Load-induced Earthquakes at Lake Kariba?II

Abstract

Summary The filling of the world's largest artificial lake at Kariba has been accompanied and followed by considerable earthquake activity. A local network of three seismographs allowed location of epicentres for 159 of 2000 tremors recorded during three years while the lake filled. The epicentres lie in the down-faulted rift valley of the middle Zambezi and it is clear that the lake has re-activated some of the existing faults, formed in a complex tectonic history since late Precambrian times. Records from permanent observatories have been used in estimates of tremor frequency and rate of energy release through 9.4 years. Calculations of stress distributions at nine lake levels have been used in estimates of the volume V, within which the maximum shear stress exceeds 1 bar, and V, as function of time is compared with the seismic activity. The latter was highly correlated with rise of V, as the lake filled and for three years after. Of total energy 2.32 x lOZ0erg some 88.4 per cent was radiated from seven events of 5.0 < m < 5-8. The main activity built up from June to September 1963, culminating in the last week of September, five weeks after the lake reached its maximum level; during that week five of the seven large shocks and many hundreds of aftershocks occurred. Activity has continued to the present, correlated with water level and more strikingly with V, until mid-1966 but not thereafter, and showing a gradual though far from continuous decline. Magnitude-frequency dependence suggests that all the tremors have been ' aftershocks ' in the sense that they have occurred in prefractured rock of low strength. Energy transformations are considered making use of calculated elastic depressions. It is shown that a vast amount of energy (5.5 x loz4 erg) is stored as elastic strain energy in the lithosphere as a result of release of gravitational energy during depression of the rock in a limited volume, and that the radiated seismic energy can be provided at plausibly low efficiency. Three modes of failure are considered to have been present. In one brief burst of about 200 tremors near the west end of the lake it is believed that the lake triggered failure of a pre-stressed, previously active fault through increment to the stress, to fluid pressure or to both. In another burst of about 50 tremors 15 to 65 km downstream from the dam only triggering of pre-stressed faults by incremental stress is believed probable. In the main activity near and under the deep Sanyati Basin, which included all the large shocks, it is argued, mainly from the correlation in time between activity and growth of V,, that the shear stress of 1 or 2 bars due to the load has triggered failure of faults initially stressed at several tens of bars. Increase of fluid pressure appears to have been