Search tips
Search criteria 


Logo of biolettersThe Royal Society PublishingBiology LettersAboutBrowse By SubjectAlertsFree Trial
Biol Lett. 2009 June 23; 5(3): 387–390.
Published online 2009 March 25. doi:  10.1098/rsbl.2009.0094
PMCID: PMC2679935

Landscape modelling spatial bottlenecks: implications for raccoon rabies disease spread


A landscape genetic simulation modelling approach is used to understand factors affecting raccoon rabies disease spread in southern Ontario, Canada. Using the Ontario Rabies Model, we test the hypothesis that landscape configuration (shape of available habitat) affects dispersal, as indicated by genetic structuring. We simulated range expansions of raccoons from New York into vacant landscapes in Ontario, in two areas that differed by the presence or absence of a landscape constriction. Our results provide theoretical evidence that landscape constriction acts as a vicariant bottleneck. We discuss implications for raccoon rabies spread.

Keywords: raccoon rabies, landscape genetics, infectious disease, modelling, Procyon lotor

1. Introduction

Molecular ecology and landscape genetics offer innovative approaches to the field of disease ecology (Archie et al. 2009). Application of these methodologies promises to improve our understanding of disease spread by characterizing factors affecting movement and connectivity among individuals in the vector populations. We apply a model-based landscape genetic approach to explore factors affecting raccoon rabies spread in the Great Lakes region of North America (figure 1). Raccoons (Procyon lotor) are the primary vector of this rabies variant (Winkler & Jenkins 1991). Southern Ontario is at high risk for rabies because it is adjacent to endemic areas in New York (NY) and Quebec. For the past decade, the shortest geographical distance for raccoon movement with disease spread into Ontario has been from the Niagara (NIA) region in NY; however, rabies has never been detected in NIA, Ontario. The first Ontario cases were detected in 1999 in the St Lawrence region (STL; Wandeler & Salsberg 1999). It is possible that the Niagara River acts as a greater barrier to raccoon movement in NIA than the St Lawrence River in STL (Rees et al. 2008a), thus, keeping the number of infected immigrants into NIA, Ontario below a threshold for which rabies is detected or becomes epizootic; however, the shape of available habitat (landscape configuration) can also affect the number of animals moving among regions (McRae & Beier 2007).

Figure 1
(a) Location of ORM landscapes in eastern North America. (b) NIA landscape and three sampling regions (A, NIAdestination; B, NIAconstriction; C, NIAorigin) for allelic diversity measures. Cell shade represents membership of one raccoon sampled for each ...

Genetic analysis of raccoons along the Ontario–NY border (Cullingham et al. 2009) revealed the presence of genetically distinct raccoon populations on either side of the Niagara River, but only one population spanning the St Lawrence River. Both regions have similar habitats and experience comparable rabies control efforts. In view of these factors, it appears that there is a greater resistance to cross-border movements in the NIA region, which we suspect is caused by the narrowing of land between NY and Ontario as bounded by lakes Ontario and Erie. Hence, we test the hypothesis that landscape configuration in NIA constricts movement of raccoons between NY and Ontario. We predict that the peninsular shape acts as a spatial bottleneck and creates two genetically distinct populations. We use the Ontario Rabies Model (ORM), an individual-based, spatially explicit stochastic genetic simulation model (see the electronic supplementary material) to simulate raccoon colonization from NY into Ontario separately in NIA and the STL, and analyse their genetic structures through time. The STL landscape acts as our control because it has no landscape constriction. Our approach (i) characterizes landscape constriction effects on gene flow and (ii) provides insight into raccoon rabies spread.

2. Material and methods

The ORM simulates raccoon population dynamics (e.g. reproduction, mortality, dispersal) of individuals residing in a virtual landscape of contiguous hexagonal cells of 10.23 km2, the approximate activity range of raccoons in Ontario (Rosatte 2000; see the electronic supplementary material). The model operates at weekly time steps (e.g. mating occurs at week 9 (end of February) and parturition occurs at week 18 (end of April)). ORM events are stochastically determined by randomly drawing a value from parameter distribution functions (Rees et al. 2008b). We used the ORM to simulate bi-parental genetic inheritance of microsatellite markers assuming no new alleles arise through mutation or recombination, since this is unlikely to be significant over the simulation period (Avise 2004). We used neutral genetic markers because they are not subject to selective pressures, thus distribute as a result of mating and dispersal processes, acting as a ‘tag’ to identify spatio-temporal patterns arising from these processes.

To test landscape configuration effects on gene flow, we modelled raccoon range expansions in NIA (43° N, 79° W) and STL (44° N, 75° W) (figure 1). The NIA study area is composed of 2255 cells representing approximately 23 070 km2 on either side of the Niagara River. STL, the control landscape, is a rectangle of 2500 cells, approximately 25 575 km2. We did not simulate the rivers in either landscape, to prevent confounding the effects of rivers and landscape constrictions on influencing dispersal movements.

In separate NIA and STL model simulations, raccoons expanded from NY to colonize ‘uninhabited’ Ontario. We expected genetic homogenization to occur more slowly in NIA because of the landscape constriction. Consequently, we ran the NIA colonization process for 2200 years and only 1000 years in STL. The NY ‘origin’ populations were tagged at time 0 with microsatellite markers at field-observed frequencies for which linkage and Hardy–Weinberg equilibrium had been confirmed (Cullingham et al. 2009). For NIA and STL, five simulations were run to capture variation from model stochasticity.

One raccoon was randomly sampled per cell from the entire landscape in NIA, subclassified as three regions (NIAorigin, NIAconstriction and NIAdestination). In the STL, we reduced analysis time and still maintaining sufficient data for objective analysis by sampling two raccoons every two to three cells across the STL landscape. This landscape was subclassified as STLorigin and STLdestination (figure 1). This sampling scheme enabled us to assess population structure over the length of the landscapes, and compare that with the structure within the constriction. We used Structure v. 2.2 (Pritchard et al. 2000) to determine the number of genetic clusters (K=1–5) 250 years after colonization (K>5 is unlikely; Cullingham et al. 2009), the approximate time period during which raccoons have colonized NIA, Ontario (Trigger et al. 2000). We continued to test for K=2 in NIA at 50–200 year intervals up to 2200 years, and checked in the STL every 100 years up to 1000 years, after the start of colonization. Structure was run using a conservative admixture model that assumed correlated allele frequencies. Three independent tests were conducted for each K, with 500 000 Markov Chain Monte Carlo cycles each for burn-in and data collection. Individuals were assigned to a cluster if average probability of ownership was 0.8 or more. We determined the most likely number of genetic population clusters using the Evanno algorithm (Evanno et al. 2005) and calculated the number of individuals assigned to each cluster. From the same sample of raccoons, we calculated mean number of alleles (allelic richness) for the aforementioned regions in NIA and STL, approximately every 100 years, up to 1000 years since colonization, using Cervus v. 3.0.3 (Kalinowski et al. 2007).

3. Results

Despite not modelling the Niagara River barrier effect, the Evanno algorithm identifies two distinct genetic clusters on either side of the Niagara River as being the most likely population structure after 250 years of range expansion (table 1), where the genetic boundary occurs at least 30 km west of the Niagara River (figure 2). Also in NIA, genetic homogenization develops in scenarios with two genetic clusters after 2000 years. At this point, there is an equal likelihood of individuals being unassigned or assigned to one of two clusters (table 2).

Figure 2
Allelic richness for NIA and STL landscapes at approximately 100-year intervals over 1000 years of colonization. Allelic richness is calculated from raccoons sampled for regions: NIAorigin (n=1491, filled circles); NIAconstriction (n=278, filled triangles); ...
Table 1
Structure results from sampling the NIA landscape 250 years after colonization: number of individuals unassigned (U) and assigned (C) to clusters with a probability of ownership of 0.8 more. (K=2 is most likely because the mean of the estimated logarithms ...
Table 2
Structure results from testing for K=2 by analysing 2255 raccoons in NIA and 560 raccoons in STL that were evenly sampled across both model landscapes. (Estimated logarithms of probability of data (est ln prob) are shown, as well as the number of individuals ...

Among all regions, allelic richness in both source populations (NIAorigin and STLorigin) are most similar to each other and similarly decreases over time at a comparable rate (figure 2). The NIAconstriction is the only other region that experiences a monotonic decline in allelic richness. In the founded populations, NIAdestination, allelic richness increases from 75 to 150 years and then declines from 250 years. This trend is similar for STLdestination, but diversity in this region is greater and does not decrease until 800 years.

4. Discussion

We undertook a landscape genetic approach to investigate NIA and STL population structures because we hypothesized that factors influencing their structures are also affecting rabies disease spread. By simulating colonization events in these landscapes, in the absence of a river barrier effect, we can assess how landscape configuration affects population genetic structure over time. In our simulations, K=2 is the most likely genetic structure in NIA, and it persists for at least 1000 years; whereas in STL, only one genetic cluster is evident through time. The NIA structure arises from the newly founded population being dominated by the genetic material of the colonizers. Expanding populations are typically lower in density along their periphery, facilitating genetic drift to fix or lose alleles, and thus increasing the likelihood of genetic differentiation from the source population (Excoffier & Ray 2008). The landscape constriction in NIA is literally a spatial bottleneck accentuating founder effects. For both NIAdestination and STLdestination, allelic richness initially increased as colonizers filled the region; however, the landscape constriction reduced the rate at which colonizers entered NIAdestination. Consequently, colonizers had a smaller influence on the NIAdestination population over time; with fewer colonizers adding to genetic diversity, genetic drift would have a stronger relative influence on the population, as indicated by earlier declines in allelic richness in NIAdestination than STLdestination.

Landscape configuration has been found to be a significant factor structuring populations in other theoretical and empirical systems (Biek et al. 2007; McRae & Beier 2007). We suggest that it affects rabies spread in NIA by reducing the rate at which infected raccoons enter Ontario below a level necessary to establish a rabies outbreak. We are not implying that the Niagara River or rabies control strategies have no effect on preventing disease spread into NIA, Ontario; however, the Canada–USA landscape constriction in NIA is an additional factor keeping NIA Ontario disease-free.

In simulation modelling, the quality of model output depends on appropriate representation of system processes. ORM has been validated (Rees 2008; Rees et al. 2008a), so we are confident of its ability to represent raccoon ecology and genetics. Our approach demonstrates the value of using landscape genetics to understand disease spread. The flexible structure of the ORM enables testing of factors hypothesized to affect animal movements, gene flow, and hence, the spread of infectious disease, from which we conclude that the narrowing of land in NIA is an important factor reducing the risk of raccoon rabies spread into NIA, Ontario.


We are grateful to the Rabies Research and Development Unit of the Ontario Ministry of Natural Resources (OMNR). This research has been supported by a strategic grant from the Natural Sciences and Engineering Research Council of Canada to B.N.W., Trent University and by a collaborative research agreement between the OMNR and Queen's University GIS Laboratory, Canada.

Supplementary Material

The Ontario rabies model:

The Ontario rabies model documentation: model structure, processes and assumptions


  • Archie E.A., Luikart G., Ezenwa V.O. Infecting epidemiology with genetics: a new frontier in disease ecology. Trends Ecol. Evol. 2009;24:21–30. doi:10.1016/j.tree.2008.08.008 [PubMed]
  • Avise J.A. 2nd edn. Sinauer Associates, Inc; Sunderland, MA: 2004. Molecular markers, natural history, and evolution.
  • Biek R., Henderson C.J., Waller L.A., Rupprecht C.E., Real L.A. A high-resolution genetic signature of demographic and spatial expansion in epizootic rabies virus. Proc. Natl Acad. Sci. USA. 2007;104:7993–7998. doi:10.1073/pnas.0700741104 [PubMed]
  • Cullingham C.I., Kyle C.J., Pond B.A., Rees E.E., White B.N. Differential permeability of rivers to raccoon gene flow corresponds to rabies incidence in Ontario, Canada. Mol. Ecol. 2009;18:43–53. doi:10.1111/j.1365-294X.2008.03989.x [PubMed]
  • Evanno G., Regnaut S., Goudet J. Detecting the number of clusters of individuals using the software Structure: a simulation study. Mol. Ecol. 2005;14:2611–2620. doi:10.1111/j.1365-294X.2005.02553.x [PubMed]
  • Excoffier L., Ray N. Surfing during population expansions promotes genetic revolutions and structuration. Trends Ecol. Evol. 2008;23:347–351. doi:10.1016/j.tree.2008.04.004 [PubMed]
  • Kalinowski S.T., Taper M.L., Marshall T.C. Revising how the computer program Cervus accommodates genotyping error increases success in paternity assignment. Mol. Ecol. 2007;16:1099–1106. doi:10.1111/j.1365-294X.2007.03089.x [PubMed]
  • McRae B.H., Beier P. Circuit theory predicts gene flow in plant and animal populations. Proc. Natl Acad. Sci. USA. 2007;104:19 885–19 890. doi:10.1073/pnas.0706568104 [PubMed]
  • Pritchard J.K., Stephens M., Donnelly P. Inference of population structure using multilocus genotype data. Genetics. 2000;155:945–959. [PubMed]
  • Rees, E. E. 2008 Approaches to modelling raccoon rabies. PhD thesis, Trent University, Canada.
  • Rees E.E., Pond B.A., Cullingham C.I., Tinline R.R., Ball D., Kyle C.J., White B.N. Assessing a landscape barrier using genetic simulation modelling: implications for raccoon rabies management. Prev. Vet. Med. 2008a;86:107–123. doi:10.1016/j.prevetmed.2008.03.007 [PubMed]
  • Rees E.E., Pond B.A., Phillips J.R., Murray D.L. Raccoon ecology database: a resource for population dynamics modelling and meta-analysis. Ecol. Inform. 2008b;3:87–96. doi:10.1016/j.ecoinf.2008.01.002
  • Rosatte R.C. Management of raccoons (Procyon lotor) in Ontario, Canada: do human intervention and disease have significant impact on raccoon populations? Mammalia. 2000;64:369–390.
  • Trigger B.G., Washburn W.E., Salomon F., Adams R.E.W., Schwartz S.B., MacLeod M.J. Cambridge University Press; Cambridge, UK: 2000. The Cambridge history of the native peoples of the Americas.
  • Wandeler A., Salsberg E. Raccoon rabies in eastern Ontario. Can. Vet. J. 1999;40:731–731. [PMC free article] [PubMed]
  • Winkler, W. & Jenkins, S. 1991 Raccoon rabies. In The natural history of rabies, 2nd edn (ed. G. M. Baer), pp. 240–325. Boca Raton, FL: CRC Press.

Articles from Biology Letters are provided here courtesy of The Royal Society