EGU2020-8640
https://doi.org/10.5194/egusphere-egu2020-8640
EGU General Assembly 2020
© Author(s) 2020. This work is distributed under
the Creative Commons Attribution 4.0 License.

A closer look at factors governing landslide recovery time in post-seismic periods

Hakan Tanyas1, Dalia Kirschbaum2, Luigi Lombardo3, and Tolga Gorum4
Hakan Tanyas et al.
  • 1Universities Space Research Association, NASA Goddard Space Flight Center, Hydrological Sciences Laboratory, Greenbelt, United States of America (hakan.tanyas@nasa.gov)
  • 2NASA Goddard Space Flight Center, Hydrological Sciences Laboratory, Greenbelt, United States of America (dalia.b.kirschbaum@nasa.gov)
  • 3Faculty of Geo-Information Science and Earth Observation (ITC), University of Twente, Enschede, Netherlands (l.lombardo@utwente.nl )
  • 4Eurasia Institute of Earth Sciences, Istanbul Technical University, Istanbul, Turkey (tgorum@itu.edu.tr)

Various mechanisms are proposed to explain landslide recovery time in the time following major earthquakes. However, research on prescribing possible recovery times following an earthquake is still relatively new. This paper provides an insight into factors governing landslide recovery time, which could be considered as a step forward in predictive modeling for landslide recovery time. To accomplish this, we examined 11 earthquake-affected areas based on the characteristics of both landslide events and landslide sites associated with diverse morphologic and climatic conditions. Our analyses indicate that the dominant characteristics of post-seismic landslide mechanisms determine the recovery time. The characteristics can be identified based on: (i) the fraction of area affected by landslides (%), (ii) mean relief and its standard deviation (m), (iii) average daily accumulated precipitation (mm) and (iv) rainfall seasonality index. If there are not enough co-seismic landslide deposits or not enough relief to trigger large deposits on hillslopes, then the recovery processes are mostly controlled by new landslides caused by a strength reduction of hillslope materials. In most of the cases, this brings a relatively quick recovery process in which the majority of post-seismic landslides may happen within a year or even in a month if sufficient intense rainfalls occur soon after the earthquake. If the predisposing factors create large co-seismic landslide deposits, then remobilization of material takes the role of the dominant mechanism and recovery may take years. Overall, our analyses show that the recovery takes relatively longer if a large amount of co-seismic landslide material is deposited within a high-relief mountainous environment where precipitation rate is low and not persistent.

How to cite: Tanyas, H., Kirschbaum, D., Lombardo, L., and Gorum, T.: A closer look at factors governing landslide recovery time in post-seismic periods , EGU General Assembly 2020, Online, 4–8 May 2020, EGU2020-8640, https://doi.org/10.5194/egusphere-egu2020-8640, 2020

Comments on the presentation

AC: Author Comment | CC: Community Comment | Report abuse

Presentation version 1 – uploaded on 28 Apr 2020
  • CC1: Type of fitting laws, Gianvito Scaringi, 04 May 2020

    Hi Hakan,

    I was curious to know if there is a specific (e.g. physically-based) reason behind the type of fitting laws you chose in your analyses or they are simply those who maximise the fit. In slide 7, you use a linear law for the fraction of landslide area, exponential for the relief, logarithmic for the accumulated precipitation, and again exponential for the seasonality index.

    About the seasonality index, I also was wondering if what other indices you considered since this one does not seem to perform great.

    Finally, I see that Wenchuan kind of consistently behaves as an outlier, or almost. I guess that, if you discard it, the regression coefficients, R2, and maybe even the type of law could change significantly. In your opinion, is there anything special (an additional factor, a specific morphology, hydrology, lithology, seismic pattern) about the Wenchuan case that we all are neglecting and makes it look different from all other cases? Or perhaps we simply do not have enough large magnitude earthquake cases that we are analysing?

    Cheers,

    Vito

    • AC1: Reply to CC1, Hakan Tanyas, 04 May 2020

      Hi Vito,

      Thank you for your questions.

      As you said, we did not use the same function for fitting. Instead, we used the function giving the best fit for each case. Let me tell you more about our logic to better answer your questions. We did not aim at proposing an empirical relation to predict recovery time. What we have done is exploring the possible contributions of various factors. In our paper, we also tested the relative contribution of each variable in a multivariate regression scheme. In addition to the one we used in our study, there are many other factors that possibly play role in recovery time. For example, vegetation recovery is an important process in terms of landslide recovery. However, if we have high precipitation rates and persistent rainfall pattern, we can already expect to have a quick vegetation recover. Therefore, we tried to consider the main factors governing the recovery process. And that is why, for example, we used seasonality as well. As you mentioned, seasonality does not seem to perform great but considering the physic behind the recovery process, seasonality is still a factor we need to address. It is also a fair comment that the Wenchuan appears as an outlier as usual. In fact, we have checked our fits by excluding the Wenchuan and there is no dramatic change in R2. Actually, it even helps to have a better fit for some variable (precipitation and seasonality). However, we do not think that R2 values are that important in our analysis because what we conclude is that there are different mechanisms governing the recovery process in different cases. Therefore, we categorized the examined landslide events based on the dominant process of landsliding (new landslides, reactivated landslides and remobilized landslides). As these mechanisms are completely different, for instance, in the Wenchuan and the Reuleut cases, they do not have to align with a nice fit following the same function. And yet, we believe our fits are important to show that there is a trend in each case.

      I hope this answers your questions. Please let me know if you have further questions/comments.

      Thank you very much for your interest in our work.

      Best,

      Hakan Tanyas

      • CC2: Reply to AC1, Gianvito Scaringi, 05 May 2020

        Hi Hakan,

        Yes, thank you very much for taking the time to reply. 

        Cheers,

        Vito