A major transition in the style of deformation of Kilauea volcano occurred during the period January 1982 through November 1983. Prior to 1983, the deformation was dominated by magma intrusions, rapid extension of the summit, and seaward movement of the south flank. The pattern of activity changed to magma eruption and slowed deformation of the volcanic edifice in January 1983. An earthquake swarm in the southwest rift zone in June 1982 and an intrusion of a large dike in the east rift zone in January 1983 prepared the volcano for the eruptive sequences that continue today. We estimate the sources of surface displacements associated with these events using a least squares nonlinear inversion technique and planar dislocations. The models are constrained with measurements of line length changes and of surface height changes from tide gauges, water wells, and spirit-level tilt data. Deformation associated with the 1982 earthquake swarm is best modeled as the emplacement of a steeply dipping dike in the upper and middle southwest rift zone extending from 2--3 km to 10--11 km depth with a width of ~1--2 m. Tilt measurements gathered at the time of the earthquake swarm record a small deflation of Kilauea's summit; however, trilateration data indicate the summit was rapidly extending in 1982. The extension and our models are consistent with seaward slip of the south flank on a low-angle basal thrust fault.

Dislocation models of the January 1983 dike intrusion yield a 14- to 15-km-long dike along the zone of observed ground rupture. The dike extends from just below the surface to the region of the most intense rift zone seismicity at 4--5 km depth. A strong nonlinear relationship among height, dip, and width results in a suite of models that explain the data equally well. Tests of parameter sensitivity indicate that the dike width and height are the least constrained parameters; the best fit average width ranges from 1.7 to 2.8 m. Our models also include a source of contraction at ~5 km depth that produces 40 cm of ground subsidence near Makaopuhi, the point of initiation of the intrusion. This location is consistent with geochemical and seismic evidence that indicate magma is stored in the rift zone. Residual errors for all models indicate that substantial line-length changes are unexplained. Seaward motions of the south flank during 1982 and 1983 are not well modeled by simple dikes and summit deflations; however, attempts to constrain low-angle faults in the south flank were only partly successful due to the limited extent of the geodetic networks.