東華大學圖書館 |

The Reconstruction and Analysis of Ocean Submesoscale Surface Data.

紀錄類型:	書目-電子資源 : Monograph/item
正題名/作者:	The Reconstruction and Analysis of Ocean Submesoscale Surface Data./
作者:	Xiao, Qiyu.
面頁冊數:	1 online resource (124 pages)
附註:	Source: Dissertations Abstracts International, Volume: 84-09, Section: B.
Contained By:	Dissertations Abstracts International84-09B.
標題:	Physical oceanography. -
電子資源:	http://pqdd.sinica.edu.tw/twdaoapp/servlet/advanced?query=29997959click for full text (PQDT)
ISBN:	9798374418156

The Reconstruction and Analysis of Ocean Submesoscale Surface Data.
Xiao, Qiyu.

The Reconstruction and Analysis of Ocean Submesoscale Surface Data. - 1 online resource (124 pages)

Source: Dissertations Abstracts International, Volume: 84-09, Section: B.

Thesis (Ph.D.)--New York University, 2023.

Includes bibliographical references

This work tries to develop a methodology to analyze the data received from the Surface Water and Ocean Topography (SWOT) satellite and future generations of observational tools with similar features, by exploiting unnoticed properties of the ocean surface data. The anticipated SWOT satellite has an unprecedented fine scale, an effective resolution of 10-15km, with global coverage. In this resolution, submesoscale activities can be partially resolved and the observations SWOT makes are expected to enrich our understanding of the ocean system.However, there are also challenges. SWOT only offers low-frequency, a 20-days cycle before another measurement at the same spot, sea surface height (SSH) data. It remains a problem of how to turn this data into a useful form and analyze it. There are at least three obstacles that motivate this work. First is that when submesoscale dynamics are involved, the geostrophic balance may not be accurate enough to use, thus there's no trivial way to convert SSH to other interested quantities like velocities. The second issue is that even if we can properly transform SSH to other quantities, how to analyze them when they are only accessible at such a low sampling rate. When we don't have observations every a few hours, we lose track of the development of submesoscale activities that last a few hours to days. We can't use low-pass filters or frequency spectrum to separate out inertial gravity wave (IGW), a component that also gets very active in this fine spatial scale. Last but not least, when we are observing only the ocean surface, our interest is not limited to that. Circulation and transportation in depth are just as crucial, but the quasi-geostrophic framework may not apply in this scale, similar to what we encounter for the reconstruction of other surface quantities from SSH.The solution proposed in this work has two parts and they are tested separately on submesoscalepermitting high-resolution simulations, given that we don't yet have access to SWOT data. In chapter 2, we present our first project that introduces joint distributions of surface kinematics, including vorticity, strain and divergence, as a tool to analyze low sampling rate surface data and induce the tracer transport in depth, trying to tackle the last two issues mentioned above. We show that the vorticity-strain joint distribution can serve as a feature and scale parser and poses few requirements on the data. Conditioning the surface divergence on it shows a similar pattern as conditioning the tracer transport in-depth, and thus it suggests that we can use surface kinematics to reveal transportation in depth. The second part of the solution, presented in chapter 3, focuses on transforming snapshot SSH to surface kinematics with neural networks. We show that neural networks outperform direct C, in particular when IGW is weak. When IGW is strong, neural networks also suffer from distortions of the true target. We analyze the reason for it based on the physical properties of IGW, and also find that divergence is a quantity that naturally filters out the IGW part when the neural network converges. We also show that pretraining with the related dataset can help the model learn fast and better when task-specific data is rare, which may be the case for real observational data.In chapter 1, we introduce features of submesoscale in more detail to help understand the importance and difficulties of this task. We also do a preliminary of neural network that we skip and assume understood when we introduce our configuration in the second project. In chapter 4, we discuss the limitations of the current work and some possible paths for future investigation.

Electronic reproduction.
Ann Arbor, Mich. :
ProQuest,
2023

Mode of access: World Wide Web

ISBN: 9798374418156Subjects--Topical Terms:

3168433
Physical oceanography.
Subjects--Index Terms:

SubmesoscaleIndex Terms--Genre/Form:

542853
Electronic books.

The Reconstruction and Analysis of Ocean Submesoscale Surface Data.
LDR:05061nmm a2200397K 4500 001 2359347
005 20230917193919.5
006 m o d
007 cr mn ---uuuuu
008 241011s2023 xx obm 000 0 eng d
020 $a 9798374418156
035 $a (MiAaPQ)AAI29997959
035 $a AAI29997959
040 $a MiAaPQ $b eng $c MiAaPQ $d NTU
100 1 $a Xiao, Qiyu. $3 3699943
245 1 4 $a The Reconstruction and Analysis of Ocean Submesoscale Surface Data.
264 0 $c 2023
300 $a 1 online resource (124 pages)
336 $a text $b txt $2 rdacontent
337 $a computer $b c $2 rdamedia
338 $a online resource $b cr $2 rdacarrier
500 $a Source: Dissertations Abstracts International, Volume: 84-09, Section: B.
500 $a Advisor: Smith, Shafer.
502 $a Thesis (Ph.D.)--New York University, 2023.
504 $a Includes bibliographical references
520 $a This work tries to develop a methodology to analyze the data received from the Surface Water and Ocean Topography (SWOT) satellite and future generations of observational tools with similar features, by exploiting unnoticed properties of the ocean surface data. The anticipated SWOT satellite has an unprecedented fine scale, an effective resolution of 10-15km, with global coverage. In this resolution, submesoscale activities can be partially resolved and the observations SWOT makes are expected to enrich our understanding of the ocean system.However, there are also challenges. SWOT only offers low-frequency, a 20-days cycle before another measurement at the same spot, sea surface height (SSH) data. It remains a problem of how to turn this data into a useful form and analyze it. There are at least three obstacles that motivate this work. First is that when submesoscale dynamics are involved, the geostrophic balance may not be accurate enough to use, thus there's no trivial way to convert SSH to other interested quantities like velocities. The second issue is that even if we can properly transform SSH to other quantities, how to analyze them when they are only accessible at such a low sampling rate. When we don't have observations every a few hours, we lose track of the development of submesoscale activities that last a few hours to days. We can't use low-pass filters or frequency spectrum to separate out inertial gravity wave (IGW), a component that also gets very active in this fine spatial scale. Last but not least, when we are observing only the ocean surface, our interest is not limited to that. Circulation and transportation in depth are just as crucial, but the quasi-geostrophic framework may not apply in this scale, similar to what we encounter for the reconstruction of other surface quantities from SSH.The solution proposed in this work has two parts and they are tested separately on submesoscalepermitting high-resolution simulations, given that we don't yet have access to SWOT data. In chapter 2, we present our first project that introduces joint distributions of surface kinematics, including vorticity, strain and divergence, as a tool to analyze low sampling rate surface data and induce the tracer transport in depth, trying to tackle the last two issues mentioned above. We show that the vorticity-strain joint distribution can serve as a feature and scale parser and poses few requirements on the data. Conditioning the surface divergence on it shows a similar pattern as conditioning the tracer transport in-depth, and thus it suggests that we can use surface kinematics to reveal transportation in depth. The second part of the solution, presented in chapter 3, focuses on transforming snapshot SSH to surface kinematics with neural networks. We show that neural networks outperform direct C, in particular when IGW is weak. When IGW is strong, neural networks also suffer from distortions of the true target. We analyze the reason for it based on the physical properties of IGW, and also find that divergence is a quantity that naturally filters out the IGW part when the neural network converges. We also show that pretraining with the related dataset can help the model learn fast and better when task-specific data is rare, which may be the case for real observational data.In chapter 1, we introduce features of submesoscale in more detail to help understand the importance and difficulties of this task. We also do a preliminary of neural network that we skip and assume understood when we introduce our configuration in the second project. In chapter 4, we discuss the limitations of the current work and some possible paths for future investigation.
533 $a Electronic reproduction. $b Ann Arbor, Mich. : $c ProQuest, $d 2023
538 $a Mode of access: World Wide Web
650 4 $a Physical oceanography. $3 3168433
650 4 $a Marine geology. $3 3173821
650 4 $a Remote sensing. $3 535394
653 $a Submesoscale
653 $a Surface water
653 $a Ocean topography
653 $a Neural network
653 $a Quasi-geostrophic
655 7 $a Electronic books. $2 lcsh $3 542853
690 $a 0415
690 $a 0556
690 $a 0799
710 2 $a ProQuest Information and Learning Co. $3 783688
710 2 $a New York University. $b Mathematics. $3 1019424
773 0 $t Dissertations Abstracts International $g 84-09B.
856 4 0 $u http://pqdd.sinica.edu.tw/twdaoapp/servlet/advanced?query=29997959 $z click for full text (PQDT)