Path effects that are unique to the source-site geometry increase the uncertainties in ground-motion prediction equations (GMPEs). A methodology is proposed to deal with this issue. First a GMPE is developed, including site and event terms. A map of the logarithmic residuals may show that there are regions in which the spatial average is different from zero. The smoothed residuals become an additional correction to the GMPE. In principle, the contour maps of these corrections are unique to each strong-motion station, but maps of nearby stations should be correlated. The potential for path corrections to significantly improve ground-motion prediction at individual stations is tested first in a simplified pilot application on a small subset of Japanese peak-acceleration and peak-velocity data. For these data, the standard deviations to GMPEs without station or path corrections are in the range of 0.76 to 0.85 (in natural-log units). With the corrections the standard deviations are reduced to a range from 0.25 to 0.35. The method is also applied more strictly to the prediction of peak acceleration and peak velocity in Guerrero, Mexico. The result is reduction of the estimated standard deviation from 0.70 to 0.35 for peak acceleration and from 0.64 to 0.40 for peak velocity. After accounting for the path terms, the standard deviation at individual stations (single-station sigma) is estimated to be 0.51 and 0.48 for peak acceleration and peak velocity, respectively.
This approach may be particularly appropriate for the design of critical structures with long lead times in their planning and design. If broadband and strong-motion stations are installed at the intended site immediately after site selection, site-specific corrections determined from background seismicity during the design stages can be used to refine and defend the design during the approval process.