Detection of breakpoints in global NDVI time series

Abstract

Continuous global time series of vegetation indices, which are available since early 1980s, are of great value to detect changes in vegetation status at large spatial scales. Most change detection methods, however, assume a fixed change trajectory – defined by the start and end of the time series – and a linear or monotonic trend. Here, we apply a change detection method which detects abrupt changes within the time series. This Breaks For Additive Season and Trend (BFAST) approach showed that large parts of the world are subjected to trend changes. The timing of the breakpoints could in some cases be related to satellite changes, but also to large-scale natural influences like the Mt. Pinatubo eruption. Shifts from greening to browning (or vice versa) occurred in 15% of the global land surface, which demonstrates the importance of accounting for trend breaks when analyzing long-term NDVI time series.