Global Analysis of Sea Surface Height and Temperature

Abstract

Altimeter data (RA) of the ERS-1 and TOPEX/POSEIDON (T/P) missions and radiometer data (AVHRR) of the NOAA mission are analysed to study the evolution of sea surface height and sea surface temperature on a variety of time and space scales. For this purpose the data are translated into monthly grids with a 1° spatial resolution. The total time span is about 3.5 years (April '92 - August '95). Fourier, EOF, and time-longitude analyses show a perfect agreement between the T/P and ERS-1 results. A significant difference, however, is encountered when estimating the tilt over the time span at issue: 0.7 mm/yr (T/P) versus 3.7 mm/yr (ERS-1). It is shown that this is not due to a systematic difference in the altimeter bias but must be caused by local differences in the media corrections. The temperature grids show a temperature rise of 0.045 K/yr. The peak of the annual cycle in the sea level lags the peak in the sea surface temperature by about ½ month in the northern hemisphere and by about 1½ month in the southern hemisphere. Dominant patterns, among which a biannual cycle (ENSO), an annual cycle (mainly steric), and a 100-days cycle (eddy activity) are extracted by estimating EOFs, of which six are needed to recover 50% of the observed variability.

Keywords: altimeter, radiometer, sea surface height, sea surface temperature, ocean circulation, sea level change, ENSO, EOF.

INTRODUCTION

In this study we pose the question "is the ERS-1 altimeter doing as well as the T/P altimeter ?". We will analyse the ERS-1 altimeter data in a same fashion as we analysed T/P data in former studies. We will then compare the ERS-1 results with the T/P results and briefly summarise some of the agreements and the differences on the basis of frequency analysis and time-longitude sections. On the fly the ERS-1 results are compared with our results of a global analysis of sea surface temperature. These temperature results are based on NOAA AVHRR radiometer data, because at the time this investigation started we did not have sufficient temperature data from the ERS-1 Along-track Scanning Radiometer (ATSR).

Data and methods

The T/P data we selected also consist of 1-Hz measurements and are extracted from the MGDR CDROMs as supplied by AVISO/CNES [AVISO, 1994]. They span the time between 17 December '92 until 10 February '96. Again the data is screened and all the standard corrections are applied: reported range stability and oscillator clock drift, JGM-3 orbits, meteorological dry troposphere, wet troposphere from on board radiometer, ionosphere based on DORIS measurements (to have a consistent data set and to be able to merge TOPEX and POSEIDON data), solid earth and pole tides, Grenoble FES95.2.1 ocean tides and loading, the sea state bias according to the latest model by Gaspar, 100% inverse barometer, calibrated bias, and again the OSU MSS95 mean sea surface.

Monthly representations of the global dynamic topography (relative to the reference mean sea surface) are obtained by interpolating/filtering one month worth of sea height residuals corrected for residual radial orbit errors to 1° resolution grids. The radial orbit error has been reduced by crossover minimisation, accounting for a bias and a 1 cycle/rev signal. By separately processing monthly batches (either 1 cycle ERS-1 from its 35-day repeat period or 3 T/P 10-day cycles), the large-scale seasonal signal is preserved. During the 2nd ice phase, when ERS-1 was in a 3-day repeat orbit (December '93 until March '94) the ERS-1 data are merged with T/P data to increase the spatial sampling and coverage. In total 3 batches of combined data were processed, i.e., crossover minimised and gridded.

Finally, the AVHRR data from the NOAA satellites have been obtained through the Physical Oceanography Archive Center of the Jet Propulsion Laboratory (PO.DAAC). These are the AVHRR weekly global 18km gridded MCSST HDF grids. They contain fully interpolated data and are derived from the emitted and reflected radiation measured by the NOAA 7,9,11 satellites. For more information the reader is referred to the Rosenstiel School of Marine and Atmospheric Science (Florida). For the purpose of this study we combined the night and day data of 5 weeks and resampled the data on a 1° resolution grid, in order to have about the same sampling characteristics as our altimeter results. In addition, we only consider the temperature results over the same period as we consider ERS-1 data.

Results and discussion

A global mean sea level and sea surface temperature is computed for each of the monthly solutions by averaging the grid values weighted by the cosine of the latitude and imposing that a certain grid point is valid throughout the time series (to avoid erroneous seasonal variability due to ice evolution). The history of these global means is given in Figure 1. From top to bottom for ERS-1, T/P, and the AVHRR. A simultaneous estimate of the annual cycle and a bias and a tilt, reveals the average rates of change: 3.7 mm/yr, 0.7 mm/yr, and 0.045 K/yr, respectively. Despite the agreements in the annual cycle (actually, the annual cycle we observe is mainly due to the 100% inverse barometer correction), the tilts significantly differ between the T/P and ERS-1 time series.

Figure 1. History of global mean sea level and sea surface temperature. From top to bottom for ERS-1, T/P, and the AVHRR.

In Table 1 we summarised some statistics.

From this table it will be clear that the annual cycles of ERS-1 and T/P agree very well and that the peak of the annual cycle in the sea level on average lags the peak in the sea surface temperature by about ½ month in the northern hemisphere and by about 1½ month in the southern hemisphere, which shows that the response in the southern hemisphere due to annual heating is one month slower than in the northern hemisphere. We also notice that the difference in tilt between the two altimeter satellites is not systematic, and might be due to local phenomena. We will investigate this somewhat further.

From the sea height and sea surface temperature time series we simultaneously estimated a biannual, an annual, and a semiannual cycle, and a trend for each grid point separately. Figure 2 shows the global distribution of the solved tilts for ERS-1, T/P, and AVHRR, respectively. Also shown is the difference between ERS-1 and T/P. We clearly see local patterns giving rise to the overall difference in sea level change estimates. The band along the equator could be explained by differences in the radiometer correction, and the big spot in the Indian Ocean by differences in ionosphere correction. Remember that the correction for ERS-1 is based on a model, and that the correction for T/P comes from the DORIS data. This needs further attention but falls outside the scope of this paper. Comparing the T/P results with the AVHRR results we find relatively high correlation for certain areas.

(a) ERS-1	(b) TOPEX/POSEIDON
(c) AVHRR	(d) ERS-1 - TOPEX/POSEIDON
mm/yr cK/yr

Figure 2. Global distribution of secular change for ERS-1 (a), T/P (b), AVHRR (c), and the difference between ERS-1 and T/P (d). Click individual images for full-size versions.

Figure 3 shows the global distribution of the resolved annual cycle of the ERS-1 and AVHRR results. On top the amplitude and at the bottom the phase. The variability in both temperature and sea level is predominantly annual and concentrated in the northern hemisphere in the western part of the ocean basins. We notice opposite phase in the northern and southern hemisphere, and locally in the sea level in the Northwest Indian Ocean (monsoon), the tropical East Pacific and tropical West Atlantic. As we noticed earlier the sea level lags the surface temperature on average by 1 month. Also the ENSO has a clear contribution to the annual cycle: strong sea level variability in the equatorial Pacific. This is not clear from the temperature plots.

(a) ERS-1 amplitude	(b) AVHRR amplitude
cm 0.5K
(c) ERS-1 phase	(d) AVHRR phase
degrees

Figure 3. Annual cycle of the sea level and the sea surface temperature. Click individual images for full-size versions.

To investigate the dominant patterns in the dynamics of the ocean surface we applied singular value decomposition (SVD) on the ERS-1 data matrix to recover the first six EOFs that explain about 50% of the signal. The results are given in Figure 4. The top panels give the geographical distribution and the bottom panel the V matrix, or time function. Clearly, the annual cycle is very dominant in the first two EOFs. In EOF #1 we notice again the 180° out of phase signature in the sea level in the two hemispheres due to the thermal expansion related to the Earth's motion about the sun. There is also evidence for a biannual cycle. The ENSO and monsoon systems are evident in EOFs #3, #4, and in particular #6. The rather high frequency of 100 days in EOFs #3 and #5 could be due to meso-scale or eddy activity. Interesting to see is that most of the sea level rise can be attributed to EOF #3, which shows traces of ENSO. This could mean that part of the rise we are monitoring is actually due to a very long-wavelength periodic signal.

cm

Figure 4. First sixs EOFs in the ERS-1 altimeter data. Click individual images for full-size versions.

To investigate ENSO in detail and also the capability of both altimeters to monitor this phenomenon, we focussed on the sea level variations in the equatorial region. Figure 5 gives time-longitude plots of the sea level anomalies for the equatorial cross-section (top) and for 5° North (bottom). Left for ERS-1 and right for T/P. Looking at the eastern equatorial Pacific between 90° W and 180° W we observe the evolution of ENSO warm and cold events (high and low anomalies, respectively). The pattern is very regular and seems to have a biannual cycle. We easily notice eastward propagation with an average phase speed of approximately 2.5 m/s, consistent with the first baroclinic mode. No significant differences are found between the ERS-1 and T/P results, which indicates that ERS-1 is really doing a good job. Also westward propagation can be identified at 5° North, of which the phase speed is about 60 cm/s, consistent with the first-mode Rossby waves. The propagation is clearly affected by natural boundaries like islands and bottom topography. There seems to be a barrier at about 175° E, which is due to the Gilbert Islands. The equatorial Atlantic also reveals an annual cycle with high levels bouncing up the east and west coast.

(a) ERS-1 (0°)	(b) TOPEX/POSEIDON (0°)
(c) ERS-1 (5° N)	(d) TOPEX/POSEIDON (5° N)
cm

Figure 5. Time-longitude sections along the equator and 5° N. Click individual images for full-size versions.

Conclusions

References

AVISO, 1994:

AVISO User Handbook: Merged TOPEX/POSEIDON Products, AVI-NT-02-101-CN, Edition 2.1, Toulouse, France, January 1994.

CERSAT, 1994:

Altimeter Products, User Manual, C1-EX-MUT-A21-01-CN, issue 2.6, Plouzane, France, February 1994.

Scharroo, R., P.N.A.M. Visser, and G.J. Mets, 1997:

TOPEX-class orbits for the ERS satellites, Submitted to J. Geophys. Res., January 1997. http://dutlru8.lr.tudelft.nl/ers/precorbs/.

Keywords: ESA European Space Agency - Agence spatiale europeenne, observation de la terre, earth observation, satellite remote sensing, teledetection, geophysique, altimetrie, radar, chimique atmospherique, geophysics, altimetry, radar, atmospheric chemistry