Effects of the sampling design and selection of parameter values on pollen-based quantitative reconstructions of regional vegetation: a case study in southern Sweden using the REVEALS model

Abstract: The need for quantification of land cover from pollen data has led to the development of a Landscape Reconstruction Algorithm (LRA). The LRA includes several models of which the REVEALS model estimates regional vegetation abundance using pollen assemblages from large sites (lakes or bogs). In this paper we explore the effects of selection and number of pollen samples, and choice of pollen productivity estimates on the REVEALS results. The effect of the size of vegetation surveys is also tested. The results suggest that the differences between two sizes of vegetation surveys have little effect on the model validation. The “characteristic radius” of regional vegetation in southern Sweden was estimated as 200 km. However, the vegetation composition in a 100 × 100 km² square matches well with that estimated by REVEALS. Whether 25, 20 (outliers excluded) or 4 pollen samples are used does not change the REVEALS reconstructions much although the error estimates are larger when outliers are included, and very large when only four samples are used. Therefore validation of the REVEALS model and REVEALS reconstructions of past vegetation can be performed using a limited number of pollen samples, although with caution. The use of many pollen samples from multiple sites is always better whenever possible. REVEALS reconstructions are closer to the actual vegetation when the Danish Pollen Productivity Estimates (PPEs) are used instead of the Swedish PPEs for Cereals, Rumex acetosa/acetosella, Plantago lanceolata and Calluna, indicating that the Danish PPEs are more reliable than the Swedish ones for those taxa. It is recommended to test more than one set of PPEs in validation and applications of the REVEALS model for a better evaluation of the results.