A closer look at regime shifts based on coastal observations along the eastern boundary of the North Pacific

• Published in: Continental shelf research Vol. 27; no. 17; pp. 2250 - 2277
• Main Author: Breaker, Laurence C.
• Format: Journal Article
• Language: English
• Published: Elsevier Ltd 01.10.2007
• Subjects: 1976–1977 regime shift; Change points; Cumulative sums (CUSUMs); Detecting regime shifts; Pacific Grove; Pattern recognition; Regime shifts; Salinity; Scripps Pier; Sea level; Sea surface temperature; Statistical significance; Sustained changes; Vancouver Island; INE, USA, California; INE, Canada, British Columbia, Vancouver I; INE, USA, California, Pacific Grove; IN, North Pacific; Scripps Pier; Sea surface temperature; Pattern recognition; Sea level; Change points; Cumulative sums (CUSUMs); Salinity; Detecting regime shifts; Pacific Grove; Vancouver Island; Statistical significance; 1976–1977 regime shift; Regime shifts; Sustained changes
• ISSN: 0278-4343; 1873-6955
• DOI: 10.1016/j.csr.2007.05.018

At least six regime shifts have been reported in the North Pacific since 1920. They occurred in 1925, 1939, 1946, 1976–1977, 1989 and 1999. The major change in 1976–1977 corresponds to a regime shift that is now widely accepted as a canonical event since it had a significant impact on virtually all climatic and ecosystem indicators. We seek to determine if daily sea surface temperature (SST) from Pacific Grove, in central California, and Scripps Pier, in southern California, and coastal observations from several other locations along the west coast of North America can be used to detect and resolve these events. Cumulative sums (CUSUMs) were initially calculated to enhance the detection process. The CUSUM trajectories during the 1976–1977 event at Pacific Grove and Scripps Pier were distinctive, highly correlated, and in phase. The turning point patterns from this event were then used to search for other events that have been reported since 1920. Turning point patterns very similar to the 1976–1977 event were detected in 1946 and 1989. The events in 1925 and 1939 were generally similar, but the CUSUM patterns for the event in 1999 departed significantly from the other events. Further examination of the 1976 and 1989 events revealed inflection points in the CUSUMs near the beginning and end of each transition that correspond to critical values or extrema in the original data. The inflection points and/or critical values provide an improved basis for determining the duration of these events. The robustness of the CUSUM approach for detecting regime shifts was examined by posing the inverse problem to determine if other possible regime changes could be detected that have not been previously reported. The period between 1946 and 1976 was examined, and one match in 1972 was found, which coincided with a large shift in the Aleutian Low Pressure Index. The CUSUM patterns associated with well-defined regime shifts may be essentially unique and thus useful in searching for other events. Whether the temperature ultimately increases or decreases following a regime shift is well-predicted by the sign of the CUSUM slope during an event. Testing regime shifts for statistical significance may be problematic, but our results suggest that when CUSUMs are employed, the detection problem becomes one that is more closely related to pattern recognition where other tests could be applied. CUSUMs often produce a distinct pattern that appears to be characteristic of regime shifts. During well-defined events such as those that occurred in 1946, 1976, and 1989, the CUSUM trajectories from Pacific Grove and Scripps Pier were highly synchronized and nearly identical in form. The CUSUM transformation allows us to identify, localize and observe how these events evolve. We have only been able to examine these events in such detail because daily observations from single locations were used. Based on the events we have examined, they have time scales that range from about 4–9 months. Salinity and sea level data were also employed in this study and were found to be less sensitive to the changes associated with regime shifts than SST. Regime shifts detected in CUSUMs of SST at two locations off Vancouver Island were found to be weaker in amplitude and less well-defined than those detected at Pacific Grove and Scripps Pier. However, they were in phase with the events observed further south. Establishing the connection between these results, and changes in the ecosystems of the North Pacific, should be given a high priority. Finally, the results of this study are related to decadal climate variability and provide additional insight into the nature of this phenomenon.