Disease or drought: environmental fluctuations release zebra from a potential pathogen-triggered ecological trap

Abstract

When a transmission hotspot for an environmentally persistent pathogen establishes in otherwise high-quality habitat, the disease may exert a strong impact on a host population. However, fluctuating environmental conditions lead to heterogeneity in habitat quality and animal habitat preference, which may interrupt the overlap between selected and risky habitats. We evaluated spatio-temporal patterns in anthrax mortalities in a plains zebra (Equus quagga) population in Etosha National Park, Namibia, incorporating remote-sensing and host telemetry data. A higher proportion of anthrax mortalities of herbivores was detected in open habitats than in other habitat types. Resource selection functions showed that the zebra population shifted habitat selection in response to changes in rainfall and vegetation productivity. Average to high rainfall years supported larger anthrax outbreaks, with animals congregating in preferred open habitats, while a severe drought forced animals into otherwise less preferred habitats, leading to few anthrax mortalities. Thus, the timing of anthrax outbreaks was congruent with preference for open plains habitats and a corresponding increase in pathogen exposure. Given shifts in habitat preference, the overlap in high-quality habitat and high-risk habitat is intermittent, reducing the adverse consequences for the population.

Keywords: anthrax; disease dynamics; ecological trap; environmental transmission; habitat selection; transmission hotspot.

Figure 1.

The study area of Etosha National Park, Namibia, in southern Africa. (a) The distribution of habitat types. The four habitat types considered included open plains (light grey), shrublands (medium grey), woodlands (dark grey) and bare areas (large salt pans; tan). White represents areas with vegetation type not available, which were removed from this study. The blue circle indicates the location of Okaukuejo station. (b) Plains zebra (E. quagga) space use shows the overall thinned readings of zebra telemetry data (yellow) and their 99.9% kernel density range (blue) which were used to generate available (not used) points for analysis of resource selection functions. (c) The percentages of the available points randomly generated within 99.9% kernel density range by habitat type.

Figure 2.

Herbivore mortalities by habitat type in Etosha National Park. Mortalities from 1998 to April 2020 with GPS locations were assigned to habitat type (open plains, shrublands or woodlands) and cause of death (anthrax or other natural mortalities) for herbivore species which had at least one anthrax mortality. The mortalities are summed for (a) all herbivore species, and then grouped by three functional foraging types, (b) grazers, including plains zebra (67.6% of total cases), blue wildebeest (Connochaetes taurinus; 9.9%) and gemsbok (Oryx gazella; 1.1%), (c) mixed feeders including springbok (Antidorcas marsupialis; 12.3% of total cases) and African elephant (Loxodonta africana; 8.5%), and (d) browsers including greater kudu (Tragelaphus strepsiceros; 0.3% of total cases) and black rhino (Diceros bicornis; 0.4%). The numbers above bars indicate ratios of anthrax mortalities to other natural causes of death. The asterisks show the significance of χ²-tests comparing the proportions of anthrax mortalities between paired habitat types. Two and three asterisks represent the Bonferroni corrected p-values less than 0.01 and 0.001, respectively.

Figure 3.

Zebra anthrax mortalities (a,b) and vegetation productivity (c,d) in Etosha National Park, by habitat type and season in 2009–2010 (a,c) and 2018–2020 (b,d). Productivity was measured as the FPAR. Error bars indicate standard deviations of spatial variation. The axes for anthrax mortalities were square-root transformed. The study period 2009–2010 was an average rainfall year; 2018–2020, a severe drought followed by an above average rainfall year.

Figure 4.

Inter-annual and seasonal variation in zebra habitat selection in Etosha National Park. (a,b) The average percentages of zebra selected locations among habitat types by season (a) in 2009–2010 (an average rainfall year) and (b) in 2018–2020 (a drought year, 2019, followed by an above average rainfall year, 2020), with error bars representing standard deviations from individual differences. (c–h) The regression coefficients indicating zebra habitat selection accounting for habitat availability by season and year. The coefficients of three contrast coding variables from seasonal individual-based RSFs represented the relative selection strength (habitat preference) for vegetated habitats over bare areas (c) in 2009–2010 and (d) in 2018–2020, for open habitats over closed habitats (e) in 2009–2010 and (f) in 2018–2020, and for shrublands over woodlands (g) in 2009–2010 and (h) in 2018–2020. Each point is an individual zebra, with males in grey and females in yellow. In 2009–2010, all the collared zebras were female. Sample sizes of individual zebras per season ranged from 5 to 17. The boxplots in (c–h) are colour-coded with the seasonal average values of FPAR at open habitats reflecting vegetation productivity differences by season, with grey to purple to blue representing low to high FPAR.

Figure 5.

Seasonal relationships between habitat dynamics and zebra habitat selection. Medians of relative habitat selection strengths in relation to average FPAR (a remote-sensing index of vegetation productivity) at open habitats (a) for vegetated habitats over bare areas, (b) for open habitats over closed habitats and (c) for shrublands over woodlands. An error bar indicates interquartile range from individual variation in each season. Grey dashed lines are best-fitting lines when linear regressions showed significant slopes. Different years/seasons are colour-/shape-coded.

Figure 6.

Linear regressions between habitat dynamics, zebra habitat selection and zebra anthrax cases. Relationships are shown between the square-root transformed anthrax case numbers by season in relation to (a) average FPAR (a remote-sensing index of vegetation productivity) at open habitats, and (b) the median of relative selection strengths for open habitats over closed habitats. Error bars indicate interquartile ranges of relative selection strengths from individual variation. Grey dashed lines are best-fitting lines from linear regressions. Different years/seasons are colour-/shape-coded.