Effects of the earthquake of March 27, 1964, on the Alaska Railroad

Abstract

In the 1964 Alaska earthquake, the federally owned Alaska Railroad sustained damage of more than $35 million: 54 percent of the cost for port facilities; 25 percent, roadbed and track; 9 percent, buildings and utilities; 7 percent, bridges and culverts; and 5 percent, landslide removal. Principal causes of damage were: (1) landslides, landslide-generated waves, and seismic sea waves that destroyed costly port facilities built on deltas; (2) regional tectonic subsidence that necessitated raising and armoring 22 miles of roadbed made susceptible to marine erosion; and (3), of greatest importance in terms of potential damage in seismically active areas, a general loss of strength experienced by wet waterlaid unconsolidated granular sediments (silt to coarse gravel) that allowed embankments to settle and enabled sediments to undergo fiowlike displacement toward topographic depressions, even in fiat-lying areas. The term “landspreading” is proposed for the lateral displacement and distension of mobilized sediments; landspreading appears to have resulted largely from liquefaction. Because mobilization is time dependent and its effects cumulative, the long duration of strong ground motion (timed as 3 to 4 minutes) along the southern 150 miles of the rail line made landspreading an important cause of damage. Sediments moved toward natural and manmade topographic depressions (stream valleys, gullies, drainage ditches, borrow pits, and lakes). Stream widths decreased, often about 20 inches but at some places by as much as 6.5 feet, and sediments moved upward beneath stream channels. Landspreading toward streams and even small drainage ditches crushed concrete and metal culverts. Bridge superstructures were compressed and failed by lateral buckling, or more commonly were driven into, through, or over bulkheads. Piles and piers were torn free of superstructures by moving sediments, crowded toward stream channels, and lifted in the center. The lifted piles arched the superstructures. Vertical pile displacement was independent of the depth of the pile penetration in the sediment and thus was due to vertical movement of the sediments, rather than to differential compaction. The fact that bridge piles were carried laterally without notable tilting suggests that mobilization exceeded pile depths, which averaged about 20 feet. Field observations, largely duplicated by vibrated sandbox models of stream channels, suggest that movement was distributed throughout the sediments, rather than restricted to finite failure surfaces. Landspreading generated stress that produced cracks in the ground surface adjacent to depressions. The distribution of this stress controlled the crack patterns: tension cracks parallel to straight or concave streambanks, shear cracks intersecting at 45° to 70° on convex banks where there was some component of radial spreading, and orthogonal cracks on the insides of tight meander bends or islands where spreading was omnidirectional. Ground cracks of these kinds commonly extended 500 feet, and occasionally about 1,000 feet, back from streams, which indicates that landspreading occurred over large areas. In areas of landspreading, highway and railroad embankments, pavements, and rails were pulled apart endways and were displaced laterally if they lay at an angle to the direction of sediment displacement. Sediment movement commonly skewed bridges that crossed streams obliquely. The maximum horizontal skew was 10 feet. Embankment settlement, nearly universal in areas of landspreading, also occurred in areas where there was no evidence for widespread loss of strength in the unconsolidated sediments. In the latter areas embankments themselves clearly caused the loss of bearing strength in the underlying sediment. In both areas, settlement was accompanied by the formation of ground cracks approximately parallel to the embankment in the adjacent sediments. Sediment-laden ground water was discharged from the cracks, and extreme local settlements (as much as 6 ft) were associated with large discharges. Landspreading was accompanied by transient horizontal displacement of the ground that pounded bridge ends with slight or considerable force. The deck of a 105-foot bridge was repeatedly arched up off its piles by transient compression. Bridges may also have developed high horizontal accelerations. One bridge deck, driven through its bulkhead, appears to have had an acceleration of at least 1.1 to 1.7 g; however, most evidence for high accelerations is ambiguous. Limited standard penetration data show that landspreading damage was not restricted to soft sediments. Some bridges were severely damaged by displacement of piles driven in sediments classified as compact and dense. Total thickness of unconsolidated sediments strongly controlled the degree of damage. In areas underlain by wet water-laid sediments the degree of damage to uniformly designed and built wooden railroad bridges shows a closer correlation with total sediment thickness at the bridge site than with the grain size of the material in which the piles were driven. Local geology and physiography largely controlled the kind, distribution, and severity of damage to the railroad. This relationship is so clear that maps of surficial geology and physiography of damaged areas of the rail belt show that only a few geologic-physiographic units serve to identify these areas: 1. Bedrock and glacial till on bedrock. No foundation displacements, but ground vibration increased toward the area of maximum strain-energy release. 2. Glacial outwash terraces. Landspreading and damage ranged from none where the water table was low and the terrace undissected to severe where the water table was near the surface and the terrace dissected by streams. 3. Inactive flood plains. Landspreading, ground cracking, flooding by ejected ground water, and damage were generally slight but increased to severe toward lower, wetter active flood plains or river channels. 4. Active flood plains. Landspreading, ground cracking, and flooding were nearly universal and were greater than on adjacent inactive flood plains. 5. Fan deltas. Radial downhill spreading and ground cracking were considerable near the lower edges of the fan deltas and were accompanied by ground-water discharge. Landslides were common from edges of deltas. Damage, landspreading, ground crack-ing, vibration, and flooding by ground water generally increased with (1) increasing thickness of unconsolidated sediments, (2) decreasing depth to the water table, (3) proximity to topographic depressions, and (4) proximity to the area of maximum strain-energy release.