Lagrangian Spatio-Temporal Covariance Functions for Multivariate Nonstationary Random Fields

  • Mary Lai O. SalvaƱa

Student thesis: Master's Thesis


In geostatistical analysis, we are faced with the formidable challenge of specifying a valid spatio-temporal covariance function, either directly or through the construction of processes. This task is di cult as these functions should yield positive de nite covariance matrices. In recent years, we have seen a ourishing of methods and theories on constructing spatiotemporal covariance functions satisfying the positive de niteness requirement. The current state-of-the-art when modeling environmental processes are those that embed the associated physical laws of the system. The class of Lagrangian spatio-temporal covariance functions ful lls this requirement. Moreover, this class possesses the allure that they turn already established purely spatial covariance functions into spatio-temporal covariance functions by a direct application of the concept of Lagrangian reference frame. In the three main chapters that comprise this dissertation, several developments are proposed and new features are provided to this special class. First, the application of the Lagrangian reference frame on transported purely spatial random elds with second-order nonstationarity is explored, an appropriate estimation methodology is proposed, and the consequences of model misspeci cation is tackled. Furthermore, the new Lagrangian models and the new estimation technique are used to analyze particulate matter concentrations over Saudi Arabia. Second, a multivariate version of the Lagrangian framework is established, catering to both secondorder stationary and nonstationary spatio-temporal random elds. The capabilities of the Lagrangian spatio-temporal cross-covariance functions are demonstrated on a bivariate reanalysis climate model output dataset previously analyzed using purely spatial covariance functions. Lastly, the class of Lagrangian spatio-temporal cross-covariance functions with multiple transport behaviors is presented, its properties are explored, and its use is demonstrated on a bivariate pollutant dataset of particulate matter in Saudi Arabia. Moreover, the importance of accounting for multiple transport behaviors is discussed and validated via numerical experiments. Together, these three extensions to the Lagrangian framework makes it a more viable geostatistical approach in modeling realistic transport scenarios.
Date of AwardJun 14 2021
Original languageEnglish (US)
Awarding Institution
  • Computer, Electrical and Mathematical Science and Engineering
SupervisorMarc Genton (Supervisor)


  • Lagrangian framework
  • advection
  • multivariate
  • geostatistics
  • nonstationary
  • space-time

Cite this