In June 1995, a flight test was carried out over the Rocky Mountains t
o assess the accuracy of airborne gravity for geoid determination. The
gravity system consisted of a strapdown inertial navigation system (I
NS), two GPS receivers with zero baseline on the airplane and multiple
GPS master stations on the ground, and a data logging system. To the
best of our knowledge, this was the first time that a strapdown INS ha
s been used for airborne gravimetry. The test was designed to assess r
epeatability as well as accuracy of airborne gravimetry in a highly va
riable gravity field. An east-west profile of 250 km across the Rocky
Mountains was chosen and four flights over the same ground track were
made. The flying altitude was about 5.5 km, i.e., between 2.5 and 5.0
km above ground, and the average flying speed was about 430 km/h. This
corresponds to a spatial resolution (half wavelength of cutoff freque
ncy) of 5.0-7.0 km when using filter lengths between 90 and 120 s. Thi
s resolution is sufficient for geoid determination, but may not satisf
y other applications of airborne gravimetry. The evaluation of the int
ernal and external accuracy is based on repeated flights and compariso
n with upward continued ground gravity using a detailed terrain model.
Gravity results from repeated flight lines show that the standard dev
iation between flights is about 2 mGal for a single profile and a filt
er length of 120 s, and about 3 mGal for a filter length of 90 s. The
standard deviation of the difference between airborne gravity upward c
ontinued ground gravity is about 3 mGal for both filter lengths. A cri
tical discussion of these results and how they relate to the different
transfer functions applied, is given in the paper. Two different math
ematical approaches to airborne scalar gravimetry are applied and comp
ared, namely strapdown inertial scalar gravimetry (SISG) and rotation
invariant scalar gravimetry (RISG). Results show a significantly bette
r performance of the SISG approach for a strapdown INS of this accurac
y class. Because of major differences in the error model of the two ap
proaches, the RISG method can be used as an effective reliability chec
k of the SISG method. A spectral analysis of the residual errors of th
e flight profiles indicates that a relative geoid accuracy of 2-3 cm o
ver distances of 200 km (0.1 ppm) can be achieved by this method. Sinc
e these results present a first data analysis, it is expected that fur
ther improvements are possible as more refined modelling is applied.