Predictive modeling of the potential natural vegetation pattern in the Podyjí National Park, Czech Republic

Abstract

The relationships between 10 types of potential natural vegetation and several environmental factors derived from a topographical map (altitude, slope inclination, curvature, potential direct solar irradiation, sky view factor), a geological map and a spring phenological map, were modeled for the Podyji National Park (Czech Republic). Different vegetation pattern models were made using spatial predictive modeling together with the Bayes formula. Compared with the field-based vegetation map, the model which included geology as a factor and all the factors derived from the topographic map, yielded an overall correspondence (i.e. percentage of grid cells in which vegetation types in the field-based map and in the model matched) of 56.8%. This measure of correspondence, however, is not dependable in areas, such as the Podyji National Park, where one vegetation type prevails. Therefore, correspondence within vegetation types, i.e. the average of correspondences calculated for each vegetation type separately, was used, thus yielding a value of 39.4%. The most important factors in this model were the sky view factor, representing the differences between the valleys and the gently undulating uplands, and altitude, as a surrogate for the macroclimatic gradient. By replacing altitude and irradiation variables related to the macroclimate and mesoclimate respectively, with the spring phenological map, overall correspondence with the original map was increased to 62.6% and correspondence within vegetation types to 50.5%. This result shows that phenological maps, which are relatively easy to create and provide a surrogate for unavailable local climatic maps, might be a powerful tool for predicting vegetation pattern on landscape scale.