• anti-predator adaptation;
  • dispersal;
  • flowering;
  • fruiting;
  • seasonality


  1. Top of page
  2. Summary
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  • 1
    We examined the factors determining synchrony in reproduction in nine Acacia and six other tree species in the Serengeti ecosystem.
  • 2
    We test two hypotheses: (i) an abiotic hypothesis where the primary determinant of synchrony is an adaptation to water availability; and (ii) biotic hypotheses where these adaptations to water can be further refined by additional adaptations to avoid predators, or attract seed and fruit dispersers.
  • 3
    Flowering and fruiting were recorded monthly for individually marked trees during 1997–2004. Flowering in different species occurs semi-annually, annually or, in the case of one species, once every 2 years. For most species synchrony of flowering was correlated with seasonal rainfall, with lags related to the mean height of the species; small species flowered during the rains while larger species flowered in the dry season. Fruiting seasons occurred at the end of the rains irrespective of the flowering season.
  • 4
    Most species showed flowering synchrony greater than expected from the distribution of rainfall. This may be related to avoidance of insect seed predators through predator satiation. Two Acacias showed multi-annual fruiting (masting), possibly as a predator avoidance mechanism. Acacia tortilis has two flowering seasons: a dry season flowering with early abortion of pods and a wet season flowering producing successful fruits.
  • 5
    Two species of Commiphora appeared to be synchronized so as to attract birds that disperse seeds. Acacia tortilis produced indehiscent pods attractive to ungulates, possibly to kill bruchid beetles during digestion and so increase seed viability.
  • 6
    Our results suggest that synchrony in these trees is caused by a strong interaction between abiotic and biotic factors. Closely related species have different reproductive patterns of synchrony that seem to be adapted to different combinations of rainfall, predators and dispersers. Rainfall is the primary determinant but the activities of predators and dispersers increase the degree of synchrony.


  1. Top of page
  2. Summary
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

Synchrony in reproductive effort among tropical trees has been a long-standing question for ecologists (Richards 1952, 1996). Interpretation of this phenomenon has focused on the importance of various abiotic and biotic environmental cues, although the primary causal mechanisms remain unclear. Abiotic hypotheses have focused on seasonal variations in the availability of moisture. Trees flower and fruit as an adaptive response to cues relating to increased stress at the start of the dry season, or increased favourable conditions at the start of the rainy season. Biotic hypotheses have focused on plant–animal interactions where simultaneous flowering leads to increased pollination and reproductive success for some species while simultaneous fruiting satiates predators, or results in a reduction of fruit predators.

Such questions are especially relevant to tropical savanna regions. In the Serengeti-Mara ecosystem of East Africa, trees have been observed to flower in most months of the year and the reasons for the timing of flowering are not obvious. Timing could be determined by the availability of water related to the seasonal rains because water is seasonally limiting in African savannas (Hesla et al. 1985; Belsky et al. 1993). In addition, trees are a major component of the savanna vegetation and canopy foliage is the food of large mammals such as giraffe and elephant, and indirectly the food source of insect-gleaning birds. Fruits are consumed, both pre- and post-dispersal, by many mammals (e.g. impalas, baboons, vervet monkeys) and birds (e.g. barbets, starlings), and flowers are not only pollinated by insects but are the food source of insects, birds and large mammals. Thus, the timing of flowering and fruiting is an important factor in the dynamics of the ecosystem and may be related to the activities of predators.

In this paper we test possible explanatory models (Table 1) that may best explain flowering and fruiting synchrony. We do so using over 41/2 years of data from trees in the Serengeti ecosystem (February 1997 to August 2001), with data for two species continuing until May 2004. If abiotic hypotheses (Table 1) are correct, we predict that flowering and fruiting phenology is an adaptive response to seasonally available resources, in our case water (hypothesis 1, the resource-matching hypothesis; Kemp 1983; Kelly 1994, Koenig & Knops 2000; Kelly & Sork 2002), or interacting with edaphic factors that affect moisture availability.

Table 1.  Hypotheses and predictions on the role of rainfall and predators as causes of synchrony in the flowering and fruiting phenology of savanna trees in the Serengeti ecosystem
Hypotheses and modelsPredictionReferences
1. Resource matchingFlowering/fruiting correlated to rainfallKemp 1983, Kelly 1994, Kelly & Sork 2002, Koenig & Knops 2000
1a. Modified by edaphic conditionsFlowering/fruiting correlated to rainfall more so in non-riverine species 
2. Predator satiation(a) More synchronized than rainfallJanzen 1974, Augspurger 1981, Norton & Kelly 1988, Kelly 1994, Kelly et al. 2000, Koenig & Knops 2000
(b) Supra-annual synchrony (masting)Ashton 1969, Ashton et al. 1988
3. Predator cleansingFruiting aborted followed by successful reproductionKelly & Sork 2002
4. DispersalFruiting synchronized in indehiscent species, or fruit producersMiller & Coe 1993, Miller 1996b

If biotic hypotheses (Table 1; hypothesis 2) are correct, then in addition to the response to rainfall, synchrony might be an anti-predator adaptation, or an adaptation for dispersal. Under this biotic category there are four potential anti-predator mechanisms that act singularly or interactively. First, a higher proportion of flowers or fruits might survive herbivory if they are produced simultaneously because predators are satiated (Janzen 1974; Augspurger 1981; Norton & Kelly 1988; Kelly 1994; Kelly et al. 2000; Koenig & Knops 2000). Secondly, predation by insects may be reduced if reproduction takes place at supra-annual intervals (termed ‘masting’) (Ashton 1969; Ashton et al. 1988). Thirdly, predator cleansing (Kelly & Sork 2002) predicts that trees reduce insect predation through abortion of fruits at an early stage of development, thus killing a cohort of insect seed predators, followed shortly by a new flowering episode with successful fruiting. Finally, Miller & Coe (1993) and Miller (1996b) propose that the synchrony of edible fruits attracts the attention of animals and promotes dispersal of the seeds.


  1. Top of page
  2. Summary
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

the study area

The Serengeti-Mara ecosystem is an area of some 25 000 km2 on the border of Tanzania and Kenya, East Africa (34–36° E; 1–3°30′ S). The climate shows a relatively constant mean monthly maximum of 27–28 °C at Seronera in the centre of the park. The minimum temperature varies from 16 °C in the hot months of October–March to 13 °C during May–August. Rain typically falls in a bimodal pattern, with the long rains during March–May and the short rains during November–December (Norton-Griffiths et al. 1975). There is a rainfall gradient from the dry south-east plains (500 mm year−1) to the wet north-west on the Kenya border (1200 mm year−1) (Sinclair 1979).

The south-east plains are generally treeless and dominated by grasses and many small dicots. Further to the west deeper soils allow larger tussock grass species that also dominate in the Acacia woodlands. The woodlands are dominated by Acacia species in all areas except for a small region in the north-west of the system where Terminalia and Combretum take over.

the tree species

We examined 15 common tree species; nine species of Acacia, two of Commiphora, and Balanites aegyptiaca (L.) Del., Ficus sp., Kigelia africana (Lam) Benth, and Terminalia mollis M. Laws (Table 2). Most are insect pollinated (Ross 1979), except Kigelia, which is probably pollinated by bats. All the Acacias except A. tortilis produce dehiscent seedpods that drop their seeds below the trees. Flowers and pods of most dehiscent species are browsed while still on the trees by primates, elephant and giraffe (Table 2). Acacia tortilis produces indehiscent pods, which are eaten rapidly by antelope and elephant as soon as they drop (Burtt 1929; Lamprey 1967; Lamprey et al. 1974). Both species of Commiphora produce seeds within a fleshy outer pericarp. Birds, especially barbets (Capitonidae) and starlings (Sturnidae), feed upon these fruits and regurgitate the seed after digesting the covering. Many birds and mammals consume Ficus fruits. Kigelia produces pods, which, when still small, are eaten by primates and giraffe, and when mature are consumed by elephants (Table 2). Terminalia produces winged seeds and is the only species in our sample dispersed by wind.

Table 2.  Location and site conditions of the species used in the study, along with their height and seed dispersal agents. Most species are insect-pollinated. Kigelia africana is also pollinated by bats and Terminalia is wind pollinated
SpeciesNumber of trees sampledLocation of highest densityConditionsMean height (m)Seed dispersal agents
Acacia drepanolobium30West, Centre, NorthDrier impeded drainage 8.8Primates, giraffe
Acacia mellifera25WestMid-slope, on sandy areas of deeper soils 6.7Giraffe
Acacia senegal20WestWell-drained ridges in stony areas with savanna 8.6Bushbabies, giraffe
Acacia seyal20WestWetter impeded drainage13.2Primates, giraffe
Acacia gerrardii20NorthMid-slope in far north-west 9.5None observed
Acacia kirkii20CentreRiver banks and in rivers15.2Primates, giraffe
Acacia robusta40UbiquitousMid-slope13.7None observed
Acacia tortilis30Centre, EastWell-drained ridges on edge of Serengeti plains12.8Antelope, elephant, giraffe
Acacia xanthophloea10CentreNear large rivers on silty soils20.2Primates, elephant, giraffe
Commiphora africana 8West, EastWell-drained ridges 5.8Birds, primates
Commiphora schimperi20West, EastWell-drained ridges 6.9Birds, primates
Balanites aegyptiaca20WestDrier impeded drainage10.5Primates, jackal, mongoose
Terminalia mollis20NorthWell-drained ridges only in far north-west19.5Wind
Ficus sp.11UbiquitousRidges and rock outcrops near springs, and near rivers15.4Birds, primates
Kigelia africana20North, Centre, EastDrainage lines and near rivers15.6Baboon, elephant

The species all prefer particular edaphic conditions and locations on the soil catena (Table 2). The catena is characterized by shallow, stony and sandy soils on the top of gentle ridges, medium depth soils on the slopes, and deep, silty soils at the bottom near rivers and in areas of impeded drainage. Although most species were widely distributed, they differed in the regions (west, north, east or centre) of the park where they were at highest densities.

data collection for flowering and fruiting

We monitored trees over the whole of the Serengeti woodlands but sampled each species throughout the regions of its highest density (Table 2). Most species grew in monospecific stands from a few hectares to many hundreds of hectares. The exceptions were Kigalia and Ficus, which grew singly in the grasslands, sometimes with a few other scattered trees close by. These two species were selected randomly throughout the region where we could see them or reach them. For trees in the monospecific stands the first tree was chosen randomly, then others were chosen at a minimum distance of 20 m from each other. Trees had to be of a size capable of flowering. Selected trees were marked with aluminium tags and located using GPS coordinates. For 12 of the 15 species, at least 20 individual trees were monitored (Table 2), but with smaller sample sizes for Ficus (n = 11), A. xanthophloea (n = 10) and C. africana (n = 8). Because flowering from start to complete disappearance of flowers lasted at least a month, we visited marked trees after an interval of 14–21 days, and this ensured that we recorded any flowering or fruiting event in a given month. In each survey we recorded all trees within the same week. We also observed whether species pods were eaten (usually by primates and some ungulates (giraffe, elephant)), when pods were still on the trees, or by ungulates on the ground. The pods of two species, A. robusta and A. gerrardii, were never observed to be eaten. Birds did not eat Acacia pods.

monitoring of flowering and fruiting

Monitoring began in February 1997. The degree of flowering and fruiting was scored as a percentage of the total surface area of the tree. We used an index for scoring from 0 to 5 (1 ≤ 5%, 2 = 6–25%, 3 = 26–50%, 4 = 51–75%, 5 ≥ 75% cover of the tree). Using this scoring index we calculated the interval that refers to the time between peaks of flowering or fruiting events of individual trees. A flowering event took less than 3 weeks, and fruiting from the first appearance of pods or fruits to maturity took a month, except in Kigelia when the very large pods took 4 months to mature. A few trees of A. robusta and A. kirkii were not recorded as flowering or fruiting until some 2 years of the study had elapsed. Records for both of these species were continued until May 2004 to accommodate the longer reproductive intervals, while records for all other species terminated in August 2001.

For each species the number of individual flowering events was summed by month over the whole study period (Σ Nfm) and divided by the total number of trees observed for each month (Σ Ntm) to give a monthly ratio (Fm). For example, if all observed trees were flowering in all five months of February (1997–2001) during the study then the ratio of flowering trees to total trees would be 1.0. Thus

  • Fm = Σ Nfm Ntm( eqn 1)

The ratios for each month were summed over the year to give the annual flowering tree/total tree ratio (Fy =Σ Fm). The percentage of annual flowering events in each month (%Fm) was therefore,

  • %Fm = 100 · Fm/Fy( eqn 2)

The same calculations were made for fruiting events. This procedure overcame bias from unequal samples of trees counted each month.


Rainfall was measured monthly from storage gauges placed in various regions of the ecosystem. We calculated the mean monthly rainfall for each region separately, and from these we obtained the proportion of annual rainfall occurring in each month.

synchrony of flowering and fruiting

Synchrony has been defined as the simultaneous occurrence of the same event in the different species being studied (Newstrom et al. 1994). However, we use synchrony in the more restricted sense that individuals within a species simultaneously flower or fruit. If the frequency distribution of flowering events is clumped and significantly different from a uniform distribution then it is synchronized with respect to the uniform distribution. By extension, the frequency distribution of these events must be significantly more clumped than that predicted by the monthly distribution of rainfall to conclude that flowering (or fruiting) is more synchronized than rainfall. This definition addresses the degree to which flowering or fruiting events are coordinated or spread over time with respect to an expected frequency distribution irrespective of the time of year.

To determine if the synchrony of flowering and fruiting was related to seasonal rains, we correlated the frequency distributions of flowering events over the months with that expected if the same sample size was distributed similar to the rainfall for the region of Serengeti that the species occurred in. Because trees may not respond immediately to rainfall, but instead could anticipate or delay their response, we correlated these events with rainfall by shifting their monthly values until the maximum correlation with the rain was obtained.

Because flowering and fruiting were recorded by month, our measure of synchrony was the degree to which observed events were distributed over the 12 months. We used the non-parametric two-sample Kolmogorov-Smirnov test, examining differences between observed distributions of flowering and fruiting events vs. (i) expected uniform distributions through the year (K-Suniform), and (ii) expected distributions determined by rainfall (K-Srain). The observed totals of flowering or fruiting events were used to calculate the expected distributions for the tests. These reproductive events should be clumped in time, that is, they should be occurring simultaneously amongst individuals of the same species. The degree of clumping was measured using Colwell's (1973) measure of contingency, M. Contingency indicates the degree to which a flowering or fruiting event is dependent upon the time of year (month). It ranges from 0 to 1 and the formulae are given in Colwell (1973). It is maximal if all reproductive events fall in the same month and minimal if these events fall evenly over all months. It makes use of information from the sample of trees that are not flowering or fruiting in each month as well as those that are.

We have defined synchrony above as the simultaneous occurrence of fruiting and/or flowering by different individuals within a species. For example, if the seasonal rainfall determines synchrony, then the monthly spread of flowering and fruiting should be similar to that of rainfall. We compared the observed proportion of flowering trees over each month with that expected if the same proportion of flowering trees was distributed identically to that of rainfall, by calculating the contingency (M) values for the two frequencies. We tested the rainfall predictions by taking the ratio of the expected M-values predicted by rainfall to the observed M-values over the 12-month period. If flowering were synchronized by rainfall then the ratio should be unity, whereas the ratio should be less than unity if flowering is more synchronized than rainfall and greater than unity if it is more spread out than rainfall.


  1. Top of page
  2. Summary
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

flowering and fruiting intervals

We calculated the interval, as defined above, between flowering (or fruiting) events (Fig. 1, Table 3). Most species show an approximately annual frequency in flowering. However, five Acacia species (A. drepanolobium, A. senegal, A. mellifera, A. seyal and A. tortilis) also show a peak at 6-month intervals. Longer intervals are shown by A. kirkii, with a small peak at 24 months. Acacia robusta has a significantly longer flowering interval (mean of 29 months) than any other species (Kolmogorov-Smirnov one-tailed test P < 0.01) and shows two peaks, one at 24 months and the other at 51 months, i.e. some trees did not flower for 4 years.


Figure 1. The frequency of flowering (solid) and fruiting (open) intervals for each tree species summed over 3-month periods (1–3 months, 4–6 months, etc., where a month is 30 days).

Table 3.  Mean interval in months between flowering peaks and between fruiting peaks. Number of intervals is the sum of all intervals over all individuals for a species
SpeciesInterval between flowering peaksInterval between fruiting peaks
MonthsNumber of intervalsMonthsNumber of intervals
Acacia drepanolobium11.9 8514.7 59
Acacia mellifera 9.4 8811.2 83
Acacia senegal 7.0 53 8.5 48
Acacia seyal 6.2131 7.3106
Acacia gerrardii11.7 4713.9 37
Acacia kirkii15.1 7020.3 33
Acacia robusta28.6 3630.4 21
Acacia tortilis 8.512211.9 81
Acacia xanthophloea11.9 3613.6 30
Commiphora africana11.8 1727.2  2
Commiphora schimperi13.3 4817.8 33
Balanites aegyptiaca 7.3 79 9.4 53
Terminalia mollis10.0 4913.1 39
Ficus sp. 7.8 47 8.3 46
Kigelia africana11.9 3112.3 21

The interval between fruiting events is generally similar to that of flowering (Fig. 1). Four species (A. tortilis, A. drepanolobium, Balanites and Terminalia) have significantly longer fruiting intervals than their flowering intervals (Kolmogorov-Smirnov one-tailed test, P < 0.006, P < 0.028, P < 0.025, P < 0.007 for the four species, respectively).

synchrony of flowering and fruiting


Flowering and fruiting frequencies (Fig. 2) were compared with that for monthly rainfall using the Pearson correlation. Most of the correlation values were high, indicating a seasonal effect, albeit with a lag (Table 4). However, three Acacia species (A. tortilis, A. xanthophloea and A. seyal) showed low correlations with rainfall distribution. Each of these has two flowering peaks, one in the wet and one in the dry season.


Figure 2. Mean monthly distribution of flowering (squares) and fruiting events (circles) expressed as a percentage of the yearly total, compared with the mean monthly rainfall (histogram) for the areas where each species of tree is found, indicated in parenthesis.

Download figure to PowerPoint

Table 4.  The maximum correlation between monthly flowering and fruiting frequency and mean monthly rainfall. The lag is measured in months before the rain peak. P-values are based on maximum correlations
CorrelationLagP <CorrelationLagP <
Acacia drepanolobium0.635 10.010.87510< 0.0005
Acacia mellifera0.621 10.010.71510   0.005
Acacia senegal0.686 20.0050.76610   0.0025
Acacia seyal0.502 20.050.604 7   0.025
Acacia gerrardii0.625 30.010.93210< 0.0005
Acacia kirkii0.691110.0050.74711   0.005
Acacia robusta0.678 20.0050.88710< 0.0005
Acacia tortilis0.271 80.250.736 8   0.0025
Acacia xanthophloea0.481 80.050.597 6   0.025
Commiphora africana0.593 40.0250.645 7   0.01
Commiphora schimperi0.656 50.010.696 5   0.005
Balanites aegyptiaca0.526 50.050.520 7   0.05
Terminalia mollis0.686 40.0050.79911   0.001
Ficus sp.0.459 10.10.33310   0.25
Kigelia africana0.578 70.0250.742 0  0.0025

Comparing species, the peak in flowering occurred at different times in relation to rainfall, with several non-Acacia species (Commiphora, Balanites, Terminalia and Kigelia) flowering in the dry season 4 or 5 months ahead of the rain. Acacia species tended to flower either just before the rain or just after it. This lag appears to be related to the mean size of the tree as indicated by height, with small species flowering before the rains and large species flowering progressively later (Fig. 3).


Figure 3. The lag in months of the flowering peak before the rainfall peak related to the height of tree for the nine Acacia species. Lag = 0.6295 x  3.368 (r2 = 0.494) where x is height in metres.

Download figure to PowerPoint

Fruiting peaks, in contrast to flowering peaks, tended to be towards the end of the rains. Most species produced fruits 2 months after the peak of rain (Table 4). The lag in fruiting with rainfall was not related to tree size. Thus, there was a longer period between the peak of flowering and the peak of fruiting in small species than in larger ones.

Frequency distributions

The majority of tree species show flowering and fruiting distributions that differ significantly from a uniform distribution (Tables 5, K-Suniform). Nevertheless, some species did produce flowers throughout the year, in particular Ficus, A. seyal and A. xanthophloea. Fruiting by most species also differed from a uniform distribution through the year.

Table 5.  Kolmogorov-Smirnov values for monthly (a) flowering or (b) fruiting distribution relative to uniform distribution (K-Suniform) and relative to that predicted by rainfall (K-Srain). Distributions are different (in bold) if P < 0.003 (Bonferroni correction)
 Sample sizeK-SuniformP <K-SrainP <
(a) Flowering
 Acacia drepanolobium2140.260.00000.160.0094
 Acacia mellifera1570.300.00000.170.0169
 Acacia senegal 980.220.01750.190.0503
 Acacia seyal3860.090.06250.150.0003
 Acacia gerrardii1220.230.00430.170.0567
 Acacia kirkii 830.340.00010.190.1162
 Acacia robusta1680.370.00000.280.0000
 Acacia tortilis2500.140.01370.160.0041
 Acacia xanthophloea 980.180.08650.310.0002
 Commiphora africana 420.500.00020.630.0000
 Commiphora schimperi1090.490.00000.550.0000
 Balanites aegyptiaca2950.160.00070.230.0000
 Terminalia mollis1490.310.00000.230.0000
 Ficus sp. 570.130.71360.350.8091
 Kigelia africana 740.330.00080.400.0000
(b) Fruiting
 Acacia drepanolobium1540.300.00000.230.0006
 Acacia mellifera3000.170.00020.150.0020
 Acacia senegal2040.220.00020.120.1112
 Acacia seyal3340.120.01090.170.0001
 Acacia gerrardii1720.240.00010.170.0172
 Acacia kirkii 510.440.00010.330.0085
 Acacia robusta2030.340.00000.270.0000
 Acacia tortilis3030.180.00010.300.0000
 Acacia xanthophloea 750.520.00000.600.0000
 Commiphora africana  80.580.71
 Commiphora schimperi1030.530.00000.610.0000
 Balanites aegyptiaca2700.110.05470.280.0000
 Terminalia mollis2880.280.00000.210.0000
 Ficus sp.1460.070.94480.230.0012
 Kigelia africana2060.120.11450.070.6940

With respect to rainfall, all Acacia species showed a peak of flowering in the period January–March (Fig. 2), which is before the April peak in rainfall. Terminalia and Balanites produce a flowering peak in the early rains (November–December), while Kigelia and Commiphora flower at the height of the dry season (August–September). Many species showed flowering frequencies that differed from that predicted by rainfall, with some highly significantly different. Only Ficus showed distributions that were similar to that of rainfall (Tables 5, K-Srain). Fruiting peaks, in contrast to flowering, tended to be towards the end of the rains, with most species producing fruits 2 months after the peak of the rain. Consequently, fruiting frequencies differed from those predicted by rainfall in all but four species.

The degree of clumping of reproductive events

We measured the degree to which individual trees of the same species flowered or fruited in the same months (clumping of events) using the contingency measure (M). Although there was flowering and fruiting in most months, all species showed some degree of clumping (Table 6). If flowering was synchronized by rainfall, then the ratio should be unity, whereas the ratio should be less than unity if flowering is more synchronized than rainfall and greater than unity if it is more spread out than rainfall. Table 6 shows that few species exhibited either flowering or fruiting that was clearly more spread out through the year than that predicted by rainfall (that is with a ratio clearly above unity). In flowering, only A. tortilis and Balanites were more spread than predicted by rainfall. For fruiting, five species were more spread out than predicted by rain (A. mellifera, A. senegal, Balanites, Kigelia and Ficus). The last three have fruits that ripen over an extended time and are eaten by large mammals. Thus most species showed a greater degree of clumping of both flowering and fruiting than that predicted by rainfall, indicating a significant degree of seasonal synchrony, as indicated by significantly different K-Srain values (Table 5). However, the relationship was different with each species. Most of the non-Acacia species tended to flower in the dry season. The Acacia species flowered at different times with respect to the rains, the smaller ones flowering before the rains, the largest ones flowering after the rains. Fruiting, in contrast, tended to occur towards the end of the rains irrespective of when flowering occurred in a species.

Table 6.  Contingency (M) values for monthly flowering and fruiting distributions, and predicted M-values if reproductive events were distributed the same as the rainfall. The ratios are computed from the expected M-values from rainfall divided by the observed M-values. M ranges from zero for uniform distribution to unity for maximum synchrony (all events in the same month). See text for explanation of the ratios
SpeciesContingency MfloweringContingency MfruitingRain MfloweringRain MfruitingFlowering Mrain/M obsFruiting Mrain/Mobs
Acacia drepanolobium0.1570.0400.0250.0180.1620.447
Acacia mellifera0.0620.0750.0470.1070.7541.417
Acacia senegal0.0620.0730.0560.1550.9052.117
Acacia seyal0.1010.1650.1430.1131.4120.686
Acacia gerrardii0.0550.0560.0260.0400.4730.713
Acacia kirkii0.1040.0910.0320.0190.3070.207
Acacia robusta0.0820.0580.0180.0230.2240.387
Acacia tortilis0.0430.1050.0570.0731.3240.691
Acacia xanthophloea0.1670.2430.0370.0270.2240.110
Commiphora africana0.1680.0510.0320.0060.1920.112
Commiphora schimperi0.1910.1770.0340.0320.1790.183
Balanites aegyptiaca0.0920.0390.1660.1491.7973.825
Terminalia mollis0.1350.3110.0720.2080.5350.667
Ficus sp.0.0510.0390.0210.0680.4131.732
Kigelia africana0.0720.0320.0230.0800.3192.502
Maximum value1111  

One interpretation of these patterns is that although rain may act to trigger reproductive events, there is some other factor that promotes more synchronized flowering and fruiting than predicted by rainfall.

seasonal availability of seedpods and fruits

Although there is a peak of seedpods produced during the rains, there are some pods available to mammal feeders year round when all Acacia species are considered together (Fig. 4a). The lowest availability of pods occurs in December following the first rains in November, and there is an abundance of pods in the middle of the dry season when food is potentially limiting to large herbivores. The peak of pod availability coincides with the peak in abundance of insects, which occurs in the middle of the rains in most insect groups, including beetles (Sinclair 1978; Sinclair et al. 2002).


Figure 4. The mean availability of (a) edible Acacia seedpods for each month, (b) all Acacia flowers for each month, and (c) Commiphora and Ficus fruits that are eaten by birds.

Download figure to PowerPoint

Acacia flowers exhibited a peak in the early part of the rains (December to March) but were present in all months (Fig. 4b). However, the lowest incidence of flowers occurred towards the end of the rains in June.

The average incidence of fruits of Commiphora and Ficus species (Fig. 4c) indicates a sharp peak in the dry season (August to October). These fruits are consumed by several frugivorous bird species, such as barbets, hornbills (Bucerotidae), mousebirds (Coliidae), turacos (Musophagidae) and bulbuls (Pycnonotidae), when there is little else available.


  1. Top of page
  2. Summary
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

Our results demonstrate that resource matching (hypothesis 1) primarily explains flowering in the Serengeti, with different trees responding to moisture cues in different ways, including interactions with soil and seasonal rainfall. Biotic influences were also highly significant, with two Acacias showing a phenology consistent with predator-avoidance mechanisms (hypotheses 2 and 3), and two species of Commiphora appeared to be synchronized so as to attract birds that disperse seeds (hypothesis 4). Both rainfall and biotic factors underlie synchronous reproductive events and we argue that rainfall is the primary trigger, and that the activities of predators and dispersers increase the degree of synchrony.

flowering and fruiting intervals

Many of the species flowered annually, five of the nine Acacias flowered twice a year, and A. robusta flowered every 2 years. Fruiting intervals reflected flowering intervals in most species. However, in both A. kirkii and A. tortilis it appeared that pods had begun growth but were aborted early in development. One explanation for the difference could be that a failure of seed-set or aborted growth during a flowering period might stimulate a new flowering episode a short time later.

causes of synchrony


We compare the results with our predictions in Table 1. Overall, rainfall-mediated flowering was pronounced but mostly for species in areas with rapid drainage where summer moisture deficits are predicted to be most severe. For species in poorly drained areas, such as A. seyal and A. xanthophloea, patterns of flowering were dissimilar to that of the rainfall distribution (Tables 5 and 6). These two species, plus Ficus, Balanites and Kigelia, showed peaks of flowering in both wet and dry seasons. All of these species live in impeded drainage, wet areas or near rivers. In contrast, all the other species showed flowering seasonality, with higher correlations to rainfall. Each of these has just one flowering peak in the year and most live on drier ground, higher on the catena. This suggests that for species with a single flowering season in the year, their reproduction is related to rainfall, and most of these species live on drier sites. For these species the results correlate with prediction 1a of the resource-matching hypothesis. These predictions did not hold for species with two flowering peaks in the year such as A. seyal, A. xanthophloea, Balanites and Ficus. The correlation of fruiting with rainfall was higher than that for flowering in most species and this may be because fruiting occurs towards the end of the rains when perhaps conditions are good for germination, whereas flowering in the larger species occurs in the dry season and that in smaller species occurs at the start of the rains. Thus, it appears that reproduction is designed to suit the resource conditions for fruiting and germination rather than for flowering. The observations for A. tortilis highlight this conclusion: fruiting was strongly correlated with rainfall because only the wet season flowering peak produced fruits, while that of the dry season aborted the pods at an early stage.

Synchrony for predator satiation

Eight species (two Commiphora species, A. xanthophloea, A. kirkii, A. robusta, A. drepanolobium, A. gerrardii and Terminalia) showed pronounced synchrony in both flowering and fruiting, and large differences in frequency distribution compared with rainfall. For these species, the results conform to the predator satiation hypothesis (prediction 2a). This mechanism of satiation involves a higher proportion of flowers or fruits surviving predator attack if they are produced simultaneously because predators have less time to impose their damaging effects, and survival of the remaining flowers or fruits increases (Augspurger 1981, 1983). Augspurger (1981) demonstrated that individuals in synchronous populations of the subtropical shrub, Hybanthus prunifolius, matured a greater number (×10) of seeds than individuals that were out of synchrony, and this was due to a much higher pollination success. In addition, fruit infestation by a damaging larva was greater in the asynchronous population. The combined effect of pollinators and seed predators was additive and produced intense selection against temporally isolated individuals.

High annual synchrony was proposed for tropical forests in Costa Rica and Malaysia as a mechanism to reduce both predation by insects before the pods leave the tree (pre-dispersal predation) (Toy 1991) and predation by vertebrates once the seeds fall (post-dispersal predation) (Janzen 1967, 1974; Walters et al. 2005). Many savanna tree species are dispersed passively by animals (birds, ungulates, rodents, termites and ants) (Tybirk et al. 1993; Miller 1994; Walters et al. 2005). Chidumayo (1997) reports that pre- and post-dispersal seed predation account for 86% and 33%, respectively, and Schupp (1988) reports that mortality on the ground after seedfall is often in excess of 75%. Because of this extensive mortality, seed and fruit predation is potentially a major ecological and evolutionary force affecting trees as individuals, populations and communities.

Predator satiation and masting

The predator satiation hypothesis also predicted (2b) supra-annual synchrony, or masting, i.e. reproductive events at multi-year intervals. Acacia robusta and A. kirkii reproduced every 2 years and some trees produced fruits only every 4 years. The combination of high synchrony and supra-annual cycles shown by these two species is more consistent with the predictions of the anti-predator hypothesis (Table 1, prediction 2b), in particular against bruchid beetle attack, than with the mechanism for sequestering resources. This latter hypothesis does not require highly synchronized fruiting. Predation by insects can be reduced if reproduction takes place at supra-annual intervals. This was proposed for the synchronized flowering of tropical dipterocarp trees every few years to avoid insect predation (Ashton 1969; Ashton et al. 1988) and mast seeding of native grasses in New Zealand (Kelly et al. 2000). A similar mechanism was proposed for the supra-annual masting cycles of conifer trees (Norton & Kelly 1988; Koenig & Knops 2001), and angiosperms such as Quercus, Fagus and Nothofagus at higher latitudes, possibly to avoid vertebrate predation (Norton & Kelly 1988; Kelly 1994; Crawley & Long 1995; Koenig & Knops 2000, 2001; Kelly & Sork 2002).

However, supra-annual intervals could also be related to the sequestering of resources, rather than to predator avoidance; growth and nutrient storage in one year alternates with reproduction in the other year. Such a strategy could also be spatially synchronized using local environmental cues (Koenig & Knops 2000). These alternative mechanisms can be distinguished by the degree of synchrony within the year; a much higher synchrony of fruiting than of resources, as recorded for A. robusta and A. kirkii, predicts predator avoidance.

Predator cleansing

Another proposed anti-predator mechanism is predator cleansing (Kelly & Sork 2002), namely the reduction of predator numbers through the early abortion of fruits (thereby killing a cohort of fruit-boring insects at little energetic cost to the plant), followed by a subsequent flowering and successful fruiting (prediction 5). One species, A. tortilis, showed this reproductive behaviour. It had a dry season flowering and early pod production that failed, followed by a wet season flowering that produced successful fruits.


Synchronized reproduction could serve to promote dispersal of fruits (hypothesis 4). Such synchrony would appear in species producing fleshy fruits or indehiscent pods. Both Commiphora species showed strong synchrony, producing fleshy fruits that are much favoured by birds that digest the fleshy outer fruit and regurgitate the seed. Several species of Commiphora and Ficus produce fruits attractive to birds and mammals (Burtt 1929; Lamprey 1967; Sharam 2005) and so they could act as seed dispersers. These fruits therefore could be synchronized so that they can attract vertebrate dispersers (Herrera et al. 1998).

Seed viability

Acacia tortilis has indehiscent pods that are rapidly consumed by impala. Lamprey et al. (1974) have shown that the viability of seeds from indehiscent Acacia species, and especially A. tortilis, increases if they pass through the ungulate's gut because the young bruchid larvae are killed before they damage the seed. This mechanism is most effective if pods are eaten shortly after they drop, and so there would be an advantage to attracting mammals quickly through synchronized shedding of pods. The pods are nutritious and attractive to ungulates (Gwynne 1969). This may be a strategy to reduce seed predators, especially because A. tortilis is attacked by at least 10 species of bruchids (Miller 1996a).

other constraints on the timing of flowering and fruiting

Flowering times affect plant success and subsequent fruit development and dispersal. An underlying assumption in each of our hypotheses is that synchrony of flowering, and fruiting, are selected traits, although we also recognize that with the resource-matching hypothesis, synchrony may in part be a response to trees tracking temporally variable resources. We have argued that the timing of flowering and fruiting is related to availability of water and the activities of predators. However, there are other constraints beyond the scope of our investigation. Other evidence indicates that flowering times may be conservative due to phylogenetic constraints (Kochmer & Handel 1986; Wright & Calderon 1995). Some proportion of the variation in timing of phenological events may be attributed to phylogeny and family membership, which influences many plant reproductive traits including flowering times (Wright & Calderon 1995) and fruiting synchrony (Gorchov 1990). For example, a number of the Acacia species have similar flowering and fruiting seasons. This overlap between species may have resulted not only because of similar responses to moisture, or anti-predator behaviour, but also because of a limited divergence of these species from ancestral patterns of flowering. Stone et al. (1998) argue that the long evolutionary history of East African savannas, and the geographically stable structure of Acacia communities, both suggest that long-term evolutionary responses, rather than ecological sorting, are responsible for the patterns seen in Tanzania. However, there can be little doubt that for most species, the phenology of flowering and fruiting is determined partly by genetic and partly by environmental factors. In addition, we cannot rule out the possibility of indirect effects. For example, rainfall appears to trigger the flowering and fruiting in most trees but it is possible that rainfall actually triggers pollinators and there may be selective pressure favouring trees that flower when pollinators are most active.

the availability of flowers and fruits for animals

The trees of the Serengeti ecosystem in a seasonal savanna habitat display a variety of flowering and fruiting strategies. Our data suggest that seasonal rainfall is the main hypothesis to explain the synchrony of reproduction, with evidence to support this in several species. Flowering in some tree species is correlated with rainfall and most of those that show more aseasonal flowering are those living in wetter habitats where the dependence on rainfall is less important. Thus, the influence of rain is modified by edaphic conditions. There is little to suggest that there is a sequence of staggered flowering or fruiting amongst species, as was shown for tropical forests (McClure 1966). Instead, most Acacia species show a relatively long period when flowers or fruits are present. Thus, the combined availability of edible pods for mammals covers most of the year.

Insects, including seed predators, reach peak abundance in the middle of the rains when there are the most seeds available for attack. Lamprey et al. (1974) proposed that Acacias have evolved adaptations to encourage consumption of pods by mammals, particularly in the indehiscent species. However, our observations show that even dehiscent pods are eaten on the tree by mammals so the beneficial effects may be more general. The seeds pass through the gut unharmed by the digestion process. This feeding has the double advantage of killing the bruchid larvae, thus increasing seed viability (from about 1% to 10–20%), as well as acting to disperse seeds (Lamprey et al. 1974; Ross 1979).

The availability of fruits eaten by birds in the commonest savanna trees (Commiphora and Ficus) occurs when there is a shortage of other foods (Sinclair 1975, 1978). The fruits are eaten rapidly, most disappearing in a few days after ripening. Thus, these trees could be supplying a food source that is critical for many bird species in this system.

In summary, reproduction of most species of Acacia shows a degree of synchrony greater than that predicted simply by the distribution of rain. The most likely function for this synchrony in many species is through some form of anti-predator adaptation. The hypotheses we tested are not mutually exclusive and they may interact in different ways. Our results suggest that synchrony in these tropical trees is caused by both abiotic and biotic factors that are inexorably intertwined. Closely related species have different reproductive patterns of synchrony that seem to be adapted to different combinations of rainfall, predators and dispersers. Rainfall is the primary determinant but the activities of predators and dispersers increase the degree of synchrony.


  1. Top of page
  2. Summary
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

We thank the directors and boards of trustees of the Tanzania National Parks and Tanzania Wildlife Research Institute for their permission to work in the Serengeti. S. Makacha, A. Nkwabi, Evelyn Turkington, Anne Sinclair and G. Sharam assisted with field data. Andrew MacDougall and Gary Bradfield provided helpful comments of earlier drafts. Support was provided by the Canadian Natural Sciences and Engineering Research Council, the National Geographic Society, and a Canada Council Senior Killam Fellowship to A.R.E.S.


  1. Top of page
  2. Summary
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  • Ashton, P.S. (1969) Speciation among tropical forest trees: some deductions in the light of recent evidence. Biology Journal of the Linnaean Society, 1, 155196.
  • Ashton, P.S., Givnish, T.J. & Appanah, S. (1988) Staggered flowering in the Dipterocarpaceae: new insights into floral induction and the evolution of mast fruiting in the aseasonal tropics. American Naturalist, 132, 4466.
  • Augspurger, C.K. (1981) Reproductive synchrony of a tropical shrub: experimental studies on effects of pollinators and seed predators on Hybanthus prunifolius (Violaceae). Ecology, 62, 775788.
  • Augspurger, C.K. (1983) Phenology, flowering synchrony, and fruit set of six neotropical shrubs. Biotropica, 15, 257267.
  • Belsky, A.J., Mwonga, S.M., Amundson, R.G., Duxbury, J.M. & Ali, A.R. (1993) Comparative effects of isolated trees on their undercanopy environments in high- and low-rainfall savannas. Journal of Applied Ecology, 30, 143155.
  • Burtt, B.D. (1929) A record of fruits and seeds dispersed by mammals and birds from the Singida district of Tanganyika Territory. Journal of Ecology, 17, 351355.
  • Chidumayo, E.N. (1997) Fruit production and seed predation in two Miombo woodland trees in Zambia. Biotropica, 19, 452458.
  • Colwell, R.K. (1973) Predictability, constancy, and contingency of periodic phenomena. Ecology, 55, 11481153.
  • Crawley, M.J. & Long, C.R. (1995) Alternate bearing, predator satiation and seedling recruitment in Quercus robur L. Journal of Ecology, 83, 683696.
  • Gorchov, D.L. (1990) Pattern, adaptation, and constraint in fruiting synchrony within vertebrate-dispersed woody plants. Oikos, 58, 168180.
  • Gwynne, M.D. (1969) Notes on the nutritive values of Acacia pods in relation to Acacia seed distribution by ungulates. East African Wildlife Journal, 7, 176178.
  • Herrera, C.M., Jordano, P., Guitian, J. & Traveset, A. (1998) Annual variability in seed production by woody plants and the masting concept: reassessment of principles and relationship to pollination and seed dispersal. American Naturalist, 152, 576594.
  • Hesla, B.I., Tieszen, H.L. & Boutton, T.W. (1985) Seasonal water relations of savanna shrubs and grasses in Kenya, East Africa. Journal of Arid Environments, 8, 1531.
  • Janzen, D.H. (1967) Synchronization of sexual reproduction of trees within the dry season in central America. Evolution, 21, 620637.
  • Janzen, D.H. (1974) Tropical blackwater rivers, animals and mast fruiting by Dipterocarpaceae. Biotropica, 6, 69103.
  • Kelly, D. (1994) The evolutionary ecology of mast seeding. Trends in Ecology and Evolution, 9, 465470.
  • Kelly, D., Harrison, A.L., Lee, W.G., Payton, I.J., Wilson, P.R. & Schauber, E.M. (2000) Predator satiation and extreme mast seeding in 11 species of Chionochloa (Poaceae). Oikos, 90, 477488.
  • Kelly, D. & Sork, V.L. (2002) Mast seeding in perennial plants: Why, How, Where? Annual Review of Ecology and Systematics, 33, 427447.
  • Kemp, P.R. (1983) Phenological patterns of Chihuahua desert plants in relation to the timing of water availability. Journal of Ecology, 71, 427436.
  • Kochmer, J.P. & Handel, S.N. (1986) Constraints and competition in the evolution of flowering phenology. Ecological Monographs, 56, 303325.
  • Koenig, W.D. & Knops, J.M.H. (2000) Patterns of annual seed production by northern hemisphere trees: a global perspective. American Naturalist, 155, 5969.
  • Koenig, W.D. & Knops, J.M.H. (2001) Seed-crop size and eruptions of North American boreal seed-eating birds. Journal of Animal Ecology, 70, 609620.
  • Lamprey, H.F. (1967) Notes on the dispersal and germination of some tree seeds through the agency of mammals and birds. East African Wildlife Journal, 5, 179180.
  • Lamprey, H.F., Halevy, G. & Makacha, S. (1974) Interactions between Acacia, bruchid seed beetles, and large herbivores. East African Wildlife Journal, 12, 8185.
  • McClure, H.E. (1966) Flowering, fruiting and animals in the canopy of a tropical rain forest. Malaysian Forester, 29, 192203.
  • Miller, M.F. (1994) The costs and benefits of Acacia seed consumption by ungulates. Oikos, 71, 181187.
  • Miller, M.F. (1996a) Acacia seed predation by bruchids in an African savanna ecosystem. Journal of Animal Ecology, 33, 11371144.
  • Miller, M.F. (1996b) Dispersal of Acacia seeds by ungulates and ostriches in an African savanna. Journal of Tropical Ecology, 12, 345356.
  • Miller, M.F. & Coe, M. (1993) Is it advantageous for Acacia seeds to be eaten by ungulates? Oikos, 66, 364368.
  • Newstrom, L.E., Frankie, G.W. & Baker, H.G. (1994) A new classification or plant phenology based on flowering patterns in lowland tropical rain forest trees at La Selva, Costa Rica. Biotropica, 26, 141159.
  • Norton, D.A. & Kelly, D. (1988) Mast seeding over 33 years by Dacrydium cupressinum Lamb. (rimu) (Podocarpaceae) in New Zealand: the importance of economies of scale. Functional Ecology, 2, 399408.
  • Norton-Griffiths, M., Herlocker, D. & Pennycuick, L. (1975) The patterns of rainfall in the Serengeti ecosystem, Tanzania. East African Wildlife Journal, 13, 347374.
  • Richards, P.W. (1952) The Tropical Rain Forest: an Ecological Study. Cambridge University Press, Cambridge.
  • Richards, P.W. (1996) The Tropical Rain Forest: an Ecological Study, 2nd edn. Cambridge University Press, Cambridge.
  • Ross, J.H. (1979) A Conspectus of the African Acacia Species. Memoirs of the Botanical Survey of South Africa No. 44. Botanical Research Institute, Pretoria, Republic of South Africa.
  • Schupp, E.W. (1988) Factors affecting post-dispersal seed survival in a tropical forests. Oecologia, 76, 525530.
  • Sharam, G. (2005) The decline and restoration of riparian and hilltop forests in the Serengeti National Park, Tanzania. PhD thesis, University of British Columbia, Vancouver.
  • Sinclair, A.R.E. (1975) The resource limitation of trophic levels in tropical grassland ecosystems. Journal of Animal Ecology, 44, 497520.
  • Sinclair, A.R.E. (1978) Factors affecting the food supply and breeding season of resident birds and movements of Palearctic migrants in a tropical African savannah. Ibis, 120, 480497.
  • Sinclair, A.R.E. (1979) The Serengeti environment. Serengeti: Dynamics of an Ecosystem (eds A.R.E.Sinclair & M.Norton-Griffiths), pp. 3145. University of Chicago Press, Chicago.
  • Sinclair, A.R.E., Mduma, S.A.R. & Arcese, P. (2002) Protected areas as biodiversity benchmarks for human impacts: agriculture and the Serengeti avifauna. Proceedings of the Royal Society, London B, 269, 24012405.
  • Stone, G.N., Willmer, P. & Rowe, J.A. (1998) Partitioning of pollinators during flowering in an African Acacia community. Ecology, 79, 28082827.
  • Toy, R.J. (1991) Interspecific flowering patterns in the Dipterocarpaceae in west Malaysia: implication for predator satiation. Journal of Tropical Ecology, 7, 4958.
  • Tybirk, K., Schmidt, L.H. & Hauser, T. (1993) Notes on soil seed banks of African acacias. African Journal of Ecology, 32, 327330.
  • Walters, M., Milton, S.J., Somers, M.J. & Midgley, J.J. (2005) Post-dispersal fate of Acacia seeds in an African savanna. South African Journal of Wildlife Research, 35, 191199.
  • Wright, S.J. & Calderon, O. (1995) Phylogenetic patterns among tropical flowering phenologies. Journal of Ecology, 83, 937948.