Discrimination of sea ice leads and floes using Deep Learning applied to Sentinel-3 Ocean and Land Colour Instrument (OLCI) imaging spectrometer

Weibin Chen; Michel Tsamados; Rosie Willatt; So Takao; Connor Nelson; Isobel Lawrence; Sanggyun Lee; David Brockley; Jack Landy; Claude De Rijke-Thomas; Dorsa Shirazi; Julienne Stroeve; Alistair Francis

doi:10.5194/egusphere-egu23-17318

Discrimination of sea ice leads and floes using Deep Learning applied to Sentinel-3 Ocean and Land Colour Instrument (OLCI) imaging spectrometer

Weibin Chen, Michel Tsamados, Rosie Willatt, So Takao, Connor Nelson, Isobel Lawrence, Sanggyun Lee, David Brockley, Jack Landy, Claude De Rijke-Thomas, Dorsa Shirazi, Julienne Stroeve, Alistair Francis

Research output: Contribution to conference › Abstract

13 Downloads (Pure)

Abstract

The Sentinel-3A and Sentinel-3B satellites, launched in February 2016 and April 2018 respectively, build on the legacy of CryoSat-2 by providing high-resolution radar altimetry data over the polar regions up to 81 degrees North. The combination of synthetic aperture radar (SAR) mode altimetry from Sentinel-3A and Sentinel-3B, and the Ocean and Land Colour Instrument (OLCI) imaging spectrometer, results in the creation of the first satellite platform that offers coincident optical imagery and SAR radar altimetry. We utilise these datasets to validate existing surface classification algorithms, in addition to investigating novel applications of deep learning to classify sea-ice from leads. This is important for estimating sea-ice thickness and to predict future changes in the Arctic and Antarctic regions. In particular, we propose the use of Vision Transformers (ViT) for this task and demonstrate their effectiveness, with accuracy reaching above 92%. We compare our automated results with human classification using the software IRIS.

Original language	English
Number of pages	1
DOIs	https://doi.org/10.5194/egusphere-egu23-17318
Publication status	Published - 15 May 2023
Externally published	Yes
Event	European Geosciences Union (EGU) General Assembly 2023 - Austria Center Vienna (ACV), Bruno-Kreisky-Platz 1, 1220, Vienna, Austria Duration: 23 Apr 2023 → 28 Apr 2023 https://www.egu23.eu/

Conference

Conference	European Geosciences Union (EGU) General Assembly 2023
Abbreviated title	EGU23
Country/Territory	Austria
City	Vienna
Period	23/04/23 → 28/04/23
Internet address	https://www.egu23.eu/

Access to Document

10.5194/egusphere-egu23-17318Licence: CC BY

EGU23-17318-printFinal published version, 288 KBLicence: CC BY

Cite this

Chen, W., Tsamados, M., Willatt, R., Takao, S., Nelson, C., Lawrence, I., Lee, S., Brockley, D., Landy, J., Rijke-Thomas, C. D., Shirazi, D., Stroeve, J., & Francis, A. (2023). Discrimination of sea ice leads and floes using Deep Learning applied to Sentinel-3 Ocean and Land Colour Instrument (OLCI) imaging spectrometer. Abstract from European Geosciences Union (EGU) General Assembly 2023, Vienna, Austria. https://doi.org/10.5194/egusphere-egu23-17318

@conference{a7aafb0f7b074d12b23f713a579bd82c,

title = "Discrimination of sea ice leads and floes using Deep Learning applied to Sentinel-3 Ocean and Land Colour Instrument (OLCI) imaging spectrometer",

abstract = "The Sentinel-3A and Sentinel-3B satellites, launched in February 2016 and April 2018 respectively, build on the legacy of CryoSat-2 by providing high-resolution radar altimetry data over the polar regions up to 81 degrees North. The combination of synthetic aperture radar (SAR) mode altimetry from Sentinel-3A and Sentinel-3B, and the Ocean and Land Colour Instrument (OLCI) imaging spectrometer, results in the creation of the first satellite platform that offers coincident optical imagery and SAR radar altimetry. We utilise these datasets to validate existing surface classification algorithms, in addition to investigating novel applications of deep learning to classify sea-ice from leads. This is important for estimating sea-ice thickness and to predict future changes in the Arctic and Antarctic regions. In particular, we propose the use of Vision Transformers (ViT) for this task and demonstrate their effectiveness, with accuracy reaching above 92%. We compare our automated results with human classification using the software IRIS. ",

author = "Weibin Chen and Michel Tsamados and Rosie Willatt and So Takao and Connor Nelson and Isobel Lawrence and Sanggyun Lee and David Brockley and Jack Landy and Rijke-Thomas, {Claude De} and Dorsa Shirazi and Julienne Stroeve and Alistair Francis",

year = "2023",

month = may,

day = "15",

doi = "10.5194/egusphere-egu23-17318",

language = "English",

note = "European Geosciences Union (EGU) General Assembly 2023, EGU23 ; Conference date: 23-04-2023 Through 28-04-2023",

url = "https://www.egu23.eu/",

}

Chen, W, Tsamados, M, Willatt, R, Takao, S, Nelson, C, Lawrence, I, Lee, S, Brockley, D, Landy, J, Rijke-Thomas, CD, Shirazi, D, Stroeve, J & Francis, A 2023, 'Discrimination of sea ice leads and floes using Deep Learning applied to Sentinel-3 Ocean and Land Colour Instrument (OLCI) imaging spectrometer', European Geosciences Union (EGU) General Assembly 2023, Vienna, Austria, 23/04/23 - 28/04/23. https://doi.org/10.5194/egusphere-egu23-17318

TY - CONF

T1 - Discrimination of sea ice leads and floes using Deep Learning applied to Sentinel-3 Ocean and Land Colour Instrument (OLCI) imaging spectrometer

AU - Chen, Weibin

AU - Tsamados, Michel

AU - Willatt, Rosie

AU - Takao, So

AU - Nelson, Connor

AU - Lawrence, Isobel

AU - Lee, Sanggyun

AU - Brockley, David

AU - Landy, Jack

AU - Rijke-Thomas, Claude De

AU - Shirazi, Dorsa

AU - Stroeve, Julienne

AU - Francis, Alistair

PY - 2023/5/15

Y1 - 2023/5/15

N2 - The Sentinel-3A and Sentinel-3B satellites, launched in February 2016 and April 2018 respectively, build on the legacy of CryoSat-2 by providing high-resolution radar altimetry data over the polar regions up to 81 degrees North. The combination of synthetic aperture radar (SAR) mode altimetry from Sentinel-3A and Sentinel-3B, and the Ocean and Land Colour Instrument (OLCI) imaging spectrometer, results in the creation of the first satellite platform that offers coincident optical imagery and SAR radar altimetry. We utilise these datasets to validate existing surface classification algorithms, in addition to investigating novel applications of deep learning to classify sea-ice from leads. This is important for estimating sea-ice thickness and to predict future changes in the Arctic and Antarctic regions. In particular, we propose the use of Vision Transformers (ViT) for this task and demonstrate their effectiveness, with accuracy reaching above 92%. We compare our automated results with human classification using the software IRIS.

AB - The Sentinel-3A and Sentinel-3B satellites, launched in February 2016 and April 2018 respectively, build on the legacy of CryoSat-2 by providing high-resolution radar altimetry data over the polar regions up to 81 degrees North. The combination of synthetic aperture radar (SAR) mode altimetry from Sentinel-3A and Sentinel-3B, and the Ocean and Land Colour Instrument (OLCI) imaging spectrometer, results in the creation of the first satellite platform that offers coincident optical imagery and SAR radar altimetry. We utilise these datasets to validate existing surface classification algorithms, in addition to investigating novel applications of deep learning to classify sea-ice from leads. This is important for estimating sea-ice thickness and to predict future changes in the Arctic and Antarctic regions. In particular, we propose the use of Vision Transformers (ViT) for this task and demonstrate their effectiveness, with accuracy reaching above 92%. We compare our automated results with human classification using the software IRIS.

U2 - 10.5194/egusphere-egu23-17318

DO - 10.5194/egusphere-egu23-17318

M3 - Abstract

T2 - European Geosciences Union (EGU) General Assembly 2023

Y2 - 23 April 2023 through 28 April 2023

ER -