Natural catastrophes are increasing worldwide. They are becoming more frequent but also more severe and impactful on our built environment leading to extensive damage and losses. Earthquake events account for the smallest part of natural events; nevertheless seismic damage led to the most fatalities and significant losses over the period 1981-2016 (Munich Re). Damage prediction is helpful for emergency management and the development of earthquake risk mitigation projects. Recent design efforts focused on the application of performance-based design engineering where damage estimation methodologies use fragility and vulnerability functions. However, the approach does not explicitly specify the essential criteria leading to economic losses. There is thus a need for an improved methodology that finds the critical building elements related to significant losses. The here presented methodology uses data science techniques to identify key building features that contribute to the bulk of losses. It uses empirical data collected on site during earthquake reconnaissance mission to train a machine learning model that can further be used for the estimation of building damage post-earthquake. The first model is developed for Christchurch. Empirical building damage data from the 2010-2011 earthquake events is analysed to find the building features that contributed the most to damage. Once processed, the data is used to train a machine-learning model that can be applied to estimate losses in future earthquake events.
Background Liquefaction induced land damage has been identified in more than 13 notable New Zealand earthquakes within the past 150 years, as presented on the timeline below. Following the 2010-2011 Canterbury Earthquake Sequence (CES), the consequences of liquefaction were witnessed first-hand in the city of Christchurch and as a result the demand for understanding this phenomenon was heightened. Government, local councils, insurers and many other stakeholders are now looking to research and understand their exposure to this natural hazard.
This paper presents on-going challenges in the present paradigm shift of earthquakeinduced ground motion prediction from empirical to physics-based simulation methods. The 2010-2011 Canterbury and 2016 Kaikoura earthquakes are used to illustrate the predictive potential of the different methods. On-going efforts on simulation validation and theoretical developments are then presented, as well as the demands associated with the need for explicit consideration of modelling uncertainties. Finally, discussion is also given to the tools and databases needed for the efficient utilization of simulated ground motions both in specific engineering projects as well as for near-real-time impact assessment.
This thesis presents the application of data science techniques, especially machine learning, for the development of seismic damage and loss prediction models for residential buildings. Current post-earthquake building damage evaluation forms are developed for a particular country in mind. The lack of consistency hinders the comparison of building damage between different regions. A new paper form has been developed to address the need for a global universal methodology for post-earthquake building damage assessment. The form was successfully trialled in the street ‘La Morena’ in Mexico City following the 2017 Puebla earthquake. Aside from developing a framework for better input data for performance based earthquake engineering, this project also extended current techniques to derive insights from post-earthquake observations. Machine learning (ML) was applied to seismic damage data of residential buildings in Mexico City following the 2017 Puebla earthquake and in Christchurch following the 2010-2011 Canterbury earthquake sequence (CES). The experience showcased that it is readily possible to develop empirical data only driven models that can successfully identify key damage drivers and hidden underlying correlations without prior engineering knowledge. With adequate maintenance, such models have the potential to be rapidly and easily updated to allow improved damage and loss prediction accuracy and greater ability for models to be generalised. For ML models developed for the key events of the CES, the model trained using data from the 22 February 2011 event generalised the best for loss prediction. This is thought to be because of the large number of instances available for this event and the relatively limited class imbalance between the categories of the target attribute. For the CES, ML highlighted the importance of peak ground acceleration (PGA), building age, building size, liquefaction occurrence, and soil conditions as main factors which affected the losses in residential buildings in Christchurch. ML also highlighted the influence of liquefaction on the buildings losses related to the 22 February 2011 event. Further to the ML model development, the application of post-hoc methodologies was shown to be an effective way to derive insights for ML algorithms that are not intrinsically interpretable. Overall, these provide a basis for the development of ‘greybox’ ML models.
Heathcote Valley school strong motion station (HVSC) consistently recorded ground motions with higher intensities than nearby stations during the 2010-2011 Canterbury earthquakes. For example, as shown in Figure 1, for the 22 February 2011 Christchurch earthquake, peak ground acceleration at HVSC reached 1.4 g (horizontal) and 2 g (vertical), the largest ever recorded in New Zealand. Strong amplification of ground motions is expected at Heathcote Valley due to: 1) the high impedance contrast at the soil-rock interface, and 2) the interference of incident and surface waves within the valley. However, both conventional empirical ground motion prediction equations (GMPE) and the physics-based large scale ground motions simulations (with empirical site response) are ineffective in predicting such amplification due to their respective inherent limitations.
Semi-empirical models based on in-situ geotechnical tests have become the standard of practice for predicting soil liquefaction. Since the inception of the “simplified” cyclic-stress model in 1971, variants based on various in-situ tests have been developed, including the Cone Penetration Test (CPT). More recently, prediction models based soley on remotely-sensed data were developed. Similar to systems that provide automated content on earthquake impacts, these “geospatial” models aim to predict liquefaction for rapid response and loss estimation using readily-available data. This data includes (i) common ground-motion intensity measures (e.g., PGA), which can either be provided in near-real-time following an earthquake, or predicted for a future event; and (ii) geospatial parameters derived from digital elevation models, which are used to infer characteristics of the subsurface relevent to liquefaction. However, the predictive capabilities of geospatial and geotechnical models have not been directly compared, which could elucidate techniques for improving the geospatial models, and which would provide a baseline for measuring improvements. Accordingly, this study assesses the realtive efficacy of liquefaction models based on geospatial vs. CPT data using 9,908 case-studies from the 2010-2016 Canterbury earthquakes. While the top-performing models are CPT-based, the geospatial models perform relatively well given their simplicity and low cost. Although further research is needed (e.g., to improve upon the performance of current models), the findings of this study suggest that geospatial models have the potential to provide valuable first-order predictions of liquefaction occurence and consequence. Towards this end, performance assessments of geospatial vs. geotechnical models are ongoing for more than 20 additional global earthquakes.
1. Background and Objectives This poster presents results from ground motion simulations of small-to-moderate magnitude (3.5≤Mw≤5.0) earthquake events in the Canterbury, New Zealand region using the Graves and Pitarka (2010,2015) methodology. Subsequent investigation of systematic ground motion effects highlights the prediction bias in the simulations which are also benchmarked against empirical ground motion models (e.g. Bradley (2013)). In this study, 144 earthquake ruptures, modelled as point sources, are considered with 1924 quality-assured ground motions recorded across 45 strong motion stations throughout the Canterbury region, as shown in Figure 1. The majority of sources are Mw≥4.0 and have centroid depth (CD) 10km or shallower. Earthquake source descriptions were obtained from the GeoNet New Zealand earthquake catalogue. The ground motion simulations were performed within a computational domain of 140km x 120km x 46km with a finite difference grid spacing of 0.1km. The low-frequency (LF) simulations utilize the 3D Canterbury Velocity Model while the high-frequency (HF) simulations utilize a generic regional 1D velocity model. In the LF simulations, a minimum shear wave velocity of 500m/s is enforced, yielding a maximum frequency of 1.0Hz.
Background This study examines the performance of site response analysis via nonlinear total-stress 1D wave-propagation for modelling site effects in physics-based ground motion simulations of the 2010-2011 Canterbury, New Zealand earthquake sequence. This approach allows for explicit modeling of 3D ground motion phenomena at the regional scale, as well as detailed nonlinear site effects at the local scale. The approach is compared to a more commonly used empirical VS30 (30 m time-averaged shear wave velocity)-based method for computing site amplification as proposed by Graves and Pitarka (2010, 2015), and to empirical ground motion prediction via a ground motion model (GMM).
The Canterbury Earthquake Sequence 2010-2011 (CES) induced widespread liquefaction in many parts of Christchurch city. Liquefaction was more commonly observed in the eastern suburbs and along the Avon River where the soils were characterised by thick sandy deposits with a shallow water table. On the other hand, suburbs to the north, west and south of the CBD (e.g. Riccarton, Papanui) exhibited less severe to no liquefaction. These soils were more commonly characterised by inter-layered liquefiable and non-liquefiable deposits. As part of a related large-scale study of the performance of Christchurch soils during the CES, detailed borehole data including CPT, Vs and Vp have been collected for 55 sites in Christchurch. For this subset of Christchurch sites, predictions of liquefaction triggering using the simplified method (Boulanger & Idriss, 2014) indicated that liquefaction was over-predicted for 94% of sites that did not manifest liquefaction during the CES, and under-predicted for 50% of sites that did manifest liquefaction. The focus of this study was to investigate these discrepancies between prediction and observation. To assess if these discrepancies were due to soil-layer interaction and to determine the effect that soil stratification has on the develop-ment of liquefaction and the system response of soil deposits.
This study explicitly investigates uncertainties in physics-based ground motion simulation validation for earthquakes in the Canterbury region. The simulations utilise the Graves and Pitarka (2015) hybrid methodology, with separately quantified parametric uncertainties in the comprehensive physics and simplified physics components of the model. The study is limited to the simulation of 148 small magnitude (Mw 3.5 – 5) earthquakes, with a point source approximation for the source rupture representations, which also enables a focus on a small number of relevant uncertainties. The parametric uncertainties under consideration were selected through sensitivity analysis, and specifically include: magnitude, Brune stress parameter and high frequency rupture velocity. Twenty Monte Carlo realisations were used to sample parameter uncertainties for each of the 148 events. Residuals associated with the following intensity measures: spectral acceleration, peak ground velocity, arias intensity and significant duration, were ascertained. Using these residuals, validation was performed through assessment of systematic biases in site and source terms from mixed-effects regression. Based on the results to date, initial standard deviation recommendations for parameter uncertainties, based on the Canterbury simulations have been obtained. This work ultimately provides an initial step toward explicit incorporation of modelling uncertainty in simulated ground motion predictions for future events, which will improve the use of simulation models in seismic hazard analysis. We plan to subsequently assess uncertainties for larger magnitude events with more complex ruptures, and events across a larger geographic region, as well as uncertainties due to path attenuation, site effects, and more general model epistemic uncertainties.
Predictive modelling provides an efficient means to analyse the coastal environment and generate knowledge for long term urban planning. In this study, the numerical models SWAN and XBeach were incorporated into the ESRI ArcGIS interface by means of the BeachMMtool. This was applied to the Greater Christchurch coastal environment to simulate geomorphological evolution through hydrodynamic forcing. Simulations were performed using the recent sea level rise predictions by the Intergovernmental Panel on Climate Change (2013) to determine whether the statutory requirements outlined in the New Zealand Coastal Policy Statement 2010 are consistent with central, regional and district designations. Our results indicate that current land use zoning in Greater Christchurch is not consistent with these predictions. This is because coastal hazard risk has not been thoroughly quantified during the process of installing the Canterbury Earthquake Recovery Authority residential red zone. However, the Christchurch City Council’s flood management area does provide an extent to which managed coastal retreat is a real option. The results of this research suggest that progradation will continue to occur along the Christchurch foreshore due to the net sediment flux retaining an onshore direction and the current hydrodynamic activity not being strong enough to move sediment offshore. However, inundation during periods of storm surge poses a risk to human habitation on low lying areas around the Avon-Heathcote Estuary and the Brooklands lagoon.
The full scale, in-situ investigations of instrumented buildings present an excellent opportunity to observe their dynamic response in as-built environment, which includes all the real physical properties of a structure under study and its surroundings. The recorded responses can be used for better understanding of behavior of structures by extracting their dynamic characteristics. It is significantly valuable to examine the behavior of buildings under different excitation scenarios. The trends in dynamic characteristics, such as modal frequencies and damping ratios, thus developed can provide quantitative data for the variations in the behavior of buildings. Moreover, such studies provide invaluable information for the development and calibration of realistic models for the prediction of seismic response of structures in model updating and structural health monitoring studies. This thesis comprises two parts. The first part presents an evaluation of seismic responses of two instrumented three storey RC buildings under a selection of 50 earthquakes and behavioral changes after Ms=7.1 Darfield (2010) and Ms=6.3 Christchurch (2011) earthquakes for an instrumented eight story RC building. The dynamic characteristics of the instrumented buildings were identified using state-of-the-art N4SID system identification technique. Seismic response trends were developed for the three storey instrumented buildings in light of the identified frequencies and the peak response accelerations (PRA). Frequencies were observed to decrease with excitation level while no trends are discernible for the damping ratios. Soil-structure interaction (SSI) effects were also determined to ascertain their contribution in the seismic response. For the eight storey building, it was found through system identification that strong nonlinearities in the structural response occurred and manifested themselves in all identified natural frequencies of the building that exhibited a marked decrease during the strong motion duration compared to the pre-Darfield earthquakes. Evidence of foundation rocking was also found that led to a slight decrease in the identified modal frequencies. Permanent stiffness loss was also observed after the strong motion events. The second part constitutes developing and calibrating finite element model (FEM) of the instrumented three storey RC building with a shear core. A three dimensional FEM of the building is developed in stages to analyze the effect of structural, non-structural components (NSCs) and SSI on the building dynamics. Further to accurately replicate the response of the building following the response trends developed in the first part of the thesis, sensitivity based model updating technique was applied. The FEMs were calibrated by tuning the updating parameters which are stiffnesses of concrete, NSCs and soil. The updating parameters were found to generally follow decreasing trends with the excitation level. Finally, the updated FEM was used in time history analyses to assess the building seismic performance at the serviceability limit state shaking. Overall, this research will contribute towards better understanding and prediction of the behavior of structures subjected to ground motion.
Damage distribution maps from strong earthquakes and recorded data from field experiments have repeatedly shown that the ground surface topography and subsurface stratigraphy play a decisive role in shaping the ground motion characteristics at a site. Published theoretical studies qualitatively agree with observations from past seismic events and experiments; quantitatively, however, they systematically underestimate the absolute level of topographic amplification up to an order of magnitude or more in some cases. We have hypothesized in previous work that this discrepancy stems from idealizations of the geometry, material properties, and incident motion characteristics that most theoretical studies make. In this study, we perform numerical simulations of seismic wave propagation in heterogeneous media with arbitrary ground surface geometry, and compare results with high quality field recordings from a site with strong surface topography. Our goal is to explore whether high-fidelity simulations and realistic numerical models can – contrary to theoretical models – capture quantitatively the frequency and amplitude characteristics of topographic effects. For validation, we use field data from a linear array of nine portable seismometers that we deployed on Mount Pleasant and Heathcote Valley, Christchurch, New Zealand, and we compute empirical standard spectral ratios (SSR) and single-station horizontal-to-vertical spectral ratios (HVSR). The instruments recorded ambient vibrations and remote earthquakes for a period of two months (March-April 2017). We next perform two-dimensional wave propagation simulations using the explicit finite difference code FLAC. We construct our numerical model using a high-resolution (8m) Digital Elevation Map (DEM) available for the site, an estimated subsurface stratigraphy consistent with the geomorphology of the site, and soil properties estimated from in-situ and non-destructive tests. We subject the model to in-plane and out-of-plane incident motions that span a broadband frequency range (0.1-20Hz). Numerical and empirical spectral ratios from our blind prediction are found in very good quantitative agreement for stations on the slope of Mount Pleasant and on the surface of Heathcote Valley, across a wide range of frequencies that reveal the role of topography, soil amplification and basin edge focusing on the distribution of ground surface motion.
