Citation

EIA: An Algorithm for the Statistical Evaluation of an Environmental Impact Assessment

Author:
Guisande, Cástor; Rueda-Quecho, Andrés J.; Rangel-Silva, Fabián A.; Ríos-Vasquez, Jorge M.
Publication:
Ecological Indicators
Year:
2018

1. Environmental impact assessment (EIA) studies are most frequently focused on assessing the possible impacts of a proposal and identifying alternatives, in order to minimize environmental damage. Little attention has been paid, however, to the development of statistical methods for evaluation of potential changes in the environment after the impact, particularly when control sites are not available.

2. Here, the EIA algorithm is presented, available as an RWizard (Guisande et al., 2014) application on http://www.ipez.es/RWizard. This algorithm provides tools for the detection of variables which change significantly, following environmental impact (EI), once the time at which the impact occurred is available.

3. In the algorithm, the following steps are taken: 1) variables are prioritized by their discrimination capacity to distinguish between the before and after of the EI; 2) estimation of the Area Under the Curve (AUC) of Receiver Operating Characteristic curves (ROC curves) is obtained for the before and after of the EI, 3) comparison of the AUC of the ROC curves (before and after the EI), with an AUC of 0.5 (the situation in which two distributions are equal), 4) identification, with a Monte-Carlo test, of the sampling sites for each sampling time, following the EI, which are significantly different than before the EI (hereinafter, outliers), 6) the augmented Dickey-Fuller test is used to determine whether there is a significant increasing trend in the percentage of outliers after the EI.

4. As a demonstration of the potential of the EIA algorithm, monthly climatic and quarterly agronomic data from sampling sites near the Sogamoso reservoir (Santander, Colombia) were used. Throughout the manuscript, the main strengths of the EIA algorithm are shown, which can be summarized as follows: 1) it allows a real, multivariate approach 2) it may be used with any kind of experimental design (Before-After, Before-After-Control-Impactor, Impact vs. Reference Sites analysis, and even designs with information from before the EI, if impact and control sites are available), 3) it overcomes the shortcomings of often-used statistical methods in EIA studies for identification of changes following an EI, and finally, 4) it allows for differentiation between natural and impact-induced variability on different temporal scales.