Cedrus atlantica pollen morphology and investigation of grain size variability using laser diffraction granulometry

Detailed information on pollen morphology and grain size is critical for palynologists to accurately identify vegetation from fossil pollen assemblages for pollen analysis. Grain size may be studied to improve the taxonomic resolution of pollen identification, for example with Pinus pollen (Desprat et al. 2015), Poaceae pollen (Radaeski et al. 2016), and indeed with Cedrus pollen (Fujiki et al. 2003). Pollen size has also been correlated with genome size and may indicate polyploidy (three or more chromosome sets) in plants (Gould 1957; Kapadia & Gould 1964; Bennett 1972; Tate et al. 2005; Knight et al. 2010; De Storme et al. 2013). Increasingly, geochemical studies utilising pollen, such as stable isotope analysis for palaeoclimate reconstructions (Amundson et al. 1997; Loader & Hemming 2004; Nelson et al. 2006, 2007; King et al. 2012; Nelson 2012; Bell et al. 2017) and biomarker analysis for UV-B reconstructions (Rozema et al. 2001, 2002; Fraser et al. 2011; Willis et al. 2011; Lomax et al. 2012; Jardine et al. 2017), require detailed knowledge of grain size for developing techniques to isolate specific grains from fossil assemblages. For example, the use of micro-sieving to concentrate pollen from sediment (Heusser & Stock 1984; Brown et al. 1989) can be modified to target grains within a specified size range to facilitate the isolation of pollen for specific species.

Traditional pollen analysis relies on visual identification of pollen grains to determine vegetation composition, while geochemical studies undertake further analysis of the grains to gain insight into environmental or climate conditions. It has also been hypothesised that the size and shape of the pollen grain itself may be influenced by climate (Ejsmond et al. 2011). Temperature has been previously linked to pollen grain size (Kurtz & Liverman 1958), and Schoch-Bodmer (1936) proposed that grain size variability in pollen is a result of fluctuations in moisture availability in ambient air during pollen development. Ejsmond et al. (2011) analysed eight Rosaceae species and found grain size increased under desiccation stress (determined by temperature, potential evapotranspiration and altitude). A positive relationship was also found between temperature and pollen size in 232 plant species from 11 taxonomic groups, suggesting possible climatic influence on size during the flowering period (Ejsmond et al. 2015) and offering potential for pollen grain size to aid in reconstructions of past environments.

Measurements of grain width on modern Nothofagus spp. pollen showed a significant relationship with mean annual precipitation (MAP), where grain size increased with reduced MAP (r² = 0.66, p < 0.0001). Applying this relationship to fossil pollen samples from Antarctica, where average grain size increased by 23% from the late Eocene to the mid Miocene, suggests a decrease in precipitation during this period (Griener & Warny 2015). This is supported by an earlier study on the same fossil samples which found stable carbon isotope discrimination (Δ¹³C) values of pollen decreased during the same period, which the authors also linked to a decrease in moisture availability (Griener et al. 2013). However, the validity of the relationship between grain size and moisture availability in this study has been questioned, citing a lack of theoretical and empirical support, with a view that it may be premature to use grain size as a moisture availability proxy (Jardine & Lomax 2017).

Atlas cedar (Cedrus atlantica (Endl.) Manetti ex Carrière) is a moisture-sensitive (Rhanem 2011; Linares et al. 2013; Ilmen et al. 2014) montane conifer endemic to semi-arid and humid areas of Morocco and Algeria (Farjon 1990), with pollen records indicating a presence in the area since at least the Last Glacial Maximum (Lamb et al. 1989; Magri et al. 2017; Zielhofer et al. 2017). The earliest work on Cedrus atlantica pollen morphology found the average size of the grain to be 61.5 ± 1.8 µm based on 100 grain measurements of pollen from a single tree in Ifrane, Morocco (Aytug 1961), while a later study recorded a size of 45.8 ± 2.4 µm measured on pollen collected from one tree in Marseille, France (Fujiki et al. 2003). In both studies, pollen samples had been collected and stored 15 years prior to measurements being taken. Cedrus atlantica pollen grains have also been measured at 75 µm on samples from Turkey (Altuner et al. 2012), and 58 µm in Vancouver, Canada (Ho 1972). In Algeria, samples from the Tell Atlas ranged in size from 60 to 63 µm, while samples from the eastern margins of the Saharan Atlas and Aurès Mountains measured 54 to 60 µm (Derridj et al. 1991).

The literature suggests that pollen from Cedrus atlantica varies in size significantly, which could possibly result from climate differences between sample locations. However, variations noted in these studies may also be due to different methodological approaches undertaken for measurement which could have affected grain size – e.g. changes resulting from chemical pre-treatments, pollen hydration, mounting media, cover-slip pressure and storage methods (Andersen 1960; Aytug 1960; Faegri & Deuse 1960; Cushing 1961; Reitsma 1969; Praglowski 1970) – so it is not possible to determine from current literature whether there is a climatic influence.

Pollen samples were collected from Cedrus atlantica trees from nine locations (Table 1) across the Middle Atlas, Morocco (n = 72), with additional samples collected from sites in Spain, France, UK and USA (n = 19). Samples were collected between September and October 2015, apart from the UK samples which were collected in 2014. Multiple strobili from each tree were collected and placed in paper envelopes on site, and freeze-dried back in the laboratory. Grains were extracted by vigorous shaking of the strobili, and collecting in sieves. Non-pollen contaminants were removed by visual inspection, and grains were stored in glass vials at 4 °C.

Measurements were carried out on untreated samples using both high-power transmitted LM and laser diffraction granulometry, with additional measurements under LM on samples treated with 10% potassium hydroxide (KOH) in a water bath at 90 ˚C for 15 minutes.

Scanning electron microscopy (SEM) images were obtained using a 120-kV FEI Tecnai 12 Twin Transmission Electron Microscope in The University of Manchester Life Sciences Faculty. Untreated and treated pollen samples were dehydrated in graded ethanol stages, then dried with Hexamethyldisilazane (HMDS) following (Chissoe et al. 1994). Grains were stuck to spurs with doubled-sided tape and sputter-coated with gold palladium for 5 minutes prior to SEM.

Untreated pollen samples were measured using a Malvern Mastersizer 2000, fitted with a Hydro 2000µP liquid sample dispersion unit for small particles (e.g. Sperazza et al. 2004) in The University of Manchester Geography laboratories. The machine was calibrated prior to measurements on pollen, using a spherical glass bead standard supplied by the manufacturer. The system was configured using the following method: Sample material was characterised as sporopollenin with a refractive index of 1.475 (Traverse 2007). Dispersant used was water with a refractive index set to 1.33 (Hecht 2002). Sample measurement time and background measurement time were set to 30 seconds. Pump speed was set to 2000 rpm, and ultrasonic treatment set to 75% with a pre-measurement period and delay of 30 seconds. Measurement repeats were set to three per aliquot with a 10-second delay. Using this method, the dispersion unit (Hydro 2000µP) was filled with deionised water using the anaerobic fill option (to remove air bubbles, which would otherwise affect the result). Once full, the pollen sample was added to the dispersion unit to reach a laser obscuration which ideally fell between 10 and 20% (this is shown on screen as the sample is added). Once the desired laser obscuration was reached, the measurement cycle was run. On completion of measurements, the system was drained, and flushed with deionised water 3 times to remove the sample. The sample can be recovered at this stage if required.

All laser particle size distribution models (Fraunhofer and Mie theory) assume the analyte particles are spherical (Syvitski 1991), and the machine will record a measurement of a particle in whichever orientation it passes through the laser. As Cedrus atlantica pollen grains are not completely spherical, and they may pass the laser in different orientations, the model will in practice provide a weighted average of the diameter of the pollen grain. Since the pollen morphology of Cedrus atlantica is consistent, an empirical relationship can be derived between the particle size measurement and LM measurements. For Cedrus atlantica pollen, the D₅₀ value is effectively equivalent to the size of the equatorial axis (EX) as measured under LM (see Results section 3.2.2.).

Data and statistical analyses were carried out using R (R Core Team 2016). Climate data was extracted from CRU TS v3.24.01 high-resolution (0.5°) gridded datasets (Harris et al. 2014). Precipitation for Middle Atlas locations was interpolated using local climate stations described in Bell et al. (2017). Aridity data was extracted from the self-calibrating Palmer Drought Severity Index (scPDSI) (Dai 2011).

Cedrus atlantica pollen are large diploxylonoid-type bisaccate grains. The grains are elongated, the corpus is typically prolate, and the cap wall is thick. Sacci are spheroidal in equatorial view, appearing more oblate in polar view. The junction of the sacci and corpus lacks a strong defining ridge, with the two parts appearing seamlessly together. Under LM the surface ornamentation of the corpus is a rough, reticulate pattern, while ornamentation is somewhat smoother on the sacci. Under SEM, surface ornamentation of the corpus is rough, appearing fossulate, with groups of irregular, spheroidal to elongated elements protruding, interspaced by deep groves. The sacci appear smooth with a scabrate–perforate surface. Surface ornamentation appears more defined following chemical treatment under both LM and SEM.

LM measurements of grain size on eight untreated pollen samples, with 240 individual grains measured (Table 2), found the average total grain size including sacci and corpus (TL) was 71.4 ± 7.6 µm, while the EX measured an average 52.2 ± 4.0 µm. There was large variation in the size of grains observed: For TL, there was a 40-µm difference between the smallest and largest grains, and for EX, there was a 20-µm difference. Full LM measurement data can be found in the online Supplementary material.

Treated pollen samples measured under LM were on average 6.9% larger in size overall compared to untreated grains (Table 3). However, the increase in size was not consistent across all the measured properties of the grain, with the largest increases to the sacci width and cap thickness (CT), while the total grain width (TW) and EX decreased in size. This resulted in the shape of the grain changing slightly, becoming subprolate.

Pollen grain size measurements taken using laser diffraction granulometry on 91 untreated pollen samples, where millions of individual grains were measured, recorded an average grain size of 59.1 ± 4.0 µm. Particle size distribution data for each sample can be found in the online Supplementary material.

The reliability of the laser diffraction particle size measurements was tested by correlation analysis with the LM measurements (Figure 2) for the samples measured using both methods (Table 4). The strongest and most significant correlation was found between the D₅₀ median size value (laser diffraction) and the LM measurement for the EX (r = 0.91, p = 0.002). During the measurement cycle for laser diffraction granulometry, pollen grains float freely in water as they pass the laser beam and are not subject to external pressure (e.g. from a cover-slip); consequently, the ‘natural shape’ of the grain is measured. In Cedrus atlantica pollen, the sacci lie underneath the corpus, protruding a short distance; effectively, the total length of the grain is slightly bigger than the equatorial axis of the corpus. Under LM, the protrusion of the sacci can appear to be greatly inflated (particularly in equatorial view) resulting in the total length of the grain appearing larger. Consequently, the best correlation with the grain size reported by laser diffraction granulometry for Cedrus atlantica pollen is with the equatorial axis as measured under a light microscope.

The eight samples measured under LM show large variability in grain size (in all measured properties) between individual grains within the same sample (Table 2). Between samples, analysis of variability (ANOVA) of the equatorial axis (measured under LM) suggests this is significant (df = 7, F = 14.97, p < 0.0001); however, Tukey's test reveals that six of these samples do not vary significantly in size from each other.

Regression analysis was performed on pollen grain size and possible influences including climate (temperature, precipitation, aridity and potential evapotranspiration), altitude, and carbon isotope discrimination values, which are an indicator of environmental moisture availability (Bell et al. 2017). Given that samples from within the same geographical area will experience the same climate conditions, and fossil pollen assemblages will comprise grains from several trees within the surrounding area (Bell & Fletcher 2016), analysis was performed testing the average grain size for each geographical sampling area, as well as individual sample grain size values (Table 5).

The regression analysis found no significant relationships with any of the climate variables or altitude to suggest a link to grain size. Similarly, multiple-regression analysis testing every possible combination of climate predictors using all subsets regression found no significant relationships or models to support a climate influence on grain size. Furthermore, there was no significant relationship with stable carbon isotope discrimination when samples were averaged by geographical location (r² = 0.10, p = 0.234) although there was a very weak significant association with individual samples (r² = 0.05, p = 0.029). However, 10-fold cross-validation of this model found the r² value could be as low as 0.003.

In another study, summer moisture availability was also ruled out as influencing pollen grain size on Cedrus atlantica populations in Algeria. Analysis showed that pollen grain size discriminated trees into different genetic and ecological groups, suggesting genetics influenced pollen size (Derridj et al. 1991). The genetic control on pollen size may stem from genome size (Bennett 1972; De Storme et al. 2013), in particular the effect of an increased number of chromosomes due to polyploidy (Kapadia & Gould 1964; Dyer et al. 2013). Knight et al. (2010) found a significant positive trend between pollen width and genome size across 464 species. However, the trend was not significant when phylogeny was taken into account, suggesting that pollen size would not be a good proxy for genome size in the fossil record. The phylogenetic relationships among Cedrus have been debated (Bou Dagher-Kharrat et al. 2001), but DNA evidence suggests Cedrus atlantica separated from a common ancestor of Cedrus libani and Cedrus brevifolia around 23 to 18 Ma BP (Qiao et al. 2007), and DNA analysis of all four Cedrus (the ‘true’ cedars) species (including Cedrus deodara) found that genome size is homogeneous among the species (Bou Dagher-Kharrat et al. 2001). Genetic diversity of Cedrus atlantica trees from Morocco was shown to be high within populations and between populations (Renau-Morata 2005; Terrab et al. 2006), particularly between populations from the Rif, and High Atlas, when compared to Middle Atlas populations (Cheddadi et al. 2009). It is possible that the variations in grain size we observe between samples could relate to genome size; however, further research is needed to confirm this in Cedrus atlantica pollen.

The influence of temperature on grain size as demonstrated by Ejsmond et al. (2015) was suggested to relate to pollen performance, whereby a trade-off exists between the size of the grain and the quantity of grains produced (Vonhof & Harder 1995). The competitive ability of pollen increases with temperature during the flowering period, and this increased competition promotes larger pollen grains (Ejsmond et al. 2015). The lack of any temperature influence on Cedrus atlantica pollen size may be due to it being a wind-pollinating species. The apparent trade-off between pollen size and quantity may not be necessary, as priority is on the quantity of grains produced in order to increase the chances of successful pollination (Whitehead 1983; Cruden 2000). Larger grains are heavier, giving increased chance of reaching ovules as they can more easily break from the airstream, while smaller, lighter grains can travel greater distances (Niklas 1985). Pollen size in wind-pollinated species consequently reflects an equilibrium between these demands (Lu et al. 2011). Large grain size variation within individual samples could therefore possibly reflect an adaptation to facilitate both demands. The observed grain size variation also implies that smaller grains found within fossil assemblages from geological archives might be an indicator of long-distance pollen transport, with larger grains indicating nearby pollen sources rather than reflecting a climate signal.

Smith (1923) noted ‘irregularity’ with the development of Cedrus atlantica pollen, where newly forming grains exist alongside mature grains, and long retention of grains after they reached maturity. Strobili typically form 2–3 months prior to pollen release, with grains developing in the later parts of this period. Pollen release is a phenological response lasting a few days, with the timing varying between individual trees and between geographical locations, depending on optimal environmental conditions including temperature, humidity levels and wind speed (Whitehead 1983; Khanduri & Sharma 2009). For Cedrus atlantica this typically occurs from early to late September and early October. This irregular development of grains may also contribute to the size variation observed, if pollen release occurs while some grains are still developing and others are fully developed.

Soil nutrient availability has also been linked to pollen grain size and production, in squash (Cucurbita pepo) plants. Grain size and the number of grains produced both increased where plants were in soil which had higher nutrient contents; this effect was greatest with increased nitrogen (Lau & Stephenson 1993), but also evident with increased phosphorus (Lau & Stephenson 1994). If soil nutrient availability influenced grain size for Cedrus atlantica pollen, then we might expect to see larger grains in the botanical garden sites and smaller grains in the relatively nutrient-poor Middle Atlas locations. However, grain size is smaller than average at two of the three botanical garden sites (Westonbirt, Paris and Boston), while grain size is larger than the average at the most southerly Middle Atlas site (Col Du Zad), an environment characterised by sparse open forest and semi-arid conditions where nutrient availability is poor. This suggests that nutrient availability does not influence grain size for Cedrus atlantica pollen.

Overall, the large variability in pollen grain size we observe within individual samples, in some cases by as much as 20 µm, suggests that grain size is influenced by a number of complex factors. The grain size variability is also not unexpected, and is well documented in other pollen (e.g. Bell 1959; Clausen 1962; Bragg 1969; Cruzan 1990; Desprat et al. 2015). Consequently, we propose that due to the size variability in Cedrus atlantica pollen, and lack of evidence for climatic influence, it would not be possible to use this as a proxy for climate or environmental reconstruction, as differences in the size of fossil pollen may simply result from the observed natural variation in size between pollen grains. Our findings contrast the suggested climatic influence on grain size observed in other species (Ejsmond et al. 2011,2015; Griener & Warny 2015), and underscore the need for further investigation of the complex controls on pollen grain size (Jardine & Lomax 2017).

We demonstrate the importance of a consistent methodological approach to grain size measurement, through the differences in grain size reported in the literature, as a result of the technique used, and between untreated and chemically treated pollen grains. It is notable that the size changes we observe in Cedrus atlantica pollen grains treated with KOH are not uniform across the entire grain, suggesting that morphological differences in pollen have different susceptibilities to chemical treatment. For example, Reitsma (1969) found the size-altering effects of KOH treatment depended on pollen type and exposure time to the treatment, while effects on pollen size from chemical pre-treatment have also been noted in Faegri & Deuse (1960), Praglowski (1970) and Charman (1992). In Cedrus atlantica the largest size increase in the pollen grain occurred in the sacci, which have a perforate surface in contrast to the rest of the grain. The small apertures allow the pollen to quickly dehydrate following pollen release (Tekleva et al. 2007), reducing its weight to assist long-distance transport (Lu et al. 2011). In a reversal of this process, the apertures may allow greater penetration of chemical treatment into the sacci compared to other parts of the grain, and may explain why they exhibited a greater increase in size. This suggests that morphological variations within pollen grains and between pollen types are affected by treatments in different ways, so relationships of grain size should always be established to specific pollen types following the same preparation and methodological protocols.

Fossil pollen grain size may also be affected by diagenesis. Although potential size-altering effects are not fully known, they are likely to differ depending on species and sedimentary setting (Mäkelä 1996), with resistance to these effects likely stemming from morphological traits of the pollen grains. Due to these effects, it may be difficult to compare grain size between fossil pollen from different sedimentary settings, implying that apparent grain size changes in a fossil sequence could result from changes to the sedimentary setting, unless it remains homogeneous throughout. Consequently, pollen grain size relationships and comparisons between pollen from a fossil setting and modern pollen samples may not be straightforward (Mäkelä 1996). Indeed, before comparison, the effects of diagenesis and sedimentary setting on fossil pollen size should first be established. Overall, all possible influences on grain size should be considered in the interpretation of results (Desprat et al. 2015), and a consistent methodological approach to measurements should be taken.

We have shown that laser diffraction granulometry provides a reliable and consistent method of determining pollen grain size, and produces results in line with existing LM methods of grain size measurement. The main benefit of laser diffraction granulometry is that it can measure thousands to millions of pollen grains (dependent on sample size) in just a few minutes, providing a very robust assessment of the typical grain size (here, D₅₀ or median) which can be repeated on multiple samples quickly. The method is also non-destructive, allowing recovery of the sample after the measurement cycle. Measurement of grains is not affected by operator subjectivity or potential inaccuracies between measurements, nor is it affected by size changes caused by mounting medium or cover-slip pressure on the grain, which can affect LM measurements. Provided the morphology of the pollen grains is consistent, an empirical relationship between laser diffraction granulometry measurements and light microscope measurements can be established. In additional testing of the method, we found similar results with Pinus pollen between the grain size reported by laser diffraction granulometry and LM measurements; however, further testing of the method is needed on other pollen types with different morphologies, and pollen types with smaller grains.

The primary use of laser diffraction granulometry for pollen size determination would be on modern samples, due to the relatively large quantity of sample material required. There is a range of possible applications, including investigating the influence of climate and environmental factors, nutrient availability, and genome size. In testing, approximately 15 mg of pollen was used to reach 10% laser obscuration, which could easily equate to millions of grains (dependent on the grain size/weight of the specific pollen type), based on estimates of pollen weight. For example, Brown & Irving (1973) weighed several pollen types, including Quercus robur, suggesting 93 grains weigh 1 µg, or that there are 1,395,000 grains per 15 mg. There are reportedly 404 maize grains per 1 µg, or 6,072,874 grains per 15 mg (Miller 1982). Bunderson & Levetin (2015) weighed different Juniperus species where approximately 273 grains weigh 1 µg, or there are 4,109,589 grains per 15 mg (averaged across species at 60% relative humidity).

We found that it is possible to use a reduced sample size, where laser obscuration only reached 0.4%, which produced grain size results in line with those at 10% laser obscuration. This equates to approximately 0.6 mg of pollen material required, which could equate to several thousand grains. Though perhaps this is still out of reach for measurements on fossil samples, it does allow measurements to be taken on modern pollen for a range of species, from small insect-pollinated flowers to large wind-pollinated trees.

Our study finds that while there is large variability in the pollen size of Cedrus atlantica, it is not significant between samples. We found no significant relationships between climate and grain size, testing temperature, precipitation, aridity and potential evapotranspiration (PET). Grain size of Cedrus atlantica may be influenced by a number of complex factors, and variability within individual samples may result from the irregular development of pollen. Our study confirms that a consistent methodological approach to grain size measurement must be taken, as there are many aspects of pollen handling and preparation stages that may influence the observed size. Size comparisons between values reported in literature, between different species/pollen types, and from different sedimentary settings may not always be possible or advisable. We have also shown a method utilising laser diffraction granulometry which can be used on modern pollen samples to accurately determine pollen size, and which produces results consistent with LM measurements. This non-destructive method of determining pollen size provides reproducible, robust results, from potentially millions of pollen grains, quickly and easily.

Fieldwork was carried out with assistance from the Haut Commissariat aux Eaux et Forêts et à la Lutte Contre la Désertification, Morocco, with thanks to Said Hajib. The authors thank Jenifer Campbell, Rachel Hurley, John Moore and Jonathan Yarwood (University of Manchester) for field and laboratory assistance. We thank Tobias Starborg (University of Manchester) for assistance with SEM imagery. Additional pollen samples were kindly supplied by Penny Jones (Westonbirt Arboretum), Martin Gardner (Royal Botanical Gardens, Edinburgh), Kathryn Richardson (Arnold Arboretum, Harvard University), Stéphanie Desprat (EPHE, Bordeaux), Daniel Gómez (Instituto Pirenaico de Ecología, Spain) and the Paris Botanical Gardens. Lastly, we thank Phil Jardine and an anonymous reviewer for their constructive comments and useful discussion to improve the final manuscript.

BENJAMIN A. BELL is a PhD student at The University of Manchester whose work focuses on the application of geochemical techniques to palynology and quaternary science, with a particular interest in Cedrus atlantica, its role within the environment and how it adapts to climate change in Northwest Africa.

THOMAS H. BISHOP is a research and teaching technician within the Geography Laboratories at The University of Manchester. He specialises in analytical techniques for palaeoenvironmental research, including chemical, physical and biological methods. His PhD research project was a palaeolimnological investigation of central Patagonia focussing on the Holocene epoch, which he completed at the University of Southampton. His current research interests are in developing novel techniques for collecting and manipulating environmental data.

WILLIAM J. FLETCHER is a palynologist whose specialist area of interest is the detection and characterisation of abrupt environmental and climatic changes in the Mediterranean region using vegetation records from Quaternary terrestrial and marine sediment archives. He obtained his PhD at the University of Cambridge and is a senior lecturer in physical geography and quaternary science at The University of Manchester.

PETER RYAN is a palynologist at The University of Manchester and is currently a teaching fellow in geography. His work focuses on the environmental impact of prehistoric people through the use of palaeoecology.

RACHID ILMEN is a professor in the Department of Hydraulic, Environment and Climate (HEC) at Hassania School of Public Works (EHTP)-Casablanca, Morocco. He received an engineer diploma from the National School of Forest Engineers, Morocco (2004), an International Certificate on Environmental Education from Shiga University, Japan (2010), a master degree on biological sciences (2009) and a PhD in climatology and climate change (2014) from the Mohammed V University, Morocco. He specialises on water, environment and sustainable development and is an author and co-author of more than 30 papers.

No potential conflict of interest was reported by the authors.

Explanatory notes for the supplementary data

Cedrus atlantica pollen size measurement data under light microscopy.

Cedrus atlantica particle size distribution data from laser diffraction granulometry.