The multi-scale structure of the Earth’s interior contains important information on plate tectonics and geological processes, making the development of reliable models a critical objective in seismological research. Compared to conventional travel time tomography based on high-frequency ray theory and limited seismic phase arrival time information, waveform inversion utilizes the dynamical properties in the waveforms such as the frequency-dependent phases and amplitudes to constrain model parameters, thus representing the most promising approach for constructing high-resolution structural models. Waveform inversion is inherently nonlinear, thus blind use of all records without proper screening during iterative inversion leads to difficulties in convergence or getting trapped into local minima. Therefore, one of the key factors for successful waveform inversions is the selection of time windows of signals to constrain the intended structural targets. Proper selection of time windows is crucial for enhancing the utilization of effective information, reducing noise interference, and improving the accuracy and stability of inversions.

To efficiently and conveniently extract effective waveform information, we develop the SeismoClip, an open-source waveform data automatic selection program. SeismoClip is a MATLAB-based interactive tool, designed for waveform inversion. It can be used not only as an independent window selector but also as a post-processing tool for other automatic window selection programs. Various auxiliary functions, such as array analysis, multi-frequency-band filtering, and short-time-average/long-time-average （STA/LTA） analysis, are also employed to significantly improve the accuracy and robustness of window selection performance. Furthermore, SeismoClip emphasizes user experience with extensive customizability, allowing researchers to make modifications to adapt program functions according to specific research needs. The core function of SeismoClip is to interactively process and select time windows from the input synthetic and observed seismic waveforms, and generate the corresponding window files. The processing flow is mainly divided into four steps: data processing, array analysis, interactive window selection, and automated adjustment.

In recent years, the widespread deployment of seismic stations has led to the availability of huge volumes of waveform records from densely distributed seismic arrays. The waveform similarity of the same seismic phase between stations and the variation of phase arrival times with epicentral distance can help us determine a reasonable interval for the seismic phase window. Therefore, we have added an array-based module to the SeismoClip to assist in the analysis and extraction of phase windows. Since recorded waveforms are affected by noise and thus have a lower signal-to-noise ratios compared to theoretical seismograms, phase information in records can easily be obscured by noise. Therefore, SeismoClip performs array analysis based on the STA/LTA curve calculated from noise-free synthetic seismograms generated from the same source parameters.

During the interactive selection process, SeismoClip displays both observed and synthetic waveforms in array form for reference. The interactive selection process is applied to single-channel waveforms. SeismoClip offers a variety of interactive selection interface options to accommodate different user needs, a including multi-frequency-band filtering single-channel selection interface, an interface highlighting weak-amplitude body-wave phases, and an interface based on existing selection results. This allows users to precisely extract seismic phase windows based on waveform data. For the interactively selected windows, SeismoClip performs a series of automated adjustments under the constraints of existing window selections to achieve a better fit. Users can choose either directly use the results of the interactive selection or allow the program to make automatic adjustments.

In this article, we compare the selection results of SeismoClip with those from the automated time-window selection algorithm FLEXWIN, and discuss the distribution of sensitivity kernels corresponding to different time-windows to highlight the importance of precise time-window selection. Comparative analysis with real data shows that the time windows selected by SeismoClip are generally more accurate and reasonable than those selected by automated selection, particularly in selecting windows from waveforms with low signal-to-noise ratios.

Overall, SeismoClip represents as a convenient and effective time-window selection tool for waveform inversion, based on array analysis of specific seismic phase tracking and multi-frequency-band analysis. Moreover, by real-time evaluation of the similarity and cross-correlation travel time differences between the observed and synthetic waveforms within the selected time window, it is possible to effectively avoid cycle skipping in waveform inversions. In cases where waveform features are not prominent, SeismoClip can be used to carefully select the time window to obtain appropriate sensitivity kernels and enhance the stability and accuracy of the inversions. For users utilizing an automated time-selection algorithm, we suggest starting with less restrictive parameters to conduct an initial window selection, and then utilizing the“interactive selection interface based on the eaisting selection results”of SeismoClip to review and refine the initial selection results.