- Ground-motion model is (for in-slab):
_{0}= 100km, c_{31}= -1.7 (fixed, see below), R_{ref}= 1km, c_{1}= 4.229, c_{2}= 0.877, c_{32}= -0.00206, c_{41}= -0.481, c_{42}= -0.152, c_{51}= 0.425, c_{61}= 0.267, c_{62}= 0.491, σ = 0.351 (intra-event), τ = 0.112 (inter-event) and ϵ = 0.369 (total). - Ground-motion model is (for interface):
where Y is in cm∕s2, c

_{31}= -1.7 (fixed, see below), R_{ref}= 1km, c_{1}= 3.945, c_{2}= 0.974, c_{41}= -0.00172, c_{42}= -0.00099, c_{51}= 0.189, c_{52}= 0.707, σ = 0.330 (intra-event), τ = 0.257 (inter-event) and ϵ = 0.418 (total). - Use 3 site classes:
- S = 1, SS = 0
- NEHRP class C site (stiff soil)
- SS = 0, S = 1
- NEHRP class D site (soft soil)
- S = 0, SS = 0
- NEHRP class B site (rock)

- Use 2 earthquake classes:
- In-slab
- Often oblique-thrust mechanism with down-dip extension and arc-parallel compression. Focal depths
(h) between 67 and 163km. 4.4 ≤ M
_{w}≤ 6.7. Generally on inner Hellenic arc. 14 earthquakes. - Interface
- Often thrust mechanism. 47 ≤ h ≤ 66km. 4.6 ≤ M
_{w}≤ 6.4. Generally on outer Hellenic arc. 7 earthquakes.

Separation made based on Atkinson and Boore (2003) and source location relative to subducting slab.

- Consider 2 locations w.r.t. arc:
- ARC = 0
- Back-arc
- ARC = 1
- Along-arc

For h < 100km, define record as back-arc or along-arc based on whether the travel path from source to site predominantly passes through back-arc or along-arc area (complex classification). For h ≥ 100km classification is based on location of station, independently of r

_{hyp}. - Locations and magnitudes selected by comparing values from several national and international institutes.
- Use data from accelerometers and broadband instruments (both permanent and temporary) from various networks from 1994 to 2008.
- Check converted M
_{w}to measured M_{w}for those earthquakes with both and find good fit (mean difference of 0.04) with low scatter (standard deviation of 0.16). - Kythera earthquake (8/1/2006, M
_{w}6.7, h = 67km) contributes large portion of data. - Note lack of data for h > 100km and M
_{w}> 5.5. - Study effects of back-arc low-velocity/low-Q mantle wedge. Note challenge in expressing in functional forms the conceptual geotectonic and wave-propagation model.
- To determine final form of model firstly fit a simple functional form: log Y = c
_{1}+ c_{2}(M - 5.5) + c_{3}log R + c_{4}R + c_{51}S + c_{5}2SS. Classify data by focal depth: 60 ≤ g < 80km, 80 ≤ h < 100km and h ≥ 100km. Also classify records into back-arc and along-arc. Use c_{2}, c_{51}and c_{52}to adjust motions to M_{w}5.5 and rock and compute running averages for the different classes. Find large difference in back-arc along-arc motions for h ≥ 100km for short periods but not for long periods. For shallower depths the difference between back-arc and along-arc motions is noticeable at long r_{hyp}. These observations motivate the final functional form. Provide a schematic presentation of the different wave propagation paths depending on depth and location w.r.t. arc. - For in-slab events, the limited range of r
_{hyp}does not allow reliable estimation of geometrical spreading due to trade-off with anelastic-attenuation coefficient. Hence regress to find geometric coefficient after fixing anelastic attenuation and using data from two magnitude bins (to examine magnitude dependency). Use average value of -1.7 as a fixed coefficient in all regressions. Try similar technique for interface events but did not find reasonable value so use the same coefficient as for in-slab events. - Examine event terms as function of M
_{w}and h and intra-event residuals w.r.t. r_{hyp}and find no trends. - Compare observed (adjusted to M
_{w}5.5 and rock) and predicted motions for earthquakes with 60 ≤ h < 100km. Find sufficient fit. - Note that the large difference between back-arc and along-arc motions may not be usual for other subduction zones.