Biggs, Hu, and Müller-Karger, revised for Deep-Sea Res II topical issue on DGoMB program

Remotely sensed sea surface chlorophyll and calculated [c1]POC flux at Deep Gulf of Mexico Benthos sampling stations

DOUGLAS C. BIGGS,* CHUANMIN HU† and FRANK E. MüLLER-KARGER†

(contributed 15 September 2005; revised 9 Nov 2006)

Table 1: longitude should be negative

Fig 4 caption: should be “January 1998 to December 2000” instead of “December 1998 to January 2000”

Fig 4 caption typo: “East of 91” instead of “ast of 91

Abstract--We used SeaWiFS ocean color data were used to make biweekly composite averages of the standing stock of sea surface chlorophyll (SSC) at 44 stations along the continental slope and rise that were the focus of May-June 2000 benthic sampling by the Deep Gulf of Mexico Benthos (DGoMB) program. At the 22 DGoMB sites north of 26oN and west of 91oW in the NW Gulf, the annual average remotely sensed SSC was 0.19 mg m-3, ranging at most locations from annual highs of about 0.3 mg m-3 in November-February to annual lows of about 0.1 mg m-3 in May-August. Comparison of three years of SeaWiFS data between Jan 1998 and– Dec 2000 showed little inter-annual variation in this pattern at these NW Gulf stations. In contrast, at the 22 NE Gulf sites north of 26oN and east of 91oW, SSC was more inter-annually variable and averaged 2.8 times higher than in the NW Gulf. Maxima in the eastern region occurred in November-February and also during summers, when eddies in the NE Gulf entrained and transported Mississippi River water east- and southward, off the margin. This summertime cross-margin transport led to apparent increases in SSC in June-August at 9 of the 22 NE Gulf stations, reaching average monthly concentrations >50% greater than in November-February. Based on a primary productivity model and a vertical flux model, tThe calculated export of particulate organic carbon (POC flux reaching the seafloor) averaged 18 mg C m-2 day-1 at the 22 NE Gulf stations, and 9 mg C m-2 day-1 at the 22 NW Gulf stations. These estimates are comparable to those measured [c2] by benthic lander So on average twice as much POC should be reaching the benthos in the NE Gulf than in the NW Gulf.

.

*Department of Oceanography, Texas A&M University, College Station, TX 77843, U.S.A.

†College of Marine Science, University of South Florida, St Petersburg, FL 33701, U.S.A.

1INTRODUCTION

Phytoplankton pigment concentration in surface waters of the Gulf of Mexico (GOM) undergoes a well-defined seasonal cycle which is generally synchronous off the shelf, where the bottom is deeper than 200 m. Müller-Karger et al. (1991) reviewed monthly climatologies of remotely-sensed near-surface phytoplankton pigment concentration (chlorophyll + phaeopigment) from multiyear series of Coastal Zone Color Scanner (CZCS) satellite images for the period 1978-1985. Using two representative 200x200 km2 sampling areas located in the southern GOM (one centered at 24°N, 86°W; the other at 25°N, 93°W), they reported that highest near-surface pigment concentration in deep waters occurs between December and February and lowest values occur between May and July. In these deep GOM waters there was only about a 3-fold variation between lowest (~0.06 mg m-3) and highest (0.2 mg m-3) near-surface pigment concentration.

Modern ocean color sensors such as the Sea-viewing Wide-Field-of-View Sensor (SeaWiFS) provide more accurate and frequent observations of the biological state of the Gulf of Mexico waters (McClain et al., 2004), thanks to improvements in technology (sensors with more spectral bands (wavelengths) and higher sensitivity) and in algorithms (atmospheric correction and bio-optical inversion). Our task objective for the Deep Gulf of Mexico Benthos (DGoMB) program was to provide a remote-sensing perspective of the potential food for the benthos, so we used SeaWiFS data to revisit the seasonal variations, and to look at detail in spatial variations in near-surface phytoplankton pigment concentrations among 44 DGoMB stations located north of 26oN and ranging in water depth from 213 m to 3146 m.

As Rowe et al (this volume) explain in their Introduction to this special issue, among the null hypotheses that DGoMB fieldwork sought to test were a) there are west-to-east differences, and b) there are nearshore-to-offshore differences, in benthic community standing stocks and species diversity. So for this paper we have summarized the west-to-east, nearshore-to-offshore, seasonal, and inter-annual variations in remotely-sensed phytoplankton pigment concentration at the DGoMB stations. We have also calculated the remotely sensed primary production (mg C m-2 day-1) and particulate organic carbon export (POC flux) that reaches the seabed (mg C m-2 day-1) by using the standard methods of Behrenfeld and Falkowski (1977) and Pace et al (1987), respectively.

2 METHODS

In early September 1997, the SeaWiFS sensor on the SeaStar satellite (Orbimage Corporation) began collecting ocean color data, providing global coverage every one to two days. The Institute for Marine Remote Sensing at the College of Marine Science at the University of South Florida (USF) downlinked, processed, and archived the regional SeaWiFS data covering the Gulf of Mexico. , and we derived weekly, biweekly, monthly, and annual composite averages of the remotely-sensed data. The SeaWiFS data/imagery provided an estimate of the surface (to one optical depth or about 30-50 m in clear water and shallower in more turbid water) chlorophyll concentration (SSC), and provided an effective means to trace the circulation, the dispersal of riverine waters, and the location of oceanographic fronts. Based on the daily observations, we derived weekly, biweekly, monthly, and annual composite averages.

In support of DGoMB oceanographic habitat characterization, we used biweekly composites for the three year period between January 1998 and– December 2000 to estimate compute the annual mean SSC at each DGoMB station, and its inter-annual variability. Table 1 gives location and water depth at each of the 44 DGoMB stations north of 26oN at which benthic sampling was carried out in May-June 2000, and shows how we have divided these stations into 22 in the NW Gulf (north of 26oN and west of 91oW) and 22 in the NE Gulf ((north of 26oN and east of 91oW).

Prior to beginning DGoMB fieldwork in May 2000, the mean Gulf of Mexico basin-wide distribution of surface SSC was calculated using the available SeaWiFS ocean color data from January 1998 through December 1999 (Biggs and Ressler, 2001). As shown in Figure 1 (color plate in this article), we have extended the data through December 2000, and then summarized the three-year mean SSC distribution in the northern GOM separately for winter (21 December – 20 March) and summer (21 June – 20 September) seasons.

For Figure 1 and for the additional characterization of DGoMB oceanographic habitat that we present in this paper, the SeaWiFS data were processed at high resolution (~ 1 km pixel size 1.0 km x 0.7 km) with the SeaWiFS Data Analysis System (SeaDAS version 2). More recent reprocessing of the SeaWiFS data (versions 3 and to 54) are now available using updated and improved calibration and data processing algorithms (atmospheric correction and bio-optical). The reprocessed data show improved SSC retrievals in coastal waters, but not in the deep waters such as those studied in the DGoMB program. Comparisons of SeaWiFS version 2 data with field measurements in 1998 and 1999 show that over the continental slope, uncertainties were less than 35%, except in river plume waters (Hu et al., 2003).

SSC around each DGoMB location was obtained by averaging over a box of variable size to help remove residual digitization-noise errors (Hu et al., 2001). All invalid pixels (as defined by the various SeaWiFS flags, such as cloud and stray light, large solar or viewing angle, etc.) were discarded. The choice of the size of the averaging box, either 3 x 3, 5 x 5, or 9 x 9, was a compromise between the number of valid pixels and the standard deviation found within the box.

Even over a two-week averaging period, a 3 x 3 pixel grid centered on the location of each DGoMB station often contained too few valid pixels due to cloud cover and cloud adjacency contamination effects. Therefore, a 5 x 5 pixel grid (n = 25) was used to derive the biweekly mean SSC concentration time series. The coefficient of variation (CV = standard deviation to mean ratio) was about 2-fold higher for NE Gulf stations than for NW Gulf stations. The average CV was about 9% (0.087) for the group of 22 DGoMB stations west of 91oW (stations along RW and W transects and AC, WC, B, and NB sites over the Louisiana continental margin), and 18% (0.185) for the 22 stations east of 91oW (stations along C, S, and MT transects, including Hi-Pro station).

Because the year-to-year variability in annual averages of SSC at DGoMB stations in the NE Gulf was higher than at those in the NW Gulf, the 1998-2000 time series was extended by adding data for the subsequent year 2001 to produce the summary of month-by-month variability for the 22 DGoMB stations in the NE Gulf that is reported inas Table 2. Each of the monthly-composite means for the NE Gulf was centered on the middle date for the month (Julian Days 15, 46, 74, 105, 135, 166, 196, 227, 258, 288, 319, and 349). Four such monthly composite means, from four different years, are shown as Figure 2 (color plate).

Weekly and biweekly composites of SeaWiFS data for the entire Gulf of Mexico are archived as png-format files that cover the 7-year period October 1997 – September 2004, on a CD-ROM appendix to a recent paper by Biggs et al (2005). Copies may be obtained from the American Geophysical Union, Geophysical Monographs division.

We followed the methodology outlined by Müller-Karger et al (2005) to compute monthly mean net primary production (NPP) using the Vertically Generalized Production Model (VGPM) [Behrenfeld and Falkowski, 1997]. The model estimates depth-integrated NPP (mg C m-2 day-1) based on remotely sensed SSC (mg m-3), surface PAR (obtained from SeaWiFS data from the NASA Goddard Space Flight Center), and sea surface temperature (SST, obtained from the advanced very high resolution radiometer or AVHRR). Sinking POC flux was in turn computed from NPP using an exponential decay model (flux(Z) = 3.523*NPP*Z-0.734) [Pace et al, 1987].

3 RESULTS

Table 1 summarizes the annual averages of SSC at NW versus NE Gulf DGoMB stations, as computed from the biweekly composite data for the 3-year period 1998-2000. The overall 3-year average SSC for the 22 stations in the NE Gulf was 0.54 mg m-3, or nearly three times greater than that for the 22 stations in the NW Gulf (0.19 mg m-3). The standard deviation about the annual mean at stations in the NE Gulf is also higher than at those in the NW Gulf. In both regions, however, the locally highest SSC concentrations were generally found at stations on the upper slope (water depths < 500 m) and locally lowest concentrations were generally found in deepest water. A negative exponential fit to the 3-year average SSC versus depth data removed 54% of the variance in SSC versus water depth at NW Gulf stations, and 37% of the variance at NE Gulf stations.

Figure 3 illustrates the seasonal cycle of SSC at the nine deepest stations in the NW Gulf (AC1, RW6, W5, W6, NB4, B2, B3, B1, and NB5). Water depth at each of these nine stations was greater than 2 km. Panel A in Figure 3 shows the three-year time series of biweekly SSC at each of the nine stations for the three years 1998-2000. There was relatively little year-to-year variability in the amplitude or phase of the annual maximum or minimum SSC. At most of the locations, SSC ranged from about 0.3 mg m-3 in November-February to minima of about 0.1 mg m-3 in May-August. Panel B of Figure 3 shows the climatological (three-year) SSC concentration (26 biweekly periods) at each of the 9 stations. These results agree with the pattern previously reported from CZCS data (Müller-Karger et al., 1991; Melo-Gonzalez et al., 2000): Deepwater SSC is lowest in spring-summer (Julian Days 100-250) and highest Nov-Feb (Julian Days 330-060).

Additional analyses not shown here confirmed that the annual cycle of SSC at most of the Far Western stations (RW1-RW6), Western Stations (W1-W6), and Louisiana Slope Stations was similar to the cycle at the nine NW Gulf deepwater stations shown above (see the figures in DGoMB interim technical reports edited by Rowe and Kennicutt, 2001 and 2002). Only at the shallowest stations (water depths < 1 km) did biweekly SSC exceed 0.4 mg m-3 in November-February.

East of 91oW, the "typical" deepwater annual cycle in SSC not only averaged more than two-fold higher but was frequently punctuated by times when SSC reached greater than average conditions for a particular season, especially during summertime (Figure 4). Table 2 shows that 9 of the 22 DGoMB stations in the NE Gulf had SSC in June, July, or August that was 50% or more greater than the average SSC for the period November-February. In particular, high summertime SSC levels occurred at MT3, MT4, HiPro, S37, S36, S35, S41, S42, and S43.

Table 3 summarizes the calculated remotely-sensed primary production and POC flux to the seabed. At NW GOM stations west of 91oW, calculated primary production averaged 0.4 g C m-2 day-1 and calculated POC flux to the seabed averaged 9 mg C m-2 day-1. At NEGOM stations east of 91oW, calculated primary production averaged 0.7 g C m-2 day-1 and calculated POC flux to the seabed averaged 18 mg C m-2 day-1. Because Table 3 shows that the annual mean SST averaged 26oC in both regions, the average two-fold differences between NW GOM and NE GOM stations are not due to Q10[c3] effects. Rather, the two-fold higher productivity and POC flux in the eastern region appears to be driven by greater off-margin export of high SSC coastal water to the eastern region, especially during summers when eddies in the NE Gulf entrained and transported Mississippi River water east- and southward, off the margin. At five stations in very blue water where DGoMB cruises did some limited box-coring and trawling in the Mexican EEZ south of 26oN, the calculated primary production averaged only about 50% that in the NE Gulf, and POC flux to the deep abyssal plain there averaged only 3 mg C m-2 day-1 (Table 3).

4 DISCUSSION

Temporal and spatial variations in SSC

The annual cycle of SSC at DGoMB stations in the NW Gulf and at the deepest stations (water depth > 2 km) in the NE Gulf showed a maximum in winter, and a minimum in summer, as previously reported (Müller-Karger et al., 1991; Melo-Gonzalez et al., 2000). At DGoMB stations in water depths of 300-1800 m and east of 91oW, however, the typical “deepwater” annual cycle in SSC was punctuated by high summertime SSC.

Remote-sensing altimetry data and at sea-truth physical oceanographic data collected in support of the DGoMB program (Jochens and DiMarco, this volume) and in support a companion field program (the Northeastern GOM Chemical Oceanography and Hydrography program, or NEGOM) explain why there was high summertime SSC in the NE Gulf. The sea surface height (SSH) altimetry data showed that in summers 1998, 1999, and 2000, warm slope eddies (WSEs) were centered over the deepwater of the DeSoto Canyon (Muller-Karger, 2000; Hu et al., 2003; Belabbassi et al., 2005; Jochens and DiMarco, this volume). A large WSE that was centered south of 28oN in summer 1999, and similar but smaller WSEs located farther north on the slope in summers 1998 and 2000, each acted to entrain low salinity, high SSC “green water” from the Mississippi River, and to transport this turbid plume seaward in July and August of each of these years. A more recent study (Hu et al., 2005) showed that such entrained low-salinity water could be transported to the Atlantic through the through the Florida Current and Gulf Stream.

High performance liquid chromatography (HPLC) analyses of the phytoplankton pigments collected from ship in these low salinity plumes during the NEGOM cruises (Qian et al., 2003) confirmed that the low salinity Mississippi plume water had locally high SSC standing stocks. These were important ground-truth data needed to understand the optical signal detected by satellite over the Mississippi River plume, since waters of various river plumes frequently absorb strongly in the blue wavelengths due to high concentrations of colored dissolved organic matter (CDOM).

CDOM fluorescence and light absorption coefficient were also measured along with surface salinity and SSC from ship during the NEGOM cruises. Strong correlations between the CDOM absorption coefficient at 443 nm (ag443, m-1) and CDOM fluorescence (330 nm excitation and 450 nm emission filters) were consistently observed (Hu et al., 2003; Nababan, 2005). Relatively high ag443 (0.1 m-1) was generally observed over the inner shelf, near the major river mouths. The ag443 signal decreased rapidly with distance from shore, except when riverine waters were entrained and transported offshore by WSEs. Because CDOM as well as SSC contributes to the ocean color signal measured by SeaWiFS, we likely over-estimated SSC at some of the shallowest DGoMB stations in the NE Gulf (Hu et al, 2003). Specifically, SSC > 4 mg m-3 shown in Figure 4 for stations MT1 and MT2 close off the Mississippi River delta and at HighPro station may be overestimated, by perhaps 50% - 100%. However, the relative synoptic patterns as well as the relative temporal variation patterns should remain valid.

The statistics (standard deviations, CVs) on the seasonal cycle and variability in SSC which we have presented in (Tables 1 and 2) should be considered only in a relative way for a number of reasons. Specifically, SSC estimates for deep stations without river plume interference have high accuracy (Hu et al., 2003), but SSC in shallow water coastal stations (depth < 30 m) and in river plumes will be overestimated. Statistics were derived from two-week and one month composites, not from the original daily data which often has significant cloud cover. For a pixel with 14 valid data values from the 2-week period, only the average value is used in the composite data, and its weight in the statistics is one. i.e., the same as another pixel with only 1 valid data value from the 2-week period.