金曜日セミナー（4月27日） Prof. Chen Ji (University of California Santa Barbara)Images of the co-seismic fault slip and corresponding uncertainties are critical in understanding the physics of earthquakes. Because the independent constraints from geophysical surface observations, such as seismic and geodetic measurements, to the underground earthquake rupture processes are generally limited, various constraints are introduced to stabilize the inversions. The uncertainty of the inverted spatiotemporal slip distributions is then caused by both the data limitation and the imprints of a priori constraints, which are difficult (if not impossible) to separate. When the inverted solutions are used to study other geophysical subjects, the uncertainties in models lead to the uncertainties of the subject that we are interested in. Here we introduce a new strategy, named subject-oriented finite fault inversion, to explore the uncertainty of the ultimate subject due to the misfit to the data. Comparing with conventional approaches, the new method includes the value of the subject as a part of the objective function that will be minimized during the inversion. The uncertainty of the subject then can be directly accessed by inspecting the trade-off curve between its value and the misfit to the data. We apply this method to study the average stress drop (average (delta tau_E)) of two large earthquakes using seismic and geodetic data individually. We find only the low bound of the average stress drop (average_min (delta tau_E)) can be constrained with seismic and geodetic data. It is 5MPa for 2014 Mw 7.9 Rat Islands earthquake and 7MPa for 2015 Mw 7.8 Gorkha Nepal earthquake. The lower bound of average (delta tau_E) leads to the lower bound of the apparent available energy and the upper bound of the seismic radiation efficiency, though the latter is also sensitive to the measure of seismic radiated energy. The radiation efficiency of the 2015 Gorkha earthquake is only 0.09-0.15, suggesting the fast rupture velocity does not necessarily imply large seismic radiation efficiency.
In the end, we advocate to use average_min (delta tau_E) as a better replacement to smooth the fault slip during source inversions.