Ecological Archives M084-012-A3
J. Grant C. Hopcraft, J. M. Morales, H. L. Beyer, Markus Borner, Ephraim Mwangomo, A. R. E. Sinclair, Han Olff, Daniel T. Haydon. 2014. Competition, predation, and migration: individual choice patterns of Serengeti migrants captured by hierarchical models. Ecological Monographs 84:355–372. http://dx.doi.org/10.1890/13-1446.1
Appendix C. Additional material and methods describing how grass nitrogen, NDVI, grass biomass, the risk of predation, access to water, and exposure to anthropogenic risk were measured.
The quality of the forage was estimated from grass nitrogen measured at 148 sites (Fig. 1) distributed in the different soil and vegetation types in Serengeti and across the rainfall gradient. The aboveground grass biomass was clipped in five 25 × 25 cm plots at each of the 148 sites (pooling all grass species) and air dried before being ground in a cyclonic grinder (Foss Cyclotec 1093) with a sieve size of 2 mm. Ground samples were further oven dried for 48 hours at 72°C and nitrogen concentrations were measured using a Near Infra-Red (NIR) spectrophotometer (Bruker MPA NIR). The NIR estimation of grass nitrogen concentrations proved to be a fast and accurate technique. The correlation between the 'true' nitrogen concentration as established from a CHNS elemental analyzer (Carlo-Erba Instruments EA1110 CHNS) and the predicted NIR nitrogen concentrations of a sub-set of samples suggest the NIR technique is highly accurate (r² = 0.97, n = 76).
The spatial distribution of grass nitrogen for the ecosystem was interpolated by regression kriging. Regression kriging (Hengl et al. 2007, Bivand et al. 2008) is an interpolation technique that takes into account the spatial autocorrelation between sampling points (as per ordinary kriging), while simultaneously accounting for the correlation between samples and an underlying predictor variable. Since grass nitrogen is inversely correlated with NDVI (r² = 0.10, slope = -0.0001, p < 0.001, n = 148), we used the 9-year mean NDVI index (2000–2009) as a covariate to more accurately predict the spatial distribution of grass nitrogen. As a separate internal accuracy check, 30 samples were selected randomly and used to test the accuracy of a second regression kriged grass nitrogen map created from the remaining 118 samples. The predicted grass nitrogen values from the second map correlated well with the 30 random samples (r² = 0.25, slope = 0.48, p < 0.01, n = 30), and verify this technique as having an acceptable level of accuracy.
Normalized Difference Vegetation Index (NDVI) is commonly used as a measure of the vegetation greenness; NDVI measures approaching 1 indicate live green vegetation while those approaching -1 suggest dry or dead vegetation. Sixteen-day NDVI composites at resolution of 250 m were downloaded from the MODIS Terra satellite for the period from 2000 to 2009 using NASA's Warehouse Inventory Search Tool website (WIST, https://wist.echo.nasa.gov/). NDVI was used as a predictor of the movement patterns of wildebeest and zebra by (a) estimating the amount of green vegetation available to grazers, and (b) to estimate the rate at which the grasses are greening or drying (dNDVI). Grasses tend to flush and senesce rapidly with rainfall and evapotranspiration, therefore large differences between current and previous NDVI scenes suggest patches are either rapidly drying or greening (positive or negative values, respectively). Conversely, small differences between the current and previous 16-day NDVI composite suggest the area is stable and neither drying nor greening.
Figure C1 illustrates the relationships between grass nitrogen, the mean annual NDVI from 2000 to 2009, and the standard deviation of NDVI from 2000 to 2009. Grass nitrogen correlates somewhat linearly with both NDVI and dNDVI, however the relationship between NDVI and dNDVI is bell-shaped. In other words, areas that are very wet or very dry all year round (i.e., areas at either extreme of the NDVI index) do not tend to vary throughout the year, whereas intermediate areas switch between a distinct green and brown phases. Therefore including both NDVI and dNDVI is warranted as they are capturing different attributes of grass quality.
Fig. C1. The relationships between grass nitrogen, mean annual NDVI, and the annual standard deviation of NDVI (calculated from 2000 to 2009), where the red lines are a spline smoother and the numbers are the linear correlation coefficient.
Grass biomass is positively correlated with rainfall and soil moisture. Under mesic conditions grasses grow tall and tend to become lignified which decreases their digestibility and their nutritional value to herbivores (Breman and De Wit 1983, McNaughton et al. 1985, Olff et al. 2002). Additionally, the species composition of grasses shifts across the mesic-arid gradient such that mesic areas tend to be dominated by lower quality species (Anderson et al. 2007). Therefore, we used rainfall and the topographic wetness index (TWI) as a proxy for grass biomass and not grass quality.
The long-term average rainfall was interpolated by regression kriging monthly rainfall records from 58 gauges for 46 years across a known southeast to northwest diagonal rainfall gradient (unpublished rainfall data the from Serengeti Ecology Department from 1960 to 2006). Rain gauges with less than three years of data were omitted.
The landscape's capacity to retain water was estimated using the topographic wetness index (TWI) which simultaneously accounts for the total water catchment area and the slope of a cell. Cells with higher TWI values tend to be flat or concave and have large catchment areas.
Since high biomass occurs on wet sites which could either be from rainfall or from catchment, we added the rainfall layer to the TWI to create a map of potential grass biomass. Highest values occur in concave landscapes with large catchment areas and high amounts of rain. Areas that are convex and do not have large catchment areas, or have low rainfall, have the lowest values.
Risk of predation
Certain landscape features such as dense woodland, embankments, predator viewsheds, river confluences, and slope facilitate predators in capturing their prey by concealing them, forming natural traps, or increasing the predator's ability to detect prey (Hebblewhite et al. 2005, Hopcraft et al. 2005, Balme et al. 2007, Kauffman et al. 2007). We used the distance to thick woody cover and the distance to drainage beds to estimate the exposure of wildebeest and zebra to predation risk (Valeix et al. 2009, Hopcraft et al. in press-a).
Woody vegetation increases the probability of lions capturing prey by providing cover required by lions to stalk their prey (Hopcraft et al. 2005). Estimates of the amount of woody cover available to predators were calculated from 1,882 samples conducted along transects over the entire ecosystem (Fig. 1). At each sampling point the mean percent woody cover greater than 0.4 m high was calculated from four equidistant measures at a radius of 15 meters (based on minimum cover requirements for lions (Elliott et al. 1977, Scheel 1993)). Sampling points were assigned to one of the 27 physiognomic vegetation classes identified by Reed (Reed et al. 2008). The mean woody cover for each vegetation class was calculated. This mean woody cover per vegetation class was then assigned to all cells of that class (cell size is 90 m (Reed et al. 2008)). The amount of horizontal woody cover available for ambush predators in a 1-km cell was calculated by averaging the mean woody cover values for all 90 m cells occurring within the 1-km block. The distance to the thickest woody cover (i.e., cells in the upper 85th percentile) was calculated in ArcGIS.
Drainage beds are often associated with erosion embankments and river confluences which conceal predators and funnel prey into natural traps. Most drainages are ephemeral freshets and do not hold water. Drainage beds include all categories of rivers in the Rivers Version 3 of the Serengeti GIS Database (www.serengetidata.org) from class 4 rivers which are swales and seepage lines with no defined bank, to class 1 rivers that flow all year round and have well defined and incised banks. We calculated the distance of each animal at every GPS location to drainages in Serengeti.
Access to free water was estimated by calculating the distance to any river that either flows continually or contains pools year-round (i.e., classes 1 and 2 of Rivers Version 3 in the Serengeti Database www.serengetidata.org). The majority of drainages in the Serengeti are ephemeral freshets and contain water only for a few weeks during the wet season (class 3 and 4), and were not considered.
Risks imposed by humans, such as illegal hunting, were estimated by calculating the distance to villages and weighting this by the estimated density of the human population.
(Log [Human Population Density +1]) / (Distance to Village (km))
Therefore, locations near densely populated villages have the highest values while locations furthest from small villages have the lowest values. Village location data comes from the Serengeti Database (www.serengetidata.org) and is the Village Version 2 shapefile. Human density data for the year 2000 were acquired from the FAO AfriCover project (www.africover.org). Human density was log transformed as values were log-normally distributed; most cells had low density estimates but a few had very high density estimates.
Anderson, T. M., M. E. Ritchie, E. Mayemba, S. Eby, J. B. Grace, and S. J. McNaughton. 2007. Forage nutritive quality in the Serengeti ecosystem: the roles of fire and herbivory. American Naturalist 170:343–357.
Balme, G., L. Hunter, and R. Slotow. 2007. Feeding habitat selection by hunting leopards Panthera pardus in a woodland savanna: prey catchability versus abundance. Animal behaviour 74:589–598.
Bivand, R. S., E. J. Pebesma, and V. Gómez-Rubio. 2008. Applied spatial data analysis with R. Springer.
Breman, H., and C. T. De Wit. 1983. Rangeland productivity and exploitation in the Sahel. Science 221:1341.
Elliott, J. P., I. M. Cowan, and C. S. Holling. 1977. Prey capture by the African lion. Canadian Journal of Zoology 55:1811–1828.
Hebblewhite, M., E. H. Merrill, and T. L. McDonald. 2005. Spatial decomposition of predation risk using resource selection functions: an example in a wolf-elk predator-prey system. Oikos 111:101–111.
Hengl, T., G. B. M. Heuvelink, and D. G. Rossiter. 2007. About regression-kriging: From equations to case studies. Computers & Geosciences 33:1301–1315.
Hopcraft, J. G. C., T. M. Anderson, S. Perez Vila, E. Mayemba, and H. Olff. in press-a. Body size and the division of niche space: Food and predation differentially shape the distribution of Serengeti grazers. Journal of Animal Ecology.
Hopcraft, J. G. C., A. R. E. Sinclair, and C. Packer. 2005. Planning for success: Serengeti lions seek prey accessibility rather than abundance. Journal of Animal Ecology 74:559–566.
Kauffman, M. J., N. Varley, D. W. Smith, D. R. Stahler, D. R. MacNulty, and M. S. Boyce. 2007. Landscape heterogeneity shapes predation in a newly restored predator-prey system. Ecology Letters 10:690–700.
McNaughton, S. J., J. L. Tarrants, M. M. McNaughton, and R. H. Davis. 1985. Silica as a defense against herbivory and a growth promotor in African grasses. Ecology 66:528–535.
Olff, H., M. E. Ritchie, and H. H. T. Prins. 2002. Global environmental controls of diversity in large herbivores. Nature 415:901–904.
Reed, D., T. M. Anderson, J. Dempewolf, K. Metzger, and S. Serneels. 2008. The spatial distribution of vegetation types in the Serengeti ecosystem: the influence of rainfall and topographic relief on vegetation patch characteristics. Journal of Biogeography 36:770–782.
Scheel, D. 1993. Watching for lions in the grass: The usefulness of scanning and its effects during hunts. Animal behaviour 46:695–704.
Valeix, M., A. J. Loveridge, S. Chamaille-Jammes, Z. Davidson, F. Murindagomo, H. Fritz, and D. W. Macdonald. 2009. Behavioral adjustments of African herbivores to predation risk by lions: Spatiotemporal variations influence habitat use. Ecology 90:23–30.
[Back to M084-012]