About me

My name is Jan Swierczek-Jereczek (insert joke about unpronouciable name). I am a PhD candidate at the Complutense University of Madrid, funded by a Marie Sklodowska-Curie Action. I mostly study past and future abrupt changes of the Antarctic Ice Sheet and therefore focus largely on climate, glaciology and glacial isostatic adjustment (GIA). Furthermore, I am interested in applying concepts of nonlinear dynamics, control theory and machine learning to Earth system modelling and analysis.

I particularly enjoy programming and therefore develop open-source softwares around the aforementioned topics. I put a particular emphasis on making them not only theoretically but also pratcially accessible. By that, I mean that an open-source licence is not enough and that thorough documentation and publication of the code is key for a code base to be useable by other researchers.

FastIsostasy.jl

The main undertaking of my PhD consists in the development of FastIsostasy.jl, a regional GIA model that captures the solid-Earth response at extremely low computational cost, even for laterally-variable Earth structures. For instance, simulating the Antarctic GIA response to the last glacial cycle takes less than 10 minutes on a low-end 2022 GPU despite using a relatively high resolution of 20 km. Compared to a computationally much more expensive 3D GIA model, the error in displacement and sea level is of less than 5% on average and 14% at most. With this tool, performing ensemble simulations that sample the uncertainties of the solid-Earth properties becomes easily feasible.

The video below depicts the bedrock displacement computed by FastIsostasy when provided with the ice loading history from ICE6G_D (main panel), as well as the perturbation of sea-surface elevation, the spatial pattern of ice thickness and the timeseries of integrated ice thickness (upper panels).

As of today, Julia code cannot be statically compiled to binaries, which can be a great obstacle to coupling it to ice-sheet models. To tackle this, Alex Robinson and I programmed a Fortran version of the code, which also includes simpler regional GIA models.

TransitionsInTimeseries.jl

In collabroation with George Datseris, I am developping TransitionsInTimeseries.jl, a Julia package that allows you to easily run a statistical analysis of time series to detect a transition or a loss of resilience of the underlying system.