Off-campus UNL users: To download campus access dissertations, please use the following link to log into our proxy server with your NU ID and password. When you are done browsing please remember to return to this page and log out.
Non-UNL users: Please talk to your librarian about requesting this dissertation through interlibrary loan.
Visualization and Modeling of Multivariate Data in Environmental Applications
With the development of new technology, it is easier to collect spatial and spatio-temporal data, common in environmental studies. First, we studied the movement of Arctic sea ice. Using the sea ice trajectories, we motivated the benefits of messy visualizations during exploratory data analysis, which allowed us to derive numerical features that summarize movement. Then, we assessed relevant features also using visualization. We used these relevant numerical features in a clustering algorithm to find clusters of like features. The boundaries between the clusters approximate the location of cracks, or leads, in the sea ice. We used the clusters to develop a non-stationary joint spatio-temporal model of the underlying process that causes the sea ice to move. The model can predict missing points along a trajectory (interpolation). We infer that missing points in a cluster would move similarly to known points in the same cluster. Understanding how the sea ice moves and where leads form provides information that affects the accuracy of climate models, as warm air from the ocean escapes into the colder atmosphere through leads. Through a simulated data set and the sea ice trajectories, we found our lead detection approach found similar leads as other methods. Additionally, our interpolation approach outperformed linear interpolation for non-linear and low-sampled data. Second, we examined visualizations displaying the relationship between crop input application and crop yield. Comprehension of this relationship is essential to find an optimal application rate, as the inefficient application of crop inputs impacts profit and the environment. Maps attempting to display the data currently exist, but these maps do not adhere to the principles of effective chart design. We identified the perceptual issues in each current visualization and proposed plots to mitigate these challenges. By documenting the process of graphical improvement to increase comprehension and ease of use, these ideas can be used with similar data in the agricultural domain to display data better. Visualizing the data more effectively makes the information presented more accessible for people and may allow further insight into the data.
Statistics|Environmental Studies|Artificial intelligence
Kleffner, Alison Ann, "Visualization and Modeling of Multivariate Data in Environmental Applications" (2023). ETD collection for University of Nebraska - Lincoln. AAI30569075.