Effects of simulated fire and browsing on the resprouting of subtropical dune thicket shrubs in the southeastern Cape Floristic Region

The expansion of subtropical thicket vegetation at the expense of more species-rich, fire-prone fynbos, potentially due to lower frequency and severity of fire and browsing, is a concern in many coastal dune landscapes of the Cape Floristic Region (CFR) where these two vegetation types co-occur. We were interested in the effects of severe fire or browsing treatments (causing complete loss of aboveground biomass) on the vigour of post-fire resprouts of dune thicket shrubs. Cape St Francis, CFR, South Africa. We used an in situ experimental approach to compare the effects of simulated fire and browsing by herbivores on mortality, resprouting vigour and resprouting rate of 10 canopy-forming dune thicket shrub species from different architectural guilds, 5 years after a previous severe wildfire. Survival was significantly lower after fire (85%) than after browsing (95%), and was significantly positively related to pre-treatment shrub size. All measures of resprouting vigour were significantly lower after fire than after browsing, and were significantly positively related to pre-treatment size. Resprouting rate was significantly lower after fire than after browsing, and was significantly positively affected by pre-treatment size. Survival and measures of resprouting vigour were generally decoupled from architectural guild and species identity. Dune thicket shrubs showed high survival after both fire and browsing treatments, suggesting that these species are resilient to frequent complete loss of above-ground biomass. Our results suggest that short interval fires (5–15 years) of high severity will merely maintain the co-occurrence of dune fynbos and thicket vegetation by setting back thicket growth rather than causing large-scale mortality of thicket shrubs.


Introduction
Fire and browsing are evolutionary pressures for driving trait acquisition in many vegetation types, such as plant architecture features and resprouting vigour (Bellingham and Sparrow 2000;Bond and Midgley 2001;Clarke et al. 2013;Lamont et al. 2017). Starting in the early Miocene (ca. 20 Myr), closed-canopy formations like subtropical thicket and dry forest became the dominant vegetation over large parts of southern Africa (Cowling et al. 2005;Neumann and Bamford 2015) and formed habitats suitable for mammalian megaherbivores (Janis 1993;Pennington et al. 2010). Top-down browsing by megaherbivores (e.g. Sivatheres and Gomphotheres) in these ecosystems likely drove trait acquisition such as resprouting, prolific branching and spinescence (Hendey 1982;Janis et al. 2000;Bond and Silander 2007;Onstein et al. 2022). Consequently, the ancestral selective force for resprouting in these plant lineages was likely defoliation by browsing. Since the late Miocene (10 Myr), cooling and drying (Janis 1993;Westerhold et al. 2005) produced fire-supporting climates, which promoted the growth and expansion of open grass-and shrub-dominated fire-prone ecosystems in the subtropics (Janis 1993;Jacob et al. 1999;Keeley and Communicated by Kyle Palmquist. 1 3 Rundel 2005;Edwards et al. 2010;Rundel et al. 2018). Subsequently, fire likely emerged as a significant evolutionary factor in filtering traits (e.g. geoxylic life form, thick bark) of thicket plants (i.e. exaptation) from thicket margins or thicket-fynbos ecotones.
Coastal dune landscapes of South Africa's Cape Floristic Region (CFR) predominantly support a vegetation mosaic of two biomes, namely Fynbos (dominated by low, fineleaved shrubs) and Subtropical Thicket (dominated by tall, broad-leaved shrubs), termed 'dune fynbos-thicket' (Cowling 1984;Tinley 1985). Here, the expansion of closed-canopy subtropical thicket into more open fynbos shrublands, which are comparatively richer in species and endemics, has raised concern among plant ecologists (Cowling et al. 1997;Cowling and Hoffman 2021). The probable cause for this expansion of thicket is a change in natural disturbance regimes, specifically a lack of historical megaherbivory and a decrease in the frequency of wildfires, the latter a consequence of urbanization leading to active fire suppression (Cowling et al. 1997;Western Cape Government 2022). The impacts of these disturbance regimes on subtropical thicket vegetation are not well understood. Other than Sharam et al. (2006), who studied broad-leaved thickets in the Serengeti of Tanzania, we know of no studies that have investigated the differential effects of fire and browsing on the survival and resprouting vigour of African thicket species. In coastal dune thicket of the CFR, the effect of fire severity on mortality and resprouting vigour of canopy-forming shrubs has been assessed by Strydom et al. (2020), who showed that smaller shrubs had higher mortality and weaker resprouting vigour than larger shrubs. While elephant impacts have been assessed in inland thicket of South Africa (Cowling and Kerley 2002;Kerley and Landman 2006), little is known about the effects of megaherbivore browsing on subtropical dune thicket species.
Here we report on the effects of experimental fire and browsing disturbance treatments on post-fire resprouting of subtropical dune thicket (hereafter 'dune thicket') shrub species in the eastern CFR. Specifically, we investigate their resilience, in terms of survival and resprouting vigour, to discrete events of complete aboveground biomass loss. While dune thicket is not a fire-dependent ecosystem, it is exposed to fire where it co-occurs with fire-dependent and highly flammable dune fynbos (Calitz et al. 2015;Msweli et al. 2020), which can burn every 10-20 years (Cowling et al. 1997;van Wilgen et al. 2010;Kraaij et al. 2013a). The position of thicket in the landscape determines the levels of fire exposure, influencing thicket species composition and structure (Strydom et al. 2022). Dune thicket has relatively low flammability (Clarke et al. 2013;Msweli et al. 2020) and is seldom exposed to fire, owing to fire suppression to protect coastal developments and alteration of fire spread patterns. Recent wildfires in long-unburnt dune thicket near Cape St Francis (2016) and Knysna (2017) showed dominant canopy species (e.g. Sideroxylon inerme Forssk., Pterocelastrus tricuspidatus Walp., Mystroxylon aethiopicum (Thunb.) Loes., Euclea racemosa L. and Searsia glauca (Thunb.) Moffett.) to have high survival rates (75-90%) and strong resprouting abilities from belowground bud banks after severe burning (Strydom et al. 2020).
The architectural composition of dune thicket is largely determined by the level of landscape-scale fire exposure (Strydom et al. 2022). Thicket that has not burnt in over ca. 100 years has an abundance of vertical growers, which tend to show weak resprouting after fire (Strydom et al. 2021). With moderate fire exposure (50-100 year fire intervals), lateral spreaders, which are strong post-fire resprouters, dominate the thicket canopies. With high fire exposure (10-50 year fire intervals), hedge formers are most abundant. Species with this architecture (e g. Euclea racemosa, Olea exasperata Jacq., Rapanea gilliana Mez and Searsia laevigata (Moffet) Moffett) are strong post-fire resprouters and include many geoxylic shrubs with enlarged woody stems belowground (White 1977;Cowling 1983;Maurin et al. 2014;Grobler and Cowling 2021). It seems likely that geoxyles evolved because of the emergence since the late Neogene of climates that could support recurrent fire, as has been demonstrated for African savanna geoxyles (Hoetzel et al. 2013;Maurin et al. 2014;Lamont et al. 2017).
Browsing regimes have influenced survival and resprouting vigour of many woody plants across multiple ecosystems, such as thicket, savanna and forest (Cowling and Kerley 2002;MacGregor and O'Connor 2004;Fornara and Du Toit 2008;Lagendijk et al. 2011). Browsing by African elephant (Loxodonta africana) on trees can be destructive (Sharam et al. 2006;Edkins et al. 2008;Morrison et al. 2016); these megaherbivores are the cause of highest levels of savanna tree mortality when compared to other browsers (MacGregor and O'Connor 2004;Sharam et al. 2006;Morrison et al. 2016). Furthermore, by reducing the stature of trees, they expose tree shoots to browsing by mesoherbivores such as kudu, Tragelaphus strepsiceros, and impala, Aepyceros melampus (Makhabu et al. 2006). In the unbroken forms of succulent thicket of the interior CFR that are not exposed to fire, elephants at high density reduce substantively the cover and stature of the vegetation when compared to elephant-free thicket and induce the proliferation of branching via epicormic resprouting in many shrub species (Cowling and Kerley 2002;Kerley and Landman 2006).
The research objectives of this study were to: (1) Understand the resilience of recently burnt, resprouting thicket shrubs to subsequent fire and browsing; we used an experimental approach to compare the effects of simulated severe fire and simulated severe browsing by herbivores on shrub survival and resprouting vigour of 10 dune thicket shrub species from different architectural guilds (sensu Strydom et al. 2021) 5 years after a previous severe wildfire. (2) Test whether resilience to fire and browsing disturbance regimes differed among architectural guilds in relation to their differential abundance in sites with different fire exposure (Strydom et al. 2022); we expect vertical growers and lateral spreaders, which grow in sites subject only to occasional fires, to have higher survival and resprouting vigour after severe browsing than after severe fire, and the hedge formers, which are most abundant in frequently burnt sites, to have a more vigorous response to fire than browsing. (3) Interpret the study findings to inform the conservation and management of dune thicket in the event of future environmental change. In particular, we aimed to: (3.1) determine whether short return-interval fire and severe fires, as was simulated in our study, could potentially limit the expansion of thicket at the expense of species-and endemic-rich dune fynbos (Hoffmann et al. 2020); and (3.2) establish the resilience of thicket shrub species to severe browsing as the increase of nature reserves, game farms and eco-estates in the coastal landscapes of the CFR, comes with the stocking of large herbivores, some extralimital to thicket (Cornelius 2011), and will likely impact thicket shrub structure and composition.

Study area
The study area is at Cape St Francis (34° 10ʹ 28ʹʹ S, 22° 82ʹ 11ʹʹ E) within the southeastern CFR of South Africa, located on Pleistocene calcarenites (Nahoon Formation) overlaid with Holocene aeolinates (Schelmhoek Formation) (Roberts et al. 2006;Cowling et al. 2019) (Fig. 1). The landscape consists of vegetated parabolic dunes and mobile sand forming a bypass dune system (Tinley 1985;Illenberger and Burkinshaw 2008). The climate of the area is warmtemperate, has year-round rainfall with a mean annual precipitation of 700 mm but with relatively dry mid-summer months (December-February) (Cowling 1984). The mean annual temperature is 17 °C with a minimum of 4 °C and maximum of 32 °C. Prevailing winds are from the westsouthwest in winter and from the east-southeast in summer. The area is subjected to gale force winds, mainly in the spring and summer months.
The vegetation of the stable dunes is predominately a mosaic of fynbos (fine-leaved shrubs that are fire prone) and thicket (broad-leaved shrubs that are less fire prone) with pockets of forest-thicket (Cowling 1984;Cowling et al. 2019;Strydom et al. 2022) (Fig. 1). Thicket shrub species are associated with three floristically and structurally distinct communities-fynbos-thicket, thicket and forest-thicket-whose distribution in the dune landscape is largely dictated by fire exposure (Strydom et al. 2022). Dune-endemic hedge-forming shrubs (e.g. Olea exasperata, Searsia crenata (Thunb.) Moffett, Euclea racemosa, Robsonodendron maritimum (Bolus) R.H.Archer) are most common in fynbos-thicket with high fire exposure (10-50 year fire intervals). Laterally spreading shrubs (e.g. Sideroxylon inerme, Mystroxylon aethiopicum, Pterocelastrus tricuspidatus) dominate thicket with moderate fire exposure (50-100 year fire intervals), while species commonly found Fig. 1 Fynbos-thicket vegetation at Cape St Francis along the southeast coast of the Cape Floristic Region (CFR) of South Africa, which experiences high (10-50 year intervals) to moderate (50-100 year intervals) levels of fire exposure. The insert shows South Africa, the CFR delineated with a green line and study location indicated with a white dot. Imagery source: QGIS.org (2022) 1 3 in forest (e.g. Dovyalis rhamnoides Engl., Psydrax obovata Bridson, Chionanthus foveolatus E.Mey. Stearn, Clausena anisata Hook.f., de Wild. & Staner) occur in forest-thicket with low fire exposure (100 + year fire intervals; Strydom et al. 2022). The study area experienced a severe wildfire in January 2016, which burnt large areas of fynbos-thicket and thicket but did not burn forest-thicket (Strydom et al. 2020).
Pre-colonial era (prior to the late seventeenth century) dune landscapes likely supported transient populations of African elephant and black rhinoceros (Diceros bicornis) (Boshoff and Kerley 2001;Radloff 2008), which may have had an impact on thicket structure. The contemporary mammalian browsing regime across the study area is dominated by bushbuck Tragelaphus scriptus with Cape grysbok Raphicerus melanotis playing a lesser role. Population densities are relatively low and unlike the herbivore-exposed thickets of the interior (Kerley and Landman 2006;Mills et al. 2014), browsing impacts are barely discernible.

Experimental design and data collection
Field experiments were undertaken during summer in January 2021, 5 years after a severe wildfire in January 2016, in fynbos-thicket vegetation which had a pre-fire age of 50 + years before the 2016 fire. The treatments applied were simulations of a high severity, short return interval fire (hereafter 'fire') and once-off severe browsing (causing near-complete loss of aboveground biomass such as induced by megaherbivores; hereafter 'browsing'), both applied 5 years post-fire ( Fig. 2). Fire was simulated through a combination of mechanical excision (using a lopper shear) of all living stems at 2 cm height from the base at ground level, and subsequent blowtorching of the remaining stumps. Entire plants were not burnt as this would increase the risk of starting wildfires. However, the removal of almost all above-ground biomass simulated the effect of severe fire, leaving only the short, burnt stumps. Removal of aboveground biomass would not preclude resprouting, as Strydom et al. (2020) showed that basal sprouting was the principal means of post-fire regeneration in these shrubs. To consistently apply high fire severity when using the blowtorch, the cut stumps were treated until the bark was consumed and the wood severely damaged or partly charred (Fig. 2b), in accordance with what was deemed to be high fire severity resulting from the 2016 fire (Strydom et al. 2020). Browsing by herbivores was simulated by means of mechanical excision (using a lopper shear) (Bergström et al. 2000) at 2 cm height from the base at ground level of all living stems. We applied these respective treatments-fire or browsing-to 10 individuals each of 10 dune thicket shrub species, of which five species were vertical growers, three lateral spreaders and two hedge formers. We selected resprouting shrubs that had a stem or base (in the case of multi-stemmed individuals) diameter at ground level of 1-10 cm. Each treated shrub was marked with a metal rod and a numbered tag to enable subsequent monitoring. Unequal availability of species representing each architectural guild did not allow for a more balanced experimental design.
We assessed the treated dune thicket species' survival (i.e. whether the plants were resprouting or not) after 12 months. To assess resprouting vigour, we measured resprouting shoot count, shoot height and canopy short and wide diameter at 12 months after treatment application with termination of Fig. 2 a Experimental fire treatment, using a gas blowtorch to burn the cut stumps, b cut stumps burnt to high severity (severe damage to the wood, which is partially consumed) (Strydom et al. 2020) and c browsing treatment, using a lopper shear to cut the stems the experiment. The resprouting rate (~ speed) was determined from % survival observed at one, three, six, nine and 12 months after treatment.

Data analysis
Resprouting vigour was expressed in terms of three measures; (i) shoot count, (ii) canopy height (cm), (iii) canopy area, taken as that of an oval (cm 2 ) (a/2 × b/2 × π), where: a = canopy diameter wide, b = canopy diameter short, and π. Pre-treatment size was calculated as the sum of the basal area (cm 2 ) of all stems of an individual, with the basal area of a stem taken as that of a circle and calculated from the measured stem diameter. We used this measure, rather than a measure of canopy dimension, to represent pre-treatment size, as Strydom et al. (2020) showed shrub basal dimension to be a significant predictor of post-fire survival and resprouting vigour in thicket shrubs.
We assessed survival and the various measures of resprouting vigour at 12 months post-treatment in relation to the predictor variables (i) treatment (categorical fixed-effect with two levels; fire or browsing), (ii) architectural guild (categorical fixed-effect with three levels; vertical grower, lateral spreader, or hedge former), (iii) pre-treatment size (continuous fixed-effect), (iv) species (categorical randomeffect with 10 species as levels) and (v) the interaction between treatment and architectural guild, using generalized linear mixed-effects models (GLMM) (Bates 2010;O'Hara 2009). There was no significant interaction between treatment and architectural guild for survival and all measures of resprouting vigour except resprouting shoot count (Online resource 1). Because of the inconsistency of a detectable interactive effect, the interaction term was excluded from final models.
These models were implemented using the lme4 package (Bates 2010) in the open-source R software version 4.1.3 (R Development Core team 2022). We assessed survival using a GLMM with binomial family and logit link function. The number of resprouting shoots (an integer-valued count variable) was assessed using a GLMM with Poisson distribution and logit link function. Resprouting shoot height (rightskewed distribution) were assessed using GLMMs with gamma family and logit link function. Resprouting canopy cover (data square-root transformed to correct left-skewed distribution) were assessed using GLMMs with gamma family and logit link function. To determine the significance of fixed-effects in the specific models, the Type II Wald chisquare test was used (Hastie and Pregibon 1992).
Resprouting rate (~ speed) was determined from the presence or absence of live resprouts on individual shrubs one month, three months, six months, nine months and 12 months after the treatment. The effect of treatment (categorical fixed-effect with two levels; fire or browsing), architectural guild (categorical fixed-effect with three levels; vertical grower, lateral spreader or hedge former) and pre-treatment size (continuous fixed-effect) on the likelihood of resprouting over time (interpreted as resprouting rate) was analysed using a Cox proportional hazard regression model (Allison 1995(Allison , 2010. The model was fitted as a factorial, testing for the effects of the factors on resprouting rate using the survival package in R (Thereau and Grambsch, 2000). The coefficient outputs of the model were analysed by ANOVA test.

Survival
Overall, the disturbance treatment type and pre-treatment shrub size significantly influenced survival, which was lower after fire (85% ± 12.8%, mean ± SE) than after browsing (95% ± 6.4%), and positively related to pre-treatment size, whereas architectural guild had no significant effect (Table 1; Fig. 3; Online Resource 2). The total proportion of variance explained by the model for survival, including both fixed-effects and the random-effect factor, was 15% (conditional R 2 values, Table 1). The fixed-effects combined explained more variance (11%, marginal R 2 values) than species as a random-effect by itself (4%). Cassine peragua L., Clausena anisata, Sideroxylon inerme and Euclea racemosa had the highest survival after both treatments, whereas Scolopia zeyheri Warb. and Dovyalis rotundifolia had the lowest survival (Online Resource 3).

Resprouting vigour
All measures of resprouting vigour (shoot count, shoot height and canopy cover) varied significantly with disturbance treatment and pre-treatment size (Table 1, Figs. 3, 4, Online Resource 4). The three resprouting measures were significantly lower after fire than after browsing, while being significantly positively related to pre-treatment size. Architectural guild had no significant effect on any measure of resprouting vigour. However, the initial models which included the interaction between treatment and architectural guild showed that this interaction was significant for resprouting shoot count (Online resource 1), with hedge formers, but not lateral spreaders nor vertical growers, being favoured by fire compared to browsing (Fig. 3f). The total proportion of variance explained by the (final) models, including both fixed-effects and the random-effect, for measures of resprouting shoot count, height and canopy cover was 57, 27 and 39%, respectively (conditional R 2 values). The fixed-effects combined explained more variation for measures of resprouting shoot count (43%), height (17%) and canopy cover (33%) (marginal R 2 values) than species as a random factor, which explained 14, 10 and 6%, of variation in these measures of resprouting vigour, respectively (Table 1). Shoot count increased two-fold after fire and browsing treatments compared to the pre-treatment state (Online Resource 4). Compared to pre-treatment states and across all species, there was a 51 and 54% recovery in height of resprouting stems and 28% and 43% recovery in plant canopy cover 1 year after fire and browsing, respectively. The resprouting vigour of Searsia glauca was notably greater after fire than browsing, and that of Cassine peragua, Clausena anisata and Mystroxylon aethiopicum were notably weaker after fire than browsing (Fig. 5); however, resprouting vigour was largely driven by pre-treatment shrub size (Table 1; Online Resource 4, 5).

Resprouting response to fire and browsing
Resprouting of dune thicket species was highly resilient to both simulated severe fire and browsing, showing high post-disturbance survival (85 and 95%, respectively). Compared to browsing, however, survival, resprouting vigour and resprouting rate were significantly lower after the fire treatment. In seven of the 10 species we investigated, resprouting after fire was delayed for up to three months compared to after browsing. However, in terms of resprouting shoot count, treatment interacted with architectural guild, showing that hedge formers, but not lateral spreaders nor vertical growers, were favoured by fire compared to browsing. Generally, resprouting after fire occurred from belowground bud banks, hosted primarily on basal burls or woody rhizomes ; in contrast, after browsing resprouts emerged epicormically from aboveground buds sensu (Pausas and Keeley 2017) (Online Resource 7). These findings suggest that the severe fire treatment destroyed the aboveground dormant buds of the thicket shrubs (Clarke et al. 2013) resulting in resprouting from the belowground buds (in woody roots or rhizomes (Nzunda et al. 2008), a phenomenon which was rarely observed after simulated browsing. Thus, the thicket shrubs can cope with both forms of disturbance, but the severe fire treatment (which essentially was equivalent to severe browsing plus severe fire) more drastically scarred plant tissue and depleted resources, whereas the browsing treatment likely resulted in a more rapid mobilization of aboveground buds and carbohydrate resources (Gurvich et al. 2005;Nzunda et al. 2008). Furthermore, if browsing, and not fire, was the main selective pressure, it may be surmised that the number of dormant buds Table 1 Output of generalized linear mixed-effects models that assessed the effects of fire and browsing treatment, pre-treatment size and species on the survival and resprouting shoot count, resprouting shoot height and resprouting cover of 10 dune thicket shrubs Fixed factors included in the generalized linear mixed-effects models were treatment (fire and browsing), pre-treatment size and architectural guilds (hedge former, lateral spreader and vertical grower), while species was included as a random factor a Chisq statistics and significance levels were obtained from deviance tables (Type II Wald chi-square tests) b Conditional R 2 indicates the proportion of variance explained by fixed and random factors combined, marginal R 2 indicates the proportion of variance explained by fixed factors alone, and R 2 (1|Species) indicates variance explained by the random factor alone  . 3 Predicted effects of a fire or browsing treatment, b pre-treatment size (see "Methods" for details), and c architectural guild (H, hedge former; L, lateral spreader; V, vertical grower) on the survival (panels a-c), resprouting shoot count (d-f), resprouting shoot height (g-i), and resprouting canopy cover (j-l) of 10 dune thicket shrub species 12 months after treatment application. The output of the Generalized Linear Mixed Effects Model (GLMM) that was used to calculate the effects is given in Table 1. Error bars and shaded bands indicate 95% confidence intervals would be fewer in the roots than in the shoots, which may partly explain reduced resprouting vigour after fire compared to browsing. However, this did not apply to hedge formers (which were favoured by fire compared to browsing), that are capable of vegetative spread from strong, belowground buds (suckers), and whose evolution likely is more closely linked to fire (particularly geoxyles, e.g. Lamont et al. 2017). Even though fire had a greater negative effect on resprouting vigour of most thicket shrubs than browsing, the impacts of continuous browsing or sustained browsing of young post-fire resprouts may have a greater negative impact on resprouting vigour than just fire, whereby stored carbohydrate reserves become depleted (Schutz et al. 2011;Ngugi et al. 2022), but this is yet to be investigated for thicket shrubs. Pre-treatment shrub size had a strong positive influence on survival and resprouting vigour after both disturbance treatments. This is unsurprising since larger plants can mobilize more carbohydrates post-disturbance (Vesk 2006;Lawes et al. 2011;Marais et al. 2014;Strydom et al. 2020;Staver et al. 2009). The smallest individuals (with the smallest basal area) were particularly sensitive to disturbance. This is consistent with a broader survey undertaken after the 2016 fire by Strydom et al. (2020), who found that smaller dune thicket shrubs had higher mortality after fire than larger shrubs, as has been observed for forest and savanna trees (Matula et al. 2019;Giddey et al. 2022). Surprisingly, architectural guild did not significantly influence survival, resprouting vigour or resprouting rate. We expected that hedge formers would be superior in terms of resprouting ability, and vertical growers inferior, corresponding to a decreasing gradient of fire exposure in the sites where these guilds dominate (Strydom et al. 2022). Although not statistically significant, some weak trends were evident; only some of the vertical growers (assumed to be weaker resprouters; Bond and Midgley 2001) exhibited strong resprouting, whereas all hedge formers were strong resprouters, and lateral spreaders were moderate-strong resprouters. Overall, species as a random factor in the models (except in the model for shoot count), explained relatively little variance in the resprouting response. Other factors that may have influenced survival and resprouting vigour include tolerance to environmental conditions (e.g. light availability or drought), soil resource availability and competition from other plants that did not experience the treatment (Cruz et al. 2002;Chew and Bonser 2009;Clarke and Knox 2009;Clarke et al. 2013;Twidwell et al. 2016;Laurent et al. 2017).

Management implications
A problem facing many fynbos-thicket dune landscapes throughout the CFR is the expansion of thicket at the expense of more species-and endemic-rich fynbos (Cowling et al. 1997). This has been hypothesized as arising from the absence of megaherbivory and a decline in fire frequency, itself a consequence of increased urbanization leading to active fire suppression (Cowling et al. 1997;Western Cape Government 2022). Accordingly, regular fires of high intensity have been recommended to curb the invasion of thicket and promote the persistence of fynbos in dune fynbos-thicket landscapes (Vlok and Euston-Brown 2002;Vlok et al. 2008). Cowling and Hoffman (2021), however, using repeat photography, showed that thicket invasion in the study area is largely restricted to hollows and lower slopes of dune ridges where the water table is relatively close to the surface; on the drier upper slopes and crests of ridges, fynbos persists despite decades without fire.
The fact that most small (≤ 10 cm basal diameter) resprouting thicket shrubs can survive and resprout successfully after severe disturbances only 5 years apart, (i.e. an intense fire in 2016 followed by severe fire and browsing treatments in 2021) suggests that these species are very resilient to frequent and severe disturbances causing aboveground biomass loss. In fact, the severity of the simulated fire treatment in this study may be considered extreme and unlikely to happen under natural conditions. This is because Fig. 5 The resprouting vigour of dune thicket species expressed as a resprouting shoot count, b resprouting shoot height, and c resprouting shoot canopy cover 12 months after application of the treatments, fire or browsing. Median (lines), 25-75 quantile ranges (boxes), 1.5* interquartile ranges (whiskers) and outliers (dots) are shown there is unlikely to be sufficient available (~ dry or dead) fuel in 5-year post-fire vegetation on nutrient-poor dune sands to sustain a severe fire that would completely consume young green plant material, even under extreme weather conditions. Our results thus suggest that thicket is likely to persist in the fynbos matrix even if subjected to frequent fires under high fire-danger weather conditions, and will likely expand in fire-free conditions, via establishment of ramets and bird dispersed propagules (Cowling et al. 1997;Cowling and Hoffman 2021). Severe burning, even at short rotation, is thus unlikely to reverse established thicket patches as the component species show strong resilience. However, wildfires or controlled burning at intervals of 10-15 years (Vlok and Euston-Brown 2002) may facilitate the persistence of populations of fire-dependent fynbos where these co-occur with thicket by providing transient niches for their reproduction (van Wilgen and Forsyth 1992; Cowling and Pierce 1988;Cowling et al. 2019). Dune swales and lower slopes commonly harbour threatened fynbos species which likely have persistent soil-stored seed banks, for example Aspalathus recurvispina and Agathosma stenopetala (Pierce and Cowling 1991;Cowling et al. 2019), rendering fire management a priority.
Unfortunately, we know nothing about the likely impacts of browsing by megaherbivores on the structure and distribution of fynbos and thicket in dune landscapes. Furthermore, the prospects of re-introducing African elephant and black rhinoceros to these dune landscapes are very slight, given the small extent of extant dune habitats and relatively high human population density of these landscapes. However, based on data from inland thicket, which does support large populations of megaherbivores, megaherbivores can have a major impact on thicket structure, reducing shrub stature and creating large gaps between thicket clumps (Cowling and Kerley 2002;Kerley and Landman 2006). These impacts may have interacted positively in dune landscapes with fire in that intense browsing may have reduced shrub vigour, and hence resilience to fire, and gaps filled by flammable fynbos may have resulted in high-intensity fires that would also have reduced thicket shrub resprouting vigour. Evidence from savanna shows that the interaction of fire and megaherbivore browsing increases tree mortality significantly (Langevelde et al. 2003;Holdo et al. 2009). However, we can only speculate about whether the same applied to dune fynbos-thicket landscapes, and whether regimes of combined fire and browsing could offer a means of managing fynbos-thicket dynamics. Similarly, the effect of multiple successive short interval fires on dune thicket is not known but we expect it to cause a slow decline in survival and resprouting vigour of thicket shrubs in the long term (Bond and Midgley 2001;Medeiros and Miranda 2008).
Given that thicket species show resilience to short interval disturbance, they will likely persist in times of global change (e.g. climate change, increase of alien invasive trees and urban expansion) which may bring about changes in the frequency and severity of fire events (Flannigan et al. 2000;Moriondo et al. 2006;Syphard et al. 2007;Archibald and Bond 2003;Kraaij et al. 2011Kraaij et al. , 2013bKelly et al. 2020). Frequent or very severe fire may be harmful to dune forest-thicket patches which occur in fire-refugia within the fynbos matrix and therefore seldom experience fire (Strydom et al. 2022). This may in the long term reduce cover of, and eventually extirpate, fire-sensitive species associated with these habitats. Dune forest-thicket and thicket species which typically experience fire once every 50-100 years (e.g. Apodytes dimidiata, Chionanthus Fig. 6 The probability of resprouting of 10 dune thicket shrub species at 1, 3, 6 and 12 months after application of the treatments, fire or browsing, as predicted by a Cox proportional hazards regression model. The resprouting probability at each time interval is indicated by a cross. Shaded bands show 95% confidence intervals foveolatus, Clausena anisata, Dovyalis rotundifolia, Cassine peragua, Scolopia zeyheri) exhibit comparatively weak survival and resprouting vigour after severe fire, especially in smaller individuals (Strydom et al. 2020). Correspondingly, small individuals of vertical-growing species exposed to a short return interval and severe fire in the current study (i.e. Clausena anisata, Dovyalis rotundifolia, Scolopia zeyheri) showed weaker survival and resprouting vigour than species from other architectural guilds. However, forest-thicket tree and shrub species rarely occupy high fire-exposed sites and are unlikely to experience high fire frequency or severity (Strydom et al. 2022).
In conclusion, our simulated fire experiment was conducted after a previous fire event, thereby testing the impact on dune thicket shrubs of two fires occurring within a 5-year period. The findings indicate that these thicket species display a remarkable resilience to even high-frequency, highseverity fires. This suggests that short return interval fires (5-15 years) will merely maintain the co-occurrence of dune fynbos and thicket vegetation by setting back thicket growth rather than causing large-scale mortality of thicket shrubs, and thereby allowing fynbos seedlings and resprouters to regenerate in the post-fire environment (van Wilgen and Forsyth 1992; Cowling and Pierce 1988;Cowling et al. 2019).