Proposed AASHTO Guidelines for Performance-Based Seismic Bridge Design (2020)

Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

58 AASHTO Seismic Map Update Background Information The current versions of both the AASHTO LRFD Bridge Design Specifications and the AASHTO Guide Specifications for LRFD Seismic Bridge Design use maps that are based on the 2002 U.S. Geological Survey seismic hazard modeling efforts, which are now considered out of date. As detailed in the two specifications, both versions use a standard spectral shape and site coefficients keyed to three mapped accelerations at 0.0 s (for PGA), 0.2 s, and 1.0 s to develop the design spectrum for any location, as illustrated in Figure 27. The original scope for the work in this phase was based on the use of the 2014 U.S. Geological Survey hazard model for the new maps and the 2015 NEHRP site factors to account for the effects of soil on the seismic hazards. However, there are some challenges associated with this approach. A new hazard model, the 2018 model, has been developed by the U.S. Geological Survey and represents the best understanding of the seismic hazard. Based on a better understanding of site effects across the country, application of the 2015 NEHRP site factors would likely involve more sites that will require site-specific ground response analyses than the previous site factors. Specifically, the updated NEHRP factors require perfor- mance of site-specific analyses for higher spectral accelerations in Site Classes D and E, as well as the previous requirement to conduct site-specific analyses for Site Class F. This represents a significant increase in the amount of site-specific analysis over what was previously required. The 2018 hazard model uses the current state of knowledge and includes significant advances that have occurred in earthquake hazard modeling since the 2014 efforts. The advances involve significant differences in the approach taken to the 2018 hazard model that will require consid- eration by AASHTO before deciding on a path forward. These include • A multi-period spectrum (22 periods) from 0 to 10 seconds is provided, rather than just 3 spectral periods. This allows direct application of the mapped values, rather than as input to a pre-determined spectrum shape. • Soil effects are directly included in the modeling, with input of the time-averaged shear wave velocity in the upper 30 meters (Vs30), resulting in the appropriately modified spectrum being supplied. The existing Site Classes A through E are further refined into a total of eight, including three additional Site Classes BC, CD, and DE, plus Site Class F. • Deep sedimentary basin effects are included for key areas that are strongly influenced by these effects. This is especially important in the Pacific Northwest and Southern California. • Updated attenuation models are included using the results of the NGA–East initiative, which will be especially important in the central and eastern United States. Figure 28 shows multi-period spectra for Salt Lake City for the refined site classes contained in the 2018 hazard model. As opposed to the flat-topped spectra currently used and shown in C H A P T E R 4

AASHTO Seismic Map Update 59 Figure 27. Current 3-period spectrum construction. Figure 28. Multi-period spectra from the 2018 hazard model for Salt Lake City. Figure 27, these spectra do not truncate the acceleration values in the region of peak spectral accelerations. Three main options were investigated for implementing updated mapping in the AASHTO specifications. The first option would be to proceed as scoped and utilize the 2014 mapping data and 2015 NEHRP site factors, including additional requirements for site-specific ground response analyses. The second option would be to utilize the 2018 mapping data but keep the 3-period spectrum construction and the separate soil factors based on the 2015 NEHRP site

60 Proposed AASHTO Guidelines for Performance-Based Seismic Bridge Design classes. The third option would be to utilize the 2018 data with the multi-period spectra and the soil effects built into the mapping data. Each option is more fully described as follows. Option 1—2014 Hazard Model 3-Period Spectrum Option 1 is a straightforward implementation of the 2014 hazard model, with updated site factors based on the 2015 NEHRP efforts. The overall spectrum construction method remains the same, and minimum updates to the specifications would be required. Figure 27 shows the current 3-point spectrum method. However, the updated site coefficients come with increased requirements for site-specific analyses due to an improved understanding that the coefficients themselves can be unconservative for poorer soils. No basin effects are included and no NGA– East attenuation models are included. Option 2—2018 Hazard Model 3-Period Spectrum Option 2 uses the 2018 hazard model, but only utilizes the same 3 periods of the spectrum as current methods to construct the design response spectrum. The remainder of the 22 spectral periods are not used. Otherwise, this option is the same as Option 1, using the 2015 NEHRP site coefficients or site-specific ground response analyses to include the effects of local soil conditions. Note that both Options 1 and 2 would need to be augmented with site-specific site response analyses for many Site Class D and E sites. This is because the site coefficients do not appropri- ately capture the response of poorer soils. Option 3—2018 Hazard Model Multi-Period Spectra Option 3 is to fully implement the multi-period spectra from the 2018 hazard model, as well as the refined site classes. Providing maps for all 22 periods of the spectrum is not practical, so the values would be only available electronically. The user would enter location, as well as the site class (or the shear wave velocity Vs30, the same value needed now to determine site class), and the multi-period spectrum adjusted for soil effects would be provided. Consideration needs to be given as to how the seismic zone (load and resistance factor design), seismic design criteria, and short period displacement magnification (guide specs) would be determined. There has been a method developed for the 2020 NEHRP Recommended Seismic Provisions, which results in equivalent Ss and S1 values, which can be used. Another option might include using the ratio of the 0.2 second and 1.0 second values from the multi-period spectrum. Hazard Models Considered Table 22 shows the pros and cons for each of the options considered. Given that it may be years before another update to the AASHTO specifications is undertaken and the significant advances represented by the 2018 hazard model, the best option for updating the seismic hazard contained in the AASHTO specifications is Option 3, a full implementation of the 2018 hazard model. There are challenges associated with this option, but this option provides AASHTO with the most accurate representation of the seismic hazard currently available. Map Development and Comparison Effort Based on the preceding analysis, Option 3 was determined to be the most appropriate for updating the hazard mapping in AASHTO. The 2018 U.S. Geological Survey hazard model was used to develop AASHTO-specific gridded data for the 22 periods on the response spectrum for

AASHTO Seismic Map Update 61 Soil Classes A, B, BC, C, CD, D, DE, and E. These data are contained in Guidelines for Performance- Based Seismic Bridge Design (to be published by AASHTO) and fully define the seismic hazard for a return period of 1000 years (actually, 7% probability of exceedance in 75 years). A comparison of design spectra from the current AASHTO provisions (based on the 2002 U.S. Geological Survey hazard model and 1997 NEHRP site factors) and the multi-period spectra from the 2018 U.S. Geological Survey hazard model has been made for 17 cities across the continental United States and is shown in Appendix B to this report. Comparisons for the cities listed in Table 23 were developed in terms of acceleration response spectra, which is typically used in seismic design, as well as the displacement response spectra, which has application when utilizing a nonlinear or pseudo nonlinear approach, such as the DDBD method. Comparisons of the PGA values are also presented, as these are often used in geotechnical evaluations of seismic slope stability and liquefaction triggering. The 2018 U.S. National Seismic Hazard Model eliminates the need to use a classical Newmark spectrum shape and separate site coefficients and thus results in a design spectrum that is more closely associated with the site-specific uniform hazard spectrum. This is evident for both the acceleration and displacement spectra. For example, the slope of the displacement spectra reduces as the period increases (which is consistent with actual ground motion records), as opposed to constant-slope displacement spectra without a corner period. Option Pros Cons Option 1 – 2014 Hazard Model, 3-Period Spectrum and Table of 2015 NEHRP Site Factors with More Site-Specific Ground Response Analyses Required • Easiest to implement • Familiar to owners and engineers • Would require site-specific ground response analyses to determine Fa for Ss > 1 sec in Site Class E and for S1 > 0.2 sec in Site Classes D and E • Does not utilize latest advances • Does not include basin effects or NGA–East Option 2 – 2018 Hazard Model, 3-Period Spectrum and Table of 2015 NEHRP Site Factors with More Site-Specific Ground Response Analyses Required • Relatively easy to implement • Maintains 3-period spectrum • Uses updated hazard modeling • Includes basin effects, although not for periods >1.0 seconds • Includes NGA–East • Would require site specific ground response analyses to determine Fa for Ss > 1 sec in Site Class E and for S1 > 0.2 sec in Site Classes D and E • 3-period spectrum does not result in truly uniform hazard • Justification for use of approximate 3-period spectrum is no longer valid Option 3 – 2018 Hazard Model, Multi-period Spectra • Utilizes most accurate hazard model and estimation of soil effects • Avoids need for site-specific ground response analysis for Site Classes D and E • Includes basin effects • Includes NGA–East • Provides a more accurate spectrum, without the need for approximations, which is more consistent with actual ground motion spectral shapes • Requires more extensive changes to the specifications to implement • Will be unfamiliar to engineers and owners (although this should be a minor point since actual ground motion spectral shapes are more similar to the 22-period construction) • Paper maps not practical to fully define hazard model Table 22. Comparison of options.

62 Proposed AASHTO Guidelines for Performance-Based Seismic Bridge Design Alaska and Hawaii were not included in the 2018 hazard model; however, approximate data sufficient to develop the 22-period spectra are being developed by the U.S. Geological Survey based on the latest hazard models available for these states and generic spectral shapes developed using the 2018 hazard model for the conterminous (or contiguous) United States. When evaluating the comparisons presented, it is useful to first compare the current AASHTO spectrum and the 2018 hazard model spectrum for Soil Class B. This is the reference spectrum, with no soil effects included. It has been referred to in the past as the “B-C boundary” hazard, and is intended to represent the level of motion at the top of a bedrock layer (i.e., bedrock outcrop at the ground surface) with a Vs30 of 760 m/s, without any amplification or damping of the motions that might occur as the motions are affected by overlying soil layers. By comparing these spectra, the differences in the reference hazard can be evaluated, separate from soil effects. Two trends become apparent in reviewing the acceleration spectra for Soil Class B: • For the locations examined in the western portion of the country, the hazard has generally (but not everywhere) decreased at the B-C boundary. • For the locations examined in the eastern portion of the country, the hazard levels are very similar between current AASHTO and the 2018 hazard model, although reductions in the long period spectral displacements are observed. The 2018 hazard model uses updated ground motion models (e.g., Next Generation Attenu- ation Relationships) and updated source models (e.g., UCERF3 and CEUS-SCC) for both the western and eastern United States. Differences in the Site Class B hazard are primarily due to the effect of using these models, which vary geographically and can compound or offset each other. Other factors contributing to the difference include updates to methods for handling of uncertainty and basin effects for four urban areas. Comparing the spectra for increasingly soft soils, from Soil Classes C through E, shows more dramatic changes with respect to the current AASHTO provisions. The method normally used Location Latitude Longitude Portland, OR 45.5 –122.65 San Francisco, CA 37.75 –122.4 Seattle, WA 47.6 –122.3 Reno, NV 39.55 –119.8 Los Angeles, CA 34.05 –118.25 Boise, ID 43.6 –116.2 Las Vegas, NV 36.2 –115.15 Missoula, MT 46.9 –114 Salt Lake City, UT 40.75 –111.9 St. Louis, MO 38.6 –90.2 Memphis, TN 35.15 –90.05 Paducah, KY 37.1 –88.6 Chicago, IL 41.85 –87.65 Evansville, IN 38 –87.6 Charleston, SC 32.8 –79.95 New York City, NY 40.75 –74 Boston, MA 42.35 –71.05 Table 23. Cities used in comparison effort.

AASHTO Seismic Map Update 63 in AASHTO to account for soil effects is to modify the reference hazard at the B-C boundary using site factors based on the properties of the soil in the top 30 meters at the three spectral periods that define the standard curve. The factors used in AASHTO are the original site factors developed in the early 1990s as a result of efforts at NEHRP agencies and are given in tables depending on the site class as well as the values of the spectral acceleration at 0.0, 0.2, and 1.0 seconds. These site factors were developed for soils in the western United States, but they are used nationwide. Furthermore, with recent work of the seismic community, it was found that both the 1997 and 2015 NEHRP site factors can be inaccurate relative to direct consideration of soil effects through the ground motion models. In contrast, the 2018 hazard model includes the effects of soil directly via the ground motion models for all 22 periods of the design spectrum, and this allows the site effects to be considered separately for each seismic source and its distance. Furthermore, the soil effects are now based on the specific regions of the country, and the differences between the western versus central and eastern United States are significant. Additionally, U.S. Geological Survey has included inter- mediate shear wave velocities, Vs30s, for use in developing a site-adjusted design spectrum, and this provides improved choices for the site input data, almost doubling the number of site classifications that can be used. Sedimentary basin effects, which are a deeper form of soil effects, have been included in the 2018 hazard model for four key regions: Seattle, San Francisco, Los Angeles, and Salt Lake City. Inclusion of basin effects elsewhere is not yet complete and will continue into future U.S. Geological Survey hazard update cycles. As a result, there are marked changes from the current AASHTO provisions, depending on which part of the country is considered. For the western locations, there is in general an increase in the design spectra for significant portions of the period range for Site Classes D and E. The shapes of the 2018 hazard model spectra are more consistent with what has been seen when performing site-specific analyses, in that short-period motions tend to be damped by the non- linear behavior of the soil, while motions at longer periods show amplification. This has not always been the case with the current methods in AASHTO. In the eastern portion of the country, there are dramatic reductions in design spectra for the softer soil sites, especially in the longer period (T > 3 s) displacement spectra. Generally speaking, this is due to the use of relationships for soil effects that are more representative of soils found in the eastern portion of the country, rather than relationships originally developed for the western United States. Comparing the PGA charts for the various locations, a similar trend is seen with lower Site Class B PGA values in the West for the 2018 hazard model but larger Site Classes D and E PGA values. For sites in the East, the PGA values for Site Classes D and E are often less than the current AASHTO values. In summary, locations in the western United States would see reductions (but not everywhere) in the hazard for firm soil and rock sites but increases in the hazard on softer soil sites. For the eastern United States, the hazard for firm soil and rock sites would be similar to what is currently specified but for softer soil sites, especially Site Class E, there could be dramatic reductions in the design spectrum. Summary of Comparisons and Recommendations The 2018 U.S. National Seismic Hazard Model represents the latest advancements in hazard modeling. It includes the latest Next Generation Attenuation Relationships for both the western and eastern United States, the effects of site soil on the hazard (which were previously

64 Proposed AASHTO Guidelines for Performance-Based Seismic Bridge Design approximated by applying 1997 NEHRP site coefficients), and basin effects for several high- population areas. The soil effects included account for the differences in soil behavior across the nation. Additionally, a 22-period construction is used to define a unique spectral shape for each site. In short, the 2018 hazard model is the best current science on defining the hazard posed by earthquake ground motions at the national level. A comparison with the current AASHTO-prescribed hazard, as defined by the mapping and design spectrum construction methods contained in the latest versions of both the AASHTO LRFD Bridge Design Specifications as well as the Guide Specifications for LRFD Seismic Bridge Design, indicates several generalizations can be made. First, the base hazard level, as defined by the design spectrum for Site Class B, is relatively similar for sites in the western United States. Second, the soil effects in the western United States tend to increase the hazard significantly over current AASHTO hazard, especially for the softer soil classifications of D and E. Third, some sites in the western United States, such as Seattle, experience a significant increase in hazard at longer spectral periods, due to the inclusion of basin amplifications. Finally, and most trouble- some, the hazard in the central and especially eastern United States is dramatically lower with the 2018 hazard model for softer soil sites. The current level of awareness related to seismic design in the central and eastern United States and the level of design that has resulted have taken many years of effort to develop. There is a concern that this would be negatively impacted by the adoption of a hazard model with substantially lower base seismic hazards for this area of the country. Merely adopting the 2018 hazard model, without other potential updates, would result in a substantial decrease in the ability of bridges designed using this model to resist seismic motions. A number of methods can be considered to attempt to alleviate this apparent problem. The adoption of a 1000 year return period for bridges, as opposed to the 2500 year return period for buildings, may be partially responsible for the situation. A short study of the 2018 hazard model results for a 2500 year return period did indicate an increase in the hazard for the central and eastern United States to a more reasonable level, but also resulted in excessive ground motion levels in the western United States. A simple increase in the return period does not appear to be a solution to the current issue. The building industry has transitioned from a pure hazard evaluation to a risk-targeted design methodology, wherein the probability of collapse when a structure is subjected to a given level of ground motion is included in the definition of the hazard. The differences between the hazard in the central and eastern United States versus the western United States were the main drivers for the adoption of this methodology. This is exactly what is seen in the comparison of the 2018 hazard model previously presented. It may be that with the exploration of the updated hazard model for AASHTO, the same conditions that prompted this action in the building community are now present for the bridge community. Consideration of a risk-targeted approach may be necessary in order to bring AASHTO in line with the latest developments in seismic design. The current AASHTO definition of seismic hazard is based on the uniform hazard approach, in which the mapped motions have an equal probability of exceedance at any location in the country. This is a rational, easily understandable approach and has served the bridge community well over the years it has been utilized. However, there are some shortcomings to this approach. The uniformity of exceedance probability is only assured at the specific return period selected. In reality, the variation in seismic hazard with return period differs substantially across the nation, with the intensity of motion at western sites rising quickly with return period then leveling out, while in the central and eastern United States, the intensity remains low until longer return periods, where it grows more rapidly. Since the choice of return period is arbitrary,

AASHTO Seismic Map Update 65 the differences in design motion across the country based on this choice, and arising from the differences in hazard with return period, also become somewhat arbitrary. A risk-targeted approach takes into account the full hazard curve and is based on the idea of an equal prob- ability of collapse or some other damage state, rather than simply the equal probability of a certain motion occurring during a specified time period (e.g., 75 years). Based on the information developed and compared as part of this project, it appears the best course of action is for AASHTO to consider a risk-targeted approach and develop the approach for bridges to a consistent level with the uniform hazard approach, make comparisons similar to what was done in this project to evaluate the effects on various locations across the country, and then decide which approach is best suited to the bridge community.