Factor weighting in DRASTIC modeling F.A.L. Pacheco a,b, ⁎, L.M.G.R. Pires c , R.M.B. Santos b,d , L.F. Sanches Fernandes c,d a Department of Geology, University of Trás-os-Montes and Alto Douro, Ap 1013, 5001-801 Vila Real, Portugal b Chemistry Research Centre, Vila Real, Portugal c Department of Civil Engineering, University of Trás-os-Montes and Alto Douro, Ap 1013, 5001-801 Vila Real, Portugal d Centre for the Research and Technology of Agro-Environment and Biological Sciences, Vila Real, Portugal HIGHLIGHTS • Compare techniques of factor weighting in DRASTIC modeling • Evaluate the impact of changing the weighting technique on the vulnerability index • Model the vulnerability of a large group of aquifer systems in continental Portugal abstract article info Article history: Received 18 May 2014 Received in revised form 28 September 2014 Accepted 28 September 2014 Available online xxxx Editor: Simon Pollard Keywords: DRASTIC aquifer vulnerability model Factor weighting Sensitivity Analysis Spearman rank-order correlation Logistic Regression Correspondence Analysis Evaluation of aquifer vulnerability comprehends the integration of very diverse data, including soil characteristics (texture), hydrologic settings (recharge), aquifer properties (hydraulic conductivity), environmental parameters (relief), and ground water quality (nitrate contamination). It is therefore a multi-geosphere problem to be handled by a multidisciplinary team. The DRASTIC model remains the most popular technique in use for aquifer vulnerability assessments. The algorithm calculates an intrinsic vulnerability index based on a weighted addition of seven factors. In many studies, the method is subject to adjustments, especially in the factor weights, to meet the particularities of the studied regions. However, adjustments made by different techniques may lead to markedly different vulnerabilities and hence to insecurity in the selection of an appropriate technique. This paper reports the comparison of 5 weighting techniques, an enterprise not attempted before. The studied area comprises 26 aquifer systems located in Portugal. The tested approaches include: the Delphi consensus (original DRASTIC, used as refer- ence), Sensitivity Analysis, Spearman correlations, Logistic Regression and Correspondence Analysis (used as adjustment techniques). In all cases but Sensitivity Analysis, adjustment techniques have privileged the factors representing soil characteristics, hydrologic settings, aquifer properties and environmental parameters, by leveling their weights to ≈4.4, and have subordinated the factors describing the aquifer media by downgrading their weights to ≈1.5. Logistic Regression predicts the highest and Sensitivity Analysis the lowest vulnerabilities. Overall, the vulnerability indices may be separated by a maximum value of 51 points. This represents an uncertainty of 2.5 vulnerability classes, because they are 20 points wide. Given this ambiguity, the selection of a weighting technique to integrate a vulnerability index may require additional expertise to be set up satisfactorily. Following a general criterion that weights must be proportional to the range of the ratings, Correspondence Analysis may be recom- mended as the best adjustment technique. © 2014 Elsevier B.V. All rights reserved. 1. Introduction The assessment of aquifer vulnerability is crucial for the protection of groundwater resources, namely against contamination with nitrate derived from agriculture. Contamination of aquifers with nitrate is a worldwide human health problem that has been reported in many studies addressing the topic from multiple standpoints (Manos et al., 2010; Pacheco, 1998a; Pacheco and Van der Weijden, 2012; Pacheco et al., 2013; Passuello et al., 2012; Sanches Fernandes and Haie, 2001; Valle Junior et al., 2014). Among the techniques in use for the evaluation and mapping of aquifer vulnerability, the most popular is the DRASTIC model (Aller et al., 1987) developed by a committee of the United States Environmental Protection Agency (EPA). The DRASTIC algorithm calculates an intrinsic vulnerability index (V) based on a weighted addi- tion of seven factors forming the DRASTIC acronym: V ¼ X p j¼1 w j X j ð1Þ Science of the Total Environment 505 (2015) 474–486 ⁎ Corresponding author at: Department of Geology, University of Trás-os-Montes and Alto Douro, Ap 1013, 5001-801 Vila Real, Portugal. Fax: +351 259 350480. E-mail address: fpacheco@utad.pt (F.A.L. Pacheco). http://dx.doi.org/10.1016/j.scitotenv.2014.09.092 0048-9697/© 2014 Elsevier B.V. All rights reserved. Contents lists available at ScienceDirect Science of the Total Environment journal homepage: www.elsevier.com/locate/scitotenv