Track 173 
Track 172 
Track 171 
Track 170 
Track 169 
Track 168 
Track 167 
Track 166 
Figure 2. Grey scale time sequence plots of anomaly heights deduced from Geosat data. 
Heights are coded as image brightnesses for a range of +_ 30cm about a mean 
value. Each rectangle contains the height anomaly data for 48 consecutive 
satellite cycles along one track from latitude 60°N to 0°N. Rectangles correspond 
to tracks for n = 173 (top, whose centre section lies over land) to 166 (bottom). 
Figure 2a (left) shows the unedited data. Figure 2b (right) shows the data after 
editing and smoothing. 
Copies of tapes of the Geosat global data set were received 
from NOAA in the USA, via MEDS in Ottawa, typically 6 
months after data collection. One 6250 bpi tape covers two 
¡ 17 day cycles. Corrections listed on the tapes for variations 
in geoid height, solid earth and ocean tides, water vapour 
and atmospheric pressure (propagation and inverse 
barometer corrections), ionosphere electron content and 
satellite attitude, were applied to the altimetry data. 
The major error remaining is due to inaccurate knowledge 
of the satellite’s absolute vertical position. This can amount 
to several meters, but is slowly varying over a spatial scale of 
several thousand kilometers, since most of the error occurs 
at a spatial frequency of one cycle per orbit. The error for 
each orbit track was compensated in the present study by 
subtracting the best fit second order polynomial function of 
I along-track distance from the observed surface heights. The 
correction was computed separately for the data in each 
satellite pass, using only data over deep water for which 
coverage extended over at least 80% of the latitude range 
0° to 60°N. 
Inaccuracy of the geoid heights listed on the Geosat data 
tapes is significant at the space scales (< 500km) being 
studied here and prevents measurement of absolute ocean 
surface heights. We therefore compute height anomalies as 
the difference between the corrected heights, and the two 
year mean heights, along each track. A constant height 
anomaly in a fixed position will therefore be removed from 
the data when the mean observed height is subtracted. This 
will not affect the detection of moving anomalies, but should 
be born in mind when interpreting the data presented 
below. 
Figure 2 shows the height anomalies computed in this way 
from Geosat data for 48 17 day cycles from November 1986 
to February 1989. The height anomalies along each track 
are plotted along horizontal lines, coded as grey scale 
variations from black (about 30cm low) to white (about 
30cm high), relative to a mean level shown as a neutral grey. 
Each rectangle consists of the 48 lines of observations, 
repeated at 17 day intervals, for one of eight passes whose 
tracks are plotted in Figure 1. The 48 lines are plotted one 
beneath the other and aligned vertically by latitude from 
60°N at the left-hand end down to the equator at the right. 
98 
The eight rectangles correspond to the eight right-most 
passes of those plotted in Figure 1. Gaps occur in the top 
two where these tracks cross the land-mass of North 
America. 
Missed passes (20 out of 384 shown here) and about 2% of 
the data which were manually deleted, have been linearly 
interpolated. The data were then smoothed by convolving 
the numbers plotted in Figure 2 with a 3 by 3 filter matrix. 
This smoothing filter is graded to 60% at the edges and 40% 
at the corners. Since it is applied to samples separated by 
20km along track, and by 17 day intervals in time, the 
effective smoothing (full width to half height of the filter) is 
over about 56 consecutive kilometers along track and about 
2.8 intervals or 48 days. 
The data in Figure 2 cover a time period of 2.3 years. Height 
variations with an annual period are visible on the left end 
of each track corresponding to positions close to the 
Alaska/Aleutian Island coast, and also at latitudes south of 
about 15°N. These larger scale phenomena have an 
amplitude of about 20cm. They may be artifacts of the 
corrections applied to the Geosat data. For the northern 
areas the tidal model is suspected. In the tropics the 
propagation delay due to the high water vapour content may 
not be correctly compensated. These features are not 
discussed further here. 
In the Gulf of Alaska, the data clearly show positive height 
anomalies with coherence scales of order 150km (along the 
tracks shown) in space, and 3 months (about 5 repeat cycles 
or 85 days) in time (Gower, 1989a). These small-scale 
anomalies are more intense closer to the Alaska/BC coast 
(n = 171-173) than in passes further offshore, and are also 
concentrated to the left (north-west) ends of the tracks, 
implying less eddy activity further south in the Gulf of 
Alaska. In the top two tracks the southern Gulf of Alaska is 
dominated by larger scale features with an annual cycle. 
Further south, between 40°N and 30°N the data again show 
increased eddy activity, and reduced activity between 30N 
and 15°N. South of 15^N the eddies are masked by the 
annual cycle.