Insights into tectonic zonation models from the clustering analysis of seismicity in southern and south-eastern Spain

<p>The southern and south-eastern parts of Spain exhibit the highest seismicity rate in the country. However, although the recently developed Quaternary Active Fault database of Iberia (QAFI, <span class="cit" id="xref_altparen.1"><a href="#bib1.bibx8">García-Mayordomo et al.</a>, <a href="#bib1.bibx8">2012</a></span>) collected the available information existing in the study area regarding fault data for their use in seismic hazard applications, this information is of limited use since data quality is very heterogeneous: few earthquakes are associated with specific fault segments, and occurrence time periods (when indicated) are affected by high uncertainties <span class="cit" id="xref_paren.2">(<a href="#bib1.bibx10">Gaspar-Escribano et al.</a>, <a href="#bib1.bibx10">2015</a>)</span>. This has motivated the definition of alternative tectonic zonation models, to be used for evaluating the seismic hazard. So far, the clustering properties have not been considered in this regard, though they can provide essential information about the features of seismic energy release, depending on the tectonic style of a region <span class="cit" id="xref_paren.3">(<a href="#bib1.bibx32">Talebi et al.</a>, <a href="#bib1.bibx32">2024</a>)</span>. This is why in this work the properties of the seismicity in terms of clustering are evaluated by applying the nearest-neighbour (NN) algorithm to the south-eastern Spain region. The scale parameters needed for the NN algorithm are optimised through the study of the <span class="inline-formula"><i>z</i></span> score and the temporal anomalies between events in the identified clusters for each run. The tree structure, using graph theory notation, has proved useful in the determination of the critical threshold that separates the background (independent) seismicity from the clustered (dependent) seismicity in the NN algorithm. Once the clusters have been identified, their properties have been quantified in terms of a selection of complexity measures: outdegree, closeness and average node depth. This procedure has been applied by considering two different completeness magnitudes: <span class="inline-formula"><i>M</i>_w</span> 3.0 (the mean completeness magnitude for the entire catalogue) and <span class="inline-formula"><i>M</i>_w</span> 2.1 (accounting for the most recent part of the catalogue). The results are similar in terms of proportion of foreshocks, mainshocks and aftershocks, and indicate a clear distinction between the western-most part (higher complexity) and eastern-most part (lower complexity). To check this result, three different zonation models have been examined and cross-compared; two of them passed the Kolmogorov–Smirnov (KS) test, meaning the distributions of the selected complexity measures are not the same for the different zones defined in the models. These zonations can be used in order to assess the seismic hazard, as they account for the influence of the tectonic setting on the patterns of earthquake occurrence, including the features of background and clustered seismicity components.</p>