Detection of Changepoints in Ocean Time Series

The aim of this thesis is to analyze climate time series for abrupt changes.
The time series are based on data provided by simulations of the Earth
System Model GFDL ESM2M and consist of a historical period and
a Representative Concentration Pathway (RCP8.5) period, where high
emissions of greenhouse gases are assumed. The annual means of the
four ocean variables sea surface temperature, sea ice concentration, net
primary production of phytoplankton, and surface partial CO2 pressure
are analyzed for abrupt changes. The analysis is based on the R package
EnvCpt, which is designed to detect structural changes in climate and
environment time series. Using post-processing R scripts, the EnvCpt
output is filtered and presented in three different spatial plots. Several regions
showing abrupt changes either caused by long-term shifts throughout
the historical period or caused by changes due to the RCP8.5 period
could be identified. The results are compared to the ones obtained by
Drijfhout et al. (2015), who also analyzed climate time series for abrupt
changes. With toy data, some limitations of EnvCpt are recognized and
some recommendations for the use of EnvCpt are given.