GNGTS 2018 - 37° Convegno Nazionale

762 GNGTS 2018 S essione 3.3 than the maximum value stored in A , then P replaces the corresponding point in A , and, by iterating this procedure, the global minimum is found. To improve the rate of convergence and the computational costs of such an algorithm, a weighted rule for the selection of the trial point P and a quadratic approximation of the objective function based on the most promising points have been introduced by Brachetti et al. (1997). Finally, to avoid cases for which a positive quadratic form cannot be defined, a mutation probability is introduced for each of the N trial points, in such a way points with a fitness nearer to a given threshold, G lim , mutate with less probability, while points with fitness better than G lim do not mutate (Bresco et al. , 2005; Di Maio et al. , 2016). In the following, the algorithm has been applied by using the objective function: where N obs is the total number of measured data, V ˆ i and V j obs are the synthetic and observed SP values, respectively. Application to multiple SP anomalies observed in a contaminated site. The study site is an open waste pond used for olive oil mill wastes (OOMW) deposition located in the Keritis river basin near Alikianos village (Crete Island, Greece). The area has been well characterized by different geophysical methods among which ERT and SP surveys (Papadopoulos et al. , 2014; Soupios and Karaoulis, 2015; Simyrdanis et al. , 2016). In particular, the SP surveys were performed along a profile of length 33 m parallel to the river with the aim to identify the flow direction of contaminants from the waste disposal site to the Keritis river. The SPmeasurements were repeated 5 times after the OOMW deposition, i.e. during the spring-summer seasons when the wastes evaporate. As an example, Fig. 1 shows the SP data observed in June and July (dotted blue lines) and the GPA curves (red lines) obtained by considering a three-source model for both the surveys, defined by the forward function: where k is the electric dipole moment, x 0j are the horizontal positions of the source center axis, z 0j are the source depths, θ i are the polarization angles, and p is the shape factor that is equal to Fig. 1 - SP data (blue dots) and GPA curves (red lines) obtained for a three-source model. Green lines are interpolated data used for the inversion procedure.