Surface Ice Velocity Retrieval From MOA Based On NCC Feature Tracking

The velocity of glacier in Antarctica is a fundamental parameter to ice dynamics and projection of sea level rise, and it is as well the key indicator of global climate change. COSI-Corr, an extension of ENVI software, was employed to acquire the horizontal velocity of ice flows throughout the whole Antarctica continent from 2003-2004 and 2008-2009 MOA (MODIS Mosaic of Antarctica) compiled by NSIDC. However, conventional tracking methods severely suffer from spurious matching resulting from ice surface's variation, illumination condition, inappropriate window size etc. So it is indispensable to correct the initial output field contaminated by noises before extracting valuable information. Usually, the low-SNR areas, which denote quite poor quality, are filtered out directly based on some roles of thumb. Here we have some experiments to test performance of FFT (Fast Fourier Transform) and SVD (Singularity Value Decomposition) of optimizing the estimation by cutting image into overlapped tiles. Validation was conducted by comparing the final result with respect to MEaSUREs in typical flow areas including inland stream and ice shelves. The primitive results shows that both methods can reduce RMSE to an extent of 20% 40% but FFT performs more robust. Our result shows that MOA datasets, which highlight true surface morphology, have potential for continental ice surface velocity's retrieval.