Aim: Grasslands occupy around 40% of the Earth’s land surface and can be regarded as the most common vegetation type in the world, with Poaceae being the most widespread angiosperm family of all.   Poaceae pollen size has previously been suggested as a proxy to reconstruct the past vegetation and climates in the Amazon area, but it is still controversial if this variable indeed can be used in broader spatial and deep-time scales. Here we set out to perform a comprehensive assessment and test the robustness of this proxy.

Location: Amazon drainage basin (ADB).

Taxon: Poaceae.

Methods: One hundred and twenty-seven specimens from 58 species (non-crops) across the Poaceae phylogeny from the Amazon drainage basin (ADB) were prepared for pollen grain size analyses, in order to explore their relationship with abiotic and biotic variables (vegetation type, soil composition, climate conditions, photosynthetic pathway, and genome size). Phylogenetic generalized least squares (P-GLS) model and linear mixed models (LMM) were applied to assess the proxy.

Results: Our measurement data show that Poaceae pollen size presents a very wide range (18.77 - 71.62 μm), not only at the genus and species levels but also within species. There is no obvious relationship between pollen size and the explanatory variables considered here, however.

Main conclusion: Poaceae (non-crop) pollen size does not respond to explanatory variables, and therefore cannot as a useful proxy to reconstruct past vegetation and climate.

1. Sample collection and processing

We sampled pollen grains 127 plant specimens from 58 species belonging to 9 different Poaceae subfamilies. 

2. Selected variables and data collection

Poaceae occur in a wide range of ecosystems in South America, ranging from mountain to lowlands, cold to temperate grasslands, floodplains, and coastline. 

2.1. Vegetation type

We assigned each sampling locality to one of the following five vegetation types: desert, montane grassland shrubland, savanna, tropical dry forest, and tropical moist forest. 

2.2. Soil type

We used the Soil and Terrain Database (SOTER) for Latin America and the Caribbean (SOTERLAC), version 2.0, at a scale of 1:5000000 to define soil categories for each site, using the WRB (World Reference Base) Soil Groups to assign each sampling locality to one of 19 soil types. DOMSOIL_UN represents the soil type corresponding to each sampled plant specimen.

2.3. Modern climate data

We used the WorldClim 2.1 database and Climate Research Unit (CRU) TS4.04 database (Harris et al., 2020) to extract the mean annual temperature. cru.MAT and cru.TAP represents the climate data corresponding to each sampled plant specimen.

2.4. C3/C4 photosynthetic pathway

Photosynthetic pathways assigned to each species came from Klink and Joly (1989), Giraldo-Canas (2010), Bremond et al. (2012), and references therein.

2.5. Genome size

We took genome size information from Kew Garden's C-values Database.

All code can be opened with RStudio.