Redes neuronales

The SIPEX’’07 experiment was held during a seven week period between September - October 2007. The RSV - Aurora Australis was in the East Antarctic sea ice zone between 110◦E and

130◦_{Egathering a suite of data related to the physics and ecology of sea ice and its interaction}

with the atmosphere and ocean.

The decision to install RAASTI in the boot of the helicopter was driven by restrictions and difculty in obtaining necessary aircraft permits for its placement inside the cabin. Unfortunately, this resulted in signicant complications with the performance of the radar during the SIPEX’’07 experiment.

RAPPLSew 19 times with the radar onboard. It was determined after the initial data analysis that the collection rate was signicantly below requisite levels during these airborne operations. The cause was found to be related to the vibrations inside the boot of the helicopter - they were affecting RAASTI’’s hard-disk drives, the only moving components located inside the radar. By way of comparison, under laboratory conditions a data rate of 40 Mbytes/s could successfully be

6.2. SEA ICE PHYSICS AND ECOSYSTEM EXPERIMENT, SIPEX’’07 107 0 0.5 1 1.5 2 2.5 3 x 10ï3 ï2 0 2 4 6 8 10 12 Time (s)

YIG driver voltage (V)

Figure 6.4: The voltage waveform used to actuate the YIG to chirp up from 2 to 8 GHz and back to 2 GHz respectively. The segment in red identies the usable segment of the chirp.

sustained; this dropped to approximately 3 Mbytes/s inight conditions and was determined to be an unacceptable rate of data acquisition.

In order to enable even minimal data to be gathered during the numerousights, only ten of the 335 chirps transmitted per second could be recorded. Considering that at its lowest maintainable speed the helicopter ies 200 m in approximately ten seconds, the data gathered at this reduced rate did not provide sufcient information for validation purposes. The suite of problems and debugging successes during theights of SIPEX’’07 are documented in appendix B.

6.2.1 Crane Experiments

Due to vibration issues encountered during airborne experiments, another approach was needed to test if the radar itself was operating correctly. In order to determine if the radar was operational, that is, sensitive to air/snow and snow/ice interfaces, it was removed from the helicopter and attached to a cage pallet, as shown in gure 6.5. Using the ship’’s aft crane, the cage pallet was swung over the side of the ship, and was slowly raised, and lowered over the snow-covered sea ice.

Figure 6.5: The radar inside a cage pallet and attached to the ship’’s crane for testing. Photo credit: J. Lieser. The radar antennas are secured to the bottom of the cage pallet.

6.2. SEA ICE PHYSICS AND ECOSYSTEM EXPERIMENT, SIPEX’’07 109 2 3 4 5 6 7 8 9 10 11 12 13 ï110 ï100 ï90 ï80 ï70 2 3 4 5 6 7 8 9 10 11 12 13 ï110 ï100 ï90 ï80 ï70 Power (dB) 2 3 4 5 6 7 8 9 10 11 12 13 ï110 ï100 ï90 ï80 ï70 Radar Height (m) air/snow snow/ice air/snow snow/ice air/snow snow/ice noise noise noise

Figure 6.6: Three individual radar records are shown, with returns due to the air/snow and snow/ice interfaces seen here to be changing their location as the altitude of the cage pallet changed. A large noise component (possibly due to cross-coupling of the antennas) is stationary.

This novel method enabled a clearer distinction to be made between radar returns due to the surface, due to subsurface properties, and returns due to noise. It was anticipated that signals indicating air/snow and snow/ice interfaces would change in range, while maintaining constant relative separation. Noise components were expected to remain static. Galin et al. [2008], summarises the results of the crane experiment, which demonstrated that the radar was indeed operational. Figures 6.6 and 6.7 show the radar return recorded. The air/snow and snow/ice interface returns are easily identiable from the background noise.

In the crane experiment, converting the time delay between the two observed peaks in the radar return directly to distance (i.e. no correction is made for the change in EM propagation speed due to the snow) provided a thickness estimate of 120 mm, whereas thein-situaverage was found to be 150 mm. Unfortunately density samples of the snow were not taken. However, assuming a dry snow pack with 300kg/m3 density leads to a snow thickness of approximately 150 mm. Due to logistical difculties only one such crane experiment was conducted during the voyage, and consequently detailed error analysis of the radar from a crane platform is not possible.

Figure 6.7: A spectrogram of the stacked radar returns during the crane tests of the radar. Two lines are seen moving across the image, identiable as the air/snow and snow/ice interface returns, whereas the a large noise signal is stationary during the experiment. The two lines of lower intensity at a large range, seeming to mirror the motion of the bright air/snow and snow/ice peaks are likely to be the 2nd harmonics of the two returns.