Applications of Deep Learning-Based Super-Resolution for Sea Surface Temperature Reconstruction

Bo, Ping; Su, Fenzhen; Han, Xingxing; Meng, Ying

doi:10.1109/jstars.2020.3042242

Cited by 13 publications

(16 citation statements)

References 45 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The National Snow and Ice Data Center (NSIDC) provides the MOD29 IST product, wherein all possible cloudcontaminated pixels are removed according to the cloud mask from the MOD35 product. However, upon visual inspection, a number of lead areas with ocean fog or plume (Qu et al, 2019;Fett et al, 1997) can be mistakenly marked as clouds, which can cause a loss of potential leads. Therefore, instead of the MOD29 IST product, the MODIS Level-1B product MOD021KM acquired by the sensor aboard the Terra satellite is used in this study, and it was obtained from the US National Aeronautics and Space Administration's Level 1 and Atmosphere Archive and Distribution System Distributed Active Archive Center (https://ladsweb.…”

Section: Modis Datamentioning

confidence: 99%

“…2a, which could impact the estimation accuracy of the THF. A median filtering method (Eppler and Full, 1992;Qu et al, 2019) was used to remove the noise from the Landsat-8 IST image caused by this type of artifact, as shown in Fig. 2b.…”

Section: Landsat-8 Datamentioning

confidence: 99%

“…Remotely sensed satellite images are a promising data source that has been often used in estimating the THF over leads (Qu et al, 2019) because in situ measurement is always difficult to conduct due to harsh weather conditions typical in polar regions. To calculate the THF over leads using the remotely sensed data, both a lead map and the corresponding ice surface temperature (IST) are required (Aulicino et al, 2018).…”

Section: Introductionmentioning

confidence: 99%

“…Specifically, in these methods, first, the relationship between an image with a fine spatial resolution and the corresponding image with a coarse spatial resolution is established through the training process with a large amount of training data, and then the trained model is used to super-resolve the testing coarse-spatial-resolution image. These methods have been successfully applied to the SR of sea surface temperature (SST) imagery (Ping et al, 2021) and SR mapping of land cover Jia et al, 2019). Therefore, the CNN-based SR methods have great potential in the field of image downscaling, especially fine-spatial-resolution IST images and lead maps that are useful in THF estimation.…”

Section: Introductionmentioning

confidence: 99%

“…This study proposes a CNN-based method for a subpixelscale estimation of the THF from MODIS TIR images named the deep-learning-based subpixel THF estimation (Deep-STHF) method. Specifically, the DeepSTHF method uses two CNNs to simultaneously produce a Landsat-like IST image and the corresponding binary map of leads from the MODIS IST images, and then it employs the generated Landsat-like IST image and lead map to estimate the THF over leads using an aerodynamic bulk formula (Marcq and Weiss, 2012;Qu et al, 2019;Renfrew et al, 2002). This study provides a new perspective for solving the mixed pixel problem in THF estimation from remotely sensed images by extracting subpixel-level information.…”

Section: Introductionmentioning

confidence: 99%

See 4 more Smart Citations

Estimating subpixel turbulent heat flux over leads from MODIS thermal infrared imagery with deep learning

Yin

et al. 2021

The Cryosphere

View full text Add to dashboard Cite

Abstract. The turbulent heat flux (THF) over leads is an important parameter for climate change monitoring in the Arctic region. THF over leads is often calculated from satellite-derived ice surface temperature (IST) products, in which mixed pixels containing both ice and open water along lead boundaries reduce the accuracy of calculated THF. To address this problem, this paper proposes a deep residual convolutional neural network (CNN)-based framework to estimate THF over leads at the subpixel scale (DeepSTHF) based on remotely sensed images. The proposed DeepSTHF provides an IST image and the corresponding lead map with a finer spatial resolution than the input IST image so that the subpixel-scale THF can be estimated from them. The proposed approach is verified using simulated and real Moderate Resolution Imaging Spectroradiometer images and compared with the conventional cubic interpolation and pixel-based methods. The results demonstrate that the proposed CNN-based method can effectively estimate subpixel-scale information from the coarse data and performs well in producing fine-spatial-resolution IST images and lead maps, thereby providing more accurate and reliable THF over leads.

show abstract

Section: Modis Datamentioning

confidence: 99%

Section: Landsat-8 Datamentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 3 more Smart Citations

Estimating subpixel turbulent heat flux over leads from MODIS thermal infrared imagery with deep learning

Yin

et al. 2021

The Cryosphere

View full text Add to dashboard Cite

show abstract

Deep learning for sea surface temperature applications: A comprehensive bibliometric analysis and methodological approach

Boufeniza,

Jingjia,

Abdela

et al. 2024

Geography and Environment

View full text Add to dashboard Cite

This study explored the potential application of deep learning techniques in sea surface temperature (SST) investigations using a mixed method, bibliometric analysis and methodological approach. CiteSpace software was utilized for a bibliometric study on 137 academic publications from 2018 to 2023. Various databases were employed for methodological analysis, which involved examining publications based on models, methodologies, applications and research areas. The data were manually organized in a relational framework of an SQL database. The analysis underscored China's prominence as a leader in the extensive research devoted to this field. The United States of America and the United Kingdom played pivotal roles in providing the essential data that served as the foundation for these studies. Moreover, the long short‐term memory (LSTM) algorithm was the predominant computational deep learning algorithm extensively used in this specific context. The analysis highlighted significant knowledge gaps in areas such as SST forecasting, modelling, satellite remote sensing, extreme events and data reconstruction. Future scientists need to show more interest in these and related subjects, while Chinese and American scientists should prioritize paper quality over quantity. Additionally, fostering stronger collaborations between universities and institutions is vital for further advancements. Ultimately, this study offers valuable insights into hotspot research areas and development processes, establishing the foundation for research and suggesting possible avenues for future development. The results of this evaluation serve as an essential guide for researchers and modellers involved in prediction initiatives using deep learning.

show abstract

End-to-End Kalman Filter in a High Dimensional Linear Embedding of the Observations

Ouala

Chapron

Collard³

et al. 2022

Mathematics of Planet Earth

View full text Add to dashboard Cite

Data assimilation techniques are the state-of-the-art approaches in the reconstruction of a spatio-temporal geophysical state such as the atmosphere or the ocean. These methods rely on a numerical model that fills the spatial and temporal gaps in the observational network. Unfortunately, limitations regarding the uncertainty of the state estimate may arise when considering the restriction of the data assimilation problems to a small subset of observations, as encountered for instance in ocean surface reconstruction. These limitations motivated the exploration of reconstruction techniques that do not rely on numerical models. In this context, the increasing availability of geophysical observations and model simulations motivates the exploitation of machine learning tools to tackle the reconstruction of ocean surface variables. In this work, we formulate sea surface spatio-temporal reconstruction problems as state space Bayesian smoothing problems with unknown augmented linear dynamics. The solution of the smoothing problem, given by the Kalman smoother, is written in a differentiable framework which allows, given some training data, to optimize the parameters of the state space model.

show abstract

Applications of Deep Learning-Based Super-Resolution for Sea Surface Temperature Reconstruction

Cited by 13 publications

References 45 publications

Estimating subpixel turbulent heat flux over leads from MODIS thermal infrared imagery with deep learning

Estimating subpixel turbulent heat flux over leads from MODIS thermal infrared imagery with deep learning

Deep learning for sea surface temperature applications: A comprehensive bibliometric analysis and methodological approach

End-to-End Kalman Filter in a High Dimensional Linear Embedding of the Observations

Contact Info

Product

Resources

About