UC Berkeley

This study has two main thrusts. The first part of the study addresses static seepage and stability of California levees as related to the presence of woody vegetation. The second part of this study addresses seismic deformations related to California levees through calibration, validation, and sensitivity analysis of a constitutive model implemented to capture seismic embankment deformations.

Two field tests were conducted to investigate the effects of seepage in the vicinity of live and decaying tree root systems to examine the effects of live and decaying root systems on levee seepage and slope stability. The first field test involved the construction of parallel trenches in the vicinity of a eucalyptus stump located along the landside of the northern levee bordering the American River adjacent to the California Exposition and State Fair. A live hackberry tree with healthy roots was present at the toe of the levee. A control set of parallel trenches was constructed away from the eucalyptus stump. During the test, the upslope trench was flooded and maintained at constant head to induce slope-parallel seepage and the downslope trench was used to make observations and collect any intercepted seepage. Piezometers and tensiometers were installed to measure positive and negative pore water pressures within the zone of flow to describe the wetting and flow patterns as they evolved within the levee. Instrumentation was installed to assess the influence of the stump and its decomposing root system. Live roots, mammal burrows, and other features added complexity to the system. In addition to instrumentation data, visual observations were recorded during the 6 day flow test. During the flow test, wetting front and water flow patterns appeared to be dominated by flow through a network of shallow mammal burrows. Physical observation of the saturation front, as seen from the lower wall, confirmed that the area below the stump was the last location to saturate during the wetting test. Ground-based tripod light detection and ranging (T-LiDAR) was used to complement traditional logging and for constructing a 3D model of the root system, burrows and stratigraphy. Preliminary computer simulations of the advance of the wetting front support the basic patterns observed in the field test.

The second field experiment was conducted along the crown of a bypassed levee along an oxbow segment of the seven mile slough on Twitchell Island in Rio Vista, California. An 8-foot deep crown trench was excavated to extend through the root system of a land side live oak tree, a water side valley oak tree, and a control section. The test was designed to evaluate the effects of seepage in the vicinity of live tree root systems. During the test, the crown trench was flooded and maintained at constant head to simulate a flood condition with water delivered from the center of the levee. Piezometers and tensiometers were installed to measure positive and negative pore water pressures, respectively, within the zone of flow to describe the wetting and flow patterns as they evolved within the levee. Burrow networks, fracturing, and gapping within levee soils, as well as variability of stratigraphic conditions across the site added complexity. Visual observations were made during the flow test to view surficial seepage and flow patterns from the surface in addition to continuous monitoring of subsurface instruments. The site contained an extensive network of muskrat burrows in addition to burrows by other species and the initial advance of the wetting front appeared to be related to burrowing activity. With increased time and saturation of levee soils, flow through macropores appeared to diminish. The levee appears to have been founded on overbank deposits comprised of lower permeability soils than the overlying levee fill. Water appeared to accumulate on this stratigraphic layer, driving seepage patterns on the landside of the levee. A break in this low permeability layer affected flow patterns while the slope of this layer toward this break appears to have added a three dimensional effect to flow patterns. Cracking was observed in the crown road along the levee crest within the first 24 hours of the flow test. After approximately 40 hours of flow, the waterside oak tree, which was initially leaning at an angle of approximately 43 degrees from horizontal, rotated an additional approximately 20 degrees into the slough, creating cracks and deformation along the waterside slope. A dye test was performed as a part of the experiment to better understand the extent of these burrows, their effect on flow patterns, and to better evaluate the role of these burrow networks in the deformations observed on the waterside slope during the flow test. Ground-based tripod light detection and ranging (T-LiDAR) was used to complement our efforts related to tracking deformations during the test.

Based on calibrated numerical simulations, trees were found not to play a significant role in seepage-induced rotational or block failure of the levee slopes. However, where trees exhibit significant lean (center of mass extends beyond the fulcrum of the root plate), horizontal roots extending into the levee may place additional loads on the levee embankment. Loading of this type can be incorporated into two dimensional slope stability analyses, using mass-averaging to capture the three-dimensional impact of the tree. Tree overturning was evaluated at the waterside oak tree. Root forces were represented as a single horizontal force and a single vertical force. Horizontally oriented tree root loading increased faster than vertical loading in response to increasing slope angle, while the reverse was true for tree lean where vertical root forces increased more rapidly with increasing tree lean. The method was implemented and successfully captured the observed landside and waterside tree responses during the Crown Trench Seepage Test.

The second part of this research focused on seismically induced permanent displacement of earth levees, embankments, and earth-fill slopes resulting from liquefaction below these earth structures. Deformations of this nature are not well captured in current seismic design practice. Ground remediation can be employed to reduce the hazards resulting from soil liquefaction for cases where the analytical tools predict poor seismic performance. There are not sufficient funds to repair all vulnerable levees in the system. Thus, robust analytical procedures are required to evaluate sections of levees where liquefiable foundation materials may lead to significant damage.

Inertially driven ground movements of intermediate levels are the primary focus of this study. In these cases, the post-liquefaction static stability of the earth slope is greater than one, and seismically induced permanent displacements result primarily from earthquake shaking after liquefaction is triggered. Limited lateral spreads involving liquefaction of medium dense sand can produce seismic displacements on the order of several centimeters to a meter or more. These levels of seismic displacements are sufficient to damage severely levees. The most commonly employed simplified method for evaluation of seismic deformation at these intermediate levels relies on the concept post-liquefaction residual shear strength. For many practical cases, residual shear strength is ill-defined due to the ever changing resistance provided by soils that undergo repetitive dilative responses during cyclic loading. Where liquefied soils are sufficiently strong to resist flow failures, engineers lack satisfactory tools to evaluate the seismic performance of earth structures that overlie liquefiable soils.

A nonlinear soil constitutive model (UBCSAND), which was developed by Professor Byrne and implemented in the finite difference program FLAC, is employed to evaluate seismic deformations of earth structures resulting from liquefaction-induced lateral movements. Analyses of one-element laboratory tests are performed first to develop trends within the UBCSAND soil model calibration parameters. The basic model parameters are correlated to (N1)60 values. The UBCSAND model also has four "fitting" parameters. Two of the four model parameters are varied in this study to evaluate the sensitivity of the results to these variations. Triggering is captured with values of the m_hfac1 parameter with a typical range of 0.5 to 2.0 depending on relative density, CSR, and initial static shear stress.

The trends identified are implemented in the back-analysis of several case histories, and the ability of the UBCSAND model within the program FLAC to capture observed deformations is evaluated. The numerical simulations of seismic performance at Moss Landing Marine Laboratory and Monterey Bay Aquarium Research Institute during the 1989 Loma Prieta Earthquake and at the Juvenile Hall Facility during the 1971 San Fernando Earthquake are shown to capture well the key features of these case histories.

The study was generalized through a broader sensitivity study to investigate the seismic performance of earthen embankments built atop potentially liquefiable soils. Several representative levee sections on differing foundations are analyzed, wherein key characteristics, such as the thickness of the liquefiable layer and its relative density, are systematically varied to develop useful insights. The thickness of the liquefiable foundation layer impacted displacements in a non-linear pattern where displacement increased more rapidly as the liquefiable material layer thickness increases. As would be expected, combinations of thicker deposits of liquefiable foundation soils combined with higher embankments yielded the maximum displacement of the cases analyzed in this study.

A suite of nine near-fault, forward-directivity, fault-normal soil earthquake ground motions and seven intermediate-field soil earthquake ground motions are used to reflect the seismic hazards most likely to control the design of levees within the San Joaquin-Sacramento delta region of California. The intense forward-directive 1994 Northridge Sylmar Converter Station motion yielded approximately twice as much displacement as the lower intensity backward-directive 1992 Landers Joshua Tree motion. The rate of increase of the calculated liquefaction-induced displacement with increasing Arias intensity was roughly linear for these embankment configurations for the entire suite of earthquake ground motions.