SEARCH

SEARCH BY CITATION

Keywords:

  •  bluetongue virus;
  • climate;
  • Culicoides imicola;
  • C. obsoletus;
  • C. pulicaris;
  • environmental envelope

Summary

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Supporting Information
  • 1
    The spread of vector-borne diseases into new areas, commonly attributed to environmental change or increased trade and travel, could be exacerbated if novel vector species in newly invaded areas spread infection beyond the range of traditional vectors.
  • 2
    By analysing the differential degree of overlap between the environmental envelopes for bluetongue, a devastating livestock disease, and its traditional (Afro-Asian) and potential new (Palearctic) midge vectors, we have implicated the latter in the recent dramatic northward spread of this disease into Europe.
  • 3
    The traditional vector of bluetongue virus, the Afro-Asian midge Culicoides imicola, was found to occur in warm (annual mean 12–20 °C), thermally stable locations that were dry in summer (< 400 mm precipitation). The Palearctic C. obsoletus and C. pulicaris complexes were both found to occur in cooler (down to 7 °C annual mean), thermally more variable and wetter (up to 700 mm summer precipitation) locations.
  • 4
    Of 501 recorded outbreaks from the 1998–2004 bluetongue epidemic in southern Europe, 40% fall outside the climate envelope of C. imicola, but within the species’ envelopes of the C. obsoletus and C. pulicaris complexes.
  • 5
    The distribution in multivariate environmental space of bluetongue virus is closer to that of the Palaearctic vectors than it is to that of C. imicola. This suggests that Palearctic vectors now play a substantial role in transmission and have facilitated the spread of bluetongue into cooler, wetter regions of Europe.
  • 6
    Synthesis and applications. The risk to Northern Europe now depends on how much of the distributions of the widespread, abundant Palearctic midge vectors (the C. obsoletus and C. pulicaris complexes) bluetongue can occupy, perhaps determined by thermal constraints on viral replication. This was highlighted by the sudden appearance in summer 2006 of bluetongue virus at latitudes of more than 50° North – approximately 6° further North than previous outbreaks in southern Europe. Future surveillance for bluetongue and for related Culicoides-borne pathogens should include studies to record and explain the distributional patterns of all potential Palearctic vector species.

Introduction

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Supporting Information

Modelling the environmental envelopes of vectors and vector-borne pathogens by statistical matching of their geographical distributions with environmental data has been used widely to (1) reveal the potential environmental factors driving their distributions and (2) predict their distributions in unsurveyed places (Rogers & Randolph 2003). The environmental envelope quantified by distribution modelling is analogous to a species’ realized niche (Austin et al. 1990; Guisan & Zimmerman 2000), i.e. the hypervolume, defined by environmental dimensions, within which that species can survive and reproduce in the face of competitors, predators and pathogens (Hutchinson 1957). More rarely these models have been used to investigate the overlap of the geographical and environmental distributions of pathogens and hosts, offering the potential to tease apart the relative roles of different host species and vectors in disease transmission cycles (Peterson & Shaw 2003; Peterson et al. 2004).

When diseases are introduced or spread into new areas by increased trade and travel, or following environmental changes, it is essential to re-assess the pieces of the epidemiological puzzle – the pathogen strains, hosts and vectors involved in transmission (Rogers & Randolph 2003). Bluetongue (BT) is a devastating midge-borne disease of ruminants that spread recently across Europe from its traditional Afro-tropical ‘home’. Historically, bluetongue virus (BTV) (the pathogen that leads to bluetongue) made only brief, rare incursions into fringe areas of Europe, with the Afro-Asian biting midge vector, Culicoides imicola s.s. Kieffer (referred to hereafter as C. imicola), being largely responsible for transmission (Mellor & Boorman 1995). Since 1998, however, six strains of BTV have entered Europe virtually simultaneously from at least two directions, and have spread across 12 countries and up to 800 km further north than reported previously (Purse et al. 2005). Early on in this process, bluetongue occurred beyond the known northern range limit of C. imicola (even though this species itself had shifted northward over the same period), thus suggesting a role for Palearctic Culicoides species as vectors. BTV spread in Europe has resulted in massive disruption of trade in animals and animal products and caused the deaths of over 1·5 million sheep since 1998 (Mellor & Wittmann 2002; Calistri et al. 2003).

A link between BTV emergence and regional climate change has already been suggested. The most compelling piece of evidence is that BTV and C. imicola expanded their ranges into those locations that have warmed the most during the 1990s, but have not done so in locations that remained cool, or decreased in temperature during the same period (Purse et al. 2005). Regional warming may have increased the importance of Palearctic vector complexes in transmission by increasing their population sizes and survival rates and by increasing their individual susceptibility through developmental temperature effects (Wittman & Baylis 2000). That these Palearctic vectors, primarily the C. obsoletus and C. pulicaris species’ complexes (hereafter, Palearctic Culicoides), have played some role in BTV transmission during the current epidemic has been established by fine-scale temporal and spatial overlap of their distributions with outbreaks (De Liberato et al. 2003; Torina et al. 2004; De Liberato et al. 2005) and by virus isolation from wild-caught Palearctic Culicoides in several sites (Savini et al. 2003, 2005; De Liberato et al. 2005). Despite this, the involvement of these species’ complexes in transmission has been generally assumed to have been relatively minor. Much vector surveillance effort has been directed towards defining C. imicola free zones (e.g. Conte et al. 2003), between which vaccinated animals could legally be moved for trade or seasonal transhumance – regardless of the abundance of the Palearctic Culicoides in those zones. It is only since 2004 that the cessation of all adult Culicoides activity, rather than simply that of C. imicola, has to be proved in a destination area before vaccinated (with live-attenuated vaccines) animals can be moved there (EC 2004).

Not only are Palearctic Culicoides common and widespread across Central and Northern Europe, but some of their field populations have been shown to have levels of vector competence equivalent to that of C. imicola (Carpenter et al. 2006). Consequently, for accurate risk assessment it is essential to determine (1) the relative extent to which Palearctic vector complexes and the major Old World vector, C. imicola, are currently involved in BTV transmission, in both geographical and environmental space and (2) to what extent and along which environmental axes had the environmental envelope of BTV extended by the early 2000s due to its transmission by Palearctic vector species. A Mediterranean-wide analysis of the current environmental drivers of C. imicola distribution (for country-specific analyses see Baylis et al. 2001; Tatem et al. 2003) is also a prerequisite for examining whether this species’ envelope has changed over time or whether the environment itself has changed, bringing about a redistribution of the species in space, over time.

Here, our first objective is to determine the climatic dimensions that best define the current environmental envelope of C. imicola across Africa and southern Europe. This vector was the target of historical surveillance efforts and is hence well-recorded, at least along the northern fringes of its range (i.e. in southern Europe), enabling us to establish the maximum extent of its geographical distribution in this region. This study is limited to consideration of the climatic dimensions of a species’ envelope but, given available data, habitat or host dimensions could be investigated simultaneously within the same model framework. Our second objective is to compare the extent of overlap of the key environmental determinants of C. imicola's distribution with areas of BTV transmission and the overlap of Palearctic Culicoides vectors with BTV transmission to shed light on the magnitude and nature of the role of these different vectors in transmission. To date, Palearctic Culicoides have been recorded mainly only across the southern portions of their Palearctic distributions, in countries affected by bluetongue since 1998. This study therefore captured only part of their full geographical distributions (although at latitudes relevant to current BTV epidemics) in Europe which we know, from sporadic historical records, to be much more widespread in areas north of the Southern Europe outbreak area.

Materials and methods

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Supporting Information

Detailed methods are given in the Supplementary material (Appendix S1). Briefly, point or commune-level distribution records for BTV and for its vectors, C. imicola and the C. obsoletus and C. pulicaris complexes (Palearctic Culicoides), were collated across southern Europe and North Africa from 1998 to 2004. At a spatial resolution of one-sixth of a degree latitude and longitude (the resolution of the environmental data), there were 180 presence and 215 absence records for C. imicola. Unusually for distribution data, these absence records were based on extensive field sampling in and near the outbreak zones. Because very few sampled absences were available for the widespread Palearctic Culicoides, presence data only (428 and 410 presence records for the C. obsoletus and C. pulicaris complexes, respectively) were used as a baseline estimate of their distributions across environmental space. There were 501 records of BTV presence, covering a substantial proportion of the total area affected during the recent bluetongue epidemic in Europe (Fig. 1). One-sixth of a degree of monthly climatic data for the period 1996–99 were obtained from the Climate Research Unit (CRU) at the University of East Anglia, UK (data set CRU TS 1·2, http://www.cru.uea.ac.uk/~timm/grid/CRU_TS_1-2.html; Mitchell et al. 2004), and temporal Fourier processed (Rogers et al. 1996) to produce a set of 17 climatic predictor variables: the annual mean, maximum and minimum of temperature and vapour pressure as well as their amplitudes and phases of the annual and bi-annual cycles and the mean annual amount, summer amount and winter amount of precipitation. The predictor variables that were most important in defining the environmental envelope of C. imicola were identified using forward stepwise discriminant analyses of clustered bootstrap samples of the presence–absence data sets (Rogers 2006). Classification accuracy was assessed using the Kappa coefficient, an index of agreement between two categorical classifications (here the observed and the predicted classes).

image

Figure 1. Recorded distribution of (a) BTV, (b) Culicoides imicola and (c) C. pulicaris complex and C. obsoletus complex in the Mediterranean Basin between 1998 and 2004.

Download figure to PowerPoint

To quantify the overlap of BTV and different vector distributions in multivariate space, defined by the predictor variables identified above, we calculated the multivariate environmental distances (Mahalanobis Distance: MD) between known outbreak areas of BTV and known areas of presence of C. imicola and of the two Palaearctic vectors, C. obsoletus and C. pulicaris. As the distribution of any pathogen, transmitted by several vectors, is composed of subsets of the distributions of each of its vectors, such measures of separation may indicate the relative importance of each vector species in disease transmission.

Results

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Supporting Information

model selection

Annual mean temperature was the key variable determining the distribution of C. imicola in Europe. It was the first variable to be selected in 494 of 500 (99%) bootstrap sample model runs and the average kappa value [± standard deviation (SD)] for single variable models using just annual mean temperature was 0·61 (± 0·045). These models achieved 76·8% (± 4·3%) correct predictions, with a mean sensitivity of 0·81 (± 0·02) and specificity of 0·84 (± 0·03). Overall, adequate predictions of the distribution of C. imicola across Europe could be achieved with three variables in the models. The addition of the second and third variables produced increments of 0·04 on average in the kappa values and 3% in the percentage of correct predictions. Addition of a fourth variable gave increments of only 0·02 in the kappa value and 1% in the percentage of correct predictions. In all 500 of the three-variable bootstrap models, six combinations of three variables were frequently chosen (Table 1), especially combinations 1 and 3. These two combinations included annual mean, annual amplitude and maximum temperature or annual mean temperature, and the bi-annual amplitudes of temperature and summer precipitation as predictors, respectively.

Table 1.  Most commonly selected three-variable model combinations in bootstrap runs of models of Culicoides imicola distribution in Europe
Combination no.Three variables in modelFrequency (%) of model's selection in bootstrap runs
1st2nd3rd
  1. Ann. = annual; Bi-ann. = bi-annual; Amp. = amplitude; Summ. = summer; Wint. = winter; Amt = amount; Temp. = temperature; Prec. = precipitation.

1Ann. mean temp.Ann. amp. temp.Ann. max. temp75 (15%)
2Ann. mean temp.Ann. amp. temp.Bi-ann. amp temp20 (4%)
3Ann. mean temp.Summ. amt. prec.Bi-ann. amp temp71 (14%)
4Ann. mean temp.Summ. amt. prec.Ann. phase. vap.21 (4%)
5Ann. mean temp.Ann. amp. temp.Ann. amt. prec.16 (3%)
6Ann. mean temp.Ann. amp. temp.Wint. amt. prec.15 (3%)

All six three-variable combinations performed well in both testing and training (Table 1, Fig. S1). Kappa values ranged from 0·655 to 0·673 in training vs. 0·610–0·641 in testing and the proportion of correct predictions (presence and absence combined) ranged from 0·828 to 0·838 in training vs. 0·807–0·823 in testing). Combinations 1 and 3 performed the best across test and training partitions and the geographical distribution of C. imicola presence predicted by each model is compared below. The key temperature and precipitation predictor variables in these combinations were used subsequently to define and investigate the climate space occupied by BTV and its vectors in Europe.

climatic predictors of the distribution of c. imicola in europe

Although the single presence cluster and two absence clusters used in the distribution models are defined statistically, and their number based on the best fits of the most parsimonious models, it is biologically informative that areas of C. imicola presence in the east and west Mediterranean Basin can all be captured in models by a single environmental cluster that has broadly consistent properties in terms of key variable means and covariances across this wide geographical area (Fig. S2). Figure 4 shows how these presence and absence clusters map geographically.

image

Figure 4. Areas of agreement and disagreement between two models of Culicoides imicola distribution in Europe overlaid with the presence of BTV in Europe from available commune and point level data only.

Download figure to PowerPoint

Overall, C. imicola occurs in Europe where temperatures are high on average (annual mean temperatures range from 12 to 20 °C) and stable throughout the year (hence low values of the annual amplitude in Table 2). The seasonal range in Fourier-fitted temperatures in presence sites (P) is approximately 13 °C, with temperatures rarely dropping below 9 °C in winter and rising to a broad peak in late summer (August), remaining at or above 23 °C for at least 3 months of the year (Fig. 2). Precipitation levels in areas of C. imicola presence are high in winter and low in summer (summer precipitation amounts range from 0 to 400 mm and winter precipitation amounts from 120 to 1200 mm).

Table 2.  Means (± SD) and medians of important environmental variables across pixels in each observed absence and presence category of the Culicoides imicola distribution in Europe
 UnitsMeans (± sd) absence Group 1 n = 180Group 2 n = 34Presence n = 181Medians absence Group 1 n = 180Group 2 n = 34Presence n = 181Kruskal–Wallis test between groups
χ2Location of pairwise differences
  • *

    All tests significant at the 0·0001 level. A1 is absence category 1, A2 is absence category 2 and P is the presence category of C. imicola. Abbreviations as per Table 1 and Vap. = vapour pressure.

Ann. mean. temp.0·1 °C127·2 (± 22·8) 137·9 (± 14·3)159·4 (± 15·4)124·7163·5277·4164·7*P greater than both A1 and A2
Ann. max. temp 231·3 (± 17·1) 214·7 (± 14·8)241·9 (± 13·9)177·4 71·3242·3 75·08*All pairs different
Ann. amp. temp. 100·7 (± 14·1)  70·5 (± 13·3) 76·5 (± 11·8)283·8 94·6132·1189·7*A1 greater than A2 and P
Bi-ann. amp. temp.   6·7 (± 2·6)  11·6 (± 2·7) 10·3 (± 2·3)117·1295·5260·2169·0*A1 lower than A2 and P
Ann. phase. vap.Decimal months  7·4 (± 0·2)   7·5 (± 0·1)  7·6 (± 0·2)123·0235·9265·5144·7*A1 lower than A2 and P
Ann. amt. prec.mm/year696·5 (± 90·5)1256·3 (± 226·9)716·1 (± 173·3)181·9375·0180·8 89·4*A2 greater than A1 or P
Summ. amt. prec. 277·6 (± 97·2) 437·7 (± 101·5)225·4 (± 85·3)214·5345·6153·9 87·6*All pairs different
Wint. amt. prec. 418·9 (± 102·3) 818·7 (± 144·8)490·7 (± 136·7)150·8369·3212·8110·4*All pairs different
image

Figure 2. Fourier fitted series of temperature across different categories of presence and absence of Culicoides imicola in Europe. A1 is absence category 1, A2 is absence category 2 and P is the presence category of C. imicola.

Download figure to PowerPoint

C. imicola is absent in the east and central Mediterranean Basin (absence category 1), where conditions are cooler (annual mean temperatures range from 0 to 18 °C; Fig. S2, Table 2) than in areas of C. imicola presence. In the summer, temperatures in absence areas reach similar highs to those in areas of C. imicola presence (around 23 °C), but winter temperatures drop much lower (as low as 4 °C). Though absence areas have the same average annual precipitation levels as C. imicola presence locations, they are wetter in summer (summer precipitation amounts range from 0 to 500 mm; Fig. S2) and drier in winter.

C. imicola is also absent in the western Mediterranean Basin (absence category 2), which also tends to be cooler on average (annual mean temperatures range from around 10–17 °C; Fig. S2) than areas of presence. In this region both presence and absence areas have a narrow seasonal temperature range, dropping below 7 °C in winter and rising to a broad peak of around 21 °C in late summer (August) (Fig. 2). Year-round precipitation levels are higher (summer precipitation amounts range from 200 to 700 mm; Fig. S2) in areas of C. imicola absence category 2 than in areas either of C. imicola presence or of C. imicola absence category 1 (Table 2).

predicted distribution of c. imicola in europe

The colours from yellow to red in Fig. 3 indicate the Bayesian posterior probability (P) (from 0·5 < P < 1) that each pixel in Europe was suitable for C. imicola presence during the late 1990s derived from the two best three-variable model combinations (1 and 3 in Table 1). The predicted geographical distributions of C. imicola are identical in 93% of sites (kappa value comparing two sets of predictions = 0·85, Fig. 4) and each corresponds closely to the observed distribution of C. imicola in the areas that have been sampled. Areas of disagreement between the two models, and the small number of false predictions made by each one, tend to be distributed along the northern range limit of C. imicola and in large parts of North Africa (Tunisia and Libya), which is predicted to be unfavourable for C. imicola by model 1 but favourable by model 3 (Fig. 4).

image

Figure 3. Predicted probability of the presence of Culicoides imicola s.s. (a) from model 1 with annual mean temperature, annual amplitude of temperature and annual maximum temperature (b) model 3 with annua; mean temperature, summer amount of precipitation, bi-annual amplitude of temperature. No prediction is made when a pixel is too environmentally dissimilar from the training presence–absence data set, defined as pixels for which the Mahalanobis Distance (MD) exceeds 30 MD units. See Fig. 2 for key to site classes.

Download figure to PowerPoint

overlap of btv and its vectors in european climate space

When the C. imicola predictions shown in Fig. 4 are overlain with data for the current distribution of BTV in Europe, constructed only from available commune- and point-level data (Fig. 4), it is striking that a large numbers of locations affected by BTV in central and eastern Europe lie north of the predicted distribution limit of C. imicola.

The overlap of the distribution of BTV and C. imicola in climate space in Europe is shown in Fig. 5a,c,e. On the two axes shown in Fig. 5a, 43·5% (218/501) of BTV records fall outside the 95% confidence ellipse of C. imicola presence points, and do so in the directions of much lower annual mean temperatures (to 6 or 7 °C) and greater ranges of annual temperature (> 9 °C). BTV transmission, below an annual mean temperature of 15 °C, also extends into locations with more than 500 mm to around 700 mm of summer precipitation. On the pairs of axes indicated in Fig. 5b,c, 44·5% (223/501) and 44·7% (224/501) of BTV records, respectively, fall outside the 95% confidence ellipse of C. imicola presence points, into wetter as well as cooler areas.

image

Figure 5. Climate space occupied by BTV and its vectors. The axes of bivariate climate space are defined by different pairings of three important determinants of Culicoides imicola distribution annual mean temperature (in 0·1 °C), annual amplitude of temperature (in 0·1 °C), and summer amount of precipitation (in mm per summer). In (a) (c) and (e) the overlap between the presence of BTV and the presence of C. imicola in environmental space is shown. In (b) (d) and (f), their overlap with the presence of other Palearctic Culicoides is also shown.

Download figure to PowerPoint

When the sampled distributions of the C. obsoletus (grey circles) and C. pulicaris (grey crosses) complexes are superimposed on these same pairs of axes (Fig. 5b,d,f), they co-occur across most of the environmental space they occupy and also overlap extensively with the climate envelope of C. imicola in Europe (represented here by the 95% ellipse of C. imicola presence) at annual mean temperatures of 12 °C to 17 °C. Importantly, however, they extend into locations that are much cooler, having annual average temperatures of 7–8 °C (particularly in locations that have high annual amplitudes in temperature; Fig. 5b,d) and wetter, having summer precipitation amounts of up to 600 mm (Fig. 5d,f). On all three pairs of axes, their distributions in climate space overlap much of the space beyond the C. imicola envelope (or ellipse) and into which BTV transmission has extended. While the Iberian and Moroccan transmission sites fell broadly inside the C. imicola envelope in two dimensions, those that fell outside were geographically distributed in central and eastern Europe – just south (in Greece, Lazio and Tuscany and Sicily) or just north (in Greece, Bulgaria and the Balkans) of the northern range limit of C. imicola.

On the basis of MDs, areas of BTV transmission were found to be approximately twice as far in multivariate space from areas of C. imicola presence as they were from areas of presence of Palearctic Culicoides, or from sites containing any vector at all (all three vector complexes/species combined) (Fig. 6a). Areas of BTV transmission (232 squares) that occurred within C. imicola's northern range limit were equally close in multivariate space to areas of presence of C. imicola, the C. obsoletus complex and the C. pulicaris complex (Fig. 6b), while those northward of this limit (269 squares) were two to three times further from areas of presence of C. imicola than they were from areas of presence of the Palearctic Culicoides, or from sites containing any vector at all (Fig. 6c).

image

Figure 6. Box-plots comparing average accuracy statistics across Mahalanobis distances (MD) between the distributions of BTV and each different vector complexes or species for (a) all BTV sites, (b) BTV sites south of Culicoides imicola's northern range limit and (c) BTV sites north (> 0·5 degrees) of C. imicola's northern range limit. In each case, MD values are calculated across 10 000 pairs of randomly selected BTV sites and vector presence sites (adjusted for the covariance of the entire distribution data set of the vector). The boxes have lines at the lower quartile, median and upper quartile values while whiskers extending from the boxes indicate the full extent of data set and red crosses indicate outliers. Notches represent a robust estimate of the uncertainty about the medians for box-to-box comparison. Boxes whose notches do not overlap vertically indicate that the medians of the two complexes differ at the 5% significance level.

Download figure to PowerPoint

Discussion

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Supporting Information

climatic predictors of the distribution of c. imicola in europe

The association of C. imicola populations with locations that are warm on average and remain warm year-round, indicated by the models presented here, is broadly consistent with recent subcontinental studies of this species’ distribution (Wittman et al. 2001; Conte et al. 2003; De Liberato et al. 2003). Wittman et al. (2001), for example, found that at least 8 months with a mean temperature greater than 12·5 °C favoured the presence of C. imicola at sites in Iberia. The precise biological basis of these associations for C. imicola is unknown, but many population processes of Culicoides are known to be highly temperature-dependent (Mellor et al. 2000; Wittmann & Baylis 2000).

C. imicola populations are also associated with locations that are reasonably dry, particularly in summer (with some populations withstanding zero precipitation in this season). This is consistent with this species’ habit of breeding in wet or damp, but not flooded, organically enriched soil (Braverman et al. 1974; Walker 1977; Braverman 1978). In Europe, the peak adult activity of C. imicola s.s., and thus the peak requirement for moist breeding sites, coincides with late summer/early autumn. This association in Europe (cf. Rawlings et al. 1998) mirrors this species’ requirements in core areas of its distribution such as South Africa, where populations are restricted to areas with rainfall of 300–700 mm per year (Meiswinkel & Baylis 1998). Areas with higher rainfall are unlikely to be suitable for C. imicola because the pupae and larvae of this species are unable to swim or float in free water, as do those of many other Culicoides species, and so drown or are washed away when the breeding sites are flooded (Nevill 1971). In comparisons in Cyprus and Israel, C. imicola was observed to breed in habitats that were generally drier than those preferred by other species of Culicoides (Braverman et al. 1974; Mellor & Pitzolis 1979).

Although clustering of the presence and absence data was employed to achieve multinormality – a statistical prerequisite for discriminant analysis – it is interesting that the presence of C. imicola was best described throughout the Mediterranean using a single presence cluster only, while two clusters or types were required to describe locations in which C. imicola is absent, corresponding broadly to a western vs. a central and eastern Mediterranean absence type. Comparing the properties of each of these clusters suggests that the combination of seasonal conditions of temperature and precipitation, rather than temperature alone (as suggested by Baylis & Rawlings 1998; Rawlings et al. 1998; Wittman et al. 2001) limits this species’ expansion northward in the Mediterranean and the species faces different challenges to further expansion in the East and West of this region. In the north-west Mediterranean (e.g. in northern Portugal), C. imicola was absent in areas that were cooler and wetter year-round than areas of C. imicola presence. Because most habitats in central and northern areas of Europe receive substantial amounts of precipitation, the risk that C. imicola breeding sites will be flooded may be uniformly high and thus limit significantly this species’ expansion northward across the continent. Locations in central and eastern areas where the species was absent (e.g. north-west Greece and Bulgaria) were as warm in summer but dropped to much lower temperatures in winter. They were also wetter in summer, when the demand for moist but not flooded breeding sites is highest, and drier in winter.

It is, of course, probable that many other factors such as soil type (Baylis et al. 1999), soil moisture availability (or a correlate, the Normalized Difference Vegetation Index – Baylis et al. 1998; Baylis & Mellor 2001; Tatem et al. 2003), slope and farm husbandry methods (Meiswinkel et al. 2000) determine the distribution of C. imicola populations at finer spatial scales. For example, within quite large areas that are too wet on average in summer, C. imicola may be able to select patches of drier microhabitats where breeding sites will not become flooded. Models based on environmental data sets at coarse spatial resolution will record such areas as ‘too wet’ rather than as ‘suitable’. Nevertheless, the availability of such dry patches for selection may be reduced within a ‘wet’ square compared to a ‘dry’ square. In fact, the predominance of macroclimate in determining distribution patterns of organisms at scales of 10 km2 and above is widely recognized (Andrewartha & Birch 1954; Currie 1991) and our study of the climatic envelope of C. imicola at a similar spatial scale has revealed consistent characteristics of the complex relationship between this species and the climate across the Mediterranean. This analysis provides a useful baseline estimate of the current (late 1990s) climatic limits of C. imicola across Europe, against which to test for past and future changes in these limits. It also allows broad-scale definition of those areas currently at risk of BTV transmission by this species.

overlap of btv and its vectors in environmental space in europe

Palearctic Culicoides are not only abundant and widespread in northern Europe, but also extend southward into North Africa (Szadziewski 1984; Baylis et al. 1997), Turkey (Jennings 1983) and the Middle East (Braverman & Galun 1973; Braverman et al. 1974). Despite their extensive geographical overlap with both C. imicola and areas of historical BTV incursions there is little evidence, either from temporal or fine-scale spatial overlap with the distribution of historical outbreaks, that Palearctic Culicoides previously played a major role in transmission within the historical limits of C. imicola in Africa (Mellor et al. 1984; Mellor et al. 1985). By contrast, our study supports a major role for these alternative vectors to C. imicola during the current epidemic. Indeed, the overall distribution of BTV transmission was closer to the distributions of the Palaearctic species complexes (both the C. obsoletus and C. pulicaris complexes), on average, than it was to the distribution of C. imicola, in both bivariate and multivariate environmental space. This indicates that these species complexes make a significant contribution to BTV's distribution across Europe, not only northward of the range of C. imicola (in south-central and south-east Europe) but also within C. imicola's range (where the distribution of BTV inside the range was not appreciably closer in climate space to the distribution of C. imicola than it was to the distributions of the Palearctic species complexes). The fact that transmission sites located northward of the range of C. imicola are not only geographically distant from known C. imicola populations but also lie well outside the climatic envelope of this species make it less plausible that the presence of this species has simply been overlooked by surveillance efforts in these areas. Moreover, this species is unlikely to move into these more northerly areas in the near future, without a substantial shift in its climatic envelope. This leaves unanswered, however, the question of why the Palearctic vectors were apparently unimportant for BTV transmission historically in Africa.

We acknowledge that BTV was recorded as present in a few restricted areas of environmental space where the Palearctic vectors were recorded as ‘absent’. Although other vectors may be important in these locations, we suspect that most of these gaps will be filled as surveillance efforts are directed at detecting these Palaearctic species. Nevertheless, this illustrates the importance of continually updating our picture of the environmental distributions of pathogens and vectors as new data become available.

The overlap in Europe between the Palaearctic Culicoides and C. imicola affords ample opportunity for ‘hand-over’ events of the virus between the traditional and novel vectors to occur (cf. Mellor 1996). Predicting where in geographical and environmental space such events are most likely is crucial. The advantage of using discriminant analysis or logistic regression to describe environmental envelopes rather than alternative programs such as garp (Peterson & Shaw 2003; Peterson et al. 2004) is that the contribution of different environmental variables to the model can be assessed clearly and used to define important dimensions in which to visualize species’ interactions most effectively (such as those between different vectors) in environmental space. This is particularly powerful when combined with temporal Fourier processing to extract information on the seasonality of climate that is crucial in driving variation in vector-borne diseases (Rogers & Randolph 2003). In addition, Mahalanobis distances can be used to assess the broad-scale overlap and shifts of species’ distributions in multivariate environmental space for any set of interacting species (e.g. herbivore–plant or predator–prey systems). As the latter reduces several environmental dimensions to a one-dimensional measure of separation, it overcomes the lack of tools for visualizing environmental space in more than three dimensions simultaneously (Soberon & Peterson 2005).

The environmental envelope modelling approach holds particular promise for examining the broad-scale geographical and environmental interactions of hosts, vectors and pathogens, because arthropod vectors are highly sensitive to climate and their environmental envelopes can probably be summarized adequately with a few important environmental dimensions. They are also dispersive, and may be expected to ‘fill’ rapidly all available environmental space in a region and respond quickly to changes in the geographical distribution of environmental space following environmental change or disturbance. This is in contrast to the rare, habitat specialists to which these models are most often applied in the field of conservation.

Most importantly, this study indicates the degree to which we need to adjust our historical picture of the seasonal conditions of temperature and precipitation that favour BTV transmission, and therefore the likelihood of its invasion elsewhere. Sellers & Mellor (1993) found that the 12·5 °C isotherm for the coldest month of the year broadly delineated the historical distribution of Culicoides-borne disease outbreaks (as well as locations where adult C. imicola could survive year-round) in the Mediterranean. Transmission now occurs in locations where the average annual (rather than winter) temperatures are as low as 12 °C. These conditions (outside the envelope of C. imicola) are within the distributions of the Palearctic Culicoides. These vectors will allow BTV to extend into colder, wetter locations, with annual average temperatures as low as 7–8 °C and summer precipitation amounts of up to 600–700 mm. Until 2004, BTV had only occupied the southern-most portion of these novel vector species complexes’ distributions. In summer 2006, however, bluetongue virus suddenly appeared at latitudes of more than 50° North, in Belgium, Germany and the Netherlands (Anonymous 2006a, b, c) – approximately 6° further North than previous outbreaks in southern Europe and, significantly, well within the core of the ranges of the Palearctic Culicoides. How far bluetongue might extend in the future, within the distributions of these Palaearctic species complexes, depends inter alia on the thermal limits to viral replication (Van Dijk & Huismans 1982), as well as on the distributions and relative capacity to act as BTV vectors, of subspecies within these complexes.

Our study highlights the dynamism of the BTV–Culicoides system in Europe. Future surveillance and research effort for bluetongue and for related Culicoides-borne pathogens (such as African horse sickness virus and epizootic haemorrhagic disease virus that have historically shared similar vectors) should aim to record and explain the distributional patterns, host and habitat preferences of all potential Palearctic vector complexes as well as the traditional, African–Asian vectors.

Acknowledgements

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Supporting Information

BVP is currently supported by BBSRC/DEFRA Research Grant (BBSRC grant no. BBS/B/00603, Defra Grant no. SE4104) ‘Epidemiology & control of orbiviral diseases in the UK, with particular reference to bluetongue & African horse sickness’ and was previously supported by EU project (Contract no. QLK2-CT-2000–00611). The authors would like to thank Prof Sarah Randolph for helpful comments on the manuscript.

References

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Supporting Information
  • Andrewartha, H.G. & Birch, L.C. (1954) The Distribution and Abundance of Animals. University of Chicago Press, Chicago.
  • Anonymous (2006a) Bluetongue in the Netherlands. OIE Disease Information, 34, 612.
  • Anonymous (2006b) Bluetongue in Belgium. OIE Disease Information, 34, 616.
  • Anonymous (2006c) Bluetongue in Germany. OIE Disease Information, 34, 618.
  • Austin, M.P., Nicholls, C.R. & Margules, C.R. (1990) Measurement of the realised quantitative niche: environmental niches of five Eucalyptus species. Ecological Monographs, 60, 161177.
  • Baylis, M., Bouayoune, H., Touti, J. & El Hasnaoui, H. (1998) Use of climatic data and satellite imagery to model the abundance of Culicoides imicola, the vector of African horse sickness virus, in Morocco. Medical and Veterinary Entomology, 12, 255266.
  • Baylis, M., El Hasnaoui, H., Bouayoune, H., Touti, J. & Mellor, P.S. (1997) The spatial and seasonal distribution of African horse sickness and its potential Culicoides vectors in Morocco. Medical and Veterinary Entomology, 11, 203212.
  • Baylis, M., Meiswinkel, R. & Venter, G.J. (1999) A preliminary attempt to use climate data and satellite imagery to model the abundance and distribution of Culicoides imicola (Diptera: Ceratopogonidae) in southern Africa. Journal of the South African Veterinary Association – Tydskrif Van die Suid-Afrikaanse Veterinere Vereniging, 70, 8089.
  • Baylis, M. & Mellor, P.S. (2001) Bluetongue around the Mediterranean in 2001. Veterinary Record, 149, 659.
  • Baylis, M., Mellor, P.S., Wittmann, E.J. & Rogers, D.J. (2001) Prediction of areas around the Mediterranean at risk of bluetongue by modelling the distribution of its vector using satellite imaging. Veterinary Record, 149, 639643.
  • Baylis, M. & Rawlings, P. (1998) Modelling the distribution and abundance of Culicoides imicola in Morocco and Iberia using climatic data and satellite imagery. Archives of Virology, Suppl. 14 (1998), 137153.
  • Braverman, Y. (1978) Characteristics of Culicoides (Diptera, Ceratopogonidae) breeding places near Salisbury, Rhodesia. Ecological Entomology, 3, 163170.
  • Braverman, Y. & Galun, R. (1973) The occurrence of Culicoides in Israel with reference to the incidence of bluetongue.Refuah Veterinaria, 30, 121127.
  • Braverman, Y., Galun, R. & Ziv, M. (1974) Breeding sites of some Culicoides species (Diptera-Ceratopogonidae) in Israel. Mosquito News, 34, 303308.
  • Calistri, P., Goffredo, M., Caporale, V. & Meiswinkel, R. (2003) The distribution of Culicoides imicola. Italy: application and evaluation of current Mediterranean models based on climate. Journal of Veterinary Medicine, Series B, 50, 132138.
  • Carpenter, S., Lunt, H.L., Arav, D., Venter, G.J. & Mellor, P.S. (2006) Oral susceptibility to bluetongue virus of Culicoides (Diptera: Ceratopogonidae) from the United Kingdom. Journal of Medical Entomology, 43, 7378.
  • Conte, A., Giovannini, A., Savini, L., Goffredo, M., Calistri, P. & Meiswinkel, R. (2003) The effect of climate on the presence of Culicoides imicola in Italy. Journal of Veterinary Medicine, Series B, 50, 139147.
  • Currie, D.J. (1991) Energy and large scale patterns of animal and plant species richness. American Naturalist, 137, 2749.
  • De Liberato, C., Purse, B.V., Goffredo, M., Scholl, F. & Scaramozzino, P. (2003) Geographical and seasonal distribution of the bluetongue virus vector, Culicoides imicola, in central Italy. Medical and Veterinary Entomology, 17, 388394.
  • De Liberato, C., Scavia, G., Lorenzetti, R., Scaramozzino, P., Amaddeo, D., Cardeti, G., Scicluna, M., Ferrari, G. & Autorino, G.L. (2005) Identification of Culicoides obsoletus (Diptera: Ceratopogonidae) as a vector of bluetongue virus in central Italy. Veterinary Record, 156, 301304.
  • European Commission (EC) (2004) Commission Decision (2004/34/EC) of 6 January 2004 amending Decision 2003/828/EC with regards to the movements of vaccinated animals from protection zones. Official Journal of the European Union, L7/47, 13/01/2004.
  • Guisan, A. & Zimmerman, N.E. (2000) Predictive habitat distribution models in ecology. Ecological Modelling, 135, 147186.
  • Hutchinson, G.E. (1957) Concluding remarks. Cold Spring Harbor Symposia on Quantitative Biology, 22, 415427.
  • Jennings, D.M. (1983) Culicoides from Western Turkey in relation to bluetongue disease of sheep and cattle. Revue d’Elevage et de Médecine Vétérinaire Des Pays Tropicaux, 36, 6770.
  • Meiswinkel, R. & Baylis, M. (1998) Morphological confirmation of the separate species status of Culicoides (Avaritia) nudipalpis Delfinado, 1961 and C. (A.) imicola Kieffer, 1913 (Diptera: Ceratopogonidae). Onderstepoort Journal of Veterinary Research, 65, 916.
  • Meiswinkel, R., Baylis, M. & Labuschagne, K. (2000) Stabling and the protection of horses from Culicoides bolitinos (Diptera: Ceratopogonidae), a recently identified vector of African horse sickness. Bulletin of Entomological Research, 90, 509515.
  • Mellor, P.S. (1996) Culicoides: vectors, climate change and disease risk. Veterinary Bulletin, 66, 301306.
  • Mellor, P.S. & Boorman, J. (1995) The transmission and geographical spread of African horse sickness and bluetongue viruses. Annals of Tropical Medicine and Parasitology, 89, 115.
  • Mellor, P.S., Boorman, J. & Baylis, M. (2000) Culicoides biting midges: their role as arbovirus vectors. Annual Review of Entomology, 45, 307340.
  • Mellor, P.S., Jennings, M. & Boorman, J.P.T. (1984) Culicoides from Greece in relation to the spread of bluetongue virus. Revue d’Elevage et de Medecine Veterinaire Des Pays Tropicaux, 37, 286289.
  • Mellor, P.S., Jennings, D.M., Wilkinson, P.J. & Boorman, J.P.T. (1985) Culicoides imicola – a bluetongue virus vector in Spain and Portugal. Veterinary Record, 116, 589590.
  • Mellor, P.S. & Pitzolis, G. (1979) Observations on breeding sites and light-trap collections of Culicoides during an outbreak of bluetongue in Cyprus. Bulletin of Entomological Research, 69, 229234.
  • Mellor, P.S. & Wittmann, E.J. (2002) Bluetongue virus in the Mediterranean Basin 1998–2001. Veterinary Journal, 164, 2037.
  • Mitchell, T.D., Carter, T.R., Jones, P.D., Hulme, M. & New, M. (2004) A comprehensive set of high-resolution grids of monthly climate for Europe and the globe: the observed record (1901–2000) and 16 scenarios (2001–2100). Tyndall Centre Working Paper, 55.
  • Monteys, V.S.I. & Saiz-Ardanaz, M. (2003) Culicoides midges in Catalonia (Spain), with special reference to likely bluetongue virus vectors. Medical and Veterinary Entomology, 17, 288293.
  • Nevill, E.M. (1971) Cattle and Culicoides biting midges as possible overwintering hosts of bluetongue virus. Onderstepoort Journal of Veterinary Research, 38, 6572.
  • Peterson, A.T., Bauer, J.T. & Mills, J.N. (2004) Ecologic and geographic distribution of Filovirus disease. Emerging Infectious Diseases, 10, 4247.
  • Peterson, A.T. & Shaw, J. (2003) Lutzomyia vectors for cutaneous leishmaniasis in Southern Brazil: ecological niche models, predicted geographic distributions and climate change effects. International Journal for Parasitology, 33, 919931.
  • Purse, B.V., Mellor, P.S., Rogers, D.J., Samuel, A.R., Mertens, P.P.C. & Baylis, M. (2005) Climate change and the recent emergence of bluetongue in Europe. Nature Reviews Microbiology, 3, 171181.
  • Rawlings, P., Capela, R., Pro, M.J., Ortega, M.D., Pena, I., Rubio, C., Gasca, A. & Mellor, P.S. (1998) The relationship between climate and the distribution of Culicoides imicola in Iberia. Archives of Virology, Suppl. 14 (1998), 93102.
  • Rogers, D.J. (2006) Models for vectors and vector-borne diseases. Advances in Parasitology, 62, 135.
  • Rogers, D.J., Hay, S.I. & Packer, M.J. (1996) Predicting the distribution of tsetse flies in West Africa using temporal Fourier processed meteorological satellite data. Annals of Tropical Medicine and Parasitology, 90, 225241.
  • Rogers, D.J. & Randolph, S.E. (2003) Studying the global distribution of infectious diseases using GIS and RS. Nature Reviews: Microbiology, 1, 231236.
  • Savini, G., Goffredo, M., Monaco, F., Di Gennaro, A., Cafiero, M.A., Baldi, L., De Santis, P., Meiswinkel, R. & Caporale, V. (2005) Bluetongue virus isolations from midges belonging to the Obsoletus complex (Culicoides, Diptera: ceratopogonidae) in Italy. Veterinary Record, 157, 133139.
  • Savini, G., Goffredo, M., Monaco, F., De Santio, P. & Meiswinkel, R. (2003) Transmission of bluetongue virus in Italy. Veterinary Record, 152, 119.
  • Sellers, R.F. & Mellor, P.S. (1993) Temperature and the persistence of viruses in Culicoides spp. during adverse conditions.Revue Scientifique et technique de l’office International des Epizooties, 12, 733755.
  • Soberon, J. & Peterson, A. (2005) Interpretation of models of fundamental ecological niches and species’ distributional areas. Biodiversity Informatics, 2, 110.
  • Szadziewski, R. (1984) Ceratopogonidae from Algeria VI Culicoides LATR. Polskie Pismo Entomologiczne, 54, 163182.
  • Tatem, A.J., Baylis, M., Mellor, P.S., Purse, B.V., Capela, R., Pena, I. & Rogers, D.J. (2003) Prediction of bluetongue vector distribution in Europe and north Africa using satellite imagery. Veterinary Microbiology, 97, 1329.
  • Torina, A., Caracappa, S., Mellor, P.S., Baylis, M. & Purse, B.V. (2004) Spatial distribution of bluetongue virus and its Culicoides vectors in Sicily. Medical and Veterinary Entomology, 18, 8189.
  • Van Dijk, A.A. & Huismans, H. (1982) The effect of temperature on the in vitro transcriptase reaction of bluetongue virus, epizootic haemorrhagic disease virus and african horse sickness virus. Onderstepoort Journal of Veterinary Research, 49, 227232.
  • Walker, A.R. (1977) Seasonal fluctuations of Culicoides species (Diptera-Ceratopogonidae) in Kenya. Bulletin of Entomological Research, 67, 217233.
  • Wittman, E.J., Mellor, P.S. & Baylis, M. (2001) Using climate data to map the potential distribution of Culicoides imicola (Diptera: Ceratopogonidae) in Europe. Revue Scientifique et Technique de l’Office International Des Epizooties, 20, 731740.
  • Wittmann, E.J. & Baylis, M. (2000) Climate change: effects on Culicoides-transmitted viruses and implications for the UK. Veterinary Journal, 160, 107117.

Supporting Information

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Supporting Information

The following supplementary material is available for this article.

Appendix S1.Detailed Methods and Culicoides distribution data.

Fig. S1. Box plots comparing average accuracy statistics across training sets and test sets for six three-variable models of C. imicola distribution in Europe.

Fig. S2. Climate space occupied by locations in presence and absence categories of C. imicola in Europe, visualised on bivariate scatter plots of important climatic determinants of distribution

This material is available as part of the online article from: http://www.blackwell-synergy.com/doi/full/10.1111/j.1365-2664.2007.01342.x

Please note: Blackwell Publishing is not responsible for the content or functionality of any supplementary material supplied by the authors. Any queries (other than missing material) should be directed to the corresponding author for the article.

FilenameFormatSizeDescription
JPE_1342_sm_FigS1.doc37KSupporting info item
JPE_1342_sm_FigS2.doc125KSupporting info item

Please note: Wiley Blackwell is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.