The Aral Sea level has been shrinking since the 60s, due to irrigation. The recession is estimated as nearly half of its volume over the last fifteen years. In 1989, the Sea split into two basins, the Great and the Small Aral. The small Aral is now stabilized, due to a dam, and its level may even be rising, but the Great Aral is still drying up. Altimetry satellites measure sea level continuously over the long term and thus enable us to monitor variations in the Aral Sea.
We will take advantage of the more than 10 years continuous Topex/Poseidon dataset, and use the GDR-M 1Hz data, since the Aral Sea is (still) a big enough body of water that high-resolution data are not absolutely mandatory to study its variations.
We will use the Broadview Radar Altimetry Toolbox to have a look at the Aral sea level variations along the years.
The exact coordinates of the actual Aral Sea have changed through the years, and the land mask provided within the data is no longer valid in that case (i.e. places flagged as “water” are now dry). Roughly, the Big (South) Aral is between 59°N to 61°N, and 44°E to 46°E
Data selection and BRAT Datasets definition
We have to select the track(s) we wish to work on (we do not need the whole cycles of T/P data).
To select the right track, you can have a look at the ground track maps that may be provided with the data handbook. You can also use a pass locator using Google Earth available through Aviso.
The “best” Topex/Poseidon track that passes over the Aral Sea is the #107 (up till cycle 365, when the satellite orbit was changed). We took all available pass 107 data file from cycle 010 to cycle 364.
For Envisat, you could use pass #126 or 253.
Name the dedicated BRAT workspace you are using for this job. Within this workspace, name your datasets; we will need 3 datasets: one for a cycle in 1992-1993 (e.g. cycle 010), one for a cycle in 2002 (e.g. 360), and one with the whole series of the pass #107.
BRAT Operations definition
In the ‘Operations’ tab, name your operation, then select your dataset and data. In ‘Data Computation’ keep ‘MEAN’ selected.
Enter your data expression: you will be computing the sea level with respect to the geoid, as for the “Amazon” data use case. The expression using T/P GDR field names is:
sat_alt – h_alt – h_geo – h_set – h_pol – iono_cor – dry_corr – wet_h_rad
(satellite altitude – altimetric range – corrections)
name it, e.g., LLH (for “Lake Level Height”).
For the X field the variable is ‘Lon_Tra’ when looking at one cycle and tim_moy_1 when working with the whole series.
As we will see, data editing is in order, to remove some less accurate data. The formula we will be using is:
geo_bad_1.water_land_distribution == 0 && is_bounded(-130,(sat_alt-h_alt),100) && nval_h_alt >= 5 && is_bounded(0,rms_h_alt,0.1) && is_bounded(-2.5,dry_corr,-1.9) && is_bounded(-0.500,wet_h_rad,-0.001) && is_bounded(-0.400,iono_cor,0.040) && is_bounded(7,sigma0_k, 30) && is_bounded(-1,h_set,1) && is_bounded(-0.150,h_pol,0.150) && is_bounded(0,att_wvf,0.4)
It is an adaptation of the Ocean data editing (provided within BRAT), with some fields removed.
Aral Sea level along the track: need of data editing
We will have a look at what we can see when plotting LLH for (e.g.) cycle 010 with respect to longitude.
If we plot “just” the level wrt longitude, in fact, we see more than the Aral Sea level. Even using the land/water flag (geo_bad_1.water_land_distribution == 0, which means that the data are considered as taken over open seas), we still have unwanted data — in part because of the changes of the Aral Sea extend, but not only.
fig 2. Lake level height for the Aral sea without data editing, T/P cycle 010 (left); The measurements over the sea is the “flat” area in the middle of the curve (between, roughly, 59.1° and 60.3°N). Right, a more problematic cycle, cycle 353, for which no such clear signal can be seen.
We will have a look at some of the data fields that are used in data editing:
- RMS of altimetric range, which measure the variability of the altimetry measurement: over an open body of water, this variability should be rather low (between 0 and 0.1 m)
- Antenna mispointing, which can come from a change in surface, the radar sensing pulse previously reflected off-nadir
- sigma0, the backscatter coefficient, which is the ration of the altimeter radar pulse power that is reflected on the surface and measured by the radar. Abrupt changes in this coefficient can also come from a change in surface reflection.
fig 3. Several data fields plotted for T/P cycle 10 over the Aral Sea area: RMS of the signal, antenna mispointing, Sigma0. They all show abrupt differences between the measurements outside and over the Sea.
Depending on the water body, fine tuning of the threshold condition used in data editing has to be done.
Aral Sea level along the track at two different time
Here we will have a look at the Aral Sea level along the track in December 1992 (cycle 010) and June 2003 (cycle 360). We compute LLH for both cycles, using the editing given in the previous page. In the ‘Views’ menu, you name your plot, give the plot a title, select both computed LLH fields and click on ‘execute’ to view them (you can restrict X from 58.5 to 61°N).
fig 4. Topex/Poseidon along-track LLH over the Aral Sea for cycle 010 (December 1992, red) and cycle 360 (June 2002, green). The slight curvature of the height is due to the geoid model inaccuracies over this area. The level is obviously much lower in 2002, and the extend of the Sea is also much smaller.
Compute the Aral sea level variation over 10 years
Using the dataset with all the pass #107 from cycle 010 to 364, we will be able to plot the evolution of the Aral Sea over the whole period.
By using the time in days (‘tim_moy_1’) as abscissa, a mean is automatically done for each day for all available data on this day. We will use geographical selection to restrict the data averaged to the Aral Sea (add ‘&& is_bounded(59.06,lon_tra,60.32) && is_bounded(44.11,lat_tra,45.52)’ in the Select expression box)