Satellite altimetry was originally intended for open oceans. Monitoring river water levels using altimetry data presents a number of problems:
– the along-track ground resolution: each radar echo is separated by approximately 580 metres, meaning that satellite altimetry is not suited to studying narrow rivers,
– environmental and geophysical corrections models (such as the wet tropospheric correction) have been optimised for open oceans, and may sometimes be nonexistent for continents,
– lastly, radar echoes are subject to perturbations from surrounding terrain (vegetation canopy, topography). When considering the Amazon basin, we have to distinguish floodplain and wetlands water from the main river.

Data used

For studying water surface heights in the Amazon Basin, we are using altimetry measurements over land with the following parameters: geoid model, dry tropospheric correction, wet tropospheric correction and ionospheric correction.
Range values should preferably be computed from waveforms to obtain improved altimetric datasets: for processing radar echoes, retracking algorithms may be adapted to the ground type under study (this is not the intention of this particular ‘Data Use Case’).
Here we are using altimetry measurements from Topex/Poseidon Geophysical Data Records (GDR-M). Other suitable products include Envisat or Jason-1 GDRs, which provide altimetry measurements directly over land, unlike the ERS-1 and 2 missions for which only waveforms are available.

There are a few advantages and disadvantages to these datasets:
– The temporal period for Envisat is 35 days, and for Jason-1 is 10 days (like T/P); consequently, Envisat’s spatial coverage is better than Jason-1,
– On the other hand, T/P was launched in 1992, which means that data series as far back as 1992 are available.

It is not possible to use CorSSH (Corrected Sea Surface Height) data here because only valid ocean measurements are available for this product.


Initial geographic extraction for GDR-M

The first step is to limit the volume of altimetry data to our area of interest.
We therefore select and extract all the available data in GDR-M within the study area (5°S-10°S, 48°E-80°E), from October 1992 to December 2000 (fig 1).
T/P GDR-M are supplied on DVD-ROM.

Distinguishing dry land data from water surface data

Waveforms are perturbed by interfering reflections due to water from wetlands, the vegetation canopy, floodplains and the main river. In this case, we have to identify water surfaces on the satellite ground track and discriminate rivers from flood areas.

Evaluating the measurement density parameter

Along with a given satellite track, all valid data are counted for the whole set of cycles over the entire period. For this purpose we consider all available measurements (at 10-day intervals) inside successive circles of a 3 km radius; circles with less than 50% valid measurements are rejected. The result is a ‘measurement density’ parameter, which is useful as an indicator of the availability of valid measurements (fig 2).

Precise geographic extraction for the Manaus area

We next focus on the T/P ground track near the Rio Negro-Solimoes confluence, where we need to know the location of the land/water boundary as precisely as possible.
To do this, we have to locate intersections of T/P ground tracks and shore boundaries. In this case, accurate and ‘up-to-date’ georeferenced datasets are essential. Information is provided either from satellite images (SPOT for example) or a GIS (fig 3).



couv_amazone_tp_sm dens_amazone_tp_sm
fig 1: T/P ground tracks in the Amazon Basin fig 2: T/P density measurement in the Amazon Basin
fig 3: Manaus area; white rectangle shows geographic extraction, T/P track 63 is in orange

We now know:
– the measurement density distribution for each ground track during the period,
– the location of T/P ground tracks, especially in the Rio Negro-Solimoes confluence area, using the most precise georeferenced geographic sources available.



Computing water surface heights

Water level h’ is computed from GDR-M using:
h’ = s – r

where s is the satellite’s altitude (orbit) and r the range value.

Subtracting geoid and geophysical effects

Altimetry data must be corrected for geoid and propagation effects as follows:
h = h’ – g – i – d – w

where g is the geoid value, i the ionospheric correction, d the dry tropospheric correction and w the wet tropospheric correction. NB: For T/P GDR-M, the wet tropospheric correction is not available for continents; thus in the present study we are unable to take this parameter into account [de Oliveira Campos et al., 2001]. Note also that depending on the area being studied, the dry tropospheric correction is fairly static, which means that sometimes it is possible to compute water surface heights without applying this correction.

Computing a mean water level

For each ground track, a mean water surface height is computed using
hmean = (sum of hi)/n

where h represents the value corresponding to the index i and n the number of altimetry measurements.

Mean water level variations in Manaus


trace_tp_sigma0_densite_sm var_amazon_sm
fig 4. Right: Water level time series in the Amazon for T/P track 63 (3.21°S-3.14°S), in metres; dots represent in situ data from the Manau station. Above: Overview of the geographic window showing the stations; the backscatter coefficient and measurement density parameter have been computed along T/P track 63.