At which temperatures do crustal earthquakes nucleate? Northwestern South America as a case study

Most crustal earthquakes occur between two depths: an upper boundary marked by the onset of seismicity and a lower one defined by the seismicity cut-off. The depth difference between them is the crustal seismogenic thickness (CST). As these boundaries are diffuse, they are usually determined from thresholds (percentiles) of the statistical distribution of earthquake hypocentral depths. This spatial earthquake distribution can be used as a proxy for the rheological conditions of the hosting rock, because earthquake generation is controlled by the mechanical rock properties, and in-situ temperature, pressure and strain rates.

Laboratory friction experiments with representative rocks and major rock-forming minerals indicate that earthquakes are expected to nucleate at <350±50°C in crustal rocks, and <700±100°C in ultramafic rocks typical of the mantle. However, the small spatial and temporal scales of these experiments hinder up-scaling their results to the geological conditions found in Nature.

In this contribution, we propose a solution for such upscaling: we use a 3D lithospheric-scale model of northwestern South America to compute the corresponding temperature field, and calculate the hypocentral temperatures of crustal earthquakes recorded in the region. The model is constrained by the integration of available geophysical data and 3D gravity modelling. For each layer, lithology-constrained thermal parameters (conductivity and radiogenic heat production) are assigned, either using direct samples when available, or from representative values. We use an S-wave tomography to set the temperature at 75 km depth as the lower boundary condition, and the measured average temperature at the Earth’s surface as the upper one. Furthermore, we use available measurements of heat flow and downhole temperatures to calibrate the model.

According to our results, most crustal earthquakes nucleated at <350°C, in agreement with laboratory experiments. The relatively few outliers are likely due to uncertainties in the Moho depths and/or in the earthquake hypocentral location. Also, they may be due to the presence of ultramafic rocks (which allow larger nucleation temperatures for seismicity) within the allochthonous crustal terranes accreted to this complex margin.

We map the depths of the upper and lower boundaries of the seismogenic crust using a spatial sampling procedure, defining them as the 10^th and 90^th percentiles of the hypocentral depths (D10 and D90, respectively). We find that D10, D90 and the resulting CST have significant spatial variations. Some of these correlate with crustal-scale faults which apparently separate crustal domains with different seismogenic behaviors.

Moreover, we point out that the two largest earthquakes recorded in the region (M_s = 7.3 and M_s = 6.8, of the Murindó sequence in 1992) nucleated at the lower boundary of the seismogenic crust, highlighting the importance of considering this lower boundary into account when characterizing seismogenic sources for hazard assessments.

Our approach could effectively bridge the scale gap between the laboratory rock friction experiments and Nature, as it enables to integrate the full geological complexity, including a realistic present-day lithospheric structure, the three-dimensional heat flow in the lithosphere, and the mantle temperatures imprint into the crustal thermal configuration.