자료유형  

 학위논문

Control Number  

0016934920

International Standard Book Number  

9798380333795

Dewey Decimal Classification Number  

551.5

Main Entry-Personal Name  

Taylor, Lindsey Michelle.

Publication, Distribution, etc. (Imprint  

[S.l.] : University of Washington., 2023

Publication, Distribution, etc. (Imprint  

Ann Arbor : ProQuest Dissertations & Theses, 2023

Physical Description  

1 online resource(98 p.)

General Note  

Source: Dissertations Abstracts International, Volume: 85-03, Section: B.

General Note  

Advisor: Hakim, Gregory.

Dissertation Note  

Thesis (Ph.D.)--University of Washington, 2023.

Restrictions on Access Note  

This item must not be sold to any third party vendors.

Summary, Etc.  

요약Forecasting on subseasonal-to-seasonal timescales and beyond requires accurate initialization of both the atmosphere and ocean. However, skillful prediction is limited by the reliability of initial conditions and the model's physics. Additionally, properly representing the fully coupled atmosphere−ocean system is computationally demanding. Model error can be reduced with ensemble modeling, which considers the average across multiple independent models. Data assimilation (DA) can also improve forecasts by combining information from observations and models to create a "best" estimate of the final state (i.e., the analysis). Here, we propose to overcome computational barriers by using coupled linear inverse models (LIMs). The LIM is a low-cost empirically-based emulator that considers the covariance statistics of data, i.e., the training data. The goal of this work is to improve atmosphere−ocean analyses and forecasts using ensembles of LIMs that sample across time and data source.We begin by investigating forecasts from LIMs that are trained on ocean-only reanalyses to quantify the impact that the uncertainty in the reanalysis has on LIM forecasts. We find a general disagreement between the ocean-only reanalyses, which contributes to unreliable LIM forecasts and motivates the use of consistent, coupled training data. We next test two types of coupled LIM ensembles: 1) an ensemble that samples randomly across time (i.e., time-sampled ensemble) and 2) an ensemble that samples randomly across time and two reanalyses (i.e., multimodel ensemble; MME). Both ensemble methods are compared to a control LIM that is trained on the full time period of a single reanalysis product from which samples of data are taken. We find that the time-sampled ensemble is able to reproduce the forecast skill of the control with a relatively low percentage of the data sampled. The MME is able to outperform the full-time control taken from either reanalysis for air temperature and sea surface temperature forecasts up to 30-days. Finally, we embed the LIM into a coupled DA framework for evaluation. For most experiments, we find that running an ensemble of DA systems and considering the average analysis leads to better verification than considering just a single DA system.

Subject Added Entry-Topical Term  

Atmospheric sciences.

Subject Added Entry-Topical Term  

Geophysics.

Subject Added Entry-Topical Term  

Meteorology.

Index Term-Uncontrolled  

Coupled atmosphere-ocean

Index Term-Uncontrolled  

Data assimilation

Index Term-Uncontrolled  

Linear inverse model

Index Term-Uncontrolled  

Multimodel ensemble

Added Entry-Corporate Name  

University of Washington Atmospheric Sciences

Host Item Entry  

Dissertations Abstracts International. 85-03B.

Host Item Entry  

Dissertation Abstract International

Electronic Location and Access  

로그인을 한후 보실 수 있는 자료입니다.

Control Number  

joongbu:641935