COS 14-7
Improving estimates of regional vegetation: Using pre-settlement vegetation data and variable wind speed to quantify pollen dispersal and source area

Monday, August 10, 2015: 3:40 PM
337, Baltimore Convention Center
Kevin D. Burke, Nelson Institute for Environmental Studies, University of Wisconsin-Madison, Madison, WI
Simon Goring, Geography, University of Wisconsin-Madison, Madison, WI
Jack Williams, Geography, University of Wisconsin-Madison, Madison, WI
Tracey Holloway, Nelson Institute for Environmental Studies, University of Wisconsin-Madison, Madison, WI
Background/Question/Methods

Pollen-based vegetation reconstructions are the primary source of information about spatio-temporal trends in vegetation dynamics at timescales of centuries to millennia. Pollen samples from individual lakes, bogs, and small hollows provide information about the vegetation within their respective source areas, which, when mapped together, generate vegetation reconstructions across spatial and temporal scales. Climate histories can also be reconstructed using empirical taxa-climate relationships. A fundamental need in these reconstructions is to estimate the pollen source area and account for the effects of intertaxonomic differences in pollen production and dispersal. Mechanistic pollen-vegetation models have continuously evolved for nearly a century, with recent advances culminating in the Landscape Reconstruction Algorithm. However, most applications of the LRA do not account for anisotropies in pollen source area introduced by varying wind speed and direction. Here we investigate the effects of wind speed and direction on pollen source area. We obtain long term estimates of dominant wind patterns for select lake sites using North American Regional Reanalysis (NARR) weather data for 1979-2012, and use pre-settlement (ca. 1810-1904) forest composition across the prairie-forest ecotone in the upper Midwestern United States and a dataset of settlement-era pollen samples compiled by the PalEON project to model pollen-vegetation relationships.

Results/Conclusions

Our results show taxon-specific dispersal relationships that suggest some taxa are less well dispersered, while others are more well dispersed. This results in a pattern of over or under-representation of vegetation across the landscape when reconstructions only consider pollen data. Comparisons of our modelled results with those of a unidirectional, homogenous model also shows a substantial effect of regionally varying winds on lake pollen loadings. Additionally, comparisons of results to existing pollen records from the Neotoma Paleoecology Database show a strong correlation (http://neotomadb.org). Variable wind speed and direction, as well as other atmospheric effects including precipitation and instability play an important role in pollen transport and deposition. By better accounting for these effects we can improve the predictive capacity of pollen-vegetation models and reduce uncertainty.