Stokes's kernel used for the evaluation of a gravimetric geoid is a functio
n of the spherical distance between the point of interest and the dummy poi
nt in the integration. Its values thus are obtained from the positions of p
airs of points on the geoid. For the integration over the near integration
zone (near to the point of interest,) it is advantageous to pre-form an arr
ay of kernel values where each entry corresponds to the appropriate locatio
ns of the two points, or equivalently to the latitude and the longitude-dif
ference between the point of interest and a dummy point. Thus, for points o
f interest on the same latitude, the array of the Stokes kernel values rema
ins the same and may only be evaluated once. Also, only one half of the arr
ay need be evaluated interest.
Numerical tests show that computation speed improves significantly after th
is algorithm is implemented. For an area of 5 by 10 arc-degrees with the gr
id of 5 by 5 arc-minutes, the computation time reduces from half an hour to
about 1 minute. To compute the geoid for the whole of Canada (20 by 60 arc
-degrees, with the grid of 5 by 5 arc-minutes), it takes only about 17 minu
tes on a 400MHz PC computer.
Compared with the Fast Fourier Transform algorithm, this algorithm is easie
r to implement including the far zone contribution evaluation that can be d
one precisely, using the (global) spectral description of the gravity field
.