Climate controls plant life‐form patterns on a high‐elevation oceanic island

Plant life‐forms characterize key morphological strategies that enable large‐scale comparisons of plant communities. This study applies Raunkiær's plant life‐form concept that was developed for temperate climate to a subtropical island flora, in parts, dominated by summer aridity. We quantify how plant life‐form patterns as well as patterns of important plant functional traits (PFTs) relate to important climate and topographic characteristics.


| INTRODUC TI ON
Plant life-forms are an established ecological approach to classify plant species based on their environmental demands. They help understand environmental and ecological characteristics of ecosystems (Klimes, 2003). Particularly, plant-climate interactions based on the location of reproductive organs in relation to the soil surface giving insights how plants can withstand and recover from disturbance and stress are well represented (Lavorel & Garnier, 2002;Raunkiaer, Fausbøll, Gilvert-Carter, & Tansley, 1934). Classification of plant species into functional groups (e.g. life-form types) provides the advantage of global comparability of research results despite regional differences in species taxonomy (Diaz, Cabido, Zak, Carretero, & Aranibar, 1999;Diaz et al., 2016). Indeed, a plant life-form approach for areas of high taxonomic distinctness due to endemism such as islands seems particularly promising to allow standardized inter-island comparisons with a functional island biogeographic perspective.
Due to its compelling simplicity, the life-form classification system of Raunkiaer et al. (1934) is the most used plant classification concept in the world (Klimes, 2003;Leuschner & Ellenberg, 2017).
It is based on key survival strategies of plants, most importantly the height of the renewable buds of plants relative to the soil surface layer during unfavourable seasons. Raunkiaer's classification system was thus developed for temperate seasonal climates where winter frosts end the growing season (Ewel & Bigelow, 1996;Mueller-Dombois & Ellenberg, 1974). Despite some additional refinements of his classification system (e.g. Cain, 1950;Ellenberg & Mueller-Dombois, 1967;Ewel & Bigelow, 1996;Mueller-Dombois & Ellenberg, 1974), five basic classes for terrestrial plants are still in wide use (phanerophytes, chamaephytes, hemicryptophytes, geophytes and therophytes). Because of the sole focus on regenerative buds during the unfavourable season, a reduced usability in warmer climates (e.g. subtropics or tropics) seems obvious. However, other seasonally occurring unfavourable conditions such as drought or heat can likewise restrict plant growth .
Taxonomy independent classifications for plant strategies in ecosystems are particularly useful for large-scale comparisons and for answering evolutionary questions related to species functioning (like convergent evolution; Chave et al., 2009;Diaz et al., 2016;Lavorel & Garnier, 2002;Reich, 2014;Wright et al., 2004).
Here we expand Raunkiaer's life-form concept to assemblages on oceanic islands, which are considered model systems for understanding the evolution of species, often called 'nature's test tubes' where ecological patterns can be observed in independently repeated evolutionary settings on different islands (Losos & Ricklefs, 2009;Whittaker, Fernández-Palacios, Matthews, Borregaard, & Triantis, 2017). Island taxa often show island-specific adaptations such as secondary island woodiness, that is, the evolution of woody life-forms on islands from predominantly herbaceous lineages (Burns, 2019;Carlquist, 1974;Darwin, 1859). On the Canary Islands, for example, secondary island woodiness appears in more than 50% of endemic species including prominent genera such as Aeonium, Argyranthemum, Crambe, Echium and Sonchus (Lens, Davin, Smets, & del Arco, 2013). Furthermore, oceanic islands are ideal study systems to answer ecological as well as evolutionary questions as they often represent climatic 'mini-continents' with steep environmental gradients (Darwin, 1859;Irl et al., 2015) along which systematic changes in the survival strategy of plant and thus the life-form spectrum can best be quantified and tested (Hoffmann et al., 2019;Klimes, 2003;Pavon, Hernandez-Trejo, & Rico-Gray, 2000). Spatial climatic heterogeneity is amplified by the influence of trade winds in many tropical and subtropical islands, which leads to an asymmetric distribution of precipitation (Garzón-Machado, Otto, & Aguilar, 2014). The large environmental heterogeneity of islands on a relatively small scale has the advantage of a clearly definable extent of the study area, ideally without external influences (but see Ibanez et al., 2019).
Furthermore, results of the study are also valuable for larger spatial scales (e.g. for continents) because the investigated responses of life-forms to climatic parameters can be easily transferred.

Examining individual functional traits offers an alternative view
to plant life-forms that are an interacting aggregation of individual traits and functions (Diaz et al., 2016). Thereby, we refer to functional traits as 'any morphological, physiological or phenological feature measurable at the individual level' that links functional traits to the performance and fitness of individuals (Violle et al., 2007).
Such traits are well expressed in plants that are able to internally store water for extended time periods to survive hygric stress (hereafter referred to as succulents, Ellenberg, 1981) and nitrogen fixers that can fix atmospheric nitrogen and have an increased water use efficiency under hygric stress as a result of their metabolism (hereafter referred to N-fixers, Reich et al., 2001). The evolution of traits associated with succulence and nitrogen fixing are thus remarkable adaptations to harsh environmental conditions, enabling these species to occupy environmental niches that are not accessible otherwise (Swenson et al., 2012).  Kreft, Jetz, Mutke, Kier, & Barthlott, 2008), where they likely reflect habitat diversity and climatic gradients. Here we analyse plant life-form patterns on very high spatial scales within islands. We expect this variability to be better reflected by climate than to small-scale topography (slope, aspect). We thus expect (a) the percentage of woody plants (phanerophytes and chamaephytes) to increase with the amount of rainfall reflecting a global tendency of tree diversity with increasing moisture conditions (Bhattarai & Vetaas, 2003). (b) The percentages of phanerophytes and chamaephytes will increase with elevation (Vazquez & Givnish, 1998) resulting from an adaptation to colder climates including frost. Non-woody plant species are likely to have a heterogeneous spatial distribution because they can strongly depend on local parameters, for example soil type and canopy cover (Bhattarai & Vetaas, 2003). Therefore, we expect (c) that the percentage of hemicryptophytes, geophytes and therophytes is unrelated to elevation and precipitation. N-fixers, in contrast have crucial adaptation mechanisms (e.g. nitrogen fixation, high morphological and physiological diversity) that allow them to cope better with harsh environmental conditions than other plant life-forms. Therefore, they are expected (d) to be more important in lower and higher altitudes where thermal and hygric stress limits plant growth of other life-forms (Haffner, 2011). Succulent plants are known to have the ability to store water and use it in times of dry periods (Willert, 1992). Because of their water use efficiency, we expect to find (e) succulent plants in arid regions of the island in the lower elevations because succulents are likely to be outcompeted when climatic conditions improve (Otto, Fernandez-Palacios, & Krusi, 2001;Vendramini et al., 2002).

| Study area
La Palma (28°26′ to 28°51′N and 18°00′ to 17°43′W) is the most northwestern island of the Canary Islands archipelago located in the Atlantic Ocean at the western continental margin of Northern Africa. The island is of volcanic origin and has its highest summit (the Roque de los Muchachos) at 2,426 m within a total island area of only 708 km 2 . On a landscape scale, La Palma shows great topographic and climatic variability . Due to the longitudinal po-  . In winter months, occasional snow falls in the summit region (Kunkel, 1993) and frost events down to 1,500 m a.s.l.
are possible (Hohenester & Welss, 1993). High solar radiation occurs at high elevations that causes high potential evapotranspiration,

| Life-form classification and environmental data
The classification system of Raunkiaer is one of the most frequently used in ecology (Ellenberg & Mueller-Dombois, 1967;Ewel & Bigelow, 1996;Mueller-Dombois & Ellenberg, 1974). As the system is based on the height of reproductive organs, it needs modifications to fit for island taxa and subtropical climate conditions. Species in the subtropics often show adaptations to unfavourable conditions like aridity, many of which are not covered by the standard Raunkiaer approach. Moreover, Raunkiaer neglected the growth phase during favourable conditions (Mueller-Dombois & Ellenberg, 1974). We therefore modified the concept allowing the classification of all species into five main plant life-forms ( Figure 2). The adapted approach classifies species according to their growth strategy. It is expected that the selected distinguishing species characteristics have an effect on the presence/absence of certain plant life-form types along environmental gradients (Klimes, 2003;Lavorel & Garnier, 2002).
Especially, phanerophytes and chamaephytes can show a high variation with precipitation (Bhattarai & Vetaas, 2003), whereas herbs such as hemicryptophytes, geophytes and therophytes are considered to be driven by the interplay of climatic factors with their local environmental components (e.g. canopy cover, disturbance regime) as well as topographic conditions such as slope and aspect (Vazquez & Givnish, 1998).
Additionally, they have to have the ability to use stored water during dry conditions. Because of this definition, the degree of succulence is not essential (Jacobsen, 1974;Willert, 1992). As a consequence, even slightly succulent plants like Rumex lunaria are considered as succulent species (Schönfelder & Schönfelder, 2012). Non-native species were not considered in this study.

| Plant species data and species distribution modelling
Information on the spatial occurrence of vascular plant species was available from the Banco de Datos de Biodiversidad (Atlantis 3.1 at www.biodi versi dadca narias.es), a long-term governmental initiative to assemble and complete all known distribution records of species on the Canary Islands (see Steinbauer, Field, Fernandez-Palacios, et al., 2016 for discussion of data quality). Occurrence records are reported in a resolution of a 500 m × 500 m grid size (3,063 grid cells covering the island of La Palma, mean number of occupied grid cells per species was 976). The data are presence-only information with much better coverage for endemic species than for native non-endemics (Steinbauer, Field, Fernandez-Palacios, et al., 2016).
Life-form type distribution was analysed after interpolating species We thus stacked individual species distribution models for each species to calculate probability-based richness for different life-forms (Grenié et al., 2020). This spatial interpolation is conceptually similar to gap-filling methodology frequently applied in macroecological studies including trait imputation that are accepted tools for reducing sampling bias (Penone et al., 2014). This partly corrects for spatial sampling biases. SDMs were implemented using generalized linear models with automated variable selection using small sample sizecorrected Akaike information criterion.
Potential explanatory variables were aspect, slope, mean annual temperature and precipitation as well as mean monthly temperature and precipitation for January, April, July and October representing

| Statistical analysis
The relation of life-form type distribution with elevation, precipitation slope and aspect (northness and eastness) was analysed by means of generalized linear models. A Poisson error distribution was used for the overall number of species associated with a specific Models were implemented as unimodal relationships (y ~ x + x 2 ) and the quadratic term dropped based on AICc using the R step function for model optimization. As a goodness-of-fit measure, pseudo-R 2 were calculated using r package modEvA version 1.3.2 (Barbosa, Brown, Jimenez-Valverde, & Real, 2016) following the approaches suggested by Guisan and Zimmermann (2000). As elevation and precipitation were much better explanatory variables for plant life from distribution and plant functional trait distribution than slope, northness and eastness, we only show results including elevation and precipitation in the main paper, while the results for slope, northness and eastness can be found in Appendices S1 and S2.
Spatial patterns in life-form type distribution were visualized using r packages raster version 2.8-4 (Hijmans, 2018). All analyses were performed in R version 3.3.2. Plant life-forms were assumed to follow climate-driven spatial pattern, although different plant life-forms showed different reactions.

| Plant life-form distribution
The number of phanerophytes showed a hump-shaped distribution both with elevation and precipitation, although the relationship with elevation was substantially right-skewed (Figures 3a and 4a). This led to a spatial pattern that showed highest values at low elevations in the northeastern zone of the island. Percentage of phanerophytes increased with elevation and precipitation (Figure 3b). Topographic variables had little effect on life-form distribution, although phanerophytes are an exception. Their number and percentage increased with the steepness of the slope (Appendix S1). This spatial pattern highlights the high contribution of trees and large shrubs to the humid laurel forest found at steep mid elevations on the northeastern side, where they contributed around 40% of species to the respective plant communities.
The number of chamaephytes showed a weak U-shaped distribution with elevation, whereas it showed a hump with precipitation ( Figures 3c and 4b). This led to a heterogeneous distribution with low values in laurel forest and high values on northern coasts and the Caldera de Taburiente complex in the centre of the island. Percentage of chamaephytes showed a strong positive relationship with elevation but a non-significant relationship with precipitation ( Figure 3d).

The number of hemicryptophyes decreased with elevation and
showed a right-skewed hump with precipitation (Figures 3e and 4c). The resulting spatial pattern revealed hot spots of the number of hemicryptophytes at low elevations, mainly in the north and east, and cold spots at high elevations. The percentage of hemicryptophytes increased with elevation and precipitation (Figure 3f). The spatial distribution revealed that hot spots were located at mid to high elevations, in some parts contributing more than 30% of species to the respective plant communities.
The number of geophytes decreased with elevation and precipitation (Figures 3g and 4d). Thus, hot spots of the number of geophytes were found at low elevations (mainly on the northern to eastern side), while the cold spots were located at high elevations.
We found a similar relationship of elevation and precipitation with % geophytes (Figure 3h). However, geophytes only contributed a maximum of around 4% of species to the respective plant communities.
The number of therophytes decreased with elevation and showed a strongly right-skewed relationship with precipitation (Figures 3i and 4e).
As a result, we found the highest number of therophytes at low elevations. Similarly, the percentage of therophytes decreased strongly both with elevation and precipitation ( Figure 3j) and also slope (Appendix S1).
Hot spots of % therophytes were found on flat ground at low to mid elevations. In some parts, theropyhtes contributed up to 45% of species, indicating the importance of this plant life-form for low and mid elevation systems that are often subject to current or historic human land use.

| Plant functional traits: Succulence and N-fixers
In order to assess the influence of climate on the spatial distribution of the two plant functional traits. succulence and N-fixers, we correlated absolute and relative values against elevation (as a proxy for temperature) and precipitation. All relationships presented in this section were strongly significant with p < 0.001.  Under such favourable growing conditions with very little seasonality in temperature and precipitation (Weigel et al., 2018), short-lived species such as therophytes and low stature species (geophytes and hemicryptophytes but also dwarf shrub chamaephytes) are outcompeted by tall-growing woody species (trees, large shrubs) that produce a dense and dark canopy (Delgado et al., 2007). Merely as a result of their large structures, phanerophytes have a competitive advantage over other plant life-forms, indicating a climate-driven selection of tall-growing plant life-forms under humid and mild conditions.

| Climate drives distribution of woody life-forms
In the cool but arid high elevation summit scrub, large shrubs with a hemispheric growth form (e.g. Adenocarpus viscosus subsp. spartioides, Genista benehoavensis, Teline stenopetala, Spartocytisus supranubius or Chamaecytisus proliferus subsp. proliferus, Echium gentianoides) dominate the vegetation. As trees are more closely coupled to atmospheric processes (especially low temperatures and high winds speeds at high elevations) and cannot profit from latent heat coming from the high solar radiation warming the soil, the growth form 'tree' reaches its elevational limit at lower elevations than shrubs or other low stature species, forming the widely studied alpine tree line (Körner, 2012).
On the Canary Islands, alpine tree line is reached at around 2,000 m a.s.l. and is modulated by local climatic processes such as aridity or the mass elevation effect . Thus, at high elevations a highly specialized endemic shrubby flora has developed  possessing many species that are often entirely restricted to this system . Indeed, the summit scrub is an evolutionary very active system as its high degree of (neo-)endemism indicates (Merckx et al., 2015;Steinbauer, Field, Grytnes, et al., 2016).
This suggests that the life-forms dominating in this system strongly reflect a climate-speciation coupling in the sense that the current pattern of plant life-forms is directly shaped by in situ climate conditions leading to the evolution and persistence of certain life-forms.

| Modulation of the non-woody plant life-form distribution
Interestingly, hemicryptophytes and therophytes show a strongly opposing pattern. The relative contribution of hemicryptophytes increases with increasing precipitation and decreasing temperature, while it is vice versa for therophytes. Two non-mutually exclusive explanations are likely: (a) On the arid leeward side of the island, where therophytes are more important, inter-annual precipitation variability is high, making precipitation events much less predictable (Jiang et al., 2017). In unpredictable environments it is advantageous to be short-lived (such as therophytes) in order to take advantage of specific precipitation events and then survive long periods of unfavourable conditions as a seed.
Nevertheless, in the driest and most arid regions of the island, the share of perennial hemicryptophytes increase again, likely because perennial hemicryptophytes might have effective perennial structures below the ground to survive water-limited periods.
(b) Pinus canariensis forests cover large areas of the leeward part of the island, especially at mid elevations, that are particularly prone to reoccurring wildfires (Molina-Terrén, Fry, Grillo, Cardil, & Stephens, 2016). This archipelago endemic pine species has up to 30 cm long needles, a thick bark and the unique ability to resprout from epicormics shoots from all above-ground organs-all remarkable features that either promote fire or enable the tree to survive fire events (Climent et al., 2004). For understorey vegetation, it makes sense to be adapted to reoccurring fire events, and being a short-lived therophyte might be a good strategy. Therophytes can take advantage of the low competition and high resource availability of post-fire conditions to complete their life cycle before other, more competitive life-forms emerge (Arévalo, Fernández-Palacios, Jiménez, & Gil, 2001). One has to note though that this system is also strongly affected by browsing damage caused by introduced herbivores (Cubas et al., 2019), which might additionally alter plant life-form patterns. In sum, we probably see a modulation of the generally climate-driven plant life-form pattern by local influences such as disturbances in the case of hemicryptophytes and therophytes. spartioides being by far the most abundant as a result of introduced herbivore browsing (Irl et al., 2012(Irl et al., , 2014. The high elevation summit scrub is climatically cool but arid (del Arco-Aguilar et al., 2010).

| Only minor influence of topography on plant life-forms and plant functional types
Thus, plant species, such as N-fixing legumes, that possess traits associated with high water use efficiency will have a competitive advantage (Reich et al., 2001). At mid elevations, N-fixers are an important species group in the undergrowth of the Pinus canariensis forest (e.g. Lotus spp., C. proliferus subsp. proliferus, Adenocarpus foliolosus, Cicer canariensis and near the tree line also S. supranubius).
This might result from the fact that the leeward pine forest is (a) relatively dry supporting species with high water use efficiency (Reich et al., 2001), and (b) the pine forest is prone to frequent reoccurring fire events (Molina-Terrén et al., 2016). Similar to therophytes, fast growing N-fixers can profit from low competition and high resource post-fire conditions, as can be seen with the endemic herbaceous N-fixer Lotus campylocladus subsp. hillebrandii that covers large areas of the pine forest floor only a few months after a fire event, turning it bright yellow (pers. obs.).
As expected, succulents are favoured at low elevations, especially on the dry leeward coast of the island. Indeed, these results are consistent with Ellenberg (1981) who found that succulents dominate at regular amounts of annual precipitation between 100 and 200 mm-approximately the average annual precipitation at the leeward coast of La Palma . Besides being able to store water internally, succulents have a higher water use efficiency than other plant functional types because many are CAM plants (e.g. many species from the Crassulacean family; Mort, Solits, Soltis, Santos-Guerra, & Francisco-Ortega, 2007). At high elevations regular winter frost and ice storms likely regulate the distribution of succulents, while at high precipitation, for example, in the perpetually humid laurel forest, being succulent does not result in a competitive advantage over other functional traits as storing water is uneconomical if water is abundantly available (Teeri, Stowe, & Murawski, 1978).
Thus, succulence on La Palma has a clearly defined climatic niche restricting it to low and dry elevations.

| CON CLUS ION
Using the strong environmental gradients offered by the climatic 'mini-continent' of La Palma, we identify strong variation in plant life-forms and plant functional traits. The strong patterns in lifeforms underline the importance to acknowledge within-island differences in theory of island biogeography. Treating an island as a single unit does not account for the fact that different parts within that island differ considerable in their connectivity to surrounding landmasses and thus in their species source pool and isolation (Steinbauer, 2017). The growth strategies reflected in Raunkiaer's plant life-forms suggest differences in species establishment and coexistence dynamics within different parts of the island's climate space. The within-island differences have considerable consequences for the processes of colonization and extinction that should reflect the here identified within-island patterns.
Our study showed that topographic variables were of subordinate importance for plant life-forms and plant functional traits on La Palma, although they have been shown to be important in other (island) contexts (Stein, Gerstner, & Kreft, 2014) together with further aspects of topography such as geology or habitat diversity (Keppel, Gillespie, Ormerod, & Fricker, 2016;Otto et al., 2016). In addition, important ecological relationships such as the speciesarea relationship can vary among plant growth forms on small islands (Schrader et al., 2020).
The strong plant life-form-climate relationship as well as the plant functional trait-climate relationship suggest that climate changethat will substantially alter island climatic conditions in the near future -will affect the future distribution of plant life-forms and plant functional traits. How this might relate to ecosystem stability and functioning as well as ecosystem services warrants further research on this topic. Overall, we show that the concepts of plant life-forms and plant functional traits are relatively easily adapted to a new flora unrelated to the original intentions of the concepts. By applying both concepts to island floras that are often quite unique as a result of the high endemism, inter-island as well as inter-archipelago comparisons become feasible and potentially very interesting for future island biogeographical studies, especially in the context of the emerging field of functional island biogeography.

ACK N OWLED G EM ENTS
C.B. was supported by the European H2020 Project 641762 ECOPOTENTIAL: Improving future ecosystem benefits through Earth Observations.

DATA AVA I L A B I L I T Y S TAT E M E N T
Data collected for this manuscript is accessible in DRYAD online repository (Irl et al., 2020)