Culicoides (Diptera: Ceratopogonidae) Abundance Is Influenced by Livestock Host Species and Distance to Hosts at the Micro Landscape Scale

Simple Summary Culicoides abundance is a critical component for transmission modelling of vector-borne diseases. It can be anticipated that Culicoides abundance is not homogeneously distributed in the landscape. Modelling parameter estimates are mostly extrapolated from indirect measurements using the most commonly used method worldwide: the light-trap (LT). However, it would be difficult to extrapolate Culicoides LT catches for modelling purposes due to the inherent biases of LT catches. The use of a sweep-net and performing collections during the hours before nightfall, when the LT is ineffective, can circumvent this bias. The aim of our study was to investigate the influence of baits (dairy cow, sheep, and LT) and the distance of measurement to hosts on Culicoides abundance using sweep-net catches. Our results indicate that cows are a far stronger attractor of Culicoides midges than sheep, and that both livestock hosts are far stronger attractors of Culicoides midges than LTs. Culicoides abundance declined significantly with increasing distance from hosts; this phenomenon was much stronger for dairy cows than for ewes. In contrast, Culicoides abundance increased with increasing distance from the LT, pin-pointing the apparent shortcomings of the LT as a surrogate host to lure midges. Abstract The vector/host ratio and host preference are important parameters for the modelling of vector-borne livestock diseases. It can be anticipated that Culicoides abundance is not homogeneously distributed in the landscape. We investigated the influence of host species (dairy cow, sheep, and a light-trap (LT) as a surrogate host) and distance of measurement to hosts on Culicoides abundance using a randomized block-design with 12 measuring days and seven 3-min aerial sweep-netting sessions per whole hour at three distances to the host (0, 10, and 25 m), from five hours before to and including one hour after sunset. Dairy cows were found to be a far stronger attractor of Culicoides midges than sheep, while both hosts were far stronger attractors of midges than the LT. Culicoides abundance declined significantly with increasing distance from the livestock hosts; this phenomenon was much stronger for dairy cows than for ewes. In contrast, Culicoides abundance increased with increasing distance from the LT, pin-pointing the apparent shortcomings of the LT as a surrogate host to lure midges. Our data indicate that livestock host species and the distance from these hosts have a profound effect on Culicoides abundance in the landscape.


Introduction
The vector/host ratio and host preference are important parameters for the modelling of vector-borne livestock diseases, but there are only sparse field data to use for estimates with which to parameterize vector-host disease transmission models [1][2][3]. Furthermore, parameter estimates are mostly extrapolated from indirect measurements using the most commonly used method worldwide-the black-light suction trap (LT)-for catching midges [4]. By experience, it was established that the presence of livestock in the The study was conducted at a dairy farm in the municipality of Ermelo, eastern Netherlands, in the May-June period (geographical location: 52 • 18 N, 5 • 43 E). The pasture used for the study measured 2.28 ha (approximately 120 m × 190 m). Three sites on a pasture, forming a rough triangle with about equal distances between the sites, were selected for the experiment and were situated approximately 150-200 m from the milking shed where 65 dairy cows were housed each night. Additionally, two Texelaar ewes were kept on a small pasture situated 50 m from the milking shed. The three experimental sites were separated from each other by a distance of approximately 80 m to avoid any one site exerting influence over any other during the trapping sessions. At each site, the following baits were used as attractants of Culicoides midges: a seven year-old dairy cow (100% Holstein-Friesian, black and white; approximate weight: 750 kg), a two year-old ewe (Texelaar sheep, white, shorn two weeks before the study started; approximate weight: 35 kg) and an Onderstepoort Veterinary Institute (OVI) black-light suction trap (LT) with an 8 W black light tube [5]. None of the animals were treated with insecticides before or during the experiment.

Study Design
The peak of Culicoides midge activity occurs around sunset [8,15]. Hence, approximately 6 h before sunset (SS − 6 h), the baits were installed at their designated positions in the pastures. The LT was installed on a tripod, 2 m above ground level. Each of the two animal hosts and the LT were randomly assigned to a site. In total, 12 replicate measurements (days) were made. The number of replicates used in our study is comparable with those of other Culicoides field experiments [8,[16][17][18][19] aimed at testing biologically relevant differences. Daily observations were made only when the following weather conditions applied: windless, no rain, and temperatures > 10 • C; all collection days were bright, at most only partly cloudy. Sunset (SS) occurred at around 22:00 h, and times varied by only a few minutes throughout the study (Central European Summer Time). In each hour from SS − 5 h to and including SS + 1 h, Culicoides midges were sweep-netted at three different distances from the baits: 0 m, 10 m, and 25 m. An overview of the study design is shown in Figure 1.
an Onderstepoort Veterinary Institute (OVI) black-light suction trap (LT) with a black light tube [5]. None of the animals were treated with insecticides before or the experiment.

Study Design
The peak of Culicoides midge activity occurs around sunset [8,15]. Hence, a mately 6 h before sunset (SS -6 h), the baits were installed at their designated posi the pastures. The LT was installed on a tripod, 2 m above ground level. Each of t animal hosts and the LT were randomly assigned to a site. In total, 12 replicate m ments (days) were made. The number of replicates used in our study is comparab those of other Culicoides field experiments [8,[16][17][18][19] aimed at testing biologically r differences. Daily observations were made only when the following weather con applied: windless, no rain, and temperatures > 10 °C; all collection days were br most only partly cloudy. Sunset (SS) occurred at around 22:00 h, and times varied b a few minutes throughout the study (Central European Summer Time). In each hou SS -5 h to and including SS + 1 h, Culicoides midges were sweep-netted at three d distances from the baits: 0 m, 10 m, and 25 m. An overview of the study design is in Figure 1. For this, circles of 10 m and 25 m distance to the baits were constructed by sett circles with white-colored wooden pegs in the pasture ground. For each whole h an experimental day, sweep-netting was randomly assigned to a distance; there down-time period of 5 min between the end of a sweep-net session at one distan the start of sweep-netting at another distance. Sweeping involved the use of a co cially available lightweight gauze insect butterfly sweep-net (diameter 46 cm; aer was made of closely woven off-white Dacron chiffon with mesh size: 70 × 70 pe insect net #7318WA with an alloy handle 2 m in length, BioQuip Products Inc. (Co CA, USA). Aerial sweep-netting was standardized with figure-8 sweeps at a calm ing pace (360 degrees; going from front to behind the sweeper) for 3 min while the tor was walking continuously circling the animals, which were tethered during the tion period, or the LT at the predetermined distance, as set out with wooden pe catching results around the animal hosts with the sweep-net consisted of a comb of midges flying toward the host in order to take a blood meal (attraction), midge away from the host after having taken a blood meal, and those that were not intere the host. However, the sweep-net result at 0 m of the LT merely provided a represe of midges flying toward the LT, because the LT traps the midges that come close to For this, circles of 10 m and 25 m distance to the baits were constructed by setting the circles with white-colored wooden pegs in the pasture ground. For each whole hour on an experimental day, sweep-netting was randomly assigned to a distance; there was a down-time period of 5 min between the end of a sweep-net session at one distance and the start of sweep-netting at another distance. Sweeping involved the use of a commercially available lightweight gauze insect butterfly sweep-net (diameter 46 cm; aerial bag was made of closely woven off-white Dacron chiffon with mesh size: 70 × 70 per inch; insect net #7318WA with an alloy handle 2 m in length, BioQuip Products Inc. (Compton, CA, USA). Aerial sweep-netting was standardized with figure-8 sweeps at a calm sweeping pace (360 degrees; going from front to behind the sweeper) for 3 min while the collector was walking continuously circling the animals, which were tethered during the collection period, or the LT at the predetermined distance, as set out with wooden pegs. The catching results around the animal hosts with the sweep-net consisted of a combination of midges flying toward the host in order to take a blood meal (attraction), midges flying away from the host after having taken a blood meal, and those that were not interested in the host. However, the sweep-net result at 0 m of the LT merely provided a representation of midges flying toward the LT, because the LT traps the midges that come close to it (i.e., they are forced by a ventilator into a beaker containing soap hanging under the device). Therefore, for a fair comparison, it was deemed reasonable to add 5% of the LT catch to the values associated with the sweep-net at 0 m around the LT. Sweep-netting was done for 3 min each whole hour, while the LT operates for 60 min each hour; with this in mind, we determined that 5% (=3 out of 60 min) of the trapping result of the LT trap should be taken into account.

Culicoides Identification
Culicoides midges were stored in 70% ethanol until they were counted. Culicoides midges were identified morphologically primarily using the works of Campbell and Pelham-Clinton [20], Glukhova [21], and Delécolle [22]. All Culicoides midges collected in the study were counted, and no subsampling was done. The C. obsoletus complex comprised two species: C. obsoletus and C. scoticus; C. chiopterus and C. dewulfi are not deemed to form part of the C. obsoletus complex and so are treated separately [23]. Female midges were age-graded as either nulliparous, pigmented parous, gravid, or freshly blood-fed (engorged) [24].

Statistical Analysis
The total numbers of Culicoides or blood-fed Culicoides per observation day collected by sweep-net during the sampling period were estimated by calculating the area under the curve (AUC), which is a function of the number of Culicoides sweep-netted at each specific time of sampling and the number of times sampling was done per day. Since we were dealing with counts, AUC values were rounded up to an integer value for further analysis.

Relationship between Baits, Number of Culicoides Sweep-Netted, and Distance
A GLMM with a Poisson distribution was fitted to explain both the daily numbers of Culicoides sweep-netted and the rate of blood-fed Culicoides sweep-netted within this daily sampling period [25]. The explanatory (fixed) variables assessed were the type of bait used to attract the midges, the distance in meters from these baits, and the interaction between bait and distance. Distance was modelled as a linear variable or by including natural cubic splines with a knot at 10 m. When modelling the rate of blood-fed Culicoides, the daily number of Culicoides caught was introduced as an offset variable.
Random effects were included to account for variations introduced by the experimental procedures. The assessed random effects were: (a) Day of sampling, which accounted for the influence of environmental variation (temperature and humidity) between sampling days and repeated measurements taken on the same hosts. Temperature and humidity were also assessed, but their influence was intrinsically explained by using day of sampling, leading to better model fits. (b) Sampler, i.e., the person performing the sweep-netting at different sampling locations (bait, distance); this accounts for variations introduced by different people performing the catches. (c) The observation effect, which accounts for daily varying overdispersion between individual Culicoides catches. This random effect was included only if the model needed correction for overdispersion. (d) Location, which assessed variation introduced by the samplings locations (which was evenly allocated to all hosts).

Relationship between the Rate of Blood-Fed Culicoides Species and Distance
To model this relationship, GLMM Poisson models were fitted for each of the different baits. In this model, the number of blood-fed Culicoides was the response variable, the daily number of Culicoides caught was introduced as an offset variable, and Culicoides species, distance (linear or with natural cubic splines with a knot at 10 m), and the interaction between Culicoides species and distance were the explanatory variables. All of the random effects explained above were also assessed for these models.

Distribution of Culicoides Midges
The total numbers of female Culicoides midges collected at a distance of 0 m were as follows: 7695 (cow), 533 (sheep), and 145 (LT) ( Table 1). With increasing distance from the cow, the number of Culicoides midges caught decreased sharply. In contrast, the number of Culicoides midges caught at a distance of 10 m from the sheep was comparable to that at 0 m, but at a distance of 25 m, the number of Culicoides midges was clearly lower compared to the numbers caught at 0 and 10 m. The situation for the LT was quite different: increasing numbers of Culicoides midges were caught with an increasing distance to the LT. The total number of male Culicoides midges collected in the study for the major species were as follows: 3 (C. obsoletus complex), 1 (C. chiopterus), 0 (C. dewulfi), 0 (C. punctatus), and 0 (C. pulicaris) for the cow; 0 (C. obsoletus complex), 0 (C. chiopterus), 0 (C. dewulfi), 0 (C. punctatus), and 0 (C. pulicaris) for the sheep; and 3 (C. obsoletus complex), 2 (C. chiopterus), 34 (C. dewulfi), 7 (C. punctatus), and 13 (C. pulicaris) for the LT. Culicoides species richness at different distances from the cow was comparable: 15, 14, and 14 different species at 0, 10, and 25 m distance, respectively. In contrast, species richness at 0 m distance from the sheep was half of that at 10 and 25 m distance: 7, 15, and 15 different species at 0, 10, and 25 m distance, respectively. Remarkably, Culicoides species richness declined with increasing distances to the LT: 17, 12 and 11 different species at 0, 10 and 25 m distance, respectively. At 0 m distance, the proportion of blood-fed Culicoides midges was 7.6%, 6.6%, and 3.2% for the cow, sheep, and LT, respectively. The top-four most abundant Culicoides species-C. obsoletus, C. dewulfi, C. chiopterus, and C. punctatus (comprising 90-95% of all the Culicoides midges collected)-were detected equally in the environment of all the baits.

Culicoides Abundance and Bait and Distance Relationship
Based on GLMM modelling, the relationship between the mean number of Culicoides midges by bait and distance to bait is visualized in Figure 2. The random effects that remained in the models were the daily sources of variation due to the influence of weather (temperature and humidity) and the daily repeated measurements on the same baits, in addition to the daily varying overdispersion (observation effects) on the Culicoides distribution (Table S1). In the presence of these two random effects, the contribution of the sampler to variations was negligible (i.e., close to zero). The location of the baits made no contribution to variations in the model and was found to be insignificant when assessed as a fixed effect.

Culicoides Abundance and Bait and Distance Relationship
Based on GLMM modelling, the relationship between the mean number of Culico midges by bait and distance to bait is visualized in Figure 2. The random effects that mained in the models were the daily sources of variation due to the influence of weat (temperature and humidity) and the daily repeated measurements on the same baits addition to the daily varying overdispersion (observation effects) on the Culicoides dis bution (Table S1). In the presence of these two random effects, the contribution of the sa pler to variations was negligible (i.e., close to zero). The location of the baits made no c tribution to variations in the model and was found to be insignificant when assessed fixed effect.

Distribution of Culicoides Midges by Sampling Hours
The mean number of Culicoides midges caught at five hours before sunset was alm nil; however, this value steadily increased to a peak at around one hour before sunset then dropped to almost zero again at one hour after sunset for all baits and at all the tances investigated (Figure 3), covering a complete-as-possible measurement window the evening. The number of blood-fed Culicoides midges collected by sweep-net stead increased from almost zero at five hours before sunset to a peak at sunset before dropp

Distribution of Culicoides Midges by Sampling Hours
The mean number of Culicoides midges caught at five hours before sunset was almost nil; however, this value steadily increased to a peak at around one hour before sunset and then dropped to almost zero again at one hour after sunset for all baits and at all the distances investigated (Figure 3), covering a complete-as-possible measurement window in the evening. The number of blood-fed Culicoides midges collected by sweep-net steadily increased from almost zero at five hours before sunset to a peak at sunset before dropping to almost nothing again at one hour after sunset for the cow and the sheep and at all the distances investigated (Figure 4). The number of blood-fed Culicoides midges collected by sweep-net in the environment around the LT was negligible. to almost nothing again at one hour after sunset for the cow and the sheep and at all the distances investigated (Figure 4). The number of blood-fed Culicoides midges collected by sweep-net in the environment around the LT was negligible.

Distribution of Culicoides Midges by Distance to Bait
The mean number of Culicoides midges sweep-netted at a distance of 10 m and 25 m distance from the cow was significantly (p < 0.001) lower, i.e., 6.5 times (95% confidence interval (CI): 4.3-9.7) and 14.7 times (95% CI: 9.8-22.2), compared to the mean number of Culicoides midges sweep-netted at a distance of 0 m, respectively ( Figure 5). The mean

Distribution of Culicoides Midges by Distance to Bait
The mean number of Culicoides midges sweep-netted at a distance of 10 m and 25 m distance from the cow was significantly (p < 0.001) lower, i.e., 6.5 times (95% confidence interval (CI): 4.3-9.7) and 14.7 times (95% CI: 9.8-22.2), compared to the mean number of Culicoides midges sweep-netted at a distance of 0 m, respectively ( Figure 5). The mean number of Culicoides midges sweep-netted at a distance of 10 m from the sheep, however, was not significantly lower (Odds ratio = 1.02, 95% CI: 0.7-1.6) compared to the mean number of Culicoides midges sweep-netted at a distance of 0 m; the mean number of Culicoides midges sweep-netted at a distance of 25 m from the sheep was significantly lower (p = 0.005), i.e., 1.9 times (95% CI: 1.

Distribution of Culicoides Midges by Distance between Baits
The mean number of Culicoides midges sweep-netted at distances of 0 m, 10 m, and 25 m from a cow was 13.2 times (95% CI: 8.8-19.8), 2.1 times (95% CI: 1.4-3.2), and 1.7 times (95% CI: 1.1-2.6) higher than that measured at the same distances from a sheep. The mean number of Culicoides midges sweep-netted at a distance of 0 m and 10 m from a cow was 57.5 times (95% CI: 36.9-89.7) and 4.1 times (95% CI: 2.7-6.2) higher than that around a LT, respectively. At a distance of 25 m from a cow and the LT, the mean number of Culicoides midges sweep-netted was not significantly different (OR = 1.2, 95% CI: 0.8-1.8).

Rate of Blood-Fed Culicoides Species and Distance
Blood-fed rates were highest at 0 m distance and declined significantly (p < 0.001)

Distribution of Culicoides Midges by Distance between Baits
The mean number of Culicoides midges sweep-netted at distances of 0 m, 10 m, and 25 m from a cow was 13.2 times (95% CI: 8.8-19.8), 2.1 times (95% CI: 1.4-3.2), and 1.7 times (95% CI: 1.1-2.6) higher than that measured at the same distances from a sheep. The mean number of Culicoides midges sweep-netted at a distance of 0 m and 10 m from a cow was 57.5 times (95% CI: 36.9-89.7) and 4.1 times (95% CI: 2.7-6.2) higher than that around a LT, Insects 2023, 14, 637 9 of 12 respectively. At a distance of 25 m from a cow and the LT, the mean number of Culicoides midges sweep-netted was not significantly different (OR = 1.2, 95% CI: 0.8-1.8).

Rate of Blood-Fed Culicoides Species and Distance
Blood-fed rates were highest at 0 m distance and declined significantly (p < 0.001) with increasing distance. No significant differences in the blood-fed rates of Culicoides species were seen between measurements around different baits. In addition, there were no significant differences in blood-fed rates among Culicoides species. No significant interactions were observed between distance and Culicoides species with respect to bloodfed rate. The random effects that remained in the model were the daily sources of variation and the sampler (Table S2).

Discussion
In a randomized trial conducted over 12 days, we quantified-on a micro landscape level-the distance dependency of Culicoides midge abundance in a pasture with different baits. An obvious conclusion is that Culicoides midge abundance under field conditions is not homogeneously distributed in the micro landscape. The non-homogenous distribution stems from the need of the midges for a blood meal, leading them to track and aggregate around warm-blooded hosts, prior to landing and feeding on them. In our study, Culicoides midge abundance clearly decreased as distance to livestock hosts increased, i.e., a sharp decrease with dairy cows and a moderate decrease with sheep. This pattern clearly reflects the attraction of Culicoides midges to these livestock hosts. Another evident inference is that a dairy cow is a far stronger attractor of Culicoides midges than a sheep, which supports the findings of recent field studies [8,18,26], in which Culicoides midges were collected by direct aspiration from the skin of the livestock hosts. This seems logical, considering the difference in size of the animals and the correlating difference in the quantity of exhaled volatiles that act as attractants. Using Kleiber's allometric scaling (i.e., weight to the power of 0.75) [17] to calibrate for a theoretical equal basal metabolism (Kleiber's allometric: 143.3 (cow), 14.3 (sheep)), there was about a 10-times difference between the cow and the sheep, which approximately reflects the observed difference in the total number of Culicoides midges collected at 0 m distance from both livestock hosts. The observation that comparable numbers of Culicoides midges were sweep-netted at 0 and 10 m distance from a sheep may indicate that the level of attraction of Culicoides midges to a sheep is quite modest. In addition, the fact that the numbers Culicoides midges sweep-netted at 25 m distance from the cow, sheep, and LT were virtually identical is a good indication that the method used and the design of this study were valid and that at this distance, there is virtually no level of attraction by the baits.
The reliance upon LT data as a vector-to-host ratio proxy is based on the assumption that the number and species of midges caught in a LT equates to that lured by the animal host [27]. However, it has become evident that the LT performs far more poorly than previously supposed [8,10,28]. Furthermore, it seems that the midges aggregating around the LT are disoriented by the UV-light and not really lured by it [10]. In contrast to the Culicoides species collected in the vicinity of livestock hosts, ornothophilic Culicoides species like C. festivipennis, C. kibunensis, C. pictipennis, and C. circumscriptum [29,30] were collected predominantly by the LT.
Manual sweep-netting is seldom used as a method to collect Culicoides midges. We know of several studies that used a sweep-net mounted on a truck to collect Culicoides midges [31][32][33][34]. In the Netherlands, sweep-netting in early summer revealed Culicoides midges to be active at least a few hours before nightfall, when the LT is not yet effective in collecting midges [8]. A recent host preference experiment with a dairy cow, a Shetland pony, and a sheep, using direct aspiration to collect midges from the skin of the animal hosts, revealed C. obsoletus complex to be largely nocturnal, while C. chiopterus and C. dewulfi were found to be predominantly diurnal [9]. The sweep-net results in this study mirror those obtained using direct aspiration during the same hours of collection in the evening. Since there is good correlation between direct aspiration (pootering) and sweep-netting results [9], and given that sweep-netting is much easier and simpler to execute than pootering, the use of a sweep-net for collecting information pertaining to the vector/host ratio and host preference may be revived.
To estimate the blood-feeding rate, direct-aspiration (pootering) from the skin of hosts is preferable to sweep-netting, since pootering is the most direct measuring method to get an impression of the blood-feeding activity for a specific host. The proportion of blood-fed midges in our study using sweep-netting was approximately 5-10% for the cow and the sheep, while in earlier, comparable studies that used pootering to collect Culicoides midges, the proportion of blood-fed midges was around 20-30% [8,26,35,36].
Little is known about the fine-scale distribution of Culicoides midges in the landscape. Predominantly, LTs are used for Culicoides surveillance; LT-derived data are widely available and are therefore used to model vector distributions and the transmission of vector-borne diseases [4,37,38]. Working with these data and models has led to the identification of major knowledge gaps, in particular in terms of the translation of LT catches to midge densities [27]. Rigot et al. [14] indicated that a Leptokurtic model-with Culicoides abundance as a negative logarithmic function of the distance from a farm-fitted their data best. Our data mirror this phenomenon with respect to the distance dependency, but on an even finer scale, indicating that such a relationship differs depending on the host species and the distance from hosts, both of which have a profound effect on the actual Culicoides abundance in the landscape.