Microseismic icequake catalogue, Rutford Ice Stream (West Antarctica), November 2018 to February 2019

Event locations were gained from the continuous passive seismic records using the QuakeMigrate (Hudson et al., 2019; J. D. Smith et al., 2020) and Nonlinloc (Lomax et al., 2000) software. A two-layer velocity model (firn layer above an ice layer) according to E. C. Smith et al. (2015) was used in the Nonlinloc location step. Based on Nonlinloc locations, we accept only events with a total root-mean-square (RMS) value of 0.02 s, a maximum azimuthal gap of 280 degrees, maximum 10% of picks with a P/S residual larger than 0.02 s/0.2 s and at least 3 P picks and 2 S picks for the final event catalogue. We account for the movement of the seismic stations relative to the bed due to ice flow by shifting each event in the final catalogue downstream. Rutford Ice Stream moved ~94 m downstream during the 90 day survey period, whereas our stations are specified at fixed locations during event location, clearly evidencing the necessity for such a shift. We perform this shift by calculating the stations' locations at the time of each individual event relative to the start of the deployment, using their GPS locations, and apply this lateral shift to that event hypocenter. This is repeated for all events in the catalogue to compensate for ice flow. Magnitudes were calculated from the far-field displacement of the P-wave (Mw; Shearer (2009); implementation of Hudson (2019)) for a 1-day data subset. A local magnitude scale (ML) was then derived based on these Mw values. The local magnitude scale follows the equation: ML = log10(A) + m x depi - t where A is the maximum amplitude of either of the two horizontal components (in instrument counts corrected for the instrument type). The distance term m accounts for the decay of amplitudes with increasing epicentral distance, depi is the epicentral distance and t is a scaling parameter that bridges the offset between Mw and ML. Focal mechanisms were calculated from first motion polarities and P to S- amplitude ratios using the HASH software (Hardebeck & Shearer, 2002, 2003). The final set of good solutions is derived based on quality criteria, which are the stability of solution upon variations of input, the azimuthal gap of the final set of stations (should be smaller than 180 degrees) used and the final number of input picks (should be larger than seven). Detailed information on catalogue creation will be contained in Kufner et al. (in prep.).

本数据集的地震事件位置通过QuakeMigrate（Hudson等人，2019；J. D. Smith等人，2020）与Nonlinloc（Lomax等人，2000）两款软件，基于连续被动地震记录提取得到。在Nonlinloc定位步骤中，我们采用了E. C. Smith等人（2015）提出的双层速度模型：表层为粒雪层（firn layer），下层为冰层。基于Nonlinloc的定位结果，我们对最终事件目录设置了严格筛选标准：仅保留均方根（root-mean-square, RMS）总值不超过0.02秒、最大方位角间隙不超过280度、P/S波波至拾取残差大于0.02秒/0.2秒的拾取占比不超过10%，且至少包含3个P波波至拾取与2个S波波至拾取的地震事件。考虑到冰流作用下地震台站相对于基岩的位移，我们将最终目录中的所有事件沿下游方向进行偏移校正。在为期90天的观测周期内，拉特福德冰流（Rutford Ice Stream）下游移动了约94米；而本研究在事件定位阶段将台站设定为固定位置，这充分证明了上述偏移校正的必要性。本次偏移校正的实现方式为：利用台站的GPS定位数据，计算每个事件发生时刻台站相对于部署起始时刻的位置，并将该横向偏移量应用至对应事件的震源位置。我们对目录内所有事件重复该流程，以抵消冰流带来的位置偏差。我们针对1天的数据子集，通过P波远场位移计算得到矩震级（Mw；参考Shearer，2009；Hudson，2019的实现方法）。基于上述矩震级结果，我们推导得到了区域震级（ML）标度关系，其表达式为：ML = log₁₀(A) + m × depi - t。其中，A为两个水平分量中任意一个的最大振幅（单位为经仪器类型校正后的仪器计数），距离项m用于修正振幅随震中距增大的衰减效应，depi为震中距，t为用于校正矩震级与区域震级之间偏移的标度参数。我们利用HASH软件（Hardebeck与Shearer，2002、2003），通过初动极性与P/S波振幅比计算得到震源机制解。最终合格的震源机制解集合基于以下质量标准筛选得到：解在输入参数变化下的稳定性、所用台站的最大方位角间隙需小于180度，以及有效拾取的总数量需大于7个。关于事件目录构建的详细信息将在Kufner等人（待刊）的研究中予以阐述。