Plant resprouting: How many sprouts and how deep? Flexible modelling of multi-species experimental disturbances

Highlights

• 39 plant species were experimentally damaged and the depth distributions of sprouts emerging after varied over >60 mm.

• Functional regression models displayed sprout distributions and revealed influences of species attributes.

• Sprout distributions were related to treatment and species growth forms and measurable traits.

Abstract

Plants can respond to damage through regrowth from meristematic tissue, commonly concentrated in buds. Knowledge of the distribution and abundance of buds within plants can assist in understanding individual, population and species responses to disturbances. Yet belowground buds can be hard to study, as they may be cryptic until initiated in response to damage. Sprouts emerging following disturbance can be studied more easily and represent a realised bud bank.

Here we aimed to characterise and compare the vertical distribution of sprouts from a previous study of clip and burn treatments for 42 species of grasses, forbs, subshrubs and woody plants in semiarid eastern Australia (Vesk et al., Oikos 2004 107:72–89). We tested whether depth distributions of sprouts could be explained by growth forms and plant traits, while also asking whether clip and burn treatments influenced those distributions. We combine function regression with a hurdle model to estimate simultaneously the probability of resprouting and the depth distribution of buds, conditional on resprouting (hereafter, sprouts). Function regression models the depth distribution of sprouts as a continuous function, rather than a single value. The hurdle model also allowed us to ask whether sprout depths and the probability of resprouting were similarly related to treatments, growth forms and traits.

Depth distributions of sprouts in successfully resprouting plants differed among growth forms. Specifically, we observed similar distributions of sprouts in forbs and woody plants, more and deeper sprouts in grasses, fewer and mainly above-ground sprouts in subshrubs. However, because the probability of subshrubs resprouting was very low, the expected number of sprouts was very low too. Burning resulted in more and deeper sprouts on successfully resprouting plants. Further, species’ traits like maximum height and specific leaf area (SLA) were correlated with different sprout distributions. Taller species had fewer shallow sprouts in general, and high SLA species had more shallow sprouts under clipping but not under burning.

This work points to experimental methods and statistical analyses to quantify the depth distribution of the buds from which plants resprout. This serves to inform comparative investigations of buds, sprouts, and disturbance responses, and is likely to improve predictions of responses to disturbances.

Introduction

Understanding of morphological and anatomical traits behind resprouting following major damage to above- or below-ground plant parts has lagged behind that of other functional traits (Klimešová and Klimeš, 2007; Ottaviani et al., 2017). Yet disturbances are a fact of plant life and provide opportunities for vegetation change. How plants equip themselves for recovery from damage is central to regeneration ecology, occupancy dynamics and species turnover (Clarke et al., 2013; Pausas et al., 2018). Recent work has shown that, in herbs, clonality and bud bank traits have significant associations with vegetation types (Klimešová and Herben, 2015) and species’ distributions in Czechia can be better predicted by clonality and bud bank traits than Leaf-Height-Seed traits (Klimešová et al., 2016).

Studies have moved to estimate the resilience of a community, or its capacity to recover from disturbance, by excavating and counting below ground buds from slabs of soil (Benson et al., 2004; VanderWeide and Hartnett, 2015). Through this approach, disturbance (grazing and fire) has been shown to have complex and apparently contradictory effects on the density of buds per shoot, with studies showing both decreases (VanderWeide and Hartnett, 2015) and increases (Fidelis et al., 2014). Bud density is one aspect of the bud bank and may work well concerning vegetation as a unit. When considering individual plants of particular species however, one may consider the number and position of buds from which regrowth occurs. Number and position may be more-representative indicators of the ultimate resprouting success of an individual plant experiencing damage (Gill, 1975). As a matter of logic, at least one bud must survive if a plant is to re-establish canopy. If all buds have an independent probability of sprouting, the greater the number of buds, the greater the chance of resprouting and survival (Malanson and Trabaud, 1988; Vesk and Westoby, 2004a). Indeed, a model positing species resprouting probability as a binomial function of the probability that at least one stem resprouts successfully explained over 40% of deviance across 43 species of semi-arid Australian plants (Vesk and Westoby, 2004b). The position of buds directly affects responses to disturbance. Buds at lower heights on the stem, or at greater depths below ground, are potentially more likely to escape damage and resprout, at least for those disturbances whose damage severity increases with height above the ground like livestock grazing and fire.

One can consider the buds that arise following damage, hereafter sprouts, as a statistical population with two attributes: number and depth (or height). All else being equal, one might expect that species that have more or deeper buds belowground should be more likely to resprout. Analysing data on buds and sprouts poses the problem of what to model. In a preliminary analysis of some of the data reported here, the maximum sprout depth was used as a species-level trait (Vesk et al., 2004). That analysis determined that species with sprouts at greater maximum depths were more likely to resprout successfully after clipping and burning. But this is unsatisfactory in several ways. First, the maximum is not necessarily a reliable indicator of the population as a whole. Second, maximum depth will be related to number; all else being equal, increases in the number of buds will increase the chances of an extreme value. Third, number and depth are distinct quantities and may reflect different processes. Last, different disturbance events or treatments might stimulate sprouting in different buds—shallow vs. deep. Ideally, one would model the whole population of buds or sprouts. Function regression provides such an opportunity, whereby the response variable is not a single value, but a smooth function, thus fitting a curve (Yen et al., 2015).

In this paper we use a function regression model, with a hurdle component, to analyse data from a field experiment of clipping and burning across 42 plant species from semi-arid NSW, Australia. We aim to model the probability of resprouting and the depth distribution of sprouts (buds, conditional on successful resprouting) in each species. We aimed to address four questions. First, do populations of sprouts differ in number and depth among growth forms? Second, does clipping allow some sprouts to emerge from above ground or at shallower depths than a combined clipping and burning treatment? Third, are species’ traits associated with the probability of resprouting or distribution of sprout depths? Fourth, are associations between traits and sprout depths coordinated with associations with the probability of resprouting? We classified species into four growth forms (forbs, grasses, chenopod subshrubs and woody plants) and characterised species on the basis of three functional traits (maximum plant height, mean number of stems and specific leaf area).

For our first question, we expected that grasses would have greater numbers of sprouts due to their growth habit based on tillers (Briske, 1996). Because grasses and forbs lack persistent stems, they grow from the plant base, relying on a bud bank, thus one might expect many sprouts emerging at or above ground level. Woody species may have buds distributed along their persistent stems and they may not be concentrated at the base of the plant. Woodiness may potentially provide protection to buds, though may also be associated with fewer buds in general, owing to the greater differentiation of cell types in wood formation and physical obstruction of sprout emergence (Vesk and Westoby, 2004a).

For question two, we expected the burning treatment to cause heat death in shallow tissues, with a naïve expectation that total loss of all canopy would stimulate the entire bud bank of a plant. Therefore, one possible outcome of our treatments would be a left-truncation of the sprout-depth distribution under the burn treatment due to the heat death of shallow buds. Alternatively, not all buds may be stimulated to sprout following disturbance, with shallower buds preferentially stimulated, and deeper buds remaining dormant if not needed. That would be reflected by sprout emergence at greater depths under the burn treatment relative to the clip treatment. This may occur because canopy re-establishment is subsidised by belowground reserves, which would be unnecessarily depleted if too many buds were stimulated.

Regarding our third question, we included maximum plant height and the number of stems because resprouters tend to be shorter and have more stems than nonsprouters (Midgley, 1996; Vesk et al., 2004; Vesk and Westoby, 2004c). We examined specific leaf area because of its role in leaf economics (Wright et al., 2004) and the fast-slow continuum (Reich, 2014) and because Hernández et al. (2011) found that Mediterranean resprouters had higher specific leaf area on average.

Lastly, for our fourth question, we expected that growth forms and traits associated with sprout number and depth in successfully resprouting plants would similarly associate with resprouting probability.