AMSR-E sea ice drift validation

1. The AMSR-E product

Since 2002, AMSR-E at 89 GHz provides temperature brightness maps with a pixel resolution of 6,25 km, which is interesting to detect small drifts. From these maps, drifts are inferred, grid spacing of the drift map is 31,25 km.

AMSR-E drifts are compared with buoys over the 2002-2003 winter, these results are also compared with Merged/buoy drifts comparison over this winter.

2. Comparison with buoys

Figure 3 shows the comparison between buoys and AMSR-E drifts. Compared with figure 1 for Merged drifts, this diagram shows a better agreement between the AMSR-E and buoys than that for Merged/buoys. The mean bias of the difference AMSR-E/buoys is about 0 km, the standard deviation is 6,7+/-0,2 km (of which 4,8 km are uncorrelated noise), lower than the 8,1+/-0,2 km (of which 5,8 km are uncorrelated noise) for the Merged/buoys comparison over the same winter. This induces an improvement of 3,3 km on the drift noise level with AMSR-E drifts (calculated with variances additivity).day-lag, winter 2002-2003. Colors are probability expressed in logarithm scale.

In ice speeds, standard deviations of the difference between AMSR-E and buoys are lower than that of the Merged/buoys comparison :

  • at 3 day-lag : 2,60+/-0,08 cm/s for AMSR-E/buoys and 3,14+/-0,09 cm/s for Merged/buoys
  • at 6 day-lag : 1,59+/-0,05 cm/s for AMSR-E/buoys and 2,08+/-0,06 cm/s for Merged/buoys

The angle difference between AMSR-E and buoys has a mean bias of 0,9+/-1,5°, the standard deviation is the same order (35,2+/-1,1°) than that of Merged/buoys (35,1+/-1,0°). Figure 4 shows the angle difference as a function of AMSR-E drifts, angle difference is important since drifts are lower than 30 km (about 5 pixels). A comparison over several winters is needed in order to show the benefit of AMSR-E on the angle comparison which is not obvious on this winter.

North and East components comparisons show a standard deviation of the difference of 6,4+/-0,2 km for AMSR-E/buoys whereas it is about 7,2+/-0,2 km for the Merged/buoys.

These results show that AMSR-E drifts are in agreement with buoys drifts, mean biases and standard deviation of the AMSR-E/buoys difference are lower than that of Merged over the same winter.