A FUZZY KNOWLEDGE-BASED DECISION SUPPORT SYSTEM FOR GROUND WATER 
VULNERABILITY ASSESSMENT* 
E. Binaghi (°), I. Cerrani(°). P. Madella (°), M.G. Montesano(°), F. Pagnoni (§), 
A. Rampini(°) 
(°) ITIM, Istituto per le Tecnologie Multimediali - C.N.R. 
Via Ampère 56, 20131 Milano (Italy) 
(§) IRRS, Istituto per Ricerche sul Rischio Sismico Rep. Telerilevamento, CNR 
Via Ampère 56, 20131 Milano (Italy) 
ABSTRACT 
Ground water vulnerability assessment has been considered a multisource pattern recognition process 
and modeled within a fuzzy logic framework. 
The experimental study has been conducted on a rice area of the Po Basin in Northern Italy where the 
impact of agro-technology on the environment is pronounced. An advanced use of an integrated 
geographic information system (IGIS) dedicated to agricultural analysis and agro-technology impact 
evaluation is included. 
Qualitative and quantitative results demonstrate the adequacy of the proposed methodology. 
1.0 INTRODUCTION 
We have addressed ground water vulnerability assessment by modeling the classification of 
multisource data obtained from the layers of a dedicated GIS with an hybrid methodology in which 
spectral data are classified separately (usually with supervised statistical methods ) and approximate 
reasoning techniques are then applied to model a knowledge-based process that explicitly describes 
dependencies of spectral features, with other multisource and contextual features, and to represent the 
expert decision attitudes in performing a selection among a set of decision classes and assigning a soft 
gradual vulnerability judgement (McKeown. 1987) (Binaghi, 1993). 
Fuzzy production rules are used as knowledge structures that describe how a combination of features 
relates to the vulnerability classes (Zadeh, 1977). 
A multistrategy learning approach for the generation and refinement of fuzzy production rules is 
proposed using a unique inferential method of empirical induction performed by two computational 
mechanisms, of fuzzy reasoning and back-propagation neural networks (Rumelhart, 1986). * 
* Presented at the International Society for Photogrammetry and Remote Sensing “Workshop on New Developments 
in Geographic Information Systems. Milan. Italy, 6-8 March 1996. 
189