In fault slip estimation based on observation of earthquakes, the model prediction errors that originate from the uncertainty of underground structure is often a major contributor of the errors between the data and the model predictions. However, most studies on slip inversions either neglect the model prediction errors or do not distinguish them from observation errors. Several methods that explicitly incorporated the model prediction errors in slip estimation commonly assumed a Gaussian distribution for the stochastic property of the model prediction errors to simplify the formulation. In this study, we propose a novel flexible Bayesian inference framework for estimating fault slips that can accurately incorporate non-Gaussian model prediction errors. This method considers the uncertainty of the underground structure, including fault geometry, based on the ensemble modeling of the uncertainty of Green's functions. We performed simple numerical experiments for estimating fault slip distribution on a 2D thrust fault using synthetic data of surface displacements. In the experiments, the dip angle of the fault plane was the parameter used to characterize the underground structure. The proposed method succeeded in estimating a posterior PDF of fault slip that is consistent with the true one, despite the uncertain and inaccurate information of the dip angle. In contrast, the estimation results obtained using a conventional approach, which does not explicitly distinguish the prediction and observation errors, included a significant amount of bias. Furthermore, the shapes of PDFs for the model prediction errors in certain observation points are significantly asymmetric, suggesting that they would not be well-modeled by Gaussian approximations. Reproduced from Agata et al., A Bayesian inference framework for fault slip distributions based on ensemble modeling of the uncertainty of underground structure - With a focus on uncertain fault dip, in press, Geophysical Journal International, doi:10.1093/gji/ggab033 / CC BY 4.0

Principal component analysis (PCA) and independent component analysis (ICA) are used to extract processes from geochemical datasets in various fields in geosciences. However, it is still not well-understood whether the extracted components truly correspond to geological processes. In this study, to verify the effectiveness of PCA and ICA as a geochemical-process extraction method, they are applied to the synthetic dataset of weathered rocks, which were generated using a simple forward model that mimics geochemical weathering. The obtained results shows that both methods can extract the leaching process of the elements during weathering as well as the heterogeneity of original-rock composition. We will also discuss the method of interpreting the extracted basis vectors as geological processes.

Earthquakes with very similar waveforms (The Moderate Repeating Earthquakes) have been confirmed in various Japanese regions. Most occur at the plate boundaries and are considered to be closely related to regular slippage between plates. Comparing earthquakes of the same magnitude, it is considered that the slip speed between plates is fast when the interval between occurrences is short, and slow when the interval is long. We are tracking the moderate repeating earthquakes that occur repeatedly at almost the same location and have very similar waveforms, using seismic records observed by the Japan Meteorological Agency's velocity-type seismographs and accelerometers. On the other hand, tsunamis and large earthquakes with damage have occurred near the plate boundary. This time, we focused on the timing of occurrence of large-scale earthquakes in areas closer to the trench than where the moderate repeating earthquakes occurred, and investigated the relationship between the moderate repeating earthquakes and large-scale earthquakes.

Spatiotemporal weighted regression (STWR) is a new tool for exploring local spatial and temporal non-stationarity in processes. Its main idea is to incorporate the rate of change of response variable into the temporal decay weighting kernel. Compared with geographically and temporally weighted regression (GTWR), it can more finely describe the change rate's spatial heterogeneity. Its initial bandwidth is smaller than that of geographically weighted regression (GWR), and consequently, the estimated surfaces of coefficients and response values produced by STWR will exhibit strong local variation. However, being an initial version of STWR, there exists some concerns about overfitting and model predictions. We designed random simulation tests for comparing the estimated coefficients of STWR and GWR. Results show that as the sample size increases, both the predicted surfaces of the response variable and coefficients of STWR and GWR tend to approach the true values. In large samples, STWR and GWR performance is close, while in small samples, STWR is obviously more suitable and better than GWR in terms of fits and predictions. The average local R-squared (R2) of GWR, sometimes, is higher than STWR, but its estimated standard error is higher than STWR. Besides, the out-of-sample predictions of response variables and coefficients are worse than those of STWR, which indicates that STWR is less likely to overfit than GWR. We also conduct experiments on real-world data on Beijing's Air Quality Index (AQI) and similar results. Interestingly, by comparing the improvement of the STWR to the general GWR of the local R2, we found that the improvement of the northern counties ( mountain counties) of Beijing is much less than that of the southern counties (plains counties), which is consistent with the change of the AQI(response variable) in the south faster than the change in the northern mountainous area. 

Slip motion along a fault largely depends on the inhomogeneity of friction that occurs between the fault interfaces. Thus, it is a crucial task to estimate the spatial-dependent frictional features from the observations of the slip motion and then to identify essential parts that contribute to the principal slip motion by quantifying uncertainties involved in the estimates. This study considers an uncertainty quantification problem of the spatially-dependent frictional features based on a fault motion model that mimics the slow-slipping region along the Bungo Channel in the southwestern part of Japan. The fault motion model employs a rate-and-state dependent friction law, in which the frictional parameters are spatially dependent. Although uncertainty quantification in high-resolution is needed to attain the above task, such quantification based on the conventional statistical ways is computationally hard since the complexity exponentially increases with the spatial resolution. This study employs a variational data assimilation method based on a second-order adjoint method to avoid such complexity. Since the data assimilation method enables a selective extraction of the uncertainty of interest, we can attain a fast uncertainty quantification of the frictional parameters in high-resolution. The application of the data assimilation to the fault motion model together with the synthetic observational data of the slip velocity quantifies the spatial dependency of the uncertainty involved in the frictional parameters in high-resolution and reveals how the amount or quality of the observational data of the slip motion influences to the spatial distribution of the frictional parameters. Such quantification provides valuable information to the observational design.

The use of Magnetotelluric method is implemented in attempt to image the subsurface structure in Pingtung Plain area of Southwestern Taiwan. The Plain is in the transition zone between the Manila subduction and Taiwan collision belt, and is considered as one of the country largest groundwater recharge area and is faced to rapid subsidence due to both tectonic activities along with excessive ground water pumping. Thick Holocene-Pleistocene sediments of less than 500m are covering most of the Plain surface. The sparse borehole data in the area provide limited geological information for the subsurface structure. In this study, we collected the Magnetotelluric data from 10 stations in the northern Pingtung Plain to explore to the depth of approximately less than 4km, divided into 2 profiles that is trending North-South and West-East. We also utilized 1 remote station, and selected 4-hours midnight timepiece to reduce the manmade artifacts and noise effects. From 1D inversion result, we obtained a sharp boundary between the unconsolidated sediments and the basement contact at 500m deep. The topmost conductive value belongs to the alluvium layer (<50Ωm) where it contacted the Miocene basement rock of resistivity reaching thousands Ω-m in the depth of approximately 500m. Most features observed in the study area shows a fault-like features and an anomalous volumetric body, thus hardly correlated with the surrounding mud diapir and volcanoes. Also, the topographical effect of SW Taiwan is assumed to be the dominant factor affecting the river course orientation rather than geological structure.

Positioning data with cm or mm level precision gives us a lot more possibilities than with low level (meter level) precision. Real-Time Kinematic (RTK) is a popular GNSS positioning method used in the case of high-precision applications. Only high-cost GNSS receivers provided better RTK solutions because the number of GNSS was not enough, but those were not affordable for many surveyors. The increase in the number of GNSS made low-cost GNSS receivers provide better RTK solutions recently, and ubiquitous networking connected to the internet enables rober-receivers to get parameters for correcting positioning signals of GNSS from station-receivers.In this study, a low cost and easy local-area RTK-GNSS system that uses existing GNSS satellites, the internet, 4G connection, and a caster server to provide real-time horizontal and vertical positioning within a few centimeters using NEO-M8P-2 module receiver with Raspberry Pi 3b+ as a station system and NEO-M8P-0 module receivers with a smartphone as a high mobility rober system, was developed and evaluated.The accuracy of the positionings (coordinates and height) was verified at each level point basically once. And to evaluate the effects of changes in environmental conditions, such as the positions of GNSS satellites, weather conditions, and 4G connection conditions, we have also conducted a series of precision surveying observations using local-area RTK-GNSS at the fixed point on the university campus over several months. It was found that this system was completely maintenance-free at least for this period and could always provide data with sufficient accuracy.

Underworld is a python-friendly geodynamics code which provides a programmable and flexible front end to all the functionality of the code running in a parallel HPC environment. This gives significant advantages to the user, with access to the power of python libraries for setup of complex problems, analysis at runtime, problem steering, and coupling of multiple problems. Underworld2 is integrated with the literate programming environment of the jupyter notebook system for tutorials and as a teaching tool for solid Earth geoscience. Underworld2 is an open-source, particle-in-cell finite element code tuned for large-scale geodynamics simulations. The numerical algorithms allow the tracking of history information through the high-strain deformation associated with fluid flow (for example, transport of the stress tensor in a viscoelastic, convecting medium, or the advection of fine-scale damage parameters by the large-scale flow). The finite element mesh can be static or dynamic, but it is not constrained to move in lock-step with the evolving geometry of the fluid. This hybrid approach is very well suited to complex fluids which is how the solid Earth behaves on a geological timescale. The development team will present the Underworld software platform with case studies including mantle convection, subduction dynamics, surface processes and ground water process modelling.

Analysis of  drill-hole  data  is  essential  for  building  the  models  that  capture  physical  variability  in  the  ore-body  and  can  be used for the optimization of mine planning and the prediction of minerals composition in the drill-core data. However, multivariate complexity and compositional data constraints can make this analysis challenging. We present a computational framework for unsupervised machine learning that uses methods such as principle components analysis (PCA) and clustering methods for analyzing drill-cores in order to provide summary statistics for drill-core evaluation/comparison. The goal of PCA is to facilitate the interpretation and explanation of the underlying data structure from drill-core data. The framework provides data visualization via clustering method such as k-means and agglomerate clustering to identify element concentrations in distinct groups and captures the relationships with other groups. We apply the method to selected drill-core data for mineral composition across different depths of Yamaranagold ore-body from Northern Australia.

Many researches showed that stress changes had important effects on the conception and triggering of earthquakes, while observations of groundwater (including water level, temperature, chemistry, soil gas, etc.) could obtain information of regional stress changes and faults activity. In this paper, the information of underground fluid and stress change were collected, the coupling mechanism of stress change and fluid change was summarized, the abnormal change characteristics of earthquakes precursor fluid in Sichuan and Yunnan regions were collected. Finally, we concluded the relationship between stress change and fluid change, and the coupling mechanism of stress change and fluid change. Combined with the analysis of the material parameters of aquifer media, it would be better to study the abnormal mechanism of precursors bred by earthquakes and had a good foundation for earthquake predictions in Sichuan and Yunnan regions.

Precursors for subsurface rock failure process (i.g. natural earthquakes and induced instabilities (NEII)) and the predictability of NEII have been debated for decades years. The main cause of this situation is due to the ambiguous relationship between the ground surface-based observed abnormal signals and the occurred NEIIs. Based on the advancing techniques and the long term exploration of observation approaches for earthquake monitoring networks, we designed a new 4D arrays of multiple disciplinary to capture signals that register the process from deformation to failure of subsurface rocks. The 4D array includes a surface seismometer array, a bore hole seismometer array, a bore hole thermometer array and a hydraulic sensor for groundwater level. At first, all the above arrays are portable on site not only from horizontal distributed points but also from vertical distributed points in a borehole. Through testing observation to get the 3D velocity profile and its change with time, the best observation strategy for permanent stations then can be determined. At least 3-4 such 3D arrays can be designed in a specific tectonic unit (tectonic block or basin and active fault system) to capture the signals from imminent deformation to failure process both from fluid media to solid rocks from different approaches. And it can supply an opportunity to study the interaction of fluid and rocks in details. This innovative idea had been evaluated by a group of specialists in 2019 through applying for an international co-operation project from the Ministry of Science and Technology of China.

Observed earthquake precursors deduced by Chinese scientific literature in past decades have been considered. Geographic location of recorded data has been compared with other geophysical parameters which includes seismicity, tectonic regimes, heat flow, crustal deformations and the occurrence of geofluids. Relations between considered geophysical parameters and fluid-related possible precursory phenomena have been discussed. Permeability variations at depth deduced by considered parameters and due to crustal deformative processes have been modelled for most relevant geological contexts of China. Engineering proper observational networks as well as data interpretation could benefit from adopted investigating methods. Most relevant suggests for Chinese context are reported. 

The 2018 Sunda Strait tsunami was generated by massive landslide following the eruption of Mount Anak Krakatau. There were large numbers of casualties as the current tsunami warning system did not cover such non-seismic tsunamis. Since then, one tsunami sensor has been installed near Krakatau but still this region is still facing tsunami treats from both seismic and non-seismic sources and corresponding tsunami warning and evacuation related issues. This study simulated tsunami hazards as the potential large seismic tsunamis including the worst scenario from Java Trench with 400 years accumulated slip and potential submarine landslides around southwest of Krakatau. Based on the simulation results, Kota Agung is found to be a location where the maximum tsunami amplitude occurs from most scenarios because of its geographical setting. When both seismic and non-seismic sources happened at the same time, combination between the first receding wave (seismic source) and the first pushing wave (non-seismic source) may increase or decrease maximum wave amplitude which as to be considered for more detail in the future. Although arrival time is different for each seismic source but earlier in overall for non-seismic sources. Focusing on a major industrial area in Ciwandan, more or less 10 m of tsunami arrives after 78 min, 40 min and 62 min for seismic, non-seismic and combined sources respectively. To increase lead time for early warning purpose, it is suggested that an observation sensor can be deployed near Panitan Island so that about 15 min more lead time can be observed. However, this will still not enough for termination of production process and horizontal evacuation of staffs in the industries. Therefore, vertical evacuation plan together with consideration of seawall construction shall be considered in the future.

Structural damage by tsunami impacts has garnered considerable interest since the advent of the 2011 Great East Japan tsunami. Structural damage is most commonly quantified by using tsunami damage fragility functions, which describe the probabilities of damage exceedance for a given flow characteristic or demand parameter. Observed inundation depths are often used an indicator of hazard intensity and are amongst the most widely used parameters to quantify damage. There are, however, other flow characteristics such as current velocity and hydrodynamic force that may offer a better explanation for structural damage. In this study, the objective is two-fold - (i) to construct damage fragility functions for port industries, using the 2011 Great East Japan tsunami as a case study and (ii)to examine the relative influence of tsunami flow parameters. For the first time, damage to port structures are being assessed for tsunami events. Approximately 5,000 port structures were assessed for damage and the damage information were catalogued into a database. Using the newly developed damage database, tsunami fragility functions were developed for eight common port industries. Numerical simulation of the 2011 tsunami was performed to derive other tsunami flow parameters such as velocity and hydrodynamic force, and these parameters were evaluated for their relative influence on the damage observed in port structures.

This study addressed two objectives based on the 2011 Great East Japan tsunami: to determine whether the maximum values of the flow depth and flow velocity are the same as their critical values and, secondly, to verify which combination of the parameters is the best predictor of the building damage level based on the decision tree classification algorithms. The data from 18,000 buildings in Ishinomaki City, Japan, which were damaged by the 2011 Great East Japan tsunami, were used to analyze. The critical variables were the simulated data at the time when the buildings collapsed. The analysis showed the accuracy of the prediction based on the group of variables. Finally, the findings showed that the combination of the critical flow depth and maximum flow velocity provided the highest accuracy for classifying the level of building damage.

Flooding frequently causes large damages with loss of life, property, and infrastructures. Reliable flood loss estimation techniques help governments in planning investments for risk reduction, and private insurers to devise policies covering flood losses, both of which advance community preparedness against the flood events. Flood inundation maps, maximum exposure values, and normalized depth damage curves (DDC) are the primary elements of flood loss estimation. A comprehensive literature survey on DDC in South East Asian countries identified that with the exception of the agriculture sector, only a few local scale DDC by residential, commercial, and industrial sectors were developed and used. The lack of a uniform DDC curve is a major constraint for estimating flood losses in cities like Jakarta, Indonesia, which experience frequent flooding and large economic losses. This study aims to derive city and national scale DDCs for residential, commercial, and industrial sectors while benchmarking such curves with the historical event reported loss for Jakarta using pre-developed flood map and maximum exposure values. These flood maps are developed using hydrology and hydraulics modelling, while maximum loss values are from Economic Exposure Data developed by Institute of Catastrophe Risk Management. It is also noted that the gradient of DDC applicable to major cities is expected to be mild compared to the rest of the country as cities are likely to follow better building standards. The results illustrate the sensitivity of the DDCs in flood loss estimation, indicating the need for land-cover based DDCs for rural and semi-urban regions. Such improved DDCs are then readily applicable in city/national scale flood loss estimation, facilitating decision making by authorities and by the insurance industry.

During flood emergencies, the most effective way to avoid human damage is evacuation to a shelter in advance, avoiding possible entrapment due to inundation. This action can be considered as proactive evacuation. However, there are instances when this behavior is not observed by majority of the population at risk. People put themselves in dangerous situations when they are left to take reactive evacuations – possibly prompted by inundation in the home. When Typhoon Hagibis struck a large part of eastern Japan, Tohoku region experienced record-breaking rainfall prompting highest level rain warnings and evacuation orders to be issued in several areas. In this study, we take the case of flood-affected towns in Miyagi Prefecture. The areas of Marumori, Osato and Osaki are located within the vicinity of major river systems that have been flooded during past heavy rains. During Typhoon Hagibis, the three areas have suffered extensive physical flood damage, and five people from Marumori lost their lives due to the inundation. Similar to previous flood events, authorities faced the challenge of getting people to evacuate. In total, only 3% of those under orders in Japan evacuated to shelters. The objective of this study is to examine the residents’ evacuation behavior. Specifically, we want to identify the factors that influence evacuation decision-making and determine the effectiveness of information sources in prompting earlier evacuations. The findings clarify which factors prompt proactive and reactive evacuations. Due to the increase in frequency of heavy rains in recent years, learning people’s behavior during emergencies becomes increasingly valuable especially to local governments. This study aims to raise people’s awareness on the importance of early evacuation in protecting their own lives.

Southeast Florida is a low-lying area and highly susceptible to permanent inundation from sea-level rise. Therefore, sea-level rise has the potential to disrupt the region’s strong Gross Regional Product, which has far outperformed that of the nation over the past 40 years. Among the four counties in Southeast Florida, Miami-Dade has the highest economic activity. The County’s roadways have the highest traffic volume per lane mile; hence, the resiliency of the county’s roadways against climate stressors is critical. The rising groundwater table due to sea-level rise is potentially damaging to the base and subbase layers of the roadways, thereby reducing the roadway service life. The existing vulnerability maps that used the bathtub method to model sea level rise can significantly underestimate the impact of sea-level rise because it does not account for rising groundwater levels and diminished drainage capacity. Enhanced vulnerability map is utilized to identify roadway segments at several categories of risk caused by different factors such as reduced subgrade strength due to groundwater table variation and traffic loading. A County-scale groundwater flow modeling results of Miami-Dade was used to identify road infrastructure subgrade layer that will be affected by rising groundwater table. Multilayer elastic pavement model was used to simulate typical pavement layers for several functional classifications of roadway and modulus resilient of the subgrade was adjusted according to rising groundwater table due to sea-level rise. An updated vulnerability map will be a crucial component to identify effective adaptation solution and resource allocation to preserve pavement service life.

This study seeks to illustrate on how both climate change and flood risk data could be utilised to better understand the impact of climate change on future flood risks and losses. Its aim is to investigate the changes in economic and property losses under future climate change, with a focus on Townsville, in Queensland, Australia. A database to identify the value of properties located in Townsville would be built using a Geographic Information System software (ArcGIS) by gathering data on the value of buildings, contents and business interruption for residential and commercial properties where possible. JBA Risk Management has established an annual damage ratio (ADR) for properties in Australia to represent the annualised cost of flood damage at any single location. An average annual loss to flood risk is derived when ADR is multiplied to the property value.  To investigate changes in the losses, daily temperature and precipitation data for Australia from the ACCESS1-0 GCM is extracted for each future period (2040-2050, 2070-2080 and 2090-2100) under the RCP4.5 and 8.5. The precipitation and temperature change factors for each future period are then applied to adjust the hazard data, used in generating the baseline ADR, to determine the changes in flood event frequency and severity. A projected property value would be generated and added to the property database set up for Townsville; the adjusted ADR would then be applied to derive future projected losses at the property level. The climate change version of ADR would allow users (e.g., insurers, financial institutes) to better consider climate risks into existing management framework and to facilitate risk-reflecting strategies in light of climate change.  

The 2004 Indian Ocean tsunami disaster devastated the community in Aceh Province, and it also accelerated the end of the 30-year civil war in this deeply religious region. In just a few hours an estimated 167,000 humans perished or dissapeared, and about 65-70% of the casualties were female. Statistical data and quantitative information are often used to describe the general characteristics of a disaster. But to truly communicate the impact of a disaster it is imperative to listen to the local communities that experienced the disaster first-hand. A discussion of a disaster of such dimensions is fraught with human, social, political, religious, communication and scientific issues.
Using the documentary "Tsunami of New Dreams" as a case study, this presentation examines methodologies and techniques for investigating a disaster from the inside out, putting qualitative information aside and focusing on testimonies from the local community that are meaningful in their local context. Insights gained from this approach can hopefully be useful to minimize the impact of future disasters.