Abstract
Dryland ecosystems are among the most vulnerable to anthropogenic climate change and
are under threat of desertification. Understanding the roles of climate and land use change
in driving desertification is essential for effective policy responses but remains poorly
quantified. This thesis focuses on improving the detection and attribution of dryland
vegetation change. Major contributions and key findings include:
1) Developing a new methodology to distinguish changes in dryland vegetation driven
by climate from both the gradual and abrupt structural ecosystem changes introduced
by land use. This novel methodology, termed Time Series Segmentation and Residual
Trend analysis (TSS-RESTREND), overcomes some of the limitations present in
existing techniques and was able to accurately capture both the timing and
directionality of ecosystem change in two regions with known histories of
degradation. TSS-RESTREND was developed into a publicly available open source R
package available through CRAN.
2) Performing the first comprehensive study to assess the impact of dataset selection on
land degradation detection. The analysis identified a large and previously
undocumented error in a widely used remotely sensed vegetation dataset. This error
has impacted many of the currently published estimates of dryland degradation. This
thesis outlines a multi-dataset ensemble approach to minimise (and quantify) the
impact of dataset errors and produce more robust estimates of dryland degradation.
3) Finally, this thesis describes the first observation-based attribution study of
desertification that accounts for climate, CO2 fertilisation and ecosystem break points
introduced by land use changes. Between 1982 and 2015, 6% of the world’s drylands
have undergone desertification driven by unsustainable land use practices
compounded by anthropogenic climate change. Despite a greening effect,
anthropogenic climate change has driven 5.28 million km2 of drylands towards
desertification, which affected 209 million people, 85% of whom live in developing
economies. In addition to the drylands experiencing desertification, there are 516
million people living in areas where land use had a desertifying effect that has been
offset by a positive anthropogenic climate change signal.
This work provides a significant step towards a robust and generally applicable method
to identify and quantify areas undergoing desertification globally.