Python Machine Learning Conference & GeoPython 2020

»TESLA-kit: An open-source python-based library for coastal risk assessment in a changing climate«
2020-09-22, 10:30–11:00, Room 2

TESLA-kit (https://github.com/teslakit/teslakit) is an unique open-source software (suite of libraries) combining tools in Python and exemplified projects in the Jupyter Notebooking ecosystem, for coastal risk assessment studies.

The evident increase in the frequency and intensity of natural hazards hitting our coasts has revealed the necessity of: (a) improving the ability of modeling coastal flooding and erosion risks; and (b) understanding and modeling the climate variability and trends of coastal dynamics. This will contribute to a better quantification and to a time-varying planning of coastal adaptation measures, therefore contributing to an increase - or at least not reduction - of the resilience of the coast. To solve this complex problem, a multi-disciplinary approach is required involving cutting-edge science of oceanographers, geologists, engineers, statisticians, mathematicians, climatologists, hydrologists, biologists, social scientists, economists, computer scientists… This is the way scientists are moving forward, combining tools (hydrodynamic and morphodynamic models, statistical models, machine learning techniques, damage models, data-driven models, optimization models), data bases (met-ocean data sets, field surveys, satellite data, topo-bathymetry, socio-economical data bases, climate change projections data sets …) and methodologies (set of modules interconnected for solving complex problems). However, data, tools and methodologies are usually non-homogeneous in terms of format files, connectivity between modules, access to the cloud,… Linking different modules for solving complex problems is itself a challenge, and many times requires a lot of resources to smooth the work flow of a particular methodology. Besides, many research projects are overlapped, resulting in a dramatic loss of efficiency and resources. Therefore, a coordinated effort to homogenize and facilitate the use of tools, the connectivity between modules and the easy access to data sets is needed.

TESLA-kit (https://github.com/teslakit/teslakit) aims to be an unique open-source software (suite of libraries / one-stop-shop), combining tools in Python and exemplified projects in the Jupyter Notebooking ecosystem, being available to all researchers and stake-holders, as the Jupyter environment is becoming the standard for doing and exchanging research. TESLA-kit is composed by different modules interconnected. These modules includes: (a) climate-based statistical downscaling techniques, non-linear data mining, multivariate extreme value models for defining the hydraulic boundary conditions; (b) functions to run hydrodynamic models (i.e. SWAN, X-Beach); and (c) access to long-term data bases (observational, hindcast and climate change projections) of atmospheric (sea level pressure fields) and oceanographic variables (sea state parameters, astronomical tides and non-tidal residuals); (d) hybrid – metamodels, surrogate models- methods for downscaling large amount of multivariate oceanographic conditions to nearshore areas.