Using self-organizing maps in airborne geophysical data for mapping mafic dyke swarms in NE Brazil

Mapping widespread dyke swarms shed light on the tectonic processes that culminate in continental fragmentation, especially in the early phases of the crustal extension and magma emplacement. Airborne magnetic data are an effective geophysical tool to clarify the extent of dyke swarms in continental-scale areas due to the expressive magnetic contrast between magmatic bodies and host rocks. However, many geological features display similar magnetic patterns, making the qualitative interpretation of magnetic anomaly maps quite subjective and ambiguous. To improve and optimize the predictive mapping of dyke swarms, this research presents a multivariate analysis of airborne geophysical surveys, applying a Self-Organizing Maps (SOM) approach using two magnetic and three gamma-spectrometric variables. This semiautomatic technique has allowed an integrated spatial analysis and the classification of the lithological units based on their magnetic and gamma-spectrometric signatures. We applied the SOM method to investigate a set of mafic dyke swarms that intruded in the Neoproterozoic Borborema Province (BP), Parnaiba Basin (PB) and the Sao Francisco Craton (SFC) in NE Brazil. These dykes are part of a large magmatic event associated with the continental breakup, which formed the Equatorial Atlantic Ocean in the Early Cretaceous, denominated EQUAMP. First, the SOM hyperparameters were defined by running the algorithm in a representative area in the central part of the BP. This training test area worked as a SOM template through which all data from the area were processed. The SOM analysis identified seven different populations, according to responses found in the five geophysical input variables. Two of these populations were associated with the mafic dykes, reducing the ambiguity of the magnetic anomaly interpretation. These populations represent high SOM quantization error, which means that these groups are the most anomalous data, evidenced in airborne magnetic data. These results were checked during fieldwork, revealing that dyke swarms occur more widely than were previously known, throughout the BP intruding the SFC, and showing some occurrences embedded in the Paleozoic sedimentary infill of the eastern border of the PB. (AU)