Aims: Historical land-use legacies and chemical soil characteristics both explain either directly or indirectly the habitat quality of Nardus grassland, which is protected under the European habitat directive. Yet the relative importance and complementarity of both sets of variables are generally unknown. This knowledge is also relevant for practical reasons, as historical land-use variables can be used in desktop spatial analyses, whereas soil characteristics require field surveys to collect samples for laboratory analyses. To this end, we aim to disentangle the relative importance of historical landuse legacies and soil chemistry for the Nardus grassland quality, and determine the potential of habitat suitability mapping for predicting potential restoration areas.Location: Natura 2000 grasslands in Flanders (northern Belgium).
Methods:We compared the model performance of three generalized additive models (GAMs), using either land-use history metrics, soil chemistry, or both as explanatory variables, with the Nardus grassland indicator species count as response.Results: All three models were able to predict areas suitable for at least three Nardus grassland indicator species with high sensitivity and specificity. However, a minimum of four indicator species are required for a favorable conservation status of Natura 2000 Nardus grasslands in Flanders. Using this threshold to detect high-priority zones, the model based on historical land-use variables resulted in a lower sensitivity than models which included soil chemistry.
Conclusions:We suggest a two-step approach, with an a priori desktop spatial analysis based on historical land-use variables subdivided in a high-priority zone and a lower-priority zone. If the targeted area for restoration or conservation can be found within the high-priority zone, additional soil analyses are only required to help guide conservation and restoration measures. If additional sites are considered within the lower-priority zone, a field survey to collect additional soil data is recommended.