Decomposition and Inference of Sources through Spatiotemporal Analysis
of Network Signals: The DISSTANS Python Package
Abstract
Regional-scale, continuously-operating Global Navigation Satellite
System (GNSS) networks are a powerful tool to monitor plate motion and
surface deformation. Since their inception, their size, density, and
length of observation record have steadily increased throughout the
world. Simultaneously, researchers have had to write accompanying
software to enable the analysis (especially the decomposition) of the
ever-increasing amount of available timeseries in an efficient way.
These codes and respective studies have individually set standards for
different subsets of the following desirable qualities: portability
(between locations), speed (code runtime), automation (avoiding or
simplifying manual inspection of each station), use of spatial
correlation (exploiting the fact that stations experience common
signals), availability (open source), and documentation (of the usage
and underlying methods). In this study, we present the DISSTANS Python
package, which aims to combine the aforementioned achievements in a
single software by offering generic (therefore portable), parallelizable
(fast) methods that can exploit the spatial structure of the observation
records in a user-assisted, semi-automated framework, including
uncertainty propagation. The code is open source, includes an
application interface documentation as well as usage tutorials, is
easily extendable, and is based on the previously published and
validated method of Riel et al. (2014). We also present two case studies
to validate our code, one using a synthetic dataset and one using real
GNSS network timeseries.