Estimation of Ocean Flow from Satellite Gravity Data and Contributions to Correlation Analysis

Abstract

This thesis, structured in two parts, addresses a series of problems of relevance in the field of Spatial Geodesy. The first part delves into the application of satellite gravity data to enhance our understanding of water transport dynamics. Here, we present two significant contributions. Both are based on satellite gravity data but stem from different mission concepts with distinct objectives: time-variable gravity monitoring and high-resolution, accurate static geoid modelling. First, the fundamental notions about gravity are introduced and a brief summary is made of the different gravity satellite missions throughout history, with emphasis on the GRACE/GRACE-FO and GOCE missions, whose data are the basis of this work. The first application focuses on estimating water transport and geostrophic circulation in the Southern Ocean by leveraging a GOCE geoid and altimetry data. The Volume Transport across the Antartic Circumpolar Current is analyzed and the resulsts are validated validated using the in-situ data collected during the multiple campaigns in the DP. The second application uses time-variable gravity data from the GRACE and GRACE-FO missions to estimate the water cycle in the Mediterranean and Black Sea system, a critical region for regional climate and global ocean circulation. The analysis delves into the analysis of the different components of the hydrological cycle within this region, including the water flow across the Gibraltar Strait, examining their seasonal variations, climatic patterns, and their connection with the North Atlantic Oscillation Index. The second part of the thesis is more focused on data analysis, with the objective of developing mathematical methods to estimate the cross correlation function between two time series that are both unevenly spaced spaced (the sampling is not uniform over time) and observed at unequal time scales (the set of time points for the first series is not identical to the set of time points of the second series). Such time series are frequently encountered in geodetic surveys, especially when combining data from different sources. The estimation of the the cross correlation function for these time series presents unique challenges and requires the adaptation of traditional analysis methods designed for evenly spaced and synchronized time series. The two main contributions in this context are: (i) the study of the asymptotic properties of the Guassian Kernel estimator, that is the recommended estimator for the cross correlation function when the two time series are observed at unequal time scales; (ii) an extension of the stationary bootstrap that allows to construct bootstrap-based confidence intervals for the cross correlation function for unevenly spaced time series not sampled on identical time points.