From 2014 to 2022, spring surveys were performed in May and June to collect host-seeking I. scapularis nymphs in eight national parks in the eastern United States: Catoctin Mountain Park (CATO), Chesapeake and Ohio Canal National Historic Park (CHOH), Fire Island National Seashore (FIIS), Gettysburg National Military Park (GETT), Manassas National Battlefield (MANA), Monocacy National Battlefield (MONO), Prince William Forest Park (PRWI), and Rock Creek National Park (ROCR) (Fig. 1). This research expands upon tick surveillance and pathogen prevalence reported by Johnson et al. (2017) from 2014–2015. At each park, between one and nine 750 m transects were surveyed each year using dragging techniques performed with a 1 m² white drag cloth for a total area of 750 m² surveyed per transect (Table 1). While vegetation features–such as ground coverage–differed visually among transects within parks, all locations were selected based on suitable tick habitat and reports of tick presence by park staff (Johnson et al., 2017), and all areas of the park were assumed to be utilized by white-tailed deer, based on anecdotal observations. Individual transect locations remained constant between years, though, not all transects were surveyed each year due to logistical limitations (Table 1; Supplementary Material I for locations). During surveys, the drag cloth was examined for ticks every 10–15 m, and all attached ticks were visually identified to life stage and species. Most transects were surveyed twice per year, once in late May and a second time approximately ten days later in an attempt to capture peak seasonal nymph activity in this region (Eisen et al., 2016; Ogden et al., 2018). Occasionally, weather prohibited either the first or second survey session from being completed. Total counts of I. scapularis nymphs collected were recorded per survey for each transect providing density estimates (number of ticks per 750 m²; Table 1). Ixodes scapularis ticks were collected in >90% ethanol and stored at −20°C.

All visually identified I. scapularis nymphs and adults were sent to the Centers for Disease Control and Prevention (CDC; Fort Collins, CO, USA) to be tested for a suite of pathogens (following Graham et al., 2016, Graham et al., 2018), including causative agents of Lyme disease (B. burgdorferi s.s. and Borrelia mayonii), relapsing fever (Borrelia miyamotoi), anaplasmosis (Anaplasma phagocytophilum), and babesiosis (Babesia microti). Prior to testing, all submitted Ixodes ticks were confirmed to be I. scapularis using an assay that differentiates I. scapularis from the morphologically similar Ixodes affinis, implementing a slight modification to the methods described by Wright et al. (2014). If pathogen testing results were inconclusive, the sample was not included in the analysis.

During most surveys, additional effort was made to increase nymph sample size, with the aim to acquire N=50 I. scapularis nymphs for B. burgdorferi s.s. testing per study site and year, in compliance with CDC surveillance guidelines (Centers for Disease Control and Prevention, 2021). This resulted in discrepancies between the maximum number of nymphs detected per transect and the number of nymphs submitted for pathogen testing each year (Table 1). The additional surveillance efforts were not accompanied by area surveyed, thus, density of infected nymph data–the gold standard metric used to measure acarological risk of human exposure to Lyme disease–were not available. Instead, the I. scapularis nymph density data and B. burgdorferi s.s. prevalence data were used as the response variables in four models–two models for each data type–that included environmental and deer density covariates (Table 2).

Ixodes scapularis nymphal host-seeking behavior is influenced by local climatic conditions, specifically, humidity and temperature (Vail and Smith, 1998; Berger et al., 2014a, Berger et al., 2014b, Burtis et al., 2016, Eisen et al., 2016). Several weather- and climate- related metrics were considered in our models to account for environmental effects on nymph density, including proportion of hot-dry-days (pHDD), monthly relative humidity (RH), monthly Palmer’s Z-index (PZ), and monthly Palmer Drought Severity Index (PDSI). Proportion of hot-dry-days was the cumulative number of days in April and May each year when the temperature was >25°C (77°F) and no precipitation was recorded (precipitation = 0), relative to the number of days when weather data were available. Hot-dry-days were identified using precipitation and temperature data collated for each transect from the nearest National Oceanic and Atmospheric Administration (NOAA) weather station (all stations were within 10–46 km of the parks, see Supplementary Material I; NOAA, 2022b). Daily relative humidity data were acquired from the NOAA surface level humidity databases (Kalnay et al., 1996), where humidity data were recorded four times per day (at 00:00, 06:00, 12:00, and 18:00 hours) at a 2.5 degree latitude by 2.5 degree longitude grid (CHOH and PRWI fell within the same grid cell, and all other parks fell into a second grid cell). Monthly relative humidity was the sum of days each month when three of the four daily relative humidity recordings were >84% (calculated for April only, May only, and April and May; Eisen et al., 2016). This coarse relative humidity metric was used in the absence of local microclimate humidity data.

To account for long-term changes in precipitation, April and May drought indices (PDSI and PZ) were considered. PDSI is a measure of drought severity for a given month based on temperature and precipitation data and accounts for the drought risk in previous and subsequent months. A PDSI value of >4 indicates wet conditions and a value <−4 indicates drought conditions (see Heim, 2002 for calculations). PZ is a derivative of PDSI and measures the deviation of a given month from the average moisture climate for that month. PZ and PDSI were obtained from the NOAA website (NOAA, 2022c,2022d), and were estimated regionally based on the contiguous United States (CONUS) Climate Divisions (Supplementary Material I; NOAA, 2022a). All environmental variables were scaled and centered in program R v.4.1.2 using the scale() function (R Development Core Team, 2016).

The National Capital Region (NCR) Natural Resources and Science Program engaged in annual fall monitoring to estimate white-tailed deer densities for parks in this region, including, CATO, CHOH, MANA, MONO, PRWI, and ROCR. The monitoring program started in 2000 and estimated deer densities using distance estimation procedures described in the NCR surveillance plan (NPS, 2005). Briefly, the distance sampling used driven line-transects to record observations and the distances of the observations from the sampling location. These data are used with a detection function to calculate the proportion of missed observations, which allows overall density to be estimated (Buckland et al., 2005). GETT and FIIS were not part of the NCR deer monitoring program but did estimate yearly deer density through their own natural resource program using the same distance sampling methodology. Deer density data were available at the park level for the NCR parks and GETT, and at within-park, regional level for FIIS (Supplementary Material II).

All parks but one (PRWI) engaged in active deer reduction strategies. Deer management was employed at the individual park level and began in different years for each park: 2013 for ROCR, 2016 for MONO, 2018 for CHOH and MANA, 2019 for FIIS (region WFE), and in 2020 for FIIS regions SH and WA. CATO and GETT had ongoing deer reduction management that initiated before the start of this research, and PRWI had no deer management. The number of deer removed by each park varied (Supplementary Material III). Deer density and management data were included in the models at a two-year lag, reflecting the two-year life cycle of I. scapularis (Fish, 1993, Wolf et al., 2020). Deer reduction efforts are expected to reduce I. scapularis reproductive events and therefore suppress the number of immature ticks in subsequent years, with the effects of deer management most likely to be observed for the nymphal life stage at a two-year lag. Two deer related variables were considered for the models: a two-year lagged deer density variable and a two-year lagged management variable. The two-year lagged deer density variable assigned the deer density from two years prior to a given surveillance year (e.g., if the tick surveillance occurred in 2020, the deer density from 2018 was assigned). The two-year lagged management variable was binary (0/1), with a value of 0 assigned if no deer management had occurred two or more years prior (referred to as “pre-management”), and a value of 1 assigned to surveys that occurred two or more years following the first year of deer management (referred to as “post-management”). The total number of years with deer density data included in both pre- and post- management categories can be found in Table 3. This method groups all nymph densities pre-management together and all nymph densities post-management. For some parks, post-management data extended 2-years post initial deer removals, and some extended 7-years past initial deer reduction.

Ixodes scapularis nymph count data were associated with 750 m² surveillance areas, thus we use “density” to describe these data, though raw counts were used in these models. Nymph densities were used in two question-focused models designed to address (i) the effect of deer density on I. scapularis nymph densities and (ii) the impact of deer reduction management on nymph densities. For both models, the response variable was the maximum nymph density observed at each transect among survey sessions within a given year. Maximum nymph density was the maximum number of nymphs observed at each 750 m² transect across all sampling sessions (i.e., maximum nymph density could be obtained from different sampling sessions for transects within a park), and all transects were included as replicates within parks. Maximum density was used to minimize variability introduced by sampling before, or after, peak nymph emergence. By using maximum density, we aimed to understand the highest potential acarological risk of encountering a nymph, using nymph density as a proxy. Model selection methods were not implemented because the models were hypothesis driven. To determine informative environmental variables, preliminary models were executed whereby maximum nymph density (nymph count per 750 m²) was the response variable, and only environmental variables were included as predictor variables (excluding all other variables). This approach was implemented to account for the influence of climate on nymph densities without overfitting the final model. Environmental variables were retained in the final model if their 95% credible interval did not overlap with zero in the preliminary model. Final environmental covariates were not highly correlated (correlation coefficient, r≤0.75).

To understand if variation in deer density affected the density of host-seeking I. scapularis nymphs, a Bayesian generalized linear model with group-specific terms was utilized (model 01). Nymph density data for transect survey i in park j, were assumed to follow a negative binomial distribution with a mean μ_ij and an overdispersion parameter φ (≥0). Negative binomial regressions account for over-dispersed data, which is often observed with count datasets. The log link function was implemented to relate the response variable to the linear combination of the predictors: the intercept (β₀), corresponding to the logged mean number of nymphs observed; the effect of the proportion of hot-dry-days (pHDD_ii) on nymph count (β₁); the effect of April relative humidity (RH_April_ii) on mean nymph count (β₂), and the effect (β₃) of park specific lagged deer density (deer_dens_ij) on mean tick density (see Table 2). To account for park level effect and multiple transect locations within parks, park identity was included as a random effect (intercept), ρ_j. Model 01 was designed to understand the general trend of deer density impact on nymph density, not the effect at the park level, thus park was included as a random effect. The final model structure was as follows: densityij~NB(μij,ϕ) (1) log(μij)=β0+β1×pHDDij+β2×RH_Aprilij+β3×deer_densij+ρj

A second Bayesian generalized linear model was used to address the impact of park-specific deer management on tick density (model 02). In this model, only parks with at least two years of nymph density data following the initiation of deer reduction efforts (post-management) were included. These parks included CHOH, MANA, MONO, FIIS (region WFE), and ROCR (Table 3; see Supplementary Material III for deer removal effort). Similar to the first model, nymph density data for transect survey i were assumed to follow a negative binomial distribution with a mean M_ij and an overdispersion parameter Φ (≥0). The log link function was implemented to relate nymph densities to the linear combination of the fixed effects: the intercept (B₀), corresponded to the logged mean density of nymphs observed at the reference park; the effect of the proportion of hot-dry-days (pHDD_ij) on nymph density (B₁); the effect of April relative humidity (RH_April_ij) on mean nymph density (B₂), and the effect (B₃) of the interaction between park identity (Park_ID_j) and deer management (binary term; deer_manage_ij; 0, no management; 1, management two or more years prior) (Table 2). Model 02 differs from model 01 in that the effect of deer removal on nymph density at each park was of interest, so park was included as an independent variable (interaction) instead of a random effect. The final model structure was as follows: densityij~NB(Mij,Φ) (2) log(Mij)=B0+B1×pHDDij+B2×RH_Aprilij+B3×deer_manageij×Park_IDj

Using the B. burgdorferi s.s. pathogen data, two models were designed to understand (i) the effect of deer density on infection prevalence and (ii) the effect of deer reduction strategies on infection prevalence. Similar to the density models, environmental variables were assessed in a preliminary model and only significant variables were included in the final pathogen model. Pathogen data were only included in the model if the number of ticks submitted for testing was N>1 for a given year (Table 1). This sample size is lower than CDC recommendation (N=50), which is intended for making conclusions about a specific prevalence estimate in a given year or locale. Our interest, however, is in making inferences across multiple years and locations for the effect of deer reductions, and removing park-years below certain sample size thresholds may bias conclusions if there is a correlation between infection prevalence and the number of ticks collected. The N>1 sample size threshold produced the same trends as a more stringent sample size (N≥10; Supplementary Materials IV) and allowed for the retention of surveillance sites that experienced declines in nymph densities following management (e.g., MANA and MONO). In the first model, the pathogen data were modeled using a Bayesian generalized linear model with group-specific terms (model 03). The response variable was a vector of successes to failures (B. burgdorferi s.s. positive versus negative nymphs) during survey i at park j (Bbss_pos_ij|Bbss_neg_ij), following a binomial distribution with a mean probability of success, ω_ij, given the number of attempts (ψ; total number of nymphs tested; sum of Bbss_pos_ij and Bbss_neg_ij). The model’s fixed effects included an intercept (η₀), corresponding to the average number of B. burgdorferi s.s. positive nymphs; the effect of May moisture anomalies (PZ_May_ii; η₁), April relative humidity (RH_April_ij; η₂), and May relative humidity (RH_May_ij; η₃) on the prevalence of B. burgdorferi s.s.; and the effect of lagged deer density (deer_dens_ij; η₄). Park identity was included as a random effect (P_j), as the general effect of deer density–not park specific effects–on pathogen prevalence was of interest. The logit function was implemented to relate the linear combination of the predictors to the response variable (Table 2). The final model structure was: Bbss_posij∣Bbss_negij~BIN(ωij,ψ) (3) logit(ωij)=η0+η1×PZ_Mayij+η2×RH_Aprilij+η3×RH_Mayij+η4×deer_densij+Pj

To address the impact of deer removal on prevalence of B. burgdorferi s.s. in nymphs, a second pathogen Bayesian generalized linear model was implemented including only parks with at least two years of pathogen data post-management (model 04). These parks included CHOH, MANA, MONO, FIIS (region WFE), and ROCR (Table 3). The response variable was a vector of successes to failures (B. burgdorferi s.s. positive nymphs, Bbss_pos_ij; B. burgdorferi s.s. negative nymphs, Bbss_neg_ij) following a binomial distribution with a mean probability of success, Ω_ij, given the number of attempts (Ψ; total number of nymphs tested; sum of Bbss_pos_ij and Bbss_neg_ij). The model’s fixed effects included an intercept (H₀), corresponding to the average number of B. burgdorferi s.s. positive nymphs at the reference park; the effect of May moisture anomalies (PZ_May_ij; H₁), April relative humidity (RH_April_ij; H₂), and May relative humidity (RH_May_ij; H₃); and the interaction between park identity (Park_ID_j) and the binary lagged deer management term (deer_manage_ij; H₄). The inclusion of an interaction between park identity and deer management allowed for the assessment of within-park deer-removal effort on nymph densities. This model had no random effect terms (Table 2). The final model structure was as follows: Bbss_posij∣Bbss_negij~BIN(Ωij,Ψ) (4) logit(Ωij)=H0+H1×PZ_Mayij+H2×RH_Aprilij+H3×RH_Mayij+H4×deer_manageij×Park_IDj

For all models, priors were weakly informative for fixed variables: N (0, 2.0; mean and standard deviation). The priors on the covariance matrices of the group-specific terms (random effects) were uninformative, with regularization, concentration, scale, and shape set to 1. Models were fit in R v.4.1.2 using function stan_glmer() in package rstanarm v.2.21.3 (R Development Core Team, 2016, Brilleman et al., 2018, Goodrich et al., 2022) and were run for 30,000 Markov chain Monte Carlo iterations with three chains, following a 30,000-iteration warmup (burn-in). Model convergence was confirmed for all models through visual inspection of parameter trace plots and ensuring the Gelman and Rubin’s potential scale reduction factor (R^) values were near to 1 (<1.1 Brooks and Gelman, 1998). Further, all variables had an effective sample size of ≥20,000. Variables were considered important (or, “significant”) predictors of the dependent variable if their posterior credible intervals did not overlap with zero. Post-hoc analyses were conducted to understand the effect of the interaction between the binary deer management variable and park identity. Estimated marginal means (Least-Squares means) were calculated using the emmeans() function in R package emmeans v.1.7.3 (Lenth, 2022).

For all predictions, the posterior_predict() function in R package rstanarm v.2.21.3 was implemented (Goodrich et al., 2022). Mean I. scapularis nymph density (number of ticks per 750 m²) was predicted using model 01 for each park, addressing the effect of white-tailed deer density at each park. Nymph densities were predicted across a range of deer densities, from 5 through 85 deer per km² at intervals of 5 deer per km², and yearly for each park using the corresponding lagged deer density for 2014–2022. Environmental covariates were held constant at their means for all predictions. These predictions were used to understand the effect of deer density in reducing tick density, specifically focusing on deer densities of 5 and 20 deer per km² (commonly cited deer density thresholds). Similarly, using model 03, the mean prevalence of B. burgdorferi s.s. infection in I. scapularis nymphs was predicted across the same range of deer density values for each park (5–85 deer per km² at intervals of five) and yearly at all parks.

To understand the impact of deer management on tick densities and B. burgdorferi s.s. prevalence, mean peak I. scapularis nymph densities (per 750 m²) were predicted using model 02 under scenarios of no white-tailed deer management 2-years prior and deer removal 2-years prior (values 0 and 1, respectively, also referred to as “pre-management“ and “post-management”) for each park. Similarly, using model 04, the mean number of I. scapularis nymphs infected with B. burgdorferi s.s. was predicted for both management scenarios (0 and 1) with the total number of nymphs (number of attempts) held constant (N=100).

The density of infected nymphs (DIN; per 750 m²) was calculated by combining the prediction estimates from tick density and pathogen prevalence models (models 01 and 03). The DIN pre- and post- white-tailed deer management was calculated by combining predictions of nymph density and B. burgdorferi s.s. prevalence from models 02 and 04. For computational purposes, values from prediction posteriors were subset to 1,000 (from 90,000).

Between 2014 and 2022, a total of 30 transects were surveyed for I. scapularis nymphs with 164 total data points (Martin et al., 2023). Transect locations were consistent across years though the number of transects surveyed varied by park (Table 1). The maximum density of I. scapularis nymphs varied across parks and transects (Fig. 2, A; Table 1), with the range of maximum nymph densities per transect between 0–153 per 750 m², and the average nymph density at 16 ticks per 750 m² (7 median, 15.9 mean, 24.3 SD). Nymph densities also varied pre- and post- deer management (Fig. 3, A, C). Pathogen data were not collected every year for each park and transect (total pathogen prevalence data points N=118; Tables 1 and 3), and after filtering by sample size (N>1 nymph submitted), 112 B. burgdorferi s.s. prevalence data points remained (sample thresholds of N>1 and N≥10 yield the same trends; Supplementary Material IV). Prevalence of B. burgdorferi s.s. in I. scapularis nymphs ranged from 0 to 50% across parks (9.2% median, 13.3% mean, 13.4% S.D.; Fig. 2, B; Table 1).

White-tailed deer densities varied across national parks and park regions through time (Fig. 2, C), and were available at the park level for all parks and at within-park regional levels for FIIS. The range (minimum and maximum) of deer densities observed for each park (deer per km²) from 2012–2020 were as follows: CATO 6–17, CHOH 6–57, FIIS 7–74 (across regions), GETT 6–16, MANA 7–38, MONO 17–82, PRWI 6–24, and ROCR 3–30 (Fig. 2, C). The average deer density across all parks and park regions between 2012 and 2020 was 26 deer per km² (26.0 mean, 19.4 S.D.). During this period (2012–2020), only one park had a deer density of <5 deer per km² (ROCR), and it was for a single year (2020); however, most parks and park regions hosted deer densities of <20 deer per km², through management efforts or by natural population fluctuations (Fig. 2, C). All but one park engaged in deer removal practices, and the number of deer removed varied among parks and years (Supplementary Material III). Deer reduction efforts varied across these parks or park regions between 2012–2020, with the number of years when removal occurred ranging from 2 to 8, and the total number of deer removed ranging from 158–659 during this period (Table 3; Supplementary Material III). All parks that engaged in deer reduction experienced a decline in deer densities following management implementation (Fig. 3, B).

The effect of deer density on I. scapularis host-seeking nymph density was assessed using model 01. This model revealed that lagged deer density was significantly positively associated with nymph density (β₃ = 0.40 [95% posterior credible interval (PCI) = 0.16, 0.66]), resulting in a 49% increase ([(exponentiated coefficient – 1) × 100]; UCLA Statistical Consulting Group) in the number of I. scapularis nymphs per 750 m² with a 1 SD increase in 2-year lagged deer density (16.2 deer per km² S.D., for data points included in model). Both the proportion of hot-dry-days and April relative humidity had significant negative effects on nymph counts (β₁ = −0.28 [95% PCI = −0.55, −0.00]; β₂ = −0.40 [95% PCI = −0.61, −0.18]), with 24% and 33% reduction in nymphs with a 1-SD increase in each, respectively (Fig. 4, A). The intercepts associated with the random effect of park ID ranged from −0.71 to 0.70, with GETT and ROCR having the largest positive intercepts (i.e., these parks had higher nymph densities than the average park) and MONO and CHOH having the largest negative intercepts (i.e., these parks had lower nymph densities than the average park; Supplementary Material V).

Predictions of tick densities at deer densities from 5 to 85 deer per km² varied across parks (Fig. 4, B). At a density of 5 deer per km², I. scapularis nymph densities ranged between 4–16 nymphs per 750 m² across all parks, with the highest densities observed at GETT and ROCR (x¯GETT=16 [95% PCI = 0, 76]; x¯ROCR=14 [95% PCI = 0, 62]) and lowest at MONO and CHOH (x¯MONO=4 [95% PCI = 0, 23]; x¯CHOH=4 [95% PCI = 0, 18]). At a deer density of 20 deer per km², densities of 5–24 nymph I. scapularis ticks per 750 m² were observed across parks, with highest predicted densities of ticks observed at GETT and ROCR (x¯GETT=24 [95% PCI = 0, 109]; x¯ROCR=20 [95% PCI = 0, 87]) and the lowest nymph densities observed at MONO and CHOH (x¯MONO=6 [95% PCI = 0, 31]; x¯CHOH=5 [95% PCI = 0, 25]).

To understand the binary effect of deer reduction practices on nymph densities, a second model was implemented (model 02). Four of five parks experienced a reduction in nymphs following deer removal. Post-hoc analysis revealed the contrast in estimated marginal mean peak tick densities when no deer management occurred (pre), and following deer reduction (post), was only significant for one park, ROCR (x¯ROCR_pre-post=1.37 [95% PCI =0.52, 2.23]). Contrasts in estimated marginal mean tick densities for the other four parks and park regions (MONO, CHOH, MANA, and FIIS [WFE]) were not significant (Supplementary Material VI). For ROCR, when no deer management was implemented, the predicted mean value of nymphs was 58 (95% PCI = 3, 206), which decreased to 14 nymphs following deer management implementation (95% PCI = 0, 47; Table 3). For MONO, MANA, and FIIS (WFE), there was a general trend for a decrease in the density of ticks following management (x¯MONO_pre=9 [95% PCI = 0, 37], x¯MONO_post=6 [95% PCI = 0, 29]; x¯MANA_pre=16 [95% PCI = 0, 55], x¯MANA_post=9 [95% PCI = 0, 34]; x¯FIIS-WFE_pre=76 [95% PCI = 4, 256], x¯FIIS-WFE_post=48 [95% PCI = 2, 180]). For CHOH, the density of nymphs increased slightly following management (Table 3). Neither of the climate variables–proportion of hot-dry-days nor April relative humidity–had significant effects on the density of nymphs (Supplementary Material VI; B₁ = 0.05 [95% PCI = −0.25, 0.37]; B₂ = −0.16 [95% PCI = −0.41, 0.09]).

The effect of deer density on B. burgdorferi s.s. prevalence in questing I. scapularis nymphs was assessed using model 03. Lagged deer density did not significantly affect the B. burgdorferi s.s. prevalence in I. scapularis nymphs (Supplementary Material IV; η₄ = 0.04 [95% PCI = −0.15, 0.23], odds ratio 1.04). April relative humidity was significantly negatively associated with B. burgdorferi s.s. prevalence (η₂ = −0.14 [95% PCI = −0.28, −0.01]) with an odds ratio of 0.87 (ratio of 1 indicating no change, >1 indicating increase, and <1 indicating decrease), but May moisture anomaly and May relative humidity were not (η₁ = −0.16 [95% PCI = −0.39, 0.07]; η₃ = −0.00 [95% PCI = −0.25, 0.23], respectively). The intercepts for the random effect of park ranged from −1.47 to 0.60, with GETT and FIIS having the largest positive intercepts (i.e., a higher proportion of B. burgdorferi s.s. positive nymphs than the average park) and PRWI and CHOH having the largest negative intercepts (i.e., a lower proportion of B. burgdorferi s.s. positive nymphs than the average park; Supplementary Material IV). The predicted prevalence of B. burgdorferi s.s. infected I. scapularis nymphs increased slightly–not significantly–with deer density when environmental variables were held constant (Supplementary Material IV).

To understand the effect of park specific deer reduction efforts on B. burgdorferi s.s. infection prevalence in I. scapularis nymphs, a second pathogen model was implemented (model 04). Post-hoc analysis of marginal means revealed that deer management did not result in significant reductions in B. burgdorferi s.s. prevalence between pre- and post- management in any park, but ROCR and CHOH did experience general declines in B. burgdorferi s.s. prevalence following deer reduction efforts. When no deer management was implemented, the predicted prevalence of B. burgdorferi s.s. at ROCR and CHOH were 17 and 4%, respectively, which was reduced to 9 and 1% following deer management. MONO and MANA experienced a general increase in B. burgdorferi s.s. prevalence following deer reductions, with MONO increasing from 10 to 22% and MANA increasing from 12 to 17%, each with wide credible intervals (Supplementary Material VI). April relative humidity significantly reduced the prevalence of B. burgdorferi s.s. in nymphs (Supplementary Material VI; H₂ = −0.70 [95% PCI = −1.19, −0.31]), with an odds ratio of 0.50 with a 1-SD unit increase in April relative humidity. May moisture anomaly and May relative humidity did not affect B. burgdorferi s.s. prevalence (H₁ = −0.42 [95% PCI = −0.93, 0.07]; H₃ = 0.22 [95% PCI = −0.32, 0.75]).

DIN was predicted for all parks at deer densities of 5 and 20 deer per km², and for each park at that park’s mean deer density from 2012–2020 (Fig. 5) using prediction results from models 01 and 03. At 5 deer per km², the mean DIN at all parks was 3 or fewer per 750 m². The highest mean DIN was at GETT, with 3.18 infected nymphs per 750 m², and the lowest at CHOH (0.12 infected nymphs per 750 m²). Similarly, at 20 deer per km², no park exceeded a mean DIN of 5 infected nymphs per 750 m². The highest mean DIN at 20 deer per km² was observed at GETT and lowest at CHOH (4.95 and 0.18 nymphs per 750 m², respectively). The predicted DIN per 750 m² pre- and post- deer management was estimated using output from models 02 and 04. Following deer management, the DIN varied among parks (Fig. 3, D; Table 3).

There has been much debate about the utility in managing white-tailed deer densities as a method to reduce nymph densities and Lyme disease risk. In the absence of complete elimination of white-tailed deer, the effect of deer management has been mixed (see review by Kugeler et al., 2016). Several foundational studies have demonstrated a positive relationship between deer densities and nymph densities at limited scales (e.g., island population; Wilson et al., 1984, Wilson et al., 1988). Here, we demonstrate a significant, positive association between 2-year lagged white-tailed deer densities and I. scapularis nymph densities across a broad (>300 km) geographic scale at eight national parks or park regions in the eastern United States.

While all parks experienced increasing nymph densities with increasing deer densities, there was substantial variation in tick densities across parks, with the highest densities observed at GETT, ROCR, and CATO, and the lowest densities observed at PRWI, CHOH, and MONO. These variations are likely due to local environmental differences among parks that are not explored here, given the focus of this study. Briefly, variation in vegetation type (Ginsberg et al., 2020) and the amount of forest fragmentation (Ferrell and Brinkerhoff, 2018), among other factors, could have influenced the variation in density of I. scapularis nymphs observed across parks. Further, there was variation within parks among survey transects within years. While all parks experienced annual fluctuations in nymph densities, the general trend was decreasing nymph densities throughout this study period, at least partly mirroring declining deer densities across parks.

White-tailed deer reduction efforts resulted in declines in tick densities at four of the five national parks or park regions that engaged in deer management; a significant reduction in tick density was observed in one of five parks (ROCR), a non-significant reduction was observed in two parks (FIIS and MANA), and no appreciable change was detected in the remaining two parks (CHOH and MONO). While only ROCR experienced a significant reduction in nymph densities following white-tailed deer removal, MONO, MANA, and FIIS (region WFE) all experienced general declines in nymph densities following management. ROCR engaged in eight consecutive years of deer removal and removed 505 deer in total, reducing deer densities from 28.3 deer per km² in 2012 to 3.47 deer per km² in 2020. MONO, MANA, and FIIS (WFE) removed 659, 451, and 259 deer over five, three, and two years, respectively. During their management periods, MONO decreased deer densities from 66.14 deer per km² in 2016 to 16.89 deer per km² in 2020; MANA reduced deer densities to 7.17 deer per km² in 2020 (first year of management density data are missing); and in one-year of management, FIIS region WFE reduced deer densities from 62.70 deer per km² in 2019 to 28.40 deer per km² in 2020. CHOH was the only park to experience a slight, non-significant, increase in nymph densities following three years of management and removal of 158 deer (deer densities reduced from 29.89 deer per km² in 2018 to 5.79 deer per km² in 2020). However, pre-management, CHOH had a low density of ticks (5 nymphs per 750 m²), which only increased by one nymph per 750 m² post-management (6 nymphs per 750 m²); a shift that could be attributed to natural, inter-annual fluctuations in tick populations (Deblinger et al., 1993).

Notably, there was no standardized deer reduction protocol implemented across parks, and substantial variability was observed in white-tailed deer removal efforts among parks and across years within parks. This study represents a real-world implementation of deer management for nymph control and revealed a general trend of reduced nymph densities following deer removal. The variability in the magnitude of this effect across parks may in part be explained by a potential delay following management implementation. Foundational studies reported gradual declines in nymph densities following deer reductions (Wilson et al., 1988, Deblinger et al., 1993), generally observable two or more years following management (3–4 years for significant reductions). While these studies used nymph counts from small mammal hosts as the metric of abundance–which is not predictive of questing nymphs (Ginsberg et al., 2020)–similar trends could be occurring in questing nymphs, and further declines may be observed in coming years for parks that recently initiated deer management (3 or fewer years). Specifically, successive years with removal of deer may result in successive reduction in nymphs, and the true magnitude of this effect may not be observed in the first few years following management and may be dampened by combining all densities post-management into one group. Our results follow this trend, with larger percent changes in nymph densities observed in years 2–3 post-management relative to the first year; however, the interpretation is limited due to three of the five parks having only three years of data post-management.

The effect of white-tailed deer management on nymph densities may be dampened in areas where environmental factors limit nymph abundance. Notably, MONO and CHOH supported very few nymphs prior to management efforts (9 and 5 per 750 m², respectively). Following deer management, MONO experienced the smallest percent decrease in nymph density among parks, and CHOH experienced a slight increase in nymph density (6 nymphs per 750 m² for each park). This suggests that lowering the density of the reproductive host (white-tailed deer) may not be a significant factor in controlling nymph densities if densities are already low or constrained due to other abiotic or biotic factors. Other factors known to influence nymph densities include vegetation and woody debris (invasive plants, Allan et al., 2010;; coarse woody debris, Larson et al., 2022), leaf litter (Schulze et al., 1995), small mammal hosts (Ostfeld et al., 2006), and forest characteristics (canopy cover, Ginsberg et al., 2020); factors that likely vary across our study range and contribute to the variable tick densities observed across parks.

Lagged white-tailed deer density did not have a significant effect on B. burgdorferi s.s. infection prevalence in this study, nor did white-tailed deer reductions change the prevalence of B. burgdorferi s.s. in parks that engaged in management. However, there was a weak positive trend of increased B. burgdorferi s.s. prevalence with increasing deer density. Currently, there is a paucity of studies examining the relationship between deer density and B. burgdorferi s.s. infection rates in I. scapularis. Literature from a similar system–I. ricinus and B. burgdorferi s.l. in Europe–have reported mixed results, where some studies reported no relationship between deer density and pathogen prevalence in ticks (Pichon et al., 1999), some found a positive relationship between deer density and B. burgdorferi s.l. (James et al., 2013), and some found a negative relationship (lower Borrelia prevalence in areas with higher deer densities; Rosef et al., 2009, Mysterud et al., 2013); though, deer density ranges among these studies likely differed. Further, we report discrepancies in B. burgdorferi s.s. among parks engaging in white-tailed deer reduction efforts. Specifically, two parks experienced a decrease in B. burgdorferi s.s. prevalence, while two experienced an increase in B. burgdorferi s.s. prevalence following management efforts. Our findings demonstrate a weak, non-significant relationship among white-tailed deer density and B. burgdorferi s.s. prevalence across a broad geographic range but reveal variable effects on B. burgdorferi s.s. prevalence following deer removal.

White-tailed deer play an important role in the reproductive stage of I. scapularis, with their large body and home range sizes supporting carriage of large numbers of ticks. Higher deer densities are assumed to facilitate high densities of immature I. scapularis in the environment. However, the consequence of high larval and nymphal burden in the landscape on B. burgdorferi s.s. transmission risk remains poorly understood. There is a possibility that high densities of immature life stages in the environment will result in larvae and nymphs utilizing a broader range of hosts, including those that do no serve as B. burgdorferi s.s. reservoirs such as white-tailed deer (Huang et al., 2019). This may theoretically reduce B. burgdorferi s.s. transmission risk to larvae; however, this association has not been made. A more plausible scenario is that higher densities of immature life stages in the landscape result in an increased density of larvae and nymphs found on individual small mammal hosts–such as white-footed mice–which could result in increased risk for transmission of B. burgdorferi s.s. among ticks during co-feeding (Levin et al., 1997, Belli et al., 2017) and may increase the risk of horizontal transmission to small mammal hosts (Eisen, 2018). Further, larval and nymphal burdens are often heterogenous among white-footed mouse hosts, with few individuals supporting high tick loads and likely responsible for the majority of B. burgdorferi s.s. transmission events (Brunner and Ostfeld, 2008, Devevey and Brisson, 2012).

Conversely, low densities of white-tailed deer should limit the density of immature I. scapularis in the landscape; however, it is unclear how this corresponds to reduced B. burgdorferi s.s. transmission risk. Reduced densities of immature life stages in the environment may not translate to lowered densities of nymphal and larval ticks on individual white-footed mice and other reservoir hosts, particularly if on-host densities are driven by host behavior (Devevey and Brisson, 2012). Reductions of white-tailed deer populations can aid vegetation regeneration and bolster small mammal communities (Rooney and Waller, 2003, Byman, 2013, Shelton et al., 2014), though diverse host communities may not reduce B. burgdorferi s.s. infection risk (no evidence for the dilution effect in some studies; States et al., 2014, Linske et al., 2018). The variable–observations of both increase and decrease–non-significant, effect of deer management on B. burgdorferi s.s. prevalence suggests that deer reductions are likely not the sole factor influencing pathogen carriage in nymphs and warrants further investigation.

It is possible that other environmental factors–including climate, weather, and abundance of host species–may play a larger role in B. burgdorferi s.s. prevalence than white-tailed deer densities. For example, in similar systems in Europe, the abundance of rodent hosts is strongly correlated with the infection prevalence of certain pathogens in Ixodes nymphs (Borrelia afzelii and Neoehrlichia mikurensis; Krawczyk et al., 2020). Further, host selection may influence B. burgdorferi s.s. transmission risk and prevalence in nymphs, though, this has mainly been observed across a north-south gradient (Ginsberg et al., 2021). In this study, host abundance for immature life stages were not available. However, weather data were incorporated. Here, within-year April relative humidity was significantly negatively correlated with B. burgdorferi s.s. prevalence; however, the connection between spring relative humidity and B. burgdorferi s.s. infection prevalence is unintuitive. Within-year climate and weather variables are not likely to affect acquisition of B. burgdorferi s.s. infection by nymphal ticks, as this should occur during the previous year at the time of the first bloodmeal as a larva. Presumably, higher spring humidity facilitates off-host survival of nymphs (Stafford, 1994), but prolonged questing periods are not expected to influence B. burgdorferi s.s. infection status (or transmission potential; Samanta et al., 2022). One hypothesis is that there is an interaction between weather and B. burgdorferi s.s. infection status that could influence survival, behavior, or detection of B. burgdorferi s.s. infected nymphs. For example, B. burgdorferi s.s. infected nymphs showed increased phototaxis and tended to quest at increased heights, which may make them more vulnerable to environmental conditions (Lefcort and Durden, 1996). However, evidence for such an interaction is currently lacking and would require additional research. More likely, environmental factors, such as humidity, may be correlated with other biotic factors not assessed here.

Previous research suggested there may be a deer density threshold below which there would be gradual declines in tick abundances (Wilson et al., 1988), with more recent suggestions that this threshold should be 3–5 deer per km² (Telford, 2017). However, there was limited scientific support for this claim. We found variable reductions in the density of infected nymphs–the most representative metric for human risk in the absence of human tick encounter or human case data–for parks reaching deer densities of 20 deer per km², while only one park attained a density of less than 5 deer per km². Specific target densities are useful for managers to strive for but may not represent thresholds, per se, such that reductions that do not reach the target may still be effective (Lloyd-Smith et al., 2005). All parks had fewer than 5 B. burgdorferi s.s. infected nymphs per 750 m² at a white-tailed deer density of 20 deer per km², which reduced to 3 or fewer infected nymphs per 750 m² at a deer density of 5 deer per km². However, given the limitations of our data (see below), the efficacy of deer density thresholds needs to be explored further.

Our results show that white-tailed deer density is correlated–though, variably–with nymph densities (as was demonstrated by Deblinger et al., 1993); however, deer density has no clear relationship with B. burgdorferi s.s. prevalence. Thus, additional factors not addressed here are likely influencing nymph densities, to some extent, and substantially contributing to B. burgdorferi s.s. prevalence. These findings suggest that managing white-tailed deer densities alone may not be effective in reducing DIN, but our interpretation is limited by the nature of our dataset, with few tick samples post-management for some locations. Further, our DIN results were estimated using two data streams–density of nymphs and infection prevalence–which vary in their relationship with deer density. Nevertheless, deer reduction efforts may be useful when implemented in combination with other management actions to target B. burgdorferi s.s. prevalence and nymph density, simultaneously. Integrated management regimes are already being tested without a clear consensus on best practices (Williams et al., 2017, Mandli et al., 2021), and we demonstrate that deer reduction may be a useful tool for inclusion in future multi-faceted management regimes.

Our study documents a regional scale assessment of the efficacy of white-tailed deer management on controlling nymph densities in the environment, and while these findings contribute additional understanding of this complex topic to the literature, they are not without limitations. Most notably, across our study sites and years, we had only one sampling point where white-tailed deer density reached ≤5 deer per km². While this was a limitation for the lowest end of the white-tailed deer density range, there were ~20 data points with corresponding deer densities between 5–10 deer per km², and thus, we believe our model projections to be robust, even at the lowest end of this range. However, additional data at densities ≤5 deer per km² will inform variability at this lower end. Another caveat of our research was that post-management data were limited for some parks that engaged in deer reductions. For three of the five parks, management was recently initiated–in 2018 or 2019–and only 2–3 years of post-management data were available. Further, low nymph densities, which were likely a result of deer removal efforts, limited our sample sizes for pathogen testing post-management. Continued surveillance in these parks will allow for a long-term assessment of the efficacy of deer management practices.