4.1 Abstract
On their own, hotspot and antipredator models were fo u n d to be insufficient in explaining the evolution o f topi leks. Contrary to prediction, a disproportionate number o f fem ales on leks was fou n d to be in oestrus. M oreover, males d id not sim ply map onto fem ale dispersion, as the sex ratio on leks was m ale-biased both outside and during the rut. The leks were not fou n d at hotspots o f exceptionally high fem ale density, nor d id they contain significant resources and fem ales spent an insignificant proportion o f time grazing there. Contrary to the prediction s o f the home range overlap model, the lek areas were not characterised by uniquely high overlap o f fem a le ranges and lek size increased rather than decreased with increasing population density; although more than one lek could be located within a fem ale home range as p red icted by the model, this result was fo u n d to be inconclusive. Thus in conclusion the home range overlap m odel was not supported. On leks fem ales were m ore likely to meet hyenas, their presum ed main predators; this held true also when encounter rate was determ ined on a p e r capita basis.
Chapter 4_____________________________Hotspots____________________________________53
f ; : " .X
Hyenas are probably the main predators of topi in the area
4.2 Introduction
One o f the major ideas about lek formation is that males position themselves on hotspots where female encounter rate is highest. Thus a range o f models can be grouped as 'hotspof models because they all share the assumption that males map onto a female dispersion that is unaffected by mating behaviour. Hotspot models have been developed over many years and numerous variations o f the theme exist, in fact also antipredator models can be seen as partly related to the hotspot idea. In this chapter I use data from topi to test a number o f key assumptions and predictions made by these models.
The original formulation o f the hotspot model, here called the sim ple hotspot model, predicts that leks form in the areas o f the highest female density, which is where resource availability is highest (Alexander, 1975; Kruijt et al., 1972). According to this idea, lek formation follows a similar logic as applies to other territorial systems where male dispersion is determined by female dispersion, which in turn is determined by resource distribution (Davies, 1991).
Assuming that individuals position themselves in an ideal free manner (Fretwell, 1972; Fretwell and Lucas, 1970), the sex ratio on and o ff lek is predicted to be similar. Furthermore, leks are expected to have higher availability o f the preferred forage. Topi select in favour o f green grass leaves rather than for total biomass or
Chapter 4______________________________H otspots_____________________________________ M
certain species (Duncan, 1975). Thus if food availability causes aggregation, it is predicted that (a) the green leaf density is higher on leks, and (b) there is a positive correlation between mating rate and the availability o f green grass leaves on territories.
The n utrient h otspot m odel, which specifically addresses lekking in savannah
ungulates, suggests that the crucial resource underlying female clustering could be trace minerals in the forage rather than food availability in general (McNaughton, 1988). Again assuming an ideal free distribution, the hypothesis predicts that females feed at least at similar rates on and o ff leks.
Without invoking any underlying cause for female dispersion, the hom e range
overlap m o d el suggests that lek formation is caused by overlap o f female home
ranges (Bradbury et a l, 1986). When running model simulations, it was found that males clustered in zones o f maximum overlap between female home ranges. Increasing density led to a decrease in lek size. Furthermore, leks could be spaced less than one female home range apart. The authors maintained that multiple leks within a female home range was not to be expected i f lek evolution had been driven by female choice; in such systems females would benefit from having only one large lek within their range enabling comparison between the maximum number o f males. Thus the model predicts that (1) leks cover the area o f highest female range overlap, (2) there is a negative correlation between lek size and density, and (3) leks may be situated less than one female home range apart.
Aggregation o f females on leks could also be explained by higher safety from predation due to either site-specific properties (Gosling and Petrie, 1990; Wittenberger, 1978) {^safe-site m o d e t) or general antipredator benefits o f grouping such as shared vigilance (Underwood, 1982), selfish herding (Treisman, 1975) or confusion effects (N eill and Cullen, 1974) ('safety-in-num bers modeF). The main predators o f topi in the study area are presumed to be hyenas. Even though an individual adult lion eats about twice as much as an adult hyena (daily meat intake: lion: 5-7 kg (Schaller, 1972), spotted hyena: 3.0 kg (Kruuk, 1972)), this difference is by far outweighed by the fact that hyenas in the Mara outnumber lions by more than 8 to 1 (Ogutu and Dublin, 1998). Furthermore, a study on a similar ungulate community in Serengeti NP found that lions did not prefer topi (Scheel, 1993), while hyenas in the area o f the present study have been found to select topi during the first half o f the year which includes the months o f the rut (Cooper et a l, 1999). Thus according to the safe-site model it is expected that females are less likely to meet
C hapter 4______________________________H otspots_____________________________________ ^
hyenas on the lek, and the safety-in-numbers model predicts that females experience a lower hyena:ungulate ratio on lek.
Finally, as all the above mentioned models make the assumption that female dispersion is determined by factors unrelated to mating activities, they share the prediction that the ratio o f females in oestrus should be similar on and o ff leks.
4.3 Methods
4.3.1 The simple hotspot model
4.3.1.1 Lek locations in relation to densities
To test whether lek locations were characterised by exceptionally high population densities, I mapped the topi density in the Burrungat Study Area on a 1 km^ scale based on 11 total counts conducted in the months between December and June in the period 1998/2000 (sec. 2.5.1). The total counts were typically conducted about a month apart (for dates, see Table 3.1). The general stability o f distributions from day to day was confirmed by altogether 22 sample counts conducted between the total counts. During a sample count, all topi within 500 m on either side o f the car were counted while driving a road transect along a 35 km loop from the Talek Gate to the Mara River and back.
I mapped the overall mean density as w ell as densities for the period preceding the rut (December-February), the rut (March-early May) and the period after the rut (late May-late June). Separate counts within each o f the seasonal sub divisions generally showed similar patterns outside the rut; this was not the case during the rut, and separate maps for each year are presented for this time. I compared the mean density in the lek vicinity with the density elsewhere, defining the lek vicinity as all 1 km^ cells that were at least partly situated within a 1 km radius o f a lek.
In addition, I mapped the distribution o f adult males and females separately onto four maps estimating food availability by the area reflecting green on LANDSAT7 satellite images taken between January and June 2000 (from w w w .usgs.gov). For January and June the photos were taken within one and four days respectively o f the closest total count; in March and April, the photos were taken in between total counts, thus I calculated the mean densities from the two closest
C hapter 4______________________________H otspots_____________________________________ 56
counts. I compared the density in the lek vicinity with the density elsewhere using the Mann-Whitney test. Furthermore, I calculated the dispersion indices for the two sexes as the variance divided by the mean (following Krebs, 1989).
4.3.1.2 Sex ratio on and off lek
For both the rut and non-rut period, I determined the adult sex ratio on the lek and at random transects through the rest o f the study area covering territorial as w ell as non territorial areas. W hile sex ratio on lek was calculated from all the adult topi present, the sex ratio in the rest o f the study area was determined by counting all the adult topi present within 500 m o f either side o f a 7 km random road transect. Counts were evenly distributed over the day.
Based on the eight total counts for which sex information was available, the spatial variation in the adult sex ratio per km^ was mapped as the ratio between the mean density o f females to the mean density o f males. Separate maps were produced for the pre-rut, the rut and the post-rut, and the proportion o f adults constituted by males was compared between the lek vicinity and elsewhere by Mann-Whitney test.
4.3.1.3 General resource availability on and off lek
I compared the wet season measures o f food availability between central lek territories, peripheral lek territories and resource territories; the measures included were the green leaf index as w ell as its components, grass cover, sward height and grass greenness (sec. 2.5.3). I also checked for correlation between the mean number o f females mated daily on territories and the territorial green lea f index.
4.3.2 The nutrient hotspot model
I compared the proportion o f time females spent grazing on the three territory types. The data were obtained by watching 24 females locally on all three territory types with a mean duration o f total watch time o f 8 / 2 hours. H alf o f the females were in
oestrus with oestrous state being determined from the reaction o f males (sec. 2.6).
4.3.3 The home range overlap model
4.3.3.1 Lek areas in relation to female range overlap
I determined the ranges o f individually recognised females in the Burrungat Study Area based on resightings during ID surveys throughout the duration o f the study.
C hapter 4 H otspots 57
During ID surveys, which typically lasted 3-4 hours, locations o f all known individuals were recorded using a GARMIN GPS 40. Care was taken to sample all areas o f the study area regularly although the ID surveys were stratified in favour o f areas known from the most recent total count to have high topi densities. A lso areas adjoining the study area were occasionally surveyed. Individual females included in the following analyses were resighted at least 20, and up to 54, times between November 1998 and June 2000.
Using the Animal Movement extension (version 2.0 Beta) (Hooge et al., 1999) to Arc V iew (version 3.2), ranges were mapped as the minimum convex polygons (MCPs) encompassing all relocation points o f individual females. The zone o f highest overlap between female ranges was mapped on a 1 km^ scale based on the number o f individual MCPs covering the area. The extent o f this zone was compared to lek locations.
Moreover, for each female, I estimated home range size as the asymptote o f the curve describing the relationship between the number o f relocations and the area o f the MCP (Jennrich and Turner, 1969) (Figure 4.1). The curve was drawn by jack-
MCP
10 13 16 10 22 26 28 31