Spatial and temporal variation in the abundance of Culicoides biting midges (Diptera: Ceratopogonidae) in nine European countries

Background Biting midges of the genus Culicoides (Diptera: Ceratopogonidae) are vectors of bluetongue virus (BTV), African horse sickness virus and Schmallenberg virus (SBV). Outbreaks of both BTV and SBV have affected large parts of Europe. The spread of these diseases depends largely on vector distribution and abundance. The aim of this analysis was to identify and quantify major spatial patterns and temporal trends in the distribution and seasonal variation of observed Culicoides abundance in nine countries in Europe. Methods We gathered existing Culicoides data from Spain, France, Germany, Switzerland, Austria, Denmark, Sweden, Norway and Poland. In total, 31,429 Culicoides trap collections were available from 904 ruminant farms across these countries between 2007 and 2013. Results The Obsoletus ensemble was distributed widely in Europe and accounted for 83% of all 8,842,998 Culicoides specimens in the dataset, with the highest mean monthly abundance recorded in France, Germany and southern Norway. The Pulicaris ensemble accounted for only 12% of the specimens and had a relatively southerly and easterly spatial distribution compared to the Obsoletus ensemble. Culicoides imicola Kieffer was only found in Spain and the southernmost part of France. There was a clear spatial trend in the accumulated annual abundance from southern to northern Europe, with the Obsoletus ensemble steadily increasing from 4000 per year in southern Europe to 500,000 in Scandinavia. The Pulicaris ensemble showed a very different pattern, with an increase in the accumulated annual abundance from 1600 in Spain, peaking at 41,000 in northern Germany and then decreasing again toward northern latitudes. For the two species ensembles and C. imicola, the season began between January and April, with later start dates and increasingly shorter vector seasons at more northerly latitudes. Conclusion We present the first maps of seasonal Culicoides abundance in large parts of Europe covering a gradient from southern Spain to northern Scandinavia. The identified temporal trends and spatial patterns are useful for planning the allocation of resources for international prevention and surveillance programmes in the European Union. Electronic supplementary material The online version of this article (10.1186/s13071-018-2706-y) contains supplementary material, which is available to authorized users.

At least 83 species of Culicoides are found in Europe (83 species reported in France [9]) but only some of these are suspected to transmit viruses: the afrotropical vector C. imicola was previously considered to be the main vector of BTV in southern European countries [10], though BTV virus was also isolated from wild specimens of Culicoides obsoletus (Meigen)/Culicoides scoticus Downes & Kettle and specimens of the Pulicaris ensemble [11][12][13][14]. During an unprecedented outbreak of BTV serotype 8 in northern Europe in 2006, it became evident that autochthonous Culicoides species of the subgenus Avaritia, specifically C. obsoletus/C. scoticus and possibly Culicoides dewulfi Goetghebuer, were transmitting the virus [15][16][17][18][19].
In 2000, the European Commission established a series of regulations for BTV control, including monitoring and surveillance in the affected countries. According to Commission Regulation (EC) No. 1266/2007, it is mandatory for member states to carry out bluetongue surveillance programmes that include vector monitoring [20]. As a result of the BTV outbreak in 2006, the northern European countries also started carrying out entomological surveillance of Culicoides vectors. Collections by light traps at ruminant farms have been a key component of these programmes in both southern and northern Europe.
Several European countries have gathered and analysed entomological data at a national level to determine the presence and abundance of different species of Culicoides [21][22][23][24][25]. In addition, vector activity during the winter has been assessed in an attempt to determine the existence of a "vector-free period" [26,27]. Determining a vector-free period might be useful for national veterinary authorities to authorize movements of test-negative ruminants. The number of national studies from European countries has increased during the last decade, but the need to quantify Culicoides vector dynamics at a continental level remains, as European legislation for vector-borne diseases is founded on joint decisions among the member states.
The aim of the present study was to generate a joint entomological database for large parts of Europe comprising different climatic zones, using surveillance and research data collected during the period 2007-2013. Nine European countries (Spain, France, Germany, Austria, Switzerland, Denmark, Sweden, Norway and Poland) agreed to share data and quantify key temporal and largescale geographical trends in the abundance of two main Culicoides species ensembles (Obsoletus ensemble and Pulicaris ensemble) and C. imicola.
Black light suction traps were placed outside stables or near animal resting sites and the coordinates of each farm were recorded. Onderstepoort traps (Onderstepoort Veterinary Institute, Pretoria, Republic of South Africa) were used to catch Culicoides from dusk until dawn in all countries [23,29,31,32,34] except Germany, where Biogents Sentinel (BG-Sentinel) traps (BioGents, Regensburg, Germany) fitted with a black light lamp were used [24], and Spain, where mini CDC model 1212 (John W. Hock, Gainesville, FL, USA) traps were used [35]. Onderstepoort traps have been reported to catch more Culicoides than the other two types of traps [36,37]. Therefore, data obtained by the BG-Sentinel and mini CDC traps were converted to the number of specimens estimated to have been collected if Onderstepoort traps had been used. For BG-sentinel traps, Venter et al. [36] calculated a conversion factor of 3.1, while Probst et al. [37] calculated a conversion factor of 3.83. We used the mean of these values (3.48) as the conversion factor in this study. Three trap efficiencies (0.404, 0.288 and 0.505) were previously reported for the CDC mini trap [36] .We used the average of these values (0.399) and used the reciprocal conversion factor of 2.51 for all vector species. A single trap per farm was used in all the countries with the exception of Germany. During the 2012-2013 campaign Germany operated 3 traps per farm, so in order to have only one observation per farm we took the median amount of Culicoides caught among the 3 traps.
Some Culicoides species are difficult to identify based on morphology, e.g. C. obsoletus/C. scoticus [13,18,[38][39][40][41]. In the data available for this analysis, specimens were identified to species level in some countries, while other countries only identified them to group level. To create a uniform database, we aggregated the species into ensembles. We use the term "ensemble" to refer to a group of sympatric species for which morphological identification is sometimes not possible or difficult, and without phylogenetic meaning. In this work, "Obsoletus ensemble" refers to the Obsoletus group and C. dewulfi together and includes the following species: C. obsoletus, C. scoticus, Culicoides montanus Shakirzjanova, Culicoides chiopterus (Meigen) and C. dewulfi. The Pulicaris ensemble includes Culicoides pulicaris (Linnaeus) and Culicoides punctatus (Meigen). As it is commonly used in the literature while authors differ in its composition, we refer to the Obsoletus group as a species group including C. obsoletus, C. scoticus, C. montanus and C. chiopterus. Based on the phylogenetic analysis of Schwenkenbecher et al. [42], we considered C. dewulfi as a separate species from Obsoletus group. We focused on these seven species, as they are considered to be farm-associated species [27,29]. Culicoides imicola specimens were identified to species level by the two countries in which they were found (Spain and France). The sampling period is shown in Additional file 1: Table S1 while the number of trapping farms per country, trap type, and national protocols of the specific country are presented in Table 1.

European temperature data
We obtained daily temperature data from Europe between 1994 and 2004 from MARS-Agri4cast. As  previously described by Beek [43], these data resulted from a linear interpolation of weather stations distributed across Europe into regular climate grids of 25 × 25 km.

Descriptive analysis and data management
We calculated the week number of each collection using the start date of the trapping. We defined week 1 of each year as 1st to 7th of January. This was done to ensure that the same dates from different years were given the same week number. We calculated the weekly and monthly mean abundance of vectors for each year. Finally, we calculated the average weekly and monthly abundance for the entire seven-year period to derive estimates for an "average year". We conducted three different analyses. In the first analysis, we divided Europe into seven latitudinal bands (A-G) of 5°width, from 35°N to 70°N (Fig. 1). Latitude range G (> 65°N) contained only two farms with just 9 observations from 4 weeks in August and September 2008, and was therefore not included in the latitudinal range analysis. To compare the seasonal variation among the seven different latitude ranges, we log transformed the trap collection data [log 10 (x + 1)] and then calculated the mean of all the trap collections for each week number and at each latitude range, based on the data for the entire 7-year period (Fig. 1). To quantify the variation of the mean abundance for each week number, we calculated the 10th and 90th percentiles. For each latitudinal range, we also calculated the weekly average of the daily minimum, mean and maximum temperatures per week for the period from 1994 to 2004, and contrasted with the Culicoides seasonal variation. We estimated the number of vectors collected in an average year within each latitudinal range by calculating the annual cumulative sum of the weekly mean abundance and multiplying this by 7 days.
In the second analysis, we calculated the average abundance on each farm for each of the 12 months [log 10 (x + 1)]. We then spatially interpolated the logtransformed monthly averages to create spatial abundance maps. The interpolation was done using Inverse Squared Distance Weight (IDW) and based on the 15 nearest trap locations in ArcGIS 10.1 (ESRI, Redlands, CA, USA). We created buffer zones of 200 km around each farm and excluded all areas beyond this limit to avoid extrapolating to unsampled areas. In addition, countries outside the area of analysis were not included in the map.
In the third analysis we examined the spatial pattern of the start of the vector season by plotting the season start date of each NUTS (nomenclature of territorial units for statistics) 3rd level polygons defined by Eurostat (1992). NUTS is a hierarchical system used to divide up the economic territory of the EU for statistical purposes. We calculated the average start of the season for the 7-year period. The start of the season was defined as the first month in which the average monthly abundance per area polygon was equal to or higher than one specimen of C. imicola, and equal to or higher than 5 specimens of the Obsoletus and Pulicaris ensemble. The threshold numbers used here were based on (but not identical to) the threshold numbers defined by the European Commission to determine the start of the season [20]. While the EU thresholds are based on individual traps, we applied the thresholds to the average vector densities of all traps within a polygon. Polygons were classified as having "no data" if (i) the polygon did not have any sampled farms, (ii) if the mean abundance of the polygon did not reach the threshold during the year, or if (iii) there were no data available for the month prior the start of the season, thus making it impossible to detect whether the season might have started earlier.

Results
The trap data made available for our analyses included a total of 8,842,998 biting midges that had been collected at 904 farms between 2007 and 2013 in nine European countries. Of these, 82.8% belonged to the Obsoletus ensemble, 12.0% to the Pulicaris ensemble and 5.1% were C. imicola. Biting midges of the Obsoletus and Pulicaris ensemble were found in all countries that were sampled, while C. imicola was only found in Spain, along the southern coast of France and in Corsica.

Culicoides temporal fluctuation by latitude range
We observed a large variation in abundance in individual traps at each latitude range and at each week number, with the weekly 10th and 90th percentiles varying by a factor of 100 or more. However, when examining the average seasonal abundance, we were still able to observe two main patterns for the two species ensembles and C. imicola. First, the annual peak abundance in the Obsoletus ensemble increased gradually from Latitude A until it reached the highest weekly average peak of over 1000 vectors per night in Latitude F (Fig. 2, left column). Secondly, the period of the vector activity became increasingly shorter at higher latitudes, lasting throughout the whole year at Latitudes A and B (Fig. 2, left column), but only from week 15 (April) to week 46 (November) at Latitude F. Despite the increasingly shorter season for the Obsoletus ensemble further north, the cumulative sum of the weekly abundance steadily increased with latitude from less than 1000 Obsoletus ensemble vectors per year on average at Latitude A to 500,000 at Latitude F (Fig. 2, right column).
We found a similar trend in the abundance of the Pulicaris ensemble vectors from south to north: the weekly mean abundance increased gradually from less than 5 in Latitude A to a peak weekly average of 100 in Latitude D (Fig. 3). However, in contrast to the Obsoletus ensemble, the Pulicaris ensemble abundance did not continue to increase beyond Latitude range D. In Latitude F, there was a marked variation in abundance across weeks (varying from low abundance one week to high peaks the next week). The cumulative sum of the weekly mean of Pulicaris ensemble abundance showed a different pattern compared to the Obsoletus ensemble, with the mean accumulated number of Pulicaris ensemble vectors peaking at mid-range latitudes (Latitude D) and reaching only 41,000 vectors per year (Fig. 3, right column). This value decreased to 15,000 vectors per year further north and south at Latitudes F and C, and to 1600 at Latitude A.

Abundance and interpolation maps
We observed a large variation in monthly abundance among farms in the same region for all two ensembles and C. imicola. However, spatially interpolation the mean monthly abundance at each farm revealed regions with a higher abundance and showed systematic variation within each latitude zone (Figs. 5 and 6). The regions with the highest monthly interpolated abundance of the Obsoletus ensemble were found in France (particularly in the north-west), Germany and southern Scandinavia (especially southern Norway). During the summer period, the monthly interpolated daily abundance was often high (1000-10,000) and occasionally very high (> 10,000) in these countries (Fig. 5). Every month, the interpolated abundance of the Pulicaris ensemble was of a lower magnitude than for the Obsoletus ensemble. During the summer months, farms with low (10-100) and medium (100-1000) abundance were often found distributed throughout the continent (with the exception of Spain), while the regions with the highest abundance were observed in Poland, Germany and Denmark with occasionally high-abundance (1000-10,000) farms (Fig. 6). The distribution in the high abundance regions were therefore found to be in more easterly areas for the Pulicaris ensemble compared to the Obsoletus ensemble.
The geographical areas where the Pulicaris ensemble and the Obsoletus ensemble showed the highest interpolated abundance during the summer months were generally also the areas where the species groups were observed earliest in the spring and latest in the autumn: western France for the Obsoletus ensemble, and Poland and Germany for the Pulicaris ensemble. In regions of low abundance, midges were first observed later and last observed earlier in the year (Figs. 5 and 6).
The highest abundance of C. imicola was found in Corsica, where a farm with extremely high abundance was found (> 10,000). In general, Spain had a medium abundance (100-1000), but high-abundance farms could occasionally be found. However, the abundance did not reach the levels seen for the Obsoletus ensemble in northern Europe (Fig. 7).

Start of vector season by NUTS 3 polygon level
We defined the start of the vector season for each NUTS 3 polygon as the first month with a mean abundance higher than or equal to one specimen for C. imicola, and higher than or equal to five specimens for the Obsoletus and Pulicaris ensembles.
According to this definition, the start of the season for the Obsoletus ensemble occurred as early as January in the west of France, some parts of Spain and Germany (Fig. 8) and as late as June in Scandinavia. In France, there was a clear spatial pattern where the Obsoletus ensemble season started early (January) in the west, and 2 to 3 months later in the east (March-April). In some provinces in Spain, the season started late (April-June).
The start of the vector season for the Pulicaris ensemble showed a spatial pattern similar to the Obsoletus ensemble, with a south-to-north gradient, where southern latitudes had an earlier start of the season (January to April in Spain and France) compared to northern latitudes (May to September in the Scandinavian countries). The start of the season for the Pulicaris ensemble occurred as early as January in some parts of Spain, Germany and Corsica, but generally occurred from March-April, 2 months later than for the Obsoletus ensemble (Fig. 9). In France, we observed the same pattern   Culicoides imicola was only recorded in Spain and France. The vector season started as early as January in southern Spain and on Corsica, while in the northern provinces of Spain, it started 5 months later (Fig. 10). However, there were also two provinces in the south of Spain where the average abundance per polygon did not reach the threshold value of one until June-August.

Discussion
The descriptive analysis presented here is based on the most extensive Culicoides dataset created for Europe to date, and represents the first combined description of Culicoides abundance and distribution for a large part of Europe. The data were gathered from a 4000 km long transect from southern Spain to the Arctic Circle in Sweden, with the most easterly collection sites in Poland. The primary aim of this descriptive analysis was to identify and quantify major geographical patterns and seasonal trends in the abundance of key Culicoides vector groups. The focus of the analysis was to identify patterns and trends important for decision making to prevent, surveillance and control of Culicoides-borne pathogens.
Specimens of the Obsoletus and Pulicaris ensemble were found in all of the countries sampled. The Obsoletus ensemble was ten times more abundant than the Pulicaris ensemble. Both groups have a Palaearctic distribution, and are widely distributed in Europe [13,[44][45][46][47]. However, the abundance of both ensembles and C. imicola varied dramatically among farms in the same region sampled during the same period, often showing a 100-fold difference between the 10th and the 90th percentile in terms of the weekly trap abundance within each latitude group. Nevertheless, distinct spatial and temporal patterns arose under the three analyses conducted in this work.
Examining the weekly data for the seven latitude ranges, we found that the mean weekly abundance of both ensembles and C. imicola varied dramatically along the south-north transect. It is interesting to note that the annual number of the Obsoletus ensemble gradually increased toward northern latitudes, despite the vector season gradually shortening by three months from southern Spain to mid Scandinavia. This suggests that the Obsoletus ensemble is better adapted to the northern European climate in the environments surrounding farms. The Pulicaris ensemble appear better adapted to central Europe, and C. imicola to southern Europe, in the dry Mediterranean regions characterized by hot summers [47]. Although in relative terms, the Pulicaris ensemble (C. pulicaris and C. punctatus) was more abundant in central Europe and the Obsoletus ensemble (composed of C. obsoletus, C. scoticus, C. montanus and C. chiopterus.) was more abundant in northern Europe, the Obsoletus ensemble was still more abundant than the Pulicaris ensemble at all latitudes.
To further explore the spatial abundance patterns, we interpolated the mean monthly abundance for each calendar month. We found areas of high abundance peaks for the Obsoletus ensemble in north-western France, which is in agreement with previous studies that also reported high numbers of Obsoletus ensemble specimens in France [9,22,23]. The Obsoletus ensemble was also found in high abundance in most parts of Germany, where similar findings have been reported by many other authors [14,19,24,47,48]. Regarding the highabundance areas found in Scandinavia in this study, these were partly driven by farms with an extremely high abundance; for example, two farms in southern Norway had more than 80,000 specimens per night (data not shown). Although the monthly mean data were log 10 transformed before interpolation in order to reduce the impact of single sites of very high abundance, these high abundance farms from Norway still influenced the observed regional pattern. These high abundance records exceeded in great magnitude, the C. obsoletus/C. scoticus abundance previously reported in Sweden by Ander et al. [33] (> 5000 specimens in suitable months). Further collections in southern Norway are needed to determine how often the Obsoletus ensemble occurs at these extremely high numbers. The traps with the highest abundance of the Pulicaris ensemble were located in Germany and Poland, with a lower abundance in France and Scandinavia [14,49,50]. In general, the Pulicaris ensemble showed a spatial pattern with a relatively more easterly distribution compared to the Obsoletus ensemble. The distribution of the regions with the highest C. imicola abundance were in accordance with the known C. imicola distribution in Europe [9,21,22,47]. Based on the interpolated monthly abundance maps, each vector group tended to appear early in the traps situated in areas that reached the highest abundance during summer, and were observed latest in the autumn. When the monthly maps are considered as a time series, the Obsoletus ensemble appears to start in western France, to increase in abundance and spread north and east until the end of August, when they retract to western France again. The Pulicaris ensemble appears to start in traps in Poland and, to a certain extent, the south-western areas of France, to grow in number and spread north and east before retracting again after August, to be observed last in Poland, northern Germany and western France. For C. imicola, the same phenomenon of areas with a strong spatial correlation between early appearance, peak abundance and a long vector season was seen in southern Corsica and southeastern Spain.
The Culicoides abundance analysed in this study is exclusively based on the abundance observed on farms.
The spatial abundance derived by interpolation therefore represents the abundance given the presence of a farm, and cannot be interpreted as an estimate of abundance in "non-farm" habitats, e.g. natural areas. Interpolation is used as tool to visualize the average abundance on farms in a larger area. To produce more detailed maps of vector abundance in future, a predictive modelling approach based on environmental and climate predictor variables will be necessary.
We analysed the start of the vector season for polygon areas in the participating countries. The European Commission currently defines the start of the vector season as the week during which the number of parous females exceeding a certain threshold (five for C. obsoletus and one for C. imicola) are caught by any trap in an area [20]. However, because there is a large variation in abundance among traps in the same area, the probability of finding a trap with an abundance exceeding the threshold will increase with the number of traps operated in the area. As a result, the vector season will tend to start earlier in regions with a higher number of traps. In this analysis, where the density of traps varies geographically, we used a more robust approach to define the start of the vector season. We took into account the number of traps in a geographical area by calculating the mean trap abundance and using five vectors per trap as the cutoff (one female per trap for C. imicola). For the Obsoletus ensemble, there was not a clear pattern in Spain, where the vector season started at different months in different polygons. Nevertheless, there was a south-north trend starting in southern France and continuing to northern Scandinavia. The length of the vector season decreased by three months at northern latitudes where the period of climatic conditions suitable for midge development is shorter. Versteirt et al. [51] investigated the start and end of the vector season for C. imicola and C. obsoletus/C. scoticus on a continental scale in Europe. Their results also showed a south-north pattern in the start and length of the vector season. However, for C. imicola, our observed data suggested that the start of the season would occur earlier in the year (January-February) compared to their results where the season started in March. The start of the Pulicaris ensemble season had a similar pattern to the Obsoletus ensemble, starting two months later at higher latitudes; however, the start of the season seems to be more homogenous across latitudes compared to Obsoletus ensemble. We found the start of the season pattern for C. imicola is similar to previous results [51], with the only difference that in our study the vector season started as early as January on Corsica and in some provinces of Spain, which we interpret as areas with C. imicola vector activity all year round. The start of the season occurred later in the year in polygons where the peak Culicoides abundance in summer was low, as the density gradually increased and the cutoff defining the start of the vector season was reached later than in areas with a high peak abundance. For example, the Pulicaris ensemble activity started as late as September in northern Sweden, despite the first individual being observed much earlier. The start of the vector season is therefore highly sensitive to the vector abundance threshold selected to define it.
The latitude range analysis showed that the vector season for both the Obsoletus and Pulicaris ensembles started in the southern latitude range at an average mean temperature of 10°C and 12°C, respectively. However, at northern latitudes, the season started before spring temperatures rose, meaning that the vector season started at mean temperatures of just 1°C and 3°C. This suggests an adaptation of the vector population to the cooler climates in northern Europe. Early EU regulations suggested that vector-free periods may be defined by a specific temperature threshold when vector surveillance data were lacking [46]. Yet the start of the vector season at successively lower temperatures toward northern latitudes found in the present study demonstrates that such a simple temperature criterion alone would be a poor proxy for the start of the vector season, and would risk the prediction of an unrealistically long vector-free season in northern Europe.
One country used CDC traps (Spain) and one used BG-sentinel traps (Germany), while the rest used Onderstepoort traps. Previous studies compared the efficacy of different trap types, and the results showed that the number of midges collected varied according to the type used. It was reported that Onderstepoort traps collected more specimens than Mini CDC and BG-sentinel traps [35][36][37]. We attempted to adjust for this using published trap comparisons, but the relative efficiency of each type trap vary considerably. However, this uncertainty in trap efficiency is likely to be of a smaller magnitude than the variation in abundance within the spatial patterns which is of at least a 10 to 100 fold magnitude. The used trap conversion factor is therefore unlikely to have affected the identified overall spatial patterns identified here.
Light traps are most efficient when collection nights are dark [10,52] and trapping in northern latitudes may therefore be less effective because nights during the summer season are shorter than in southern Europe [53]. The use of light traps in this study may therefore have underestimated the vector abundance at higher latitudes during the summer period.
We here focus on estimating the host seeking vector population. The vectors collected in light traps near stables and animal resting sites are likely to be predominantly host-seeking, while vectors that are already bloodfed are less likely to be collected in traps. The results from this analysis cannot directly be used for estimating the total vector population consisting of both host seeking and non-host seeking vectors. Because the blood meal digestion time in Culicoides is relatively slower at low temperatures [54], the proportion of the Culicoides population digesting blood and developing eggs and not attracted to traps may therefore be relatively larger at low temperatures. Due to the increasingly lower temperatures toward the north of the transect, the total vector population estimated from trap collections at higher latitudes, will be underestimated compared to the total vector population estimated from trap collections in southern Europe unless the blood meal developing time is taken into account.
Females of the species belonging to the Obsoletus group are difficult to separate based on morphological characters [13,18,[38][39][40][41]55] and therefore they are often grouped into the Obsoletus group or complex. The same occurs with the species of the Pulicaris group which are often merged into the Pulicaris group [25,31,56,57]. Aggregating species into groups might represent a problem for identifying accurate temporal and spatial patterns, as different species from the same group might exhibit different seasonal trends [58,59]. The seasonal fluctuation of individual species of the Obsoletus group remains unknown. However, Searle et al. [27] analysed the phenology (start, end and duration of the vector season) of the male specimens for each species and the authors did not find a significant difference between the start and the end of season among the species. Nevertheless they found the length of the season period was different among them. Analyses regarding temporal fluctuation at species level of the Obsoletus group would be necessary as the species might present different temporal fluctuation patterns undetected when the species are grouped. Little is known about potential differences in vector competence for BTV for individual species of the Obsoletus and Pulicaris ensemble [11]. A study on the subject includes Carpenter et al. [60] who analysed vector competence for the Obsoletus group in the UK. Vector competence at species level of the Obsoletus group for SBV can be found in [60], Balenghien et al. [61] and Ségard et al. [62]. In this work, despite analyzing the data at ensemble level, we consider that the identified patterns and trends identified here still represent a useful and relevant overview of transmission potential in Europe.

Conclusions
This is the first report in which a dataset this size and covering a large part of Europe has been analysed. We identified and quantified the main mean spatial and temporal differences of three Culicoides species groups. Understanding the spatial and seasonal patterns of key vector groups or species facilitates the planning of preventive strategies and allows the development of more cost-effective vector and disease surveillance programmes by veterinary authorities in the European Union. The monthly abundance of the Obsoletus ensemble increased gradually from northern Spain to mid Scandinavia. The vector season also became increasingly shorter toward the north, starting three months later in mid Scandinavia compared to southern Spain. Nevertheless, the annual accumulated abundance of the Obsoletus ensemble increased steadily with latitude to 500,000 vectors per trap per year in mid Scandinavia. The Pulicaris ensemble was more frequent in central Europe, peaking in Germany and Poland with about 40,000 vectors per year, and with a more easterly distribution compared to the Obsoletus ensemble. For each of the species groups, there were areas in which the vectors appeared early, reached the highest mean peak abundances and lasted the longest. The Obsoletus ensemble was more abundant and had a longer season than the Pulicaris ensemble, whereas C. imicola appeared as a strictly southern species with a long vector season but with an abundance level that did not reach the peak abundance observed for the Obsoletus ensemble. This study suggests that future collaboration and data sharing between European countries may further improve our understanding of the spatio-temporal abundance of Culicoides vectors.