Jökull - 01.12.1994, Side 24

Fig. 3 Locations of points of
the Hengill GPS network.
Point 7134 (Núpafjall) was
occupied each day. Lines in-
dicate faults of the SISZ
(from Jóhannesson, 1989).
3. mynd. Mœlipunktar fyrir
gervitunglastaðsetningar á
Hengilssvœðinu. Mælt var
daglega í punkti 7134. Línur
sýna misgengissprungur á
Suðurlands- skjálftasvœðinu.
Fig. 4 Map showing the lct
horizontal error (scaled form-
al error) ellipsoids for points
relative to the point 7134, for
the best final solution for the
entire network i.e., the ambi-
guity-fixed solution.
4. mynd. Kort af óvissu í lá-
réttri staðsetningu mœli-
punkta, miðað við að stað-
setning Núpafjalls (punkts
7124) sé föst. Sporbaugar
sýna eitt staðalfrávik óviss-
unnar í bestu heildarlausn
útreikninga fyrir allt svæðið.
point determinations were scaled by the reduced chi-
squared value determined by the network adjustment
(the normalised RMS). This is the RMS of the re-
peatability: formal error ratios for the individual
points, and expresses the amount by which the formal
error underestimates the measurement repeatability.
In the absence of a more accurate, independent check
on the coordinates of our points from another geodet-
ic technique, these scaled formal errors are the best
estimates of the true accuracy of our results available.
The effects of ocean loading are expected to be
appreciable only in the case of the long lines connect-
ing point Núpafjall to the regional points. For these
lines variations in the vertical component only are ex-
pected to be significant. Therefore, we studied varia-
tions in this component for those lines. Ambiguity-
free, ionosphere-free solutions were used because of
the difficulty of fixing the L5 ambiguities in the long
lines. This was a consequence of the long line lengths
which suffer from errors in the modelled orbit of the
satellites, and high ionospheric noise during sunset
and sunrise. A-priori point coordinate positions were
obtained by processing two 6 hour windows from the
full 18 hour sessions that were of high enough quality
to fix the L5 ambiguities. Two-hour discrete intervals
were then processed separately to calculate the varia-
tion in the vertical components.
22
JÖKULL, No. 44