Representational bias in phytoliths from modern soils of central North
America: Implications for paleovegetation reconstructions
Ethan Hyland, , Selena Y. Smith, Nathan D. Sheldon

http://www.sciencedirect.com/science/article/pii/S0031018213000837#

Abstract
Understanding localized patterns and community compositions of
vegetation in an environment is critical to the reconstruction of
climatic and ecological conditions across all spatiotemporal scales.
One of the most accurate and useful ways to characterize vegetation,
and therefore to describe the climatic and ecological conditions of a
location, is through the plant fossil record. Phytoliths (plant silica
microfossils) are often preserved in the absence of other
paleobotanical data and are becoming more widely used for deep-time
vegetation reconstructions. Significant work has been done to
standardize the analytical methodology of phytolith extraction,
statistical analysis, and interpretation, but more detailed
investigations are needed to understand how well a given soil
assemblage represents the actual aboveground plant biomass for a given
ecosystem.

We present results from paired soil phytolith assemblages and local
vegetation assemblages across the central United States, including
temperate forest, grassland, and rangeland/scrubland ecosystems.
Phytolith assemblages obtained via extractions from soil A-horizons
were compared to percent cover of species and plant biomass estimates
obtained via in situ field observations and aerial estimates of tree
cover to analyze differences in the relative abundance of forest/woody
vegetation vs. grasses. Soil phytolith assemblages from all sites
average a 29% bias toward the grass morphotypes as compared to actual
aboveground biomass observations, and comparisons to percent cover
yielded broadly comparable bias figures. Percent bias estimates do not
show significant correlations to most environmental factors
(temperature, precipitation, local elevation), however, an extremely
strong correlation (p < 0.001) was observed with soil order type. This
is likely due to the fact that soil order reflects both vegetation
type and chemical factors known to affect overall phytolith
assemblages; therefore, soil order is a proxy that aggregates several
sources of bias. As a result, we suggest further research into the
development of correction factors between phytolith sample assemblages
and their interpreted past counterpart ecosystems based on estimates
derived from modern analyses of each major soil order type. Such
background corrections are essential to the continued use of
phytoliths as a proxy for past vegetation and ecological
reconstructions of temperate ecosystems throughout the Phanerozoic
record.