- Ground-motion model is:
_{0}= -2.376, c_{1}= 1.818, c_{2}= -0.1153, c_{3}= -1.752, σ_{intra}= 0.28, σ_{inter}= 0.24 and σ_{total}= 0.37. - Adjusts data to NEHRP B/C boundary (V
_{s,30}= 760m∕s) using site amplification term of Boore et al. (2013). Suggests that this site amplification term can be used to evaluate model for other site conditions. - Uses data from Next Generation Attenuation-West 2 with 3 ≤ M
_{w}≤ 6 and r_{hypo}≤ 40km that pass selection criteria of Boore et al. (2013). - Aim is a model for use in evaluating hazard from induced seismicity but use natural seismicity data because induced seismicity data lacking. Assumes that ground motions of the two types are similar.
- Little data from r
_{hypo}< 10km, which means that control on distance saturation is uncertain. Hence use h_{eff}from Yenier and Atkinson (2014) from stochastic modelling of global earthquakes with M_{w}≥ 6. Suggests that h_{eff}= 1km (corresponding to M_{w}4) is the lower limit that should be used. Tries alternative functions for h_{eff}and note that they fit the data almost as well but lead to different very near source predictions. This implies an epistemic uncertainty of up to a factor of two in the near-source area. - Chooses functional form as simplest likely to be applicable.
- Finds that standard least-squares regression gives similar results.
- Plots predictions against observations grouped into 0.5M
_{w}width bins. Finds good fit. - Plots residuals w.r.t. r
_{hypo}and finds no trends, which means that a more complex functional form is not required. - Note that model is strictly applicable only to about 40km because of lack of anelastic attenuation term.
Proposes adding a additional term c
_{4}R with c_{4}calibrated using the behaviour of the Boore et al. (2013) model beyond 40km. For PGA c_{4}= -0.002 is proposed.