Pollen–vegetation relationships along steep climatic gradients in western Amazonia Dunia H. Urrego, Miles R. Silman, Alexander Correa-Metrio & Mark B. Bush Keywords Bioclimatic envelopes; Moss polsters; Pollen calibration; Probability density functions; Vegetation plot data. Received 10 October 2010 Accepted 16 February 2011 Co-ordinating Editor: Otto Wildi Urrego, D.H. (corresponding author, d.urrego@epoc.u-bordeaux1.fr); Correa-Metrio, A. (acorrea@my.fit.edu) & Bush, M.B. (mbush@fit.edu): Department of Biological Sciences, Florida Institute of Technology, 150 West University Boulevard, Melbourne FL 32901, USA Silman, M.R. (silmanmr@wfu.edu): Department of Biology, Wake Forest University, 1834 Wake Forest Rd., Winston-Salem NC 27106, USA. Abstract Question: How accurately do Amazonian montane forest pollen spectra reflect the vegetation? Can compositional changes observed in the vegetation along environmental gradients be identified in the pollen spectra? How well do herbarium collection data and bioclimatic envelopes represent abundance changes along elevation gradients? Location: Amazonian montane forests, Peru. Methods: Moss polsters collected along five altitudinal transects spanning over 3000 m a.s.l. were used to characterize pollen spectra. Vegetation plot data from a network of 15 1-ha permanent plots were used to correlate pollen spectra with present-day vegetation. Probability density functions (PDFs) fitted to pollen and plot data allowed comparisons using Spearman correlation coeffi- cients. Ordination analyses were used to summarize changes in pollen spectra. Correlations between pollen-based PDFs and previously-published herbarium collection PDFs were also evaluated. Results: Pollen spectra closely reflected changes in species composition along elevation gradients. A mid-elevation shift in pollen spectra was identified using ordination analyses. Pollen spectra from the driest forest in our data set were statistically different from those of wet forests. Pollen abundance PDFs along the altitudinal gradient were significantly correlated (P o 0.01) with PDFs fitted to plot abundance, basal area and herbarium collection data for ten out of 11 taxa analysed. Conclusions: Pollen spectra closely reflected the vegetation composition of Amazonian montane forests. The differentiation of pollen spectra from dry localities showed the potential of genus-level pollen data to reflect precipitation gradients. Pollen spectra also reflected mid-elevation compositional changes well along the lower elevation limit of ground cloud formation. Despite collection biases, herbarium-based bioclimatic envelope PDFs also represented well forest compositional changes along elevation gradients. Introduction Neotropical montane forests support an exceptionally large proportion of global biodiversity and have special significance because of their high levels of endemism (Leo 1995; Pounds et al. 1999; Still et al. 1999; Foster 2001). Both high biodiversity and endemism in these forests result from strong environmental gradients in tempera- ture and precipitation, which are mostly determined by the Andes and their topographic complexity. Many tropical montane species are believed to be habitat spe- cialists that occupy tightly constrained niches (De Oliveira & Daly 1999; Ter Steege et al. 2003). It is this dense species packing and the high levels of specialization and ende- mism that makes tropical montane forests especially vulnerable to current climate change (Pounds et al. 1999; Still et al. 1999; Orme et al. 2005). Understanding species distributional ranges has been a focus in vegetation science that has led to several Journal of Vegetation Science 22 (2011) 795–806 Journal of Vegetation Science Doi: 10.1111/j.1654-1103.2011.01289.x r 2011 International Association for Vegetation Science 795