the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 30 Mar 2026
Looking for Seismic Signatures in the Landscape: A Landslide-Based Record of Holocene Fault Ruptures in the Puget Lowland
Abstract. Earthquake-triggered landslides pose a major hazard in tectonically active regions and may leave lasting imprints in the landscape that reflect past seismic activity. In the Puget Lowland of Washington State, an urban corridor transected by multiple Holocene-active crustal faults, we investigate whether deep-seated landslides retain a record of prehistoric earthquakes. Using a regional inventory of more than 2,000 deep-seated landslides mapped from high-resolution lidar, we reconstructed a 4,000-year landslide chronology by relating deposit surface roughness to age, calibrated with16 radiocarbon-dated landslides. Temporal clustering of landslides was assessed by identifying peaks in landslide frequency that exceeded a steady-state landslide production model. Those peaks were then compared to earthquake-based scenarios incorporating known ruptures on the Tacoma Fault Zone (TFZ), Seattle Fault Zone (SFZ), Southern Whidbey Island Fault Zone (SWIFZ), and Darrington-Devil’s Mountain Fault Zone (DDMFZ). Our reconstructed landslide history reveals clustering of landslides 1000, 1250, 1900, and 2800–3200 years before present (ybp), coinciding with the timing of major Holocene earthquakes around 1000, 2000, and 3000 ybp. A multi-fault earthquake model reproduces these elevated periods more closely than a steady-state scenario. Frequency Ratio (FR) analyses show persistent fault-proximal landslide clustering, particularly along the SFZ and TFZ, where FR decreases with distance from the fault during time intervals containing well-constrained surface rupturing earthquakes. Interpretation of landslide clustering on the DDMFZ is complicated by geomorphic predisposition; on the other hand, weaker signals near SWIFZ likely reflect preservation bias within the landslide record. Nonetheless, temporal and spatial patterns show that Holocene crustal earthquakes generally leave detectable signals in the landslide record of the Puget Lowland. These results show the usefulness of roughness-calibrated landslide chronology as an independent paleoseismic indicator. Overall, this study underscores the value of integrating paleolandslides in reconstructing past earthquake activity and refining hazard assessment in landslide-prone, seismically active landscapes.
- Preprint
(1994 KB) - Metadata XML
-
Supplement
(939 KB) - BibTeX
- EndNote
Status: final response (author comments only)
- RC1: 'Comment on egusphere-2025-6555', Anonymous Referee #1, 22 Apr 2026
-
RC2: 'Comment on egusphere-2025-6555', Anonymous Referee #2, 15 May 2026
This is a review of “Looking for Seismic Signatures in the Landscape: A Landslide-based Record of Holocene Fault Ruptures in the Puget Lowland” by Ozioko and coauthors. This study modifies an available landslide inventory for the Puget Lowlands in Washington state and compares the temporal and spatial distribution of landslides with Holocene fault activity to assess if landslides could be coseismically triggered when compared to a steady-state landslide model. Landslide age was estimated using a surface-roughness model calibrated with limited radiocarbon ages. Fault activity was compiled from existing studies. The results indicate a broad temporal correlation with paleoearthquake timing and landslide-estimated ages, but some complications with spatial-temporal correlations of landslides for some of the fault zones. The conclusions will be of interest to a broad audience interested in cascading hazards, including landslide and earthquake hazards, in the Pacific Northwest and beyond. Overall, I found the study thoughtful with interesting implications, but the manuscript requires clarification for the reader to better understand the background, methods, and key interpretations. Below I outline several broader comments, followed by minor comments by line number and on the figures. I look forward to eventually seeing this work published.
Major comments:
- Consider adding more descriptions of the fault zones, including the fault lengths, orientations, fault kinematics, and clearly stating any information on how the most recent event or paleoearthquake history is known, as this is important background information for the reader to evaluate the robustness of the temporal correlations and possible ground motions. I note below that several key references per fault, in addition to earthquake timing, are not fully mentioned or considered. In addition, which fault trace is actually used in the spatial analysis, as many of the faults are multi-stranded or represent secondary faults above deeper folds? And what is the source of the fault linework?
- Throughout the manuscript, the authors use of Cal BP, cal BP, Cal YBP, and ybp to describes ages. I suggest choosing one and be consistent.
- The methods section may benefit from some example figures of landslides that are smooth and older vs. rougher and younger, to demonstrate this metric to the reader. Consider adding a figure with a few of the dated landslides and a figure of the landslides with shaded relief or roughness to illustrate this.
- The landslide database contains different types of mapped landslides and mass movements (L237-238). Are all of these landslide or mass movement types expected to result from ground-shaking? Consider adding a few sentences to discuss this limitation in the dataset if needed, and adding citations to support that these variable types of mass movement can result from coseismic shaking.
- The manuscript would really benefit from a broader discussion or comparison to other regions. The manuscript mentions some of these techniques have been used elsewhere – with the same or different success? Have these other studies been able to tie landslides to upper crustal earthquake histories? Expanding beyond a detailed discussion of these 4 fault zones in the Cascadia region may help broaden the interest of the manuscript to an international audience.
- The writing in the manuscript can be tidied up a bit, as it is repetitive and at times contradictory. Please check for repetitiveness between the background, results, and discussion. In addition, some paragraphs address multiple topics and the main message isn’t quite clear – I suggest editing carefully for clarity and focusing on flow and organizing relevant details in the same parts of the text. I give a few examples below but this list is not exhaustive.
- I struggled a little to follow the argument that the landslide age inventory agrees with the paleoearthquake histories but doesn’t always work out spatially – is a implication here that perhaps other (unknown or undated) faults may have similar paleoearthquake timing within uncertainty (as seen with the Seattle-Tacoma faults which ruptured at similar times) to explain this spatial disconnect? Or that other triggers could be responsible for the landslides? Along these lines, I think the manuscript would really benefit from a short discussion on the late Holocene climate as a possible triggering mechanism if not all landslides spatiotemporally align with faults that had earthquakes during a known time.
Minor comments:
L37-39: Consider adding the magnitude of these earthquakes.
L42: Define “shallow depths” here, as typical seismogenic crustal thicknesses are between 10-30 km or so.
L44: Define Mw at its first appearance in the manuscript.
L54: This example would benefit from more description to help the reader understand some of the differences between subduction zone and crustal fault earthquakes, such as the hypocentral depths of these earthquakes, whether or not they were onshore or offshore, etc. Also, some subduction events do create landslides of note – see Yin et al., 2026 and Grant and Collins 2025 for landslides during the 2025 M8.8 Kamchatka earthquake, but less than predicted by landslide models. Consider explaining some of these nuances in more detail.
L65-79: This paragraph would benefit from a better topic sentence or breaking it into 2 paragraphs.
L80-81: And perhaps more importantly, high-resolution satellite imagery, right?
L93: Consider editing to “timeline of prehistoric landslides”?
L93-98: What about cosmogenic exposure age dating, which has been successfully used to date some rockfalls and landslides? Consider adding a few sentences about this technique and how it fits into these issues with dating landslides.
L122: Consider changing “unobserved” to “prehistoric” or “Holocene”.
L140-141: This sentence is important information but feels out of place in a paragraph describing the geology. Perhaps this should go with the geochronology background/discussion?
L160: I suggest starting a new paragraph here as the focus shifts from the surficial geology and soil conditions to fault zones.
L167: What kind of fault is the Seattle Fault Zone? Did the entire fault zone rupture in this most recent event, or only one strand based on the available data (this may have implications for where ground-shaking occurred)? Also, this statement needs a citation (I think Bucknam et al., 1992 would be appropriate?)
L170-172: It may also be important to state that studies looking at the last 11 ka only determined 1 regional uplift event at 900-930 A.D. on the Seattle Fault (Sherrod et al., 2000; Davis et al., 2026), as this ties into the spatiotemporal analysis discussed later.
L171: Label Bainbridge Island to Figure 1.
L173: Consider also citing Kelsey et al., 2008, which also supports the uplift event on the Seattle fault zone ~1.1 ka.
L184: What type of fault is the Tacoma Fault Zone? How was the most recent event on this fault dated?
L194: What type of fault is the Southern Whidbey Island Fault zone? Note that Kelsey et al. (2004) only provides evidence of the 2800-3200 earthquakes and limited evidence of 1 or 2 earthquakes on a different strand – please add a citation for the statement that the fault has hosted 4 earthquakes in the Holocene. It looks like this may come from Sherrod et al. 2008, who note 4 earthquakes since ~16.4 ka, and that the most recent earthquake may be younger than 2.7 ka. Please consider these details in your analysis.
L198: Please add more details on the Darrington-Devils Mountain Fault zone, including type of fault, length, orientation, etc.
L216-218: Consider expanding this discussion, as it seems relevant. How does this study differ from the two mentioned here?
L252: Consider adding a citation to support this statement.
L337-344: It is great to discuss this limitation of the alternative triggering mechanisms, but it’s unclear how this was discussed or considered in the interpretation? Are there any paleoclimate datasets that may help determine periods of higher or more intense precipitation events in the past? Consider expanding this discussion.
L445-456: this paragraph switches from using ybp to cal BP? I suggest using one and being consistent.
L535: What is “rmsle”? Please define.
L545-547: This statement is hard to evaluate given the overlapping lines and confidence intervals in Figure 4. I suggest making two separate parts for this figure, one to illustrate the steady-state model vs. landslide inventory and the other for the earthquake model and landslide inventory.
L553-554: The timing may agree, but do the landslides with the appropriate surface roughness correspond spatially to these fault zones?
L568-574: I don’t quite follow this argument on the individual fault zones sometimes agreeing with the landslide inventories?
L688-690: This statement is a little misleading because the spatial-temporal model doesn’t align, so the landslides near these faults don’t have the “right timing” to be associated with the paleoearthquakes on these faults. Consider clarifying.
L737-739: I suggest adding more details for these USGS ShakeMap scenarios, such as the magnitude of the earthquake being modeled.
L768-770: I would urge caution at interpreting these landslides that occur hundreds of years after the earthquake as a post-seismic response, especially drawing to the comparison of post-seismic landslides on an annual to decadal scale. Has this been observed elsewhere? It seems that at these timescales following an earthquake, other triggering mechanisms could be plausible as well.
L773: If one is invoking high-intensity precipitation events, why is that not the trigger? Why must the trigger be an earthquake hundreds of years earlier?
L873-875: Consider adding some citations to support this statement.
Figure 1: Cite the source for the fault traces. Consider also showing their fault kinematics. Add a citation for the lidar dataset. Check the basemap carefully, we there are duplicate city labels in the background. Also consider adding a white box to outline the legend I the lower left, it’s hard to read some of the text because it’s overlapping the DEM and other words. Add a citation for the Hamma Hamma landslide, if possible. Consider also adding ages for the new landslides, this could be done by color coding the symbol for a specific age range. If possible, add an inset that shows this site in the larger context of the entire US west coast or North America, to provide a better reference for an international audience.
Figure 2: Add scale to field photographs (how long is the shovel?) Also consider adding latitude and longitude coordinates to the map.
Figure 3: this exponential relationship seems largely dependent on the 2014 Oso landslide data. Consider performing a sensitivity test to assess how much this one datapoint affects the modeled line.
Figure 4: these lines all overlap and it’s hard to tell when something goes above the steady-state model in the background. Consider making this figure 2 panels.
Table 1 was hard to read in the current PDF format for review, but I assume it will be reformatted in any final publication.
Citation: https://doi.org/10.5194/egusphere-2025-6555-RC2
Viewed
| HTML | XML | Total | Supplement | BibTeX | EndNote | |
|---|---|---|---|---|---|---|
| 425 | 159 | 30 | 614 | 62 | 21 | 25 |
- HTML: 425
- PDF: 159
- XML: 30
- Total: 614
- Supplement: 62
- BibTeX: 21
- EndNote: 25
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
The supplementary material demonstrates that the main clustering patterns persist across a range of bin widths. Given the importance of binning choices for peak identification, I suggest explicitly stating in the main text that the principal results are robust to reasonable changes in temporal binning, for instance 50 to 200 years, and that the interpretations do not rely on a single bin configuration.