This study addressed the interesting question of how the timing and spatial pattern of earthflows relates to salmon habitat in Washington, USA.  Spatially, the authors found that wide valleys tend to be present upstream of earthflows, which implies that they enhance salmon habitat.  For the temporal dimension, the authors used radiocarbon dates from charcoal and estimates of lake sedimentation rates to constrain maximum and minimum ages, respectively, of several earthflows.  They then used arguments about surface roughness to hypothesize about ages of other earthflows and noted that those ages may be broadly consistent with inferred trends in salmon populations.

Although I found the motivation for this study quite interesting, the data unfortunately aren’t sufficient to support the inferences regarding the timing of earthflows and their relationship to temporal trends in salmon population.  First, to test the hypothesis that high earthflow frequency corresponds to times of large salmon populations, the salmon population would need to be independently known over the Holocene.  I’m not a salmon expert, but that information doesn’t seem readily attainable, and the manuscript doesn’t provide any independent references that have determined it.  Instead, there is a suggestion that salmon populations may have stabilized 4-5 ka based on very broad scale inferences about climatic and tectonic changes.  Second, the study did not present a significant, data-derived relationship between earthflow age and surface roughness, which would be necessary in order to use surface roughness to infer age of undated earthflows.  So, although I wouldn’t recommend this manuscript for publication at this time, I did find the research question quite compelling, and hope the authors will consider some more detailed comments if they choose to continue pursuing work on this study:

• Age vs. activity. Since age (~time since a landslide happened) and activity (~how active the landslide is currently) both affect a landslide’s surface roughness, it can be quite challenging to disentangle these two effects for earthflows in particular.  This may be one reason why there is not a relationship between landslide age and roughness for the study site.  Fig. 5 is a bit misleading on this point, since it shows a modeling result, which by design will smooth the landslide deposit over time.  If you add the age and roughness values from Table 1 to the figure, it is clear that the radiocarbon-based ages and roughnesses are not correlated and do not follow the model-predicted trend.  Those data are so scattered that that general inferences about mid or late Holocene timing aren’t supported either (e.g. the smoothest landslide is the youngest (Dickey Cr.) and the oldest is the roughest (Indian Cr.)).  Perhaps the fact that age and roughness don’t correlate is a useful observation for thinking about what other factors control surface roughness of these earthflows.
• Landslide dates. As stated in the manuscript, charcoal provides only a maximum age, and often this age is not a close maximum.  E.g. see Struble et al. (2020), GSA Bull. where charcoal from a landslide’s deposit is up to thousands of years older than the landslide.  Also in the present manuscript, the discrepancy between the minimum and maximum ages of the Rye Creek earthflow (346 and 4,353 yrs, respectively) suggests that either the earthflow’s age isn’t accurately enough known to determine an age-roughness model, or that a single age for the earthflow may not be representative of its long term behavior, such as persistent movement and/or reactivations.
• An exciting, if somewhat technical, finding of this work is that the chosen roughness metric decreases linearly with model simulation time. Previous studies have found an exponential age v. roughness relationship, which means that absolute uncertainties of predicted ages are quite large for the older landslides on the steeper part of that curve.  A relatively simple fix of using this or a similar roughness metric may dramatically reduce the uncertainty on predicted ages of old landslides.

This manuscript investigates the timing of earthflows in the Teanaway basin, WA, and seeks to link earthflow activity with climate, geology, and the development of salmon habitat. The paper presents a relative dating technique (MADstd) for earthflows as well as 6 radiocarbon and 3 lake sedimentation dates for earthflows in the basin to constrain the absolute ages to the mid-late Holocene. They show that earthflows occur more frequently on southwest facing slopes, which suggests a structural control on earthflow location as the local formations dip southwest. They also measure valley width and show that earthflows impact valley width in the study area and that more active earthflows (those with higher MADstd) are correlated with locations with anomalously narrow valleys. They hypothesize in the discussion that earthflow activity is climate-modulated and that, because salmon populations stabilized in the mid-Holocene, these earthflows must not have negatively affected habitat.

I found this manuscript to be overall scientifically sound and well-written, but a bit confusing and in need of some figure revision and tightening around a clear hypothesis.

One thing I found confusing is the discrepancy between the model MADstd values and the MADstd values for earthflows in the Teanaway basin. I don’t understand how mid-Holocene ages were estimated for the older earthflows in the Teanaway when MADstd values are much lower than the lowest in the model. Perhaps this can be expanded. Unlike reviewer #1, I don’t think it matters that there is poor correlation between the absolute ages presented here and the MADstd – I think it’s clear how much uncertainty there is in the earthflow ages, particularly the older ones. That said, I think it might be nice to show in a figure that where you do have field evidence of relative ages (e.g., cross-cutting relationships), it does work. Also, I think this part of the discussion should be moved to the results.

There is too much focus on salmon habitat. Although I think it’s a good motivation for investigating the timing and controls on earthflows and how they impact valley width, the focus on it (e.g., section 2.2) implies there will be a solid conclusion relating to it. In the end, the claim is that given that earthflows were active when salmon populations stabilized, they don’t seem to have negatively impacted salmon habitat. It’s not a big conclusion, but reviewer 1 may be right that the data still doesn’t support it. We can never know what fish populations would be if there were no earthflows. I think all you can say is that earthflows contribute to topographic heterogeneity and that non-catastrophic disturbances and topographic heterogeneity are generally good for biodiversity.

The hypotheses about climate control on earthflow activity seem like a bit of a stretch given the uncertainty on earthflow ages. But, if you’re going to discuss this, Bennett 2016 probably ought to be referenced (see below). Perhaps though, instead, more focus should be on the main results: earthflows are active in the Teanaway basin, are structurally controlled, and act to modify valley width and hence floodplain habitat, as well as sediment flux and most likely grain size. Much of the discussion could instead be used to discuss limitations on the techniques employed and areas of future work.

I suggest deleting everything about rotational and translational slides to better focus on earthflows.

Figure comments

Fig. 3 I think this could be moved to a supplement

Fig. 7 I think this could be moved to a supplement

Fig. 8 Why isn’t valley width plotted against discharge as in May, 2013? It would also be really nice to see some lidar hillshade images of these constrained and upstream widened reaches.

Fig. 9 I think this figure should be moved up to the methods or results

References

1. L. Bennett,J. J. Roering,B. H. Mackey,A. L. Handwerger,D. A. Schmidt,B. P. Guillod, 2016, Historic drought puts the brakes on earthflows in Northern California, GRL