You are on page 1of 6

The Holocene 13,4 (2003) pp.


A simple and effective methodology for sampling modern pollen rain in tropical environments
William D. Gosling,1 * Francis E. Mayle,1 Timothy J. Killeen,2 ,3 Marcelo Siles,3 Lupita Sanchez3 and Steve Boreham4
(1Department of Geography, University of Leicester, University Road, Leicester LE1 7RH, UK; 2Centre for Applied Biodiversity Science, Conservation International, 2501 M Street, NW, Suite 200, Washington, DC 20037, USA; 3 Museo de Historia Natural ‘Noel Kempff Mercado’, Avenida Irala 565, Casilla 2489, Santa Cruz de la Sierra, Santa Cruz, Bolivia; 4Department of Geography, University of Cambridge, Downing Place, Cambridge CB2 3EN, UK)
Received 6 February 2002; revised manuscript accepted 15 July 2002

Abstract: To gain a better insight into the nature of palaeovegetation change in tropical ecosystems, more information needs to be gleaned from the limited number of fossil pollen records that exist. To achieve this, a detailed understanding of modern tropical ecosystems and the pollen they produce is required. To facilitate this, a practicable and effective mechanism for sampling modern pollen rain from the tropics is required. This paper presents a modi ed eld methodology based upon three years of trapping experience in Noel Kempff Mercado National Park, Bolivia, and improved laboratory preparation methodologies. We demonstrate here a simple and very effective way to sample modern pollen rain in tropical environments using a funnel trap mounted on a stake containing cotton bre as the trapping medium. Key words: Pollen analysis, pollen rain, pollen trap, tropics, modern analogue, Bolivia.

Extracting detailed palaeoecological data from fossil pollen records in the tropics provides insights into palaeoclimatic change, biodiversity, evolution and biogeography, all of which can contribute to a better understanding of the likely impact of current climatic change on tropical ecosystems. However, the interpretation of fossil pollen records from the tropics has been shown to be dif cult. This is, in part at least, due to a lack of understanding of the relationship between different tropical ecosystems and their modern pollen spectra. For example, it is unclear whether the ‘forest’ pollen signal present at the last glacial maximum in the records from the Amazon fan (Hoorn, 2001) and Lake Pata (Colinvaux et al., 1996) is that of a semi-deciduous dry forest or humid evergreen forest. This distinction is critical as these ecosystems occur under markedly different climatic conditions (Pennington et al., 2000). In fossil pollen investigations of the tropics to date, ecosystems have principally been separated into
*Author for correspondence (e-mail:

‘forest’ and ‘savanna’ categories, based principally upon the dominance of Poaceae (grass) pollen in savannas and a limited number of other indicators. However a high grass-pollen signal may not necessarily be indicative of savanna, but could instead be attributable to aquatic or shoreline grasses (Bush, 2002). It is therefore fundamentally important that a better understanding of modern tropical pollen spectra is achieved. To facilitate this, a reliable and effective method of sampling modern pollen rain in the tropics is required. There are two key problems encountered when relating fossil pollen assemblages from the tropics to modern ecosystems. Many tropical genera contain species from different ecosystems, but the fact that tropical pollen is rarely identi able to species level means that most ecosystems cannot be characterized by an indicatorspecies approach. Therefore, to identify an ecosystem, a pollenassemblage signature must be known; i.e., the relative proportions and/or pollen-accumulation rates (PARs) of a suite of pollen taxa. Multivariate statistics (e.g., Maddy and Brew, 1995) are generally required to distinguish these different assemblages. There is an increasing number of studies from tropical areas 10.1191/0959683603hl649rr

Ó Arnold 2003

Weng et al.. 2000. (2002). This reduced size means that traps are more dif cult to relocate in the eld (Figure 2). or to a stake driven into the ground. 1998. Elenga et al. (a) Cross-sectional view of the head of the pollen trap. 2000).. 2001) and surface lake sediments. Bush and Rivera. such as savannas and deciduous dry forests. variations in lake size and basin morphometry can be major complicating factors. Therefore.. or using a wire (Figure 2) mounted to any convenient location within the ecosystem. 2000. with uorescent tape (Figure 2b). Caution should be taken if mounting a trap to a tree. held in place by a coarse mesh.. redrawn from Bush (1992). it is essential to mark the trap location well. Behling et al. Soil samples have been more widely used than other types of natural traps in the tropics because of the ease of sampling and their ubiquitous occurrence. Bush and Rivera. 1998. Barboni and Bonne lle. seasonally inundated savanna and semi-deciduous dry forest. 2001). the traps can be positioned at any place or height within the ecosystem to optimize the spatial resolution of sampling and allowing patterns of pollen rain to be studied in three dimensions.. There are also two fossil pollen records from the park. allowing direct quantitative pollen-vegetation relationships to be investigated. they allow pollination biology to be investigated in terms of PARs and interannual variability. because there is a likelihood of overrepresentation of pollen of the ‘host’ tree due to its close proximity. The modi ed Old eld trap of Bush (1992) has been favoured over other types of constructed trap. Behling et al. personal communication). Third. we also found that the pollen within these dry samples showed no evidence of oxidation reducing the need to nd an effective moisture transfer mechanism. anticipated to improve retention of pollen in the bres and prevent oxidation (M. the bottle half of the trap can be considered super uous and consequently omitted. These provide three key bene ts over natural traps. Clearly the evaporation of water alone is insuf cient to keep the bres moist in these environments. Bush et al. Ten pollen traps were set up within each of 17 permanent vegetation study plots (20 3 500 m).. The local environment should be taken into consideration when positioning the traps. (2003) and Grabandt (1980) have shown that pollen from moss polsters provide a very effective means of distinguishing plant communities along the eastern ank of the tropical Andes above 500 m elevation.g.g... Bush (1992) states that a plastic bottle should be present under the funnel to catch and store rainfall. Many modern pollen studies in the tropics have used constructed traps (e. arti cial traps provide the potential for consistency in sampling strategy and methodology among different ecosystems. Bush et al. 2001). Second. The funnel trap described by Bush (1992) has been shown to be an effective and practicable mechanism for sampling modern pollen in tropical environments (Behling et al. The traps can either be positioned on the ground as per Bush (1992). However.g. Therefore. Bolivia: terra rme humid evergreen forest.614 The Holocene 13 (2003) using a variety of methods to sample the modern pollen rain. typically the modi ed Old eld trap described by Bush (1992) (e. First. 2001). However. Hicks and Hyvarinen ¨ (1986) and Hicks et al. such as soil (e. Based upon three years’ eldwork in a broad range of tropical ecosystems. The study plots provide excellent spatial and compositional information about the ecosystems in NKMNP (Pan l. we describe improvements to the trap design of Bush (1992) and present an improved methodology for extracting pollen from the trapping medium. Cundill (1986). Lakes are ideal natural traps as their surface samples provide an exact analogue of the depositional environment of any fossil pollen cores taken from a lake. or constructed traps. Construction of traps and methodology eld The trap described by Bush (1992) (Figure 1) consists of a funnel (70 mm diameter) containing a glass. Bush and Rivera... generally at 50 m intervals. we found that bres remained dry even when bottles were full of water. It is possible that some kind of wick system may improve moisture transfer. The depth of seasonal inundation is a particularly important consideration in many tropical environments. seasonally inundated evergreen forest. the funnel design can deal with the high rainfall of tropical environments whereas the Tauber and Cundill traps are liable to become swamped. thus providing a source of moisture for the bres during dry periods. 1997. Second. The main problems with the use of soil samples are that in most environments oxidation occurs resulting in poor pollen preservation and rates of sedimentation are not ordinarily known (Wright. providing a detailed vegetation and climatic history of the past 50 000 years (Mayle et al. To avoid traps becoming submerged in such environments. However. Furthermore. First. Omitting the bottle greatly reduces the amount of equipment that needs to be transported into the eld. but this has not been tested. Knowledge of modern pollen-production rates could be important when interpreting the fossil pollen record. 2001. unpublished data). we recommend that the maximum depth of inundation be used as the ‘base level’ above which traps should a) 7 cm b) 1 litre bottle Mesh Whatman GF/D filter paper Drain hole Viscose Rayon Staple Figure 1 Schematic of modi ed Old eld trap. while collecting traps from NKMNP during the dry season. contain few mosses. Bush et al. 1998. Furthermore. such as those used by Tauber (1974). 1997.. cerrado (upland) savanna. (b) The trap reservoir. for use in tropical regions for two reasons.. e.g. These studies have tended to use either natural traps. Vincens et al. 1967). . Therefore we chose this trap design for a three-year study of the ve undisturbed tropical ecosystems within Noel Kempff Mercado National Park (NKMNP). the spatial distribution of lakes rarely matches that of the ecosystems under study and the extent to which bioturbation has occurred is usually unknown. NKMNP is therefore an exceptional location for carrying out modern pollen studies and allows the effectiveness of the trapping method to be assessed in a variety of different ecosystems. mounted in a plastic bottle and placed on the ground. this trap is light and easy to construct and therefore ideal for use in the tropics where sites are often remote and dif cult to access. Bush.bre lter paper and viscose rayon staple.. Moss polsters are not an appropriate method of sampling in much of the tropics as many tropical ecosystems.. Killeen. 1997.

bre equivalent) using a buchner funnel with suction. It is unnecessary to add preserving chemicals in the short term.: Technique for sampling modern pollen rain in the tropics 615 a) 7 cm b) Trapping medium Sealant Other symbols as Fig. personal communication). burnt or covered by leaf litter or other dead vegetation. adapted for speci c use with pollen traps that use bre as the medium of collection. years) then it may be advisable to add water to prevent oxidation. Discard the ltrate and use tweezers to transfer the lter papers into 50 ml centrifuge tubes. The trap described above can hold ~250 cm3 of dry viscose rayon staple. ‘Washing’ method for viscose rayon staple This protocol is designed for the extraction of pollen from viscose rayon staple.. that pollen rain is being sampled from the same level in the air. . be modi ed accordingly if sample sizes differ. in 7% hydrochloric acid (HCl) and add the solution to the bag if PAR is to be calculated. Wash repeatedly with copious quantities of water to separate the pollen from the bre. However. These should be kept in darkness at 3°C to prevent microbial growth. (a) Schematic cross-section view of the head and mounting of the pollen trap. Note mesh is 2 mm mosquito netting. C. Vigorous teasing and kneading of the wool is required to free material. Centrifuge the sample for ve minutes at 3000 rpm. (1) Extraction from viscose rayon staple (a) Add 5% potassium hydroxide (KOH) or sodium hydroxide (NaOH) to the bag containing the viscose rayon pad (50 ml is usually suf cient to saturate the bre). The use of a ‘washing’ methodology is necessary as no effective chemical procedure has been found which is capable of destroying the viscose rayon staple while leaving the pollen in a countable state (Bush. Dissolve exotic marker pollen. the ground can be taken as ‘base level’. Alison.g. as the lter paper can become clogged. if samples are to be stored for a long period of time (e. After collection. The quantities of chemical recommended here are for use with samples of this size and the protocol should (2) Separation of ‘washed’ sample from buchnar funnel lter paper(s) (a) Use concentrated (40%) hydro uoric acid (HF) to dissolve the glass. or overnight cold. 1 Stake Wire Figure 2 Modi cations presented in this paper to the trap described by Bush (1992). (1997) and Bush and Rivera (1998) advocate the use of viscose rayon staple in pollen traps and recommend a ‘wash’ methodology for sample preparation. (b) (c) Preparation methodology We present two methodologies. Traps should be positioned at a set height (e. the bre. the rst for those traps containing viscose rayon staple as the trapping medium and the second for those with cotton bre. Multiple lter papers may be needed for large samples. In environments not subject to seasonal ooding. e. be positioned. (1991). Gosling et al. (1997). lter and mesh should be separated from the funnel and sealed in a plastic grip-lock bag. 50 cm) above the ‘base level’ to ensure. The following method is modi ed from that described by Behling et al. Leave overnight to allow the pollen to be freed from the bre. as far as is possible. 1992. Typically. Note the small size of the trap and the uorescent marker tape. The lter papers now hold all the material washed from the bre. We suggest that it is preferable to mount traps on stakes even in ecosystems that are not inundated as it makes them more easily visible and less likely to be damaged by animals.bre lter paper(s) and also remove any silicates that may have entered the sample. These are modi cations from the standard pollenpreparation procedures as described in Faegri and Iversen (1989) and Moore et al.William D.. and decant the supernatant. Lycopodium tablet(s).g. (b) Photograph of a funnel trap mounted on a stake. It may be necessary to repeat this process if a number of lter papers have been used or there is a large quantity of silicates in the sample. 4–5 washes are sufcient to remove all organic material. Then lter the sample through a Whatman GF/D lter paper (or other glass. This procedure can be carried out over two hours by heating the sample in the HF to 90°C. This makes the sample alkaline which takes any humic acids within the sample into solution.g. Behling et al.

The percentage data should show a 1:1 relationship if there is no biasing towards certain pollen types between washes.992 0.1055 Linear equation relating in ux rates y y y y y = = = = = 0.8241x + 0.13 76. Once counted (~300 grains). The degree to which the extraction has been successful is determined by the extent to which the ‘un-recovered’ data set is a subset of the ‘recovered’ data set.(NB: since acetolysis mixture is explosive if it comes into contact with water.8 159. The relationship between the concentration data should show the effectiveness of the washing protocol at extracting the pollen from the bres.1979 1.3303 0. Place in hot 7% HCl for 20–30 minutes.2036x + 0. samples can be run concurrently in separate tubes. relationship could suggest that certain pollen types were more easily washed from the bre than others. Acetolyse the sample. It was conceivable that different pollen types might have been preferentially released or held within the bres.231 63. bre should Figure 3 The effectiveness of the washing method in extracting pollen from viscose rayon staple (semi-deciduous dry forest sample). The cotton bre should be separated from the lter paper and mesh parts of the trap. a and b). Then centrifuge and decant. (b) Total concentration pollen-assemblage data. Centrifuge and decant. . Figure 3.2206 1. This lter paper was then processed separately (hereafter referred to as ‘un-recovered’ sample). the two assemblages were plotted against each other (Figure 3.995 0. Recombine with the rest of the sample after extracted from the cotton bre. Then a clean lter paper was placed in the buchnar funnel and one additional wash performed.8707x + 0.0855x – 0. Destruction method for cotton bre The quantities of chemical recommended here are for use with samples of cotton bre with a volume of 250 cm3 when dry.616 The Holocene 13 (2003) Table 1 Test for biasing of samples through washing preparation: equations describe the relationship between pollen assemblages ‘recovered’ using the washing protocol described in the text and a ‘un-recovered’ assemblage recovered from a subsequent wash Ecosystem Terra rme evergreen forest Inundated evergreen forest Cerrado savanna Inundated savanna Semi-deciduous dry forest Linear equation relating % y y y y y = = = = = 0. adding the exotic marker tablets if required. (1991) and mount onto slides.1411x + 0. The effectiveness of the ‘washing’ preparation method at extracting the pollen grains from the bres had to be established to allow con dence in any results produced. (3) Final treatment (a) Acetolyse the sample following the standard procedure described in Moore et al. To dry. place sample in 7% HCl and heat to 90°C for 20– 30 minutes to remove silicate residues and uorosilicates (heating increases the solubility).2482x + 0.563 R2 0. according to the above protocol (hereafter referred to as ‘recovered’ sample). 2 = three-pored Moraceae. Sieve at 250 mm to remove any large fragments of mesh that may remain.565 0.781 0. dependent on grain morphology or size. (a) Percentage pollen-assemblage data. Any outliers to a strong positive (e) (2) Extraction of sample from cotton bre (a) Acetolyse the sample to destroy the cotton bre and any other cellulose and/or polysaccharides. Part (d) should be run concurrently with the cotton bre (see section 2 of this protocol) and is therefore described in detail below. To test for biasing. Samples must be subjected to the same process as the cotton bre described below.768 (b) (c) Once all fragments of lter paper and silicates have been removed.0893x – 0. This relationship can be tested using a simple linear equation (Table 1. Numbers on graph indicate the position of pollen types: 1 = two-pored Moraceae. (a) (b) (c) (d) Dissolve in HF. a and b). using a ‘whirlimix’ and stirring rod to ensure that the entire sample is mobilized and washed effectively. the cotton bre must be dry beforehand.2947x – 0.2043x – 1149 2151.1884 0. Wash sample in water. (1) Extraction of sample from trap lter paper and mesh Parts (a) and (b) of this section of the protocol basically follow the steps described above (see parts (a) and (b) in section 2 of the above protocol) and therefore only a summary is presented here.776x + 0. a sample was washed ve times.

In the case of all these samples. Negrelle. However.A. However. Moreno. Grabandt.ngdc. Asanza. Santa Cruz. Colinvaux. the slight differences are well within the statistical error expected with the count size —— 1992: A simple yet ef cient pollen trap for use in vegetation studies. and Tinsley.B. 239–58. P. 275–76. a and b). Henry Hooghiemstra (University of Amsterdam). samples should be checked regularly until dry. 40°C.html (last accessed 31 January 2003). The eldwork was funded by The Royal Society (FEM). D. M. Ammann. This test demonstrates that although the washing technique does not remove all the pollen from the bre. R.B. 729–43. Vincens. and Colinvaux. Bush. 2000: Use of plots to de ne pollen-vegetation relationships in densely forested ecosystems of Tropical Africa. Review of Palaeoecology and Palynology 114. 83–98. P. Wash with water and recombine all parts of cotton bre sample if applicable. C.. and Iversen. Review of Palaeobotany and Palynology 29. the ‘Fundacion Amigos de la Naturaleza’ for permission to work within NKMNP. Palaeoclimatology.) To prepare the acetolysis mixture combine 9 parts of acetic anhydride (i.E.. M. the British Ecological Society (FEM) and the Quaternary Research Association (WDG). and Hyvarinen. Nick Tate (University of Leicester). de Namur. 287–99. 85–87. and Rivera. 5–17. 1997: Modern pollen rain data from the tropical Atlantic rain forest. Global Ecology and Biogeography Letters 7. H.. an anonymous referee. 2001: Pleistocene deposits of the Amazon Fan: a palynolog- Recommendations (1) We recommend that a simple funnel trap (Figure 2) mounted. Hoorn. a and b). Review of Palaeobotany and Palynology 97. M. WDG gratefully acknowledges a University of Leicester PhD studentship. 5 ml). R. This demonstrates that this method does not extract all the pollen from the bres. M. A. Mark Bush (Florida Institute of Technology) for discussions on techniques of sampling modern pollen rain.: Technique for sampling modern pollen rain in the tropics 617 (b) (c) (d) be placed in a beaker and put into a drying oven at c. the ‘unrecovered’ sample still contained tens of thousands of pollen grains indicating that there was still a substantial quantity of pollen being retained within the bres even after repeated vigorous washing (Figure 3b). Faegri. 45 ml) with 1 part of concentrated sulphuric acid (i. It may be necessary to add extra steps or modify certain existing steps depending upon sample characteristics. Gosling et al. 2002: On the interpretation of fossil Poaceae pollen in the lowland humid neotropics. Miller. 379–92. Pardoe. Hicks. Elenga. Grimsby) for discussions on the destruction of viscose rayon staple. Centrifuge for ve minutes at 3000 rpm. thus ensuring complete recovery of all pollen grains and spores. M. S... for logistical support. thus showing that the washing methodology was not selectively biasing the recovery of certain pollen taxa. This is probably because of higher diversity in pollen taxa identi ed in these samples. Journal of Biogeography 13. 1998: Pollen dispersal and representation in a neotropical rain forest.. sieve at 250 mm. South Brazil. local guides Juan Surubi and Pastor Soliz for invaluable assistance in the eld. 1996: A long pollen record from lowland Amazonia: forest and cooling in glacial times. Review of Palaeobotany and Palynology 112. de Oliveira. and Bonne lle. (3) Final treatment (a) (b) Recombine with the sample extracted from the lter paper and mesh.R. Acknowledgements We acknowledge the following: colleagues at the Museo de Historia Natural ‘Noel Kempff Mercado’. Discussion As with all pollen preparation methodologies. New York: Wiley. noaa. P. P. Palaeogeography. Journal of Tropical Ecology 17. H. 1986: A new design of pollen trap for modern pollen studies.B.and under-represented in the semi-deciduous dry forest sample (points 1 and 2 in Figure 3. P. Bush. H. K. 65–147. Journal of Vegetation Science 3.. Bolivia. Figure 3. then decant supernatant away. The PAR values for the ‘un-recovered’ sample were typically 20% of those of the ‘recovered’ sample.J. and Schwartz.. very similar sized Moraceae grains were found to be both over. as long as the exotic marker pollen is added at the beginning of the preparation procedure. 2001: Precipitation signal in pollen rain from tropical forests. S. Place suf cient mixture into each tube to cover the dry cotton bre and incubate in water-bath at 90°C for three minutes. Katy Roucoux and Ian Lawson (University of Cambridge) for constructive comments which improved the manuscript.. V-P.. P. Mario Saldias and Nelson Rodriguez.. Science 274. mechanism for sampling modern pollen rain and its spatial characteristics. Palaeoecology 177. J. especially Mario Suarez Riglos. However. Pollen et Spores 28. Hicks. Chris Alison (‘Acordis’. Moreno. 1986: Sampling modern pollen deposition ¨ by means of ‘Tauber traps’: some considerations. and mount onto slides. de Oliveira. The R2 values produced clearly show that there is no signi cant difference in the composition of the pollen assemblages between the ‘recovered’ and ‘un-recovered’ samples from the evergreen forest. 1980: Pollen rain in relation to arboreal vegetation in the Colombian Cordillera Oriental. Latalowa.B.William D.A. M.E. Roux. 2001: The in uence of biogeographic and ecological heterogeneity on Amazonian pollen spectra. South India. 219–42. D. B. on a stake is an effective and reliable . The outliers to the linear relationships established for the data sets appear not to be a product of biasing caused by differences in grain morphology or size. For example. and Bush. and Colinvaux. 79–96.A..E. Behling.B. The cerrado savanna ecosystems show a weaker relationship between the two sets of data.. (2) We recommend that cotton bre should be used as the trapping medium as it is possible to completely remove this from the sample. Reserva Volta Velha. C.A. H. Treat sample with 10% NaOH or KOH to remove humic acids. both the preparation methodologies described in this paper have been shown to provide statistically reliable pollen counts. E. Cundill.e.e. M.C. J. We recommend that large samples are split and acetolysed in separate tubes to be recombined after destruction of the cotton bre. this methodology should be viewed as a recipe. R. it is still be possible to calculate PARs reliably. semi-deciduous forest and inundated savanna ecosystems (Table 1. 2002: http://www. After three minutes remove sample from water-bath and top up with glacial acetic acid to cool the sample and stop the acetolysis reaction. Bush.. this is not a cause for concern as the pollen-assemblage composition of the ‘unrecovered’ sample is not statistically different from that of the ‘recovered’ sample. References Barboni. where possible. R. E. 1989: Textbook of pollen analysis (fourth edition).

H. and Killeen. 359–69. University of Georgia. Journal of Biogeography 27. and Pendry. Houston. USA. 1996. PhD dissertation.K. J. Dallas. 944A. editors. H. Jr 1967: The use of surface samples in Quaternary pollen analysis.T. D. Burbridge. Achoundong. American Association of Stratigraphic Palynologists Foundation. Guillet. 321–30.E. and Collinson. R.A.E. Wright. Dubois.B. 2291–94. Webb. USA. Maddy. F.. M.. In Goodman. P. Mayle. and Riera. Tauber.N. New Phytologist 73. Oxford: Blackwell Science. T... 2003: An analysis of modern pollen rain on an elevational gradient in southern Peru. A. S. 2000: Neotropical seasonally dry forests and Quaternary vegetation change. C. R. M.T.618 The Holocene 13 (2003) ical analysis of holes 940A.K..J. V.A. R.. 2000: Millennial-scale dynamics of southern Amazonian rain forests. Beyala.. 1991: Pollen analysis (second edition). Review of Palaeobotany and Palynology 110.K. C. Texas.. and 946A (ODP LEG 155). M. Bush. C. Texas.A. and Silman. Proceedings of the IX international palynology congress. in press. D. 191–208. Buchet. Review of Palaeobotany and Palynology 2. Journal of Tropical Ecology. Cambridge: Quaternary Research Association.. Science 290. 261–73. and Brew. G. editors 1995: Statistical modelling of Quaternary science data: technical guide 5. D. J. de Namur.. Weng..S. Vincens. and Ciance. Prado. 2000: Pollen-rainvegetation relationships along a forest-savanna transect in southeastern Cameroon. Pan l.E. B.D. R. 2001: Late Holocene forest and savanna diversity and dynamics across an Amazonian ecotone. 329–38. . 1974: A static non-overload pollen collector. Pennington. Moore.. B. G.E.