5.1. Seasonal Bio-Thermal Patterns
 Our inference of the Gulf's seasonal bio-thermal cycle from these satellite data is consistent with previous studies of this basin [Walsh et al., 1989; Muller-Karger et al., 1991]. The period of winter mixing and thermal energy loss from the surface ocean to the atmosphere is accompanied by an increase in near-surface chlorophyll concentration. The mechanistic hypothesis is that light penetration is sufficiently deep and mixing depth, although increased, is nonetheless still sufficiently shallow such that light-limitation of the phytoplankton does not generally occur; phytoplankton may immediately assimilate nutrients entrained into the surface mixed layer. Thermal stratification then occurs during the spring transition to net upper ocean heating, and this process restricts the surface mixed layer from deep-ocean nutrients. The resulting high-light, low-nutrient surface environment is then characterized by a decline in surface chlorophyll concentration.
 The comparison of seasonal chlorophyll concentration trends for the deep Gulf with the QAA-estimated phytoplankton absorption coefficients further suggests an additional change in the phytoplankton absorption efficiency (a*) that may result from a shift in the surface phytoplankton community composition. This satellite-based inference is consistent with in situ observations. Qian et al.  examined phytoplankton pigment distributions between the Mississippi River and Tampa Bay. While their study was largely confined to the continental shelf, these authors report that the proportional prokaryote abundance on the outer shelf and slope stations declined during the winter cruises. Furthermore, analysis of chlorophyll fractionation, phytoplankton species counts, and pigment biomarkers made during the 1992–1994 Louisiana-Texas Shelf Circulation and Transport Process Study confirm that pico- and nanophytoplankton (<20 μm cell diameter) comprised the majority of deep water phytoplankton, except during the winter cruise when large diatoms were instead abundant [Lambert et al., 1999; Biggs and Ressler, 2001, and references therein].
 In addition to phytoplankton absorption trends, a strong seasonal correspondence was identified between the satellite estimates of %CDM, the CDM absorption coefficient, and surface chlorophyll. This finding is of particular importance because the processes constraining chlorophyll concentration may not be the same as those constraining CDM absorption. In situ data from the Gulf [Werdell and Bailey, 2005] and other ocean basins [Siegel et al., 2002b] indicate that the majority of CDM absorption is due to the dissolved rather than the particulate phase. Thus the discussion of CDM absorption variability may be further illuminated by an examination of the potential sources and sinks of colored dissolved organic matter (CDOM) in the surface waters overlying the deep Gulf. For greater clarity, we discuss CDOM first in terms of potential autochthonous sources from within the deep Gulf, and reserve the discussion of allochthonous materials for the section on lateral advection (5.2).
 CDOM has been previously described as the “wastewater” left behind from past oceanic processes [Siegel et al., 2002b]. The exact pathways of CDOM production are not clear, although there is evidence that CDOM is produced via bacterial degradation of noncolored organic matter derived from phytoplankton [Rochelle-Newall and Fisher, 2002; Nelson et al., 2004]. Regardless of the specific pathways of production, primary productivity presumably leads to an accumulation of wastewater and this implicates two in situ CDOM sources: (1) increased CDOM production in winter as a consequence of elevated levels of primary productivity; and (2) the deep chlorophyll maximum, often observed in subtropical basins below the seasonal thermocline and near the nitrate depletion isotherm, may also be a location of CDOM accumulation during the stratified periods, and these materials are subsequently brought to the surface during the winter overturn. This latter mechanism was proposed to explain the observed seasonal CDOM cycle in the Sargasso Sea near Bermuda [Michaels and Siegel, 1996; Nelson et al., 1998], and other tropical/subtropical latitudes [Chen and Bada, 1992; Coble et al., 1998; Siegel et al., 2002b; Simeon et al., 2003]. In the Gulf, the large seasonal barrier layer between the stratification period ILD (∼30-m) and the nitrate depletion isotherm (∼70-m; Figure 2a) may be such a location of CDOM accumulation. Subsurface CDOM maxima coincident with chlorophyll-a maxima have indeed been observed in the Gulf [Chen et al., 2004]. To be sure, however, CDOM production in the deep ocean may also arise from the decomposition of nonliving organic matter over much longer timescales than the seasonal cycle emphasized here [Nelson et al., 2007].
 Phytoplankton are consumed by herbivores, particulates may sink, yet there must be some additional removal process to account for seasonal CDOM depletion, particularly since CDOM largely consists of humic substances that may be recalcitrant to microbial consumption. The most likely removal process candidate is photochemical degradation [Mopper et al., 1991; Moran and Zepp, 1997; Mopper and Kieber, 2000; Del Vecchio and Blough, 2002]. Photochemical degradation may entail the direct oxidation of CDOM to dissolved inorganic carbon, the formation of other potentially smaller and more labile organic compounds, or the alteration of CDOM optical properties, i.e., photobleaching.
 Collectively, these photochemical processes act to reduce apparent CDOM absorption coefficients and may further explain the persistent relationship identified between CDM absorption and upper ocean thermal energy if one considers that the increased vernal shortwave irradiant flux (red line, Figure 3b) may have both photochemical and thermal consequences. The increased irradiant flux results in net upper ocean heating (positive net heat flux; Figure 3a), and the thickness of the surface mixed layer is consequently reduced (Figure 2a). Colored dissolved materials become confined to this restricted surface layer as they are simultaneously exposed to an increasing intensity of ultraviolet (UV) irradiance (280–400 nm), which is the spectral range that largely stimulates photochemical processes [Kieber, 2000]. As summer stratified conditions progress, UV-stimulated CDOM depletion may rapidly outpace CDOM production within the increasingly warm and nutrient-depleted surface waters. Conversely, a winter decline in the total irradiant flux, combined with wind-driven (latent) losses of heat energy, may stimulate convective overturn and turbulent mixing. Primary productivity increases during this period with the potential for a proportional increase in the CDOM production rate relative to that of CDOM depletion.
 Despite these potentially divergent constraints for CDM absorption and chlorophyll, the two variables are nonetheless highly correlated (Table 2a). A potential explanation for this tight CDM-chlorophyll coupling is that photoadaptation of surface phytoplankton may decouple some of the apparent variability in chlorophyll from changes in phytoplankton biomass, as has been previously suggested for subtropical ocean basins [Siegel et al., 2005]. Indeed, phytoplankton carbon to chlorophyll ratios, a proximate indicator of the photoadaptive response, may vary by a factor of six or more [Geider, 1987; Sakshaug et al., 1989; MacIntyre et al., 2002]. Given that CDM absorption appears to account for the majority of constituent blue-light absorption in the Gulf and in much of the global oceans [Siegel et al., 2002b], we may speculate that increasing intracellular quantities of chlorophyll (as a photoadaptive response to a light field largely attenuated by CDM) may further reduce the phytoplankton absorption efficiency due to the package effect [Morel and Bricaud, 1981; Kirk, 1994]. Accordingly, there may be a subtle interaction between chlorophyll, phytoplankton absorption efficiency, and CDM absorption that is discernable within satellite data.
5.2. Mesoscale Circulation and Lateral Advection
 Our seasonal mechanistic inferences are intentionally described thus far in a one-dimensional (vertical) context. The role of lateral advection becomes clearer as one considers the mesoscale variability. The Loop Current delivers surface waters to the Gulf from the Caribbean Sea that are comparatively lower in chlorophyll concentration [Muller-Karger et al., 1989; Bidigare et al., 1993; Melo González et al., 2000]. This pattern persists into the mesoscale (∼months, hundreds of kilometers) because isotherms, and by inference nutrient isopleths, are depressed within the centers of convergent circulation. In contrast, isotherms are uplifted within divergent, cyclonic circulation features, resulting in local “biological oases” [Biggs and Ressler, 2001], i.e., local areas of enhanced primary productivity.
 Analogously, our satellite estimates of CDM absorption coefficients in the surface waters of the northwestern Caribbean Sea are generally ∼25% lower than those estimated for the surface waters overlying the deep Gulf. Given that both the total irradiant flux and UV intensity increase towards the equator [Whitehead et al., 2000], it is not surprising that an initially poleward flowing surface current would deliver comparatively high-thermal energy, low-CDOM waters. Since there is ostensibly no significant CDOM source within convergent circulation features due to depressed nutrient isopleths and, consequently, a depressed rate of primary productivity, this pattern is maintained within the Gulf. Indeed, humic substance concentrations in the Loop Current are much lower than those typically observed elsewhere in the Gulf [Harvey et al., 1983; Carder et al., 1989]. In contrast, the biological oases provided by cyclonic eddies may also be local areas of CDOM production; although entrainment of deeper high-CDOM waters into the surface mixed layer may also occur. Here again, the tight CDM absorption to chlorophyll coupling may be due, in part, to a photoadaptive response.
 Further deviations from these patterns may indicate lateral advection of shelf waters or potential interactions of the deep Gulf with the continental shelf/slope. The Mississippi River plume, the Suwannee River plume, and other rivers and estuaries have a notable impact upon the Gulf's coastal ocean color variability and have been implicated in shelf water export events [Muller-Karger et al., 1991; Gilbes et al., 1996; Muller-Karger, 2000; Del Castillo et al., 2001; Gilbes et al., 2002; Jolliff et al., 2003; Toner et al., 2003]. Furthermore, coastal river plumes generally have CDOM absorption values one or more orders of magnitude larger than those observed in the deep Gulf [Green and Blough, 1994; Del Castillo et al., 2000; Hu et al., 2003]. Whether these allochthonous colored dissolved materials mix conservatively into the deep Gulf or are instead subjected to rapid photochemical degradation within buoyant surface plumes [Kieber et al., 1990; Miller and Zepp, 1995; Amon and Benner, 1996; Hedges et al., 1997; Vodacek et al., 1997; Andrews et al., 2000; Morell and Corredor, 2001; Del Vecchio and Blough, 2002; Jolliff et al., 2003] warrants further study.
 Nevertheless, the potential role of freshwater effluent in our basin-wide analysis of the deep Gulf deserves additional clarification. The Mississippi River drains 41% of the conterminous United States (the third largest drainage basin in the world [Van der Leeden et al., 1990]) and is responsible for the majority of the total freshwater discharge into the Gulf basin [Dunn, 1996; Goolsby et al., 2001]. Seasonal peaks in Mississippi River discharge rate often occur in spring (March–May [Walker, 1994]), and maximum seasonal CDOM absorption values associated with low-salinity plumes around the Mississippi River Delta extend well into the summer (July–August [Hu et al., 2003]). This pattern appears to be out of phase with the seasonal thermal/CDM absorption cycle for the deep Gulf identified here.
 Thus export of freshwater effluent to the deep Gulf may be tentatively identified in the satellite record as anomalous deviations from the expected bio-thermal patterns. For example, the daily mean CDM absorption values (as well as the bio-thermal model hindcast values) are extracted for the continental slope (300–2000 m depth) immediately south of the Louisiana and Mississippi coasts (Figures 15a and 15b). Although the potential influence of river plumes during the mixing periods cannot be excluded, the seasonal pattern of elevated CDM absorption values during winter months is consistent with the Gulf-wide bio-thermal variability as well as the local NCEP estimate of latent heat flux, and by inference, vertical mixing (Figures 15b and 15c). In contrast, increasing thermal stratification and shortwave irradiance (Figure 15c) should correspond to lower CDM absorption values, consistent with the photochemical CDOM depletion hypothesis. This is clearly not the case during the summers of 2003 and 2004 when significant deviations from the expected patterns occur (Figure 15b). It is likely that these anomalous CDM absorption peaks do, in fact, correspond to episodes of low-salinity water (<34 PSU) export to the deep Gulf.
Figure 15. (a) Map of extracted area (shaded region 300–2000 m depth). (b) The time series of the mean CDM absorption values (at 412 nm) from the shaded area is shown as black triangles; the solid black line corresponds to the low-pass filtered values. The mean of the CDM absorption values hindcast by the empirical bio-thermal model are also shown (red triangles; Figure 15b). The red dashed line corresponds to the positive anomaly criterion (thermal hindcast value +100% log-space error). The percentages of low-pass filtered SeaWiFS-estimated values that exceed this positive anomaly criterion for each thermal period are shown; values exceeding 10% are shown in red. (c) The local NCEP reanalysis estimate of latent heat flux. (d) The NCEP shortwave heat flux component is shown as the red line; the average MODAS-estimated STR-30 values are also shown (black triangles; STR-30 × 2000).
Download figure to PowerPoint
 There may also be a more subtle correspondence between these bio-optical anomalies and the thermal fields. Enhanced thermal stratification during summer (Figure 15d) may effectively decouple surface and bottom frictional boundary layers over the shelf–thereby enhancing Ekman drift in the surface layers [Kudryavstev and Soloviev, 1989; Savidge, 2002] and increasing the cross-shelf extent of water mass exchange [Lentz, 2001]. In addition to eddy intrusion over the continental slope, this stratification effect may augment summer time off-shelf river plume dispersal in the northern Gulf [Walker, 1996; Salisbury et al., 2004]. Indeed, entrainment of low-salinity waters into deep Gulf mesoscale circulation features appears to be much more frequent during the thermally stratified periods [Morey et al., 2003]. Thus it would appear that if the Mississippi River was the dominant source of colored dissolved materials to the deep Gulf, the seasonal CDM absorption and %CDM cycles should instead peak during the stratified periods. Since this appears not to be the case (Figure 6), the seasonal bio-thermal patterns identified herein likely arise from the cycling of autochthonous materials. Furthermore, fluxes of heat and fluvial freshwater into the deep Gulf may have distinctly opposite bio-optical consequences, and these fluxes may potentially be assessed using satellite-based data sets.