Microsoft word - itm_kyushu_20091117.doc

14 Conclusions
The Kyushu Case Study has helped to extend and test the ITM Methodology further, from the basis developed in the Tohoku Case Study. However, it is emphasised at the start of these conclusions that both Case Studies were intended to be development tools, rather than demonstrations of a completed, mature and systematic technical methodology that was already available and widely tested. However, despite this caveat, we believe that the two Case Studies together have produced all the main elements of such a methodology and illustrated comprehensively how it can be used to provide probabilistic assessments of tectonic hazard. The methodology is now ready, at a ‘beta-test’ level, for deployment in any areas that may arise in NUMO’s repository siting programme. The methodology will need to be refined, adjusted to the scale of the area under consideration and adapted to the geological and tectonic complexity of the region. In particular, those early Steps of the methodology that use expert elicitation to develop and weight appropriate models will need to be managed rigorously – and with the input of considerably more resources and local expert knowledge than was possible in our methodology development project. In these conclusions, we look first at how the methodology performed when extended from the relatively simple Tohoku application and applied to the considerably more tectonically varied terrain of Kyushu. We make observations about the limitations that are inevitably evident at this ‘working prototype’ level of development. These observations lead on to suggestions for how the methodology would need to be upgraded for actual application in making siting decisions at ‘real sites’ and in feeding quantitative information to safety assessment work. Finally we look at two aspects of extension of the methodology for such use: how it might be downscaled to a single site level and how it might address time periods that are longer than those of principal concern in siting (i.e. beyond 100,000 years). 14.1 Performance of the ITM Methodology in Kyushu
The Kyushu and Tohoku Case Studies differ in a number of respects. Tohoku is a region that has a relatively stable plate tectonic setting, while the Kyushu region is one of the most dynamic and rapidly changing plate boundaries in the world. The situation in Kyushu is intrinsically more complex. The datasets are also less complete or confident in Kyushu compared to Tohoku. GPS rates appear to be higher in Kyushu, while paradoxically the active faults and seismicity are much lower in Kyushu. The apparent discrepancy in Kyushu between GPS data and uptake of deformation on active faults may reflect some combination of the historical seismicity sampling too short a time period, the presence of active faults that have not yet been identified, deformation by aseismic creep, and that some of the deformation is taken up by magmatic intrusions in regions of extensional tectonics. The various modes of strain accommodation in the Kyushu are not yet fully understood, as reflected in these multiple explanations. The style of volcanism and geochemistry of the magmas varies considerably across Kyushu, but is reasonably simple arc volcanism in Tohoku. Predictability of future volcanism and faulting in Kyushu is less certain because geological setting is evolving very rapidly, while the fundamental assumption in Tohoku that the plate motions driving intraplate deformation are stable over periods up to a hundred thousand years are not appropriate in Kyushu. Overall, the Methodology has performed very well in Kyushu, with good agreement between the essentially seismotectonic domains and the volcano chemistry dataset, good general agreement between probability surfaces and volcanism (although the volcanism is too young and the tectonics too dynamic to establish steady state patterns), and good integration between volcanism and tectonism, owing to the use of coherent geological models on which the probabilistic procedures are based. We note that the differences observed between the different rock deformation datasets are mark of the success of the methodology, in the sense that it highlights specific issues that would need to be resolved before one could be sure of the level of confidence in the geological model of the region. 14.2 Limitations of the Current Version of the Methodology
The Kyushu Case Study has revealed a number of limitations to the methodology, which should be addressed when it is being deployed in an actual siting programme. Depending on the area being evaluated and the stage of the site evaluation programme, there may be a requirement to look at the likelihood of lower impact volcanic events (than vent formation at the repository site), such as dike or sill intrusion, the development of large-scale hydrothermal activity or, possibly, events that could affect surface facilities over the multi-decade operating life of a repository (e.g. tephra deposition). Some of the rock deformation datasets used in this study are inherently regional to sub-regional in their utility – GPS for example. Here, a denser network of observations and data-points can refine uncertainties, but it cannot yield variations in strain rates at <10 km spatial dimension, unless the deformation is very shallow in the crust. Faults and seismicity are currently assessed as independent datasets but it may be more appropriate in future to merge them, as is done in PSHA for nuclear facilities. This would provide areal continuity, although it mixes temporal scales. The seismicity dataset requires further assessment – it shows uniformly low strains and it is unclear whether this is a data limitation or a methodological problem. At a site scale, the methodologies cannot be a substitute for site-specific deterministic studies of the nature, history and impacts of tectonic processes as evidenced by field investigation data. However, these site data should be consistent with, and further inform, the range of alternative geological models that underpin the ITM Methodology. At the PIA and DIA stages of the project the ITM probabilistic methods will support the site investigations and the results of site investigation can be used to modify the weightings of various parameters in the probabilistic model – perhaps providing more confidence (and weight) to one or other geological model. 14.3 Limitations of the Kyushu Dataset
Significant data limitations have been identified in the study. These include active fault parameters, inland uplift data, the very limited range of recorded historical seismicity (which is generally of low magnitude events that nowhere approach a Mmax event) and uncertainty in historical earthquake catalogues. Inverting the seismicity rates suggests the return time for an Mmax event somewhere in Kyushu is about 80 years. No M 7.5 events are known and only two events in the 400-year period of the record are near M7 (1912 in the Kagoshima Graben and 2006, near Fukuoka). There are many opportunities for further datasets, especially geophysical, to be used to inform alternate viable geological models of both rock deformation and volcanism For example, it is not clear at present whether the gap in the volcanic arc is related to the subducting slab or whether there are upper plate or mantle processes that control the extent of the active arc. An obvious question is whether the gap will continue to be an anomaly out to, say, 1 Ma. A related question is whether volcanism in the Beppu-Shimabara graben and in the Kagoshima trough is localised owing to tectonic extension, or whether slab geometry or mantle constraints are the fundamental driver. Also, we have not yet explored how much of the geodetic strain observed in Kyushu may be related to volcanism in various parts of the region. 14.4 Opportunities for Further Methodology Development
Prior to, or in the course of, applying the Methodology in an actual site evaluation programme, whether in Kyushu or elsewhere in Japan, it would be reasonable to carry out some further developments. Itemised below is an initial checklist of topics that could be considered, although it is suggested that this is revisited at the time of deployment. There are opportunities for considerably more analysis of how various logic tree options and branches should be weighted and further analysis of what elements of the probabilistic formulation constitute aleatory uncertainties (natural, earth system variability) and epistemic uncertainties (knowledge uncertainties that can be reduced with further investigation). It is probable that different weighting profiles could be used in different time periods, reflecting confidence in the applicability of data over different periods of interest. Further opportunity exists to assess whether strain has the same annual probability for all return periods and the form that the strain will manifest itself in. At some stage of site evaluation, it will be essential to explore the range of forms that strain release may take in a volume of rock, which will depend on the geological context. Only by considering the manifestation of the strain can the estimated strain rates and likelihoods be factored properly into safety assessment models for a repository. One aspect that needs to be considered when looking at strain manifestation is how much strain is taken up by volcanic intrusion across a region. When repository design becomes more specific, then a useful strain parameter may be displacement. This measure can be developed from the probabilistic procedures already in place. Further development of analysis methods pertaining to the seismicity catalogue may be warranted. The uniformly low seismicity strain rates in Kyushu may be related to the use of the maximum likelihood method of developing seismicity parameters “a” and “b”. This, coupled with the strong effect in the model of smoothing by placing highest weights on a uniform source characterisation, strongly implies that earthquakes of similar magnitude and rate can occur anywhere in Kyushu. This represents a strong preference for “gap-filling”. The alternative is that earthquake occurrence is localised to crustal processes and structure and that future earthquakes are more likely to occur in places where they have occurred in the past. There are many opportunities for further analysis of uncertainties and sensitivities in the probabilistic methodologies. The figures produced to demonstrate the application of the ITM probabilistic methodologies for future volcanism and rock deformation in Kyushu often represent preferred or mean estimate models, but the sensitivity and uncertainties of these mean estimates is only rather poorly assessed at present. The potential for uplift of the repository into the oxidising zone could be a significant issue in some locations if the safety case extends to time periods of 1 Ma. Uplift rates as low as 0.2 mm/a may represent significant risk if the requirement for geological stability extends to 1 Ma. Very detailed work is required to assess uplift and uncertainty confidently, at this level of precision. 14.5 Refining the Methodology for Downscaling to Sub-regional or Site
The ITM Methodology has been developed principally to consider hazard from future volcanism and tectonic activity on a regional scale. In the context of Japan this means regions of the order 105 to > 106 km2. The methodology has been developed for a variety of purposes, including comparison of the relative hazard at competing sites, testing the adequacy of exclusion criteria, investigating the uncertainty in the assessment of hazard at specific places, and mapping out hazard across the region to identify places of low hazard and therefore with promise as site location. The hazard is expressed in terms of strain rates, locations of major tectonic structures, and for volcanism the probability of a new volcano forming in a specified time period. Such hazard assessments will inform the PIA stage. Once a potential site or volunteer community has been identified then the detail with which the hazards are assessed needs to be commensurately improved. While the same essential questions are being asked, in essence what is the probability of a site being disrupted or compromised by volcanic or tectonic activity, the change in scale will require some refinement of the methodology. Perhaps the most important aspect of down-scaling is that there needs to be an increase in the volume and completeness of data. Existing data can be gathered during the PIA stage and gathering of new data with associated interpretive analysis may be essential as the programme moves towards specific site investigation. The more detailed data and analysis of all the tectonic and volcanic datasets will be needed principally to reduce or constrain the uncertainties. For volcanic hazards the questions to be addressed are likely to change with reduced scale. At large-scale whole regions of order 105 to > 106 km2 have been considered using the ITM methodology. The hazards have been expressed as probability of a volcanic event defined as a volcano at a specific site over a specified period of time. The ITM Methodology thus far has considered specific sites to mean areas of 5 x 5 km and time-scales of 104 to 106 years. The modification of the methodology at intermediate and site specific scales is now discussed. At sub-regional scale (~104 to 103 km2) the probability of new volcanoes forming needs to investigated not just at the site itself but in areas that are close enough to the site that they might pose direct or indirect effects. The “15 km” exclusion criterion provides a yardstick but indirect effects (such as hydrothermal systems) might indicate larger distances should be considered. The method could, for example, be modified to look at the probability of a new volcano forming in an area being affected by a volcano sited at a distance of 15 km from the site. At this intermediate scale, the different kinds of hazards may need to be considered individually. The most important hazard is intrusive disruption of the site. Thus, models may need to be developed which look in more detail at the probability of intrusive disruption across the region, taking into account local tectonic, geophysical and geological data in more detail. It is likely that much more detailed geochronological, paleovolcanological and palaeosiemological studies will be needed within the sub-region to characterise recurrence rates of volcanism, styles of volcanism and past history of faulting. Deterministic assessments are likely to also be important. For example, probability of rock deformation or volcanism might change discontinuously across a major fault between two different tectonic domains, and using smoother probability algorithms may give misleading results. As another example, probability of intrusion might be represented as some functional form of the distance from the centre and the hazard of intrusion might be some combination of intrusive hazard related to multiple volcanic events. At a scale close to the repository footprint itself (~ 102 to 101 km2) the very local factors will start to affect probabilities and a more deterministic approach may be essential. Site evaluation should be based upon standard geological principles, but guided by “potential issues” identified via coherent geological models at regional and sub-regional scale. An example from the rock deformation perspective would be found when the fault process zone width is highly dependent on factors such as the sense of fault movement and lithology of the surrounding rock. For instance, the zone may be wider for a reverse fault in weak rock versus a strike-slip fault in hard rock. Alternatively, the fault zone complexity may reflect fault maturity. Therefore the distance from a known fault will have to be carefully defined (in the current Kyushu Case Study we have taken a rough centreline for the fault position) and the likelihood of future changes to the fault zone width should be assessed. Another example is that a site located along strike of a region of young volcanic vents and related tectonic features might be expected to have higher hazard than the other sites located in the same distance from volcanoes but in a direction normal to the volcanic lineament. Local geological or tectonic fabrics are likely to have a major influence on the nature of future volcanism and its interaction with the repository. At the scale of the repository the hazards analysis is likely to be strongly linked to consequence analysis. In regions like Kyushu, the arc volcanism and active fault systems are exceptionally young and unsteady and therefore harder to predict. Down-scaling will require incorporation of understanding of how the large scale changes in plate configuration and tectonics relate to the smaller scale relationships at the sub-region scale and specific sites. A key aspect of sub-regional and site scale characterisation will be to understand stress distributions and strain rates. Monitoring of strain rates down bore-holes is an example of the kind of technique that could be applied. 14.6 Refining the Methodology for the 100 k years to 1 million years
If the requirement for forecasting geological stability were to be extended to 1 million years it would present formidable challenges. Plate motions and configurations may not remain constant in space and time. For example, the volcanic front can be expected to migrate and subducted features that may affect the intensity, style, and location of volcanism and faulting, such as sea floor island chains or ridges, may change their position beneath Kyushu. The forcing factors in back-arc volcanism are poorly understood with apparent volcanic “flare-ups” such that the hazard is not constant in time. Other datasets would be needed to inform coherent geological models to overcome the requirement to drop the constraint of plate tectonic stability over the time period of interest. Potential unsteadiness in tectonics and volcanism will lead to greater uncertainty in the probabilistic models. A wider range of viable geologic models could accommodate increased uncertainty of the temporal and spatial evolution of faulting and volcanism. For volcanism, this time period presents formidable problems in that very large volcanic systems, such as calderas, can develop on time scales that are well within this time period, Aso being a very good example. Any location that is within the footprint of the volcanic arc or back arc region could be affected by future volcanism. Back arc regions are less prone to major changes, but over a million years the possibility of any given location within the back arc region with no previous record of volcanism developing monogenetic volcanism cannot be ruled out. Slab roll back provides issues for assessing fore-arc regions, especially close to the contemporary volcanic front. It is clear that significant adjustments in plate configurations and regional tectonics are possible in a period of 1 million years, so assessment over such a long period would necessarily have large uncertainties.

Source: http://www.numo.or.jp/en/reports/pdf/TR-09-02-14.pdf