Physics
Permanent URI for this communityhttp://hdl.handle.net/1903/2269
Browse
5 results
Search Results
Item Combining Physics-based Modeling, Machine Learning, and Data Assimilation for Forecasting Large, Complex, Spatiotemporally Chaotic Systems(2023) Wikner, Alexander Paul; Ott, Edward; Physics; Digital Repository at the University of Maryland; University of Maryland (College Park, Md.)We consider the challenging problem of forecasting high-dimensional, spatiotemporally chaotic systems. We are primarily interested in the problem of forecasting the dynamics of the earth's atmosphere and oceans, where one seeks forecasts that (a) accurately reproduce the true system trajectory in the short-term, as desired in weather forecasting, and that (b) correctly capture the long-term ergodic properties of the true system, as desired in climate modeling. We aim to leverage two types of information in making our forecasts: incomplete scientific knowledge in the form of an imperfect forecast model, and past observations of the true system state that may be sparse and/or noisy. In this thesis, we ask if machine learning (ML) and data assimilation (DA) can be used to combine observational information with a physical knowledge-based forecast model to produce accurate short-term forecasts and consistent long-term climate dynamics. We first describe and demonstrate a technique called Combined Hybrid-Parallel Prediction (CHyPP) that combines a global knowledge-based model with a parallel ML architecture consisting of many reservoir computers and trained using complete observations of the system's past evolution. Using the Kuramoto-Sivashinsky equation as our test model, we demonstrate that this technique produces more accurate short-term forecasts than either the knowledge-based or the ML component model acting alone and is scalable to large spatial domains. We further demonstrate using the multi-scale Lorenz Model 3 that CHyPP can incorporate the effect of unresolved short-scale dynamics (subgrid-scale closure). We next demonstrate how DA, in the form of the Ensemble Transform Kalman Filter (ETKF), can be used to extend the Hybrid ML approach to the case where our system observations are sparse and noisy. Using a novel iterative scheme, we show that DA can be used to obtain training data for successive generations of hybrid ML models, improving the forecast accuracy and the estimate of the full system state over that obtained using the imperfect knowledge-based model. Finally, we explore the commonly used technique of adding observational noise to the ML model input during training to improve long-term stability and climate replication. We develop a novel training technique, Linearized Multi-Noise Training (LMNT), that approximates the effect of this noise addition. We demonstrate that reservoir computers trained with noise or LMNT regularization are stable and replicate the true system climate, while LMNT allows for greater ease of regularization parameter tuning when using reservoir computers.Item Transport in Rayleigh-Stable Experimental Taylor-Couette Flow and Granular Electrification in a Shaking Experiment(2015) Nordsiek, Freja; Lathrop, Daniel P; Physics; Digital Repository at the University of Maryland; University of Maryland (College Park, Md.)This dissertation consists of two projects: Rayleigh-stable Taylor-Couette flow and granular electrification. Taylor-Couette flow is the fluid flow in the gap between two cylinders rotating at different rates. Azimuthal velocity profiles, dye visualization, and inner cylinder torques were measured on two geometrically similar Taylor-Couettes with axial boundaries attached to the outer cylinder, the Maryland and Twente T3C experiments. This was done in the Rayleigh stable regime, where the specific angular momentum increases radially, which is relevant to astrophysical and geophysical flows and in particular, stellar and planetary accretion disks. The flow substantially deviates from laminar Taylor-Couette flow beginning at moderate Reynolds number. Angular momentum is primarily transported to the axial boundaries instead of the outer cylinder due to Ekman pumping when the inner cylinder is rotating faster than the outer cylinder. A phase diagram was constructed from the transitions identified from torque measurements taken over four decades of the Reynolds number. Flow angular velocities larger and smaller than both cylinders were found. Together, these results indicate that experimental Taylor-Couette with axial boundaries attached to the outer cylinder is an imperfect model for accretion disk flows. Thunderstorms, thunder-snow, volcanic ash clouds, and dust storms all display lightning, which results from electrification of droplets and particles in the atmosphere. While lightning is fairly well understood (plasma discharge), the mechanisms that result in million-volt differences across the storm are not. A novel granular electrification experiment was upgraded and used to study some of these mechanisms in the lab. The relative importance of collective interactions between particles versus particle properties (material, size, etc.) on collisional electrification was investigated. While particle properties have an order of magnitude effect on the strength of macroscopic electrification, all particle types electrified with dynamics that suggest a major role for collective interactions in electrification. Moreover, mixing two types of particles together does not lead to increased electrification except for specific combinations of particles which clump, which further points towards the importance of collective phenomena. These results help us better understand the mechanisms of electrification and lightning generation in certain atmospheric systems.Item Information Synthesis Across Scales in Atmospheric State Estimation: Theory and Numerical Experiments(2015) Kretschmer, Matthew; Ott, Edward; Hunt, Brian R; Physics; Digital Repository at the University of Maryland; University of Maryland (College Park, Md.)This thesis studies the benefits of simultaneously considering system information from different sources when performing ensemble data assimilation. In particular, in Chapter 2 we consider ensemble data assimilation using both a global dynamical model and climatological forecast error information, and, in Chapters 3 and 4, using both a global dynamical model and at least one higher-resolution limited-area dynamical model. Focus is given to applying data assimilation for atmospheric state estimation. Introductory material on ensemble forecasting is given in Chapter 1. In Chapter 2, I first investigate how the forecast background-error climatology can be used to help improve state estimates, and subsequent forecasts initialized from those state estimates. ``Climatological perturbations'' derived from an estimate of the background-error covariance matrix are added to the dynamic ensemble that has been forecasted from the previous analysis time, enlarging the space of possible analysis increments. Numerical experiments on a one-dimensional toy model test this method and illustrate that climatologically augmenting the dynamical forecast ensemble during the analysis has a positive impact on state estimation and forecast accuracy. Chapter 3 studies data assimilation that considers state information from various spatial scales. In practice, it is common for regional-scale weather forecasts to be created using limited-area atmospheric models which have relatively high spatial resolution. Limited-area model forecasts require lateral boundary conditions, which often come from a lower resolution forecast model (with different model physics) defined over a larger, often global, domain. Here I describe how data assimilation may be performed on a composite forecast state containing information from all available forecast models, and show results from numerical experiments that detail the benefits of this approach. Chapter 4 of this thesis explores forecast model bias, which is the result of uncertain, unknown or incorrect model physics. I adapt a strategy for correcting forecast model bias to use when performing data assimilation using the composite state method described in Chapter 3. In numerical experiments, I test this bias correction strategy for differently biased global and limited-area models, and observe that analysis and forecast accuracy is dramatically improved when compared to forecasts made without bias correction.Item Neutral Gas and Plasma Interactions in the Polar Cusp(2012) Olson, David K.; Moore, Thomas E.; Coplan, Michael A.; Physics; Digital Repository at the University of Maryland; University of Maryland (College Park, Md.)When the solar wind interacts with the Earth's magnetosphere, both energy and matter can be transferred across the magnetopause boundary. This transfer gives rise to numerous phenomena, including ion outflow and neutral upwelling in the polar cusps. These processes are caused by a transfer of energy to the ionospheric plasma and neutral gas through various mechanisms. The heated plasma or gas expands, increasing the density of the atmosphere at high altitudes by as much as a factor of two, and injecting ionospheric plasma into and even outside of the magnetosphere. These two phenomena are examined in two ways: A novel high energy (0.1--10 keV) spectrograph for ionospheric cusp ions was designed as part of the Rocket Experiment for Neutral Upwelling (RENU), a sounding rocket campaign carried out at the northern polar cusp to observe the electrodynamic properties of the cusp during a neutral upwelling event. This instrument is called the KeV Ion Magnetic Spectrograph (KIMS). Ion outflow in the ionosphere has shown evidence of correlation with both Poynting flux and soft electron precipitation in the cusp. The heat input from these energy sources might also affect neutral gas in the ionosphere, contributing to upwelling phenomena seen at the dayside cusp. Using data from the Fast Auroral Snapshot Explorer (FAST) and the Challenging Minisatellite Payload (CHAMP) satellites, correlations of electromagnetic and particle energy inputs are examined with both ion outflow and neutral upwelling in the cusp. The added ability to process large quantities of data quickly and reference the data between separate satellites in this statistical survey gives clues to the consistency of the observed correlations with ion outflow over time and to the relative importance of these energy sources in the neutral upwelling phenomenon. It also provides the ability to understand these connections in a broad spectrum of conditions of the Sun and solar wind as well as in the Earth's magnetosphere.Item Information flow in an atmospheric model and data assimilation(2011) Yoon, Young-noh; Ott, Edward; Physics; Digital Repository at the University of Maryland; University of Maryland (College Park, Md.)Weather forecasting consists of two processes, model integration and analysis (data assimilation). During the model integration, the state estimate produced by the analysis evolves to the next cycle time according to the atmospheric model to become the background estimate. The analysis then produces a new state estimate by combining the background state estimate with new observations, and the cycle repeats. In an ensemble Kalman filter, the probability distribution of the state estimate is represented by an ensemble of sample states, and the covariance matrix is calculated using the ensemble of sample states. We perform numerical experiments on toy atmospheric models introduced by Lorenz in 2005 to study the information flow in an atmospheric model in conjunction with ensemble Kalman filtering for data assimilation. This dissertation consists of two parts. The first part of this dissertation is about the propagation of information and the use of localization in ensemble Kalman filtering. If we can perform data assimilation locally by considering the observations and the state variables only near each grid point, then we can reduce the number of ensemble members necessary to cover the probability distribution of the state estimate, reducing the computational cost for the data assimilation and the model integration. Several localized versions of the ensemble Kalman filter have been proposed. Although tests applying such schemes have proven them to be extremely promising, a full basic understanding of the rationale and limitations of localization is currently lacking. We address these issues and elucidate the role played by chaotic wave dynamics in the propagation of information and the resulting impact on forecasts. The second part of this dissertation is about ensemble regional data assimilation using joint states. Assuming that we have a global model and a regional model of higher accuracy defined in a subregion inside the global region, we propose a data assimilation scheme that produces the analyses for the global and the regional model simultaneously, considering forecast information from both models. We show that our new data assimilation scheme produces better results both in the subregion and the global region than the data assimilation scheme that produces the analyses for the global and the regional model separately.