Georisk special issue in Honour and Memory of Professor Tien H. Wu

Professor Emeritus Tien H. Wu passed away on 7 June 2018. This Georisk special issue was initiated by Georisk and ISSMGE TC304 to commemorate the passing of a founding member of the georisk community who has been instrumental in bringing probabilistic methods to geotechnical practice over a distinguished career of more than 50 years. He is fondly addressed as TH by many of his friends and colleagues.

The invited papers in this special issue will also be presented in a Special Session in Honour and Memory of Professor Tien H. Wu in the 7th International Symposium on Geotechnical Safety and Risk (7th ISGSR), Taipei, 11–13 December 2019.

TH was a Professor Emeritus of The Ohio State University (he retired from OSU in 1994). His distinguished academic career includes 12 years of services as Assistant Professor, Associate Professor, and then Professor of Civil Engineering at Michigan State University, and 30 years of services as Professor of Civil Engineering at Ohio State University. During these years, he was also a visiting professor at various institutions including the Norwegian Geotechnical Institute, the Royal Institute of Technology at Stockholm, the National University of Mexico, and the Forest Research Institute of New Zealand. In addition, he served as a United Nations Consultant in India and worked on World Bank projects in China and Ecuador.

For his life-long contributions to the profession of Civil Engineering, TH was conferred as an Honorary Member of ASCE in 2003. This is the highest honour one can receive from ASCE.

TH’s research work covers a broad spectrum of geotechnical subjects including strength properties of soil and rock, glaciology in Alaska and Antarctica, stability of embankments and natural slopes, ground water and seepage, soil-structure interaction of buried structures, risk and reliability assessment for foundations and slopes, and soil reinforcement. His work is of very high quality; among the many awards TH has received is the ASCE State-of the-Art Award in 1990 for his paper on “Reliability of Offshore Foundations”. Throughout his career, TH has published many outstanding practical papers that documented case histories. Examples include a paper on “prediction and mapping of landslide hazard” (Canadian Geotechnical Journal, Vol. 37, pp. 781–795, 2000) and another paper on “stability of shale embankment and slopes” (ASCE Journal of Geotechnical Engineering, Vol. 119, pp. 127–146, 1993). His strength and interest in documenting case histories and innovative solutions to geotechnical problems perhaps could be attributed to his early and continuous learning from Professor Peck. After all, TH was Professor Peck’s first PhD from University of Illinois at Urbana-Champaign (UIUC).

TH is a pioneer in probabilistic methods in geotechnical engineering. His work in this field started in the early 1960s and extended over the next five decades. The lead paper in this special issue by Baecher and Christian (2019) unveils a captivating historical overview of TH’s influential role in shaping the early development of geotechnical reliability. His early papers in the area of probabilistic methods in geotechnical engineering such as the “Probability of foundation safety”, “Safety analysis of slopes”, “Probabilistic analysis of seepage”, and “Uncertainty, safety, and decision in soil engineering” provided important readings that guided the study of geotechnical reliability for decades. The introduction of probabilistic methods to the geotechnical engineering community has not been easy, and the adoption of the probabilistic approaches in geotechnical practice has been very slow. The adoption of probabilistic approaches can only be enhanced if more good examples, such as TH’s case histories, are made available. In this regard, TH’s persistence in his research and publication on this subject over the last five decades has made a tremendous contribution. Many of the case histories that were documented in his reports and papers can be used as key examples toward a common goal of improving the use of probabilistic tools in our profession. His research was recognised by the Ralph B. Peck Award from the Geo-Institute (2008) among other accolades.

The guest editors invited contributions that emphasise how case histories and data can enhance decision making when mediated by probabilistic methods that account for more realistic features of geotechnical data such as property uncertainties, model uncertainties, spatial variabilities, multivariate correlations between properties, and others. We believe TH’s spirit will live on in this type of use-inspired research grounded on data that he has tirelessly championed in his professional career. In the era where data is recognised as the “new oil”, it makes sense for us to lean towards decision making strategies that are more responsive to data, particularly if we have zettabytes coming our way (Phoon 2020). It does not make sense to reduce one zettabyte to a few bytes describing a single cautious value. It does not make sense to expect big data to be “perfect” as demanded by the classical deterministic world view. Site-specific data are multivariate, uncertain and unique, sparse, incomplete, and possibly corrupted (MUSIC) (Phoon 2018) while generic data from multiple sites are big and not directly applicable to any one site (Big Indirect Data or BID) (Phoon, Ching, and Wang 2019). There is value in data of any kind (good or not so good quality, or right or wrong fit to a physical model) and the challenge is for the new generation of researchers to uncover this value, be it using probabilistic, machine learning, or other data-driven methods, and to re-imagine the role of the geotechnical engineer in an immersive environment likely to be imbued by machine intelligence. Perhaps for the first time, geotechnical engineers are getting re-acquainted with data and beginning to see value beyond inputs for physical models, measured responses from load tests for design checks, or monitoring information for the observational approach. In our opinion, we are at the threshold of an exciting future where digital and physical realities merge. The work of TH and other pioneers in geotechnical reliability established a solid foundation for the next revolution to take place.

This special issue consists of 15 papers authored by some of the leading researchers in the field. The lead paper by Baecher and Christian (2019) entitled “TH Wu and the origins of geotechnical reliability” traced TH’s seminal insights that shaped the early evolution of geotechnical reliability through the case histories he published. The authors observed that TH was one of the first Bayesianist in geotechnical engineering. The frequentist approach dominated much of the literature in the mid-twentieth Century until the relatively recent emergence of deep learning that re-ignited interest in the power of the Bayesian approach to “learn” from uncertain data (Deep Bayesian). One could say that TH was ahead of his times. In an email to the lead guest editor, Professor Baecher reflected that:

TH Wu was a pioneer and significant contributor to the early practical uses of probabilistic and statistical methods in geotechnical engineering, for which he is – we think – inadequately remembered, and this contribution attempts to refresh our memories of his many and varied contributions. Unlike other early contributors, TH’s work was influenced by important case studies, many of which were local to him in Ohio and the American Midwest.

The remaining 14 papers can be grouped into statistical analysis of soil properties (Bilgin, Arens, and Dettloff; Feng and Vardanega; Knuuti and Länsivaara; and Uzielli and Mayne), characterisation and application of model uncertainties (Chahbaz, Sadek, and Najjar; Lesny; Phoon and Tang), and probabilistic case studies (Abdulla, Sousa, Einstein, and Awadalla; Hicks, Varkey, van den Eijnden, de Gast, and Vardon; Huang, Zeng, and Kelly; Low; Mostofi, Gilbert, Montgometry, and Wartman; Shen, Juang, Ku, and Khoshnevisan; Zhu and Zhang). The key highlights are briefly presented below.

1. Statistical analysis of soil properties

Bilgin, Arens, and Dettloff (2019) examined the Standard Penetration Test (SPT) and Cone Penetration Test (CPT) results obtained from two sites located in the State of Ohio in the United States. Six CPT soundings were conducted adjacent to the SPT locations. N₆₀ values estimated from CPT results were compared to the measured SPT-N₆₀ values, with 64 data points. The results revealed that the CPT estimated N₆₀ values, on average, were 20% higher for sandy soils and 40% lower for cohesive soils compared to the SPT measured N₆₀ values. The results also showed a good correlation between the average CPT-N₆₀/SPT-N₆₀ ratios sorted by CPT soil behaviour type (SBT).

Feng and Vardanega (2019) used the database FG/KSAT-1358 to investigate the best fit probability density functions for the negative natural logarithm of saturated hydraulic conductivity as well as for void ratio, liquid limit and water content ratio. FG/KSAT-1358 contains 1358 experimental values of saturated hydraulic conductivity representative of a wide range of fine-grained soils. The values are obtained from a variety of test methods: falling head; constant head; flow pump and consolidation testing. Using the Akaike Information Criterion, the best fit distributions found were: the lognormal for the void ratio; loglogistic for the liquid limit and water content ratio and logistic for −ln[k_sat(m/s)]. The statistical distribution of saturated hydraulic conductivity is a critical input for some geotechnical stochastic analyses such as slope stability studies.

Knuuti and Länsivaara (2019) studied the transformation uncertainties for some transformation models for the determination of the undrained shear strength from a piezocone (CPTU). The study included four test sites in Finland on soft clay (Lempäälä, Masku, Murro and Perniö), containing 3–5 borings each, with a total of 4450 data points. The results showed generally low uncertainties for all models, but models based on the net cone resistance and pore pressure gave lower uncertainties than a effective cone resistance based model.

Uzielli and Mayne (2019) developed new probabilistic correlations for assigning design values of effective friction angle of clean to silty sands from cone penetration test data applying quantile regression to data from a high-quality database including results of 57 triaxial tests from 27 different sands from sites in Ireland, USA, Italy, Taiwan, Poland, Hong Kong, Norway, Japan, Canada, New Zealand and the North Atlantic. Specimens were obtained primarily from special one-dimensional freezing methods as well as newer gel samplers, Mazier tubes, and other techniques. Their new correlations allow the selection of design values of effective friction angle for any desired probability of exceedance and, thus, for any desired level of conservatism.

2. Characterisation and application of model uncertainties

Chahbaz, Sadek, and Najjar (2019) quantified the uncertainty in the bond stress mobilised at target values of anchor slip to aid the serviceability limit state design of anchored wall systems. A database of 70 anchor tests in three geologic settings (32 in Limestone, 26 in Marl, and 12 in Clay) was assembled from 28 sites around Beirut. The bond stress – displacement relationship is modelled with a hyperbolic model and the statistics of the model are derived for the geologic units. The statistics of the model parameters can be used to simulate bond stress – slip curves that represent the expected variability in the anchor response.

Lesny (2019) presented a case study on probability-based derivation of resistance factors for ultimate limit state (ULS) prediction of shallow foundations under combined loading. The underpinning of this study was the model uncertainty assessment with a comprehensive database including more than 500 load tests mainly for non-cohesive soils from field and laboratory tests under different scales and conditions. Identified dependencies of the model factor statistics on the type of loading and the soil friction angle led to different resistance factors established from Monte Carlo Simulation for natural and for controlled soil conditions and different loading situations that were implemented in a LRFD code.

Phoon and Tang (2019) evaluated the performance of the Chin-Davisson method to interpret pile axial capacity and characterise the statistics of the model factor that is a ratio between the interpreted capacity and the calculated one. For steel H-piles under compression, 177 static load tests were used to calculate the model statistics where each measured curve was artificially truncated at the maximum load equal to 50%, 75%, and 90% of the Davisson’s capacity. The results demonstrated that extrapolation from load test terminated 75% or higher of the Davisson’s capacity is reasonable.

3. Probabilistic case studies

Abdulla et al. (2019) presented a probabilistic, nonlinear subsurface characterisation model, based on 3D Gaussian distributions. The model is characterised by its ability to provide a comprehensive uncertainty representation with minimal involvement of the user required. It was applied to the case of the Masdar City Subsurface in Abu Dhabi, United Arab Emirates for the identification of Sabkha, a problematic geologic layer prone to dissolution. The results are in the form of probabilistic Sabkha location profiles and maps, which can serve as basis of hazard and risk analysis and can aid decision makers to properly design the infrastructure and mitigate potential risks.

Hicks et al. (2019) reported a case study involving the assessment and re-design of an existing dyke in the Netherlands. They compared a reliability-based random finite element analysis consistent with the requirements of Eurocode 7 with a deterministic analysis based on 5-percentile property values. The results showed that a consideration of the spatial nature of soil variability generally leads to higher computed factors of safety and, for those dyke sections requiring remedial action, to more economic designs. Back-figured characteristic values were shown to be considerably higher than the 5-percentile soil properties; hence, a reduction in over-conservatism was achieved.

Huang, Zeng, and Kelly (2019) used Bayesian updating to predict the long-term settlement of the Teven Road trial embankment built for the Pacific Highway upgrade project in the coastal regions of Eastern Australia. The embankment was built in two stages with total filling of 1.8 and 3.8 m respectively. The prediction based on posterior is better than that based on prior information. However, because the first stage (settled for 6 years) did not carry the information for the second stage (settled for another 1.6 years), using only monitored data from the first stage could not predict well the second stage settlement.

Low (2019) analysed probabilistically an underwater slope in San Francisco Bay Mud that failed during excavation and a rock slope in Hong Kong which had been identified as potentially unstable, using the first-order reliability method (FORM) and reliability-based design (RBD). It is emphasised that the design point in RBD-via-FORM automatically reflects complexly intertwined and case-specific parametric sensitivities, correlations, and hidden subtleties, thereby complementing partial factor design approaches like LRFD and Eurocode 7 by overcoming some limitations of LRFD and Eurocode 7. Such merits of RBD-via-FORM may not be possessed by other probabilistic approaches like FOSM and Monte Carlo simulations.

Mostofi et al. (2019) applied a new methodology based on the Theory of Decision Entropy to assess the recurrence probability of the 2014 landslide near Oso Washington, a relatively rare but consequential natural hazard. Given the information available today, if the risk for another landslide is accepted, then the expected time between occurrences of massive landslides at this location is about 2000–3000 years, the mean occurrence rate tends to increase with time since the last occurrence, and the alternative of avoiding the risk is preferred if the present worth cost to prevent development for 100 years is less than about 1/6 the cost of another massive landslide.

Shen et al. (2019) demonstrated how simple statistical methods could be used for assessing the efficacy of dynamic compaction (DC) in mitigating the liquefaction hazard through a case study. A rare data set of 27 CPT soundings each before and after DC was collected and analysed in this study. The results showed that the efficacy of DC on liquefaction mitigation could be effectively demonstrated and communicated through maps of liquefaction probability generated by geostatistical methods coupled with visualisation technique. This work demonstrated the value of statistical methods, even in a simple form, in a geotechnical project.

Zhu and Zhang (2019) proposed a random field model to characterise hydrological influences of vegetation on the stability of soil slopes and quantify the influence of uncertainty in vegetation transpiration on the pore-water pressure distributions in and the stability of a vegetated slope. The maximum transpiration rates for a wide range of sites and vegetation species are summarised. Three idealised types of root geometry, i.e. uniform, triangular and parabolic types, are considered. The uniform root geometry shows the most significant impact on enhancing the slope factor of safety through transpiration, compared with the triangular and parabolic geometry types.

The Guest Editors would like to thank Professor Limin Zhang, Editor-in-Chief of Georisk and Professor Jianye Ching, Chair of ISSMGE TC304 for supporting this special issue. The excellent contributions from the invited authors, including the heartfelt words penned by some of his friends and colleagues in a section “What do I remember about Professor Wu” at the end of some papers, are deeply appreciated. Last but not least, this special issue owes much of its success to the meticulous management of the review process by Professor Zijun Cao, Assistant Editor of Georisk, and the generous assistance of many experts who provided timely and helpful reviews at short notice.

ORCID

Kok-Kwang Phoon http://orcid.org/0000-0003-2577-8639