30 years of change in water-limited ecosystems

Abstract:

Water availability limits plant growth and production in most terrestrial ecosystems. However these ecosystems do not show the same sensitivity to changes in precipitation. Water-limited ecosystems are defined here as ecosystems where rainfall is the dominant climate constraint to plant growth. Drought-prone and often characterized by increasing human pressure on land and natural resources, these regions are amongst the most vulnerable on Earth. Despite the many years of research, a clear understanding of changes in vegetation dynamics and species distribution, as well as related drivers, has not been reached yet.

In this research we take advantage of the 30years time span offered by the GIMMS FAPAR3g dataset to investigate abrupt and gradual changes in Rain-Use Efficiency (RUE). Using the piece-wise regression method implemented in BFAST (Breaks For Additive Season and Trend) and BFAST01, a change type classification scheme is produced for water-limited ecosystems. Compared to classical non-parametric trend analysis, this approach allows detecting trend shifts during the study period.

This global scale analysis revealed that for more than 50% of the cases no significant changes in RUE were registered between 1982 and 2011. Whereas when significant changes were registered, monotonic increase was the predominant type of changes. Large patches of reversing trends were also observed, notably in Asia (China, Kazakhstan), in the Sahelian region (Sudan, Ethiopia, Senegal), and in South America (Peru and Argentina). Even though reversing trends appear to be more frequently observed in regions with high land cover change dynamics, a comprehensive attribution of drivers for all recorded changes is still under discussion. Indeed the co-occurrence of global drivers (such as change in climate and in extreme events) and local drivers (such as land-cover changes) makes it a very delicate task.