the Creative Commons Attribution-NonCommercial 4.0 International License.
the Creative Commons Attribution-NonCommercial 4.0 International License.
A process-based framework for regional landslide debris inundation hazard assessment: application to the West Coast region of New Zealand
Abstract. Regional landslide debris inundation hazard assessments commonly rely on scenario-based formulations that require explicit enumeration of discrete landslide events. While effective for local studies, these approaches are computationally intensive, sensitive to scenario design choices, and difficult to apply consistently at regional scales where landslide catalogues are incomplete and long-term behaviour is poorly constrained. In parallel, geomorphological models provide insight into landslide-driven sediment production over long timescales, but are rarely formulated to generate spatially explicit hazard metrics relevant to runout and inundation.
Here we present the Landslide Source-to-Inundation Process Model (LS-IPM), a reduced-complexity, process-based framework that reframes landslide debris inundation hazard as a flux-based intensity problem. Rather than simulating discrete landslide scenarios, LS-IPM couples spatially distributed landslide sediment production to mass-conservative, topography-controlled debris routing, producing continuous fields of landslide-derived debris flux. Landslide magnitude and frequency are implicitly linked through a constraint on sediment production, allowing hazard to be evaluated everywhere in the landscape without explicit specification of individual failure events.
We apply the LS-IPM to the ~25,000 km² West Coast region of New Zealand at 25 m resolution, generating hazard estimates for present-day conditions, future climate scenarios (RCP6, 2081–2100), and a scenario Alpine Fault Mw 8.0 earthquake. Results show highest debris flux on steep slopes, at the base of high-relief landforms, and along convergent topography, consistent with geomorphological expectations. Approximately 80 % of the region is affected by some degree of landslide debris inundation hazard, with future climate conditions expected to increase mean annual debris flux by ~50 %.
By integrating geomorphological constraints with reduced-complexity runout modelling, the LS-IPM provides a scalable framework for regional landslide debris inundation hazard assessment that is compatible with subsequent probabilistic risk analysis.
Status: open (until 20 Oct 2026)
- RC1: 'Comment on egusphere-2026-4888', Anonymous Referee #1, 20 Sep 2026 reply
-
CC1: 'Comment on egusphere-2026-4888', Syed Azeem Inam, 20 Sep 2026
reply
1. Please demonstrate that the sediment production constraint yields identifiable frequency-magnitude relationships rather than multiple equally plausible event populations with identical debris flux.
2. Please justify the suspended sediment multiplier of 1.5 using a spatially explicit uncertainty model rather than agreement with selected regional denudation estimates.
3. Please propagate uncertainty jointly from erosion rate, bulking factor, density, susceptibility, volume distribution, source geometry, mobility, and movement mode into every reported hazard metric.
4. Please assess whether summing rainfall and earthquake probabilities is valid when triggers are non-independent and when repeated failures alter local susceptibility through time.
5. Please replace the binary kinematic feasibility criterion with a probabilistic formulation that represents uncertainty in source geometry, strength, depth, and spatially variable lithology.
6. Please test whether the assumed log-normal source volume distributions remain valid after inventory incompleteness, mapping resolution, censoring of small failures, and source runout separation are explicitly corrected.
7. Please conduct independent component-wise validation of source probability, mobilized volume, runout extent, deposition, and debris flux rather than validating final outputs using reconstructions based on the same GPP framework.
8. Please calibrate and validate the PCM parameters and the avalanche versus flow weights against geographically independent West Coast observations instead of transferring classifications from Cyclone Gabrielle.
9. Please demonstrate numerical and physical convergence across DEM resolution, routing iterations, volume class discretization, integration interval, smoothing scale, and sink filling settings before presenting regional hazard percentages.
Citation: https://doi.org/10.5194/egusphere-2026-4888-CC1 -
RC2: 'Comment on egusphere-2026-4888', Anonymous Referee #2, 21 Sep 2026
reply
Review of “A process-based framework for regional landslide debris inundation hazard assessment: application to the West Coast region of New Zealand”
This paper presents the LS-IPM, a reduced-complexity, flux-based framework for regional landslide debris inundation hazard assessment. It couples landslide source susceptibility, long-term sediment production, and downslope debris routing, avoiding the need to enumerate discrete landslide scenarios. The manuscript is well structured and beautifully written. The reviewer has no critical concerns but a few major/minor comments.
Figures 1 and 2. It would be helpful to label latitude and longitude on the maps, so that readers unfamiliar with the local geography can orient themselves more easily. The authors may consider adding a zoomed-out map of the region for easier reference.
Line 612. Just out of curiosity, will the future land use be considered?
Line 586. Should be kyr-1 not kyr.
The Abstract and Conclusions state that future climate conditions increase mean annual debris flux by ~50%, but Section 4.3 reports a ~50% increase in sediment production, not debris flux. These are not the same quantity: debris flux also depends on routing, deposition, and connectivity, and Section 4.3 explicitly describes the flux response as spatially non-uniform. The Abstract and Conclusions should be corrected to avoid conflating sediment production with debris flux.
Citation: https://doi.org/10.5194/egusphere-2026-4888-RC2 -
RC3: 'Comment on egusphere-2026-4888', Anonymous Referee #3, 22 Sep 2026
reply
This manuscript presents a valuable framework (LS-IPM) for regional hazard assessment based on a flux-driven representation of landslide processes. The conceptual approach is quite interesting and very well motivated, and the West Coast case study demonstrates its potential applicability. However, the manuscript is too lengthy and repetitive in several sections, and the highlight on methodological novelty is not fully matched by the level of validation presented. I suggest a major revision of the manuscript. Comments are presented below.
1. Manuscript structure
- The manuscript could benefit from significant streamlining and reducing redundancy. Several concepts, particularly the distinction between LS-IPM and scenario-based approaches, are repeated throughout Introduction, Discussion, and Conclusions.
- The Introduction is verbose and could be condensed by focusing more clearly on the knowledge gap, limitations of existing approaches, and the objectives of the study.
- Methodology section,especially sub-section 2.3.4, contains extensive mathematical formulation. A conceptual workflow figure presenting the relationships between LSS, LSP, and LSD would greatly improve readability.
2. Performance assessment and model validation
- The manuscript would benefit from a stronger and more explicit validation strategy. Currently, validation relies mainly on qualitative agreement with geomorphic expectations or comparisons with model-derived products, thus, where possible, additional comparisons against independent observations of landslide occurrence, inundation extent, sediment volumes, mapped deposits should be included to better illustrate model performance and predictive capability.
- The distinction between model validation, model evaluation, and plausibility checks should be clarified throughout the manuscript.
3. Assumptions, Uncertainty and Sensitivity
- Several key assumptions supporting the framework require stronger justification, including sediment-production constraint, adopted volume distributions, movement-mode weighting, and the treatment of future climate scenarios.
- The uncertainty related to these assumptions should be discussed more in detail, including how they could affect the resulting hazard estimates.
- A concise sensitivity analysis, or at least a summary of the sensitivity of model outputs to major parameters, would improve the robustness of the framework.
4. Existing literature
- Some statements regarding the novelty of the proposed framework and the limitations of existing approaches appear stronger than necessary. The authors may wish to moderate these claims and more clearly position LS-IPM as a complementary approach within the broader landscape of regional landslide hazard modelling.
- In the discussion, the distinction between methodological elements that are genuinely novel and those adapted from existing modelling frameworks should be made clearer.
5. Results and Interpretation
- The results section would benefit from additional quantitative analysis and synthesis. Currently, much of the discussion remains descriptive.
- More emphasis could be placed on quantifying the relative contributions of different triggers, landslide-volume classes, and topographic connectivity to the final hazard patterns.
- In the discussion section a more detailed consideration of transferability of the framework to other regions could be added, considering different geomorphic, climatic, and tectonic settings.
Citation: https://doi.org/10.5194/egusphere-2026-4888-RC3 -
RC4: 'Comment on egusphere-2026-4888', Anonymous Referee #4, 25 Sep 2026
reply
1. This is the point I keep coming back to. Figure 7 is presented as the main check on the model, but as the authors themselves note in Sect. 4.4, the fluxes reconstructed for the Murchison and Inangahua events are produced with the same GPP routing that generated the predictions. So the comparison says something about the magnitude of the debris budget coming out of the LSS and LSP components, and essentially nothing about whether the LSD routing puts the debris in the right places. For a paper whose central claim is that runout-informed hazard can be mapped continuously at regional scale, the absence of any direct spatial check against mapped inundation is a real gap. The sentence in the Conclusions about modelled flux extending to within one 25 m grid cell of mapped fan margins is exactly the kind of evidence that is needed, but it is reported in passing. I would like a short validation subsection, even for a handful of well-mapped fans or catchments, comparing predicted inundation extent to observed debris extents — hit rates, false-positive rates, something quantified. That would address the "not independent" criticism far more directly than a histogram comparison.
2. About the headline figure: "approximately 80 % of the region is affected" appears in the abstract, the introduction and the conclusions, but it is not clear what it means operationally. The flux legend in the maps starts at 0.0000 m³ m⁻² yr⁻¹, and at that threshold essentially any cell that has ever been crossed by debris counts as affected. A percentage computed that way is less a statement about hazard than about topographic connectivity, and it risks being read as "80 % of the West Coast is at risk." Please state the flux threshold behind the 80 % figure, and more usefully show an exceedance curve of affected area against flux threshold, so readers can see what fraction of the region actually falls into the high-flux classes. Without that, the number will be quoted far beyond what the model supports.
3. The results are more sensitive to the movement-mode treatment than the paper suggests, and the inputs behind it are the least constrained. The 66/33 avalanche-to-flow split for rainfall events is transferred from Cyclone Gabrielle on the North Island (Sect. 3.3.3), the 83/17 split for earthquakes is assumed, and the PCM parameters come from Swiss debris-flow work. Figure 7 shows that the movement mode changes the affected area from roughly 72 % to 80 % at the lowest thresholds and reshapes the upper tail — a first-order effect driven by inputs that have no local calibration and no sensitivity analysis behind them. I would like a one-at-a-time sensitivity run over the movement-mode weights, the PCM/fahrböschung parameters, and the production factors (ωL = 1, ηL = 1.3), reporting how the spatial pattern, the 80 % figure and the ~50 % climate change all respond. If the headline conclusions survive that, the paper is considerably stronger; if they do not, we need to know now rather than after publication.
4. One small thing on the literature. In Sect. 1.2 the argument that regional landslide assessments tend to stop at source occurrence or susceptibility, and rarely carry through to debris inundation, is well made but rests almost entirely on the New Zealand context. A couple of recent examples from outside that setting would make the point land harder — for instance work on extreme-rainfall and typhoon-induced landsliding that combines susceptibility mapping with early-warning thresholds (Xiao et al., 2026a, https://doi.org/10.1016/j.envsoft.2026.107034; Xiao et al., 2026b, https://doi.org/10.5194/nhess-26-611-2026), which is precisely the kind of assessment that stops at the source/warning stage.
Citation: https://doi.org/10.5194/egusphere-2026-4888-RC4
Viewed
Since the preprint corresponding to this journal article was posted outside of Copernicus Publications, the preprint-related metrics are limited to HTML views.
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 187 | 0 | 4 | 191 | 0 | 0 |
- HTML: 187
- PDF: 0
- XML: 4
- Total: 191
- BibTeX: 0
- EndNote: 0
Viewed (geographical distribution)
Since the preprint corresponding to this journal article was posted outside of Copernicus Publications, the preprint-related metrics are limited to HTML views.
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
The topic of this manuscript falls within the core scope of NHESS, namely natural hazard processes, regional-scale hazard assessment, and landslide/debris-flow hazards under climate change and earthquake scenarios. However, the current version is closer to a “methodological framework description + preliminary application” and is not yet sufficient as an independently reproducible, validated, and critically evaluable NHESS paper. I recommend a “very major revision” before re-evaluation. My main comments are as follows: