The spatiotemporal distribution of pollen traits related to dispersal and desiccation tolerance in Canarian laurel forest

Questions: Pollen traits (e.g., size, wall thickness, number of apertures) have been suggested to be relevant in terms of pollination mechanisms and the ability of the male gametophyte to withstand desiccation. We examined the spatiotemporal distribution of pollen traits related to dispersal (ornamentation and dispersal unit) and desiccation tolerance (wall thickness, presence of furrows and pores and pollen size). Specifically, we address two questions: how are the pollen traits distributed in relation to different levels of aridity? And, how did the pollen trait composition change with changing past environmental conditions? Location: forests La Gomera and Tenerife (Canary Islands). Methods: We used pollen rain from 19 forest plots on an elevational gradient of 1050 m and all laurel forest types (cold, dry, humid and ridge crest) to quantify pollen trait composition using community- weighted means. In addition, we used fossil pollen to examine the composition of pollen traits over 9600 years in response to known intervals of regional past climate change. Results: Our results demonstrated increased prevalence of desiccation tolerance-related pollen traits over drier areas of the laurel forest distribution. We also found increased prevalence of rich pollen grain ornamentation in the core of the laurel forest distribution. Holocene pollen functional diversity increased during a trend towards drier conditions as did the proportion of pollen grains with apertures and thicker walls to indicate desiccation tolerance. Conclusions: Our study provides the first step towards understanding the role of pollen traits when quantifying the dynamics of different plant communities.


| INTRODUC TI ON
Identifying plant traits that enable plant species to cope with environmental change has been highlighted as an important research goal (Aguirre-Gutiérrez et al., 2022). It allows assessing how plant communities respond to shifts in climate and to identify those plants that might be more resilient to past, current and future environmental change (Díaz et al., 2016;Kühn et al., 2021). However, pollen traits have been largely overlooked. Our paper contributes to this research through the analysis of pollen morphological traits related to desiccation tolerance and dispersal from laurel forest trees and shrubs on the islands of La Gomera and Tenerife, Canary Islands ( Figure 1). The overall aim of the paper is to shed new light on how the distribution of pollen dispersal and desiccation traits vary over space in the Canarian laurel forest and through time using palaeoecologic records.
Viable pollen grains are important for plant dispersal and fitness (Dafni & Firmage, 2000). It has been suggested that specific pollen traits might be significant for plant fitness (Franchi et al., 2011;Pacini & Hesse, 2012;Abrego et al., 2017;Seddon et al., 2021). In particular, changes in the moisture level may differentially affect the viability of pollen with high and low desiccation tolerance, which finally impacts plant dynamics and survival (e.g., Franchi et al., 2011). For example, thin pollen walls have been documented for taxa living in environments of high relative humidity as thinner pollen walls provide an advantage in terms of more rapid germination due to shorter rehydration times (Osborn et al., 2001;Pacini & Hesse, 2012). Another example is the relationship between the number of apertures (e.g. furrows and pores) and desiccation tolerance (Franchi et al., 2011). In pollen grains with low desiccation tolerance (recalcitrance), furrows are absent and there may be an absence of pores too (Franchi et al., 2011).
The absence of apertures prevents the loss of water from the pollen grain, enabling immediate germination if moisture becomes available (Franchi et al., 2011). On the contrary, pollen grains with high desiccation tolerance (orthodoxy) present furrows that allow changes in pollen volume during dehydratation and rehydration phases (Tweddle et al., 2003). In addition, a thicker pollen wall and larger pollen size with a lower surface-to-volume ratio are thought to protect the grain from desiccation and are suggested to be prevalent in the forest edges, where ultraviolet (UV) light is more intense and environmental conditions are harsher (Pacini & Franchi, 1999;Ejsmond et al., 2011;Norros et al., 2015). We are also interested in analysing the relationship between the variation in the exine sculpturing, pollen size and shape and the dispersal mode. However, this relationship is a subject of debate and a source of uncertainty (Vaknin et al., 2000;Norros et al., 2015;Abrego et al., 2017). There is some evidence of a relationship between rich ornamentation and entomophily, as it is thought that ornamentation plays an important role in the attachment to insects and to the stigma of the flower, but see Hesse et al. (2000) and Konzmann et al. (2019) for further discussion. Pollen grains of windpollinated species often lack elaborate sculptures and appear smooth (i.e., psilate) and/or they have sac-like attachments (saccate) to help in their transport through the air (e.g., Pinus and many other coniferous species) (Linder, 2000;Williams, 2010).
Finally, we apply the pollen trait approach to palaeoecological data sets as a new aspect in the reconstruction of past vegetation dynamics. The change recorded in the laurel forest communities was hypothesized to be related to a regional climatic shift towards drier conditions (de Menocal et al., 2000;Kröpelin et al., 2008;Nogué et al., 2013). Whereas changes in the forest community from La Gomera were related to the increasing drier condition during the end of the African Humid Period (Nogué et al., 2013), the Tenerife core covered only the last 4700 calibrated years before the present (cal.  (Nogué et al., 2013).

F I G U R E 1 (a) Location of study sites in
Here, we assess whether past regional intervals of drought were reflected in the profile of pollen traits by an increase in desiccation tolerance pollen traits.
We aim to examine patterns of pollen morphological traits related to desiccation tolerance and dispersal in the Canarian laurel forest to examine the spatiotemporal distribution of pollen traits.
Specifically, we address two research questions: 1. How are the pollen traits distributed across study sites? We hypothesise that plant communities growing in drier environments would have a higher preponderance of species that have pollen traits associated with desiccation tolerance. We also hypothesise that pollen dispersal traits would follow a gradient across laurel forest types.
2. How did the pollen trait composition change with changing past environmental conditions? We hypothesise that past regional intervals of drought were reflected in the profile of pollen traits.

| Laurel forest pollen rain collection
The pollen rain was collected from the 13 study sites in Garajonay National Park to supplement the data collected from the six study

| Pollen identification and quantification
A known quantity of spores of the exotic clubmoss Lycopodium spp.
was added to pollen samples as a quantitative marker, before samples were passed through a 150μm sieve to remove plant material, treated for 1 min with an acetolysis mixture involving acetic anhydride and sulphuric acid in a ratio of 9:1, dyed with safranin and dehydrated using tert-butyl alcohol. Pollen samples were mounted on slides in silicone oil (Bennett & Willis, 2001

| Pollen rain composition
For each sampled study site, we analysed 21 pollen taxa of trees and shrubs. We identified the pollen taxa at least at genus level with the exception of Lauraceae that was included at family level (see Appendix S5,S6). In each sample, we calculated for each taxon the pollen percentage of total pollen in the sample. Note that although Pinus canariensis is not present in the laurel forest study sites, Pinus pollen was included in the analyses due to the presence of pollen grains in our modern pollen survey. Pinus canariensis is abundant on Tenerife, but less so on La Gomera, where there is currently no pine zone habitat above the laurel forest (del Arco Aguilar et al., 2010).

| Pollen trait community-weighted means
For each of the 19 study sites, we calculated the community-weighted mean (CWM) for each pollen trait (Lavorel et al., 2008 Appendix S4). We calculated the proportion of pollen traits over time for the taxa included in this study and we compared the trends with regional palaeoclimatic data. The palaeoclimatic data are based on the quantitative reconstruction of western Saharan precipitation derived from leaf wax isotopes from marine core GC37 (Tierney et al., 2017) and marine core GC27 for the reconstruction of the aeolian dust pro-

| Pollen rain composition
The number of pollen grains per pollen trap in La Gomera ranged from 252 (ACE, humid laurel forest) to 781 (SEC, dry laurel forest) (Appendix S1,S4). The best represented pollen taxa were Morella and Erica. The relative abundance of Lauraceae was generally low but it was present in all study sites, ranging from 0.1% (SEC) to 1% (NOR, cold laurel forest), consistent with the known low pollen production dispersibility of these entomophilous species (Appendix S6, Figure S1). Pollen composition and abundance varied between laurel forest sites in Garajonay National Park. PAL (humid laurel forest) site featured a high percentage of Erica pollen (44%) and a low percentage of Morella pollen (17%) (Appendix S1).

| Pollen trait community-weighted means
Pollen taxa with more furrows and pores were more prevalent in dry laurel forest sites than in humid, crest and cold forests (Figure 2).
Both pollen taxa with simple and complex ornamentations were better represented than moderate ones. Overall, pollen taxa of medium and small size were better represented than pollen of larger sizes. The highest prevalence of medium-sized pollen grains was found in the dry laurel forest sites. A larger proportion of small pollen was found in the humid laurel forest and of larger pollen in the cold laurel forest.
Across study sites, pollen taxa with thicker pollen walls were better represented than those with thin pollen walls. However, the ratio of thick to thin pollen walls declined from dry forest sites to ridge crest, humid and cold. Pollen taxa dispersing as monads were the best represented for all forest types, with bisaccate pollen was the least represented. Finally, in all site types, most pollen taxa were spheroidal and there were no differences between laurel forest types ( Figure 2).

| Holocene change of pollen traits focusing on desiccation tolerance
When looking at the distribution of different pollen traits through time, we found that the proportion of pollen grains with apertures increased towards the present. In addition, pollen grains with thicker walls increased towards the present for La Gomera, but in Tenerife remained stable (Figure 3a-c). For the La Gomera pollen record these changes occurred during a time interval for which palaeoclimatic data suggested that the regional climate became drier at around 5500 years ago (Figure 3e, Appendix S3). In addition, we also found that for the La Gomera palaeoecological record, pollen functional trait space was greater from 5000 cal. year BP to the present than in the previous pe- There is no evidence suggesting precipitation-driven changes in the laurel forest at La Laguna during 4700-4200 cal. year BP.

F I G U R E 2
Box plots representing the community-weighted mean (CWM) pollen trait value for each laurel forest type (dry, humid, cold, ridge crest). The four box plot components are midline, median, maximum and minimum values. For the continuous traits (furrows and pores), this is calculated as the mean trait value for all taxa in the community. For categorical traits (dispersal unit, pollen size, ornamentation and wall thickness), this is calculated as the abundance of each individual class.

| DISCUSS ION
Our results are pointing towards a potential filtering process, where plants producing pollen grains with advantageous traits (e.g., highdesiccation tolerance traits in drier environments) were more represented in the community during the drier periods (Pacini et al., 2006;Franchi et al., 2011) (Figure 3). We would like to note that this paper does not aim to track evolutionary adaptations to the island environment nor to analyse whether or not pollen properties are conservative traits of selective value.

| Pollen desiccation tolerance traits: Apertures, wall thickness and pollen size
By analysing a selected set of pollen traits related to desiccation tolerance in an environmental gradient from humid to dry we were able to investigate whether laurel forest communities growing in different environments displayed distinct pollen trait assemblages.
We found that laurel forest communities growing in a dry environment (450-800 m a.s.l.) produced pollen grains with apertures (furrows and pores), which are indicators of high desiccation tolerance ( Figure 2). These pollen traits are thought to confer resistance to dry environments due to harmomegathic efficiency, which is the mechanism that allows variation of the pollen volume with changing moisture conditions (Franchi et al., 2011). Pollen grains with low desiccation tolerance were more prevalent at the laurel forest sites The function of the wall is considered to be mainly protection against adverse environmental conditions such as desiccation and UV radiation (Pacini & Hesse, 2012). We found that in general, dry laurel forest showed pollen grains with thicker walls and cold laurel forest sites displayed pollen grains with thinner walls (Figure 2). It has previously been discussed that a possible advantage of thin pollen walls to plants in humid environments is that it allows rapid germination (Pacini & Hesse, 2012). In addition, it has also been suggested that to minimise the rate of water loss due to desiccation, plants produce larger pollen grains (Muller, 1979;Stroo, 2000;Ejsmond et al., 2011). In our study sites, while larger pollen grains are more prevalent on the most humid sites of the laurel forest, our results are not conclusive as medium and small-sized pollen grains were the most abundant in all forest types. Therefore, we need to treat these preliminary data with caution since our results were not conclusive.

| Pollen dispersal traits: Ornamentation, shape and dispersal unit
Another interesting result of our pollen trait analysis, specifically the presence or absence of pollen ornamentation, is related to laurel forest dispersal strategies. It has been described that pollen wall sculpturing may play a role in the attachment to insect pollinators and also to the stigma of the flower (Eriksson & Bremer, 1992;Vaknin et al., 2000;Hesse 2000;Pacini & Franchi, 2020). Over 60% of the plant taxa included in this study are classified as insect-pollinated (Olesen et al., 2007) (Appendix S6). This level of entomophily is similar to that in tropical forests where around 80% of taxa are animal-pollinated (Bush & Rivera, 1998;Weng et al., 2004;Fernández-Palacios et al., 2017). With this level of insect pollination, we would expect a dominance of pollen grains displaying complex wall ornamentation. However, the CWM analysis is not conclusive as both pollen grains with simple and complex ornamentations are equally represented across the study sites. However, we found some generalisations when focusing on the different laurel forest types. Pollen grains with simple ornamentations appear to be more prevalent in the dry laurel forest. In comparison, the CWM analysis showed that the humid laurel forest displayed the highest prevalence of pollen grains with complex and moderate ornamentations, most often associated with insect pollination (Hu et al., 2008). Our results also suggest that ornamented pollen grains are more prevalent in the core of the laurel forest that overlaps with the humid and cold laurel forest plots (Figures 1 and 2).
Finally, the analysis of dispersal units shows that the majority of plant taxa disperse as monads followed by tetrads. Similarly, the CWM analysis for pollen shape showed that the majority of pollen grains are spheroidal and do not follow any site-specific trend.
The lack of conclusive results suggests that the systems we have analysed lack a sufficient range in characteristics to reveal such relationships.

| Temporal variation of pollen traits related to desiccation tolerance
Data on past environmental change from available palaeoecological records (pollen and charcoal) and palaeoclimatic data (leaf wax isotopes and aeolian dust) suggest that the laurel forests of La Gomera and Tenerife have been subject to past climate-induced and anthropogenic shifts (de Nascimento et al., 2009;Nogué et al., 2013;McGee et al., 2013aMcGee et al., , 2013bTierney et al., 2017;Nogué et al., 2021; (Appendix S3). The next question is whether the regional shift towards dryer conditions (the end of the African Humid Period) was reflected in the fossil pollen trait composition. remain viable in seed banks for a long time as they cannot germinate under the shade of closed forest canopies (Fernández-Palacios & Arévalo, 1998). For Tenerife, the palaeoecological record is shorter (past 4700 years) and not covering the shift towards drier conditions. However, the main trend on the pollen trait profiles aligns with the previous results of more presence of plant taxa producing pollen grains with apertures and thick pollen walls.
Second, one of the most interesting conclusions is that current laurel forest taxa produce a high proportion of desiccation-tolerant pollen traits, a pollen-trait community that established around 5000 years ago and is potentially linked to a regional shift towards drier conditions. While high-desiccation tolerance pollen grains have a longer survival rate under low relative humidity enabling them to disperse further and to cross between different partners, low-desiccation pollen grains are known to survive for fewer hours after anther opening and to germinate immediately if moisture is available. If moisture is not available the survival rate is low (Franchi et al., 2011). However, results also showed that plants producing pollen grains with low desiccation tolerance prevailed during the periods of drought. This may indicate that the environmental change would not have been strong enough to filter out such traits.
Finally, the remaining question is whether pollen trait functional space changed over time or stayed constant. On the island of La Gomera fossil pollen functional trait space was small prior to 5000 cal. year BP and increased in functional dispersion from 7500 to 5000 cal. year BP (Figure 3d). Ecological theories such as niche packing (Pellissier et al., 2018) and environmental filtering (Kerkhoff et al., 2014) have suggested that the range of functional trait space in stressful environments is reduced. On the other hand, an increase in functional trait space, as shown in our analysis, might be related to a shift in the balance of species in the forest. Accordingly, our results indicate that laurel forest communities have currently a wider range of pollen traits (e.g., both high and low tolerance for desiccation) than in the past, and thus a greater diversity of trait responses (Winfree & Kremen, 2008;Pellissier et al., 2010;Matteodo et al., 2013;Kerkhoff et al., 2014;Pellissier et al., 2018). Future research should assess the relationships between pollen functional traits and other plant traits in laurel forest taxa, aiming to understand the entirety of the plant functional space and its role in forest adaptation to shifting climatic regimes.

| Limitations
The analysis of pollen traits holds great promise for improving our understanding of community and ecosystem responses to environmental change (Carvalho et al., 2019;van der Sande et al., 2019). However, there are several limitations. First, the larger amount of pollen grains released by anemophilous (i.e., high grain production) plant types when compared to entomophilous (i.e. low production) plant types (Ackerman, 2000;de Nascimento et al., 2015) may influence our findings. Lauraceae trees, which are animal-pollinated, are well known to produce a limited pollen signal (Connor et al., 2012;de Nascimento et al., 2015). For example, the 'Los Noruegos' (NOR) plot, located in the humid laurel forest type, featured the highest percentage of Lauraceae pollen grains, amounting to just 1% of the total pollen taxa analysed (Appendix S1). By comparison, in 'La Meseta' (SEC) site located in dry laurel forest, the abundance of Lauraceae pollen grains was just 0.1%. Second, the presence of Pinus pollen grains, despite an absence of this taxon at the study sites, requires an explanation. Pinus pollen is known to disperse over large spatial scales and displays several morphological traits related to wind dispersal, such as the presence of air-filled sacs (Williams, 2010) (Figure 1). Pinus canariensis, the sole native pine species in the Canaries, is abundant, especially in the upper regions of Tenerife, growing in a broad belt above the laurel forest. We therefore explain the presence of Pinus pollen grains as being dispersed from nearby trees. Third, it is likely that the robustness of the methodological framework presented here will improve with an increase in availability of pollen trait data from a wider range of plant taxa and field studies, for example, the standardization and integration of pollen traits and their interactions such as the presence or absence of pollen kit, chemical elements and phylogenetic studies.

| CON CLUS I ON S AND NE X T S TEPS
Our results indicate that pollen traits contain relevant information on plant responses to environmental change, plant survival in different environments and plant fitness over time. We suggest the following next steps: 1. To integrate pollen traits with other plant traits (e.g., leaves, roots and seeds): we suggest to integrate further pollen trait data sets from different ecosystems worldwide with plant trait data. This integration may be needed before drawing firm global conclusions concerning the drivers of the spatiotemporal distribution of pollen traits, and to enable analysis of key ongoing questions in vegetation dispersal, persistence and resilience (Franchi et al., 2011;Carvalho et al., 2019).
2. To improve the quantity and quality of the pollen trait data: new developments in computer vision and pattern recognition applied to palynology have huge potential for the analysis of pollen traits.
Among the advantages of including artificial intelligence-based pollen trait classification, the most important is the improvement in accuracy and robustness of the data set. This type of analysis could be carried out on microscopic images of fossil pollen slides allowing direct measurements of pollen traits for each time-step (Viertel & König, 2022).