Reduction of Rain-Induced Errors for Wind Speed Estimation on SAR Observations Using Convolutional Neural Networks
Synthetic aperture radar (SAR) is known to be able to provide high-resolution estimates of surface wind speed. These estimates usually rely on a geophysical model function (GMF) that has difficulties accounting for nonwind processes, such as rain events. Convolutional neural network, on the other ha...
Main Authors: | , , , , |
---|---|
Format: | Article |
Language: | English |
Published: |
IEEE
2023-01-01
|
Series: | IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing |
Subjects: | |
Online Access: | https://ieeexplore.ieee.org/document/10168970/ |
_version_ | 1797330403679272960 |
---|---|
author | Aurelien Colin Pierre Tandeo Charles Peureux Romain Husson Ronan Fablet |
author_facet | Aurelien Colin Pierre Tandeo Charles Peureux Romain Husson Ronan Fablet |
author_sort | Aurelien Colin |
collection | DOAJ |
description | Synthetic aperture radar (SAR) is known to be able to provide high-resolution estimates of surface wind speed. These estimates usually rely on a geophysical model function (GMF) that has difficulties accounting for nonwind processes, such as rain events. Convolutional neural network, on the other hand, have the capacity to use contextual information and have demonstrated their ability to delimit rainfall areas. By carefully building a large dataset of SAR observations from the Copernicus Sentinel-1 mission, collocated with both GMF and atmospheric model wind speeds as well as rainfall estimates, we were able to train a wind speed estimator with reduced errors under rain. Collocations with in-situ wind speed measurements from buoys show a root mean square error that is reduced by 27% (respectively, 45%) under rainfall estimated at more than 1 mm/h (respectively, 3 mm/h). These results demonstrate the capacity of deep learning models to correct rain-related errors in SAR products. |
first_indexed | 2024-03-08T07:19:21Z |
format | Article |
id | doaj.art-64c582826b93417ca954563025d5aaf8 |
institution | Directory Open Access Journal |
issn | 2151-1535 |
language | English |
last_indexed | 2024-03-08T07:19:21Z |
publishDate | 2023-01-01 |
publisher | IEEE |
record_format | Article |
series | IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing |
spelling | doaj.art-64c582826b93417ca954563025d5aaf82024-02-03T00:01:39ZengIEEEIEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing2151-15352023-01-01168586859410.1109/JSTARS.2023.329123610168970Reduction of Rain-Induced Errors for Wind Speed Estimation on SAR Observations Using Convolutional Neural NetworksAurelien Colin0https://orcid.org/0000-0002-4158-4933Pierre Tandeo1https://orcid.org/0000-0003-1647-8239Charles Peureux2https://orcid.org/0000-0002-1384-0944Romain Husson3https://orcid.org/0000-0001-8275-433XRonan Fablet4https://orcid.org/0000-0002-6462-423XIMT Atlantique, Lab-STICC, UMR CNRS 6285, Brest, FranceIMT Atlantique, Lab-STICC, UMR CNRS 6285, Brest, FranceCollecte Localisation Satellites, Brest, FranceCollecte Localisation Satellites, Brest, FranceIMT Atlantique, Lab-STICC, UMR CNRS 6285, Brest, FranceSynthetic aperture radar (SAR) is known to be able to provide high-resolution estimates of surface wind speed. These estimates usually rely on a geophysical model function (GMF) that has difficulties accounting for nonwind processes, such as rain events. Convolutional neural network, on the other hand, have the capacity to use contextual information and have demonstrated their ability to delimit rainfall areas. By carefully building a large dataset of SAR observations from the Copernicus Sentinel-1 mission, collocated with both GMF and atmospheric model wind speeds as well as rainfall estimates, we were able to train a wind speed estimator with reduced errors under rain. Collocations with in-situ wind speed measurements from buoys show a root mean square error that is reduced by 27% (respectively, 45%) under rainfall estimated at more than 1 mm/h (respectively, 3 mm/h). These results demonstrate the capacity of deep learning models to correct rain-related errors in SAR products.https://ieeexplore.ieee.org/document/10168970/Deep learningoceanographysynthetic aperture radar (SAR)wind |
spellingShingle | Aurelien Colin Pierre Tandeo Charles Peureux Romain Husson Ronan Fablet Reduction of Rain-Induced Errors for Wind Speed Estimation on SAR Observations Using Convolutional Neural Networks IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing Deep learning oceanography synthetic aperture radar (SAR) wind |
title | Reduction of Rain-Induced Errors for Wind Speed Estimation on SAR Observations Using Convolutional Neural Networks |
title_full | Reduction of Rain-Induced Errors for Wind Speed Estimation on SAR Observations Using Convolutional Neural Networks |
title_fullStr | Reduction of Rain-Induced Errors for Wind Speed Estimation on SAR Observations Using Convolutional Neural Networks |
title_full_unstemmed | Reduction of Rain-Induced Errors for Wind Speed Estimation on SAR Observations Using Convolutional Neural Networks |
title_short | Reduction of Rain-Induced Errors for Wind Speed Estimation on SAR Observations Using Convolutional Neural Networks |
title_sort | reduction of rain induced errors for wind speed estimation on sar observations using convolutional neural networks |
topic | Deep learning oceanography synthetic aperture radar (SAR) wind |
url | https://ieeexplore.ieee.org/document/10168970/ |
work_keys_str_mv | AT aureliencolin reductionofraininducederrorsforwindspeedestimationonsarobservationsusingconvolutionalneuralnetworks AT pierretandeo reductionofraininducederrorsforwindspeedestimationonsarobservationsusingconvolutionalneuralnetworks AT charlespeureux reductionofraininducederrorsforwindspeedestimationonsarobservationsusingconvolutionalneuralnetworks AT romainhusson reductionofraininducederrorsforwindspeedestimationonsarobservationsusingconvolutionalneuralnetworks AT ronanfablet reductionofraininducederrorsforwindspeedestimationonsarobservationsusingconvolutionalneuralnetworks |