Machine learning-based temporal mixture analysis of hypertemporal Antarctic sea ice data

Hypertemporal image (HTI) is often used to exploit the seasonal characteristics of environmental phenomena such as sea ice concentration (SIC). However, it is difficult to analyse the long-term time series acquired at high temporal frequencies and over extensive areas. This study performed temporal mixture analysis (TMA), which is algebraically similar to spectral mixture analysis (SMA), but occurs in the time domain instead of the spectral domain. TMA was used to investigate the temporal characteristics of Antarctic sea ice. Because endmember (EM) selection is critical to the success of both SMA and TMA, it is important to select proper EMs from large quantities of HTI. In this study, a machine learning (ML) technique is incorporated in identifying EMs without prior information to address the limitations of previous research. A fully linear mixing model was then implemented in an attempt to produce more robust and physically meaningful abundance estimates. Experiments that quantitatively and qualitatively evaluated the proposed approaches were conducted. A TMA of high-temporal-dimensional data provides a unique summary of long-term Antarctic sea ice and noise-whitened reconstruction images via inverse processing. Furthermore, comparisons of regional sea ice fractions from experimental results enhance the understanding of the overall Antarctic sea ice changes.