Imputation of environmental variables for vegetation plots based on compositional similarity
Journal
Journal of Vegetation Science
Journal Volume
21
Journal Issue
1
Pages
88-95
Date Issued
2010
Author(s)
Abstract
Question: Large databases contain many plots, but few subsets with measured environmental data. To obtain broader datasets, researchers use expertbased indicator values as surrogates; alternatively, these can be estimated by imputation. Does imputation provide more exact approximations than indicator values? Location: West Carpathians (Slovakia, Poland, Czech Republic) and Bulgaria. Methods: We developed a simple imputation method based on plot similarity that estimates missing environmental variables for plots - MOSS (mean of similar samples). This was tested for pH and conductivity, important environmental factors influencing vegetation composition and structure within wetlands, on two datasets of 485 (West Carpathians) and 118 (Bulgaria) plots for which directly measured values were available. The West Carpathian dataset was used for calibration. Imputation was based on calculating mean of the measured factor from a group of most similar plots. Using pre-defined similarity criteria, we selected subsets of both datasets and compared estimated and measured values. Using root mean-squared error of prediction, we compared predictive power of MOSS with Ellenberg indicator values and other recent methods. Results: Within one study region, MOSS predicts sample pH and conductivity more precisely than Ellenberg and similar calibration methods. Predictive power slightly decreased when MOSS was transferred to a distant region. Conclusions: Imputation using MOSS appears to accurately predict pH and conductivity from existing composition data within a single geographical region, and thus increases number of replicates. MOSS does not require expert-based indicator values, which may be imprecise. We provide examples where MOSS can be utilised without risk of circular reasoning or introducing pseudo-replications. © 2009 International Association for Vegetation Science.
SDGs
Type
journal article