Displacement time histories from double‐integrated accelerograms typically cannot be used to recover near‐field terms because of noise in the acceleration traces. To minimize this problem, empirical mode decomposition (EMD) is used to derive a baseline correction scheme. The scheme is tested against several models composed of a single frequency or two frequencies and offsets in acceleration. It is verified against real displacement time history using Global Positioning Systems (GPS) measurement. For single‐ and double‐frequency‐content models, obvious discontinuities are found at drift times for lower intrinsic mode function (IMF) components. A drift model, however, can be clearly found from summing higher IMFs and the residuals of simple frequency‐content waveform models. On the other hand, results show that the lesser frequency‐content signal has the greater decomposed result. Therefore, a suitable corner frequency for low‐pass filtering is first implemented to reduce frequency content. In this case, the trend in the given drift model is easily found by summing higher IMFs for a complex frequency content model (model 3 of this study). A suitable corner frequency for low‐pass filtering is attained using a grid‐search method. A new semiautomatic EMD‐derived baseline correction scheme is tested. The corrected coseismic deformation value (CDV, denoted hereafter as from preevent displacement to the final offset during strong motion), peak ground displacement, and displacement time history from this method have good agreement with 1‐Hz continuous GPS measurement for the 2011 Tohoku earthquake.

You do not currently have access to this article.