Using a hybrid Monte Carlo/Genetic Algorithm Slip Estimator to produce high resolution models of paleoearthquakes from geodetic data

Identifying fault sections where slip deficits have accumulated may provide a means for understanding sequences of large megathrust earthquakes. Stress accumulated during the interseismic period on locked sections of an active fault is stored as potential slip. Where this potential slip remains unreleased during earthquakes, a slip deficit can be said to have accrued. Analysis of the spatial distribution of slip during antecedent events along the fault will show where the locked plate has spent its stored slip and indicate where the potential for large events remains. The location of recent earthquakes and their distribution of slip can be estimated instrumentally. To develop the idea of long-term slip-deficit modelling it is necessary to constrain the size and distribution of slip for pre-instrumental events dating back hundreds of years covering more than one ';seismic cycle'. This requires the exploitation of proxy sources of data. Coral microatolls, growing in the intertidal zone of the outer island arc of the Sunda trench, present the possibility of producing high resolution reconstructions of slip for a number of pre-instrumental earthquakes. Their growth is influenced by tectonic flexing of the continental plate beneath them allows them to act as long term geodetic recorders. However, the sparse distribution of data available using coral geodesy results in a under determined problem with non-unique solutions. Instead of producing one definite model satisfying the observed corals displacements, a Monte Carlo Slip Estimator based on a Genetic Algorithm (MCSE-GA) accelerating the rate of convergence is used to identify a suite of models consistent with the data. Successive iterations of the MCSE-GA sample different displacements at each coral location, from within the spread of associated uncertainties, producing a catalog of models from the full range of possibilities. The suite of best slip distributions are weighted according to their fitness and stacked to produce a final estimate of the distribution of slip for a particular earthquake. Examination of the slip values in the stacked models allows areas of high confidence to be identified where the standard deviation is low. Similarly, areas of low confidence will be found where standard deviations are high. These high resolution models can be used to reconstruct a history of slip along the fault, both identifying and quantifying of slip deficits and constraining confidence in the accuracy of the modelled information. This presentation will demonstrate the ability of the MCSE-GA to produce accurate models of slip for instrumentally recorded earthquakes and show estimates for slip during paleoearthquakes along the Sunda Megathrust.