Earthquakes are processes in which the internal workings (such as the accumulation of tectonic stress) are only observed indirectly, although the final effects are all too observable! Hidden Markov models (HMMs, a general statistical framework for modeling partially observed systems) are an intuitively attractive idea for analyzing seismicity. I will briefly introduce the idea of using HMMs to investigate earthquake cycles, and then focus on one case study incorporating GPS data into earthquake forecasting.
A new model we developed, the Markov-modulated Hawkes process with stepwise decay (MMHPSD), can capture the cyclic parent-generating- offspring feature of the temporal behavior of earthquakes. The decomposition of the earthquake cycle motivated the construction of a non-linear filter measuring short-term deformation rate-changes to extract signals from GPS data. This filter was applied to a) deep earthquakes in central North Island, New Zealand, and b) shallow earthquakes in Southern California. The study examines the use of HMMs to extract possible precursory information that indicates an elevated probability of large earthquakes occurrence. This study is controversial and still requires further tests. Japan is an ideal place to carry out this test.