What is the contribution of stress redistribution in earthquake swarm dynamics?

Earthquake swarms are generally interpreted as resulting from the redistribution of stresses within the crust. Swarms develop in response to fluid flow and poro-thermo-elastic stresses in reservoirs, aseismic slip on major faults, or during magmatic events in volcanic areas. However, our ability to quantify stress changes at depth from the observation of earthquake swarms is still very limited.  In his seminal study (Dieterich, 1994) was able to develop a model leading to a quantitative relationship between stress and seismicity rate. This model, based on non-interacting spring-and-slider systems undergoing rate-and-state friction was successful in determining stress conditions from seismicity rate in several active areas involving both tectonic and magmatic processes. This approach nevertheless relies on very strong assumptions, one of them being that no stress redistribution occurs following an earthquake. Stress redistributions are however known to drive earthquake sequences as observed during foreshock aftershock sequences. Ignoring this contribution might lead to wrong estimations of stress conditions at depth from seismicity rate.
In order to evaluate the role of stress redistribution in earthquake swarm dynamics, I present a new physics based earthquake simulator extending Dieterich's model. It consists of a set of planar rate-and-state frictional faults distributed in a 3D homogeneous elastic medium, and loaded by a prescribed stress history. Faults can have any size and orientation. Stress redistributions are thus fully accounted for.
The model is then used to investigate the relationship between seismicity rate and stressing history under different loading conditions (constant tectonic stressing, periodic loading) and fault properties (initial stress, frictional properties, relative distance between faults). In many cases, Dieterich's theory ignoring stress transfers captures many features of the seismicity rate patterns. This is particularly true for periodic loading, which generates frequency dependent seismicity modulation: at low frequency, seismicity rate scales exponentially with the loading stress, while at higher frequencies it tracks the stressing rate. The period separating the two modulation regimes is correctly predicted by Dieterich's theory. Under constant loading, seismicity rate is also constant (as predicted by Dieterich's theory) if the sequences are analysed over long enough time series involving several seismic cycles on each fault. At a shorter time scale however, significant clustering (not predicted by Dieterich's approach) arises, in particular for compact fault distributions enhancing the stress redistributions. 
More generally, the approach presented here allows to define the mechanical conditions leading to a significant contribution of stress transfers in the development of earthquake swarms.