Assume that the mass density is smooth in some sense on the scale of the dominant pressure wavelengths of concern. This leads us to examine the relative magnitude of the two terms on the left hand side of (13):

(18) |

It is convenient to normalize the two terms by and scale
by the Fresnel length *L*_{f}, since spatial changes in properties are
only meaningful with respect to the spatial wavelengths and frequency
components in the propagating wavefield:

(19) |

Examining these two terms, it is apparent that (1) is valid when

(20) |

and

(21) |

As a matter of practical interest, consider the following example.
Assume a propagating wavefield in a region with a phase velocity *v _{2}* of
3 km/s, an average density of 2.0 g/cc,
and a dominant frequency content of 30 Hz. The dominant
spatial wavelength will be on the order of 100 m, and we will assume that
the wave energy will be coherent over a quarter-wavelength Fresnel zone
of

(22) |

where
is an average over the Fresnel zone and is the total
density contrast over the Fresnel zone, I evaluate the density
gradient term *R _{2}* to be %.
Similarly, for the displacement gradient term,

(23) |

or, %. For this realistic exploration example,
by
about three orders of magnitude! In fact, scaling this result implies that
a density gradient of up to 16 g/cc/km over a 25 m Fresnel zone,
or a total density contrast of 0.4 g/cc/*L*_{f}, will result
in an error of less than 10% if the variable velocity acoustic equation
(1) is used:

(24) |

On the other hand, this result also implies that very large contrasts in density and very high frequency data may be sufficient to invalidate the use of (1). The validity condition (20) is likely to be violated at strong geologic discontinuities such as shale/carbonate boundaries or at basement volcanic unconformities given sufficiently high frequency data.

11/18/1997