society, or other naturalist organisations in the region.
When considering the distribution patterns of odonate species, intriguing trends can be observed. While some species in this survey were found throughout the province, others are limited to a distinct region. While some species distributions can be explained by the extent of a habitat type (Precambrian Shield lakes, for example) other species show distinct distribution boundaries that do not match to discrete landscape features. For these species -such as Aeshna canadensis, Erythemis simplicicollis or Enallagma civile -range boundaries appear to be limited primarily by a maximum latitudinal extent. It may be that climatic patterns are the limiting factor in the distribution of these species; We propose, in future work focusing on species such as these and employing a larger and more robust dataset for North American odonates, to further and more rigorously explore the relationship between odonate distributions and climate change