the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 08 Jul 2025
Offshore Crustal Thickness Variation along the Palu–Koro Strike–Slip Fault in the Sulawesi region from OBS Receiver Function Analysis
Abstract. The North Sulawesi Subduction Zone is one of the youngest active subduction systems in the western Pacific. In western Sulawesi, the Palu–Koro strike–slip fault connects with the westward-extending North Sulawesi Trench, forming a distinctive subduction–transform fault system. Understanding the crustal structure beneath the Celebes Sea and the geometry of the Palu–Koro fault is crucial for assessing regional deformation, rupture dynamics, and seismic hazards. In this study, we analyse data from nine ocean bottom seismometers (OBSs) deployed across the Palu–Koro fault using the receiver function H–κ stacking method to estimate crustal thickness. Our results reveal a shallow Moho (~8 km depth) beneath the Celebes Sea, in contrast to significantly greater depths (~25 km) beneath eastern Kalimantan and northern Sulawesi. Sharp variations in Moho depth near the Palu–Koro fault suggest the juxtaposition of two distinct crustal blocks. Combining S-wave velocity structures and local seismicity catalogue, we infer that the Palu–Koro fault is a left-lateral, supracrustal strike–slip fault extending into the Celebes Sea. These findings provide new geophysical constraints on the interplay between strike–slip faulting and subduction retreat, with implications for the generation of tsunamis by submarine earthquakes in this tectonically complex region.
- Preprint
(1521 KB) - Metadata XML
-
Supplement
(1566 KB) - BibTeX
- EndNote
Status: final response (author comments only)
-
RC1: 'Comment on egusphere-2025-3105', Anonymous Referee #1, 28 Jul 2025
This study shows the depths of Moho and crustal Vp/Vs around the Palu-Koro fault in the Sulawesi region, using receiver functions obtained from ocean bottom seismometer (OBS) data. The authors applied the H-kappa method to receiver functions to obtain those parameters, and also conduct nonlinear inversions to obtain one-dimensional Vs profiles beneath OBSs. In particular, the Moho depths obtained in this study show a large lateral variation, and the authors interpreted that this variation reflects different crustal blocks in this region. The results are interesting. However, I have concerns about the methods, including the calculation of receiver functions and the robustness of the inversion results. Moreover, it seems that there are several discrepancies between the results from the H-kappa method and inversions. These points should be clarified. I hope that my comments below contribute to improve this manuscript.
Major comments
–Methods
Did the authors constrain the Vs-profile inversions by the thickness and Vp/Vs obtained in the H-kappa method? Even using the NA approach, it is generally difficult to constrain trade-offs between the absolute Vs and thickness of each layer when using only receiver functions at a frequency range. To overcome this problem, seismologists often use joint inversions between dispersion curves of surface waves and receiver functions. In this case, using the thickness and Vp/Vs obtained in the H-kappa method may work to constrain such trade-offs. In cases that there are no constrains for the trade-offs, it is necessary to add synthetic tests showing that trade-offs between thickness and Vs of each layer are constrained well in the current inversions.
Moreover, in the inversion results, water layers are not displayed, but they should be displayed in the inversion results (Figs. 4 and S6–S11), and 0 km in the vertical axis corresponds to the sea surface. Did the author take into account the water layer in the forward calculation of receiver functions during the inversions? This should be taken into account, because multiple P reflections from the sea surface are contaminated in the vertical component, and Ps waves impinging to the seafloor are transmitted to the sea water as P waves. Such effects influence obtained receiver functions and eventually results.
–Results from H-kappa and inversions
There are several inconsistencies between H-kappa and inversion results. For example, in line 168 “Notably, the S-wave velocity …”, it is unclear which one corresponds to Moho steps. At C28F, the largest step is obtained around 10 km depths, but Table 2 shows a Moho depth at 25.10 km. At M03F, large steps can be seen at 4 km and 28 km depths, and Table 2 shows 24.80 km depth. In this case, the 28 km step may correspond to the Moho. However, in the H-kappa result in Fig. 3b, H = 16.28 km is obtained at this station (M03F).
Moreover, at C09G (Fig. S7), a gradual velocity change can be seen at 5–10 km depth, but H = 12.16 km. At C12F (Fig. S8), a gradual velocity change occurs at 4–10 km depth, but H = 11.60 km. At M02F (Fig. S11), a velocity step is identified around 30–32 km depth, but H = 23.97 km. There are several inconsistencies, and I wonder which ones are more reliable. I wonder such discrepancies are caused by trade-offs between the thickness and Vs of each layer in the inversions: trade-offs cannot be resolved in the inversions. More explanations on this point should be needed.
Other comments
Line 55
offshore off-shore -> offshore
Line 61
Remove “it offsets”.
Line 85
Figs. S4–S5 are referred before Fig. S3 (line 98). Please reorder the numbering.
Lines 90–99
More explanations for clock corrections, azimuthal orientations, and the processing are needed in the main text or Supplement. For example, how did the authors calculate the cross-correlation functions in Fig. S1? No bandpass filters? No suppressions on earthquake signals in the continuous records? It seems that Fig. S1a shows a clock deviation of ~200s. Please state this explicitly in the text.
Moreover, in Fig. S2, it seems that short-period components are used for teleseismic events. Although there are potentially some anisotropies at shallow depths beneath the sensors and topographical variations of the seafloor, which may produce energies in the transverse component, did the authors estimate azimuthal estimation errors using multiple teleseismic events coming from different azimuths? The robustness of the horizontal orientations of OBSs should be confirmed.
In lines 95–97, do the authors want to state about low qualities of teleseismic signals observed by OBSs, or want to state that some signals can be observed even in OBS data? According to this sentence, what the authors want to say is unclear. In line 97, although the authors show running spectra in Fig. S3, what do the authors indicate by showing this figure? I think that the authors applied demeans, detrends, rotations of horizontal components, and bandpass or high-pass filters. The running spectra at relatively high frequencies should show similar features between left and right panels. Did the authors use the same color palette for left and right panels?
The above points are examples that should be explained, and more explanations for this paragraph should be added. Also, in Fig. S2a, senor -> sensor.
Table 1.
How did the authors estimate Time_Err (s)? More details are needed. Also, please state the sampling rate of OBSs. Dt_Err (s) for each sampling point is listed, but the sampling rate is unclear.
Fig. 2b
Please define GCARC. Also, it is necessary to state in the main text or the caption of Figs. 1 and 2 whether land stations are broadband or not.
Line 123
Did the authors calculate receiver functions in the time or frequency domain? If they are calculated in the frequency domain, did the authors apply a water level method? If so, please explain more details for the receiver function estimations. Also, multiple P reflections from the sea surface should be contaminated in the vertical component. How did the authors treat this problem when calculating receiver functions?
Line 129–137
This can be moved to section 2, because this paragraph explain the method. Also, what Vp is used for H-kappa method should be noted. In this technique, Vp averaged over H (or crust) is assigned. Did the author refer to local or regional Vp tomographic studies, and change those values at individual stations?
Line 133
What is “a widIn”?
Line 136
Do the authors remove a sentence after “while the complete”? The results are shown in Fig. S12, and this is stated in line 137.
Line 129–137
It is necessary to explicitly state that the thickness obtained in this study includes layers of the marine sediment and igneous rock above the Moho, and Vp/Vs is a value averaged over the two layers.
L155
How did the authors treat Vp and density during iterations of inversions? In particular, are the Vp during the inversions consistent with the Vp assigned in the H-kappa method?
Line 162
What kind of parameter corresponds to Voronoi cells in this study? Please define it more details.
L168
Figs. S6–S11 are referred after Fig. S12 (line137). Please reorder the numbering.
Line 187
“Sulawesi”, “Borneo”, and “Makassar Strait” should be noted in Fig. 5.
Line 204
“represents subducting oceanic”
Please define the corresponding station. C08F?
Line 205
“The ~19 km Moho depth…”
The meaning of this sentence is unclear.
Line 207
Please specify the Sunda block and micro blocks associated with the Indo-Australian Plate in some maps.
Line 220
Please specify the locations of a Mw7.5 earthquake and Palu city in some maps.
Fig. 6
Again, the H-kappa results at M03F among Fig. 3, Table 2, and Fig. 6 are inconsistent. In Fig. 6a, please define the location of panel (b) by a rectangle.
Line 268
toward east -> toward west?
Line 269
“the wide-angle refraction study in the 90s”
Please add citations.
Fig. S5
What is “MODE”? Please define it.
Citation: https://doi.org/10.5194/egusphere-2025-3105-RC1 -
RC2: 'Comment on egusphere-2025-3105', Andrew Frederiksen, 11 Aug 2025
General comments:
This paper presents receiver-function results from a deployment of OBS instruments in the Celebes Sea. The experiment encountered serious difficulties, with only nine stations usable for this analysis out of twenty-seven deployed. Add to this the difficulties of receiver-function analysis under typical OBS constraints (high noise level and short deployment time), and the authors have done a truly remarkable job getting the most out of this challenging data set. The main results, a Moho step near the Palu-Koro Fault and a low-velocity upper crust near the fault trace, are interesting and look to be robust (the Moho depths match very well between methods, for instance), and the authors sensibly don't place too much weight on less robust results, like the Vp/Vs values retrieved from H-k stacking. As noted below, the writeup is a bit brief and could stand to be clearer about some methodological aspects; I would consider this a minor revision.
(Note that I have deliberately not looked at other comments on this manuscript before writing this; it's best for reviews to be independent.)
Specific comments:
1. Some details on the OBS processing are missing. For instance, did the OBSes have hydrophones? If so, were they used to correct the seismic data in any way?
2. For the receiver function deconvolution, what time-domain method was used? From context I would guess the iterdecon technique, but it's not stated explicitly.
3. In figure 3: what order are the RFs presented in? Chronological, or by epicentral distance, or something else? Also, it would be helpful to see an overlay of the expected arrival times for the best-fit H and k, to see what arrivals are being used.
4. The text on page 7 refers to an anticorrelation between H and surface topography, but this isn't shown directly. A plot might make this more convincing (free-air gravity could also optionally be included). Figure 6 does show this, but doesn't include all stations.
5. I assume that the traces were stacked before the RF waveform inversion, since a fit to only one trace is shown, but this isn't stated explicitly. Also, was a moveout applied before stacking? And in the inversion, were Vp and density held fixed? If so, where did the values come from?
6. In Figure 5, the H results look spatially coherent, but I don't think the k results do. The authors implicitly recognize this by not basing much interpretation on the k results, but it should be clearly stated in the text that a decision was made not to use them (this is not unusual for H-k results from noisy/limited data -- H is more robust).
7. The earthquake relocation work described on page 11 is a new result, so there should be a bit more detail on how it was obtained. What stations were used? Were the OBS data included?
Technical corrections: Please see my annotations to the manuscript and supplement.
-
EC1: 'Comment on egusphere-2025-3105', David Snyder, 20 Aug 2025
Both reviewers have raised substantive concerns about the consistency, quality and reliability of the combined velocity-depth modelling: the manuscript needs major, more detailed clarifications about the processing and analyses of the limited OBS data presented here. Reviewer 2 acknowledges the poor data quality, but Reviewer 1 would like to see more consistency in the methods' results. Carefully consider using one method's results as a starting condition/model for the application of the second method or provide a detailed explanation why such an approach would be inappropriate. Challenging conditions for OBS data acquisition is common, so many readers could benefit from "lessons learned" during your study.
Citation: https://doi.org/10.5194/egusphere-2025-3105-EC1
Viewed
| HTML | XML | Total | Supplement | BibTeX | EndNote | |
|---|---|---|---|---|---|---|
| 529 | 41 | 11 | 581 | 13 | 33 | 44 |
- HTML: 529
- PDF: 41
- XML: 11
- Total: 581
- Supplement: 13
- BibTeX: 33
- EndNote: 44
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1