Abstract
The impacts of global change on ecological systems are multiple and difficult to explain. Consequently, linear descriptions of these changes dominate studies in ecology and conservation. Ignoring non-linear temporal dynamics to describe and characterise the fate of biological systems is therefore a weakness in conservation strategies. This is particularly true in the case of abrupt shifts, manifested by rapid, large-scale changes, which may indicate profound upheavals that are difficult to reverse. In this thesis, I sought to systematise the identification of abrupt shifts in time series by developing an approach that would cover a wide range of trajectories relevant to ecological data. Using simulated data, I assessed the performance and conditions of application of this classification, and then applied it to globally distributed time series of fish stock productivity. This approach provided a picture of the prevalence of abrupt shifts relative to other trajectories, at global and regional scales and over time. The results show that abrupt shifts, both negative and positive, are far from negligible, occurring in almost a quarter of the stocks considered. I have also shown that abrupt declines in productivity were more numerous in regions that experienced the most rapid increases in ocean surface temperature. Furthermore, only some of the steep declines were followed by stock collapses. Finally, I investigated the question of the synchrony of abrupt shifts between stocks by revealing simultaneous declines in productivity in time series over a large part of the Northern Hemisphere during the decade 1980-1990. The results of this thesis work underline the need to consider all the dimensions of temporal dynamics, and in particular its non-linear components, in order to monitor and assess the fate of living organisms exposed to the contemporary regime of exploitation and destruction.