DEAD RETENTION IN :

A SOUTH AFRICAN PERSPECTIVE

Thesis for a BSc.(Hons) degree in Biological Sciences at the University of Cape Town 2016

Completed by: Alexandra Connolly Student Number: CNNALE002 Contact: [email protected]

Supervisor: Prof. Jeremy J. Midgley Biological Sciences Department, University of Cape Town

A amplexicaulis flower surrounded by dead and alive . Photo taken in Jonaskop, by Alexandra Connolly.

Acknowledgements

I would like to extend my sincerest thanks to Jeremy Midgley, my supervisor, for guidance and encouragement throughout this process and for having unwavering faith in my abilities. I would also like to express my gratitude for this opportunity which has been made possible by the Biological Sciences Department and the University of Cape Town. Lastly, I could not have reached this point without my family and I am incredibly humbled by their support throughout my academic career. Special thanks to my mother, Sandra Connolly, without whom none of this would be possible.

Plagiarism Declaration

I, Alexandra Connolly, know that plagiarism is wrong. Plagiarism is to use another’s work and pretend it is your own. Any statements in this thesis derived from the work of others has been attributed to them, and cited and referenced as such. The work in this thesis is my own and I have not copied phrases, sentences, or longer tracts word-for-word from other sources.

Signature: ______Alexandra Connolly 11/11/2016

ABSTRACT

Dead leaf retention has been observed and investigated in the genus of Australian Proteaceae, with the two main hypotheses suggesting that dead leaves provide an advantage with regards to selfish fertilization and increased flammability. Protea amplexicaulis, a rodent pollinated found in the Cape Floristic Region of South Africa, has been observed to retain dead leaves for up to six years, the adaptive significance of which has yet to be explored. Using this species as a South African case study, this research explores the two previously suggested hypotheses by means of evaluating post-fire seedling growth (selfish-fertilization hypothesis) and by recording the nearest neighbour to 60 P.amplexicaulis individuals and theoretically assessing the effect of increased fire intensity on them (Flammability hypothesis). There was no evidence in support of either of these hypotheses. A novel hypothesis is proposed and preliminarily explored in this study, which suggests that dead leaf retention allows for reduced visual detection of flowers and enhanced below canopy shelter, thereby potentially providing shelter-enrichment for rodent pollinators or decreasing the prevalence of nectar robbing by illegitimate pollinators such as local sunbirds. Dead leaves were found contribute to 12.8% of below-canopy shelter when below- canopy images were analysed before and after the removal of all dead leaves. Furthermore, the effect of dead leaves on the visibility and accessibility of flowers within the P. amplexicaulis shrub was investigated by assessing changes in flower location with dead leaf removal, and was found to significantly increase the proportion of flowers completely hidden from view. In conclusion, there is preliminary evidence to suggest that marcescence developed in P. amplexicaulis as a strategy to reduce nectar robbing by illegitimate pollinators, and potentially as a strategy for increasing the time spent pollinating by rodents via means of shelter-enrichment. The applicability of this finding is explored with regards to other Cape Protea species and to Australian Banksia.

KEY WORDS Crypsis, Leaf-Longevity, Marcescence, Proteaceae, Therophily.

INTRODUCTION

Marcescence whereby leaves remain on stems after they senesce, also known as dead leaf retention, is observable in many species and is thought to provide a variety of functional advantages. In New Zealand, dead leaves surrounding the trunks of Cabbage trees (Cordyline australis) were found to be an adaptation to frost, with the marcescent skirt functioning as insulation against chilling temperatures (Harris et al. 2004). A similar functional advantage was suggested for Espeletia spp (Smith 1979). Dense marcescent skirts were also found to produce higher relative humidity at the surface of the stem thereby protecting the stem from desiccation in drought periods (Smith 1979). Suggestions pertaining to nutrients cycling and competition have also been put forth. Otto and Nilsson (1981) hypothesised that dead leaf retention during winter months allows for the preservation of nutrients in marcescent leaves until the spring growth period, at which time nutrients from litter can be rapidly recycled by the tree instead of lost to soil leaching. However, it has since been found that marcescence is unlikely to be attributable to a nutrients cycling strategy (Escudero & del Arco 1987). Another hypothesis suggested that staggered litter fall would create a thick blanket of dead leaves on the soil thus restricting or completely inhibiting growth of competing and seedlings (Nilsson 1983). While many of these hypotheses are likely explanations for observed marcescence, they have limited applicability, mostly restricted to trees or species in areas with strong climatic fluctuations such as tropical alpine regions. The adaptive significance of dead leaf retention in other habitats and types has been poorly researched.

Marcescence has recently been observed in the Proteaceae family in both Australia and the Cape Floristic Region (CFR) in South Africa. While very rarely observed in Cape Protea species, marcescence has been observed in 25 species of the Banksia genus in Australia and is thought to have proliferated in the Miocene approximately 25.9 million years ago (He et al. 2011). The adaptive significance of this trait in Proteaceae has remained uncertain, however due to the similarities between the CFR and Australian Proteaceae species and habitats, the selective pressure was likely the same across the family. Given the nutrient poor, often water stressed, environment in which these species occur, retaining dead leaves likely incurs costs relating to nutrients cycling and mechanical support (Escudero & del Arco 1987; Midgley & Bond 2011). Marcescence is thus likely to have a selection-based evolutionary explanation, such as providing a functional advantage, rather than representing an ecophysiological trade- off. However, as leaf abscission time is a fundamental ecological plant characteristic, and

correlated with multiple physiological and morphological traits (Reich et al. 1992), as well as habitat (Escudero & del Arco 1987), marcescence could be a result of the plants overall leaf longevity strategy. Dead leaf retention could thus correlate with live leaf retention time, which has been found to correlate with the degree of retention of serotinous cones (Midgley & Enright 2000).

So far, only two hypotheses have been put forward regarding the adaptive significance of marcescence in Proteaceae, both of which were suggested by He and colleagues (2011) seeking to explain the trait in Australian Banksia. Fire is an ancient and major driver in Australian ecosystems, with many plant species thought to have developed certain traits, such as pyriscent serotiny, as fire-adaptations (He et al. 2011; Keeley et al. 2011). However, whether these traits are actually fire adaptations rather than exaptations is still debated (Mutch 1970, Snyder 1984). Plants in the CFR are also thought to be fire-adapted, with many species dependant on fire for seed germination and flowering (De Lange & Boucher 1990; Brown et al. 2003; Lamont & Downes 2011). Flammability is one such trait which is thought to have evolved in fire prone systems, with increased individual flammability causing a localized spike in fire intensity. According to the kill thy neighbour hypothesis (Bond & Midgley 1995), this will benefit the individual by killing the neighbouring plants and thereby providing a space of low competition in which its seedlings can establish. The low water content of dead leaves would undoubtedly impact local fire intensity, and this is the basis of the flammability hypothesis put forward for marcescence (He et al. 2011). However, this hypothesis predicts that neighbouring plants are mostly fire-avoiders rather than fire-adapted reseeders or resprouters which would senesce in low-intensity fires anyway. He and colleagues (2011) also suggest that the complete incineration of dead leaves would result in a highly localized release of nutrients into the soil below the canopy of the mother plant, and thus create a fertilized micro-climate in which seedlings can establish and thrive. If this is correct, one would expect more seedlings to survive beneath the canopy of a burnt mother plant and that they would exhibit increased growth due to the additional nutrients. The selfish-fertilization hypothesis assumes extremely limited dispersal of nutrients and seeds post fire, for which there is evidence to the contrary (Smith 1970; Grier 1975; (Slingsby & Bond 1983; Bond 1988; Auld & Denham 1999; He et al. 2004).

Protea amplexicaulis is a low sprawling shrub with serotinous cones which is endemic to the CFR and found mostly on north-facing sandstone slopes, often forming extensive dense stands (Rourke 1980). It has been observed to retain its dead leaves for up to six years

(Midgley & Bond 2011), and is thus one of the few marcescent Cape Protea species. This species is classified as a rodent sugarbush (Rebelo 1995), possessing traits thought to be adaptations to non-flying mammal pollination such as bowl shaped heads, copious production of sucrose-rich nectar, and a distinctive yeasty or musty odour (Turner 1982; Rebelo & Breytenbacht 1987). With this in mind, I suggest an alternative selective force for the development of marcescence in this species and potentially in the broader Proteaceae family. The retention of dead leaves notably contributes to the fullness of the shrubbery by which the understory of the plant is almost completely hidden from view. It is therefore plausible that marcescence allows for increased below-canopy shelter for both flowers and rodent pollinators. This hypothesis thus suggests that marcescence contributes to the overall crypsis of flowers in rodent-pollinated (therophilous) Proteaceae. Crypsis refers to a reduction in visual detection of flower heads, and is considered an explanation of the inflorescence positioning common in therophilous (Rourke & Wiens 1977; Wiens et al. 1983; Rebelo & Breytenbacht 1987). The adaptive significance of crypsis has yet to be determined, however Wiens and colleagues (1983) suggested that inflorescent positioning which aids to conceal flowers, hence a strategy of crypsis, may have evolved as a means to protect rodent pollinators from avian predators or in order to decrease the prevalence of nectar robbing by illegitimate bird pollinators.

Following along similar lines, if the retention of dead leaves increases the density of the plant’s canopy, thus providing increased below-canopy protection, rodents would likely spend more time beneath the plant canopy due to enhanced perceived safety. The preference for shelter-enriched microhabitats by rodents has been repeatedly observed (Sih 1980; Longland & Price 1991; Bowers & Dooley 1993; Manson & Stiles 1998; Muñoz et al. 2009) and shown to cause differential seed selection (Perea et al. 2011), seed fate (Sivy et al. 2011) and overall foraging behaviour (Brown et al. 1988; Kotler et al. 2002). As such, an increase in perceived shelter would likely result in more time spent pollinating, and hence affect the seed set and overall reproductive success of the plant. The second crypsis hypothesis put forward by Wiens and colleagues (1983) is also potentially applicable to marcescence. As P. amplexicaulis, as well as an array of other Proteaceae, is strictly therophilous (Wiens et al. 1983), visitation by illegitimate pollinators such as birds would result in a loss of nectar without effective pollination. Nectar robbing is known to decrease the reproductive fitness of plants, which can be measured by assessing seed set (Traveset et al. 1998; Irwin et al. 2001; Burkle et al. 2007; Irwin et al. 2010). Therophilous Proteas are thought to have an effective

10mm stigma-nectar distance, which is perfectly tailored to suit the rostrum length of the main rodent pollinators (Wiens et al. 1983), but as such the prized nectar is also accessible to local birds such as the orange-breasted sunbird (Anthobaphes violacea). The retention of dead leaves may not only increase the crypsis of flowers within the shrub, hence preventing the attraction of birds, but could further deny avian access to flowers due to the dense shrubbery surrounding flower heads. This nectar robbing hypothesis is also applicable to illegitimate insect pollinators which rely on visual cues.

Crypsis of flowers may also provide protection from nectar predation by larger mammals, such as the Cape Baboon (Papio ursinus). Baboons is the region are known to rely mostly on seeds and herbaceous leaves for nutrition, with the nectar from flowers constituting only 3% of their usual diet (Davidge 1978; Johnson et al. 2013). However, substantial baboon predation was observed in P. amplexicaulis stands in the Bainskloof area of the , South Africa. As this area recently burned, the local troop likely experienced a food shortage and hence turned to the sucrose-rich nectar of local sugarbush Proteas in order to make up for the loss of herbaceous matter. However, if this substantial nectar predation is a fairly common occurrence, it is possible that the crypsis strategy of P. amplexicaulis is an attempt to conceal flowers from baboons and thus reduce nectar predation.

This research aims to broadly explore the adaptive significance of dead leaf retention in Proeteaceae by using the marcescent P. amplexicaulis as a South African case study. I will first explore the extent of marcescence in Cape Proteas and assess whether there is evidence for dead leaf retention being a result of an overarching leaf longevity strategy. The applicability of the flammability and selfish-fertilization hypotheses put forth by He and colleagues (2011) will be investigated with regards to the CFR, as well as their theoretical applicability in Australia. The further objective of this study is to preliminary investigate the basis of a crypsis hypothesis for marcescence by assessing the effect of dead leaf retention on below canopy shelter enrichment and on flower crypsis with regards to visual cues and accessibility to nectar robbers. Lastly, I will investigate the extent of baboon predation on P. amplexicaulis flowers and assess whether marcescence may provide a crypsis advantage in this case.

METHODS

Sampling Sites Two sampling sites were chosen in which P. amplexicaulis is known to occur in dense stands, along with many other South African Protea species. The first was on the south-west facing slopes of Bainskloof pass, Western Cape, South Africa (-33.62818 S; 19.09990 E, approximately 586m above sea level). The vegetation here is mostly mountain fynbos with little presence of alien species, the habitat is fairly undisturbed, although pollution does occur close to the road. The majority of sampling was completed on the steep slope to the west of the R301 near the main rest stop, as this area is inhabited by mature Proteas. The east side of the road at the same location provides a post-fire environment as a large area of the Bainskloof vegetation burnt in a natural fire in late April 2015. The second site of research was on the north facing slopes of Jonaskop, a well-known mountain in the Riviersonderend mountain range (-33.58107 S; 19.30219 E, approximately 1000m above sea level). Mature P. amplexicaulis plants are found scattered on the relatively flat area below the summit, locally known as Jonasplaats, which occurs along the east side of the access road. The vegetation here is protected, with restricted access, and as such is undisturbed. Sampling took place at both localities during P. amplexicaulis late-winter flowering in August 2016.

Leaf Longevity Strategies The leaf retention strategies of seven Cape Protea species (P. amplexicaulis, Protea. humiflora, Protea. laurifolia, Protea lorifolia, Protea nana, , and ) were assessed with regards to the maximum age of retained leaves both dead and alive. Ten individuals from each species were randomly selected and assessed per site, except for those only found at one site (P. humiflora and P. lorifolia only found in Jonaskop) for which only 10 individuals were assessed in total. Leaf age can be inferred, with an accuracy of 1 year, from the annual stem growth of the plant, which is marked by node scarring or by branching events in Cape Proteaceae (J.J. Midgley pers. comm.) A strict rule was enforced, by which only nodes with more than three fully expanded leaves attached to the relevant stem would be included in leaf longevity counts.

Nearest Neighbour Investigation In order to investigate whether the kill thy neighbour hypothesis (Bond & Midgley 1995) is applicable to P. amplexicaulis individuals, the nearest neighbours of 30 P. amplexicaulis

plants were recorded per site. The nearest neighbouring species was only recorded if it occurred within a 2-meter radius of the plant as anything further would not be directly affected by increased flammability. Only woody, canopy plants were recorded as neighbours as these would provide the competition needed for the kill thy neighbour hypothesis. Other P. amplexicaulis plants were recorded as nearest neighbours if applicable.

Post-fire Seedling Growth Protea amplexicaulis seedling establishment and growth post-fire was investigated to determine whether seedlings do better beneath the canopy of burnt mother plants, as predicted by the selfish-fertilization hypothesis (He et al. 2011). Sampling took place at the Bainskloof site, in an area which burnt 16 months prior. A burnt P. amplexicaulis individual (See figure 1) was chosen randomly within the matrix. The P. amplexicaulis seedlings (See figure 2) occurring within a 1m2 transect around the center of the burnt individual’s canopy were then measured for absolute height using a calliper. The same measurements were then taken for seedlings falling within a 1m2 transect 1.5 meters (center -center point) horizontally to the right of the parental transect. This was done as to assure no change in water and nutrients flow down the slope and to avoid sampling bias. If another burnt P. amplexicaulis individual was found to the right, then the transect was moved to the left of the original burnt individual. This was repeated for a total of 36 paired transects.

Figure 1: The remains of a burnt P. Figure 2: A P. amplexicaulis seedling at the amplexicaulis individual in Bainskloof, 16 Bainskloof site, 16 months post-fire. months post-fire.

Changes in Below-Canopy Shelter The effect of dead leaf retention on below-canopy exposure in P. amplexicaulis was assessed by quantifying the change in light exposure before and after the pruning of dead leaves. At the Jonaskop site, 20 randomly selected P. amplexicaulis were sampled. Using a GoPro HERO4 camera, an image from the base of the plant, facing the sky, was taken and analysed using Image.J (Schneider et al. 2012) in order to estimate the percentage of sky visible. With the camera kept in place, the shrub was then stripped of all dead leaves, and another image was taken and analysed. Before the repeat photo was taken, the branches were encouraged to fall back into their natural positions as they were often moved around during defoliation.

Crypsis of Flower Heads Before defoliation of the 20 P. amplexicaulis plants used for image analysis, the locations of flowers produced this year were categorised in terms of crypsis. All fresh flowers within the shrubs were located and categorised into: 1) Completely exposed with the center of the flower clearly visible and accessible without tampering of branches or leaves (Figure 3A); 2) Partially covered whereby the flower is somewhat visible but would not be accessible without moving branches and leaves (Figure 3B); and 3) Hidden within the canopy of the plant and only found with extensive searching. This categorisation was then redone once dead leaves had been pruned from the plant. An additional ten individuals were sampled in Jonaskop for this analysis.

Figure 3: Examples of the “flower location” categories used during sampling of P. amplexicaulis. Flowers which were fully visible were labelled as ‘exposed’ (A); visible but not accessible without movement of branches as ‘partially covered’ (B); and completely concealed from view as ‘hidden’ (not depicted here).

Baboon Predation Protea amplexicaulis was observed to be heavily predated upon by Chacma baboons (Papio ursinus), with evidence of fresh flower heads that were ripped off bushes and the nectar eaten, in the Bainskloof sites. The rate of predation by baboons on P. amplexicaulis individuals in Bainskloof was quantified by counting the number of ripped-off fresh flower heads laying within a 3m2 transect in dense P. amplexicaulis areas. The total number of this- year’s flowers was also quantified for all P. amplexicaulis individuals whose canopy lay within the transect. Ten transects were randomly chosen and sampled. Within the same site, 30 individuals displaying evidence of baboon predation were assessed with regards to flower location, as previously described. Furthermore, in order to quantify the rate of baboon predation on other Proteas which produce substantial amounts of nectar, the ratio of picked to intact flowers was estimated for P. repens and P. nana within close proximity to the P.amplexicaulis transects. A random survey of 100 P. repens flowers was conducted within the Bainskloof site, as well as a survey of 39 P. nana flowers. The sample size of P. nana flowers was limited by the number of individuals present within the site.

Data Analysis All formal statistical testing was completed in R statistical software (R Core team, 2016). The maximum age of dead and alive leaves of the seven sampled Cape Proteas were averaged per species. The live-leaf ages of P. amplexicaulis and P. humiflora were then separately compared to the overall live-leaf longevity of the other five species using Mann-Whitney- Wilcoxon tests for independent, non-normal samples (Hollander & Wolfe 1973). Using a Wilcoxin signed-rank test for paired samples (Hollander & Wolfe 1973), the number of seedlings found beneath mother-plant canopies and within the matrix was compared. The difference in the average height of seedlings between parental and matrix samples was assessed in the same way, however the sample size was drastically reduced (n=12) as many paired transects did not have seedlings present in both transects. Outputs from the Image.J analyses were used to investigate the difference between below-canopy light with and without dead leaves, using a students paired t-test. The flower locations of the 30 individuals before and after dead leaf removal were compared by means of a Chi-Squared test with three categories. Similarly, the flower locations of the 30 individuals within the baboon predation transect in Bainskloof were compared to the 30 pre-defoliation individuals in Jonaskop.

RESULTS

Of the seven Protea species which were assessed for their leaf longevity strategy, only P. amplexicaulis and P. humiflora were found to retain dead leaves. These species retained live leaves for a similar period of time, with P. amplexicaulis retaining live leaves for 2.7 ± 0.16 years and P. humiflora for 2.9 ± 0.16 years. However, their dead leaf retention strategies seem to differ, with P. humiflora retaining dead leaves for longer than live leaves on average (mean dead leaf retention = 3.5 ± 0.24 years), while P. amplexicaulis has a similar live and dead leaf strategy with a mean dead leaf retention of 2.8 ± 0.18 years (Figure 4). The live leaf retention strategies of P. amplexicaulis and P. humiflora were not found to be statistically different to the average of the other five species (W = 986, n1 = 20, n2 = 80, p = 0.097; W =

451.5, n1 = 20, n2 = 80 and p = 0.51 respectively). P.laurifolia’s leaf longevity strategy stands out from the other 6 species, with live leaves retained for up to 8 years (mean = 4.6 ± 0.37 years) (Figure 4).

Figure 4: Leaf retention strategies of seven Protea species sampled at the Bainskloof and Jonaskop sites. Bars represent the maximum age in years of dead and alive leaves, averaged over n individuals per species. Error bars represent standard error.

As shown in Figure 5, P. repens was by far the most common woody species found within 2 meters of P. amplexicaulis individuals at both sites, with the second most common being another P. amplexicaulis plant. There were two occurrences of unidentified woody neighbours, and three occurrences of no woody neighbour within a 2-meter radius of P. amplexicaulis individuals at the Jonaskop site (Figure 5). Analysis of P. amplexicaulis seedling establishment in a post-fire environment showed no significant difference in the number of seedlings established beneath the canopy of a burnt individual (42 total seedlings)

and within the vegetation matrix 1.5m away (46 total seedlings) (, V = 128, n = 36, p = 0.771). The average height of the seedlings per transect were also found to be statistically similar (V = 41, n = 12, p = 0.910), with an overall mean height of 50.2 ± 17.58 mm (n = 42) in the transects below burnt P. amplexicaulis individuals and 50.5 ± 17.96 mm (n = 46) in the matrix (Figure 6).

Figure 5: The nearest woody plant species within a 2m radius of P. amplexicaulis individuals sampled at the Bainskloof and Jonaskop sites (n= 30 P.amplexicaulis individuals per site).

Figure 6: The height (mm) of established P. amplexicaulis seedlings averaged over 36 paired 1m2 transects below burnt P. amplexicaulis individuals and 1.5m away in the matrix, as sampled at the Bainskloof post-fire site in August 2016. Error bars represent standard error.

The removal of dead leaves from P. amplexicaulis individuals resulted in a significant 12.8 ±

7.17 % mean increase in below canopy light exposure (t(19, 1) = -8.02, p < 0.001) (Figure 7)

2 and a significantly different distribution with regards to flower exposure (χ (2) = 156.59 p < 0.001). The percentage of flowers completely hidden within the canopy decreased from 68% to 38%, while the number of fully exposed flowers increased by 13% and partially covered increased by 17% (Figure 8).

A B C

Figure 7: One of the P. amplexicaulis individuals sampled at the Jonaskop site (A) and the below- canopy sky visibility before (B) and after (C) the removal of dead leaves.

Figure 8: The effect of dead leaf removal (‘Defoliated’) on the distribution of flowers within P. amplexicaulis shrubs, with regards to visual exposure. ‘Exposed’ = entire flower easily accessible without moving branches or leaves; ‘Partially Covered’ = visual evidence of a flower but not fully visible or accessible without moving branches or leaves; ‘Hidden’ = completely cryptic within the canopy. Overall proportions were averaged for 30 individuals at the Jonaskop site (n= 327 flowers).

There was evidence for substantial predation on P. amplexicaulis flowers by baboons at the Bainskloof site as seen by the littering of ripped off flower heads. Within this site, the number of evidently picked fresh flowers was almost half of this year’s flowering output (mean = 47.5 ± 2.01 %; n= 763). This is much higher than the baboon predation evident on other sugarbushes cohabiting this area, with P. repens having a 19% predation rate (n=100), and

P. nana showing no evidence of flower predation (n=39). In order to assess whether baboon predation was affected by flower location, the exposure distributions were compared between predated upon individuals in Bainskloof and untouched individuals in the Jonaskop site (n1 =

349 flowers/ 30 individuals; n2 = 347 flowers/ 32 individuals, respectively). There was no significant difference between the flower location distribution of plants subjected to high

2 levels of baboon predation and those subjected to no predation (χ (2 = 2.86, p = 0.239). The exposure distribution of flowers in baboon predated individuals was similar to that of the ‘normal’ distribution in the Jonaskop site as shown in figure 4, with the biggest shift being a 2% reduction in the number of fully exposed flowers.

DISCUSSION

The results of this study support the hypothesis that marcescence in Cape Protea species is decoupled from the plant’s leaf longevity strategy and further suggests that this trait may have developed as a crypsis-related adaptation. Only P. amplexicaulis and P. humiflora, the two non-flying mammal pollinated Protea sampled, were observed to retain their dead leaves (Figure 4). Furthermore, the live-leaf longevity of P. amplexicaulis and P. humiflora was not statistically different to that of the average live-leaf longevity of the other six sampled species, which implies that the live leaf strategies of these two species are not a function of their dead leaf retention. Ultimately, this indicates that the observed trait of marcescence in Cape Proteaceae is a unique trait which is unlikely to represent an ecophysiological trade-off. As such, the trait must have been selected for due to a fitness advantage.

The two previously suggested hypotheses for the development of marcescence in Proteaceae, albeit originally suggested for Australian Banksia species (He et al. 2011), have little support in the Cape. The suggestion that dead leaves are retained as they increase flammability (He et al. 2011), and thus cause a momentary spike in fire intensity killing neighbouring plant competitors and providing space for post-fire seedling establishment (Bond & Midgley 1995), is an unlikely explanation for marcescence in P. amplexicaulis. This species is found neighbouring P. repens for the most part, and the other 22% of the sample population was found in dense P. amplexicaulis stands with its closest neighbour being another P. amplexicaulis plant or a grassy species (Figure 5). As both P. repens and P. amplexicaulis are reseeders and not fire-avoiders, there is no real advantage of increased flammability because these species would die and reseed in a relatively low-intensity fire. Many Australian plant

species are thought to be highly adapted to the ancient fire regime of Australia (He et al. 2011), and hence the plants found neighbouring marcescent Banksia species would likely also have fire adaptations such as the ability to reseed or resprout post-fire. This lack of fire- avoiding species would make the selfish flammability hypothesis an unlikely explanation for marcescence in Banksia as well. In order to fully investigate this hypothesis, one could conduct flammability experiments on individuals with and without dead leaves and assess the effect of the resulting fire intensity on neighbouring species in both South Africa and Australia.

The second hypothesis put forward by He and colleagues (2011) is that the nutrients retained in dead leaves provides a direct reproductive benefit because when the leaves burn they release nutrients into the soil below, thus creating an ideal micro-habitat for seedling establishment. However, the data shows no support for this as there was no difference between the number of P. amplexicaulis seedlings established beneath mother-plant canopies in comparison to the vegetation matrix, and furthermore there was no seedling growth benefit beneath the maternal canopy (Figure 6). This investigation into seedling growth below burnt mother plants is notably robust as there in no way to link seedlings to parents and as such the seedlings found below one individual may have been dispersed from another individual. Nutrient-rich ash is often dispersed by wind and water post-fire (Smith 1970; Grier 1975) and Proteaceae seeds commonly show adaptations for multiple dispersal strategies (Slingsby & Bond 1983; Bond 1988; Auld & Denham 1999; He et al. 2004). Therefore, an explanation based on the idea of highly limited ash and seed dispersal is not applicable to the habitats of both Australian and South African Proteaceae. Although unnecessary, this hypothesis could be further tested by assessing soil nutrient contents below burnt individuals at various times since last fire.

Marcescence as a strategy for reduced flower predation by illegitimate pollinators is the most likely explanation given the results of this study. Due to the dense, shrub-like canopy of P. amplexicaulis, illegitimate pollinators such as birds would have limited access to flower heads, and as such only be able to steal nectar from fully exposed, and to a lesser extent partially exposed, flowers. The significant shift in flower exposure once dead leaves had been removed from P. amplexicaulis individuals (Figure 8) highlights the potential role of marcescence in reducing the prevalence of nectar robbing. Both before and after defoliation, the majority of flowers were fully hidden within the canopy. While this is commonly observed in many rodent pollinated species and likely attributable to the need for flowers to

be close to the ground and accessible to rodents (Rebelo & Breytenbacht 1987), it is plausible that this provides an additional benefit of increased crypsis, especially considering that the rodents in question are known to be good climbers. Furthermore, a 30% decrease in the number of flowers potentially exposed to birds was due solely to the cover provided by dead leaves. The foundation for the nectar robbing hypothesis has therefore been preliminarily supported. However, this study cannot conclusively support the hypothesis as both an observational study and an assessment of the effect of flower crypsis on seed set is required.. It is worth noting that a reduction of nectar predation by illegitimate pollinators is applicable to insect pollinators as well, and is worth investigation. The nectar robbing hypothesis may be equally applicable to marcescent Banksia as many of the species observed to retain dead leaves (He et al. 2011) possess traits synonymous with non-flying mammal pollination (Hopper 1980; Cunningham 1991; Goldingay et al. 1991; George 1984; Hackett & Goldingay 2001) and are potentially predated upon by illegitimate bird pollinators.

While marcescence may provide protection from bird predators, there is no evidence that it reduces the rate of predation by baboons. The notably high rate of predation by baboons in Bainskloof may be attributable to a reduction in food resources due to the recent fire and potentially due to the current drought. Although P. amplexicaulis was subjected to substantially higher baboon predation than surrounding sugarbushes, it exhibits no indication of preventative strategies with regards to the crypsis of flowers. This may be because significant baboon predation is a rare event or because baboons would likely gain access to hidden flowers by disturbing the shrubbery. Investigating the shelter-enriching potential of marcescence indicated that the retention of dead leaves provided a 12.8% exposure reduction for rodents beneath the canopy. While this is a significant change, it seems rather low. However, it is possible that this change is enough to encourage increased time spent pollinating by rodents as it is the perceived safety from predators that is in question, and not the absolute change in exposure. To properly investigate whether marcescence promotes increased time spent pollinating by non-flying mammals, a study which assess the effect of marcescence on seed set is required. This would need to take place roughly three months after prime flowering and hence did not fall within the timeline of this research. Shelter enrichment has been shown to affect a wide-range or rodents (Brown et al. 1988; Manson & Stiles 1998; Kotler et al. 2002) and would therefore be a potential explanation for marcescence in the rodent pollinated Proteaceae of both the Cape and Australia.

Beyond investigating the basis of multiple hypotheses relating to marcescence in Proteaceae, this study highlights a need for interdisciplinary thought when considering evolutionary mechanisms. As the research by He and colleagues (2011) was rooted in the field of fire ecology, the resulting explanations were naturally based on knowledge of fire and fire- adapted systems and hence were limited to a fire-adaption mechanism for the evolution of the retention of dead leaves. However, if a biologist from a different field of study were to consider the trait, they may see potential links to a multitude of other mechanisms. This study is an example of such interdisciplinarity as the observation of marcescence in P. amplexicaulis by Midgley and Bond (2011) was considered from a pollination point of view and later linked to ideas regarding flower crypsis originating with Wiens and colleagues in 1983. If the crypsis hypotheses are in fact further supported by future research, this may have implications for other species with potentially crypsis related traits. Protea recondita, for example, is a rodent pollinated Protea found in the Cape and has been observed to have extended bracts which curve around the inflorescent, somewhat concealing the flower. Following along the lines of the crypsis hypotheses suggested in this study, this trait may also represent a strategy for increased crypsis as could a number of other unexplained traits in non-flying mammal pollinated plants.

In conclusion, there is preliminary evidence to suggest that marcescence developed in P. amplexicaulis as a strategy to reduce nectar robbing by illegitimate pollinators such as birds by obscuring flowers within the canopy, and potentially as a strategy for increasing time spent pollinating by rodents via means of shelter-enrichment. This finding is in support of the hypotheses regarding crypsis suggested by Wiens and colleagues (1983). More research is required to conclusively support both the nectar robbing and shelter enrichment hypotheses, as well as to link this to marcescence in other Cape Proteas such as P. humiflora. These explanations should also be further explored in marcescent Australian Banksia, as the previously suggested selfish-fertilization and flammability hypotheses have scant support in South Africa and questionable mechanistic validity in Australia.

REFERENCES

Auld, T.D. and Denham, A.J.,1999. The role of ants and mammals in dispersal and post- dispersal seed predation of the shrubs (Proteaceae). Plant Ecology, 144(2): 201- 213. Bond, W.J. and Midgley, J.J., 1995. Kill thy neighbour: an individualistic argument for the evolution of flammability. Oikos, 73: 79–85. Bond, W.J., 1988. Proteas as 'tumbleseeds': wind dispersal through the air and over soil. South African Journal of Botany, 54: 455- 460. Bowers, M. A. and Dooley, J. L., 1993. Predation hazard and seed removal by small mammals: microhabitat versus patch scale effects. Oecologia, 94: 247-254. Brown, J.S., Kotler, B.P., Smith, R.J., and Wirtz II, W.O., 1988. The effects of owl predation on the foraging behavior of heteromyid rodents. Oecologia, 76(3): 408-415. Brown, N.A.C., Van Staden, J., Daws, M.I., Johnson, T., and van Wyk, A.E., 2003. Patterns in the seed germination response to smoke in plants from the Cape Floristic Region, South Africa. South African Journal of Botany, 69(4): .514-525. Burkle, L.A., Irwin, R.E., and Newman, D.A., 2007. Predicting the effects of nectar robbing on plant reproduction: implications of pollen limitation and plant life-history traits. American Journal of Botany, 94: 1935-1943. Davidge, C., 1978. Ecology of Baboons (Papio Ursinus) at Cape Point. Zoologica Africana, 13(2): 329-350. Cunningham, S.A., 1991. Experimental evidence for pollination of Banksia spp. by non- flying mammals. Oecologia, 87(1): 86-90. De Lange, J.H. and Boucher, C., 1990. Autecological studies on Audouinia capitata (Bruniaceae). I. Plant-derived smoke as a seed germination cue. South African Journal of Botany, 56(6): 700-703. Escudero, A. and Del Arco, J.M., 1987. Ecological significance of the phenology of leaf abscission. Oikos, 11-14. George A.S., 1984. The Banksia book. Kangaroo Press, Syndey, Australia. 240p. Goldingay, R.L., Carthew, S.M., and Whelan, R.J., 1991. The importance of non-flying mammals in pollination. Oikos, 79-87. Grier, C.C., 1975. Wildfire effects on nutrient distribution and leaching in a coniferous ecosystem. Canadian Journal of Forest Research, 5(4): 599-607. Hackett, D.J. and Goldingay, R.L., 2001. Pollination of Banksia spp. by non-flying mammals in north-eastern New South Wales. Australian Journal of Botany, 49(5): 637-644. Harris, W., Beever, R.E., Parkes, S.L., and McSeveny, T., 2004. Genotypic variation of dead leaf retention by Cordyline australis (Lomandraceae) populations and influence on trunk surface temperatures. New Zealand Journal of Botany, 42(5): 833-845. He, T., Krauss, S.L., Lamont, B.B., Miller, B.P., and Enright, N.J., 2004. Long‐distance seed dispersal in a metapopulation of inferred from a population allocation

analysis of amplified fragment length polymorphism data. Molecular ecology, 13(5): 1099- 1109 He, T., Lamont, B.B., and Downes, K.S., 2011. Banksia born to burn. New Phytologist, 191: 184–196.. Hopper, S.D., 1980. Bird and mammal pollen vectors in Banksia communities at Cheyne Beach, Western Australia. Australian Journal of Botany, 28(1): .61-75. Irwin, R.E., Brody, A.K., and Waser, N.M., 2001. The impact of floral larceny on individuals, populations, and communities. Oecologia, 129:161-166. Irwin, R.E., Bronstein, J.L., Manson, J.S. and Richardson, L., 2010. Nectar robbing: ecological and evolutionary perspectives. Annual Review of Ecology, Evolution and Systematics, 41(271): e292. Johnson, C.A., Raubenheimer, D., Rothman, J.M., Clarke, D., and Swedell, L., 2013. 30 Days in the Life: Daily Nutrient Balancing in a Wild Chacma Baboon. PLoS ONE 8(7): e70383. Keeley, J.E., Pausas, J.G., Rundel, P.W., Bond, W.J. and Bradstock, R.A., 2011. Fire as an evolutionary pressure shaping plant traits. Trends in Plant Science, 16(8): 406-411. Kotler, B.P., Brown, J.S., Dall, S.R., Gresser, S., Ganey, D. and Bouskila, A., 2002. Foraging games between gerbils and their predators: temporal dynamics of resource depletion and apprehension in gerbils. Evolutionary Ecology Research, 4(4): 495-518. Lamont, B.B. and Downes, K.S., 2011. Fire-stimulated flowering among resprouters and geophytes in Australia and South Africa. Plant Ecology, 212(12): 2111-2125. Longland, W.S. and Price, M.V., 1991. Direct observations of owls and heteromyid rodents: can predation risk explain microhabitat use? Ecology, 72(6): 2261-2273. Manson, R. H. and Stiles, E. W., 1998. Links between microhabitat preferences and seed predation by small mammals in old fields. Oikos, 82: 37-50. Midgley, J.J. and Bond, W.J., 2011. Pushing back in time: the role of fire in plant evolution. New Phytologist, 191:5–7. Midgley, J.J. and Enright, N.J., 2000. Serotinous species show correlation between retention time for leaves and cones. Journal of Ecology, 88: 348-351. Muñoz, A., Bonal, R., and Díaz, M., 2009. Ungulates, rodents, shrubs: interactions in a diverse Mediterranean ecosystem. Basic and Applied Ecology, 10: 151-160. Mutch, R.W., 1970. Wildland Fires and Ecosystems‐‐A Hypothesis. Ecology, 51(6): 1046- 1051. Hollander, M. and Wolfe, D.A., 1973. Nonparametric Statistical Methods. John Wiley & Sons, New York, USA. Pages 27–75. Otto, C. and Nilsson, L.M., 1981. Why do beech and oak trees retain leaves until spring? Oikos, 387-390. Perea, R., González, R., San Miguel, A. and Gil, L., 2011. Moonlight and shelter cause differential seed selection and removal by rodents. Animal Behaviour, 82(4): 717-723.

R Core Team. 2016. R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. Rebelo, A.G., 1995. Proteas: a field guide to the Proteas of Southern Africa. Fernwood Press, Cape Town, South Africa. Rebelo, A.G. and Breytenbacht, G.J., 1987. Mammal Pollination in the Cape Flora. Pp 109- 125 in Rebelo, A.G., editor. Preliminary synthesis of pollination biology in the Cape flora. National Scientific Programmes Unit: CSIR, Pretoria, South Africa. Reich, P.B., Walters, M.B., and Ellsworth, D.S., 1992. Leaf life‐span in relation to leaf, plant, and stand characteristics among diverse ecosystems. Ecological monographs, 62(3): 365-392. Rourke, J.P., 1980. Proteas of Southern Africa. Purnell, Cape Town, South Africa. Rourke, J. and Wiens, D., 1977. Convergent floral evolution in South African and Australian Proteaceae and its possible bearing on pollination by nonflying mammals. Annals of the Missouri Botanical Garden, 1-17. Schneider, C.A.; Rasband, W.S., and Eliceiri, K.W., 2012. NIH Image to ImageJ: 25 years of image analysis. Nature methods, 9(7): 671-675. Sih, A., 1980. Optimal behaviour: can foragers balance two conflicting demands? Science, 210: 1041-1043. Sivy, K.J., Ostoja, S.M., Schupp, E.W. and Durham, S., 2011. Effects of rodent species, seed species, and predator cues on seed fate. Acta oecologica, 37(4): 321-328. Slingsby, P. and Bond, W.J., 1983. Of ants and Proteas. African Wildlife, 36: 104-107. Smith, A.P., 1979. Function of dead leaves in Espeletia schultzii (Compositae), and Andean caulescent rosette species. Biotropica, 43-47. Smith, D.W., 1970. Concentrations of soil nutrients before and after fire. Canadian. Journal of Soil Science, 50: 17-29. Snyder, J.R., 1984. The role of fire: Mutch ado about nothing? Oikos, 404-405.

Traveset, A., Willson, M.F. and Sabag, C., 1998. Effect of nectar‐robbing birds on fruit set of Fuchsia magellanica in Tierra del Fuego:a disrupted mutualism. Functional Ecology, 12(3): 459-464. Tuner, V., 1982. Marsupials as pollinators in Australia. In Armstrong J. A., Powell J. M., and Richards A. J., editors. Pollination and evolution. Royal Botanical Gardens, Syndey, Austraila. Wiens, D., Rourke, J.P., Casper, B.B., Rickart, E.A., LaPine, T.R., Peterson, C.J., and Channing, A., 1983. Nonflying mammal pollination of southern African Proteas: a non- coevolved system. Annals of the Missouri Botanical Garden, 1-31.