Surface ocean p CO 2 seasonality and sea-air CO 2 flux estimates for the North American east coast
2013; Wiley; Volume: 118; Issue: 10 Linguagem: Inglês
10.1002/jgrc.20369
ISSN2169-9291
AutoresSérgio R. Signorini, Antonio Mannino, Raymond G. Najjar, Marjorie A. M. Friedrichs, Wei‐Jun Cai, Joe Salisbury, Zhaohui Aleck Wang, Helmuth Thomas, Elizabeth H. Shadwick,
Tópico(s)Atmospheric and Environmental Gas Dynamics
ResumoJournal of Geophysical Research: OceansVolume 118, Issue 10 p. 5439-5460 Regular ArticleOpen Access Surface ocean pCO2 seasonality and sea-air CO2 flux estimates for the North American east coast Sergio R. Signorini, Corresponding Author Sergio R. Signorini Science Applications International Corporation, Washington, D.C., USACorresponding author: S. R. Signorini, NASA Goddard Space Flight Center, Code 616.2, Bldg 28/Rm W168, 8800 Greenbelt Rd., Greenbelt, MD 20771, USA. ([email protected] or [email protected])Search for more papers by this authorAntonio Mannino, Antonio Mannino NASA Goddard Space Flight Center, Ocean Ecology Laboratory, Greenbelt, Maryland, USASearch for more papers by this authorRaymond G. Najjar Jr., Raymond G. Najjar Jr. Department of Meteorology, The Pennsylvania, University Park, Pennsylvania, USASearch for more papers by this authorMarjorie A. M. Friedrichs, Marjorie A. M. Friedrichs Virginia Institute of Marine Science, College of William & Mary, Gloucester Point, Virginia, USASearch for more papers by this authorWei-Jun Cai, Wei-Jun Cai School of Marine Science and Policy, University of Delaware, Newark, Delaware, USASearch for more papers by this authorJoe Salisbury, Joe Salisbury Institute for the Study of Earth, Oceans and Space, of New Hampshire, Durham, New Hampshire, USASearch for more papers by this authorZhaohui Aleck Wang, Zhaohui Aleck Wang Department of Marine Chemistry & Geochemistry, Hole Oceanographic Institute, Woods Hole, Massachusetts, USASearch for more papers by this authorHelmuth Thomas, Helmuth Thomas Department of Oceanography, University, Halifax, Scotia, CanadaSearch for more papers by this authorElizabeth Shadwick, Elizabeth Shadwick Antarctic Climate and Ecosystems Cooperative Research Centre, University of Tasmania, Hobart, Tasmania, AustraliaSearch for more papers by this author Sergio R. Signorini, Corresponding Author Sergio R. Signorini Science Applications International Corporation, Washington, D.C., USACorresponding author: S. R. Signorini, NASA Goddard Space Flight Center, Code 616.2, Bldg 28/Rm W168, 8800 Greenbelt Rd., Greenbelt, MD 20771, USA. ([email protected] or [email protected])Search for more papers by this authorAntonio Mannino, Antonio Mannino NASA Goddard Space Flight Center, Ocean Ecology Laboratory, Greenbelt, Maryland, USASearch for more papers by this authorRaymond G. Najjar Jr., Raymond G. Najjar Jr. Department of Meteorology, The Pennsylvania, University Park, Pennsylvania, USASearch for more papers by this authorMarjorie A. M. Friedrichs, Marjorie A. M. Friedrichs Virginia Institute of Marine Science, College of William & Mary, Gloucester Point, Virginia, USASearch for more papers by this authorWei-Jun Cai, Wei-Jun Cai School of Marine Science and Policy, University of Delaware, Newark, Delaware, USASearch for more papers by this authorJoe Salisbury, Joe Salisbury Institute for the Study of Earth, Oceans and Space, of New Hampshire, Durham, New Hampshire, USASearch for more papers by this authorZhaohui Aleck Wang, Zhaohui Aleck Wang Department of Marine Chemistry & Geochemistry, Hole Oceanographic Institute, Woods Hole, Massachusetts, USASearch for more papers by this authorHelmuth Thomas, Helmuth Thomas Department of Oceanography, University, Halifax, Scotia, CanadaSearch for more papers by this authorElizabeth Shadwick, Elizabeth Shadwick Antarctic Climate and Ecosystems Cooperative Research Centre, University of Tasmania, Hobart, Tasmania, AustraliaSearch for more papers by this author First published: 29 August 2013 https://doi.org/10.1002/jgrc.20369Citations: 73AboutSectionsPDF ToolsRequest permissionExport citationAdd to favoritesTrack citation ShareShare Give accessShare full text accessShare full-text accessPlease review our Terms and Conditions of Use and check box below to share full-text version of article.I have read and accept the Wiley Online Library Terms and Conditions of UseShareable LinkUse the link below to share a full-text version of this article with your friends and colleagues. Learn more.Copy URL Abstract [1] Underway and in situ observations of surface ocean pCO2, combined with satellite data, were used to develop pCO2 regional algorithms to analyze the seasonal and interannual variability of surface ocean pCO2 and sea-air CO2 flux for five physically and biologically distinct regions of the eastern North American continental shelf: the South Atlantic Bight (SAB), the Mid-Atlantic Bight (MAB), the Gulf of Maine (GoM), Nantucket Shoals and Georges Bank (NS+GB), and the Scotian Shelf (SS). Temperature and dissolved inorganic carbon variability are the most influential factors driving the seasonality of pCO2. Estimates of the sea-air CO2 flux were derived from the available pCO2 data, as well as from the pCO2 reconstructed by the algorithm. Two different gas exchange parameterizations were used. The SS, GB+NS, MAB, and SAB regions are net sinks of atmospheric CO2 while the GoM is a weak source. The estimates vary depending on the use of surface ocean pCO2 from the data or algorithm, as well as with the use of the two different gas exchange parameterizations. Most of the regional estimates are in general agreement with previous studies when the range of uncertainty and interannual variability are taken into account. According to the algorithm, the average annual uptake of atmospheric CO2 by eastern North American continental shelf waters is found to be between −3.4 and −5.4 Tg C yr−1 (areal average of −0.7 to −1.0 mol CO2 m−2 yr−1) over the period 2003–2010. Key Points Seasonal and interannual variability of surface ocean pCO2 and sea-air CO2 flux N. Amer. east coast continental shelf is a sink of atm. CO2 at 3.4 to 5.4 TgC/yr Satellite-based pCO2 algorithms for the continental shelf 1. Introduction [2] Coastal oceans, despite covering a small fraction of the earth's surface, are important in the global carbon cycle because rates of carbon fixation, remineralization, and burial are much higher than the global average. A crucial difference between the coastal ocean and the open ocean is the proximity of sediments to the sea surface, providing a close coupling in space and time of the pelagic and benthic environments. Thus, the shallow water column in coastal regions constitutes a close link between surface sediments and the atmosphere allowing relatively direct interactions between both the sedimentary and atmospheric compartments [Borges et al., 2005; Thomas and Borges, 2012; Thomas et al., 2009; Thomas, 2004]. An additional characteristic of the coastal seas and continental shelves is the high temporal and spatial variability of CO2 fluxes [Borges et al., 2005, 2008; Cai et al., 2006; Frankignoulle and Borges, 2001; Shadwick et al., 2010, 2011]. The driving factors often vary within the system at seasonal time scales, and the deduction of general patterns remains difficult, typically requiring detailed case studies. [3] The work of Borges [2005] was the first to compile a global coastal shelf sea-air CO2 flux based on limited observed systems and using an upscaling scheme. Borges [2005] showed that the inclusion of the coastal ocean increases the estimates of CO2 uptake by the global ocean by 57% for high latitude areas, and by 15% for temperate latitude areas, while at subtropical and tropical latitudes the contribution from the coastal ocean increases the CO2 emission to the atmosphere from the global ocean by 13%. Cai et al. [2006] conducted a study of sea-air carbon exchange in ocean margins by grouping the numerous heterogeneous shelves into seven distinct provinces. Their results showed that the continental shelves are a sink of atmospheric CO2 at mid-high latitudes (−0.33 Pg C yr−1) and a source of CO2 at low latitudes (0.11 Pg C yr−1), with a net uptake of −0.22 Pg C yr−1. Laruelle et al. [2010] evaluated the exchange of CO2 between the atmosphere and the global coastal ocean from a compilation of sea-air CO2 fluxes scaled using a spatially explicit global typology of continental shelves. Their computed sink of atmospheric CO2 over the continental shelf areas (−0.21 ± 0.36 Pg C yr−1) is at the low end of the range of previous estimates (−0.22 to −1.00 Pg C yr−1). Laruelle et al. [2010] also concluded that the sea-air CO2 flux per surface area over continental shelves, −0.7 ± 1.2 mol CO2 m−2 yr−1, is twice the value of the open ocean based on the most recent CO2 climatology at the time. More recently [Cai, 2011] showed that the continental shelves are sinks of atmospheric CO2 (∼0.25 Pg C yr−1, but still with large uncertainty), accounting for ∼17% of open ocean CO2 uptake (−1.5 Pg C yr−1, Takahashi et al. [2009]). The largest uncertainty of these scaling approaches stems from the availability of CO2 data to describe the spatial variability, as well as to capture the relevant scales of temporal variability. [4] Given that relatively large amounts of carbon are exchanged via the sea-air interface in coastal seas and continental shelves, the knowledge of the seasonal and interannual variability of the sea-air CO2 flux in coastal oceans is a very important component of the carbon budget, which requires comprehensive regional studies. In general, the coastal ocean is characterized by a high variability in carbon cycling, which presents significant challenges in determining spatial and temporal integrals of relevant quantities, such as the sea-air CO2 flux. Therefore, innovative methods are needed for scaling up relatively sparse field measurements, in this case surface ocean pCO2, into the required temporal and spatial resolutions to effectively derive regional sea-air CO2 flux estimates. One method for obtaining such regionally integrated fluxes is through the use of biogeochemical circulation models, which can be evaluated using the sparse field measurements, and then used to compute the mean and variability associated with these regional fluxes [Hofmann et al., 2011]. Satellite data, because of their high temporal and spatial resolution, provide another very promising asset to accomplish this goal. For example, Lohrenz and Cai [2006] conducted a satellite ocean color assessment of sea-air fluxes of CO2 in the northern Gulf of Mexico. They used principal component analysis and multiple regression to relate the surface ocean pCO2 to SST, salinity, and chlorophyll and used retrieval of corresponding MODIS-Aqua products to assess the regional distributions of pCO2. [5] In this paper, we use multiple regression analysis to relate surface ocean pCO2 to environmental variables (SST, surface salinity, and chlorophyll) and use the resulting equations with inputs from corresponding satellite products to provide an assessment of the spatial and temporal variability of the surface ocean pCO2 and sea-air CO2 flux for the North American east coast. A brief description of the biological/physical setting of the study region is provided in section 2. The processing of in situ and satellite data sets and the development of regionally specific empirical pCO2 algorithms are described in section 3. The algorithm evaluation and the estimates of sea-air flux from the available pCO2 binned data and algorithm are provided in section 4, as well as a sensitivity analysis of parameters that influence the surface ocean pCO2 seasonal and interannual variability. Finally, we provide a summary and discussion of suggested future work in section 5. 2. Physical and Biological Setting [6] The temporal and spatial variability of the surface ocean pCO2 on continental shelves are influenced by a combination of physical and biogeochemical factors, including surface temperature-driven solubility, biological processes, fall-to-winter vertical mixing, ocean circulation, river runoff, and shelf-ocean exchange [Wang et al., 2013]. Here we provide a summary of the physical and biological factors that are potentially important in shaping the pCO2 variability in the North American east coast continental shelf. [7] The definition of the coastal ocean is elusive, as it can be related to bathymetry, hydrography, or distance from shore; and some features, such as river plumes and coastal biomass maxima, can be ephemeral. Community efforts to standardize this definition to a fixed distance from shore, such as Hales et al. [2008] as adopted by the Surface Ocean CO2 Atlas (SOCAT; http://www.socat.info/), extend seaward from the North American continent beyond what we feel represents the reach of coastal processes. As a result, we have used the outer boundaries of the regions defined by Hofman et al. [2008, 2011] to define the extent of the coastal ocean. The North American east coast (Figure 1) encompasses three large regions of diverse physical and biological characteristics: the southeast U.S. continental shelf, also known as the South Atlantic Bight (SAB), the northeast U.S. continental shelf, and the Scotian Shelf (SS). Within the northeast U.S. continental shelf there are four subregions: the Middle Atlantic Bight (MAB), Georges Bank (GB), Nantucket Shoals (NS), and the Gulf of Maine (GoM). For this study, we combined the GB and NS regions into a single region (GB+NS) for simplicity and because these two regions share many similar physical and biogeochemical attributes [Fox et al., 2005; Shearman and Lentz, 2004; Thomas et al., 2003]. These North American continental shelf subregions are defined in Hofmann et al. [2011] with the GB+NS region separated from the GoM as in Hofmann et al. [2008]. The 58 coastal subregions shown in Hofmann et al. [2008] were developed based on a combination of bathymetry, SST fronts, stratification, and biological properties. For simplicity, here we consolidate the very fine regional domains into five major subregions described above. However, we recognize that previous studies have adopted other methods to identify regional domains [Hales et al., 2008, 2012]. For example, a self-organizing mapping method has been adopted to subregionalize the North American Pacific Coast [Hales et al., 2012]. The method relies on an artificial neural network to identify biogeochemical regions within the target study area. Figure 1Open in figure viewerPowerPoint Regional domains for analysis adapted from Hofmann et al. [2008] and Hofmann et al [2011]. The white circles show the locations of the NDBC buoys within each regional domain. The white star shows the location of the Sable Island meteorological station and the white square the location of the Carioca buoy. [8] Our focus is on the continental shelf that we operationally define as depths less than 200 m since the depth of the actual shelf break varies. Bathymetric variation in our study area is large. Portions of GB and NS are only several meters below the sea surface, whereas in the GoM and areas of the SS, water depths exceed 200 m. Our study area is also at the “crossroads” of the north flowing Gulf Stream and the southwest flowing slope water-Labrador current [Rossby, 1987]. Chapman and Beardsley [1989] suggest that glacial melt and runoff from Western Greenland generates a buoyancy-driven coastal current that flows over the SS and GB and eventually into the MAB. This coastal current is an important driver to the distribution of the marine CO2 system, including surface pCO2 along its flow path [Wang et al., 2013], i.e., the Gulf of St Laurence, the SS, the GoM, and the MAB. There is little exchange of water between the MAB and SAB along the narrow shelf at Cape Hatteras. In the SAB, the Gulf Stream is close to the shelf break and has a direct influence on the outer SAB shelf [Signorini and McClain, 2007], readily identifiable by the warm and salty signature shown in seasonal maps of sea surface temperature (SST), sea surface salinity (SSS), and chlorophyll (Chl) of Figure 2 (see section 3 for methodology), whereas north of Cape Hatteras, the influence of the Gulf Stream is more indirect. Here anticyclonic warm core rings result from landward meanders of the Gulf Stream [Joyce et al., 1992]. The rings are carried in the southwestward flow of slope water where they interact with the outer shelf from GB to Cape Hatteras, frequently entraining phytoplankton-rich shelf water [Joyce et al., 1992]. Near Cape Hatteras, the warm core rings may be reabsorbed into the Gulf Stream, a process readily apparent in daily time series animations of chlorophyll (Chl) and SST. In the SAB, the outer shelf waters are warmer (Figure 2) in summer and autumn than winter and spring due, in part, to the proximity of the Gulf Stream as a result of the expansion of the subtropical gyre [Signorini and McClain, 2007]. Figure 2Open in figure viewerPowerPoint Seasonal climatology maps of SST, SSS, and Chl. Upper row: SST composites from MODIS Aqua; middle row: SSS composites from World Ocean Data 2009; bottom row: Chl composites from MODIS Aqua. Refer to section 3 for details. The MODIS SST and Chl seasonal climatologies are based on the period 2002–2011. The seasons are defined as Dec-Jan-Feb (DJF), Mar-Apr-May (MAM), Jun-Jul-Aug (JJA), and Sep-Oct-Nov (SON). [9] The pCO2 variability in riverine-plume systems is a result of complex biogeochemical interactions. In the Gulf of Maine for instance, labile riverine carbon is responsible for sustaining supersaturated pCO2 conditions in late fall, while at other times of the year phytoplankton productivity, most likely driven by inputs of riverine dissolved inorganic nitrogen, is responsible for pCO2 undersaturation [Salisbury et al., 2008]. The North American east coast continental shelf is influenced by the discharge of several major rivers and estuaries (Chesapeake Bay, Delaware Bay, and Gulf of St Lawrence, for example) that contribute to complex physical and biogeochemical interactions that influence the seasonal and interannual variability of the surface ocean pCO2, an important parameter for the determination of the sea-air CO2 flux. Vandemark et al. [2011] showed that the observed pCO2 and CO2 flux dynamics in the Gulf of Maine are dominated by a seasonal cycle, with a large spring influx of CO2 and fall-to-winter efflux back to the atmosphere. They also showed that in the western Gulf of Maine the ocean is a net source of carbon to the atmosphere (+0.38 mol CO2 m−2 yr−1) over a period of 5 years, but with a moderate interannual variation where years 2005 and 2007 represent cases of regional source (+0.71 mol CO2 m−2 yr−1) and sink (−0.11 mol CO2 m−2 yr−1) anomalies, respectively. Comparison of results with the neighboring Middle Atlantic and South Atlantic Bight shelf systems showed that the Gulf of Maine differs by enhanced pCO2 control factors other than temperature-driven solubility, such as biological drawdown, fall-to-winter vertical mixing, and river runoff [Salisbury et al., 2008; Shadwick et al., 2010]. [10] Shadwick et al. [2011] investigated the seasonal variability of pCO2 in the Scotian Shelf and concluded that the region acts as a net source of CO2 to the atmosphere on an annual basis (1.4 mol CO2 m−2 yr−1). On a seasonal basis, there is a reversal of the flux only when a pronounced undersaturation of surface waters is reached for a short period during the spring bloom. Outside of the spring bloom period, the competing effects of temperature and biology influence on surface pCO2 are nearly equal and opposite. DeGrandpre et al. [2002], based on measurements of surface ocean pCO2 during the Ocean Margins Program [Verity et al., 2002], concluded that the MAB is a sink of atmospheric CO2 with an annual mean of −1.0 ± 0.6 Tg C yr−1, or an area average of −1.1 ± 0.7 mol CO2 m−2 yr−1. A significant portion of this atmospheric uptake is a result of the annual cycle of heating and cooling combined with strong winds during the winter undersaturation period. [11] Jiang et al. [2008] showed that on an annual basis the SAB is a relatively small net sink of atmospheric CO2 (−0.48 ± 0.21 mol CO2 m−1 yr−1). Seasonally, the SAB shifts from a sink of atmospheric CO2 in winter to a source in summer. The annual cycle of sea surface temperature plays an important role in controlling the seasonal variation of pCO2. The combination of stronger wind speeds during fall winter, when CO2 undersaturation is significant due to lower SSTs, results in a net annual CO2 sink. Other important factors controlling the pCO2 variability in the SAB are the marsh export of organic carbon and dissolved inorganic carbon (DIC) in the warm months (June-November), which directly supports CO2 outgassing in these months via organic carbon decomposition and increase in DIC [Jiang et al., 2013; Wang et al., 2005]. In addition, the marsh areas in the SAB also export alkalinity, another important factor influencing the variability of pCO2 and sea-air flux [Wang et al., 2005; Wang and Cai, 2004]. [12] The seasonal Chl climatology from MODIS Aqua (Figure 2) shows that the maximum Chl in the GoM, GB, and NS occurs during spring (March-April-May, MAM). The GB region has the highest Chl in spring, but it is maintained at concentrations above 2.5 mg m−3 in all seasons due to vigorous tidal mixing. Figure 2 also shows that the low-salinity nearshore waters along the entire east coast coincide with regions of elevated Chl, an indication of the influence of nutrient-rich riverine waters. On the MAB shelf, there is a high-Chl region during winter (December-January-February, DJF) in the nearshore and outer-shelf waters, but the fall bloom (SON) dominates between approximately the 40 and 60 m isobaths. The high satellite-derived “Chl” in winter may be in part colored dissolved organic matter flowing out from rivers, plus photoacclimation by phytoplankton (higher Chl-a due to low surface solar radiation and a well-mixed water column). [13] The minimum surface Chl over much of the MAB occurs during summer (JJA) when highest SST (Figure 2), peak stratification and a pronounced subsurface Chl maximum layer occur [O'Reilly and Zetlin, 1998]. Summer mixed-layer depths of ∼3.5 to 10 m are typical for MAB shelf waters. The spring bloom (MAM) is clearly shown by the elevated Chl concentrations in the MAB, GB, and GoM (Figure 2). Figure 2 also shows that the SAB Chl has its largest changes in the outer shelf, with a maximum in DJF and lowest values in JJA under the influence of the oligotrophic waters of the Gulf Stream. 3. Data Sets and Methods Processing of In Situ and Satellite Data Sets [14] The surface ocean pCO2 data are obtained from SOCAT, combined with additional available data from regionally specific field experiments (see Appendix A) and binned by month for each year (1978–2010) into 0.15° × 0.15° grid cells. The SOCAT data [Pfeil et al., 2012] holds 6.3 million quality-controlled surface ocean pCO2 from the global oceans and coastal seas covering the period of 1968–2007. These data were put together following uniform format and a strict protocol that included quality control with clearly defined criteria performed by a team of international experts. [15] The MatLab function bin2d, developed by J. Nielsen and available at the Nansen Environmental and Remote Sensing Center (NERSC) from www-2.nersc.no/∼even/, was used to bin all data sets into the study grid. First, all the available data within 24°N to 46°N and 82°W to 56°W were selected for binning. These included 416,261 colocated surface ocean pCO2, SST, and sea surface salinity (SSS) values from SOCAT from the period 1978–2007, 11,628 from the 2006 SAB cruise (only 2005 cruises are included in SOCAT), and 309,665 from the GoM (2004–2010). The binned pCO2 data were then adjusted to reference year 2004 using an atmospheric growth rate of 1.68 µatm yr−1 [Le Quéré et al., 2010] and assuming that the surface ocean pCO2 is trending at the same pace as the atmosphere. All the adjusted pCO2 data were then binned into 12 individual calendar months, each containing the average of all data within a particular month and grid bin. The data were then divided into regional study domains following the boundaries shown in Figure 1. [16] The available pCO2 data were divided into two individual sets, one dedicated to algorithm development (data bins covering more than 6 months) and one dedicated to algorithm evaluation (data bins covering less than 6 months). Surface ocean pCO2 data from underway (UW) transects across the Scotian Shelf and pCO2 time series from the CARIOCA buoy located at 44.296°N and 63.257°W [Shadwick et al., 2010] were also used for algorithm evaluation, together with SOCAT data on the Scotian Shelf not used for the algorithm development. Figure 3a shows color-coded SOCAT surface ocean pCO2 cruise tracks and Figure 3b shows corresponding coastal binned data with associated color-coded temporal coverage in months. The highest temporal coverage corresponds to the most traveled routes (in orange to red), i.e., most frequent destination ports (Boston, New York, Norfolk, Miami) used by the Volunteering Observing Ships (VOSs). The VOS ships according to map available at the CDIAC web site (http://cdiac.ornl.gov/oceans/VOS_Program/) are: the Skogafoss, A. Companion, Oleander, Falstaff, and Explorer of the Seas. The SOCAT data set also includes transects occupied by research vessels. Figure 3 clearly shows that the surface ocean pCO2 data have spatial and temporal distribution gaps that may be potentially responsible for biases in the calculation of sea-air fluxes. Figure 3Open in figure viewerPowerPoint (a) Color-coded SOCAT surface ocean pCO2 cruise tracks and (b) corresponding coastal binned data with associated color-coded temporal coverage in months. The highest temporal coverage corresponds to the most traveled routes (in orange to red), i.e., most frequent destination ports (Boston, New York, Norfolk, Miami) used by the Volunteering Observing Ships. The SOCAT data set also includes transects occupied by research vessels. The SS, GoM, GB+NS, MAB, and SAB regional boundaries are overlaid as black lines. [17] Monthly sea-surface salinity (SSS) climatology was interpolated and gridded onto the 0.15° × 0.15° study domain grid using the World Ocean Database (WOD) 2009 station data and the method of Kriging. The Interactive Data Language (IDL) function KRIG2D was used for this purpose. Monthly climatologic mixed layer depth (MLD) was derived from WOD 2005 for the entire East Coast based on temperature profiles using 0.5°C temperature difference criterion [Hofmann et al., 2008]. The MLD data were binned into the same 0.15° × 0.15° study domain grid. [18] Both data and algorithm sea-air CO2 flux estimates were obtained using gridded (0.25° × 0.25°) winds from the Jet Propulsion Laboratory Cross-Calibrated Multiple Platforms (CCMP) [Atlas et al., 2011] product (ftp://podaac-ftp.jpl.nasa.gov/allData/ccmp/L2.5/flk). Monthly wind climatology was derived using data from 1999 to 2008, a period approximately centered on 2004, the reference year adopted for the adjusted surface ocean pCO2 data. The climatologic and interannual CCMP monthly winds were regridded (0.15° × 0.15°) and extrapolated nearshore using the function “surface” from Generic Mapping Tools (GMT) [Smith and Wessel, 1990; Wessel and Smith, 1991], which is based on an adjustable tension continuous curvature surface gridding method. High-frequency (10 min) winds from 10 NOAA National Oceanographic Data Center NDBC buoys (http://www.nodc.noaa.gov/BUOY/) and hourly winds from Sable Island were used to obtain correction coefficients to account for nonlinearities in the gas exchange parameterization resulting from the use of monthly climatologic winds. The method for deriving these coefficients is described under section 3.3. [19] All parameters used to develop the pCO2 algorithm and to derive the sea-air CO2 flux, including all satellite data products (SST and Chl), SSS, and the CCMP wind speed, were also binned monthly into the same grid. The satellite data products consisted of 9 km, level 3 mapped, MODIS Aqua (MODISA) climatologic and interannual monthly composites of SST and Chl obtained from the NASA ocean color distribution archive (http://oceancolor.gsfc.nasa.gov/). A validation between log-transformed MODISA Chl retrievals versus all available in situ observations (SAB to GoM, depth<=200 m, N = 404), conducted using the SeaBASS (SeaWiFS Bio-optical Archive and Storage System: http://seabass.gsfc.nasa.gov/) data search and validation tools, showed good matchup agreement (r2 = 0.75, RMSE = 0.30, APD = 35.8%). For the algorithm development, we used the available binned surface ocean pCO2, SST, and SSS derived from the in situ data, combined with monthly climatologic satellite Chl binned at the same grid points as no in situ concurrent Chl measurements are available. For the algorithm application, we used monthly interannual (2003–2010) satellite SST and Chl, and monthly climatologic SSS derived from WOD 2005 data. [20] Seasonal maps were constructed by averaging the monthly data and derived products into four 3 month composites, defined as: winter (December-January-February, DJF), spring (March-April-May, MAM), summer (June-July-August, JJA), and autumn (September-October-November, SON). Development of Regional pCO2 Algorithms [21] The algorithm development is based on binned in situ pCO2, SST, and SSS, and satellite-derived Chl monthly climatology, as well as day of the year (Julian day). The algorithm was developed through the multiple linear regression (MLR) analysis based on all spatial bins containing more than six available monthly occurrences of the in situ data (remaining data were reserved for evaluation), and is represented as (1) [22] The first terms in brackets represent the surface ocean pCO2 corrected to the year 2004 and the last term is a correction factor for different years to account for the rise of surface ocean pCO2 due to the uptake of anthropogenic CO2. The input for “Day” (Julian day) was normalized sinus
Referência(s)