UC Berkeley

Natural gas transmission pipelines can be affected by earthquakes from traveling seismic waves and earthquake-induced ground failure from liquefaction or landslides. Case histories of earthquake effects to natural gas transmission pipelines in California show that ground failure poses the most acute risk as no modern gas pipelines in California have been shown to rupture due to seismic waves. The OpenSRA (Open Seismic Risk Assessment) software tool has been developed through the contributions of several members of a large multidisciplinary research team to assess the seismic risk to natural gas infrastructure including below ground transmission pipelines. The tool implements the Pacific Earthquake Engineering Research (PEER) center’s Performance Based Earthquake Engineering (PBEE) methodology for assessing risk. The PEER PBEE framework assesses seismic performance at the system level by probabilistically quantifying an intensity measure (IM), such as the peak ground acceleration (PGA), and the response of the system to the IM in terms of seismic displacement or other engineering demand parameters (EDP). The EDPs are used with fragility relationships to estimate the damage to the system in terms of longitudinal pipe strain or other damage measures (DM). The DMs are used to evaluate decision variables (DV) such as the probability of pipeline rupture. Research performed through this study identified data and methods for assessing the seismic permanent ground displacement (PGD) EDP at the statewide to site-specific scales in the OpenSRA tool. Due to the differences in the types of data and methods available for estimating seismic displacement at the statewide versus the site-specific scale, four data and analysis levels were created: Level 1 analyses assess the seismic risk to natural gas transmission pipelines from ground failure using data available at a uniform resolution across the state of California. The Level 1 methods estimate potential liquefaction and landslide-induced displacements using proxies for geologic, geotechnical, and groundwater data and have very high aleatory variability (due to inherent randomness) and epistemic uncertainty (due to uncertainty that the model is correct). Level 2 analyses utilize data available at regional scales at higher resolution compared to Level 1 analyses including larger scale geologic maps, limited, generic subsurface geotechnical data, and better groundwater information. Level 2 analyses have high uncertainty, but it is less than at Level 1. Level 3 analyses utilize site-specific data including subsurface data from geotechnical borings or cone penetration tests (CPTs). The methods for estimating seismic displacement at the site-specific scale have less epistemic uncertainty and are more reliable compared to the methods utilized at Levels 1 and 2. Due to the higher quality and resolution of the data available at Level 3 and the reduced epistemic uncertainty of the Level 3 methods for estimating seismic displacement, Level 3 analyses have less uncertainty than Level 2 analyses. Level 4 analyses employ state-of-the-art numerical simulations and require advanced laboratory testing to calibrate the material constitutive models. Level 4 analyses are beyond the current scope of the OpenSRA project. Due to a lack of suitable methods available in the literature for estimating liquefaction-induced lateral spread displacement at Level 2, research focused on developing a new method for probabilistically estimating potential lateral spread displacement at regional scales. The new method collects CPTs across a region and sorts them into distinct surficial geologic deposits. The lateral displacement index (LDI) is then calculated for 225 unique combinations for peak ground acceleration (PGA), earthquake moment magnitude (Mw), and depth to groundwater (GWT). Models conditioned on the surficial geology, PGA, Mw, and GWT are developed to estimate the probability that LDI is negligible (i.e., equals “zero,” which is defined as LDI less than three) and the non-zero LDI and its uncertainty. LDI is assumed to be distributed as a mixed-random variable whereby there is a mass probability that LDI equals “zero” and a distribution of non-zero LDI. An estimated distribution of LDI is converted to a distribution of lateral spread displacement using correlations of LDI to lateral displacement conditioned on the topographic slope for gently sloping sites far from a free-face or the free-face ratio for sites near a free-face feature. The method is shown to estimate reasonably both the spatial extent and magnitude of lateral displacements for the 1989 Mw 6.9 Loma Prieta earthquake in the San Francisco Bay area of California and the 2010 Mw 7.1 Darfield and 2011 Mw 6.2 Christchurch earthquakes in the Christchurch area of New Zealand. Other research of this study focused on the longitudinal strain response of the pipelines to the seismic PGD experienced at Balboa Boulevard in the San Fernando Valley of Southern California during the 1994 Mw 6.7 Northridge earthquake. Eight pipelines, including five natural gas transmission pipelines, a natural gas distribution line, and two pressurized water trunk lines crossed the liquefaction-induced ground deformation zone produced by the 1994 Northridge earthquake. The Old Line 120 natural gas transmission pipeline, the gas distribution line, and the Granada and Rinaldi Trunk Lines broke in both tension and compression during the 1994 Northridge earthquake. The New Line 120, Line 3000, and Line 3003 natural gas transmission pipelines and the Mobil Oil Line M70 crude oil transmission pipeline did not break in 1994. No PGD was observed and no pipelines failed at the site during the 1971 Mw 6.6 San Fernando earthquake. Evidence suggests the groundwater was lower in 1971 than in 1994, which reduced the likelihood of liquefaction-induced ground movements at Balboa Boulevard in 1971. The longitudinal strains were assessed in a conventional manner using an analytical model typically used in engineering practice. The pipelines were analyzed with mean values for the soil-pipeline system properties, including: the steel yield strength, the shape of the steel stress-strain curves, the soil-pipeline interface shear stress, the pipe geometry, the length of the ground deformation zone, and the amount of seismic PGD. Critical strains for tensile rupture and compressive buckling were estimated. The results of this modeling show good agreement between the expected and observed performance of the pipelines. The pipelines that failed developed the highest longitudinal strains and the pipelines that did not fail developed significantly lower strains. In the case of one pipeline (i.e., Line 3000), however, the longitudinal strain developed in it was estimated to be sufficient to cause buckling in the compressive deformation zone, but it did not fail. There are several possible reasons for this discrepancy. The soil-pipeline interaction in the analytical model depends not only on the length of the soil block displacement, but also on the shear force conveyed to the pipeline by the adjacent soil. There is uncertainty in the soil-pipeline interface shear stress and small variations can significantly affect the strain estimate. The critical compressive strain plays an important role in predicting pipeline failure. Uncertainty in the critical strains was not evaluated in the conventional analysis. The longitudinal strain is also sensitive to the pipe steel yield stress, which was assumed to be equal to its specified minimum value, and the amount of ground displacement, for which there is significant uncertainty. The longitudinal strain response of the pipelines at Balboa Boulevard was also assessed probabilistically in the manner in which OpenSRA assesses the seismic risk of natural gas pipelines. The aleatory variability and epistemic uncertainty for each of the soil-pipeline system parameters was estimated and Monte Carlo simulations of the longitudinal strain were calculated with the validated analytical model. New fragility functions for assessing tensile rupture and compressive buckling are developed, including their aleatory variability and epistemic uncertainty. Sampling the distributions for each of the system parameters allows for a distribution of the longitudinal strain to be estimated for each pipeline. Assessing the longitudinal strain distributions with the new fragility functions results in distributions for the probability of tensile rupture and the probability of compressive buckling for each pipeline. The results of this study show good agreement between the expected and observed performance of the pipelines. The probability of compressive buckling distribution for Line 3000, which was expected to fail in the conventional analysis, varies from low to high, demonstrating the significant uncertainty in the assessment of this line. The methodology employed in the OpenSRA software is judged to be reasonable in its application to assessing the seismic performance of buried pipelines.