GNGTS 2019 - Atti del 38° Convegno Nazionale

554 GNGTS 2019 S essione 3.1 of homogeneous sources (e.g. Spector and Grant, 1970; Fedi et al. , 1997), or random and uncorrelated sources (e.g. Blakely, 1995) or fractal source distributions (e.g. Maus et al. , 1997; Pilkington and Todoeschuck, 1993; Bansal et al. , 2011). Instead, the depth to the top of intermediate depth sources and the depth to the bottom, including the Curie isothermal surface, are the most difficult tasks. Nonetheless, different techniques have been developed to estimate the depth to the bottom of the source bodies. The spectral peak method and the centroid method (e.g. Battacharyya and Leu, 1975; Tanaka et al. , 1999) are the most commonly used techniques to estimating the depth to the bottom assuming a random source distribution. More recently, different methods have been developed to estimate the depth to the bottom of sources assuming a fractal model (Fig. 1b): nonlinear inversion (Maus et al. , 1997; Bouligand et al. , 2009), and the modified centroid method (Bansal et al. , 2011). Spectral analysis of potential field data may be used to analyze data in the form of maps or profiles. However, spectral analysis in the form of maps is the most common technique, which is performed by subdividing the whole map into different overlapping windows, whose size depends on the map dimension and on the target of study. Several authors have suggested the optimal size of the window (e.g. Blakely, 1995; Maus et al. , 1997; Bouligand et al. , 2009) allowing a consistent characterization of the source depth. Even though a lot has been done to understand the optimum window size for depth estimation, there is still no agreement on the minimum extent of the survey area required to get a reliable depth estimate. Fig. 1 - Synthetic magnetic anomaly produced by complex source ensembles (a) and an example of 3D fractal magnetization distribution model (b). In this work, we review spectral methods for depth estimation owing to different source models (i.e. statistical ensemble sources, random and uncorrelated sources, and fractal/scaling sources) and discuss practical constraints on the depth estimation and inherent assumptions/ limitations of the different approaches. To the sake of usefulness, we decided to organize this work not according to the different methods but in relation to the main different tasks shared by all the methods. So, the main section will regard: a) depth to the top estimation; b) depth to the bottom estimation and Curie depth; c) selection of the window size and wavenumber range. Assuming a statistical ensemble of homogeneous sources (Fig. 1a), a new technique is also proposed to estimate the source depth to the bottom by solving for the thickness factor. Using synthetic data, we also discussed the effect of window size on depth estimation assuming different source models. We show that the window size depends not only on the complexity of geology but also on the type of method utilized. In fact, despite the several different approaches all the methods give quite consistent and often similar estimates of the source depths. However, due to ambiguities on the correction spectral factor, the best estimates are obtained if this factor is constrained by a priori information. Acknowledgements. This project is funded by Eni as part of the Eni 2017 award: Debut in Research-Young Talents from Africa Prize .