Interannual correlations between sea surface temperature and concentration of chlorophyll pigment off Punta Eugenia, Baja California, during different remote forcing conditions

Abstract. Interannual correlation between satellite-derived sea surface temperature (SST) and surface chlorophyll a (Chl a) are examined in the coastal upwelling zone off Punta Eugenia on the west coast of the Baja California Peninsula, an area than has been identified as having intense biological productivity and oceanographic transition between midlatitude and tropical ocean conditions. We used empirical orthogonal functions (EOF) analysis separately and jointly on the two fields from 1997 through 2007, a time period dominated by different remote forcing: ENSO (El Nino–Southern Oscillation) conditions (weak, moderate and strong) and the largest intrusion of subarctic water reported in the last 50 years. Coastal upwelling index anomalies (CUI) and the multivariate ENSO index (MEI) were used to identify the influence of local (wind stress) and remote (ENSO) forcing over the interannual variability of both variables. The spatial pattern of the individual EOF1 analysis showed the greater variability of SST and Chl a offshore, their corresponding amplitude time series presented the highest peaks during the strong 1997–2000 El Nino–La Nina cycles and during the 2002–2004 period associated to the intrusion of subarctic water. The MEI is well correlated with the individual SST principal component (R a 0.67, P 0.4) mainly regulated by ENSO cycles. This was spatially revealed when we calculated the homogeneous correlations for the 1997–1999 El Nino–La Nina period and during the 2002–2004 period, the intrusion of subarctic water period. Both, SST and Chl a showed higher coupling and two distinct physical–biological responses: on average ENSO influence was observed clearly along the coast mostly in SST, while the subarctic water influence, observed offshore and in Bahia Vizcaino, mostly in Chl a. We found coastal chlorophyll blooms off Punta Eugenia during the 2002–2003 period, an enrichment pattern similar to that observed off the coast of Oregon. These chlorophyll blooms are likely linked to high wind stress anomalies during 2002, mainly at high latitudes. This observation may provide an explanation of why Punta Eugenia is one of the most important biological action centers on the Pacific coast.


Introduction
Continuous oceanographic observations carried out on the west coast of Baja California by the CalCOFI (California Cooperative Oceanic Fisheries Investigations) and IMECOCAL (Mexican Research of the California Current) programs have helped define the region of Punta Eugenia (Fig. 1) as oceanographic transitional zone (Durazo 5 and Baumgartner, 2002), where the southern part of the California Current (CC) and the North Equatorial Current (NEC) interact at a global scale. The area is also influenced by warm and dense water originating in the Gulf of California (Parés-Sierra et al., 1997), creating a complex mixing zone between coastal and oceanic flows and intense mesoscale variability characterized by a complex pattern of filaments, meanders, 10 and semi-permanent eddy structures (Gallaudet and Simpson, 1994). These structures carry nutrient-rich coastal waters to deep areas, causing important seasonal variability and inter-annual and very long-term changes (Espinosa-Carreón et al., 2004).
Seasonal wind forcing over the Punta Eugenia area is controlled regionally by the position and intensity of the North Pacific high pressure and the California semi-Introduction  SAGARPA, 2001), and extraction of salt by solar evaporation (7000 t yr −1 ; SEMARNAT, 1997) contributes to the economy of the entire peninsula. Oceanographic features off the west coast of the Baja California Peninsula are dramatically affected by global-scale ENSO and interdecadal variability. El Niño events have a negative effect on fisheries: an increase in SST, high sea level, change in composition of the zooplankton community, and microbial pollution (Strub and James, 2002;Lavaniegos et al., 2002;Boehm et al., 2004). Larval reproduction and embryonic development of spiny lobster (Panulirus interruptus) are heavily impacted because the onset and duration of larval development is accelerated or delayed, which dramatically reduces the captures in this region (Vega, 2003). Carreón-Palau et al. (2003) and Muciño-10 Días et al. (2004), describe overfishing of abalone because survival, growth, and larval recruitment are heavily dependent on cooler environmental conditions. These relationships between biological and environmental factors demonstrate strong physicalbiological coupling in this region.
Climate effects on SST and Chl a in this coastal environment are documented here 15 for more than a decade of satellite measurements (1997)(1998)(1999)(2000)(2001)(2002)(2003)(2004)(2005)(2006)(2007), period defined by Behrenfeld et al. (2006) as a permanent El Niño conditions that include the different ENSO conditions (weak, moderate and strong) beginning with an strong El Niño/La Niña cycles between 1997 and 1999, followed by weak El Niño between 2002Niño between -2004, and finally a moderate El Niño during 2006-2007. Additionally, during the 2002 the California Current System (CCS), remains in the cold phase, a state it has had since the 1999 La Niña phase with the constant permanence of the largest intrusion of subarctic water reported in the last 50 yr characterized as a cold and fresh anomaly in the upper halocline (Huyer, 2003;Goericke, et al., 2005). In this study, we explored the interannual covariance between SST and chlorophyll a 25 (Chl a) off Punta Eugenia, an adequate area as a reproductive habitat due to high levels of biological production and its responses to two different large scale processes; the ENSO cycles and the intrusion of subartic water. Individual and joint empirical orthogonal functions (EOF) analyses were used to extract the principal modes of interannual The initial resolution of 9 km × 9 km for Chl a was sampled at a 4 km × 4 km cell size to generate monthly averages with the same spatial resolution for SST, rotating and orienting the images along the coast.

15
Indices of the intensity of large-scale, wind-induced coastal upwelling (CUI) are generated by the NOAA/NMFS Pacific Fisheries Environmental Laboratory (PFEL) at 15 standard locations along the west coast of North America (Schwing et al., 1996;http://www.pfeg.noaa.gov/). We used the CUI centered on 27 • N, 116 • W as representative of the Punta Eugenia coastal region, filtering out the seasonal cycle by subtract-Introduction indices were normalized and smoothed using standard deviation and a running mean of three months to reduce small scale and short term variability. The monthly composite images of SST and Chl a, were arranged in matrices M (x, t) for each variable, where x stands for the cells of each image and t stands for each month between September 1997 and December 2007 (NT images). To remove the 5 annual and semiannual signals by subtracting a fitted (least squares) periodic function from each pixel time series, the periodic function was defined as: where A 0 is the annual mean, A 1 , w 1 , and ϕ 1 are the amplitude, frequency, and phase of the annual signal and A 2 , w 2 , and ϕ 2 are the equivalent of the semi-annual signal.
Next, we obtained the matrices of SST and Chl a anomalies (interannual variability), subtracting from each time series of each cell, its corresponding fitted periodic function (Eq. 1) as: Each matrix was transformed into normalized anomalies NA(x, t) by dividing A(x, t) se-15 ries by its standard deviation (series were re-scaled to make them comparable). The resulting NA(x, t) matrices of both variables were used in the individual EOF analysis to identify the dominant modes of SST and Chl a interannual variability and its evolution over time (principal components). Joint SST and Chl a EOF are calculated from the covariance matrix constructed from both variables to highlight how they covary with 20 each other, forcing them to have the same temporal variability (Wilson and Adamec, 2001). To isolate and spatially localize important coupled modes of variability, homogeneous correlation (Bretherton, et al., 1992)  MEI and CUI series were obtained from the NOAA Climate Prediction Center (CPC-NCEP-NOAA) website (http://www.cpc.ncep.noaa.gov) and from the Pacific Fisheries Environmental Group (PFEG) website (http://www.pfeg.noaa.gov/). We note that, though these indices summarize variability over different regions and the correlation between them are not large, wind data used to build monthly CUI anomalies in this region 5 are strongly affected by ENSO events and are not completely independent (Storch and Zwiers, 1999). Additionally, to observe the presence of the intrusion of subartic water and its relationship with the wind forcing from the west coast of North America Chl a, and CUI (space-average SST(x, t) and Chl a (x, t) and CUI monthly time series), all dominated by strong a seasonal peak, mainly affecting SST. Peaks of Chl a and CUI occurring in phase during the spring (March-Jun). The annual cycle for each of the three parameters reveals significant overlap between them. Figure 3 shows the percentage of variance explained by the first three modes for 5 the different EOF analyses (the individual EOFs and the joint EOF analysis). In the individual analysis, the first three EOFs for both parameters accounted for 92 % and 68 % of the total variance, respectively. The interannual variability of SST and Chl a off Punta Eugenia is clearly defined by the EOF 1 , which accounted for 78 % and 45 % of the variance respectively, which were statistically different because their error bars 10 do not overlap with their neighbors (North et al., 1982). The correlations between the principal components of each EOF 1 (SST and Chl a) and the MEI and CUI anomalies are presented in Table 1. The correlation between the EOF 1 of SST and MEI were high (R = 0.68, P < 0.01) and low for Chl a (R = −0.23, P < 0.01), and the correlation between the EOF 1 of SST and CUI was moderate (R = −0.38, P < 0.01) and low 15 with Chl a (R = 0.25, P < 0.01), suggesting that the interannual correlation of both variables is strongly dominated by SST and forced principally by different ENSO conditions (weak, moderate, and strong). Figure 4 shows the spatial patterns and principal components for mode 1 of the individual EOF analyses (surface plot of the EOF 1 loadings), which account for 78 % and 20 45 % of the total variance for SST and Chl a (with the sign-reversed) respectively. Both spatial patterns showed strong gradients along the coast and high loadings were observed offshore (deep region). The pattern of high SST variability (Fig. 4a), coincide with the pattern of high Chl a variability (Fig. 4b), whereas the low SST and Chl a variability is showed near to the coast. Overlaid on the spatial component maps are 25 contours of the homogeneous correlation (Bretherton et al., 1992), which is the correlation between the time series of the data at each point and the amplitude time series of the EOF 1 scores (Fig. 4c). High homogeneous correlations indicate the region that contributes the most to the temporal variability. Only correlation contours above 0.7 are OSD 10,2013 Interannual correlations between SST and Chl a H. Herrera-Cervantes et al.  (Venrick et al., 2003;Lynn, 2003, andDurazo et al., 2005;Goericke, et al., 2005), which masked the presence of the weak El Niño event of this period.
10 Figure 5 shows the spatial pattern of the first mode of the joint EOF 1 analysis for SST and Chl a (Fig. 5a, b), accounting for 80 % of the total variance. The amplitude time series corresponding to the joint EOF 1 is not shown since they are identical to those of individual EOF 1 in Fig. 4c ( Table 1). The SST and Chl a spatial pattern differ to those of the individual EOF 1 (Fig. 4a, b). The primary difference between the indi-15 vidual and joint spatial modes of the SST and Chl a is that the coastal and northern region dominates more in the joint EOF. In the joint mode low variability corresponds to the southern and depth region as a continuous feature. The high SST along the coast coincident with the area where coastal trapped waves generated during El Niño event continued its poleward propagation along the west coast of the peninsula (Durazo and 20 Baumgartner, 2002;Dever and Winant, 2002), while high Chl a can be associated with remote forcing of northern origin. The spatial distribution of the SST-Chl a correlations (Fig. 5c), present the area where both parameters strongly co-vary. This area is enclosed with the highest joint EOF 1 scores (> 2.0) of the SST (Fig. 5a). High correlations (|R| > 4) occur in a band parallel to the coast, region where intense coastal up- 25 welling and complex mesoscale variability are found. A consistent inverse relationship is observed between SST and Chl a coastal anomalies; an increasing SST anomalies is coupled to decreasing Chl a anomalies (during El Niño events), implying increases in sea level that deepen the nutricline, reduce the supply of nutrients and limits Chl a 861 OSD 10,2013 Interannual correlations between SST and Chl a H. Herrera-Cervantes et al. production. High positive Chl a anomalies associated with La Niña conditions, may elevate the nutricline, decreasing SST and increase the effects of coastal upwelling. Figure 6 shows contours of mean homogeneous correlation (HCs) calculated for three subset of the time series in both SST and Chl a respectively overlaid on (a) the spatial correlation pattern and (b) the chlorophyll mode to assess what regions domi-5 nate during the El Niño period (September 1997 to December 1998), La Niña period (September 1998to December 2000 and the intrusion of subarctic water period (January 2002to December 2003. For both SST and Chl a, a narrow area confined along the coast dominates during El Niño and La Niña: (solid and dotted contours in Fig. 6a). The combination of the mean sets of HCs distributions is virtually identical to the spatial 10 correlation and the joint EOF patterns, meaning that the physical-biological coupling in the coastal region shown in the Fig. 5, is driven by both increase and decrease of SST and chlorophyll during ENSO cycles.
In contrast, the region that dominate significantly during the intrusion of subartic water period for both SST and Chl a is showed by the mean HCs of > 0.6 (solid contours 15 in Fig. 6b) are all in the deep zone, they occur in the same region as during the entire time series (Fig. 4) and where the individual EOF does show the high variability mainly in the chlorophyll mode. These results indicate that the individual variability pattern of both parameters were driven more by the intrusion of subarctic water that by ENSO cycles, contributing little to the coupled mode. The three sets of HCs in Fig. 6 delineate 20 the regions that contribute the most to the individual and coupled mode during different remote forcing; the coastal area and deep region. Figure 7 shows the time series of the MEI index (bars) and CUI anomalies (black curve), both compared with the temporal evolution of SST and Chl a coastal anomalies plotted in two Hovmöller diagrams (i.e., time series of coastal pixels plotted as contours)  showed a pronounced decline off study area. High positives anomalies of Chl a that stalled off Punta Eugenia presented a pattern that is only compared to that bloom off Northern California (40 • N), where the wind pattern and the intense coastal upwelling processes, produce the most important Biological Action Centers (BAC) of the western coast of North America characterized by levels of pigment concentration above the 5 average for the coastal zone (Lluch-Belda et al., 2000).

Discussion
In this study, we used EOF analysis over satellite-drived SST and Chl a anomaly data off Punta Eugenia, which has allowed us a wide view of the biophysical coupling during several ENSO events and the largest intrusion of subarctic water reported during 2002-

10
2003. In addition, we related each SST and Chl a series within ∼ 40 km closest to the coast off Punta Eugenia to two independent variables, MEI and CUI and we followed the coastal signal of the Chl a related to the intrusion of subartic water (2002)(2003) from 22 • N to 45 • N. To our knowledge, this is the first time that physical-biological covariability driven by remote forcing (tropical and subartic origin) off Punta Eugenia 15 has been examined at this level of detail. The spatial patterns of individual EOF 1 ( Fig. 4a and b) showed that the interannual variability of SST and Chl a are not homogeneous and have a strong gradient that defines one area with contours of HCs values > |0.6| offshore indicate the region that contribute the most of the temporal variability different to that showed by Espinosa-  Table 1). The time series of both principal components compared with the MEI and CUI anomalies (Fig. 4c) 25 indicate that these were forced by ENSO-related events and mainly, at least during the 2002-2003 by the large intrusion of subarctic water (Durazo et al., 2005). During this OSD 10,2013 Interannual correlations between SST and Chl a H. Herrera-Cervantes et al.  (Fig. 5a, b) and correlation patterns (Fig. 5c) show that the SST-Chl a interannual covariation off Punta Eugenia is the ENSO mode, an intense variability in the coastal band of high negative correlation (∼ 40 km band closest to the coast), suggesting a high biophysical coupling in this near shore band during ENSO cycles 10 (Espinosa, et al., 2004). SST and Chl a are negatively correlated because signals generated by El Niño events in the equatorial region propagate poleward as coastal trapped waves (Dever and Winant, 2002;Jacobs et al., 1994), accompanied by positive SST anomalies, increases in sea level highs that deepen the nutricline and reduce the availability of nutrients to the euphotic zone. While the signals associated with La Niña 15 events are accompanied by positive anomalies of Chl a and negative anomalies of SST associated with an intensification of the north wind, from which the upwelling events raise the nutricline and thus increase the availability of nutrients in the coastal region of Baja California (Chavez et al., 1999;Wooster and Hollowed, 1995;Espinosa-Carreón et al., 2004). The correlation pattern (Fig. 5c), suggests that remote forcing 20 associated to ENSO event, drive a strong physical-biological covariability in a narrow coastal band. This response is demonstrated by the joint EOFs analysis applied in the SST and Chl a, suggesting that the different physical-biological responses to events as the ENSO oscillations arise from a combination of ecological and physical dynamics (Wilson and Adamec, 2001 the intrusion of subartic water period. Both spatial patterns showed similar behavior, different to that showed by Espinosa-Carreón et al. (2004), where the most of the signal of nonseasonal variability for pigment concentration is in a narrow coastal band. Neither of this patterns have very large spatial loadings near to the coast, and the HCs > 0.6 delineate the regions where the temporal trend of the data correspond the best to the 5 principal components for that mode and hence delineate the regions that contribute the most to the mode, in this case the individual variability mode is defined as the subarctic water mode. In contrast, the first co-variability mode is the ENSO mode; its temporal components (not shown) are identical to those of the individual modes (Fig. 4) and theirs spatial 10 patterns for SST and Chl a (Fig. 5a, b) are virtually identical to those of the spatial distribution of the SST-Chl a correlations (Fig. 5c), mainly with the SST having spatial correlation of −0.98, the joint spatial pattern for Chl a differ little by having high loadings more evenly centered northern of Punta Eugenia, including Bahí a Vizcaíno. In a similar fashion we overlaid on the spatial correlation pattern the mean HCs calculated for three 15 subsets of the time series of SST and Chl a corresponding to the periods El Niño, La Niña and the intrusion of subartic water (Fig. 6). There are not significant changes between the mean HCs distribution during the El Niño and La Niña time period (solid and dotted contours), the region dominated for both events is a narrow coastal band that coincide to those of the spatial distribution of the high correlations (|R| > 4).

20
The HCs distribution calculated for both SST and Chl a during the intrusion of subartic water delimited a different area. The mean HCs for both variables dominate a broad area centered in the western basin where the depth is ∼ 1000 m off the coast (Fig. 6c), region that contribute significantly to the SST and Chl a individual mode when analyzed over the entire time series. During the intrusion subartic water, there are mean HCs of 25 > 0.6 that contribute little to the joint mode and overall correlation pattern. The coastal zone in the joint modes (Fig. 5a, b), unlike the individual mode, contribute significantly to the mode and coincide to those of high HCs during El Niño and La Niña time period, OSD 10,2013 Interannual correlations between SST and Chl a H. Herrera-Cervantes et al. which reflects the stronger representation of the coastal zone in the ENSO mode and in the interannual covariation of SST and Chl a. The SST and Chl a anomaly evolutive diagrams (Fig. 7, middle and lower panels) showed that the signals associated to weak 2002-2004 El Niño was masked in surface by the intrusion of subartic water. Both, MEI and CUI anomalies do not showed rela-5 tionship with the intrusion of subartic water during this period. Only during the 1997-2000 and 2005-2006 periods, positives Chl a coastal anomalies are in agreement with positives CUI anomalies. Espinosa-Carreón et al. (2004) suggested that on an interannual timescale, changes in the monthly CUI anomalies do not appear to be the primary source of variability in the oceanic parameters like that SST and Chl a, since removing the seasonal component results in low correlation values (see Table 1). The chlorophyll bloom of 2002-2003 was greater than during the strong 1999-2000 La Niña, evidencing that the distribution of biological groups in areas as far south as 28 • N are influenced mostly by the intrusion of water of northern origin than by events of both local and equatorial origin (upwelling and ENSO event). Durazo et al., 2005 observed 15 the presence of zooplankton groups (chaetognaths) at depths of 100 m associated with water of tropical origin that accompanies El Niño events, but also zooplankton groups (salps) associated with an intrusion of subarctic water.  positives Chl a anomalies occurs again during summer 2003 coincident with the peaks of amplitude time series of the Chl a EOF 1 scores (Fig. 4c).
Our study supports the notion that changes in Chl a concentration could be due to other factors such as decadal variability, changes in grazing pressures on the phytoplankton (Lavaniegos et al., 2006;Wilson and Adamec, 2001) that may cause impact 5 on biological communities located along the coast. Gaxiola et al. (2008), using data obtained during IMECOCAL cruises (2001)(2002)(2003)(2004)(2005)(2006)(2007), observed that, the intrusion of subartic water represented by an anomalous low salinity condition in the southern sector of the California current appears to be coupled with the 2002-2006 warm phase of the Pacific Decadal Oscillation Index (PDO). Neither the local upwelling (represented 10 by the CUI anomalies) nor the zonal Ekman drift velocity used as a proxy for coastal upwelling (Gaxiola et al., 2008) showed signals associated with the presence of the intrusion of subartic water. In this case, the positives signals of wind stress developed along the coast, were agreement with the bloom of the Chl a.

15
The temporal and spatial resolution of SST and Chl a satellite data, allowed for observation both separately and jointly on the two fields of the physical-biological coupling during large scale processes that affected the CC region off Punta Eugenia, dynamic region of eddy variability, intense biological productivity and oceanographic transition highly influenced by equatorward flows (e.g. anomalous intrusion of subarctic water) 20 and by subtropical signatures triggered by poleward flow (e.g. ENSO cycles). The individual SST and Chl a interannual variability were forced mainly by an unusual enhanced onshore transport of subarctic water from the offshore CC during the of equatorial origin (ENSO events) tightly observed in the alongshore band of ∼ 40 km wide. Although, ENSO events dominate along the time period, the presence of the intrusion of subarctic water off Punta Eugenia as a principal remote forcing dominate the individual interannual variability of both variables. These individual patterns present a 5 similar area were this event dominated, an offshore region where the mean HCs calculated for the intrusion of subarctic water time period in both parameters, where in agreement with the high EOF 1 loadings. This remote forcing combined with the positives coastal wind stress signals, results in a large-scale chlorophyll bloom that extend along more than 3000 km of the Northeast Pacific coast between Oregon and Baja Unlike the data gathered during oceanographic cruises, the data used in this study came from 4 km × 4 km quadrants, and may properly represent the very near shore 15 environments. The 2002-2004 El Niño was masked by the presence of subarctic water, reversed biological-physical response in the surface mainly north of Punta Eugenia. This condition resulted in an atypical situation, such as abnormal cooling of sea surface together with significant high pigment concentration occurring during a warm ENSO phase (Hereu et al., 2003). Furthermore, this abnormal situation was not associ-20 ated with upwelling-favorable winds (Durazo, et al., 2005) since monthly CUI anomalies showed in Fig. 4c and zonal Ekman drift velocity (Gaxiola, et al., 2008) were negatives during this period. This results successfully complements the reveled previously by hydrographic surveys (CaLCOFI and IMECOCAL programs) off Baja California. Introduction

Tables Figures
Back Close

Full Screen / Esc
Printer-friendly Version
OSD 10,2013 Interannual correlations between SST and Chl a H. Herrera-Cervantes et al.