Beautiful, very impressive and convincing report! Only comment is the report focuses exclusively on the circum-Pacific subduction zones. Is this intentional (I didn't read the report in detail)? I see no mention of the Hellenic Arc that represents the most significant seismic and tsunami hazard in the Mediterranean region. Our Greek colleagues have recently had the arc selected as a "supersite", there has been a major effort for some yearts to instrument the arc and backarc, and there are efforts underway or being planned for tsunami warning systems and other hazard mitigation strategies. If you want information on this program, please contact the Suoersite Coordinator, Alexandros Savvaidis Greek Supersite Coordinator: Alexandros Savvaidis ITSAK (Institute of Engineering Seismology and Earthquake Engineering) Courier Address: Dasyliou Str, Elaiones, 55535 Pylaia, Thessaloniki, Greece Postal Address: P.O. Box 53, 55102, Finikas, Thessaloniki, Greece Tel: +30 6974482593 email: alexandros@itsak.gr

p.34 Singapore and Indonesia are in Asia, not Oceania p.42. I'm confused by this figure. For Sumatra, the gap is marked at the Enggano section of the Sunda megathrust, as well as far western Java. Why only far western Java, when the whole megathrust offshore Java has had no history of a large earthquake? I would say the whole thing is a gap. And for Sumatra, why is the Mentawai patch not marked as a gap, since it is known to be locked and overdue. For the Enggano section, however, it's unclear that it's capable of generating a great earthquake. It's quite confusing, actually, as the GPS data suggest that it is creeping, but we do see drowned trees on the island of Enggano that would suggest it has been locked in the past. Also, I don't see any blue dots, only white dots.

I think the report is very well put together. It makes a very compelling case for the importance of an integrated scientific effort to understand the science that underlies subduction zone hazards. It is somewhat lacking in detailed specifics of what a program might be (e.g., the details that would allow NSF to estimate the likely cost), but I think that accurately reflects the state of the discussions at the workshop. My biggest concern comes in section 9. To make SZ4D a success, a substantial investment in infrastructure will be required. The report text (p. 55, right column) explains that the infrastructure could be "grouped into a single, large, coordinated program or be a more loose federation of smaller, right-sized, mid-scale projects". However, Table 9.1 on page 54 lists only Mid-Scale Infrastructure/MRI proposals in this option and I would judge that the use of the term "right-sized" in the text indicates a preference for this approach. The practical problem with this is that unless there are development of which I am aware, NSF has no mechanism to fund mid-scale infrastructure. As I understand it MRI's top out at $7M and MRE's start at $70M (I am not certain of these numbers but there is a big gap) and while there are discussions at NSF as to how to remedy this, they have been ongoing for at least a decade with no evidence from the outside of much progress. Furthermore, the current incarnation of the MRI solicitation seems to limit MRIs to single instruments, thus excluding them as a mechanism to fund infrastructure for facilities that might operate networks of instruments. Even if NSF solves this mid-scale infrastructure program, it strikes me that the success of a program that is dependent on integrated rather than isolated disciplinary observations is going to be governed by the success of a the weakest infrastructure proposal. I spent nearly 20 years of my career in the RIDGE and RIDGE2000 program and this effort could never fully achieve one of its goals, integrated observations of seafloor volcano-hydrothermal systems, because of challenges of getting multiple proposals funded to make observations at the same time. It is a little different for this effort, but if the goal is integrated observations, it seems really odd to fund all the components of the infrastructure separately when that is harder path to follow. So, I would suggest that at the very least the report discuss the challenges of funding mid-scale infrastructure in NSF and identify that as a topic for discussion at the first infrastructure planning workshop.

Dear writing team, Congratulations on a fantastic report that well captures the discussions at the meeting. Great diagrams and images. Thank you for your hard work on this. Here some comments: I very much appreciate the gas chemistry is on diagram 3.1. You may change to "gas chemistry and flux" if it fits into the available space on the figure. p. 34 Columbia needs to be changed to Colombia Section 5: Link between hazards and science: no volcano hazards mentioned in this section. This section should also emphasize needed improvements of probabilistic methods for volcano eruption forecasting and hazards. These models need to be based on physical and chemical measurements in near real time to produce a physical volcano model as you also show in Box 7.6. Section 8.4 (p.52) International capacity building. An excellent example is also the Swedish 'Network of atmospheric and volcanic change (NOVAC)' that has been implemented about 10 years ago and is still going and growing in many countries around the world to monitor volcanic gas emissions. This network has resulted in close collaboration between volcano observatories, universities and government agencies to monitor volcanoes and assess hazards. Also the DCO DECADE network has resulted in a international volcanic gas monitoring network for science and for eruption prediction. In both cases capacity is built locally by supplying monitoring instrumentation that is maintained at the local level and training of local scientists in maintenance and data interpretation. Thank you for considering these comments and congratulations again on excellent work. Tobias

The SZ4D plan is spectacular in its breadth and is overall an exciting document. Of course, I see a couple areas that could be strengthened. One is that this plan under-sells the contribution that structural seismology will make to understanding the deeper parts of subduction systems. The primary constraints we have on geometry, composition, and arguably temperature and fluid distribution come from seismic imaging. Seismology remains the only tool that can provide relatively unambiguous depths to boundaries or other structures within the earth. The field has seen great advances over the last decade, with high-density array studies targeting upper mantle structure, a much better theoretical framework for interpreting seismic parameters, and impressive results emerging from EarthScope. Yet the document is startling in having no figures showing earthquake-seismic imaging results, for example among spectacular recent imaging from Japan, several parts of the Andes, New Zealand, Alaska, Cascadia, Mexico, Central America, and the Aegean (among many). Images would complement well the text in 2.2, and/or some of the boxes on international efforts. I would be happy to help identify something. There is ample room, e.g. in 4.5 (Frontiers) to emphasize that these technologies that have great potential here to contribute. Also, Figure 3.1 somehow misses seismic imaging as a useful tool - it uses “seismology” to mean “seismicity” but nothing at the time scales of 1E11 – 1E16 yr where imaging provides a primary constraint. The lists of “facilities” in sections 7.1-7.2 would benefit from mentioning a broader host of seismic, geodetic, electromagnetic, analytical and experimental on-shore facilities seem likely to be needed, beyond the offshore ones. Overall, the report seems odd in its neglect of this large and thriving field. A related benefit of a clearer presence of imaging would be to strengthen the motivation for models that examine processes at long time scales (>>decadal) and great depths. I would think a central part of SZ4D would be to actually test models and develop a tight interplay between observations and modeling; this requires thinking hard about what observations are possible and what are not. Beyond that, there is a tremendous opportunity to integrate the updip thrust zone and deeper parts of subduction zones where magmas originate; that could be strengthened. This requires thinking about the plate interface between the megathrust locked zone and the arc source region in probably some new ways, as the EU “ZIP” program is starting to do. The surface processes section (section 2.4) seems to be using “subduction zone” in a very different way than the rest of the document: that section talks about collisional, orogenic processes not the oceanic subduction prevalent throughout the rest of the document (which usually does not produce mountains). It seems unrelated, the way it is written. The MARGINS experience with Source-to-Sink showed that a disconnected surface process component is tough to maintain. This is a community with a very exciting set of approaches and problems and would be good to have them integrated into core investigations in SZ4D. Overall this is an exciting plan, and it is good to see it is moving forward. I suspect that the kinds of science I am discussing will be part of SZ4D regardless, I just think there is an opportunity to make use of their potential contributions to make the overall case stronger.

The efforts to include capacity building (CB) efforts and education and outreach (EO) from the beginning of the SZ4D initiative is applauded. There is much opportunity to weave these efforts within the science efforts especially of existing groups and organizations engaged in CB/EO. I note that while section 8 is devoted to a framework for CB/EO, these topics are not a part of section 9 that outlines things that can be done now to advance the ideas in the workshop report, nor articulated in table 9.1. There is a recommendation to develop a planning committee. It is critical that any planning committee include members or representatives with expertise in CB and EO. NSF currently funds organizations and varied PIs to support CB/EO activities. These existing organizations should be brought to the table to ensure that any efforts to support Broader Impacts move forward in tandem with the SZ4D science plan.

I think the SZ4D Initiative is a great opportunity for Geophysical research to start answering the big questions society demands nowadays about the devastating events that have occurred recently and that have killed so many. I agree that the effort must be worldwide in order for it to be effective. The Earth is the only laboratory we have to test our hypothesis about what is occurring in subduction zones, the places where those disasters are originated. However, I have major concerns about the theoretical background of the initiative relating to seismic occurrence. Firstly, the assumption of the seismic gap hypothesis as the starting point. That hypothesis has been rejected in a simple statistical test several times [Kagan and Jackson, 1994; Rong et al., 2003]. From my point of view, that hypothesis should be removed from the proposal. Mainly because the initiative is sufficiently sound without it. The fact that we need to understand subduction zones much better is reason enough for the initiative. I think that introducing a theoretical background that has been shown not to pass simple statistical tests could diminish the strength of the proposal. A throughout review of earthquake predictability has been made by Jordan et al. [2009], and I think that could be a better starting point for establishing the theoretical background on seismic occurrence in this proposal. Our basic understanding of earthquake physics is that the stress is being accumulated on certain regions due to the plates' motion, and that, whenever that stress surpasses the strength of the material, it ruptures, producing an earthquake. To study how the stress is building up in subduction zones, to monitor stress and strain production, and observing as best we can the properties of the materials in those regions is the goal, I think, of this initiative. As you mention in page 8, “Some seismologists hold the view that any subduction zone is capable of producing an M9 earthquake, contrary to the more orthodox view that different subduction zones have different properties that lead to different seismic outcomes. Some creep slowly, some fail in small earthquakes, and still other are capable of catastrophic large earthquakes. Segment boundaries appear to be persistent in some cases, yet at other times ruptures proceed right across apparent segment boundaries, even for the same subduction systems.” The question you pose “Are major earthquakes more likely to occur as time since the last major event increases?” is going to be answered by observing those quantities, without the need of introducing the seismic gap hypothesis. As you mention, sometimes they occur on reported seismic gaps, but sometimes they don't. Of course that happens, since the hypothesis has been rejected already. You would have some instances where they match the hypothesis, but that does not mean that it is statistically sound. The fact that all those factors are going to be monitored is enough for this proposal to be strong, because they are physical quantities that have been identified as important in the seismic process, in lab experiments and in actual earthquakes. Another issue that concerns me is the treatment of precursory phenomena. In Jordan et al. [2009], again, a review of where seismologists stand on that topic is thoroughly explained. Giving examples for Tohoku and Iquique is important, but those precursory phenomena are, as always, identified in hindsight. There is not usually distinction between foreshocks, mainshocks and aftershocks in seismic models, because these definitions depend on what we understand by mainshock. It is usually the biggest event, but that does not mean is the originating one. Actually, it would all be a continuum of events triggering others by introducing stress changes on the neighbouring faults. So, in general, while I agree on the necessity of understanding what happens on those regions where the most disastrous events occur, by observing directly what happens, I think a more careful statistical analysis should be given to this initiative. I understand that the proposal includes two different approaches: the first one is the long-term observation of subduction zones, where no foreseeable outcome in terms of models can be done (although some informed guesses can be made), and then controlled experiments. I think it is of utmost importance to emphasize the statistical tools that will be used to evaluate the outcomes of these experiments. The CSEP (Collaboratory for the Study of Earthquake Predictability) project is a good example of how to formalize those tests. I clearly see how the seismic gap hypothesis can be easily understood by the population, since it is an intuitive way of explaining what is happening in the Earth. However, scientific rigour should not be disregarded in order to fulfil a desire to be taken into account for funding with a loose explanation. I think that deters the strength of the initiative. Moreover, whenever any of those gaps identified as “more likely” to rupture are not the ones that finally produce a major disaster, the whole initiative could be in danger of discredit because of it. Actually, the complexity of the whole thing makes it so difficult to forecast where the next major event is going to occur. It is what makes checking “seismic gaps” so inefficient to be prepared for such threats. We don't want another Parkfield experiment where, sure enough, much insight and theoretical advances has produced, but that was thought of as the definite project to understanding earthquake predictability. I don't know how you could explain again to society the failure of an experiment where we already know that could happen. I think you can avoid that from the very beginning by being rigorous and careful about what you say. To explain why it is important to study those places in the long term, and to layout a set of statistical tools for testing those experiments before they are performed is, from my point of view, the optimal way of convincing governments and sponsors to provide support to this initiative in the long term. I think that, apart from all the other topics that are covered in the initiative, a special task would be providing sound statistical tools for testing all the models that will be proposed with the new observations. I believe that would give strength and scientific rigour to the proposal.

I read the report with great interest. My specialization is the magnetic field of the Earth’s lithosphere. I am a postdoc at GFZ, Section 2.3 Earth’s magnetic field. My comment concerns the fact that no mention is made on using the Earth’s magnetic field as a tool in deciphering subduction zones. I understand this to be due to the fact that no research has been done recently on this topic. However, there have been several relevant papers in the ‘80s and ‘90s based on MAGSAT satellite data (e.g., Frey, GRL, 1982; Clark et al., GRL, 1985; Arkani-Hamed and Strangway, Tectonophysics, 1987), on aeromagnetic data (e.g., Blakely et al., Geology, 2005) and on marine magnetic data (Okubo et al., Tectonophysics, 1991; Okubo and Matsunaga, JGR, 1994). According to these studies, subduction zones generate a strong magnetic field signal due to the magnetization contrast between the cold, subducted oceanic crust and the surrounding hotter nonmagnetic mantle and due to local variations of the Curie isotherm. Serpentinization plays also a key role. By fitting even simple shape models to the observed magnetic anomalies, properties of the subduction zones such as its thermal state, its lengths and dip can be inferred. Since the ‘90s, the geomagnetic community has made significant progress both in acquiring higher resolution satellite data (satellites Oersted, CHAMP, current satellite mission Swarm) and in compiling together all available aeromagnetic and marine magnetic data (World Digital Magnetic Anomaly Map (WDMAM) project, www.wdmam.org). This, in combination with advances in data processing and modelling techniques, has led to global magnetic field models of much higher spatial resolution. Advances in the interpretation of the magnetic field signal in terms of sources’ properties have also taken place. Based on the above, in my opinion, geomagnetism is currently in a good state, when combined with input from other disciplines already mentioned in the report, to contribute to our understanding of subduction zones. Personally, I am very much interested in this topic and would be glad if I could contribute to your SZ4D Initiative.