- Ground-motion model is:
where Y is in cm∕s2, a = 0.6241, b = -0.001623, c = 0.01134646, d = 0.1694, SB = -0.5930,
SC = -0.1206, SD = 0.0830, SE = 0.1597 and σ = 0.489.
- Use 4 NEHRP site classes:
- 760 < V s,30 ≤ 1500m∕s. 11 Malaysian stations. Use SB.
- 360 < V s,30 ≤ 760m∕s. Use SC.
- 180 < V s,30 ≤ 360m∕s. Use SD.
- V s,30 < 180m∕s. Use SE.
For Japanese stations based on measurements, for Malaysian stations based on geological descriptions and
for Iranian stations, because of lack of information, based on topographic slope using approach of Wald
and Allen (2007) (note uncertainty in these classifications).
- Focal depths between 51 and 305km with 11 out of 13 earthquakes ≤ 105km (contributing all but 5
- Purpose of model is to assess seismic hazard in Peninsular Malaysia from distant intraslab subduction
- Use data from Peninsular Malaysia (34 records), Java (1 record), Japan (474 records) and Saravan-Iran (22
records) because limited data from Malaysia and since no detectable differences between ground motions
in different subduction zones. Based on previous studies believe that attenuation rate in Malaysia and
Japan is similar.
- Classify earthquakes as intraslab based on their locations and mechanisms using approach of Atkinson and
Boore (2003): if depth < 50km and extensional then intraslab; reverse-faulting earthquakes are classified
as intraslab if depth > 50km or if on steeply-dipping planes.
- Use locations, Mw and mechanisms from Global Centroid Moment Tensor.
- Data reasonably well distributed w.r.t. Mw, rhypo and site class.
- Data from Malaysian Meteorological Department (note that some doubts about accuracy of these records),
K-Net, KiK-Net and Building and Housing Research Center.
- Records baseline adjusted and acausal filtered using 4th-order Butterworth filters after zero padding at
both ends. For late-triggered records taper using a raised cosine to first and last 5% of record.
- Decide on functional form using trial and error.
- Plot residuals w.r.t. rhypo grouped into classes of 0.5Mw wide. Find no apparent trends.
- Plot histograms of residuals and find good match with normal distribution.
- Compare predictions and observations (including additional PGAs from Singapore) for various magnitude
ranges and focal depths w.r.t. rhypo. Find good match.
- Note that more data from Peninsular Malaysia would make the model more reliable.
- Note that the maximum-likelihood or two-step approaches should be used to resolve related to data
distribution and to evaluate the different components of variability.