"There has been a flurry of recent research on statistical indicators of impending ecological regime shifts, e.g. rising variance, changing skewness, rising autoregression," researcher Oonsie Biggs, who is now at Stockholm University, told environmentalresearchweb. "While it has been shown theoretically that a number of these new statistical indicators will give advance warning of a regime shift, no one has yet investigated whether such warning could theoretically come early enough to allow society to take action to avert an undesirable regime shift."

Biggs and colleagues at the University of Wisconsin, Madison, modelled a fisheries ecosystem consisting of a piscivorous (fish-eating) fish and its planktivorous prey fish, which also preys on juveniles of the piscivorous species. The system had two possible regimes, one dominated by piscivores and the other dominated by planktivores. The two key anthropogenic factors acting to cause a regime shift were angling and shoreline development. Shoreline development affects the amount of refuge habitat, such as fallen trees in shallow water, where juveniles can escape predation.

The team found that the speed at which the manmade factor could be adjusted made a huge difference to the amount of early warning needed to avert a regime shift. For example, for a regime change caused by angling, which it was easy to stop quickly, the system could be 10 years into a regime shift, but cutting harvesting could still prevent a permanent change. But, for a shoreline development-driven shift, habitat restoration would need to be started at least 45 years before the onset of the shift in order to avert it.

"The current set of indicators may only be useful in detecting and averting regime shifts driven by variables that can be rapidly manipulated," said Biggs. "Our work points to a need for further research into more sensitive indicators that may provide greater advance warnings of regime shifts, which could then potentially be used to detect and avert regime shifts driven by variables that can only be gradually manipulated through management action."

Now the team at Wisconsin is testing whether such indicators can be observed in practical field settings. "Other groups, such as the group at Wageningen University, are looking into the possibility of using space as a substitute for time, which may allow the indicators to be calculated from much shorter data time series," said Biggs.

The researchers reported their work in PNAS.