S. Afr. 1. Bot., 1988,54(5): 455-460 455 as 'tumbleseeds': Wind dispersal through the air and over soil

W.J. Bond Saasveld Forest Research Centre, Private Bag X6515, George, 6530 Republic of South Africa Present address: Department of Botany, University of Cape Town, Private Bag, Rondebosch, 7700 Republic of South Africa Accepted 17 May 1988

The fruits of the genus have long stiff trichomes forming a pappus-like structure. In serotinous species the achenes (seeds) are released only after fire when vegetative barriers to dispersal are minimized. The dispersal of seed by free-fall from the cones to the ground (phase I) was compared with subsequent dispersal by rolling over the substrate (phase II). Seed shadows observed in the field as well as seed release under controlled conditions suggest that phase I dispersal is seldom more than 30 m. However phase II dispersal distances measured both from seed shadows and controlled-release experiments were much greater, frequently exceeding 50 m with a maximum over 500 m. The most important biological determinant of both phase I and phase II dispersal was the size of the tuft of hairs and wing loading of the seed. Substrate roughness was a major physical determinant of phase II dispersal distance. In rocky areas, phase II dispersal can be effectively discounted. In smooth areas, pre­ vious estimates of Protea migration rates may be an order of magnitude too low. Other Cape with hairy seeds and serotinous cones occur in Au/ax and a few species of . This convergence suggests that long-distance phase II transport may have adaptive value.

Die sade van die genus Protea is bedek met lang trigome (hare). Die saad van saadhoudende spesies word eers na vuur vrygestel en versprei wanneer die plantaardige versperrings minimaal is. Die verspreiding van die saad direk uit die keels grondwaarts (fase I), is met die daaropvolgende verspreiding, die rolbewegings oor die grond (fase II), vergelyk. Die verspreiding van saad om die moederplant (saadskaduwee) soos in die veld waargeneem en saad­ vrystelling onder beheerde toestande dui daarop dat fase I verspreiding seide meer as 30 m is. Fase II ver­ spreidingsafstande soos gemeet deur beide die saadskaduwee en beheerde saadvrystellingseksperimente was egter baie verder en dikwels meer as 50 m met 'n maksimum van meer as 500 m. Die mees bepalende biologiese faktore in beide fase I en fase II verspreiding was die lengte van die trigome (hare) en die gevolglike vlerklading van die saad. Vir die fase II verspreidingsafstande was die bepalende fisiese faktor die ruheid van die substraat. In rotsagtige gebiede kan fase II verspreiding geignoreer word. Dit blyk dat in areas met gladde oppervlaktes, Protea­ migrasie-afstande vantevore onderskat is. Verdere voordele van spesies met saadhoudende keels in die Kaapse Proteaceae word gevind in die genera Aulax en Leucadendron. Hierdie samelopende ontwikkeling dui daarop dat langafstand-fase II verspreiding aanpasbaarneidswaarde mag he.

Keywords: , Proteaceae, seed dispersal, serotiny

Introduction Previous studies of Protea dispersal have all been indi­ Serotinous proteas in Cape fynbos are remarkably sensi­ rect and none have explicitly considered phase II dispersal tive to fire regime. Populations can be eliminated by short or the dependence of dispersal distance on soil roughness. fire intervals (van Wilgen 1981) or by burning in late win­ Bond (1980) noted that propagules could travel long dis­ ter or spring (Jordaan 1949; Bond et al. 1984; van Wilgen tances by rolling over the ground but that small barriers & Viviers 1985). Even in summer and autumn, the light­ such as roads or streams could prevent dispersal. He cited ning fire seasons, seedling recruitment may fall below very different seedling densities on opposite sides of a replacement levels (Bond et al. 1984). Such extreme den­ road, and a stream bisecting a fire as evidence. Manders sity fluctuations in the dominant woody elements appear (1986) recorded the distribution of 7-year-old in the to be unique among fire-prone shrublands. They imply vicinity of an older, isolated Pro tea laurifolia individual unstable populations with the risk of frequent local extinc­ and found that 95% were within 15 m with a maximum tion. Since these species do not store seed in the soil, their dispersal distance of 30 m. Brits (1987) established isola­ persistence in a landscape must depend either on long­ ted Protea bushes into a barren area and then recorded distance seed dispersal (to recolonize areas where re­ subsequent seedling distributions in relation to these. He generation has failed) or a high reproductive capacity (of also reported short dispersal distances, not exceeding 20 m the few surviving individuals). Some knowledge of seed from point sources. Brits noted that seedling distribution dispersal in Protea is therefore essential for predicting in his study was correlated with the dominant wind direc­ rates of recovery from local population crashes and asses­ tion of the first few days after seed release. He suggested sing the need for active rehabilitation in conservation that the free-flight stage was therefore the most important areas. for seed dispersal. Brits (1982) argued that the limited dis­ Protea seed is dispersed by wind. Dispersal by wind can persal of Protea seeds may have adaptive value in variable be divided into two stages. Firstly propagules fall from the habitats subjected to stress. infructescence to the soil surface (the primary trajectory In this study I attempt to separate the role of first and or phase I dispersal), and secondly, they may be moved second phase dispersal in Protea seed movement. I ask the across the ground surface by physical or biotic agents questions: (phase II dispersal) (Watkinson 1978). Phase II dispersal (a) How far do seeds move in free flight? (phase I dis­ may be particularly important in Pro tea species with ser­ persal). iotinous cones since seeds are released onto a surface (b) How much seed rolls beyond the maximum distances partly cleared by fire and exposed to wind. If phase II attainable by free flight? How does this depend on movement is important then dispersal would depend substrate type? (phase II dispersal). strongly on the nature of the substrate (eg. Mortimer 1974 (c) How important are seed characteristics for dispersal? in Harper 1977; Watkinson 1978). Are site characteristics, especially surface roughness, 456 S.-Afr. Tydskr. Plantk., 1988, 54(5)

more important determinants of dispersal distance? was measured. Seeds were counted and weighed within (d) What realistic rates can be expected for recoloniza­ successive 5 m radii from the release point. tion of an area by seed dispersal? Phase II dispersal Methods To study the further dispersal of seeds after they have I studied dispersal of Protea rep ens achenes in the labora­ fallen to the ground, I observed propagule movement of tory and under natural conditions. has Pro tea repens on a windy day along a wide sandy beach. achenes ('seeds') ca. 10 mm long covered in 10-20 mm The surface of smooth, dry, soft sand approaches that of stiff golden brown trichomes with an 80-mm persistent recently burnt coastal fynbos or stone less mountain fynbos style that often remains on the mature fruit (Figure 1). soils and provides an upper estimate of phase II dispersal The fruits (hereafter referred to as seeds) are borne in ser­ distances. I released batches of 20 seeds on the sand at otinous woody heads and are released within a few days of windspeeds between 6 and 9 m S-1. Immediately after re­ fire. Seeds are blown out of the flower heads by wind lease the seeds were tracked and their resting position (phase I dispersal). On reaching the ground they may be marked. blown over the burnt soil surface (phase II dispersal). The Following the same methods, I assessed the importance long trichomes hold the seeds above the soil so that they of trichomes in phase II dispersal by releasing Protea intercept the wind. aurea seeds simultaneously with all hairs removed, hairs trimmed level with and perpendicular to the base of the Phase I dispersal style and untreated seeds. I also observed Phase II disper­ The free-flight dispersal of a wind-dispersed propagule is a sal in Aulax cancellata (hairy achenes), Leucadendron function of its equilibrium rate of descent (terminal rub rum (parachute device), L. galpinii (hairy achene), velocity) which is primarily determined by seed weight and L. conicum (compressed achene) (nomenclature fol­ and projected surface area (Burrows 1973; Sheldon & lows Bond & Goldblatt 1984). Burrows 1973; Green 1980; Augspurger & Franson 1987). Wing loading (the mass of a seed divided by the surface Field studies area of the pappus) is the most important seed char­ Patterns of dispersal over different substrates were ob­ acteristic affecting dispersal (Sheldon & Burrows 1973). I served near a block of burnt proteoid fynbos in the nor­ measured fall rates over a height of 5 m in a stairwell lined thern slopes of the Groot Swartberg. The proteoid layer with paper to reduce air currents. I assume that the aver­ was dominated by P. repens with rare P. lorifolia, P. exi­ age rate of fall over this distance approximates terminal mia, and P. punctata. The block was surrounded by non­ velocity. I weighed seeds and measured pappus area proteoid fynbos burns in the same fire (March 1987) ex­ (Figure 1) to determine wing loading and its effect on seed cept for the north side which was bounded by a road. The fall. substrate varied from smooth to rocky. An unburnt fire Dispersal distances under windy conditions were break ran parallel to and ca. 50 m from the eastern edge of assessed by releasing seeds from sun-dried P. repens the block (Figure 2). The firebreak had low graminoid flower heads held at 1,75 m above the ground in an open vegetation which readily traps seeds. I could estimate field to simulate natural conditions. Seeds were anchored long-distance dispersal by determining the proportion of where they fell by lawn grass so that only phase I transport all seeds in the transect arriving at this graminoid barrier. I could also compare the effects of different substrate types on dispersal distance by censusing at different points along a. the firebreak. Seeds were counted in 2-m x I-m quadrats along eight tran­ sects laid at right angles to the burnt Protea. In addition

a B persistent style A Ltttt

c ,''',hlt'' t1tD E

pd SMOOTH * ** * * D ROAD * * * '" '" ... UNBURNT ... '" e. '" ... '" Cl\ ... F W F Figure 2 Diagram of the Swartberg site for studying the effect of Figure 1 Proteaceae fruits used in this study. (a) Protea repens substrate on seed dispersal. All the area south of the road was burnt (x 1,5), pd = pappus diameter used in calculating pappus area; (b) except for a firebreak (labelled un burnt) to the east. Burnt proteas Aulax cancellata (x 2) (adaxial side); (c) Leucadendron ga/pinii are indicated by asterisks on the west. Profiles of transects are given (x 2) (abaxial side); (d) Protea aurea (x 2); (e) Leucadendron con­ across rough, stony (A-B), smooth sandy (C-D) and vegetated (E-F) icum (x 2) substrates. South is towards the top of the Figure. S. Afr. J. Bot., 1988, 54(5) 457

four transects were laid north of the block in unburnt that were dispersed further than 20 m were trapped where vegetation in an old firebreak. T counted seedlings on the the substrate changed at the edge of the beach. Dispersal same transects in the burnt area in October to compare distances would have been even greater in a more exten­ the relationship between seed distribution and seedling sive habitat. Surprisingly, the persistent style did not emergence (for example, are seeds dispersed long dis­ appear to influence dispersal distance since seeds with tances viable?). styles were recovered 200 m from the source. The duration of phase II dispersal was short. Typically Results seeds were anchored by sand in the lee of small rills or Phase I dispersal were trapped in debris within minutes of release. Once Phase I dispersal distances of Protea repens are shown in trapped, many seeds were quickly buried by as much as 20 Figure 3. The dispersal curve has a shape typical of wind­ mm of sand and were not re-exposed. dispersed seeds (Harper 1977). The right hand side of the The presence of long trichomes is important for phase II curve approximates a negative exponential of the form dispersal. Both trichome removal and trimming signifi ­ Qd = Ce -kd where Q" is the density of seeds at a given cantly reduced the distances dispersed by Protea aurea distance, C is the initial density of seeds, d is distance from (Table 1). Leucadendron and Aulax seeds with trichomes the source and k is the decay constant. The value k is a or parachutes were also blown significant distances measure of species dispersal capacity and can be estimated whereas Leucadendron seeds without trichomes scarcely from the slope of a log transformed regression (McClan­ moved (Table 2). ahan 1986). In this case k = -2,94 (,2 = 0,83, P ~ 0,01). The maximum distance was 31 m in the prevailing condi­ (a) tions of 6,7 m S-l north winds. 4 3,8 0 As in previous studies (Harper 1977; Green 1980; Aug­ 3,6 spurger & Franson 1987) seed dispersal distances under 3,4 0 0 0 windy conditions were correlated with fall rates in still air 3,2 0 0 (r2 = 0,745, df = 111, P < 0,001). Thus the behaviour of 'I 3 seeds in still air can be used to interpret seed aspects im­ (/) 00 0 0 0 0 .s 2,8 00 0 portant in dispersal in the field. OJ 0 0 1ii 2,6 The relationship between Pro tea rep ens terminal velo­ 2,4 00 aD 0,05). Under windy 1,4 conditions however heavier seeds dispersed significantly 0 200 400 600 shorter distances (r2 = 0,592, df = 99, P < 0,001 seed Pappus area (mm2) weight vs. distance). (b) The long persistent style did not appear to be an ob­ 4 stacIe to wind dispersal. No difference was detected be­ 3,8 tween terminal velocities of seeds with styles present or re­ 3,6 moved (0,398 m S-l style +; 0,40 m S-I styles -). 3,4 3,2 I 0 0 0 Phase II dispersal (/) 3 .s a 0 DO a a Windspeeds during the beach trials were comparable to CD 2,8 0 DO 1 winds prevailing during phase I experiments (6-9 mss ). 1ii 2,6 00

Figure 4 The relationship between (a) pappus area (b) wingload­ 40 ing, and fall rate of Prolea repens seeds in still air.

~ 30 >- u c Table 1 The influence of trichomes on phase II dispersal OJ ::l U of Protea aurea. Values are number of seeds recovered [l! 20 u. from two separate trials

Trichomes (n = 40) 10 Distance (m) Intact Trimmed Removed

0 <1 4 24 2,5 7,5 12,5 17,5 22,5 27,5 32,5 40 <10 7 9 Dispersal distance (m) < 50 4 > 50 11 1 Figure 3 Phase I dispersal of Prolea repens seeds. Windspeed was Not recovered 14 6 o 6,7 m S-I; n = 185 . 458 S.-Afr. Tydskr. Plantk. , 1988, 54(5)

Field trials the rocky substrate reached the firebreak (Table 3) . Re­ The Swartberg field trials allow an assessment of the ef­ gression curves fitted to the pooled data for sandy and fects of surface roughness on dispersal distance. There stony transects conform with the usual negative ex­ were large differences in seed movement over different ponential (McClanahan 1986) for the first 32 m from the substrates which resulted in different spatial patterns of edge of the population (Table 4). Though the exponent is seedling emergence (Figure 5) . Over a quarter of all seeds more negative in stony substrates (Table 3 & 4), the differ­ counted on the sandy substrate reached the unburnt fire­ ence in the regression constants was not significant for the break 54 m from the nearest source. None of the seeds on two substrate types (F = 1,89, df = 1,122, P > 0,05). This suggests equivalent phase I dispersal over this dis­ tance. It is only over the longer distances that dispersal Table 2 Comparison of phase II dispersal in over the two substrates diverges markedly. Reference to Figure 5 indicates the marked effect of a Leucadendron and Au/ax species with different seed barrier on phase II dispersal. The barrier in this case was a types. Seed type t = 'tumbleweed', r = ridged nut, firebreak consisting of low 30 cm tall) unburnt gram­ p = parachute. n = number of seeds released. Note that « some seeds (those dispersing longer distances) were not inoid veld. Seeds penetrated less than 2 m into the un­ recovered burnt veld (Figure 5). In the burnt area, seeds were also generally found trapped by small obstacles, burnt grass Distance moved (m) tussocks in the sand transects or rocks in the stony Seed Not Species type n < I < 10 < 50 > 50 recovered transect.

A ulax cancellala 12 3 6 3 0 Dispersal over un burnt vegetation L. conicum 15 15 0 Seed movement over unburnt vegetation was in marked L. rubrum p 20 0 6 2 10 2 contrast to that observed in burnt areas. The vegetation L. galpinii (+ hairs) t + 15 0 3 1 6 5 formed a moderately dense, low herbland or restios and L. galpinii (- hairs) t - 10 9 1 0 grasses. Phase II transport is extremely unlikely under

T1. Sandy T2. Sandy

100 +++ 100

80 80 • • ••••• • • • • • • • • • • • • • • ••••• • • • • 60 ••••• 60 • • • • • • % • % • • • • 40 40 • +++++++++++++++ • ++++ • +++++++-+++++-++ • -+-++++++-t 20 ++ ++ + 20 • + -+ • + + + O L-----~------~------L-----~------L----~ o + o 10 20 30 40 50 60 o 10 20 30 40 50 60 Distance (m) Distance (m)

T3. Stony T4. Stony

100 +++++++++++!it**·~·· 100 • ••••••• +-+i···•• .. ••• • 80 • 80 • •• -.+ + 60 60 % %

40 40

20 20

OL-----~------L ~ ~ ~ ~ OL-----~------~------~----~------L----~ ______o 10 20 30 40 50 60 o 10 20 30 40 50 60 Distance (m) Distance (m)

Figure 5 The effect of substrate on dispersal distance of Prolea repens in the Swartberg, southern Cape. Points represent the cumulative percentage of seeds (dots) and seedlings (crosses) at increasing distances from a burnt 28-year-old stand of Proteaceae. Arrows indicate the location of an unburnt firebreak acting as a barrier. Tl and T2 are sandy substrates with no stones, T3 and T4 are stony substrates. No. of seeds: Tl = 898, T2 = 614, T3 = 467, T4 = 302. S. Afr. J. Bot., 1988, 54(5) 459

these conditions yet seeds were found up to 85 m from the times occur due to strong updraughts. The dispersal of source (Figure 6). In this case seeds were mostly found be­ seed 85 minto un burnt fynbos (Figure 6) would seem to tween vegetated tussocks suggesting Phase I transport by be an example. There is no information on the frequency an updraft of sufficient strength to exceed the terminal of such events but they are probably rare. velocity of seeds (ca. 4 m S-I or 1,5 km h- I ). This study differs from previous studies (Brits 1987; Manders 1986) in demonstrating the potential for long­ Discussion distance dispersal in hairy-fruited Proteaceae. Both the This study confirms two distinct dispersal phases in Pro­ experiments on sandy beaches and the indirect field obser­ tea, Aulax and Leucadendron species which have seeds vations in the Swart berg indicate that a large fraction of similarly covered with long, dense hairs. Phase I dispersal (fertile) seeds can be transported over 50 m and a few up (free-flight movement) seldom exceeds 30 m at wind­ to 500 m if substrates are favourable. Indirect evidence for speeds characteristic of the Cape mountains. This is com­ the importance of phase II dispersal can often be seen in parable to the distances reported by Brits (1987) and Man­ the dense piles of seeds which accumulate at natural pro­ ders (1986). Dispersal over longer distances may some- jections and the verges of mountain tracks which act as barriers to seed movement after fire. The latter form the familiar Proteaceous 'avenues' bordering many mountain Table 3 Protea repens dispersal over different roads. substrates, Swartberg, southern Cape. k and Care The distribution of seeds in the Swartberg site suggests regression estimates of seed density: Qd = Ce -k~ where that the seed shadow is strongly substrate dependent. In d is distance from modal densities rocky sites or rough vegetated areas, dispersal patterns are Sandy Stony Un burnt primarily created by free-flight. Subsequent redistribution is unimportant for dispersal distance. Phase II dispersal k 1,60; 1,29 1,71; 1,44 1,19; 1,27 will only be significant in smooth cleanly burnt areas. 1,03; 1,02 1,50; 1,21 C 6,43; 6,07 5,83; 4,88 4,29; 4,47 influences on dispersal 5,0 ; 4,76 5,31; 4,24 R2 0,78; 0,84 0,83 ; 0,88 0,71 ; 0,78 Dispersal distances in Protea seeds are a function of wing 0,64; 0,70 0,84; 0,64 loading - the ratio of the projected surface area of the Distance including 90% of seeds (m) propagule and its weight (Figure 4). Surface area is a func­ > 54 > 54 12; 14 24; 54 tion of the length of trichomes, their density ('porosity') > 54 > 54 14; 10 and the angle at which they are held. The height at which Distance including 95% of seeds (m) seeds are released is also an important biological determi­ > 54 > 54 42 ; 36 36; 65 > 54 > 54 18; 18 nant of dispersal distance (Burrows 1973; Levin & Kerster 1974). The length, density and angle of the trichomes vary in Pro tea seeds (Rourke 1980; Vogts 1982) and, together Table 4 Regression statistics for seed density (no. of with seed weight, may affect inter- and intra-specific dif­ seeds 2 m - 2) as a function of distance (m) in stony vs. ferences in dispersal. sandy substrates. The regressions are calculated for the 10rdaan (1972) and Brits (1982) have suggested that the pooled values or replicated transects up to 32 m from the trichomes of Protea seeds limit dispersal by trapping sand grains and binding seeds together in clumps. This might source population. n in each case = 64 appear to be the case in the beach sand experiments since Sand phase II dispersal was of very brief duration and trichomes log seed number = 6,156 - 1,488 (log distance), r 2 = 0,723 did trap sand grains once the seed had settled. However Stone the results of this study show that trichomes enhance dis­ log seed number = 5,885 - 1,702 (log distance), r2 = 0,821 persal. Long dense trichomes are important both for phase I dispersal where they are the main determinant of the rate of descent in P. repens (Figure 4), and phase II dispersal, since trimming their length curtails dispersal dis­ tance (Table 1). T5. Unburnt Are Protea seeds adapted for short-distance dispersal? 100 • ••••••• • Brits (1982) has suggested that Pro tea dispersal is topo­ ••••••••• chorous ie. seeds are adapted for short-distance dispersal. •••••• 80 •••• He argues that topochory is well suited to small variable • habitats characteristic of fynbos since the seed is not lost to • • unfavourable sites. Although mechanisms for long-dis­ 60 • tance dispersal are generally rare in fynbos (Siegfried % • 1983; Slingsby & Bond 1985), this study does not support 40 Brits' contention that Protea seeds are only topochorous. On the contrary, serotinous Protea seeds are also capable of dispersing long distances (a) through being released 20 after fire when vegetative barriers to wind dispersal are minimal and (b) through their ability to roll over the soil 0 surface, also cleared by fire. Substrate roughness will 0 20 40 60 80 100 strongly limit phase II dispersal regardless of seed Distance (m) morphology but long, dense trichomes will always favour Figure 6 Dispersal of Protea rep ens into un burnt fynbos in the longer dispersal through their effect on wing loading. Swart berg. Points represent cumulative frequency of seeds from a Appropriate evidence for topochory is thus the absence of road bordering a burnt Protea stand into short graminoid fynbos. trichomes on some taxa rather than their presence. Protea n = 218. magnifica, for example, has a very large seed with a very 460 S.-Afr. Tydskr. Plantk. , 1988, 54(5)

sparse ring of trichomes in contrast to the well-developed Acknowledgements pappus of P. repens (Vogts 1982, pers. obs.) which has the I am grateful to H. Homan for pointing out the Swartberg widest distribution in Protea (Rourke 1980). Leaucaden­ study site, and for his critical interest which helped initiate dron species in the sub-section Villosa have hairy seeds in this study. I thank W.M. Bond, 1. Midgley and R. coastal fynbos but glabrous seeds in mountain areas America for assistance in the field and 1. Vlok, 1. Mid­ (Williams 1972). Interestingly, L. brunioides, which has gley, 1. Breytenbach, D. Richardson and P. Manders for the widest habitat range in the sub-section Villosa, has helpful discussions and comments on the manuscripts. E . both hairy- and glabrous-seeded populations (Williams Breytenbach drew Figure 1 and, together with M. Viviers, 1972). translated the abstract. This study was supported by the conservation research programme of the Directorate of Recolonization rates Forestry, Department of Environment Affairs. Manders (1986) suggested a potential migration rate on the order of 30 m per generation for serotinous Prot­ References eaceae. On this basis a population would advance only 1 AUGSPURGER, e.K. & FRANSON, S.E. 1987. Wind dispersal km in 500 years at a mean fire frequency of 15 years. of artificial fruits varying in mass, area and morphology. Ecology These values are appropriate for seeds dispersed only by 68: 27--42. phase I dispersal and excluding occasional long-distance BOND, W.J. 1980. Fire and senescent fynbos in the Swartberg, dispersal by turbulent updrafts (Figure 5; Brits 1987) . southern Cape. S. Afr. For. 1. 114: 68-71. However where substrate allows effective phase II dis­ BOND, P. & GOLDBLATT, P. 1984. Plants of the Cape flora. A descriptive catalogue. Jl S. Afr. Bot. Supp\. 13. persal, migration rates may be an order of magnitude BOND, W.J. , VLOK, J. & VIVIERS, M. 1984. Variation in seed­ greater. For example, it would take one or two fire cycles ling recruitment of Cape Proteaceae after fire. i . Ecol. 72: and thus 30 to 60 years rather than 500 to advance 1 km 209-221. over smooth substrates in Mander's (1986) example. BRITS, G.J. 1982. Some adaptations in seed regeneration of fynbos Rapid migration by phase II dispersal can be affected by Proteaceae. Paper read at the 8th annual S.A.A.B. congress. the construction of roads (Figure 5; Bond 1980) and by the BRITS, G.J . 1987. Short-distance seed dispersal in three serotinous Cape Proteaceae. Paper presented at the 9th Annual Research size of fires. A mosaic of small patch burns would be the Meeting of the Fynbos Biome Project, 1987. least effective method for promoting recolonization of loc­ BURROWS, F.M. 1973. Calculations of the primary trajectories of ally extinct areas. Dispersal between burnt patches and ac­ plumed seeds in steady winds with variable convection. New ross intervening un burnt barriers is very unlikely. Fires Phytol. 72: 647---{j64. should therefore be large enough to include surviving GREEN, D.S. 1980. The terminal velocity and dispersal of spinning stands of mature plants to act as a seed source. samaras. Am. 1. Bot. 67: 1218-1224. HARPER, J.L. 1977. Population biology of plants. Academic Press, London. Conclusions JORDAAN, P.G. 1949. Aantekeninge oor die voortplanting en Seeds dispersed by rolling over the ground ('balloon' brandperiodes van Protea mellifera Thunb. Jl S. Afr. Bot. 15: seeds, tumbleweeds or 'chamaechores') are characteristic 121-125. of open environments such as sparsely vegetated shrub JORDAAN, P.G. 1972. Die generatiewe voortplanting van die Pro­ teaceae. Bull. Bot. Soc. Sth. Afr. 58: 48-56. steppes and arid lands (van der Pijl 1982). Their occur­ LEVIN, D.A. & KERSTER, H .W. 1974. Gene flow in seed plants. rence in dense fynbos shrub lands is linked to serotinous Evol. BioI. 7: 139-220. species which likewise release seeds en masse into an open MANDERS, P.T. 1986. Seed dispersal and seedling recruitment in post-fire environment. Depending on substrate, such . S. Afr. 1. Bot. 52: 421--424. seeds can potentially travel far greater distances than McCLANAHAN, T.R. 1986. Seed dispersal from vegetation myrmecochores and other seed types released between islands. Ecol. Modelling 32: 301-309. MORTIMER, A.M. 1974. Studies of germination and establish­ fires. They can apparently also disperse as far or further ment of selected species with special reference to the fates of than winged or compressed seeds of fynbos Proteaceae. seeds. Ph.D. thesis, Univ. of Wales. The presence of both short-distance (myrmecochory) ROURKE, J.P. 1980. The proteas of southern Africa. Purnell, and long-distance (wind-dispersed) propagules in Cape Cape Town. Proteaceae suggests interesting differences in the role of SHELDON, J.e. & BURROWS, F.M. 1973. The dispersal ef­ immigration for preventing local population extinction. In fectiveness of achene-pappus units of selected Compositae in myrmecochores, populations decimated by poor recruit­ steady winds with convection. New Phytol. 72: 665-75 SIEGFRIED, W.R. 1983. Trophic structure of some communities ment after 'bad' fires would be unlikely to recover by im­ of fynbos birds. il S. Afr. Bot. 49: 1--43. migration. To persist, they should be well buffered against SLINGSBY, P. & BOND, W.J. 1985. The influence of ants on the recruitment failures, possibly by producing seeds that can dispersal distance and seedling recruitment of lie dormant in the soil for many years. In Protea , depend­ conocarpodendron (L.) Buek (Proteaceae). S. Afr. 1. Bot. 51: ing on substrate roughness, populations could survive 30-34. large fluctuations in recruitment through long-distance VAN DER PIJL, L. 1982. Principles of dispersal in higher plants. 3rd edn, Spring-Verlag, Berlin. dispersal. However the geographic area in which such VAN WILGEN, B.W. 1981. Some effects of fire frequency on long-distance dispersal is important may be quite restric­ fynbos plant community composition and structure at Jonkers­ ted since many Protea species occur in rocky mountain hoek, Stellenbosch. S. Afr. For. 1. 118: 42-55. habitats which would not favour phase II dispersal. VAN WILGEN, B.W. & VIVIERS, M. 1985 . The effects of season The present study illustrates merely the potential pat­ of fire on serotinous Proteaceae in the western Cape and the im­ terns of phase I and II dispersal. Actual dispersal patterns plications for catchment management. S. Afr. For. 1. 133: 49-53. VOGTS, M. 1982. South Africa's Proteaceae. Struik, Cape Town. of Pro tea propagules under the influence of variable wind WATKINSON , A.R. 1978. The demography of a sand dune annual: speed, wing loads and especially substrate roughness and Vulpia fasciculata. II I. The dispersal of seeds. i. Ecol. 66: 483--498. their relevance for population biology and biogeography WILLIAMS, I.J.M. 1972. A revision of the genus Leucadendron remain open questions. (Proteaceae), Contr. Bal. Herb . 3: 1--425.