Most paleoseismic studies are at low to moderate latitudes. Here we present results from a high-latitude (61°30′ N) trenching study of the Castle Mountain fault in south-central Alaska. This fault is the only one known in the greater Anchorage, Alaska, area with historical seismicity and a Holocene fault scarp. It strikes east-northeast and cuts glacial and postglacial sediments in an area of boreal spruce-birch forest, shrub tundra, and sphagnum bog. The fault has a prominent vegetation lineament on the upthrown, north side of the fault. Nine trenches were logged across the fault in glacial and postglacial deposits, seven along the main trace, and two along a splay. In addition to thrust and strike-slip faulting, important controls on observed relationships in the trenches are the season in which faulting occurred, the physical properties of the sediments, liquefaction, a shallow water table, soil-forming processes, the strength of the modern root mat, and freeze-thaw processes. Some of these processes and physical properties are unique to northern-latitude areas and result in seismic disturbance effects not observed at lower latitudes.

The two trenches across the Castle Mountain fault splay exposed a thrust fault and few liquefaction features. Radiocarbon ages of soil organic matter and charcoal within and overlying the fault indicate movement on the fault at ca. 2735 cal. (calendar) yr B.P. and no subsequent movement. In the remaining seven trenches, surface faulting was accompanied by extensive liquefaction and a zone of disruption 3 m or more wide. The presence of numerous liquefaction features at depths of <0.5–1.0 m indicates faulting when the ground was not frozen—i.e., from about April to October. Sandy-matrix till, sand, silt, gravel, and pebbly peat were injected up to the base of the modern soil, but did not penetrate the interlocking spruce-birch root mat. The strength of the root mat prohibited development of a nonvegetated scarp face and colluvial wedge. In only one trench did we observe a discrete fault plane with measurable offset. It lay beneath a 2-m-thick carapace of liquefied sand and silt and displayed a total of 0.9–1.85 m of thrust motion since deposition of the oldest deposits in the trenches at ca. 13,500 yr B.P. We found liquefaction ejecta on paleosols at only one other trench, where there were bluejoint (Calamagrostis canadensis) tussocks that lacked an extensive root mat. From crosscutting relationships, we interpret three paleoliquefaction events on the main trace of the Castle Mountain fault: 2145–1870, 1375–1070, and 730–610 cal. yr B.P. These four earthquakes on the Castle Mountain fault in the past ∼2700 yr indicate an average recurrence interval of ∼700 yr. As it has been 600–700 yr since the last significant earthquake, a significant (magnitude 6–7) earthquake in the near future may be likely. Paleoseismic data indicate that the timing and recurrence interval of megathrust earthquakes is similar to the timing and recurrence interval of Castle Mountain fault earthquakes, suggesting a possible link between faulting on the megathrust and on “crustal” structures.