A MULTIRESOLUTION INVERSION OF ONEDIMENSIONAL MAGNETOTELLURIC DATA

For magnetotelluric (MT) inversion, when initial guesses are far from the global minimum, the results are often trapped by some local minimums by using linear inversion methods. This moviated us to develop a new method, called multiresolution inversion (MI) method, to overeome the troubles mentioned above. The basic idea of MI is tha the original MT inverse problem can be decomposed into a set of inverse problems at different scales by multiresoluhon analysis (MRA), then the inverse problem of the coarest scale is solved firshy and its final result is implemented as an initial model of the inverse problem of the second coarest scale. This process is repeated unhl the inverse problem of the finest scale is solved. Two numericalexamples of ID MT theorehcal model and a real MT profile interpretation show the algorithm is stable, less dependent on the initial guesses, and can improve the inverseresolutions.