Sources of seismic events in the cooling lava lake of Kilauea Iki, Hawaii

Seismic surveys conducted in recent years revealed a surprisingly high and sustained activity of local seismic events originating in the partially frozen lava lake of Kilauea Iki crater, Hawaii. About 8000 events per day were counted in 1976 at the center of the lake with a seismograph having a peak magnification of 280,000 at 60 Hz. The activity was found to be uniform over the whole area above the inferred magma lens and very weak in the periphery of the lake. The frequency-amplitude relation for these shocks obeys the Ishimoto-Iida or Gutenberg-Richter law very well, with a b value of 1.19 (±0.06). Locations of a few selected events indicate that they occur both above and below the layer of melt, although the seismic activity appears to be much higher in the upper crust. Whenever clear, the first motion is always outward from the source, suggesting that a crack opening under tensile stress owing to cooling is the responsible source mechanism. A simple model of a circular tensile crack nucleating at a point and growing at subsonic velocity can match the far-field P wave from these sources fairly well. Typical parameters for a large event inferred from the model are the following: radius, 2.7 m; maximum static tensile displacement between crack faces, 2.9μ; cavity volume, 4.4×10−5 m3; and a seismic moment tensor with diagonal elements only, having the values 3.8×1012, 4.5×1012, and 3.8×1012 dyn cm. The magnitude of the event is about −1, and its stress drop is of the order of 0.01 bar. A Q as low as 10 is required to satisfy the shape of the observed wave forms. Using the observed b value, the total cavity volume integrated over all cracks which is generated daily in the upper crust of Kilauea Iki is of the order of 1–20 m3, in rough agreement with the amount of volume contraction expected on the basis of our knowledge of the thermal history of the crust. As an alternate interpretation of the data, a model is proposed which views the seismic activity as reflecting the extension by up to several tens of centimeters of long existing cracks rather than the formation of new cracks. Discrimination between the two types of models does not appear to be feasible with the present available data.