BayesOpt-Antarctica

In this repository Bayesian Optimisation over the Antarctica ice sheet.

Contents

Open the html version to avoid having to run code locally to see results and visualisations.

2DBayesianOptimisation.ipynb
2DBayesianOptimisation.html

Simple Bayesian Optimisation loop over modelled RACMO2.3p3 ANT27 surface mass balance (smb) data. Input variables are restricted to x1 and x2 coordinates, projected using EPSG:3031 WGS 84 / Antarctic Polar Stereographic projection. Implementation using BoTorch package.

Data

surface mass balance (modelled with regional climate model)

name	description
model version	RACMO2.3p3
file name	smb_monthlyS_ANT27_CONsettings_197901_201812.nc
file source	Zenodo
file format	.nc NetCDF (Network Common Data Form)
key model publication	van Dalum et al. 2022
spatial resolution	27 km gridded
projection	rotated pole
spatial coverage	Antarctica
temporal resolution	aggregated (but monthy available)
temporal coverage	1979 - 2019 available
target variable	SMB (surface mass balance)
uncertainty estimates	-
unit	mm w.e. a-1 (millimeter water equivalent per year)
	equivalent to kg m-2

Contact

Kim Bente Data Science PhD student, School of Computer Science, University of Sydney
kim.bente@sydney.edu.au
www.kimbente.com