Global Change Biology Global Change Biology (2014) 20, 2117–2123, doi: 10.1111/gcb.12457
From silk to satellite: half a century of ocean colour anomalies in the Northeast Atlantic DIONYSIOS E. RAITSOS* , YASWANT PRADHAN†, SAMANTHA J. LAVENDER‡§, I B R A H I M H O T E I T ¶ , A B I G A I L M C Q U A T T E R S - G O L L O P k, P H I L L I P C . R E I D § k* * and ANTHONY J. RICHARDSON††‡‡ *Plymouth Marine Laboratory (PML), Prospect Place, The Hoe, Plymouth PL1 3DH, UK, †Met Office, FitzRoy Road, Exeter EX1 3PB, UK, ‡Pixalytics Ltd, 1 Davy Road, Tamar Science Park, Derriford, Plymouth Devon PL6 8BX, UK, §School of Marine Science & Engineering/Marine Institute, University of Plymouth, Drake Circus, Plymouth PL4 8AA, UK, ¶King Abdullah University for Science and Technology, Thuwal, Saudi Arabia, kSir Alister Hardy Foundation for Ocean Science (SAHFOS), The Laboratory, Citadel Hill, Plymouth PL1 2PB, UK, **Marine Biological Association of the UK, The Laboratory, Citadel Hill, Plymouth PL1 2PB, UK, ††Centre for Applications in Natural Resource Mathematics (CARM), School of Mathematics and Physics, University of Queensland, St Lucia Queensland 4072, Australia, ‡‡Climate Adaptation Flagship, CSIRO Marine and Atmospheric Research, Ecosciences Precinct, GPO Box 2583, Dutton Park, Queensland 4102, Australia
Abstract Changes in phytoplankton dynamics influence marine biogeochemical cycles, climate processes, and food webs, with substantial social and economic consequences. Large-scale estimation of phytoplankton biomass was possible via ocean colour measurements from two remote sensing satellites – the Coastal Zone Colour Scanner (CZCS, 1979–1986) and the Sea-viewing Wide Field-of-view Sensor (SeaWiFS, 1998–2010). Due to the large gap between the two satellite eras and differences in sensor characteristics, comparison of the absolute values retrieved from the two instruments remains challenging. Using a unique in situ ocean colour dataset that spans more than half a century, the two satellite-derived chlorophyll-a (Chl-a) eras are linked to assess concurrent changes in phytoplankton variability and bloom timing over the Northeast Atlantic Ocean and North Sea. Results from this unique re-analysis reflect a clear increasing pattern of Chl-a, a merging of the two seasonal phytoplankton blooms producing a longer growing season and higher seasonal biomass, since the mid-1980s. The broader climate plays a key role in Chl-a variability as the ocean colour anomalies parallel the oscillations of the Northern Hemisphere Temperature (NHT) since 1948. Keywords: chlorophyll, Northeast Atlantic, northern hemisphere temperature, ocean colour, phytoplankton variability Received 13 May 2013 and accepted 22 October 2013
Introduction Phytoplankton that form the base of the marine food web and produce ca. 50% of global primary production, are sensitive to climate variability and environmental change (Beaugrand, 2005; Hays et al., 2005). It is thus important to monitor phytoplankton biomass through time, although this has been challenging considering the sparse long-term information. Even for the available datasets, there is uncertainty on the trends of phytoplankton biomass, with studies reporting contradictory findings (summarized by Chavez et al., 2011). For example, Boyce et al. (2010) report a decline in phytoplankton biomass in the North Atlantic based on ocean transparency measurements and in situ chlorophyll data. By contrast, Wernand et al. (2013), McQuattersGollop (2011) and Antoine et al. (2005) have shown an increase in phytoplankton biomass in the North Correspondence: Dionysios E. Raitsos, tel. +44 01752 633406, fax +44 01752 633 101, e-mail: [email protected]
© 2013 John Wiley & Sons Ltd
Atlantic. This uncertainty in the historical changes of phytoplankton makes it difficult to predict changes in the future as the climate continues to warm. As phytoplankton ultimately supports most of the world’s fisheries, uncertainties in its historical and future response to climate change hamper policy decisions [IPCC (Intergovernmental Panel on Climate Change), 2007]. One way to increase our knowledge of change in phytoplankton biomass is to re-analyse the few available long-term in situ datasets in conjunction with data from satellite remote sensing. In the Northeast (NE) Atlantic Ocean and North Sea, ocean colour is estimated via two distinct methods (over large spatial and temporal scales): through satellite remote sensing and with the Continuous Plankton Recorder (CPR). Platt & Stuart (1997) reported that ‘until the advent of satellite remote sensing, the CPR provided the only means of collecting plankton data at large spatial scales: it was the remote sensing of the day’. In the late 1970s, the CZCS ocean colour satellite offered the first global view of phytoplankton biomass 2117
2118 D . E . R A I T S O S et al. patterns. The CZCS mission ended in the mid-1980s and no other satellite with ocean colour measurement capabilities was launched until the mid-1990s (McClain, 2009), leaving a nearly 12-year gap. Due to this gap and methodological differences compared with modern sensors, CZCS and contemporary satellite datasets cannot be directly compared to assess decadal changes (Antoine et al., 2005). The Phytoplankton Colour Index (PCI) from the CPR dataset has been collected consistently since 1948 (Richardson et al., 2006), and covers the period before and during the satellite eras. These data could serve as a link between the two sensor eras, as its validity compared with satellite derived Chl-a has been demonstrated previously (Batten et al., 2003; Raitsos et al., 2005; McQuatters-Gollop et al., 2007; Henson et al., 2009a). Here, we aim to bridge the temporal gap between the two incompatible satellite sensors using the CPR PCI, to enable a consistent cross-comparability. Using this novel synergistic re-analysis method, we aim to assess concurrent multidecadal ocean colour changes in the NE Atlantic over a half-century. We finally evaluated the observed ocean colour alterations in relation to changes seen in Northern Hemisphere Temperature (NHT).
Materials and methods
Datasets Ground truth ocean colour dataset. Towed behind ‘ships of opportunity’, the CPR has consistently sampled the upper ocean (top 10 m) using a virtually unchanged methodology since 1948 (Richardson et al., 2006). Water enters the CPR body through a small aperture and plankton is filtered onto a constantly moving (powered by an impeller) band of silk mesh. The preliminary examination of CPR samples, prior to taxonomic analysis, involves the visual estimation of the green colour of the silk mesh that filters surface waters. The silk mesh is assessed for ‘colour’ based on a relative scale of greenness (four categories) and is determined by reference to a standard colour chart. Green chlorophyll pigments, derived from chloroplasts of intact and broken cells and small unarmoured flagellates, are responsible for this colouration (Edwards et al., 2001). To carry out a robust statistical analysis of the PCI, this scale is converted into semiquantitative values so that the colour intensity of the PCI categories can be compared. Colebrook (1960) described an experimental procedure for producing numerical values for each PCI category based on acetone extracts of the samples. Recently, a field trial was undertaken to assess the numerical values assigned by previous investigations for each category of the greenness (Raitsos et al., 2013). Using modern methods, the results of this experiment confirmed the validity of this long-term in situ ocean colour data set. This in situ ocean colour estimate is a proxy of phytoplankton biomass and has been extensively used to describe
the major spatiotemporal patterns of phytoplankton, and is used here (Batten et al., 2003; Raitsos et al., 2005, 2013).
Satellite ocean colour data Level-3 monthly binned products (9 km resolution) of SeaWiFS (Sea-viewing Wide Field-of view Sensor) (version R2010.0) and CZCS reprocessed Chl-a data produced by the Ocean Biology Processing Group were acquired from the NASA Oceancolor website (http://oceancolor.gsfc.nasa.gov/).
Temperature data NHT anomalies for 1948–2007 were expressed as anomalies relative to the mean for the 1961–1990 reference period (Jones et al., 2012). NHT data were produced by the Climatic Research Unit and the Met office Hadley Centre (Jones et al., 2012).
Data analysis To perform a monthly time-series comparison between the satellite derived and in situ ocean colour measurements, all data were area-averaged for the NE Atlantic and the North Sea (Fig. 1). We acknowledge that the monthly means derived from averaging an area are potentially biased by the number of data points used. Thus, only the first 3 years of the CZCS data were used since observed records from 1982 onwards had large data gaps, which reduced the quality of the monthly means. There was similar data patchiness in both space and time in the SeaWiFS dataset from 2008 to 2010, therefore, only records until Dec 2007 were used. In addition to the above limitation, the ocean colour data used in this work can be limited by several factors: e.g. sampling frequency for the CPR, and cloud cover for satellite observations. However, we base our conclusions on the most well-sampled area of the CPR survey. Since 1948, the CPR has collected >110 000 samples, with an average of >160 samples each month. For the case of satellite data, each monthly mean is derived from several
Fig. 1 Location of Continuous Plankton Recorder samples (n = 111 803) in the Northeast Atlantic Ocean and North Sea for 1948–2007. Red points represent the CPR samples during the CZCS period (1978–1981, n = 6121), whereas blue data points represent CPR samples during the SeaWiFS era (1998 – 2007, n = 23 174). © 2013 John Wiley & Sons Ltd, Global Change Biology, 20, 2117–2123
N O R T H E A S T A T L A N T I C O C E A N - C O L O U R A N O M A L I E S 2119 thousand data points (on average >19 000 for CZCS, and >27 000 for SeaWiFS). Initially, we examined the variability of the satellite-derived Chl-a (CZCS and SeaWiFS) in relation to PCI (comparison based on their monthly means). The PCI was then calibrated against the SeaWiFS Chl-a, and PCI arbitrary values were transformed to mg Chl-a per cubic metre (mg m3). To enable a consistent cross-comparability of all three datasets, the CZCS Chl-a product was also adjusted based on SeaWiFS. We finally examined concurrent phytoplankton biomass alterations (decadal variability and phenology) in relation to changes seen in the temperature of the northern hemisphere. Pearson correlation (r) analysis was used to examine relationships between the two ocean colour datasets (in situ and satellite). The probability of significance of the correlation was adjusted to correct for temporal autocorrelation (PACF) in the datasets. This method was designed to account for autocorrelation in correlation analysis between biological and environmental data, and here it was used to calculate the probabilities with consideration of temporal autocorrelation, and thus calculate a robust significance level (Pyper & Peterman, 1998). We calculated standardized anomalies in the reconstructed Chl-a (1948–2007) to remove seasonality (and thus reduce temporal autocorrelation) and also to make these data comparable to NHT anomalies. The Chl-a standardized anomalies have been computed as the difference between the value of a given month minus the overall mean (climatological monthly mean), and divided by the climatological monthly SD. A regime shift index (RSI) combined with an automatic sequential algorithm (Rodionov, 2004) was used to examine statistically the existence, timing and significance (P < 0.05) of abrupt changes in the long-term ocean colour and NHT. The absolute value of RSI represents the magnitude of the shift(s), while its sign determines the change in direction of the mean between regimes (see Rodionov, 2004 for more information).
This statistical tool was used to identify concurrent abrupt changes between temperature and the biological data series. In addition, the cumulative sum method was used to further decompose the different trends of the decadal variability of both datasets (ocean colour and NHT). This technique emphasizes major changes in the local mean values along the timeseries and smooths high frequency variability. Successive positive anomalies yield an increasing slope, whereas successive negative anomalies yield a decreasing slope.
Results The first step was to assess the relationship between the standard satellite-derived ocean colour products (CZCS and SeaWiFS Chl-a) and in situ PCI data. Monthly CZCS Chl-a measurements and PCI data averaged for the NE Atlantic (including the North Sea) were significantly correlated (Fig. 2a, r = 0.66, PACF = 0.0001). The monthly time-series between SeaWiFS and PCI and their linear regressions Eqn (1) are presented in Fig. 2b. Based on >10 years of data, there were similar intraand interannual variation in monthly time-series of SeaWiFS and PCI (Fig. 2b), with a clear linear relationship (r = 0.67, PACF = 0.0001). However, there is a noticeable difference in the intensity of Chl-a values (mg m3) between the two sensors, as the CZCS Chl-a concentrations appeared to be higher than those of SeaWiFS. This is due to inconsistencies in sensor characteristics, and thus, the absolute values retrieved from the two instruments are not comparable. To overcome these difficulties and to investigate decadal changes in satellite (CZCS and SeaWiFS) and in situ (PCI) ocean colour data, the three different
(a)
(b)
Fig. 2 Relationships between ocean colour datasets (monthly averages) in the Northeast Atlantic Ocean and North Sea. (a) Phytoplankton Colour Index (red line) of the CPR survey vs. CZCS Chl-a (black line, Nov 1978 to Dec 1981) and (b) Phytoplankton Colour Index of the CPR survey (red line) vs. SeaWiFS Chl-a (black line, Sep 1997 to Dec 2007). © 2013 John Wiley & Sons Ltd, Global Change Biology, 20, 2117–2123
2120 D . E . R A I T S O S et al. datasets were adjusted for comparability. For instance, PCI is an in situ estimate of ocean colour, not a measurement of Chl-a, while CZCS Chl-a is not directly comparable with SeaWiFS Chl-a. Three steps were taken to adjust the ocean colour datasets. First, the semiquantitative values of PCI were converted to Chl-a (mg m3) based on the SeaWiFS measurements using Eqn (1). The 10-yr relationship between PCI and SeaWiFS [Fig. 2b, and Eqn (1)] was then used to retrospectively calculate PCI_Chl-a back to 1948. Second, a new relationship between the transformed PCI_Chl-a and original CZCS Chl-a data was derived. Last, the equation derived from the second step Eqn (2) was applied to the original CZCS Chl-a product. This adjustment enables the three different datasets to be comparable, without affecting the patterns and trends originally present. PCI Chl-a ¼ 0:3276 ðPCI original þ 0:469Þ
ð1Þ
CZCS adj ¼ 0:1657 ðCZCS original þ 0:4384Þ
ð2Þ
Decadal changes of phytoplankton biomass in the three datasets (PCI_Chl-a, CZCS_adjusted, SeaWiFS) are shown in Fig. 3a–b. In the upper panel (A), the
annual means can be observed, whereas below (B) the six-month running averages of all the datasets are depicted for the NE Atlantic and North Sea from 1948 to 2007. It is evident that the phytoplankton biomass of the study area has undergone substantial changes, which are apparent in both the in situ and remotely sensed Chl-a. The PCI Chl-a increased by 23% after the regime shift in the mid-1980s (and 25% based only on the difference between the satellite periods). When all three time-series products were plotted together, not only did the CZCS and SeaWiFS datasets show the phytoplankton biomass increase, but also similar magnitudes of Chl-a changes (Fig. 3a–b). Based on our study, CZCS retrievals overestimated Chl-a by nearly 50% on average. Figure 3b suggests that, in addition to the abrupt increase of Chl-a biomass, it is also evident that the range of the Chl-a variation increased rapidly after the mid-1980s, shifting in both satellite and PCI datasets (Fig. 3b). The Chl-a range has increased after the shift, with greater differences in Chl-a between summer and winter. To assess whether the observed phytoplankton variability in the NE Atlantic and North Sea is related to broader climate oscillations, the monthly Chl-a anomalies were plotted against NHT anomalies since 1948 (Fig. 4). Monthly anomalies of PCI Chl-a and NHT
(a) (a)
(b) (b)
Fig. 3 Decadal changes of ocean colour time-series derived from the PCI transformed to Chl-a, CZCS_adjusted Chl-a and original SeaWiFS Chl-a, depicting the variability of Chl-a in the Northeast Atlantic Ocean and the North Sea since 1948. (a) Chla annual means based on the three datasets. (b) Six-month moving average from the three ocean colour datasets. The vertical grey line depicts the significant abrupt change in the regime shift index (RSI) in 1987, as detected by the RSI algorithm in the annual variation of both timeseries (Chl-a: RSI = 0.894, P < 0.0001, and NHT: RSI = 0.538, P < 0.0001).
Fig. 4 Chlorophyll-a standardized anomalies in the Northeast Atlantic and the North Sea compared with Northern Hemisphere Temperatures (NHT) anomalies since 1948. (a) Monthly mean Chl-a anomalies vs. NHT anomalies and (b) Cumulative sum method applied to summarize the major change in the annual mean of both datasets (ocean colour and NHT anomalies). The vertical grey line depicts the significant abrupt change in 1987, as detected by the regime shift index (RSI) algorithm. © 2013 John Wiley & Sons Ltd, Global Change Biology, 20, 2117–2123
N O R T H E A S T A T L A N T I C O C E A N - C O L O U R A N O M A L I E S 2121 significantly covary (Fig. 4a: r = 0.41, PACF = 0.00002, n = 600). Both ocean colour and NHT anomalies had similar variability between 1948 and 1985, with both time series then showing a rapid concomitant increase after the mid-1980s. To further decompose the signal and differentiate the two distinct periods with disparate trends, cumulative sums of the annual means of both datasets were calculated (Fig. 4b). During the first 35 years of the time series, the cumulative sums show negative successive anomalies, whereas in the mid-1980s (in particular during 1987), there was an abrupt increase in both datasets. In addition to the overall positive relationship, successive changes (both negative and positive) seen in the NHT series are also mirrored in the Chl-a variation, suggesting a link between phytoplankton biomass dynamics in the NE Atlantic and hemispheric climate variability. To investigate phenological changes, the PCI seasonal cycle was compared with that of original satellite Chl-a products. The seasonal cycles of both datasets (in situ and satellite) were computed based on the same years and averaged over the same area of the NE Atlantic and North Sea. The left panel of Figure 5 illustrates the monthly PCI means from 1948 to 2007. Two features are apparent from this contour plot: an abrupt increase of PCI in the mid-1980s; and
(a)
(a)
(b) (b)
Fig. 5 Contour plot of the monthly time-series of the Phytoplankton Colour Index of the CPR survey in the Northeast Atlantic and the North Sea (1948 – 2008). a) Seasonal cycles of SeaWiFS Chla and PCI from 1997 to 2007, b) seasonal cycles of CZCS and PCI from 1978 to 1981. © 2013 John Wiley & Sons Ltd, Global Change Biology, 20, 2117–2123
a distinct phenological change portraying the merging of the bimodal spring and autumn seasonal cycle to a united one. Seasonal cycles of both PCI and remotely sensed Chl-a concentrations show similar patterns (Fig. 5a and b). During the CZCS years (late 1970s and early 1980s), two distinct phytoplankton seasonal peaks are evident, whereas during the SeaWiFS period (from the late 1990s onwards) the spring and autumn seasonal phytoplankton blooms are clearly merged to form a continuous spring-summer bloom (Fig. 5 a–b).
Discussion Here, a unique in situ ocean colour dataset was used as a link to bridge the gap between the two incompatible satellite sensors, so that multidecadal and phenological changes could be assessed. A simple equation Eqn (2) was produced, which can be applied to the CZCS Chl-a data to make it compatible with SeaWiFS. Using this equation, and based on the synergistic use of two independent datasets (a half century of in situ and remotely sensed ocean colour measurements) concurrent ocean colour alterations were shown in the NE Atlantic and North Sea. It is evident that the phytoplankton biomass has increased markedly in both satellite and PCI datasets. Assessing the phenology (bloom timing) of phytoplankton is important not only because it is a sensitive indicator of climate change (Hughes, 2000), but also because alterations in timing can lead to a mismatch between trophic levels and functional groups (Edwards & Richardson, 2004). Results from the synergistic re-analysis reflect a merging of the two seasonal phytoplankton blooms producing a longer growing season and higher seasonal biomass, since the mid-1980s. The spring bloom has increased in strength during the study period, and most notably post-regime shift. This could be an indication of the difficulty in predicting the fate of phytoplankton after abrupt changes, and highlights the importance of historical information on phytoplankton biomass variability for comparison. The substantial increase in the duration of the high-Chlorophyll period, could ultimately lead to considerable trophic changes. The results presented here are of mutual benefit for both the validity of the remotely sensed phytoplankton decadal changes and confirmation of the CPR decadal variability. In accordance with our results, Antoine et al. (2005) suggested that part of the increase of the North Atlantic Chl-a is due to seasonal alterations, and particularly the increased biomass of the spring and summer blooms during the SeaWiFS years. In addition, Gregg & Conkright (2002) reported that the spring bloom peak was
2122 D . E . R A I T S O S et al. higher in the present decade compared with the 1980s in the North Atlantic. The initiation of the North Atlantic phytoplankton blooms is strongly related with insolation and sea surface temperatures (Raitsos et al., 2011; Racault et al., 2012). However, these factors are not necessarily responsible for the phenological changes documented here. For example, a recent study suggested that the deepening of the mixed layer depth and the intensification of winter winds are responsible for the stronger NE Atlantic spring bloom (Martinez et al., 2011). In contrast with our results, using ocean transparency measurements along with in situ chlorophyll data (not including PCI), Boyce et al. (2010) reported a decline in global phytoplankton, as well as in the North Atlantic Ocean. The study by Boyce et al. (2010) contradicts past studies based on satellite (Gregg & Conkright, 2002; Antoine et al., 2005; Martinez et al., 2009), in situ (Reid et al., 1998; Edwards et al., 2001; McQuatters-Gollop, 2011; Wernand et al., 2013) and modeled datasets (Henson et al., 2009a) that provide evidence for an increase in phytoplankton biomass in the NE Atlantic and the North Sea. Using remotely sensed Chl-a (during CZCS and SeaWiFS eras), Antoine et al. (2005) showed that one of the most pronounced areas of phytoplankton increase was in the NE Atlantic (45°N–20° W). They also stated that the annual mean Chl-a concentration in the North Atlantic has increased by 23% between the two satellite eras. Here, based on a different methodological approach and dataset, similar results were found, as the in situ PCI Chl-a appeared to have increased by 23% after the mid-1980s. Further evidence indicated that although a large portion of global ocean Chl-a has decreased from the CZCS to SeaWiFS era, the NE Atlantic and North Sea phytoplankton biomass has clearly increased (Gregg & Conkright, 2002). Analysis of biogeochemical models indicates that Chl-a is increasing in the NE Atlantic and North Sea (Henson et al., 2009a); however, although the model reproduced the interannual variability, it could not simulate the abrupt phytoplankton shift. No matter which methodological approach is used, using in situ, satellite, modelled and/or blended datasets, our results and almost all the literature support the assertion that phytoplankton biomass is increasing in the NE Atlantic and the North Sea. To gain further confidence on the validity of the reconstructed multi-decadal ocean colour timeseries, we assessed the observed variability in relation to broader climate oscillations (as seen from the temperature of the northern hemisphere). Ocean colour anomalies of the NE Atlantic and the North Sea significantly parallel the NHT oscillation since 1948 (at a monthly and annual temporal scale). However, the exact
mechanism (top-down and/or bottom-up) behind this relationship remains unclear and an in depth investigation is out of the scope of this study. In general, oceanic warming at higher latitudes (nutrient-rich, light-limited areas in temperate and polar regions) contributes to reduced mixing that may lead to increased phytoplankton growth (Richardson & Schoeman, 2004; Doney, 2006). Martinez et al. (2009) reported that although the regions of opposite Chl-a and sea surface temperature (SST) patterns account for 60% of the global ocean, the NE Atlantic Ocean is characterized as an area of increasing SST and higher phytoplankton abundance. Significant changes in the North Sea and the NE Atlantic have been documented across various trophic levels (Reid et al., 2003; Beaugrand, 2004; Edwards & Richardson, 2004; Beaugrand et al., 2008). In the North Sea, long-term plankton alterations have been attributed to warmer-than-average SST, a positive phase in the North Atlantic Oscillation (NAO) index and increased oceanic inflow from the North Atlantic (Beaugrand, 2004). Waters in the North Sea are also becoming less turbid, a situation that allows the normally light-limited coastal phytoplankton to use lower concentrations of nutrients more effectively (McQuatters-Gollop et al., 2007). In the North Atlantic Ocean, phytoplankton blooms and community structure variability have been strongly associated with different phases of the NAO index (Henson et al., 2009b, 2012). However, a unified ecological response to the documented environmental changes should not be expected throughout the entire North Atlantic (Beaugrand, 2009), as it is composed of several different complex biomes (Longhurst, 2007) and these may react differently to climatic variations. Forecasting the potential response of phytoplankton to climate variability and change requires information provided by the synergistic re-analysis of the few available long-term in situ datasets along with the remotely sensed ones. With satellite sensors having limited lifespans, the CPR colour index has already outlasted CZCS, SeaWiFS and MERIS, and with the approaching end of MODIS, it is imperative that we maintain the monitoring of phytoplankton from in situ ocean colour datasets (e.g. CPR or Forel-Ule) so we have a long-term baseline that accompanies the pioneering satellite missions.
Acknowledgements The authors gratefully acknowledge the data sources for this study: CZCS and SeaWiFS Project – NASA/Goddard Space Flight Center (http://oceancolor.gsfc.nasa.gov). We are also grateful to past and present staff of SAHFOS who have contributed to the maintenance of the CPR time series. Particular thanks to David Johns for his advice on CPR datasets and methods.
© 2013 John Wiley & Sons Ltd, Global Change Biology, 20, 2117–2123
N O R T H E A S T A T L A N T I C O C E A N - C O L O U R A N O M A L I E S 2123
REFERENCES
Longhurst AR (2007) Ecological Geography of the Sea (2nd edn). pp. 1–542. Elsevier Inc, London. ISBN: 978-0-12-455521-1.
Antoine D, Morel A, Gordon HR, Banzon VF, Evans RH (2005) Bridging ocean color
Martinez E, Antoine D, D’Ortenzio F, Gentili B (2009) Climate-driven basin-scale decadal oscillations of oceanic phytoplankton. Science, 326, 1253. Martinez E, Antoine D, D’Ortenzio F, de Boyer Montegut C (2011) Phytoplankton spring and fall blooms in the North Atlantic in the 1980s and 2000s. Journal of Geophysical Research, 116, C11029. McClain CRA (2009) Decade of satellite ocean color observations. Annual Review of Marine Science, 1, 19–42.
observations of the 1980s and 2000s in search of long-term trends. Journal of Geophysical Research, 110, C06009. Batten SD, Walne AW, Edwards M, Groom SB (2003) Phytoplankton biomass from continuous plankton recorder data: an assessment of the phytoplankton colour index. Journal of Plankton Research, 25, 697–702. Beaugrand G (2004) The North Sea regime shift: evidence, causes, mechanisms and consequences. Progress in Oceanography, 60, 245–262. Beaugrand G (2005) Monitoring pelagic ecosystems using plankton indicators. ICES Journal of Marine Sciences, 62, 333–338. Beaugrand G (2009) Decadal changes in climate and ecosystems in the North Atlantic Ocean and adjacent seas. Deep Sea Research Part II, 56, 659–673. Beaugrand G, Edwards M, Brander K, Luczak C, Ibanez F (2008) Causes and projections of abrupt climate-driven ecosystem shifts in the North Atlantic. Ecology Letters, 11, 1157–1168. Boyce DG, Lewis MR, Worm B (2010) Global phytoplankton decline over the past century. Nature, 466, 591–596. Chavez FP, Messie M, Pennington JT (2011) Marine primary production in relation to climate variability and change. Annual Review of Marine Science, 3, 227–260. Colebrook JM (1960) Continuous plankton records: methods of analysis, 1950–59. Bulletin of Marine Ecology, 5, 51–64. Doney SC (2006) Plankton in a warmer world. Nature, 444, 695–696. Edwards M, Richardson AJ (2004) Impact of climate change on marine pelagic phenology and trophic mismatch. Nature, 430, 881–884. Edwards M, Reid PC, Planque B (2001) Long-term and regional variability of phytoplankton biomass in the Northeast Atlantic (1960-1995). ICES Journal of Marine Sciences, 58, 39–49. Gregg WW, Conkright ME (2002) Decadal changes in global ocean chlorophyll. Geophysical Research Letters, 29, 1730. Hays GC, Richardson AJ, Robinson C (2005) Climate change and marine plankton. Trends in Ecology and Evolution, 20, 337–344. Henson S, Raitsos D, Dunne JP, McQuatters-Gollop A (2009a) Decadal variability in biogeochemical models: comparison with a 50-year ocean colour dataset. Geophysical Research Letters, 36, L21601. Henson S, Dunne JP, Sarmiento JL (2009b) Decadal variability in North Atlantic phytoplankton blooms. Journal of Geophysical Research, 114, C04013. Henson S, Lampitt R, Johns D (2012) Variability in phytoplankton community structure in response to the North Atlantic Oscillation and implications for organic carbon flux. Limnology and Oceanography, 57, 1591–1601. Hughes L (2000) Biological consequences of global warming: is the signal already apparent? Trends in Ecology and Evolution, 15, 56–61. IPCC [Intergovernmental Panel on Climate Change] (2007) Climate Change 2007: Impacts, Adaptation and Vulnerability. Cambridge University Press, Cambridge. Jones PD, Lister DH, Osborn TJ, Harpham C, Salmon M, Morice CP (2012) Hemispheric and large-scale land surface air temperature variations: an extensive revision and an update to 2010. Journal of Geophysical Research, 117, D05127.
© 2013 John Wiley & Sons Ltd, Global Change Biology, 20, 2117–2123
McQuatters-Gollop A, Raitsos DE, Edwards M, Pradhan Y, Mee LD, Lavender SJ, Attrill MJ (2007) A long-term chlorophyll dataset reveals a regime shift in North Sea phytoplankton biomass unconnected to increasing nutrient levels. Limnology and Oceanography, 52, 635–648. McQuatters-Gollop A et al. (2011) Is there a decline in marine phytoplankton? Nature, 472, E6. Platt T, Stuart V (1997) Sir Alister Hardy: an ocean colour pioneer (1896–1985). Backscatter, 8, 1. Pyper BJ, Peterman RM (1998) Comparison of methods to account for autocorrelation in correlation analyses of fish data. Canadian Journal of Fisheries and Aquatic Sciences, 55, 2127–2140. Racault MF, Le Quere C, Buitenhuis E et al. (2012) Phytoplankton phenology in the global ocean. Ecological Indicators, 14, 152–163. Raitsos DE, Reid PC, Lavender SJ, Edwards M, Richardson AJ (2005) Extending the SeaWiFS chlorophyll data set back 50 years in the northeast Atlantic. Geophysical Research Letters, 32, L06603. Raitsos DE, Lavender SJ, Maravelias CD, Haralabous J, McQuatters-Gollop A, Edwards M, Reid PC (2011) Macroscale factors affecting diatom abundance: a synergistic use of Continuous Plankton Recorder and satellite remote sensing data. International Journal of Remote Sensing, 32, 2081–2094. Raitsos DE, Walne A, Lavender SJ, Licandro P, Reid PC, Edwards M (2013) A 60 year ocean colour dataset from the Continuous Plankton Recorder. Journal of Plankton Research, 35, 158–164. Reid PC, Edwards M, Hunt HG, Warner A (1998) Phytoplankton change in the North Atlantic. Nature, 391, 546. Reid PC, Edwards M, Beaugrand G, Skogen M, Stevens D (2003) Periodic changes in the zooplankton of the North Sea during the twentieth century linked to oceanic inflow. Fisheries Oceanography, 12, 160–169. Richardson AJ, Schoeman DS (2004) Climate impact on plankton ecosystems in the northeast Atlantic. Science, 305, 1609–1612. Richardson AJ, Walne AW, John AWG, Jonas TD, Lindley JA, Sims DW, Witt M (2006) Using continuous plankton recorder data. Progress in Oceanography, 68, 27–74. Rodionov SNA (2004) Sequential algorithm for testing climate regime shifts. Geophysical Research Letters, 31, L09204. Wernand MR, van der Woerd HJ, Gieskes WWC (2013) Trends in ocean colour and chlorophyll concentration from 1889 to 2000 worldwide. PLoS ONE, 8, e63766.
From silk to satellite: half a century of ocean colour anomalies in the Northeast Atlantic DIONYSIOS E. RAITSOS* , YASWANT PRADHAN†, SAMANTHA J. LAVENDER‡§, I B R A H I M H O T E I T ¶ , A B I G A I L M C Q U A T T E R S - G O L L O P k, P H I L L I P C . R E I D § k* * and ANTHONY J. RICHARDSON††‡‡ *Plymouth Marine Laboratory (PML), Prospect Place, The Hoe, Plymouth PL1 3DH, UK, †Met Office, FitzRoy Road, Exeter EX1 3PB, UK, ‡Pixalytics Ltd, 1 Davy Road, Tamar Science Park, Derriford, Plymouth Devon PL6 8BX, UK, §School of Marine Science & Engineering/Marine Institute, University of Plymouth, Drake Circus, Plymouth PL4 8AA, UK, ¶King Abdullah University for Science and Technology, Thuwal, Saudi Arabia, kSir Alister Hardy Foundation for Ocean Science (SAHFOS), The Laboratory, Citadel Hill, Plymouth PL1 2PB, UK, **Marine Biological Association of the UK, The Laboratory, Citadel Hill, Plymouth PL1 2PB, UK, ††Centre for Applications in Natural Resource Mathematics (CARM), School of Mathematics and Physics, University of Queensland, St Lucia Queensland 4072, Australia, ‡‡Climate Adaptation Flagship, CSIRO Marine and Atmospheric Research, Ecosciences Precinct, GPO Box 2583, Dutton Park, Queensland 4102, Australia
Abstract Changes in phytoplankton dynamics influence marine biogeochemical cycles, climate processes, and food webs, with substantial social and economic consequences. Large-scale estimation of phytoplankton biomass was possible via ocean colour measurements from two remote sensing satellites – the Coastal Zone Colour Scanner (CZCS, 1979–1986) and the Sea-viewing Wide Field-of-view Sensor (SeaWiFS, 1998–2010). Due to the large gap between the two satellite eras and differences in sensor characteristics, comparison of the absolute values retrieved from the two instruments remains challenging. Using a unique in situ ocean colour dataset that spans more than half a century, the two satellite-derived chlorophyll-a (Chl-a) eras are linked to assess concurrent changes in phytoplankton variability and bloom timing over the Northeast Atlantic Ocean and North Sea. Results from this unique re-analysis reflect a clear increasing pattern of Chl-a, a merging of the two seasonal phytoplankton blooms producing a longer growing season and higher seasonal biomass, since the mid-1980s. The broader climate plays a key role in Chl-a variability as the ocean colour anomalies parallel the oscillations of the Northern Hemisphere Temperature (NHT) since 1948. Keywords: chlorophyll, Northeast Atlantic, northern hemisphere temperature, ocean colour, phytoplankton variability Received 13 May 2013 and accepted 22 October 2013
Introduction Phytoplankton that form the base of the marine food web and produce ca. 50% of global primary production, are sensitive to climate variability and environmental change (Beaugrand, 2005; Hays et al., 2005). It is thus important to monitor phytoplankton biomass through time, although this has been challenging considering the sparse long-term information. Even for the available datasets, there is uncertainty on the trends of phytoplankton biomass, with studies reporting contradictory findings (summarized by Chavez et al., 2011). For example, Boyce et al. (2010) report a decline in phytoplankton biomass in the North Atlantic based on ocean transparency measurements and in situ chlorophyll data. By contrast, Wernand et al. (2013), McQuattersGollop (2011) and Antoine et al. (2005) have shown an increase in phytoplankton biomass in the North Correspondence: Dionysios E. Raitsos, tel. +44 01752 633406, fax +44 01752 633 101, e-mail: [email protected]
© 2013 John Wiley & Sons Ltd
Atlantic. This uncertainty in the historical changes of phytoplankton makes it difficult to predict changes in the future as the climate continues to warm. As phytoplankton ultimately supports most of the world’s fisheries, uncertainties in its historical and future response to climate change hamper policy decisions [IPCC (Intergovernmental Panel on Climate Change), 2007]. One way to increase our knowledge of change in phytoplankton biomass is to re-analyse the few available long-term in situ datasets in conjunction with data from satellite remote sensing. In the Northeast (NE) Atlantic Ocean and North Sea, ocean colour is estimated via two distinct methods (over large spatial and temporal scales): through satellite remote sensing and with the Continuous Plankton Recorder (CPR). Platt & Stuart (1997) reported that ‘until the advent of satellite remote sensing, the CPR provided the only means of collecting plankton data at large spatial scales: it was the remote sensing of the day’. In the late 1970s, the CZCS ocean colour satellite offered the first global view of phytoplankton biomass 2117
2118 D . E . R A I T S O S et al. patterns. The CZCS mission ended in the mid-1980s and no other satellite with ocean colour measurement capabilities was launched until the mid-1990s (McClain, 2009), leaving a nearly 12-year gap. Due to this gap and methodological differences compared with modern sensors, CZCS and contemporary satellite datasets cannot be directly compared to assess decadal changes (Antoine et al., 2005). The Phytoplankton Colour Index (PCI) from the CPR dataset has been collected consistently since 1948 (Richardson et al., 2006), and covers the period before and during the satellite eras. These data could serve as a link between the two sensor eras, as its validity compared with satellite derived Chl-a has been demonstrated previously (Batten et al., 2003; Raitsos et al., 2005; McQuatters-Gollop et al., 2007; Henson et al., 2009a). Here, we aim to bridge the temporal gap between the two incompatible satellite sensors using the CPR PCI, to enable a consistent cross-comparability. Using this novel synergistic re-analysis method, we aim to assess concurrent multidecadal ocean colour changes in the NE Atlantic over a half-century. We finally evaluated the observed ocean colour alterations in relation to changes seen in Northern Hemisphere Temperature (NHT).
Materials and methods
Datasets Ground truth ocean colour dataset. Towed behind ‘ships of opportunity’, the CPR has consistently sampled the upper ocean (top 10 m) using a virtually unchanged methodology since 1948 (Richardson et al., 2006). Water enters the CPR body through a small aperture and plankton is filtered onto a constantly moving (powered by an impeller) band of silk mesh. The preliminary examination of CPR samples, prior to taxonomic analysis, involves the visual estimation of the green colour of the silk mesh that filters surface waters. The silk mesh is assessed for ‘colour’ based on a relative scale of greenness (four categories) and is determined by reference to a standard colour chart. Green chlorophyll pigments, derived from chloroplasts of intact and broken cells and small unarmoured flagellates, are responsible for this colouration (Edwards et al., 2001). To carry out a robust statistical analysis of the PCI, this scale is converted into semiquantitative values so that the colour intensity of the PCI categories can be compared. Colebrook (1960) described an experimental procedure for producing numerical values for each PCI category based on acetone extracts of the samples. Recently, a field trial was undertaken to assess the numerical values assigned by previous investigations for each category of the greenness (Raitsos et al., 2013). Using modern methods, the results of this experiment confirmed the validity of this long-term in situ ocean colour data set. This in situ ocean colour estimate is a proxy of phytoplankton biomass and has been extensively used to describe
the major spatiotemporal patterns of phytoplankton, and is used here (Batten et al., 2003; Raitsos et al., 2005, 2013).
Satellite ocean colour data Level-3 monthly binned products (9 km resolution) of SeaWiFS (Sea-viewing Wide Field-of view Sensor) (version R2010.0) and CZCS reprocessed Chl-a data produced by the Ocean Biology Processing Group were acquired from the NASA Oceancolor website (http://oceancolor.gsfc.nasa.gov/).
Temperature data NHT anomalies for 1948–2007 were expressed as anomalies relative to the mean for the 1961–1990 reference period (Jones et al., 2012). NHT data were produced by the Climatic Research Unit and the Met office Hadley Centre (Jones et al., 2012).
Data analysis To perform a monthly time-series comparison between the satellite derived and in situ ocean colour measurements, all data were area-averaged for the NE Atlantic and the North Sea (Fig. 1). We acknowledge that the monthly means derived from averaging an area are potentially biased by the number of data points used. Thus, only the first 3 years of the CZCS data were used since observed records from 1982 onwards had large data gaps, which reduced the quality of the monthly means. There was similar data patchiness in both space and time in the SeaWiFS dataset from 2008 to 2010, therefore, only records until Dec 2007 were used. In addition to the above limitation, the ocean colour data used in this work can be limited by several factors: e.g. sampling frequency for the CPR, and cloud cover for satellite observations. However, we base our conclusions on the most well-sampled area of the CPR survey. Since 1948, the CPR has collected >110 000 samples, with an average of >160 samples each month. For the case of satellite data, each monthly mean is derived from several
Fig. 1 Location of Continuous Plankton Recorder samples (n = 111 803) in the Northeast Atlantic Ocean and North Sea for 1948–2007. Red points represent the CPR samples during the CZCS period (1978–1981, n = 6121), whereas blue data points represent CPR samples during the SeaWiFS era (1998 – 2007, n = 23 174). © 2013 John Wiley & Sons Ltd, Global Change Biology, 20, 2117–2123
N O R T H E A S T A T L A N T I C O C E A N - C O L O U R A N O M A L I E S 2119 thousand data points (on average >19 000 for CZCS, and >27 000 for SeaWiFS). Initially, we examined the variability of the satellite-derived Chl-a (CZCS and SeaWiFS) in relation to PCI (comparison based on their monthly means). The PCI was then calibrated against the SeaWiFS Chl-a, and PCI arbitrary values were transformed to mg Chl-a per cubic metre (mg m3). To enable a consistent cross-comparability of all three datasets, the CZCS Chl-a product was also adjusted based on SeaWiFS. We finally examined concurrent phytoplankton biomass alterations (decadal variability and phenology) in relation to changes seen in the temperature of the northern hemisphere. Pearson correlation (r) analysis was used to examine relationships between the two ocean colour datasets (in situ and satellite). The probability of significance of the correlation was adjusted to correct for temporal autocorrelation (PACF) in the datasets. This method was designed to account for autocorrelation in correlation analysis between biological and environmental data, and here it was used to calculate the probabilities with consideration of temporal autocorrelation, and thus calculate a robust significance level (Pyper & Peterman, 1998). We calculated standardized anomalies in the reconstructed Chl-a (1948–2007) to remove seasonality (and thus reduce temporal autocorrelation) and also to make these data comparable to NHT anomalies. The Chl-a standardized anomalies have been computed as the difference between the value of a given month minus the overall mean (climatological monthly mean), and divided by the climatological monthly SD. A regime shift index (RSI) combined with an automatic sequential algorithm (Rodionov, 2004) was used to examine statistically the existence, timing and significance (P < 0.05) of abrupt changes in the long-term ocean colour and NHT. The absolute value of RSI represents the magnitude of the shift(s), while its sign determines the change in direction of the mean between regimes (see Rodionov, 2004 for more information).
This statistical tool was used to identify concurrent abrupt changes between temperature and the biological data series. In addition, the cumulative sum method was used to further decompose the different trends of the decadal variability of both datasets (ocean colour and NHT). This technique emphasizes major changes in the local mean values along the timeseries and smooths high frequency variability. Successive positive anomalies yield an increasing slope, whereas successive negative anomalies yield a decreasing slope.
Results The first step was to assess the relationship between the standard satellite-derived ocean colour products (CZCS and SeaWiFS Chl-a) and in situ PCI data. Monthly CZCS Chl-a measurements and PCI data averaged for the NE Atlantic (including the North Sea) were significantly correlated (Fig. 2a, r = 0.66, PACF = 0.0001). The monthly time-series between SeaWiFS and PCI and their linear regressions Eqn (1) are presented in Fig. 2b. Based on >10 years of data, there were similar intraand interannual variation in monthly time-series of SeaWiFS and PCI (Fig. 2b), with a clear linear relationship (r = 0.67, PACF = 0.0001). However, there is a noticeable difference in the intensity of Chl-a values (mg m3) between the two sensors, as the CZCS Chl-a concentrations appeared to be higher than those of SeaWiFS. This is due to inconsistencies in sensor characteristics, and thus, the absolute values retrieved from the two instruments are not comparable. To overcome these difficulties and to investigate decadal changes in satellite (CZCS and SeaWiFS) and in situ (PCI) ocean colour data, the three different
(a)
(b)
Fig. 2 Relationships between ocean colour datasets (monthly averages) in the Northeast Atlantic Ocean and North Sea. (a) Phytoplankton Colour Index (red line) of the CPR survey vs. CZCS Chl-a (black line, Nov 1978 to Dec 1981) and (b) Phytoplankton Colour Index of the CPR survey (red line) vs. SeaWiFS Chl-a (black line, Sep 1997 to Dec 2007). © 2013 John Wiley & Sons Ltd, Global Change Biology, 20, 2117–2123
2120 D . E . R A I T S O S et al. datasets were adjusted for comparability. For instance, PCI is an in situ estimate of ocean colour, not a measurement of Chl-a, while CZCS Chl-a is not directly comparable with SeaWiFS Chl-a. Three steps were taken to adjust the ocean colour datasets. First, the semiquantitative values of PCI were converted to Chl-a (mg m3) based on the SeaWiFS measurements using Eqn (1). The 10-yr relationship between PCI and SeaWiFS [Fig. 2b, and Eqn (1)] was then used to retrospectively calculate PCI_Chl-a back to 1948. Second, a new relationship between the transformed PCI_Chl-a and original CZCS Chl-a data was derived. Last, the equation derived from the second step Eqn (2) was applied to the original CZCS Chl-a product. This adjustment enables the three different datasets to be comparable, without affecting the patterns and trends originally present. PCI Chl-a ¼ 0:3276 ðPCI original þ 0:469Þ
ð1Þ
CZCS adj ¼ 0:1657 ðCZCS original þ 0:4384Þ
ð2Þ
Decadal changes of phytoplankton biomass in the three datasets (PCI_Chl-a, CZCS_adjusted, SeaWiFS) are shown in Fig. 3a–b. In the upper panel (A), the
annual means can be observed, whereas below (B) the six-month running averages of all the datasets are depicted for the NE Atlantic and North Sea from 1948 to 2007. It is evident that the phytoplankton biomass of the study area has undergone substantial changes, which are apparent in both the in situ and remotely sensed Chl-a. The PCI Chl-a increased by 23% after the regime shift in the mid-1980s (and 25% based only on the difference between the satellite periods). When all three time-series products were plotted together, not only did the CZCS and SeaWiFS datasets show the phytoplankton biomass increase, but also similar magnitudes of Chl-a changes (Fig. 3a–b). Based on our study, CZCS retrievals overestimated Chl-a by nearly 50% on average. Figure 3b suggests that, in addition to the abrupt increase of Chl-a biomass, it is also evident that the range of the Chl-a variation increased rapidly after the mid-1980s, shifting in both satellite and PCI datasets (Fig. 3b). The Chl-a range has increased after the shift, with greater differences in Chl-a between summer and winter. To assess whether the observed phytoplankton variability in the NE Atlantic and North Sea is related to broader climate oscillations, the monthly Chl-a anomalies were plotted against NHT anomalies since 1948 (Fig. 4). Monthly anomalies of PCI Chl-a and NHT
(a) (a)
(b) (b)
Fig. 3 Decadal changes of ocean colour time-series derived from the PCI transformed to Chl-a, CZCS_adjusted Chl-a and original SeaWiFS Chl-a, depicting the variability of Chl-a in the Northeast Atlantic Ocean and the North Sea since 1948. (a) Chla annual means based on the three datasets. (b) Six-month moving average from the three ocean colour datasets. The vertical grey line depicts the significant abrupt change in the regime shift index (RSI) in 1987, as detected by the RSI algorithm in the annual variation of both timeseries (Chl-a: RSI = 0.894, P < 0.0001, and NHT: RSI = 0.538, P < 0.0001).
Fig. 4 Chlorophyll-a standardized anomalies in the Northeast Atlantic and the North Sea compared with Northern Hemisphere Temperatures (NHT) anomalies since 1948. (a) Monthly mean Chl-a anomalies vs. NHT anomalies and (b) Cumulative sum method applied to summarize the major change in the annual mean of both datasets (ocean colour and NHT anomalies). The vertical grey line depicts the significant abrupt change in 1987, as detected by the regime shift index (RSI) algorithm. © 2013 John Wiley & Sons Ltd, Global Change Biology, 20, 2117–2123
N O R T H E A S T A T L A N T I C O C E A N - C O L O U R A N O M A L I E S 2121 significantly covary (Fig. 4a: r = 0.41, PACF = 0.00002, n = 600). Both ocean colour and NHT anomalies had similar variability between 1948 and 1985, with both time series then showing a rapid concomitant increase after the mid-1980s. To further decompose the signal and differentiate the two distinct periods with disparate trends, cumulative sums of the annual means of both datasets were calculated (Fig. 4b). During the first 35 years of the time series, the cumulative sums show negative successive anomalies, whereas in the mid-1980s (in particular during 1987), there was an abrupt increase in both datasets. In addition to the overall positive relationship, successive changes (both negative and positive) seen in the NHT series are also mirrored in the Chl-a variation, suggesting a link between phytoplankton biomass dynamics in the NE Atlantic and hemispheric climate variability. To investigate phenological changes, the PCI seasonal cycle was compared with that of original satellite Chl-a products. The seasonal cycles of both datasets (in situ and satellite) were computed based on the same years and averaged over the same area of the NE Atlantic and North Sea. The left panel of Figure 5 illustrates the monthly PCI means from 1948 to 2007. Two features are apparent from this contour plot: an abrupt increase of PCI in the mid-1980s; and
(a)
(a)
(b) (b)
Fig. 5 Contour plot of the monthly time-series of the Phytoplankton Colour Index of the CPR survey in the Northeast Atlantic and the North Sea (1948 – 2008). a) Seasonal cycles of SeaWiFS Chla and PCI from 1997 to 2007, b) seasonal cycles of CZCS and PCI from 1978 to 1981. © 2013 John Wiley & Sons Ltd, Global Change Biology, 20, 2117–2123
a distinct phenological change portraying the merging of the bimodal spring and autumn seasonal cycle to a united one. Seasonal cycles of both PCI and remotely sensed Chl-a concentrations show similar patterns (Fig. 5a and b). During the CZCS years (late 1970s and early 1980s), two distinct phytoplankton seasonal peaks are evident, whereas during the SeaWiFS period (from the late 1990s onwards) the spring and autumn seasonal phytoplankton blooms are clearly merged to form a continuous spring-summer bloom (Fig. 5 a–b).
Discussion Here, a unique in situ ocean colour dataset was used as a link to bridge the gap between the two incompatible satellite sensors, so that multidecadal and phenological changes could be assessed. A simple equation Eqn (2) was produced, which can be applied to the CZCS Chl-a data to make it compatible with SeaWiFS. Using this equation, and based on the synergistic use of two independent datasets (a half century of in situ and remotely sensed ocean colour measurements) concurrent ocean colour alterations were shown in the NE Atlantic and North Sea. It is evident that the phytoplankton biomass has increased markedly in both satellite and PCI datasets. Assessing the phenology (bloom timing) of phytoplankton is important not only because it is a sensitive indicator of climate change (Hughes, 2000), but also because alterations in timing can lead to a mismatch between trophic levels and functional groups (Edwards & Richardson, 2004). Results from the synergistic re-analysis reflect a merging of the two seasonal phytoplankton blooms producing a longer growing season and higher seasonal biomass, since the mid-1980s. The spring bloom has increased in strength during the study period, and most notably post-regime shift. This could be an indication of the difficulty in predicting the fate of phytoplankton after abrupt changes, and highlights the importance of historical information on phytoplankton biomass variability for comparison. The substantial increase in the duration of the high-Chlorophyll period, could ultimately lead to considerable trophic changes. The results presented here are of mutual benefit for both the validity of the remotely sensed phytoplankton decadal changes and confirmation of the CPR decadal variability. In accordance with our results, Antoine et al. (2005) suggested that part of the increase of the North Atlantic Chl-a is due to seasonal alterations, and particularly the increased biomass of the spring and summer blooms during the SeaWiFS years. In addition, Gregg & Conkright (2002) reported that the spring bloom peak was
2122 D . E . R A I T S O S et al. higher in the present decade compared with the 1980s in the North Atlantic. The initiation of the North Atlantic phytoplankton blooms is strongly related with insolation and sea surface temperatures (Raitsos et al., 2011; Racault et al., 2012). However, these factors are not necessarily responsible for the phenological changes documented here. For example, a recent study suggested that the deepening of the mixed layer depth and the intensification of winter winds are responsible for the stronger NE Atlantic spring bloom (Martinez et al., 2011). In contrast with our results, using ocean transparency measurements along with in situ chlorophyll data (not including PCI), Boyce et al. (2010) reported a decline in global phytoplankton, as well as in the North Atlantic Ocean. The study by Boyce et al. (2010) contradicts past studies based on satellite (Gregg & Conkright, 2002; Antoine et al., 2005; Martinez et al., 2009), in situ (Reid et al., 1998; Edwards et al., 2001; McQuatters-Gollop, 2011; Wernand et al., 2013) and modeled datasets (Henson et al., 2009a) that provide evidence for an increase in phytoplankton biomass in the NE Atlantic and the North Sea. Using remotely sensed Chl-a (during CZCS and SeaWiFS eras), Antoine et al. (2005) showed that one of the most pronounced areas of phytoplankton increase was in the NE Atlantic (45°N–20° W). They also stated that the annual mean Chl-a concentration in the North Atlantic has increased by 23% between the two satellite eras. Here, based on a different methodological approach and dataset, similar results were found, as the in situ PCI Chl-a appeared to have increased by 23% after the mid-1980s. Further evidence indicated that although a large portion of global ocean Chl-a has decreased from the CZCS to SeaWiFS era, the NE Atlantic and North Sea phytoplankton biomass has clearly increased (Gregg & Conkright, 2002). Analysis of biogeochemical models indicates that Chl-a is increasing in the NE Atlantic and North Sea (Henson et al., 2009a); however, although the model reproduced the interannual variability, it could not simulate the abrupt phytoplankton shift. No matter which methodological approach is used, using in situ, satellite, modelled and/or blended datasets, our results and almost all the literature support the assertion that phytoplankton biomass is increasing in the NE Atlantic and the North Sea. To gain further confidence on the validity of the reconstructed multi-decadal ocean colour timeseries, we assessed the observed variability in relation to broader climate oscillations (as seen from the temperature of the northern hemisphere). Ocean colour anomalies of the NE Atlantic and the North Sea significantly parallel the NHT oscillation since 1948 (at a monthly and annual temporal scale). However, the exact
mechanism (top-down and/or bottom-up) behind this relationship remains unclear and an in depth investigation is out of the scope of this study. In general, oceanic warming at higher latitudes (nutrient-rich, light-limited areas in temperate and polar regions) contributes to reduced mixing that may lead to increased phytoplankton growth (Richardson & Schoeman, 2004; Doney, 2006). Martinez et al. (2009) reported that although the regions of opposite Chl-a and sea surface temperature (SST) patterns account for 60% of the global ocean, the NE Atlantic Ocean is characterized as an area of increasing SST and higher phytoplankton abundance. Significant changes in the North Sea and the NE Atlantic have been documented across various trophic levels (Reid et al., 2003; Beaugrand, 2004; Edwards & Richardson, 2004; Beaugrand et al., 2008). In the North Sea, long-term plankton alterations have been attributed to warmer-than-average SST, a positive phase in the North Atlantic Oscillation (NAO) index and increased oceanic inflow from the North Atlantic (Beaugrand, 2004). Waters in the North Sea are also becoming less turbid, a situation that allows the normally light-limited coastal phytoplankton to use lower concentrations of nutrients more effectively (McQuatters-Gollop et al., 2007). In the North Atlantic Ocean, phytoplankton blooms and community structure variability have been strongly associated with different phases of the NAO index (Henson et al., 2009b, 2012). However, a unified ecological response to the documented environmental changes should not be expected throughout the entire North Atlantic (Beaugrand, 2009), as it is composed of several different complex biomes (Longhurst, 2007) and these may react differently to climatic variations. Forecasting the potential response of phytoplankton to climate variability and change requires information provided by the synergistic re-analysis of the few available long-term in situ datasets along with the remotely sensed ones. With satellite sensors having limited lifespans, the CPR colour index has already outlasted CZCS, SeaWiFS and MERIS, and with the approaching end of MODIS, it is imperative that we maintain the monitoring of phytoplankton from in situ ocean colour datasets (e.g. CPR or Forel-Ule) so we have a long-term baseline that accompanies the pioneering satellite missions.
Acknowledgements The authors gratefully acknowledge the data sources for this study: CZCS and SeaWiFS Project – NASA/Goddard Space Flight Center (http://oceancolor.gsfc.nasa.gov). We are also grateful to past and present staff of SAHFOS who have contributed to the maintenance of the CPR time series. Particular thanks to David Johns for his advice on CPR datasets and methods.
© 2013 John Wiley & Sons Ltd, Global Change Biology, 20, 2117–2123
N O R T H E A S T A T L A N T I C O C E A N - C O L O U R A N O M A L I E S 2123
REFERENCES
Longhurst AR (2007) Ecological Geography of the Sea (2nd edn). pp. 1–542. Elsevier Inc, London. ISBN: 978-0-12-455521-1.
Antoine D, Morel A, Gordon HR, Banzon VF, Evans RH (2005) Bridging ocean color
Martinez E, Antoine D, D’Ortenzio F, Gentili B (2009) Climate-driven basin-scale decadal oscillations of oceanic phytoplankton. Science, 326, 1253. Martinez E, Antoine D, D’Ortenzio F, de Boyer Montegut C (2011) Phytoplankton spring and fall blooms in the North Atlantic in the 1980s and 2000s. Journal of Geophysical Research, 116, C11029. McClain CRA (2009) Decade of satellite ocean color observations. Annual Review of Marine Science, 1, 19–42.
observations of the 1980s and 2000s in search of long-term trends. Journal of Geophysical Research, 110, C06009. Batten SD, Walne AW, Edwards M, Groom SB (2003) Phytoplankton biomass from continuous plankton recorder data: an assessment of the phytoplankton colour index. Journal of Plankton Research, 25, 697–702. Beaugrand G (2004) The North Sea regime shift: evidence, causes, mechanisms and consequences. Progress in Oceanography, 60, 245–262. Beaugrand G (2005) Monitoring pelagic ecosystems using plankton indicators. ICES Journal of Marine Sciences, 62, 333–338. Beaugrand G (2009) Decadal changes in climate and ecosystems in the North Atlantic Ocean and adjacent seas. Deep Sea Research Part II, 56, 659–673. Beaugrand G, Edwards M, Brander K, Luczak C, Ibanez F (2008) Causes and projections of abrupt climate-driven ecosystem shifts in the North Atlantic. Ecology Letters, 11, 1157–1168. Boyce DG, Lewis MR, Worm B (2010) Global phytoplankton decline over the past century. Nature, 466, 591–596. Chavez FP, Messie M, Pennington JT (2011) Marine primary production in relation to climate variability and change. Annual Review of Marine Science, 3, 227–260. Colebrook JM (1960) Continuous plankton records: methods of analysis, 1950–59. Bulletin of Marine Ecology, 5, 51–64. Doney SC (2006) Plankton in a warmer world. Nature, 444, 695–696. Edwards M, Richardson AJ (2004) Impact of climate change on marine pelagic phenology and trophic mismatch. Nature, 430, 881–884. Edwards M, Reid PC, Planque B (2001) Long-term and regional variability of phytoplankton biomass in the Northeast Atlantic (1960-1995). ICES Journal of Marine Sciences, 58, 39–49. Gregg WW, Conkright ME (2002) Decadal changes in global ocean chlorophyll. Geophysical Research Letters, 29, 1730. Hays GC, Richardson AJ, Robinson C (2005) Climate change and marine plankton. Trends in Ecology and Evolution, 20, 337–344. Henson S, Raitsos D, Dunne JP, McQuatters-Gollop A (2009a) Decadal variability in biogeochemical models: comparison with a 50-year ocean colour dataset. Geophysical Research Letters, 36, L21601. Henson S, Dunne JP, Sarmiento JL (2009b) Decadal variability in North Atlantic phytoplankton blooms. Journal of Geophysical Research, 114, C04013. Henson S, Lampitt R, Johns D (2012) Variability in phytoplankton community structure in response to the North Atlantic Oscillation and implications for organic carbon flux. Limnology and Oceanography, 57, 1591–1601. Hughes L (2000) Biological consequences of global warming: is the signal already apparent? Trends in Ecology and Evolution, 15, 56–61. IPCC [Intergovernmental Panel on Climate Change] (2007) Climate Change 2007: Impacts, Adaptation and Vulnerability. Cambridge University Press, Cambridge. Jones PD, Lister DH, Osborn TJ, Harpham C, Salmon M, Morice CP (2012) Hemispheric and large-scale land surface air temperature variations: an extensive revision and an update to 2010. Journal of Geophysical Research, 117, D05127.
© 2013 John Wiley & Sons Ltd, Global Change Biology, 20, 2117–2123
McQuatters-Gollop A, Raitsos DE, Edwards M, Pradhan Y, Mee LD, Lavender SJ, Attrill MJ (2007) A long-term chlorophyll dataset reveals a regime shift in North Sea phytoplankton biomass unconnected to increasing nutrient levels. Limnology and Oceanography, 52, 635–648. McQuatters-Gollop A et al. (2011) Is there a decline in marine phytoplankton? Nature, 472, E6. Platt T, Stuart V (1997) Sir Alister Hardy: an ocean colour pioneer (1896–1985). Backscatter, 8, 1. Pyper BJ, Peterman RM (1998) Comparison of methods to account for autocorrelation in correlation analyses of fish data. Canadian Journal of Fisheries and Aquatic Sciences, 55, 2127–2140. Racault MF, Le Quere C, Buitenhuis E et al. (2012) Phytoplankton phenology in the global ocean. Ecological Indicators, 14, 152–163. Raitsos DE, Reid PC, Lavender SJ, Edwards M, Richardson AJ (2005) Extending the SeaWiFS chlorophyll data set back 50 years in the northeast Atlantic. Geophysical Research Letters, 32, L06603. Raitsos DE, Lavender SJ, Maravelias CD, Haralabous J, McQuatters-Gollop A, Edwards M, Reid PC (2011) Macroscale factors affecting diatom abundance: a synergistic use of Continuous Plankton Recorder and satellite remote sensing data. International Journal of Remote Sensing, 32, 2081–2094. Raitsos DE, Walne A, Lavender SJ, Licandro P, Reid PC, Edwards M (2013) A 60 year ocean colour dataset from the Continuous Plankton Recorder. Journal of Plankton Research, 35, 158–164. Reid PC, Edwards M, Hunt HG, Warner A (1998) Phytoplankton change in the North Atlantic. Nature, 391, 546. Reid PC, Edwards M, Beaugrand G, Skogen M, Stevens D (2003) Periodic changes in the zooplankton of the North Sea during the twentieth century linked to oceanic inflow. Fisheries Oceanography, 12, 160–169. Richardson AJ, Schoeman DS (2004) Climate impact on plankton ecosystems in the northeast Atlantic. Science, 305, 1609–1612. Richardson AJ, Walne AW, John AWG, Jonas TD, Lindley JA, Sims DW, Witt M (2006) Using continuous plankton recorder data. Progress in Oceanography, 68, 27–74. Rodionov SNA (2004) Sequential algorithm for testing climate regime shifts. Geophysical Research Letters, 31, L09204. Wernand MR, van der Woerd HJ, Gieskes WWC (2013) Trends in ocean colour and chlorophyll concentration from 1889 to 2000 worldwide. PLoS ONE, 8, e63766.
