"Regime shifts or red noise?"
Understanding decadal-scale variability of the North Pacific marine ecosystem has been an important issue because some strong environmental changes have occurred at this time scale. Studies of many physical and biological time series lead to a speculation that a regime or ocean condition may persist for 2 to 3 decades and then undergo a rapid shift to another state. Many nonlinear mechanisms exist that can cause a system to switch abruptly among multiple stable states. However, zero-mean Gausian red-noise time series have long runs without a zero crossing might be easily mistaken as regimes separated by abrupt shifts. The notion of regime shifts has had widespread influence on oceanographic researches although the debate continues between “regime shifts” and “red noise” schools. Therefore, there is a need to distinguish regime shifts from red noise in oceanographic time series with a quantitative method. One possible way to detect whether or not a “regime” signal exists as a product of an underlying dynamic instability is to see if the physical and biological time series contain a nonlinear signature. Uncovering a nonlinear signature in the time series is a “necessary” condition. The idea behind determining nonlinearity of a time series is to examine whether or not there is a significant improvement in out-of-sample forecast performance with an equivalent nonlinear versus a linear forecast model. In this research we applied nonlinear time series analyses on physical and biological time series in the North Pacific. Physical data include the Pacific Decadal Oscillation Index, North Pacific Index, and Scripps pier sea-surface temperature; biological data include the ichthyoplankton time series collected in the California Cooperative Oceanic Fisheries Investigations surveys, the diatom time series collected under the Scripps pier, and fish catch and recruitment data in the Northeast Pacific. We found that the physical time series cannot be distinguished from red noise. This is in contrast to the biological time series that show nonlinearity. We suggest that the biological response to external linear physical forcing and/or internal species interaction is nonlinear. This implies that biota do not simply track physical signals. To achieve better ecosystem management, an understanding of underlying dynamic instability is necessary.