IE DATA VALIDATION
The IE data has been analysed by the two consortiums (see project members page) and interesting results were released already. The primary outcome of this initial validation was the exciting and unexpected finding that far more information was available in the burst echoes than had previously been considered possible. Several problems were also found, and were investigated and finally solved. At the beginning the reconstruction of the individual echoes appeared to be correct, but some anomalies were present in the reconstructed burst echo sequences. The bursts of individual echoes (in groups of 2000 waveforms) were plotted as shown in next figures, such that the x axis was showing the echo number (ranging from 1 to 2000) while the y axis ranged from 0 to 127 (i.e. samples number). Several examples are shown here below:
This first burst corresponds to some Sahara desert echoes showing low power 'fat' echoes.
The 2nd burst and 3rd bursts are located over the Arctic (ocean/ice echoes). Some nice detail can be recognized that would be lost on averaging.
The next three plots show echoes over moderately sloping terrain in Iran (next first plot), snow cover echoes over Greenland (second plot) and wetlands echoes over Northern Russia (third plot):
What DMU had suspected from their work over land surfaces that over diffuse surfaces that are not very bright to radar, the information is being returned from a small subset of the footprint, is now confirmed and it can be seen that the information content in the burst sequence changes rapidly.
By examining the pulse-to-pulse correlation of the individual echoes three main types of anomalies were identified, that had a great impact on the scientific usefulness of the data (see next three plots: first one with "crenulations" in the window position, second one with "saw-tooth" effect in the window position and third one with discrete jumps in power)):
The jumps in the echo sequences were found to be direcly related to jumps in the window delay (or rx distance parameter).
The majority of the bursts, especially those acquired where the terrain is changing rapidly, showed discontinuities 'crenellations' WITHIN the burst when the sequence was visualised. Obviously, if these bursts had been retracked using any algorithm these discontinuities would have appeared in the height values UNLESS the users had known that they had to pick up the window delay and calculate a range with it at 1800 Hz and use this range for their calculations. This range contained the same jumps as the burst echoes, a consequence of the way the instrument works.
The phase was also affected significantly every time a jump occurred.
The solution was applied in Fourier space, and required the processor to be run twice, once to get the window delay and again to use this information over the same burst to re-apply the fourier transforms.
This removed the problem as it can be shown in the next plots.
The solution was officially implemented in the processor of one of the consortiums (see "raies processor enhancements technical note v. 1.3" for details on the algorithmic changes) and is currently used to generate the IE processed data.
The samples of data present in the 'Data' page have also been generated with the upgraded version of the processor.