Ectoparasites causes harm to animals and humans and may spread disease. With ongoing climate change, affecting the life cycles and distribution ranges of many parasites , and the increase in population density of many large ungulates in Europe , there is an urgent need to understand what causes variation in prevalence and infestation intensity. Our results revealed a deer ked infestation prevalence of 100%, similar to that Paakkonen et al.  report from Finnish moose. Despite different histories of colonization  and a higher density of definitive hosts in Norway compared to Finland , the deer ked is apparently an ectoparasite well-adapted to Fennoscandian moose and seems able to exploit the entire moose population within its distribution range. The generality of our results are further supported by the 100% deer ked infestation prevalence reported from Polish moose examined in 1988 .
Ked infestation intensity in relation to moose age and gender
Although all moose were infested with keds, a wide range in deer ked intensity (0.004 – 1.405 keds/cm2) was observed, with the highest intensity in male yearlings (Figure 2). Our study does not reveal the mechanism behind this wide range, but differences between age and sex classes may be explained by physical and behavioral differences. Firstly, Kadulski  and Kortet et al. argue that swarming keds prefer large body size when choosing hosts. As calves follow their mother during their first autumn this may imply that cows will attract higher numbers than their accompanying calves. Hence, calves at foot may partly be protected against deer ked attacks. However, body size alone cannot explain the pattern, since yearlings had higher infestation than adults. Secondly, moose calves are born with a reddish coat and their moulting period is July to September, in contrast to adults moulting between April and July . Hence, deer ked swarming and calves moulting coincides, which most likely hamper deer ked establishment in the calves` coat. Kortet et al.  did not find any difference in ked preferences for black and red, mimicking cow and calf colors, stating that coat color is not important for ked host selection. Thirdly, locally acquired resistance to sheep keds (Melophagus ovinus) was demonstrated in an experiment with lambs (Ovis aries) . Resistance was gradually lost over the following weeks after infestation, but the experiment demonstrated that repeated infestation in the same test area reduced time to onset of resistance . This suggests that adult moose exposed to several swarming seasons may develop resistance more quickly than yearlings, resulting in higher ked intensity in the latter age class, fitting our results. Fourthly, calves are less active than yearlings, cows without calves and especially bulls, during autumn. Kortet  argues that movement is the main cue in ked host selection and therefore calves might be less exposed to winged keds sitting in the vegetation, waiting to flying onto any moving object passing by. In a closely related study in Finland, Paakkonen et al. stated that bulls had about three times the intensity of keds compared to cows. Our model also predicted significantly higher ked intensity in bulls compared to cows, but the sex difference was less pronounced (Figure 2). Heavier infestation in bulls in autumn is consistent with expectations from the life history theory as a tradeoff may exist between resources required for parasite resistance and reproduction. Accordingly, parasite levels are found to increase in males during the rutting season .
During close physical contact between cow and calf, direct transfer of wingless keds may also affect the infestation rate. Small , citing Tetley 1958, states that sheep keds migrate to the surface of the fleece in response to increased ambient temperature. Deer keds displayed similar migrating behavior from skin surface to the tip of hair when moose pelts were brought from subzero temperature into a heated room (20°C) (unpublished, Madslien), indicating that transfer of deer keds can occur between moose in close physical contact. Similarly, Davis et al.  and Samuel and Trainer  showed that Neotropical deer ked (Lipoptena mazamae) could be transferred from doe to fawn in white tailed deer (Odocoileus virginianus). However, we regard it likely that direct transfer is relatively small compared to direct settlement of winged keds.
Effect of habitat on infestation intensity
The preferred habitats for pupal development and survival, as well as winged imagines host acquisition, are discussed in the literature . In our study, we expected to find a positive correlation between preferred moose habitats during autumn and winter, i.e. areas where the deer ked pupae will be deposited during the reproductive season of this insect, and deer ked intensity. Our study indicates that moose living, or at least shot, in a habitat dominated by Scots pine, an important species for moose browsing during late autumn and winter, had the highest infestation loads, hence consistent with our first hypothesis. Haarløv et al. argued that puparia and winged imagines prefer woody areas, due to the soils loose structure, moderate moist and limited wind all year round. Haarløv et al. also found that red deer (Cervus elaphus) had higher deer ked intensity than fallow deer (Dama dama) and explained the difference with habitat preferences. Fallow deer prefer open grasslands, where pupae that drop from the coat will be exposed to extreme and possibly fatal conditions, whereas red deer favor more protected woody areas. In our model, bogs were negatively correlated with ked intensity, possibly explained by adverse conditions for pupal survival in humid substrates or because these areas are avoided by moose due to their low production of feed. Samuel et al. speculated that flooding in lowlands, resembling the condition in a bog, killed a high number of soil-dwelling Neotropical deer keds in Texas, which substantiates the notion that very damp soil has a negative impact on pupal development and survival. Lack of high vegetation as vantage points for winged keds searching for hosts and being exposed to heavy winds may also prevent bogs from being a suitable habitat for winged keds. On the other hand, Darling et al. explained high intensity of keds at wallows with the rubbing behavior of red deer during moulting in April and May, resulting in a large number of pupae released from the coat simultaneously. If this is the case in adult moose, which are moulting in early summer , preferred habitats during this period should have an increased deer ked intensity. In this part of the year, moose are typically feeding on emerging deciduous leaves in deciduous and young spruce forests . Hence, the lack of any association between infestation intensity and the coverage of spruce or deciduous forest lend little support to this hypothesis in our study.
Effect of moose population density on infestation intensity
Because moose is the preferred definitive host of deer keds in Norway, we hypothesized a strong, positive correlation between moose population density and deer ked intensity, as Balashov  reported from moose in northwestern Russia. Balashov  monitored human deer ked infestation intensities in three different geographical areas for 8 consecutive years and observed on average a 8 to 29 fold decrease in ked intensity from 1991 to 1995. During this period, a corresponding dramatic reduction in moose population density was claimed to be the cause of decreased ked attacks on human study objects walking in the forest. A similar tendency, although weaker, was detected in our model (Table 2), hence partly supporting our second hypothesis. However, the observation of an association weaker than expected might have been influenced by little variation in moose population density (average; 0.50 moose seen per hunter-day, SD ± 0.20, range; 0.15 to 1.21). Moreover, as density was measured by its proxy moose seen per hunting effort, there is also a chance that varying habitat composition and hunting methods may have affected the precision of this variable. Given our current results, however, it is unlikely that a small regulation of moose numbers by managers will cause large effects on deer ked infestation.
Although not included in our study, seasonal migration by moose can be expected to affect deer ked intensity in other parts of Scandinavia. Moose in the study area are mostly stationary , contrary to populations further north and west, where a varying part of the population migrate between summer areas at higher latitudes and low altitude winter areas. Because pupae are mainly shed in the winter areas and swarming occurs in the summer area, only a proportion of the population will be exposed to keds during the swarming season. Hence, high deer ked intensities in stationary moose could partly be attributed to an accumulation of keds in moose habitats utilised throughout the year and the fact that the entire moose population is exposed to keds during the swarming season.
Effect of latitude, longitude and altitude on infestation intensity
Our best model predicted a negative association between ked intensity and latitude and a positive association between ked intensity and longitude (Table 2), supporting our third hypothesis of a decreasing gradient in ked intensity from southeast to northwest. Härkönen et al.  also demonstrated decreasing off-host survival and pupal development of keds towards higher latitudes, but explained the results by reduced summer temperatures along an 1000 km long geographical gradient. In our study, the geographical gradient in latitude is only 70 km, which means that alterations in temperature within the study area due to difference in latitude alone are not likely to occur. As an alternative, we suggest that the decreasing ked intensity from southeast to northwest is due to the keds` main direction of dispersal and the possibility that the parasite population has not yet reached the carrying capacity in the whole study area.
Altitude in the study area ranges from about 100 to 466 m a.s.l. and killing sites were found between 111 and 455 m a.s.l, indicating that moose utilize habitats in the entire range of elevation gradients. A tendency of negative correlation between ked intensity and altitude was supported in our data (Table 2). Kovanci et al.  explained a negative correlation in cherry fruit fly (Rhagoletis cerasi) intensity and altitude by about 0.5°C decrease in temperature per 100 m increase in altitude and not by the elevated altitude per se. Since temperature is regarded as an important factor for off-host survival and pupal development of deer keds [41, 43], an indirect effect of altitude, through decreasing temperature, may explain the effect of altitude in our data. However, we cannot exclude the possibility of a slight contraction of moose space use from higher to lower altitudes during winter (e.g. due to more favorable snow conditions). Hence, the higher ked intensity may also be a result of more pupae being shed at lower altitudes.
Methods (sampling period, anatomical site and procedures)
We chose to restrict sampling to the first week of the hunting season for two reasons. First, we know from experience that about half of the total numbers of moose are shot during the first week of hunting, hence this period was best to maximize sample size. Moreover, allowing a longer sampling period than a week may bias the intensity results by a time-dependent, cumulative effect of swarming keds attaching to the coat.
Previous studies in moose [9, 28] and red deer  have shown that deer keds are distributed all over the hide, with aggregates around the axilla-neck and groin-anal regions. Neck region, in contrast to the groin, is covered by long guard hairs which protect keds from being torn out of the coat, both during moose movement through dense vegetation and post-killing transport and handling by humans. Hence, sampling from the dorsal neck region of hunted moose was chosen.
Three methods of quantitative collection of deer keds (L.cervi and L.mazamae) in the coat are described in the literature; cutting the hair coat with scissors , combing  and digestion with KOH . We developed our own method by removing the coat with a wool clipper, followed by careful inspection of hair and naked skin for keds.
In hindsight, this method was suboptimal due to the risk of cutting parasites into pieces, and hence biasing the number of keds per sample. To avoid this, we only counted the anterior part of demolished keds.
Welch et al. found that random sampling of 15% of a moose hide total area is sufficient to estimate densities of winter ticks (Dermacentor albipictus). Similar studies are not performed with deer keds, but using this study as a guideline, we inspected about 13.4% (20 × 20 cm = 400 cm2/29.833 cm2 average skin area for adults ) of total skin area in adults and 21.4% (400 cm2/18.666 cm2) average skin area for calves. Based on Welch`s  study, the size of our skin samples were sufficient for estimating total ked densities in calves, but somewhat small for the same estimation in adults.
Since unfavorable weather conditions , habitat  and predation [13, 14] may affect emergence success of pupae, survival of pupae through winter, spring and summer is important for ked abundance. However, within its core distribution range, pupal survival rates are assumed to be high  and therefore we did not include proxies of pupal survival as variables in our study.