Monitoring a Sentinel Species from Satellites: Detecting Emiliania huxleyi in 25 Years of AVHRR Imagery - Satellite-based Applications on Climate Change - page 273

SST (Reynolds and Smith 1994 ). Mixed-layer depths were acquired from Simple
Ocean Data Assimilation Reanalysis (SODA) Version 2.0.2 (Carton and Giese
2008 ). This blended ocean product combines the advantages of ground-truth
observations with the uniformly mapped regularity of numerical simulation.
The annual bloom extent was also compared to several climate indices which
have been linked to ecosystem changes: El Ni ˜ o Southern Oscillation (ENSO)
(Wolter and Timlin 1998 ), Pacific Decadal Oscillation (PDO) (Trenberth and
Hurrell 1994 ), and North Atlantic Oscillation (NAO) (Hurrell 1995 ). The trend
toward more positive NAO values over the past 30 years coincides with changes in
marine and terrestrial ecosystems (Hurrell and Deser 2009 ). Over short time scales,
warm PDO eras have been linked to enhanced coastal ocean productivity near
Alaska and inhibited productivity to the south, with the opposite patterns during
cold eras (Mantua et al. 1997 ).
18.3 Results and Discussion
18.3.1 Extent of E. huxleyi Blooms
The timing of E . huxleyi blooms detected in AVHRR showed good correspondence
to those identified in SeaWiFS (Fig. 18.3 ). Some discrepancies in magnitudes
were evident, but were expected due to differences in sensor sensitivity and
image coverage between the two satellites. During 10 years of overlapping data,
the Bering Sea had the greatest fraction of bloom coverage, followed by the
Norwegian Sea and the area south of Iceland. Our AVHRR bloom detection method
did not match SeaWiFS as well on the Patagonian Shelf; there was some disparity in
magnitudes, and many times SeaWiFS detected a bloom but AVHRR did not.
In general, using AVHRR, E . huxleyi blooms were detected more frequently in
Bering and Norwegian Seas. Using SeaWiFS, blooms were detected more fre-
quently in the Patagonian Shelf and south of Iceland.
For the entire AVHRR series (Fig. 18.4 ), some years show greater bloom
coverage over a week or two, while some smaller fractional bloom areas are
maintained over several weeks. Normalized annual cumulative bloom coverage
values account for this effect. E . huxleyi blooms in all regions achieved their annual
cumulative maxima around 1992 and 1996 (Fig. 18.4 ). Since then, annual values
clearly show a trend toward less bloom area. Blooms in all regions increased
slightly in 2006 after decreasing over the last decade. A linear fit to the annual
values shows a marked decrease in all regions; this trend is highly statistically
significant for the Northern Hemisphere regions and to 94% confidence for the
Patagonian Shelf (Table 18.1 ). E . huxleyi blooms were identified in the Bering Sea
in 1997 with the advent of SeaWiFS ocean color, although they had been present
prior to that as evidenced in AVHRR as well as in situ data collected during field
studies in the early 1990s (Merico et al. 2003 ). The Patagonian Shelf had the
smallest area covered by E . huxleyi blooms which were drastically reduced since
2000, suggesting a possible regime shift. For all regions, there was no evidence for
any poleward shift by the species.