Since the 2017 discovery of established populations of the Asian longhorned tick, (Haemaphysalis longicornis Neumann) in the United States, populations continue to be detected in new areas. For this exotic and invasive species, capable of transmitting a diverse repertoire of pathogens and blood-feeding on a variety of host species, there remains a lack of targeted information on how to best prepare for this tick and understand when and where it occurs. To fill this gap, we conducted two years of weekly tick surveillance at four farms in Tennessee (three H. longicornis-infested and one without) to identify environmental factors associated with each questing life stage, to investigate predictors of abundance, and to determine the likelihood of not collecting ticks at different life stages. A total of 46,770 ticks were collected, of which 12,607 H. longicornis and five other tick species were identified. Overall, abundance of H. longicornis was associated with spring and summer seasons, forested environments, relative humidity and barometric pressure, sunny conditions, and in relation with other tick species. The likelihood of not collecting H. longicornis was associated with day length and barometric pressure. Additional associations for different life stages were also identified and included other tick species, climatic variables, and environmental conditions. Here, we demonstrated that environmental variables can be useful to predict the presence of questing H. longicornis and provide ideas on how to use this information to develop a surveillance plan for different southeastern areas with and without infestations.

Tick Collection. Multiple tick species were collected at four privately owned farms in three eastern Tennessee counties. At each farm, 100 m transects were identified, representing three habitat types (forest, forest-field edge, and open pasture).We dragged for ticks along transects with a corduroy drag and checked the cloth every 10m along each transect for the presence of any tick species or life stage. All ticks found on the drag cloth were then stored in a tube containing 80% ethanol or RNA later; one tube was used for each transect. Larvae were collected with labeled lint roller sheets and stored in plastic bags. Once a site was identified and confirmed as infested with H. longicornis, we began to drag weekly from epidemiological weeks 20–40 and then every other week for epidemiological weeks 41–19. Ticks were identified to species and life stage using a dissecting microscope and dichotomous keys. 

Variable Selection and Acquisition. Potential environmental predictors of H. longicornis abundance (larval stage, nymph stage, females, and total collected) were considered for investigation. At the beginning of each transect, we used a Kestrel 3000 pocket weather meter-heat stress monitor (Scientific Sales Inc, Lawrenceville NJ USA) to measure environmental conditions for each collection event (transect). These continuous variables included temperature, relative humidity, barometric pressure, and wind speed. We also identified categorical variables including farm, county, daylength (number of minutes between sunrise and sunset), season based on the equinox, habitat (identified at the site as edge, field, or forest), observed sun cover (cloudy, partly cloudy, full sun), and the number of habitat-sharing ticks of different life stages, sexes, and species obtained during collection. These included counts of larval stage, nymphal stage, males, females, and total collected for the species Haemaphysalis leporispalustris Packard, Amblyomma americanum L., Dermacentor albipictus Packard, Dermacentor variabilis Say, and Ixodes scapularis L.

Descriptive Analysis. All statistical analyses were performed using STATA Version 16.1. Relative activity, similar to seasonality, of each H. longicornis life stage was calculated by determining the mean percent of ticks collected during each epidemiological week of the study. Normality of the continuous variables was assessed using the Shapiro-Wilk test, implemented in the STATA swilk command. The continuous variables included daylength, temperature, relative humidity, barometric pressure, wind speed, and total number of ticks collected. All assessed variables were non-normally distributed; thus, median and interquartile ranges were used as the measures of central tendency and dispersion, respectively. Distribution of the categorical variables (collection year, county, farm site, season, habitat, and sun cover at collection start) and their 95% confidence intervals were computed in (SAS/R).

Investigation of Predictors of Abundance. A series of negative binomial models, each built in two steps, was used to investigate the predictors of H. longicornis abundance. First, the STATA glm command was used to investigate the univariable associations between each potential predictor and the outcome under a Poisson distribution, applying a relaxed alpha level of 0.10. Since there was evidence of significant overdispersion of all Poisson models (Pearson dispersion parameters for life stages: 1.79–173.82; P < 0.05), as well as excess zero counts of the outcome variables, zero-inflated negative binomial (ZINB) models were fit to the data using the STATA zinb command. The Poisson and ZINB models were compared using their Akaike Information Criterion (AIC) values, with the model having the lowest value considered the best fitting model. Potential predictor variables with P < 0.10 were considered for inclusion in the multivariable models. Potential predictor variables identified in either the negative binomial or logit portions of the univariable zero-inflated models were used to develop multivariable models for each H. longicornis life stage.

The second step in model building involved development of a multivariable ZINB model for each age/sex category using a manual backwards elimination process using a critical alpha of 0.05. The coefficients of all variables were reviewed at each step for evidence of confounding. In situations where the removal of a variable resulted in a change of 20% or more in the coefficients of any of the variables in the model, the removed variable was considered a confounder and retained in the model, regardless of its statistical significance. No biologically plausible two-way interaction terms were identified for testing in the models. Incidence risk ratios (IR) and their 95% confidence intervals were computed for all variables retained in the final negative binomial part of the zero-inflated models. Odds ratios (OR) and their 95% confidence intervals were computed for all variables retained in the final logit part of the zero-inflated models.