Importance of Considering Near‐Surface Attenuation in Earthquake Source Parameter Estimation: Insights from Kappa at a Dense Array in Oklahoma

Separating earthquake source spectra from propagation effects is challenging. The propagation effect contains a site‐dependent term related to the high attenuation of shallow sediments. Neglecting the site‐dependent attenuation can cause large biases and scattering in the corner‐frequency (⁠$fc$⁠) estimates, resulting in significant stress‐drop deviations. In this study, we investigate shallow attenuation at the LArge‐n Seismic Survey in Oklahoma (LASSO) and site‐related biases and scattering in source parameter measurements due to simplified attenuation models. We measure the high‐frequency spectral decay parameter kappa on the vertical acceleration spectra of regional earthquakes (125 km away). The site‐dependent kappa (⁠$κ0,acc$⁠) suggests that attenuation increases rapidly at shallow depth and is highly site‐dependent. 10%–75% of the attenuation is site‐dependent for S waves and even larger for P waves. The quality factor for S waves (⁠$QS$⁠) ranges from 10 to 100 in the upper 400 m. $QP$ for P waves is mostly below 10 within the same depth. The Quaternary sediments tend to be more attenuating (⁠$QS<30$⁠), but the Permian rocks also can have high attenuation. We demonstrate that using a non‐site‐dependent attenuation model in single‐spectra fitting leads to large scattering in $fc$ estimates among stations with apparent good fits. The apparent $fc$ can significantly deviate when the range of site‐dependent kappa is large or with a higher assumed source spectral fall‐off rate n. The biases in apparent $fc$ depend on site condition and distance; however, the correlation between $fc$ and these factors might not be obvious, depending on model assumptions. An apparent increase of stress drop with magnitude in a previous study for local microearthquakes (⁠$1.3<Mw<3.0$⁠) can be largely negated by including a site term, restricting to sites on the higher Q formations, and fixing n. This study highlights the importance of considering near‐surface attenuation when modeling source parameters.