final_program_abstracts[1]

11 IMSC Session Program
Use of pollen data to investigate past climates: Spatial and
ecological sources of uncertainty
Tuesday - Plenary Session 3
Mary Edwards and Heather Binney
University of Southampton, UK
Pollen data are the most abundant form of information about past terrestrial
environments. Many hundreds of localities have been studied, particularly across
North America and Europe. For the period 21,000 yr BP to present, tens of thousands
of individual pollen spectra are recorded, many linked to a radiocarbon chronology.
This considerable archive of pollen data, both modern and fossil samples, provides the
basis of continental-scale reconstructions of past vegetation, and, either directly or
indirectly, climate. Various algorithms link current patterns of pollen distribution
with climate variables and are used to infer past climate conditions for different points
in time at each locality. These all suffer from a so-called ‘no-analogue’ situation when
fossil pollen spectra do not closely resemble any modern counterparts and the climatepollen
relationship breaks down. The ‘biomization’ approach circumvents this
problem by basing pollen classification on functional types, which are physiologically
linked to climate and which characterize various ‘biomes’ or major vegetation units.
Vegetation models can use the same vegetation units driven by the same bioclimatic
relationships. This paves the way for an inverse approach to climate reconstruction in
which a climate simulation drives a vegetation model, and the resultant vegetation
map is compared with pollen-based biomes to assess the effectiveness of the
simulation.
Two main types of spatial error or bias characterize the use of arrays of pollen data.
First, the relationship between abundance of a pollen taxon at a locality and the
abundance of the plants that produced it for a given source area is not linear, and, for
a number of physical and ecological reasons, the relationship varies for each pollen
type. The theory behind this is well developed; while it is difficult to implement
corrections in detail, a simple algorithm can be applied to reduce this kind of bias.
Second, pollen data comprise information points in a largely unpopulated space, and
the extrapolation of taxon abundances, biome extent, or reconstructed climate
variables away from the measured points remains a considerable challenge.
Abstracts 89