Plant species as predictors of soil pH: Replacing expert judgement with measurements

Abstract

Question: The use of expert-based indicator values to estimate abiotic conditions from vegetation is widespread. However, recent research has shown that expert judgement may contain considerable bias and thereby introduces a large amount of uncertainty. Could expert based indicator values be replaced by indicator values based on field measurements? Location: Europe. Methods: We developed a method to estimate species response based on measured physical data, and a method to predict abiotic conditions from the vegetation composition using these responses. This method was tested for soil pH. Results: We were able to estimate the pH response of 556 species of the Dutch flora. Ca. 20% of the responses were, at least, bimodal and many responses had a very wide range. The simplest method ('raw mean') yielded the best prediction of pH; the indicator value of a species is the mean of the soil pH values of the sites where it was observed. A list of all rawmean estimates per species is given. The predicted pH of a new site is the mean of the indicator values of the present species. The estimated species responses were validated on independent Dutch and European data sets. Older successional stages seem to be predicted better than younger stages. Conclusions: Our method performed better than the popular Ellenberg indicator system for the Dutch data set, while being just as easy to use, because it only needs a single value per species. We foresee that, when more data become available, our method has the potential to replace the Ellenberg system.

Ecological indicator values of forest communities in Zonguldak-Çitdere region

“…For this reason, occurrence and abundance of different plant species enables ecologists to make statements about the prevailing environmental conditions. This method, which has been used for many years in Europe, is the main approach used to predict an environmental variable by means of the flora without measuring them (Jean-Claude and Eva, 2003;Wamelink et al, 2005). One of the most widely used indicator species system was proposed by , which describes the response of a species to edaphic and climatical parameters in Central Europe.…”

Experience with using Ellenberg’s R indicator values in Slovakia: Oligotrophic and mesotrophic submontane broad-leaved forests

“…Ellenberg's indicator values (EIV) have been exhaustively discussed by many authors (e.g., Mucina 1985;Jurko 1986;Klimeš 1987; and recently by Schaffers & Sýkora 2000; Wamelink et al 2002Wamelink et al , 2005, and several aspects of their usage have been criticised.…”

“…Also the effects of cover weights and indifferent species have been extensively examined by Käfer & Witte (2004) and Ewald (2003). Other bioindication methods such as the Maximum Amplitude Overlap method (PepplerLisbach 2008) and the Maximum Likelihood method (Wamelink et al 2005) can also be considered. However, these last two methods cannot be used for EIVs since they operate solely with species tolerances or with complete information of species response curves.…”

“…Therefore various methods of statistical calibration have been introduced to calculate indicator values directly from field measurements (e.g. Huisman et al 1993;Wamelink et al 2005;Peppler-Lisbach 2008). These are commonly using weighted averaging (ter Braak & Barendregt 1986) or modelling species response curves against measured physical data (e.g., ter Braak & Looman 1986;Austin et al 1994;Huisman et al 1993), where both the species IVs and tolerances are statistically defined (e.g., ter Braak & van Dam 1989;Gégout & Pierrat 1998;Schröder et al 2005).…”

“…These are as follows: type I -a flat linear relationship, type IIa monotone relationship, type III -a plateau-like relationship, type IV -a symmetric Gaussian relationship, and type V -a skewed unimodal relationship. This method of statistical calibration improves bioindication and supposedly circumvents bias in EIVs (Wamelink et al 2005). An alternative calibration procedure of adjusting the original EIVs with respect to a measured gradient using the HOF model was suggested by Lawesson et al (2003).…”