The potential of REVEALS-based vegetation reconstructions using pollen records from alluvial floodplains

Abstract

The two-step Landscape Reconstruction Algorithm—comprising the models REVEALS and LOVE—translates fossil pollen assemblages into quantitative reconstructions of past vegetation cover. The REVEALS model has been applied to lakes and bogs of varied sizes, but application to records from alluvial sites is currently lacking and could further extend these reconstructions to areas where lakes and bogs are absent. In addition, since alluvial sites are often located in areas that experienced early anthropogenic impact, such data will provide more insight into land cover change in the context of agricultural development. In this paper, we test the performance of the REVEALS model using pollen records from multiple alluvial sites obtained from the floodplains of the Belgian Dijle catchment. The modelled vegetation cover is compared to the observed vegetation cover based on a historical land cover map of ca. ad 1775 (ca. 175 cal bp). The discrepancy between the modelled and the observed regional vegetation cover is relatively small and can largely be explained by local differences in land cover surrounding the sites, based on the results of the LOVE model. This study concludes that the LRA approach accurately reconstructs the regional vegetation cover of the river catchment based on pollen obtained from alluvial floodplains, and shows that it is able to reflect major local differences in land cover as well. Including pollen data obtained from alluvial floodplains could lead to a substantial increase in spatial coverage of future LRA-based land cover reconstructions.