Surface-wave amplitudes from explosion sources show less variation for a gi
ven event than body wave amplitudes, so it is natural to expect that yield
estimates derived from surface waves will be more accurate than yield estim
ates derived from body waves. However, yield estimation from surface waves
is complicated by the presence of tectonic strain release, which acts like
one or more earthquake sources superimposed on top of the explosion, Moment
-tensor inversion can be used to remove the tectonic component of the surfa
ce waves, however moment-tensor inversion for shallow sources is inherently
non-unique so the explosion isotropic moment cannot be determined with the
necessary accuracy by this means. Explosions on an island or near a mounta
in slope can exhibit anomalous surface waves similar to those caused by tec
tonic strain release. These complications cause yield estimates derived fro
m surface waves to be less accurate than yield estimates from body waves re
corded on a well-calibrated network with good coverage. Surface-wave amplit
udes can be expressed as a surface-wave magnitude M-s, which is defined as
the logarithm of the amplitude plus a distance correction, or as a path cor
rected spectral magnitude, log M-0', which is derived from the surface-wave
spectrum. We derive relations for M-s vs. yield and log M-0' vs, yield for
a large data set and estimate the accuracy of these estimates.