Repository logo

Browsing by Author "van Leeuwen, Peter Jan, committee member"

Now showing 1 - 4 of 4
  • Thumbnail Image
    ItemOpen Access
    A novel smoother-based data assimilation method for complex CFD
    (Colorado State University. Libraries, 2024) Hurst, Christopher L., author; Gao, Xinfeng, advisor; Guzik, Stephen, advisor; Troxell, Wade, committee member; van Leeuwen, Peter Jan, committee member
    Accurate computational fluid dynamics (CFD) modeling of turbulent flows is necessary for improving fluid-driven engineering designs. Traditional CFD often falls short of providing truly accurate solutions due to inherent uncertainties stemming from modeling assumptions and the chaotic nature of fluid flow. To overcome these limitations, we propose the integration of data assimilation (DA) techniques into CFD simulations. DA, which incorporates observational data into numerical models, offers a promising avenue to enhance predictability by reducing uncertainties associated with initial conditions and model parameters. This research aims to advance our understanding and application of DA for CFD modeling of highly chaotic dynamical systems. This dissertation makes several novel contributions in DA and CFD: i) A novel DA algorithm, the maximum likelihood ensemble smoother (MLES), has been developed and implemented to provide better model parameter estimation and assimilate time-integrated observations while addressing nonlinearity, ii) Multigrid-in-time techniques are applied to enhance the computational efficiency of the MLES by improving the optimization processes, and iii) The MLES+CFD framework has been validated by classical test problems such as the Lorenz 96 model and the Kuramoto-Sivashinsky equation. The effectiveness of the MLES has been demonstrated through a few test problems featuring chaos, discontinuity, or high dimensionality.
  • Thumbnail Image
    ItemOpen Access
    Bayesian models and streaming samplers for complex data with application to network regression and record linkage
    (Colorado State University. Libraries, 2023) Taylor, Ian M., author; Kaplan, Andee, advisor; Fosdick, Bailey K., advisor; Keller, Kayleigh P., committee member; Koslovsky, Matthew D., committee member; van Leeuwen, Peter Jan, committee member
    Real-world statistical problems often feature complex data due to either the structure of the data itself or the methods used to collect the data. In this dissertation, we present three methods for the analysis of specific complex data: Restricted Network Regression, Streaming Record Linkage, and Generative Filtering. Network data contain observations about the relationships between entities. Applying mixed models to network data can be problematic when the primary interest is estimating unconditional regression coefficients and some covariates are exactly or nearly in the vector space of node-level effects. We introduce the Restricted Network Regression model that removes the collinearity between fixed and random effects in network regression by orthogonalizing the random effects against the covariates. We discuss the change in the interpretation of the regression coefficients in Restricted Network Regression and analytically characterize the effect of Restricted Network Regression on the regression coefficients for continuous response data. We show through simulation on continuous and binary data that Restricted Network Regression mitigates, but does not alleviate, network confounding. We apply the Restricted Network Regression model in an analysis of 2015 Eurovision Song Contest voting data and show how the choice of regression model affects inference. Data that are collected from multiple noisy sources pose challenges to analysis due to potential errors and duplicates. Record linkage is the task of combining records from multiple files which refer to overlapping sets of entities when there is no unique identifying field. In streaming record linkage, files arrive sequentially in time and estimates of links are updated after the arrival of each file. We approach streaming record linkage from a Bayesian perspective with estimates calculated from posterior samples of parameters, and present methods for updating link estimates after the arrival of a new file that are faster than fitting a joint model with each new data file. We generalize a two-file Bayesian Fellegi-Sunter model to the multi-file case and propose two methods to perform streaming updates. We examine the effect of prior distribution on the resulting linkage accuracy as well as the computational trade-offs between the methods when compared to a Gibbs sampler through simulated and real-world survey panel data. We achieve near-equivalent posterior inference at a small fraction of the compute time. Motivated by the streaming data setting and streaming record linkage, we propose a more general sampling method for Bayesian models for streaming data. In the streaming data setting, Bayesian models can employ recursive updates, incorporating each new batch of data into the model parameters' posterior distribution. Filtering methods are currently used to perform these updates efficiently, however, they suffer from eventual degradation as the number of unique values within the filtered samples decreases. We propose Generative Filtering, a method for efficiently performing recursive Bayesian updates in the streaming setting. Generative Filtering retains the speed of a filtering method while using parallel updates to avoid degenerate distributions after repeated applications. We derive rates of convergence for Generative Filtering and conditions for the use of sufficient statistics instead of storing all past data. We investigate properties of Generative Filtering through simulation and ecological species count data.
  • Thumbnail Image
    ItemOpen Access
    Cloud property retrievals using polarimetric radar: untangling signals of pristine ice and snow
    (Colorado State University. Libraries, 2020) Kedzuf, Nicholas J., author; Chiu, J. Christine, advisor; van Leeuwen, Peter Jan, committee member; DeMott, Paul, committee member; Chandrasekaran, V., committee member
    Ice and mixed phase clouds are critical components of Earth's climate system via their strong controls on global precipitation distribution and radiation budget. Their microphysical properties have been characterized commonly by polarimetric radar measurements. However, there remains a lack of robust estimates of ice number concentration, due to the difficulty in distinguishing embedded pristine ice from snow aggregates in remote sensing observations. This hinders our ability to study detailed cloud ice microphysical processes from observations. This thesis presents a rigorous method that separates the scattering signals of pristine ice and snow aggregates in scanning polarimetric radar observations to retrieve their respective abundances and sizes for the first time. This method, dubbed ENCORE-ICE, is built on an iterative ensemble retrieval framework. It provides number concentration, median volume diameter, and ice water content of pristine ice and snow aggregates with full error statistics. The retrieved cloud properties are evaluated against in-situ aircraft measurements from a UK field campaign. For a stratiform cloud system with embedded convective features associated with observed ice number concentration of 0.1–10 L–1 and ice water content from 0.01–0.6 g m–3, the retrievals are mainly in the range of 1.0 –15 L–1 and 0.003–0.6 g m–3. To investigate the ice property evolution in a Lagrangian sense, the retrieval method is also applied to along-wind scanning radar measurements from an Atmospheric Radiation Measurement (ARM) campaign in Finland. For the cases presented, snow aggregates are typically of 5–10 mm size in diameter, which is ~10 times larger than pristine ice and thus dominates radar reflectivity. However, the partitioning in ice water content between pristine ice and aggregates varies and largely depends on ice number concentration. More importantly, the retrieved pristine ice number concentration exceeds the predicted concentration of primary ice nuclei at a mid-cloud temperature of –15°C by two orders of magnitude, suggesting possible secondary ice production, one of the outstanding issues in cloud physics. This highlights the potential of using ENCORE-ICE to identify secondary ice production events and understand their trigger mechanisms.
  • Thumbnail Image
    ItemOpen Access
    Understanding and quantifying the uncertainties in satellite warm rain retrievals
    (Colorado State University. Libraries, 2022) Schulte, Richard, author; Kummerow, Christian D., advisor; Bell, Michael, committee member; Boukabara, Sid, committee member; Reising, Steven, committee member; van Leeuwen, Peter Jan, committee member
    Satellite-based oceanic precipitation estimates, particularly those derived from the Global Precipitation Measurement (GPM) satellite and CloudSat, suffer from significant disagreement over regions of the globe where warm rain processes are dominant. Part of the uncertainty stems from differing assumptions about drop size distributions (DSDs). Satellite radar-based retrieval algorithms rely on DSD assumptions that may be overly simplistic, while radiometers further struggle to distinguish cloud water from rain. The aim of this study is to quantify uncertainties related to DSD assumptions in satellite precipitation retrievals, contextualize these uncertainties by comparing them to the uncertainty caused by other important factors like nonuniform beam filling, surface clutter, and vertical variability, and to see if GPM and CloudSat warm rainfall estimates can be partially reconciled if a consistent DSD model is assumed. Surface disdrometer data are used to examine the impact of DSD variability on the ability of three satellite architectures to accurately estimate warm rainfall rates. Two architectures are similar to existing instrument combinations on the GPM Core Observatory and CloudSat, while the third is a theoretical triple frequency radar/radiometer architecture. An optimal estimation algorithm is developed to retrieve rain rates from synthetic satellite measurements, and it is found that the assumed DSD shape can have a large impact on retrieved rain rate, with biases on the order of 100% in some cases. To compare these uncertainties against the effects of horizontal and vertical inhomogeneity, satellite measurements are also simulated using output from a high-resolution cloud resolving model. Finally, the optimal estimation algorithm is used to retrieve rain rates from near-coincident observations made by GPM and CloudSat. The algorithm retrieves more rain from the CloudSat observations than from the GPM observations, due in large part to GPM's insensitivity to light rain. However, the results also suggest an important role for DSD assumptions in explaining the discrepancy. When DSD assumptions are made consistent between the two retrievals, the gap in total accumulation between GPM and CloudSat is reduced by about 25%.