Abstract's details
Unsupervised classification of multi-mission altimetry data for open water detection in the Greenland Sea
CoAuthors
Event: 2016 Ocean Surface Topography Science Team Meeting
Session: Instrument Processing: Measurement and retracking (SAR and LRM)
Presentation type: Type Poster
Contribution: PDF file
Abstract:
Estimating sea surface heights by satellite altimetry in the Greenland Sea is challenging because of rapid changing ice coverage and ocean conditions. To obtain reliable and accurate sea surface heights in this region, it is necessary to identify altimeter echoes reflected by leads or polynyas, small open water areas between the sea ice. Reflections from those areas of smooth water are more or less specular, depending on the fraction of sea ice in the altimeter footprint.
In the present investigation, we analyze reflected altimeter pulses, also called waveforms, from different conventional pulse-limited satellite missions in order to distinguish between open water returns and waveforms contaminated by ice. For this purpose, we implement an unsupervised classification approach that does not need any training data. The classification process is based on a set of parameters derived from the waveforms' shapes, for example the peakiness or the maximum power.
After waveform clustering, the classification is validated by several SAR images near the north-east coast of Greenland with small time lags between altimetry pass and the SAR measurements. Moreover, gridded sea ice motion observation are considered whenever necessary. To allow for a quantitative validation an automated approach for detecting open water in the SAR images has been developed.
The classification results will be used to study variations of sea-ice cover in the arctic ocean. Moreover, it enables reliable sea surface height computations for a variety of applications.
In the present investigation, we analyze reflected altimeter pulses, also called waveforms, from different conventional pulse-limited satellite missions in order to distinguish between open water returns and waveforms contaminated by ice. For this purpose, we implement an unsupervised classification approach that does not need any training data. The classification process is based on a set of parameters derived from the waveforms' shapes, for example the peakiness or the maximum power.
After waveform clustering, the classification is validated by several SAR images near the north-east coast of Greenland with small time lags between altimetry pass and the SAR measurements. Moreover, gridded sea ice motion observation are considered whenever necessary. To allow for a quantitative validation an automated approach for detecting open water in the SAR images has been developed.
The classification results will be used to study variations of sea-ice cover in the arctic ocean. Moreover, it enables reliable sea surface height computations for a variety of applications.