This article is published in Environmetrics as a special issue on Quantitative approaches to ecosystem service evaluation, edited by R. I. Smith, Centre for Ecology and Hydrology, UK; E. M. Scott, School of Mathematics and Statistics, University of Glasgow, UK; J. McP Dick, Centre for Ecology and Hydrology, UK.
Special Issue Paper
A causal modelling approach to spatial and temporal confounding in environmental impact studies†
Article first published online: 18 APR 2011
Copyright © 2011 John Wiley & Sons, Ltd.
Special Issue: Quantitative Approaches to Ecosystem Service Evaluation
Volume 22, Issue 5, pages 626–638, August 2011
How to Cite
Paul, W. L. (2011), A causal modelling approach to spatial and temporal confounding in environmental impact studies. Environmetrics, 22: 626–638. doi: 10.1002/env.1111
- Issue published online: 4 JUL 2011
- Article first published online: 18 APR 2011
- Manuscript Accepted: 1 FEB 2011
- Manuscript Revised: 23 JAN 2011
- Manuscript Received: 20 OCT 2009
- before-after control-impact (BACI) design;
- causal modelling;
- environmental impact study;
- spatial and temporal confounding
Deciding whether an anthropogenic disturbance has caused a change in an ecosystem is problematic because of spatial and temporal confounding. The generally accepted approach to this problem has been to use a before-after control-impact (BACI) type of design. However, advances in graph theory and causal modelling now provide a formal basis for selecting the covariates that need to be observed in order to control confounding bias, and the solution provided by these advances is simpler than that provided by any of the BACI-type designs including BACI itself as well as BACIP, Beyond-BACI and MBACI designs. Based on an explicit description of the nature of spatial and temporal confounding in causal models for two hypothetical environmental impact studies, it is argued that confounding can be controlled by adjusting directly for spatial or temporal location in a before-after (BA) or control-impact (CI) study. It is further argued that there is no advantage in combining these designs in a BACI-type study, either from a causal modelling perspective or from the perspective of the assumptions implicit in BACI-type designs. Copyright © 2011 John Wiley & Sons, Ltd.