Application of the ACE algorithm on hydrogeological monitoring data from the Bükk Mountains

Abstract

This paper introduces the alternating conditional expectation (ACE) algorithm of Breiman and Friedman (1985) in multiple regression problems in groundwater mon- itoring data analysis. This special inverse nonparametric approach can be applied easily for estimating the optimal transformations of different groundwater monitor- ing data from the Bukk Mountains to obtain maximum correlation between ob- served aquifer variables. The approach does not require ap rioriassumptions of a mathematical form, and the optimal transformations are derived solely based on the groundwater data set. The advantages and applicability of the proposed approach to solve different multiple regression problems in hydrogeology or in groundwater man- agement are illustrated by means of case studies from a Hungarian karst aquifer. It is demonstrated that the ACE method has certain advantages in some fitting problems of groundwater science over the traditional multiple regression. In the past, different groundwater monitoring data (like groundwater level, ground- water temperature and conductance, etc.) had been used for groundwater manage- ment purposes in the Bukk Mountains. One of the difficulties in earlier approaches has been the need to make some kind of assumption of the expected mathematical forms among the investigated reservoir and petrophysical variables. By using non- parametric regression, the need to assume a specific form of model is avoided, and a clearer vision of the relationships between aquifer parameters can be revealed in the Bukk Mountains, where karst water is the main source of potable water sup- ply. Complex monitoring data from the Bukk Mountains were analyzed using the ACE inverse method, and results were verified successfully against quantitative and qualitative field observations.