Variation in ultrafiltered and LMW organic matter fluorescence properties under simulated estuarine mixing transects: 2. Mixing with photoexposure

Authors


Abstract

[1] Ultrafiltered and low molecular weight dissolved organic matter (UDOM and LMW-DOM, respectively) fluorescence was studied under simulated estuarine mixing along with moderate photoexposure using Delaware, Chesapeake, and San Francisco Bays (USA) natural organic matter. UDOM was produced by tangential flow ultrafiltration (TFF) from the marine (>33 PSU), mid-estuarine (∼16 PSU), and freshwater (<1 PSU) members. TFF permeates (<1 kDa) were used to create artificial salinity transects nominally ranging from ∼0 to ∼36, with 4 PSU increments. UDOM or permeate (as control) from freshwater and mid-estuary was added to each salinity mix in the artificial transect to determine the impact of mixing behavior on optical properties. Three-dimensional fluorescence excitation-emission matrix (EEMs) spectra were generated for each end-member permeate (LMW fraction) and UDOM through the full artificial mixing transect. Fluorescent properties representing standard-identified peaks, fluorescence ratios and excitation-emission characteristics were assayed as previously reported. However, in this study, each sample was additionally photobleached for three days (nominally) to determine the coupled effect of estuarine mixing and photobleaching on LMW and UDOM fluorescence. Permeates, except Delaware Bay samples, were more bleached at lower salinities (<16). This effect was especially noticeable for mid-estuarine LMW organic material which was highly bleached at low salinities. Humic-type UDOM was generally bleached less at low salinities, maximally at mid-salinities, and less as it mixed toward the ocean end-member. As with mixing alone experiments, the B peak showed virtually no variability in the LMW and UDOMs fraction and was not significantly bleached. The N and T peak behaved similarly to one another and were significantly bleached. PCA and PARAFAC models confirmed trends for individual peaks. A four-dimensional PARAFAC model with pre- and post-bleached as the fourth dimension showed increases in the T peak fluorescence after photobleaching (with some overlap of the B and N peak). Results from this study indicate that coupled mixing and photobleaching can alter CDOM fluorescence in ways which might increase the difficulty in using CDOM as a proxy for DOM in regional carbon cycling biogeochemical models.

1. Introduction

[2] Terrestrially derived organic matter leached out of soils and sediments into streams and rivers ultimately travels toward the ocean end-member. Almost the entire terrestrially derived OM pool entering the ocean through estuaries is degraded within several thousand years, indicating rapid biogeochemical processing [Deuser, 1988; Hedges et al., 1997]. Although flocculation, photo- and microbial degradation have been suggested as removal processes [Bushaw-Newton and Moran, 1999; Kieber et al., 1989, 1990; Kieber and Mopper, 1990; Moran et al., 2000; Sholkovitz, 1976], scaling this mechanistic information to the global scale has been difficult [Cauwet, 2002; Hedges et al., 1997; Mopper and Kieber, 2002].

[3] Colored or chromophoric dissolved organic matter (CDOM) has been used as terrestrially derived organic matter tracer through rivers, estuaries and the coastal ocean as it is generally thought to undergo little biological degradation, and often shows conservative mixing [Blough and del Vecchio, 2002]. While CDOM shows little degradation under relatively static conditions (i.e., long-term bottle incubation), its signal becomes scarce in the open ocean. As with bulk DOM, mechanisms for CDOM removal must exist and be functional at very short geologic timescales. Photochemical alteration has been proposed as a significant mechanism impacting CDOM absorption and fluorescence [Moran et al., 2000]. However, with low light penetration in many estuaries, photodegradation effects are presumed to be more important in the coastal ocean than perhaps in estuaries [Andrews et al., 2000; Mopper and Kieber, 2002]. Controlled experiments show that natural organic matter can be converted to lower molecular weight organics [Kieber et al., 1989; Kieber and Mopper, 1990] and energy transfer between singlet oxygen, superoxides and peroxides can secondarily alter organic matter in coastal waters [Andrews et al., 2000; O'Sullivan et al., 2005].

[4] Assessing terrestrially derived DOM lability through estuaries has been difficult as although operationally defined, having little definitive chemical composition information (temporally or spatially) makes modeling chemical transformations problematic. The assumption that a component of the natural pool (operationally equated with CDOM) is resistant to microbial attack forms the basis of conservative coastal mixing models. For instance, CDOM fluorescence and CDOM absorption have been shown to decrease linearly with salinity until regions offshore of the continental shelf are sampled [De Souza Sierra et al., 1997; Vodacek et al., 1997]. Surveys from different regions, however show nonlinear relationships between CDOM and salinity [Blough and del Vecchio, 2002]. Recently, more mechanistic studies have been performed to begin determining what components of the natural organic matter pool are photochemically altered [Dalzell et al., 2009].

[5] Changes in ionic strength and perhaps divalent ion content might alter natural organic matter conformation and thus its optical properties. Using Chesapeake Bay waters, quantum yields were shown to increase during photobleaching (surface water, natural UDOM and humic acid treatments), but no specific dissolved iron effect was noted [Osburn et al., 2009]. Other metals, like copper, have been shown to alter quantum yield [Witt et al., 2007]. This finding has relevance when attempting to model optical properties, assumed to be static, across an ever-changing environment like an estuary. Reported here are the results of a systematic study showing considerable changes in CDOM fluorescence under natural mixing conditions, both within the low molecular weight (LMW) and (HMW) DOM pool. In this study, we hypothesize that CDOM conformational changes might differentially impact photolability thereby playing a role in altering CDOM loss or gain in a manner different from what would be calculated assuming CDOM mixes (and photodegrades) conservatively through an estuary.

2. Study Sites and Methods

2.1. Shipboard Analysis and Sampling

[6] Samples were obtained during cruises through the San Francisco Bay (2003) aboard the R/V Point Sur and in the Delaware and Chesapeake Bays from (2003, 2004) aboard the R/V Cape Henlopen. Sample locations, dates and a site map are presented in this volume [Boyd et al., 2010].

2.2. Mixing Experiments

[7] The ultrafiltration permeates were used to create an artificial salinity gradient (every 4 PSU) by mixing proportions of the freshwater (F) and mid-estuarine (E) (to create ∼0–16 PSU) and mid-estuarine (E) to marine (M) (to create ∼16–36 PSU). Initial salinities were measured using the ship's CTD system at the point of collection [Boyd et al., 2010]. UDOM from freshwater and mid-estuarine sites was added at a 1:10 dilution to the entire permeate salinity mix as was freshwater and mid-estuarine permeates (for LMW characterization and as UDOM only property controls). Experimental design is explained fully in [Boyd et al., 2010].

2.3. Exposures

[8] Sample mixtures (permeates and UDOM + permeates) were aseptically (flame sterilization) transferred to clean baked (450°C for 4 h minimum) quartz bottles with ground glass stoppers. The bottles were overfilled to exclude bubbles, stoppers were inserted and the tops were wrapped with parafilm. Quartz bottles were then transferred to a shallow Plexiglass box on the NRL Chemistry Building roof. Running water from the city water tap was introduced into the holding box and ran during the course of the exposure to keep temperature constant. The bottles were placed on their sides to maximize exposure. A set of dark controls (mixed but not exposed) were also analyzed [Boyd et al., 2010]. A Biospherical Instruments PUV-500 profiling radiometer was placed next to the incubation box and was run (data collected at 1 s temporal resolution) during the course of the rooftop experiment to record total exposure. Samples were nominally exposed for three days under sunny conditions. Total sunlight exposures were calculated from the PUV measurements using a model which interpolates solar spectra [Osburn et al., 2001] and ranged from 1.3 to 1.71 X 106 Joules m−2 (Table 1).

Table 1. Sample Exposure Doses
SamplesExposure Dates (2005)Ozone (Dobson Units)Exposure (×106 Joules m−2)
  • a

    Estimated value: PUV failed during exposure.

DB Mar03 F and E permeate and UDOM14–16 Jun2971.43
SF May03 F permeate and UDOM21–23 Jun3291.37
SF May03 E permeate and UDOM24–27 Jun3161.41
CB Sep03 F permeate and UDOM28–30 Jun3091.27
CB Sep03 E permeate and UDOM2–5 Jul3191.53
CB Mar04 F permeate and UDOM8–11 Jul3181.55
CB Mar04 E permeate and UDOM15–18 Jul3031.37
CB Aug04 E permeate and UDOM30 Sep to 3 Oct2741.71
CB Aug04 F permeate and UDOM14–17 Oct2571.30a

2.4. Optical Measurements

[9] Each individual sample was analyzed by UV-Vis absorbance spectroscopy on a Shimadzu UV-1601 spectrophotometer and excitation-emission matrix (EEM) fluorometry on a Shimadzu RF-5301-PC spectrofluorometer as previously described [Boyd and Osburn, 2004]. Daily instrument calibrations were performed with Milli-Q water. Absorbance was measured using a 5 cm cell from 250 to 600 nm. EEMs were collected using 5 nm excitation wavelength steps ranging from 250 to 450 nm. The emission intensity was measured from 250 to 600 nm in 1 nm steps at 300 nm min−1. Slit widths were 5 nm for the excitation and 1.5 nm for the emission. The highest sensitivity settings (in software) were used to obtain the scans. No samples were additionally diluted (aside from what is described for mixing experiments) before optical measurements. Instrumental measurement error was estimated at 1.7% [Boyd et al., 2010]. Experimental replication precision was determined for photobleaching experiments by running triplicate EEMs on each UDOM dilution after exposure. Standard errors ranged from 0.90 to 20.2% (for the C peak at 35 PSU) and had an overall average of 7.14%. UV-vis spectra had standard errors ranging from 0 to 0.45% with an average of 0.16%. Highest errors were at longer wavelengths (where signal was lowest). Shorter wavelengths were most important in inner filter corrections, so UV-vis spectral errors were considered negligible during mathematical processing. Additive experimental and instrument error was thus estimated as ±10%.

2.5. Data Analysis

[10] UV-Vis spectra and 3-D fluorescence data were exported to ASCII files and processed and corrected with a MATLAB™ script as previously reported [Boyd and Osburn, 2004]. As with mixing alone experiments [Boyd et al., 2010], UDOM fluorescence spectra were corrected for permeate fluorescence by mathematical subtraction after correcting permeate spectra by re-performing the inner-filter correction using the permeate plus UDOM absorbance spectra. This allowed scaling for the predicted fluorescence quenching by UDOM added to the permeate. Exposed EEMs were subtracted from their requisite mixing alone EEMs to determine photoexposure effects on CDOM (by difference). Excitation and emission wavelengths were output for the literature value regions identified in Table 2 [Blough and del Vecchio, 2002; Coble, 1996; Parlanti et al., 2000]. Fluorescence ratios were calculated as previously described [Boyd et al., 2010].

Table 2. CDOM Components Found in Representative EEMsa
Excitation (nm)Emission (nm)PeakaType
330–350420–480C, αhumic
250–260380–480A, αhumic
310–320380–420M, βmarine humic
310412C-Mintermediate C-M
280370Nunknown
270–280300–320B, γtyrosine-protein like
270–280320–350T, δtryptophan-protein, phenol like

[11] Parallel Factor Analysis (PARAFAC) was performed on “stacked” pre- minus post-exposed EEM spectral data. EEMs were stacked with the third dimension being salinity. Statistical processing routines were performed in Matlab™ using PLS_Toolbox (Eigenvector Research, Wenatchee, WA). PARAFAC analysis was constrained by setting excitation and emission wavelength values and loadings along the salinity axis to nonnegative numbers and the sample mode was set to observe salinity. Four compiled data sets were made from all spectra from each experiment; initial minus exposed freshwater permeate mixes (all seasons), initial minus exposed estuarine permeate mixes (all seasons), initial minus exposed freshwater UDOM mixes (all seasons), and initial minus exposed estuarine UDOM mixes (all seasons). All models were checked for convergence using random initialization. Seven components were chosen for the initial model (e.g., Table 1). Random subset cross-validation was used for all model runs and selected components were sequentially lowered until core consistency for each component was at least 50%. Convergence was tested using four runs (with random initialization) to confirm identical fits.

[12] A 4-D PARAFAC analysis was conducted with pre-and post-exposed EEMs in the fourth dimension in which case all modes were constrained to non-negativity. The models were examined for outliers by eliminating samples with extraordinary sum squared residuals and T2 values. The number of factors for each model was determined by fitting several models at different numbers of factors and considering the minimum of iterations needed for fit and decreasing variance captured of the last factor. Core consistency analysis was greater than 50% for the 4-component model.

3. Results

3.1. Photobleached Fluorescence Peak Intensity Maxima and Fluorescence Ratios

[13] There were no discernable trends in excitation or emission wavelength maxima for permeates or UDOM in response to salinity. Freshwater LMW fluorescent organic matter in permeates was more bleached in lower salinities (thus higher remaining signal when subtracting post exposure to pre-exposure) and aside from DB Mar03 samples, there was a general decrease toward the ocean end-member for remaining humic-type peaks (Figure 1a). The remaining B peak was anomalous and showed no general trend but considerable variability over the salinity range (Figure 1b). The remaining T peak was quite different from the remaining B peak and decreased toward the ocean end-member with variability at lower salinities between sampling and sites much like occurred with mixing studies alone (Figure 2a). The N peak showed less decrease, but was more bleached at lower salinities (Figure 2b).

Figure 1.

Removed peak intensity (pre minus post exposed) (a) A peak intensity, freshwater permeate and UDOM; (b) B peak intensity, freshwater permeate and UDOM; (c) A peak intensity, estuarine permeate and UDOM; (d) B peak intensity, estuarine permeate and UDOM. Legend: solid green squares, DB Mar03 F UDOM; solid blue circles, SF May03 F UDOM; orange asterisks, CB Aug04 F UDOM; solid purple stars, CB Sep03 F UDOM; solid black diamonds, CB Mar04 F UDOM; green open triangles, DB Mar03 E permeate; blue open inverted triangles, SF May03 E permeate; orange right-facing open triangles, CB Aug04 E permeate; purple open left-facing triangles, CB Sep03 E permeate; black open six-point stars, CB Mar04 E permeate.

Figure 2.

Removed peak intensity (pre minus post exposed) (a) T peak intensity, freshwater permeate and UDOM; (b) N peak intensity, freshwater permeate and UDOM; (c) T peak intensity, estuarine permeate and UDOM; (d) N peak intensity, estuarine permeate and UDOM. Legend: solid green squares, DB Mar03 F UDOM; solid blue circles, SF May03 F UDOM; orange asterisks, CB Aug04 F UDOM; solid purple stars, CB Sep03 F UDOM; solid black diamonds, CB Mar04 F UDOM; green open triangles, DB Mar03 E permeate; blue open inverted triangles, SF May03 E permeate; orange open right-facing triangles, CB Aug04 E permeate; purple open left-facing triangles, CB Sep03 E permeate; black open six-point stars, CB Mar04 E permeate.

[14] Estuarine LMW fluorescence was similar to freshwater, but humic peaks showed an increase in bleaching toward lower (<12) salinities (Figure 1c). This extremely variable fluorescence at low salinities was observed in mixing alone experiments using mid-estuarine permeates [Boyd et al., 2010]. The remaining B peak intensity was extremely variable and showed no obvious trend with salinity (Figure 1d). The remaining T peak signal appeared more like humic peaks than the B peak (Figure 2c), while the N peak was the highest bleached for some samples at low salinities (CB samples) and less bleached at low salinities for the faster flushing SF and DB samples (Figure 2d). As with freshwater LMW material, there was no discernable trend for fluorescence remaining in the B peak region (Figure 1d).

[15] Permeate peak fluorescence intensity before and after photobleaching was modeled using linear regression. Peak intensities were modeled against salinity for each peak. Goodness of fit statistics (r2) were high (about 0.80) for C, M, and C-M peaks for both freshwater and estuarine permeates and in all but several cases (B peak) were above 0.50. The assumption was made that equal permeate fluorescence photobleaching across the salinity gradient would not change the linear regression slope. A decrease in slope indicated higher bleaching at lower salinities, while a positive change in slope was interpreted as increased photobleaching at higher salinities. All sites/seasons showed a decrease in slope except for DB Mar03 which had increasing slope for the A, M, C-M, T and N peak with both freshwater (Figure 3a) and estuarine permeate (Figure 3b). The B peak showed no consistent shift (and generally did not regress with salinity).

Figure 3.

Slope change for pre- minus post-photobleaching for peak abundances. (a) Freshwater permeate and (b) estuarine permeate.

[16] Aside from the remaining B peak fluorescence, UDOM from both freshwater and estuarine systems showed remaining fluorescence which appeared lower at the salinity extremes, and higher in mid salinities (Figures 1a and 1d). While variations were often within the 10% error range, fluorescence bleaching appeared more pronounced at mid-salinities. This was less pronounced with freshwater DB Mar03 UDOM, which showed much less variability in the humic-type peaks and showed higher photobleaching (e.g., Figures 1a and 1c).

[17] LMW fluorescence in photobleached permeates resulted in uninterruptable peak ratios, likely due to extremely low signal (which was sometimes negative after subtraction). However, for most ratios, removed UDOM fluorescence allowed calculating peak ratios which fell within reasonable ranges. For freshwater UDOM, the McKnight fluorescence index (370F450/500) remained within a 0.1 unit range across the salinity gradient, but showed some variability between the 0–8 and 28–36 salinity ranges (Figure 4a). With removed E UDOM fluorescence, this ratio was considerably more variable, particularly for the CB Aug04 and CB Sep03 samplings (Figure 5a). With freshwater UDOM, the fluorescence index remained below 1.0, characteristic of terrigenous DOM [Del Castillo et al., 2001]. However, estuarine UDOM bleached at lower salinities had fluorescence index values higher than 1.0 (Figure 5b). With the exception of DB Mar03, the 305F430/540 ratio for photobleached freshwater UDOM increased systematically or in punctuated steps from freshwater to the ocean end-member indicating more terrigenous signal (Figures 4c and 5c). As with most spectral features, there was considerable variability at the salinity extremes. The less specific 355F450 fluorescence intensity index showed fluorescence signal least removed at salinity extremes and mid salinities (Figures 4d and 5d).

Figure 4.

Photobleached fluorescence ratios for freshwater UDOM mixes. (a) The 370F450/500 ratio (McKnight aromaticity index), (b) the 265F375/400 ratio, (c) the 305F430/540 ratio, and (d) the intensity for 355F450. Legend: solid green squares, DB Mar03; solid blue circles, SF May03; orange asterisks, CB Aug04; solid purple stars, CB Sep03; black solid diamonds, CB Mar04.

Figure 5.

Photobleached fluorescence ratios for estuarine UDOM mixes. (a) The 370F450/500 ratio (McKnight aromaticity index), (b) the 265F375/400 ratio, (c) the 305F430/540 ratio, and (d) the intensity for 355F450. Legend: solid green squares, DB Mar03; solid blue circles, SF May03; orange asterisks, CB Aug04; solid purple stars, CB Sep03; black solid diamonds, CB Mar04.

3.2. Photobleached Fluorescence PARAFAC Analysis

[18] PARAFAC analysis was conducted in two fashions. In the first analysis, EEMs representing each complimentary salinity were directly subtracted (pre-exposure-post-exposure) as in all above analyses. For each sampling event, the resultant EEMs were stacked by salinity (a 10 EEM stack representing EEMs at 0,4,8, …36 salinity) and classed by sampling event (see methods). In the second analysis, a four dimensional data space was created with excitation, emission, salinity, pre- and post-exposure. Data were classed by sample (site/season). Freshwater and estuarine UDOM samples were respectively combined. Under the first analysis, a 4-component model was validated for both subtracted (pre-post exposed) freshwater and estuarine permeate mixes. Subtracted (pre-post exposed) EEMs validated for only 2 components with freshwater UDOM and 3 components for estuarine UDOM. Excitation and emission maxima are annotated in Table 3. For the permeate samples, the N and A or C peaks had loadings which overlapped for several sites/seasons. In general, these scores decreased along the salinity gradient, indicating enhanced LMW fluorescence photobleaching at lower salinities (Figures 6a–Figures 6d). Loadings which overlapped or were near the N peak showed considerable variability and in the case of estuarine permeate, demonstrated extreme shifts at low salinities (Figure 6b). Permeate PARAFAC analysis was likely somewhat signal limited.

Figure 6.

Removed permeate PARAFAC scores for (a) N peak, freshwater permeate, (b) N peak, estuarine permeate, (c) A-C peak, freshwater permeate, and (d) A-C peak, estuarine permeate. Legend: open green triangles, DB Mar03 permeate; open blue inverted triangles, SF May03 permeate; open orange right-facing triangles, CB Aug04 permeate; open purple left-facing triangles, CB Sep03 permeate; open six-point black stars, CB Mar04 permeate.

Table 3. CDOM Components Found in PARAFAC Loadings EEMs
SampleComponentExcitation (nm)Emission (nm)Peak
F permeate1290365N
 2250360A (α′)
 3310–330415C (α)
 4260–270/375465A (α′)
E permeate1270–300365N
 2250/335360A (α′)
 3250/320400C (α)
 4265/350465A (α′)
F UDOM1260/325–340430A (α′), C (α)
 2275/370370–400N
E UDOM1260–280/300–325470A (α′), C (α)
 2260/320410C-M
 3250/295365N

[19] PARAFAC analysis for photobleached UDOM was difficult to interpret. With freshwater UDOM, loadings which corresponded to the N peak demonstrated considerable variability (Figure 7a), while much less variability was seen in loadings which overlapped A or C peaks (Figure 7b). The DB Mar03 scores falling into the A or C peak region demonstrated preferential bleaching at lower salinities (aside from an anomalous high value at 0 salinity for CBM04). As with estuarine permeates, the photobleached estuarine UDOM N peak scores (when found) appeared to decrease at very low salinities (except DB), while the humic-type peak scores (A-C) where more conserved and tended to show some increased bleaching at mid salinities relative to the salinity extremes (Figures 8a and 8b). The Intermediate C-M peak showed considerable bleaching at lower salinities (Figure 8c).

Figure 7.

Removed UDOM PARAFAC scores for (a) N peak, freshwater UDOM and (b) A-C peak, freshwater UDOM. Legend: solid green squares, DB Mar03; solid blue circles, SF May03; orange asterisks, CB Aug04; solid purple stars, CB Sep03; solid black diamonds, CB Mar04.

Figure 8.

Removed UDOM PARAFAC scores for (a) N peak, estuarine UDOM, (b) A-C peak, estuarine UDOM, and (c) C-M peak estuarine UDOM. Legend: solid green squares, DB Mar03; solid blue circles, SF May03; orange asterisks, CB Aug04; solid purple stars, CB Sep03; solid black diamonds, CB Mar04.

[20] The second PARAFAC model (4-D) provided loadings which overlapped the A, C, M, Intermediate C-M, B and T peak regions. This model primarily described how these components behaved (collectively across sites/seasons) under photoexposure. Four components overlapping peak regions were identified for both freshwater and estuarine UDOM (Figures 9a–Figures 9d). Component one primarily overlapped the A peak region (Figure 9a). Component two overlapped the A, M, and Intermediate C-M peak regions (Figure 9b). Component three represented the A and C peak regions (Figure 9c), and component four represented the T, and to a far lesser extent, the B and N regions (Figure 9d). Unsurprisingly, humic-type peak fluorescence (A, C, M, Int C-M) was decreased due to photoexposure (Figure 9e). However, component four, representing T, B and partially N increased in response to photoexposure (Figure 9e).

Figure 9.

PARAFAC loadings for pre- and post-bleached EEM PARAFAC stacks. Combined model with both freshwater and estuarine UDOM. (a) Component 1, (b) component 2, (c) component 3, (d) component 4, and (e) component behavior due to photobleaching.

4. Discussion

4.1. Peak Excitation and Emission Maxima, Peak Intensities, and Peak Ratios

[21] Humic-type peaks within permeate samples, representing LMW terrigenous organic matter, appeared almost universally to bleach more at lower salinities (Figures 1 and 2), the principal exception was freshwater permeates from DB Mar03. There was often (especially with estuarine permeate) a drastic increase in fluorescence photobleaching below 8 salinity (Figures 1 and 2). This fluorescence photobleaching effect appears to overlap with an anomalous increase in fluorescence within this region due to mixing effects (assumed to be ionic strength-related) alone [Boyd et al., 2010]. One could speculate that either processed end-member (freshwater perhaps) or in situ produced LMW organic matter is impacted by modest salinity changes (ionic strength) which may induce either conformational changes or alter complexation which in turn impacts photodegradation potential. The LMW organic matter would presumably consist of smaller biomolecules, amino acids, and small peptides [Yamashita and Tanoue, 2003], but the fluorescence signal might be dominated by fulvic acids [Leenheer et al., 1995a, 1995b]. B and T peak fluorescence (discussed further in this section) has been presumed to be derived from tyrosine and tryptophan residues within proteinaceous components; however, recent evidence suggests a LMW DOM source for this fluorescence [Yamashita and Tanoue, 2003]. While largely confined to the cell biochemistry community, reports are available which describe considerable fluorescence shifts in small peptides under different divalent ionic strength regimes such as KI [Weljie and Vogel, 2000]. We hypothesize that components of the mid-estuarine LMW organic matter pool are closely associated with divalent cations, which in turn modulate their fluorescence properties and photoreactivity. There is evidence that divalent metals can significantly impact the fluorescence properties of LMW organic matter during sea-ward mixing off the West Florida shelf [Zanardi-Lamardo et al., 2004]. In this study, the authors also noted a higher relative fluorescence for lower molecular masses. While perhaps more relevant to UDOM, decreases in photoreactivity have been noted with divalent metal-binding ligands [Witt et al., 2007]. With humic-type material, photochemical “protection” was noted when complexed with divalent copper [Brinkmann et al., 2003], although these elements are not likely in high abundance in the sampled estuaries. Dissolved element measurements during DOM photodegradation also argues for tightly coupled organic matter–trace element behavior in estuaries [Shiller et al., 2006].

[22] For freshwater and marine humic-type UDOM, there was slight increase in fluorescence photobleaching at mid-salinities and a decrease toward the freshwater and ocean end-members (Figures 1 and 2). The exception to this general trend was freshwater UDOM from DB Mar03 (Figures 1 and 2). Humic-type components were separated by type in a study from Isa Bay in Japan [Yamashita et al., 2008]. Type II humic materials were described as having non-conservative mixing behavior when plotted against salinity. While not separated into LMW and HMW fractions, type II humics were thought to be produced mid-estuary, possibly as a product of type I (C and A peak) photo- or bio-degradation. Another possible suggestion for type II humic creation was POM dissolution [Yamashita et al., 2008]. As we are aware of no other photobleaching experiments on LMW and UDOM from the same sample matrix, it is not possible to directly relate this data set to other analyses. It is likely that at very low salinities, LMW fluorophores, perhaps dominated by fulvic acids, are highly photoreactive. Their photoreactivity is diminished upon mixing with a small amount of seawater, possibly due to divalent ion concentrations increasingly complexing them. Lower salinity waters appear to have a smaller fluorescence photobleaching effect on UDOM but there appears perhaps to be a photolabile component of UDOM at mid-salinities. Based on peak and PARAFAC score changes with salinity, photolabile material may consist of HMW humic acids [Osburn et al., 2009].

[23] Quite possibly, a bi-phasic estuarine mixing scenario may exist in estuaries wherein LMW organic matter becomes increasingly complexed and more photo-resistant toward mid salinities while higher molecular weight organic matter (relatively photo-resistant at lower salinities) becomes more photo-reactive as it mixes into mid-estuary. This might be consistent with the mid-estuarine humic source explanation offered by Yamashita et al. [2008], particularly if UDOM represented a more refractory DOM component derived from riverine upper reaches. This scenario is consistent with earlier reports for the Chesapeake Bay system [Osburn et al., 2009].

[24] EEMs resulting from photoexposure had B and T peaks which differed considerably in character, both in terms of LMW and UDOM fractions. While these peaks have traditionally been thought to represent proteinacous material with tyrosine and tryptophan residues, respectively, their behaviors are beginning to be seen as different in natural waters [Yamashita et al., 2008]. Certainly in this series of experiments, both in terms of mixing alone [Boyd et al., 2010], and mixing with photoexposure, the T peak most closely resembled the N peak, which has been classified as non-humic and biologically labile [Coble et al., 1998; Cory and McKnight, 2005]. The tracking of these two peaks with salinity caused Yamashita et al. [2008] to classify them together as “type III” components. Photobleached fluorescence in both T and N peaks showed no trend in low salinity regions (<16), sometimes increasing and sometimes decreasing with decreasing salinity. Because the T (and N) peak showed different behavior based on size (LMW and UDOM), it is likely that they may represent proteinaceous material (HMW), fulvic acids, or phenolic precursors of humic material [Maie et al., 2007], each, of course, with different reactivity.

[25] The B peak showed no trend with salinity mixing alone [Boyd et al., 2010] and little to no photobleaching in either the LMW or UDOM samples. Yamashita et al. [2008] classify the B peak fluorescene as a type IV component and argue it represents relatively ubiquitous, highly degraded (thus refractory) LMW organic matter [Yamashita et al., 2008; Yamashita and Tanoue, 2003]. The B peak may represent the recalcitrant LMW fraction of DOM reported elsewhere [Amon and Benner, 1996].

[26] As fluorescence peak ratios are often used to track organic matter from various sources in coastal systems, it is interesting to note that for surveyed ratios (with UDOM), there were significant changes based on the salinity gradient. With freshwater UDOM, the McKnight aromaticity index varied by at least 0.02 units for all sites/seasons with the lowest variation coming from DB Mar03 (Figure 3a). There appeared, as with other parameters, significant shifts at the end of salinity ranges (0–12, 28–36). For estuarine UDOM, “dry season” samples from CB (Aug04, Sep03) showed extreme variation (up to 0.5 units) depending on salinity. Mechanistically, it is not possible without further chemical analysis to determine if fluorescence photobleaching changed the amount of aromaticity within the UDOM, but the presence of positive and negative fluctuations, especially in estuarine UDOM (Figure 5a), argues that small changes in spectral regions used for ratio calculations can have a significant impact on their values. The increase in this ratio at the ocean end-member for freshwater UDOM seems consistent within sites/seasons (with the common exception of DB Mar03). The 265F375/400 ratio has been used to infer a terrigenous vs ocean origin for natural organic matter. As with peak intensities, this ratio was extremely variable at low salinities, in the case of estuarine UDOM, well within the “marine” range of greater than 1.0 (Figures 4b and 5b). The 305F430/540 ratio also demonstrated considerable variability at freshwater end-member salinities and tended to be lower at the low salinity range and remain relatively constant at mid salinities toward the ocean-member. Without mixing and photoexposure history, it might be difficult to absolutely apportion organic matter fluorescence to a given “origin” solely using fluorescence ratios. The less specific CDOM fluorescence signal 355F450 also showed variation due to photobleaching at different salinities (Figures 4c and 5c), often mimicking humic-type peak variations (with higher fluorescence photobleaching at mid salinities). Again, without molecular-level knowledge of the mechanism altering UDOM during estuarine mixing and fluorescence photobleaching, it may be difficult to apply fluorescence ratios as a static tracer of CDOM origin.

4.2. PARAFAC Analyses

[27] PARAFAC analysis on bleached EEM stacks was able to identify several factors which corresponded to known fluorescence peak regions. However, PARAFAC on bleached EEM stacks produced scores which in many cases failed to overlap known peak areas (e.g., Table 2). In many PARAFAC runs, peaks corresponding to a recently reported region (component 3 [Yamashita et al., 2008], SQ1 [Cory and McKnight, 2005], or peak 2 [Stedmon and Markager, 2005]) were evident but were not universal. In general, peaks identifiable in PARAFAC scores followed peak intensity characteristics: linear or exponential decreae across the salinity gradient for humic-type permeate peaks (Figures 6c and 6d), and extremely variable scores in low or high salinities with UDOM (Figure 7b).

[28] The most enlightening PARAFAC analysis came from stacking the mixing alone and photobleached EEMs into a 4-dimensional model wherein changes across the fourth dimension (bleached vs non-bleached) could be evaluated. In this PARAFAC model, four components were modeled: One, overlapping with the A peak region; two, overlapping the A, M and Intermediate C-M region; three, overlapping the A and C region; and four, overlapping (bisecting) the T and to a lesser extent, B peak region (Figures 9a–Figures 9d). The major finding in this analysis is that while the humic-type peaks all decrease in abundance from pre- to post-exposure, the T-B peak component actually increases due to photobleaching (Figure 9e). Interpretation of a 4D model takes one more level of complexity to understand. The loadings underlying those scores show a profile of behavior, certain EEM components decreasing with photoexposure for components 1 through 3, and increasing with exposure for component 4. A positive slope in the loading for the T-B factor in the exposure mode could mean that the T-B peak in the exposed EEMs is decreasing less quickly with salinity than in the unexposed EEMs. Alternatively there could be real photochemical conversion of material to this florescent region. Clearly, as indicated by other studies, there is a different dynamic affecting the T-B peak with respect to salinity and photobleaching than with the humic regions of the EEM. Considering the biological importance of amino acids for instance, this should not be surprising.

5. Conclusion and Carbon Cycle Implications

[29] As with arguments made in this paper's companion study on mixing alone, current net terrigenous and autochthonous estuarine carbon flux estimates are difficult to obtain using discrete seasonal shipboard measurements. Spectral properties determined through extensive riverine biogeochemical sampling will increasingly be refined through remote sensing approaches. Efforts have been made over the past 30 years or so to link in-water optical properties to indices which can be measured over greater spatial and temporal scales (e.g., via satellite). This goal is predicated on the ability to definitively link conservative optical CDOM optical properties to both biogeochemical and remote sensing data. Empirical fluorescence and absorbance assessments have been performed to determine the net flux of CDOM from rivers, and coastal estuaries to the sea. The overriding question however remains as to what fate besets terrigenous and estuarine CDOM. Its signal is virtually absent from oceanic DOM and rates of input constrain a sink process working on the order of several hundred or less years. Laboratory and field studies suggest both conservative and non-conservative mixing behavior for optically active DOM in coastal waters [cf. Blough and del Vecchio, 2002]. This duplicity may be an indication that the dynamics of DOM mixing are highly complex. Several conclusions are evident from this work. First, LMW DOM fluorescence is likely attributable to fulvic acids which are known to be highly photoreactive and whose fluorescence would likely be quenched significantly with modest saline mixing. Second, the “proteinacious” component (ascribed primarily to the T peak and possibly containing components such as phenolics and fulvic acids), behaves differently from the humic-type components under photodegradation. Therefore mixing models for optically active DOM might need to consider each group of compounds, represented as factors in the EEMs, individually with unique responses to each environmental variable. This possibility shows that more work needs to be done tracing these factors through estuaries at various depths and longitudinal position. Finally, it can be hoped that those studies can produce robust models that will correlate with measurements that can be taken at higher temporal and spatial resolutions.

Acknowledgments

[30] This research was performed while Bethany Barham held a National Research Council Research Associateship Award at the Naval Research Laboratory. The Office of Naval Research supported this work. PARAFAC modeling was greatly assisted by Kevin J. Johnson. The authors also wish to thank the captain, crew, and support personnel of the R/V Cape Henlopen and R/V Pt. Sur and N. V. Blough for shared ship-time.

Ancillary