The Over-Prediction of Seismically Induced Soil Liquefaction during the 2016 Kumamoto, Japan Earthquake Sequence

Round 1

Reviewer 1 Report

The paper is interesting and deals with a critical topic in many areas around the world, i.e. the mis-prediction of liquefaction susceptibility in pyroclastic and transitional soils. It is well written and includes interesting filed and lab data. Provided you take into account the comments listed below (minor revision) the paper can be published.

·         Lines 446-447: You obtained soil samples using a piston sampler, which may cause some disturbance in loose (and collapsible, in the case of pyroclastic grains) soils. You mention this issue in your conclusions (lines 792-793) but it would be convenient a more detailed comment in the text on the possible effect of this disturbance. It would be also very interesting to compare the relative density guessed on the base of the close by CPT with the one measured in lab on the (theoretically) undisturbed samples. Even though it is a comparison between a semiempirical formulation and a measurement, it may add some evidence of interest.

·         Lines 512-521: you do not say anything on the way you reconstituted the specimens, which is a relevant topic considering its effect on CRR (please check the work by Cubrinovski and Ishihara in the ‘90s who gave clear evidence of that). Reconstituting soil specimens may therefore have an effect more relevant than the one you are considering, obviously depending very much on the geological depositional set up. Especially in your case, in which the micro structure may be metastable because of the high angularity of the grains, as often is the case with pyroclastic silty sands and sands, you may end up with a different micro structure by reconstituting the specimens, likely overestimating CRR. Since this may affect your conclusions, it deserves a comment in the paper.

·         Lines 785-787: you mention plasticity as a possible reason of the higher resistance of pyroclastic soils. But in some cases it is rather low. I understand you have no info on grains crushability and meta-stable structure, but aren’t you overstating the effect of plasticity? In my experience, in pyroclastic soils CRR based on in situ tests is often overpredicted even when plasticity is extremely low or nihil, and this is the reason why I think the microstructure should be investigated in more detail.

Author Response

Reviewer #1:

Comments and Suggestions for Authors

The paper is interesting and deals with a critical topic in many areas around the world, i.e. the mis-prediction of liquefaction susceptibility in pyroclastic and transitional soils. It is well written and includes interesting filed and lab data. Provided you take into account the comments listed below (minor revision) the paper can be published.

• Lines 446-447: You obtained soil samples using a piston sampler, which may cause some disturbance in loose (and collapsible, in the case of pyroclastic grains) soils. You mention this issue in your conclusions (lines 792-793) but it would be convenient a more detailed comment in the text on the possible effect of this disturbance. It would be also very interesting to compare the relative density guessed on the base of the close by CPT with the one measured in lab on the (theoretically) undisturbed samples. Even though it is a comparison between a semiempirical formulation and a measurement, it may add some evidence of interest.

(Response): We agree with the reviewer that adding some additional clarification on the likely sample disturbance (including particle grain crushing) will improve the paper. We have added this clarification. We also like the idea of comparing the in-situ relative density to the correlated relative density for the adjacent CPT. However, there is little meaningful correlation with actual in-situ conditions that can be inferred from the laboratory specimens due to the manner of sampling and sample preparation that was performed. It is well established that the “best’ method for obtaining undisturbed samples in sands is by freezing and coring the sand, which was well beyond our scope and budget for this research. We attempted to compensate by draining the tubes containing sandy specimens in an attempt to create negative pore water pressure and temporary “apparent cohesion” in the sand during extrusion and sample preparation for triaxial testing. We have clarified in the text that the samples used for triaxial testing at Site 2-1 were almost certainly disturbed to some degree and the test results should be interpreted with caution. For this reason, we are choosing to forego comparisons between the triaxial specimens and correlate in-situ properties such as relative density. We believe that the principal value of the triaxial testing is to provide a precursory understanding of how close the soils at Kumamoto may have come to liquefying in the 2016 KES.   

• Lines 512-521: you do not say anything on the way you reconstituted the specimens, which is a relevant topic considering its effect on CRR (please check the work by Cubrinovski and Ishihara in the ‘90s who gave clear evidence of that). Reconstituting soil specimens may therefore have an effect more relevant than the one you are considering, obviously depending very much on the geological depositional set up. Especially in your case, in which the micro structure may be metastable because of the high angularity of the grains, as often is the case with pyroclastic silty sands and sands, you may end up with a different micro structure by reconstituting the specimens, likely overestimating CRR. Since this may affect your conclusions, it deserves a comment in the paper.

(Responses): We thank the reviewer for his/her comment. In reviewing the data from these tests to prepare a response, we discovered that the reconstitution of specimens was performed for a different test not related to this study. We therefore have revised the manuscript to clarify that the specimens were NOT reconstituted, and provided the following detailed description on the sample preparation and test procedure:

“Cyclic undrained triaxial testing was also performed on soil specimens from Site 2-2. In some specimens that were mainly sand with no cohesion, water was drained from the plastic tubes to generate temporary capillary forces (negative pore water pressure) in the sand specimen. These temporary forces minimized sample disturbance during transportation and trimming. The soil specimens that had some cohesion were produced in the laboratory by first carefully pulling the soil from the plastic tubes. After being vertically extruded from the tube, each intact specimen was cut with a thin sharp edge to a height of two sample tube diameters, according to ASTM D5311. The diameter of the triaxial specimen obtained in this way was not reduced by trimming. It remained the same as it was after the extrusion. The specimen then was slid carefully onto the fine-grained porous stone with filter paper fixed onto the triaxial cell pedestal. The top porous stone was placed on the top of the specimen with filter papers, and a membrane was placed around it by means of a membrane stretcher and vacuum. The vacuum was then released so that the membrane became slightly pressed against the specimen and also grasped both porous stones. The triaxial top cap was first mounted by fixing it on the rigid columns located inside the cell, and the top specimen cap was then lowered onto the top porous stone already sitting on the specimen. Finally, the cell was closed and sealed by vacuum grease. Then the triaxial cell was transferred from the preparation table to the cyclic testing frame. Here, the cell was fixed to the base of the frame and attached to the vertical actuator and transducer, volume change burette, and pore water pressure transducer then the triaxial cell was filled with de-aired.

“After the soil specimens were prepared based on the procedure mentioned above, the specimens were slowly flushed with fresh de-aired water from the bottom to the top of the specimen. After the flushing stage, back-pressure saturation was conducted. Back pressure and cell pressure increments equal to 20 kPa and an initial mean effective stress equal to 20 kPa were used at saturation stage until the pore pressure coefficient B reached a value of 0.95 or higher. Final B values are reported for each triaxial test in saturation figure. Volume and height changes were recorded for each specimen during both the flushing and back pressure saturation stages using an external LVDT, which was securely connected to the axial actuator and piston rod. The volumetric strain was measured precisely for each specimen during the flushing and back pressure saturation stages using a calibrated electronic volume measuring apparatus. The time interval required to achieve saturation ranged from 1 to 2 hours depending on the fines content of the tested soil. After back pressure saturation, the specimens were subjected to isotropic mean effective stress equal to 100 kPa and were used in all the undrained cyclic tests. The time required to complete the isotropic compression stage was about 1 hour and when both the volume change and vertical strain reached the critical state.

“The cyclic strain-controlled tests were conducted in the sinusoidal mode with the servo-hydraulic closed loop system with calibrated digital instrumentation for measuring load, displacement, and pressure. These tests were conducted following the specifications of ASTM D5311 and BS 1377. During Cyclic axial loading, a cyclic load was applied to the saturated specimens and the variation of axial stress, excess pore water pressure, and axial strain of the specimen were continuously recorded during cyclic loading. The intensity of the cyclic load was varied in such a way as to produce a wide range of cyclic stress ratios that simulate or mimic the cyclic stress ratios that the soil exposed in the field and corresponding number of cycles required to cause initial liquefaction. It should be noted that initial liquefaction occurs when the excess pore water pressure becomes equal to the initial consolidation stress of the specimen. The cyclic loading in the cyclic triaxial tests was applied with frequency equal to 1 Hz.

“All cyclic triaxial tests were conducted in strain-controlled mode because (1) strain-controlled tests cause less water content redistribution in soil samples before initial liquefaction occurs and provides more realistic predictions of in-situ pore pressures than those obtained from stress-controlled tests; and (2) stress-controlled triaxial tests are less accurate due to the development of different strains during compression and extension phases. They also increase strains as the number of cycles progresses, which makes the determination of shear modulus and damping ratio values more difficult.”

• Lines 785-787: you mention plasticity as a possible reason of the higher resistance of pyroclastic soils. But in some cases it is rather low. I understand you have no info on grains crushability and meta-stable structure, but aren’t you overstating the effect of plasticity? In my experience, in pyroclastic soils CRR based on in situ tests is often overpredicted even when plasticity is extremely low or nihil, and this is the reason why I think the microstructure should be investigated in more detail.

(Response): We agree with the author that our conclusion regarding the effect of soil plasticity being the primary cause of the high CRR in the pyroclastic soils is likely overstated. We should not state anything conclusively that we did not actually test. We have modified the language in the conclusions to state: “Liquefaction was over-predicted for sandy deposits that were naturally placed. Volcanic soils, especially prevalent in the northern half of the Kumamoto Plain, were observed in our study to have medium-to- high fines content, moderate to high plasticity, and high organic content. These properties likely contributed to the increased resistance to soil lique-faction. Likewise, the crushable nature of the volcanic soils may have led to artificially low penetration resistance readings for the SPT and CPT. The crushability and metastable structure of the volcanic soils also certainly would have contributed (perhaps even dominated) the resistance of the soil to liquefaction triggering as has been observed in previous studies by other researchers. The less volcanic soils of the southern half of the Kumamoto Plain experienced much more sporadic minor liquefaction. Volcanically derived sands in general also showed a higher resistance to liquefaction based on our cyclic triaxial results.” We also note that our last bullet point in the Conclusions clearly states that we did not investigate the grain crushability of the soil. 

Reviewer 2 Report

The paper investigates an interesting topic such as the nature as to why soil liquefaction did not occur as prevalently during the 2016 KES as predicted by subject-matter experts. The paper includes a literature review assessment of the known liquefaction characteristics and an assessment is made to the validity of current liquefaction triggering models used commonly by geotechnical engineers. The methodology is pertinent and English is good.

Several issues need to be considered. 

1 Section 1 is too long and needs to be separated into several ones

2 Literature review: the numerical modelling on liquefaction needs to be considered in details. Please axpand this part

3 Section 4 : this "However, we emphasize that this study did not report the effects of liquefaction-induced ground settlement or bearing capacity failure, which would have been significant using conventional design  methods, even for liquefaction predicted at deeper depths" is a very limiting point. The authors should explain why they did this choice and also if the paper may be considered significant even if this assumption.

4 Conclusion: the limitation that the study did not measure or quantify the potential amount that grain crushability or the potential for the soil to be unsaturated due to the high organic content needs to be discussed and the authors need to clarify why they did not consider this potential important parameter.

Author Response

Reviewer #2:

The paper investigates an interesting topic such as the nature as to why soil liquefaction did not occur as prevalently during the 2016 KES as predicted by subject-matter experts. The paper includes a literature review assessment of the known liquefaction characteristics and an assessment is made to the validity of current liquefaction triggering models used commonly by geotechnical engineers. The methodology is pertinent and English is good.

Several issues need to be considered. 

1 Section 1 is too long and needs to be separated into several ones

(Response): We appreciate the reviewer’s opinion that Section 1 is too long and needs to be separated into smaller sections. We chose to follow the recommended paper format by Geosciences, which dictates that all introductory materials be placed in Section 1. According to this recommended paper format, Section 2 corresponds to Materials and Methods. Therefore, we believe that no change to the manuscript is necessary.  

2 Literature review: the numerical modelling on liquefaction needs to be considered in details. Please axpand this part

(Response): While we appreciate the reviewer’s enthusiasm for a discussion of numerical modeling of liquefaction, we do not agree with his/her assessment that it “needs to be considered in detail,” or even mentioned, in this manuscript. The intent and designed scope of this study and manuscript were to understand why existing indicators (e.g., aerial imagery, modern liquefaction hazard mapping, simplified triggering analyses using publicly available borings) all produced “false positives” of predicted liquefaction damage from the 2016 KES. At no time did any of the engineering reconnaissance experts associated with our effort (either Japanese or American) use or rely upon numerical modeling of liquefaction to inform their predictions of liquefaction severity in the days, weeks, and even months following the 2016 KES. Numerical modeling of soil liquefaction in pyroclastic soils is indeed a very interesting topic and worthy of study. However, we do not believe this manuscript is the place for that.

3 Section 4 : this "However, we emphasize that this study did not report the effects of liquefaction-induced ground settlement or bearing capacity failure, which would have been significant using conventional design  methods, even for liquefaction predicted at deeper depths" is a very limiting point. The authors should explain why they did this choice and also if the paper may be considered significant even if this assumption.

(Response): We have revised this statement to read as follows: “However, we emphasize that this study did not report the effects potential overprediction of liquefaction-induced ground settlement or bearing capacity failure, which would have been significant using conventional design methods, even for liquefaction predicted at deeper depths. We did not investigate these effects from liquefaction be-cause their investigation would have required the collection of additional critical data (e.g., geotechnical explorations and laboratory testing) directly adjacent to or beneath affected structures, which we did not have the time, budget, or authorizations to collect.” We believe that this paper is significant for engineers holistically assessing liquefaction hazard in areas with pyroclastic sediments. All effects from soil liquefaction (e.g., settlement, lateral spread, bearing capacity) are conditional upon triggering. Therefore, a study that makes conclusions regarding the triggering of liquefaction will similarly apply to all of the subsequent effects of liquefaction triggering.

4 Conclusion: the limitation that the study did not measure or quantify the potential amount that grain crushability or the potential for the soil to be unsaturated due to the high organic content needs to be discussed and the authors need to clarify why they did not consider this potential important parameter.

(Response): We have clarified in the bullet point that we did not measure these properties “because those properties were beyond the scope of our research and funding.” At the time that our research proposal to the US NSF was prepared and budgeted, those properties were not being considered by our team. It was only through the course of our study that we came to understand the potential importance of those parameters in increasing the soil’s resistance to liquefaction.

Reviewer 3 Report

The Authors present an interesting study on the overprediction of soil liquefaction in the city of Kumamoto during the 2016 Japan earthquake sequence. Based on the post-earthquake engineering reconnaissance and the analysis of in-situ and laboratory tests, current simplified liquefaction triggering procedures are commented on. The importance of considering geology and soil mineralogy for the evaluation of liquefaction in volcanic soils is highlighted in the paper. The research is well presented and supported by a detailed description of the site and available data. The paper can be accepted in this present form.

Author Response

Reviewer #3:

The Authors present an interesting study on the overprediction of soil liquefaction in the city of Kumamoto during the 2016 Japan earthquake sequence. Based on the post-earthquake engineering reconnaissance and the analysis of in-situ and laboratory tests, current simplified liquefaction triggering procedures are commented on. The importance of considering geology and soil mineralogy for the evaluation of liquefaction in volcanic soils is highlighted in the paper. The research is well presented and supported by a detailed description of the site and available data. The paper can be accepted in this present form.

(Response): We thank the reviewer for his/her review of our manuscript and positive feedback.

Round 2

Reviewer 2 Report

The authors did not consider the requests I asked for.

Therefore, the paper needs to be modified as I (the reviewer!) requested.

Regards