Estimating the yields of the Amchitka tests by waveform intercorrelation

Summary. A new method of estimating the yields of events at a nuclear test site is introduced and tested against the Amchitka experience. The intercorrelation method is a relative waveform comparison technique, which involves convolution of the first few seconds of each short‐period P‐wave recorded at a given station for two events with estimates of the effective source function (including pP ) for the other event. This procedure accounts for common path and receiver effects as well as differences in time functions and near‐source surface interactions. If the source parameters for a master event are determined independently, the assumed source parameters for a second event can be adjusted to minimize the residual differences in the inter‐correlated signal pairs for a large number of stations simultaneously. Yield estimates can then be made using empirical relations between the source parameters and yield. The source time function representation used is ψ( t ) =ψ{1 − exp(− Kt )[1 + Kt + ( Kt ) 2 /2 − B ( Kt ) 3 ]}. Due to the narrowband character of short‐period P‐waves and the possibility of yield dependence of overshoot of the source spectrum (controlled by B) , it is generally not possible to resolve uniquely the static source strength, ψ ∞ . However; for fixed values of B , an alternate parameter, ψ′ ∞ , can be determined, which gives the correct source strength in the period range of the signals. For the Amchitka events ψ′ ∞ is found to have a simple power law relation with yield. With B = 1, intercorrelation of MILROW with CANNIKIN gives a ψ′ ∞ of 4.0 ± 0.6 × 10 11 cm 3 for CANNIKIN, compared with 4.5±0.5 × 10 11 cm 3 determined from near‐field data. Intercorrelations of MILROW and CANNIKIN with LONGSHOT give consistent ψ′ ∞ values of 2.0 ± 0.3 × 10 10 cm 3 for LONGSHOT. In all cases the pP lag times and amplitudes obtained in the procedure are consistent with previous determinations. The effects of underparameterization of the effective source functions are investigated and found not to bias the ψ′ ∞ determinations strongly. The use of complete waveform information is shown to have several advantages over standard amplitude comparisons for estimating yield.