Chronic wasting disease (CWD) is a neurodegenerative disease of cervids caused by a misfolded prion protein (PrP^CWD), through contact with other infected deer, from the environment as well as transmitted vertically during birth (1, 2). Infected deer, including white-tailed deer (Odocoileus virginianus), mule deer (O. hemionus), elk (Cervus canadensis), and moose (Alces alces) exhibit symptoms of wasting, display behavioral changes such as decreased social interaction, loss of awareness, and loss of fear for humans, and exhibit polydipsia (increased drinking), polyuria (increased urination), and hypersalivation (excessive salivation) (3). Infection will lead to premature death, which leads to concerns for conservation as well as loss of revenue to rural economies that rely on big game hunting (4, 5). Although currently there is no evidence, there is additionally a looming concern that CWD spillover to humans could occur in the future (6, 7) as venison consumption is common among hunters and their families, which may result in human suffering similar to that seen with mad cow disease outbreak in the 1990s. Such negative perception will lead to a decline in big game hunting, which will have a serious economic impact and consequences for deer conservation and management. There are no vaccines or effective preventive options available for managing this disease among free-ranging or captive cervids at the present time.

The geographic distribution of CWD-positive cervids has steadily increased over time. First noted in Colorado in 1967 at a research captive deer facility (1), CWD-positive captive and free-ranging cervids have since been detected in at least 35 states in the US and 5 provinces in Canada, mainly in the Great Plains region (8). Independent of the North American spatial distribution, the disease has also been reported among cervids in Norway, Sweden, and Finland, with an unclear origin, and in captive cervids in South Korea, which was imported from Canada (9, 10).

The past two decades have revealed a number of factors that influence the persistence of CWD infection among cervids, particularly white-tailed deer, and these factors could potentially promote spatial spread (11–13); although, the ecology and epizootiology of this disease remains yet to be fully understood. In Kansas for instance, the focus area for the present study, CWD was first noted in the year 2005 in Cheyenne county, the northwestern most county in the state, and every successive year since then the disease has been detected among white-tailed deer and mule deer from additional counties during annual surveillance. The current potential spatial distribution extent of the disease in the state, and any of its potential drivers are not known. Predicting the spatial distribution and expansion, and determining what factors, if any, in the environment could be contributing to the geographic spread of diseases, is important for planning appropriate management and surveillance strategies and for informing policy makers and public.

As it is typical for many wildlife diseases, much of the information we currently have on CWD in Kansas is based on opportunistic surveillance efforts, which rely on samples submitted for CWD testing by volunteers, viz., hunters, taxidermists, meat processors, and wildlife biologists. Conducting spatial pattern analyses and subsequent interpretations based on such data is problematic because they are likely to be a biased representation of underlying disease spatiotemporal prevalence. Novel methods have been developed over the years that addresses some of these concerns. For a disease like CWD with a highly heterogeneous prevalence across different regions in Kansas, for which surveillance data is sparsely available and collected non-uniformly over time and space, it is ideal to use occupancy methods that are flexible enough to accommodate for spatiotemporal heterogeneities in sampling and do not rely on host ecological data such as density, distribution, or population dynamics (14, 15).

Free-ranging and farmed ungulates in N. America are highly susceptible to CWD, and since its first detection the disease has numerically increased over the years and has spread relatively rapidly to newer geographic areas (8). Despite control efforts, the disease has proven difficult to contain; possibly due to the many different ways by which it is transmitted, environmental reservoirs, including soil and potentially plant materials in which the prions can remain viable for several years (16, 17). Additionally, the regulatory factors that govern big game hunting and stakeholder, landowners’ preferences influence executable management options (18, 19). The steady, almost exponential growth in white-tailed deer populations over the past decades (20), and potentially varying levels of their genetic susceptibility (21, 22) and cervid social behavior (23, 24) could also be contributing factors. Fundamental to proper management of wildlife diseases is to understand how far the disease has spread geographically, what/if there are any discernable distributional patterns, and the potential factors that contribute to such patterns. In this effort, we evaluated these aspects for CWD in Kansas for the first-time using surveillance data collected since 2005.

In this study, we used a dynamic patch-occupancy model in a Bayesian framework to quantify the spatial distribution of CWD in Kansas. Additionally, we utilized elevation, and environmental covariates derived from the land cover/land use and soil survey in order to estimate which of these factors play a role in the spatial distribution.

From 2005 to 2020, there were 18,421 samples collected for CWD surveillance in Kansas. There were 418 that were CWD positive, 17,861 that were CWD negative, and 142 that were indeterminant. There was a low of zero positive samples in 2006 to a high of 114 positive samples in 2019. Samples were from five species (Table 2). The majority (88%) of samples were from white-tailed deer (Odocoileus virginianus), 10% were from mule deer (O. hemionus), 1% from elk (Cervus canadensis), and cumulatively less than 1% of samples were from fallow deer (Dama dama), sika deer (Cervus nippon), or the species was unknown (Table 2).

The spatiotemporal occupancy patterns of CWD-infected deer in Kansas (Figure 2) illustrate the geographic expansion of CWD in the state, progressing from the northwest corner toward the central and eastern areas over the years. This pattern largely reflects a natural spatial progression of CWD in the state due to deer movement, rather than a pattern of accidental introductions over time into new areas aided by hunters or other agents. If the latter were the case, it would have likely resulted in a non-uniform spatial pattern with prominent, heterogeneous spatial foci of higher occupancies. The initial infections in Kansas likely originated from deer that migrated from the western states of Colorado and Wyoming, and Nebraska to the north, where the disease was established a number of years prior (44, 45). The number of positive CWD cases in the central and eastern side of the state remains low at the present time, despite higher prevalence of CWD infected deer in Missouri to the east [e.g., (46)]. Prevention and/or management strategies such as targeted culling when new cases occur in the central and eastern region Kansas may prevent the further spread of CWD in the state.

Areas in the state with higher soil pH values (alkaline) are associated with higher CWD occupancy. Additionally, CWD occupancy changed non-linearly with increasing clay content in the soil, peaking slightly above 20% and gradually declining beyond that level, with an optimal range between 18 and 25% by weight. These observations are similar to those made in previous studies [e.g., (28, 29, 47)]. Dorak et al. (28) found that the potential exposure to prions in Illinois soils increases when the soil pH is above 6.6 and the clay content is below 18%, which aligns with our current findings. The similarity is remarkable, given the differences between the two studies in terms of modeling approaches and soil environmental datasets. We found lower depths to restrictive layer in the soil were associated with higher CWD occupancy, suggesting that the areas with soils that have shallower distance to claypan, or bedrock are potentially more suitable for prion persistence and subsequent transmission to deer. This aligns with our observation that regions with soils possessing higher available water storage (AWS) capacity also exhibit higher CWD occupancy. Soils with shallow restrictive layers generally have lower AWS capacity. Higher bulk density indicates more compacted soil with less pore space, which can affect prion persistence, bioavailability, and movement assisted by wind and water. We observed that CWD occupancy is non-linearly associated with bulk density over a narrow range of values between 1.2 and 1.6 g/cm³. Areas with lower bulk density values favor higher CWD occupancy, but this occupancy quickly declines over a short range before increasing again.

The heterogeneities in soil physiochemical properties that affect CWD occupancy are an indication of landscape-level factors that can be targeted for management. Understandably, it is impracticable nor advised to amend soil properties on such large scales to prevent CWD transmission. However, this understanding allows us to focus management strategies on 20 km² areas, or contiguous groups of 20 km² areas, which are likely to have higher CWD occupancy and also possess high-risk soil properties. One previously proposed approach is targeted culling in areas with higher deer density and riskier soil physiochemical properties (28, 29, 48). However, the practice of targeted or local culling must be continuously evaluated for its effectiveness in reducing CWD infection among cervids (49, 50). Carcass removal from the landscape in areas where CWD is present and soil properties are high risk, and restricting feeding or baiting in these areas to minimize deer congregation and thereby concentrated prion environmental deposition have also been proposed as management actions (cite). In naïve, uninvaded high-risk areas, wild and captive cervid movement could be restricted along with restrictions placed on hunting activity and carcass disposal, among others, which may prevent the dispersal and invasion of CWD prions. The diagnostic detection of CWD prions in environmental samples has become more accessible and standardized with the recent adaptation of the RT-QuIC assay. This advancement enables the establishment of monitoring programs to track CWD prion levels across different management zones and monitor prion movement, for instance through watersheds or other catchment areas.

The environmental correlations observed in this study may represent important relationships with CWD occupancy in the broader region; however, it should be noted that some of these associations might be spurious and/or related with the happenstance of the invasion. For instance, we observed that higher elevations were related to higher CWD occupancy, but this is likely due in large part to the fact that there is a higher concentration of cases in the relatively high-altitude western portion of Kansas where initial invasion likely came from Colorado and Wyoming. Infected cervids are known to be present in lower elevation area as in the case of their wider presence in Illinois and Missouri, in relatively flat terrains. Therefore, elevation at the scale of our analysis may not be a good predictor of future CWD occurrence within Kansas, but more of an explanation of where it came from. Similarly, mixed forest cover was the only habitat-specific covariate that was suggested to be important for CWD occupancy in Kansas. More mixed forest cover was less associated with CWD occupancy. However, the percent of mixed forest cover only ranged from 0 to 3% cover within the 20 km² grids we considered across the state. This may be another example of instances where the happenstance of the invasion is suggesting a relationship that is not necessarily a driver of CWD occurrence in Kansas. Future studies evaluating the association of environmental covariates with CWD, and other wildlife diseases may benefit from using derived remote sensing variables of factors that affect land cover land use change [e.g., (51–53)] in addition to the crude estimates as used in the present study.

The 20 km² spatial units for analysis in this study is somewhat artificial and it is not the scale in which deer are managed in the state of Kansas—which are several kilometers wide/deep, and they are based on road networks for borders (54). Although strictly not reflective of the home ranges of white-tailed-and mule deer in the region, which varies by season, region, and sex, and can range from 1 to 5 km² in Texas (55–57) the spatial unit we used for modeling the occupancy status allowed us to potentially account for uncertainties associated with harvest locations reported by sample submitters, as well as deer movement during the hunting season. Compared to the current deer management units, the finer, more uniformly sized 20 km² spatial units we used in the present study are potentially more suitable when it comes to deciding, implementing CWD management strategies, such as localized and targeted culling. This is particularly valuable in the newly invaded areas in the central and eastern sides of the state, which may not be predictable or visible at coarser scales of spatial analysis. Also, in the present work, we did not allow detection probability to vary, however, it would be possible with a modification to our current approach to estimate detection probabilities separately, for instance if differences are expected for different sources of surveillance or diagnostic methods, which may help inform management and/or surveillance decisions [e.g., (58)].

The probability of detecting CWD in our study exceeded our expectations (average detection probability given CWD was present was 0.22), given the inherent challenges of modeling spatiotemporal patterns for wildlife diseases (58, 59), and it was also higher than that reported in similar studies. For instance, our own studies assessing wildlife rabies in North America, which incorporated various landscape-level covariates similar to the present study, typically reported detection rates ranging from 0.02 to 0.15 at the higher end (58–60), compared to up to 0.83 in the present study. The standard errors noted for occupancies in the present study are low and did not raise concerns. Further, the AUC values for all the models considered in the study were within 15% of each other and close to 0.9, suggesting an excellent model fit (Table 3). This indicates that the random effect terms and the inclusion of fixed environmental covariates in the study adequately captured the variability in CWD data.

One of the epizootiological aspects of cervid CWD that is not yet fully understood is the potential prevalence of CWD prions in various scavenger and predator species, along with their exact role in the disease’s spread (61, 62). The soil properties observed in this study that influence CWD occupancy also indirectly affect various ecological parameters, including nutrient availability, plant growth rate, and vegetation density. These factors, in turn, influence the habitat quality and suitability for different scavenger species, affecting their composition, diversity, density, and movement patterns across the landscape. Together, these factors potentially impact CWD dispersal in the environment. Future studies that evaluate the role of scavengers in CWD epizootiology will benefit by considering the underlying landscape-level soil characteristics. For instance, spatially informed sampling of scavenger species may be conducted in areas where certain soil conditions indicate potential for higher CWD prion levels. In conclusion, this study further shows that occupancy class models are a valuable tool in wildlife disease spatiotemporal modeling and for evaluating environmental drivers of such diseases. The spatiotemporal patterns and environmental factors identified in the present study will be useful in managing this disease in Kansas and potentially other geographic regions.

Some potential defects in our models and the resulting occupancy predictions must be mentioned. The choice of spatial extent (20 km² grid) for analysis could impact our predictions; too coarse of a scale would result in overestimation of occupancy, whereas too fine a resolution could have the opposite effect. We also assumed that a site’s occupancy status remains constant over the annual time step, which may be unrealistic, particularly in areas with fewer CWD-positive deer harvested, especially since surveillance was limited to the deer hunting season, lasting approximately 7–8 weeks each year. We also assume that our surveillance samples all have equal probability of detection, which may or may not be the case and needs a second assessment with different model specifications.