Partitioning of pollinators during flowering in an African Acacia community

Ecology, Dec, 1998 by Graham N. Stone, Pat Willmer, J. Alexandra Rowe

Tests of structure in temporal patterns of pollen release across species

Character displacement resulting from interspecific competition (or competitive displacement; Brown and Wilson 1956, Pleasants 1980, Williams 1995) reduces the overlap among species along a given resource axis, resulting in species values that are more regularly spaced than predicted by chance (e.g., Armbruster et al. 1994). Several methods have been developed to test temporal structure in flowering seasons (Poole and Rathcke 1979, Pleasants 1980, Cole 1981, Gleeson 1981, Fleming and Partridge 1984, Kochmer and Handel 1986), each comparing the observed spacing of flowering seasons to the null prediction that species' blooming periods are randomly distributed through time. The most appropriate statistic with which to test character displacement is the Var statistic, originally proposed by Poole and Rathcke (1979) (Pleasants 1994, Williams 1995). Use of Var in its original form is made complex by the need to perform a long series of randomizations in order to assess the statistical significance of values obtained. Williams (1995) provides a table of critical values of a statistic V, derived from Var, which greatly simplifies the detection of significant character displacement, and we use the V statistic here. For a given set of species values along a resource axis, calculation of V requires estimation of (1) the distances between successive species (first to second species, second to third species, etc.) and (2) the difference between the greatest and least values shown by the set of species in question, or range. V is then given by the expression:

V = Sum of squares of the distances/(number of species - 1) x [(range).sup.2].

The V statistic can be applied to any resource axis, and we use it to examine the spacing of daily peaks of pollen availability within and among Acacia species. Character displacement, if it has occurred, is predicted to result in two patterns in timing of pollen release: (1) intraspecific aggregation and (2) interspecific divergence. Both of these predictions can be tested using Williams' V statistic (Williams 1995).

Intraspecific variation. - If individuals of a species release their pollen at similar times, peaks of pollen availability for individuals will be aggregated in daily time. Testing for aggregation requires an assumption about the time period within which individuals may release their pollen, since the null hypothesis to be tested is that individual peaks of pollen availability are randomly distributed within this period. Because all of the flower visitors observed at Acacia trees were diurnally active species, and none of the Acacia species release its pollen at night, we assume the period available for visitor activity to be from sunrise to sunset ([approximately]0600-1800).

Interspecific variation. - Analysis of the spacing of flowering phenologies is often carried out between first and last species peaks (Pleasants 1980), which is appropriate where there are no a priori limits to the resource axis between which species values should be dispersed. Where limits can be identified, spacing is analyzed between limiting values along the resource axis. We apply both methods to interspecific patterns of pollen release, using dawn and dusk as biologically meaningful a priori limits to the resource axis, as justified above. The null hypothesis tested in each case is that the peaks of pollen availability are randomly distributed between the delimited values of time.