PIs : Paulina PINTO (UMR 1137) et Ingrid SEYNAVE (UMR 1092)
Co-applicants: J-L Dupouey et P Montpied (EEF), J-C Gégout, Ch. Piedallu, J-C Pierrat (LERFOB)
Collaborations: M Legay, C Richter, V Boulanger (ONF), S Cavaignac (FCBA), E Paillassa, P Gonin (IDF), Y Dumas, S Perret (IRSTEA)
Context — The presence of plant species has long been used as an indicator of environment characteristics and forest potentialities. In France, forest management is based on a mapping of forest sites, which are predominantly determined by the site flora. However, the definition of site types remains qualitative. Accordingly, forest site classification cannot be connected to models determining distribution or growth of tree species nor to evolution models of nutritional conditions in a context of global change. Since the end of the 20th century, formalized methods have been developed to evaluate species’ indicator values in relation to environmental parameters. In France, indicator values for acidity (pH) and for nitrogen nutrition (C/N) have been determined for more than 500 forest species. However, these indicator values for estimating the pH and C/N of a site using floristic information are seldom used in forest management, because of the time and botanical expertise needed for flora inventories.
Objectives — Optimize flora inventories to predict nutritional characteristics of forest sites using bioindication methods.
Approaches — Firstly, using the random forests method, we will seek a reduced pool of species that maximize the accuracy of predictions of pH and C/N. Secondly, using data sets with floristic timed inventories and soil analyses, we will seek a compromise between the accuracy of predictions and time of data acquisition. Finally, we will study the effect of the number and size of plots on the bioindication quality. These analyses will focus on pH and C/N by comparing bio-indicated values to measured values from soil samples.
Expected results and impacts — To provide recommendations for the establishment of vegetation survey protocols and thus make bioindication methods useful for forest management and characterization of experimental plots.