Dissolved organic carbon in coral-reef lagoons, by high temperature catalytic oxidation and UV spectrometry

Abstract Two surveys were carried out on ten atolls in the Tuamotu archipelago (French Polynesia, Pacific Ocean). In vitro UV (250–400 nm) spectra of water samples gave absorption at 254 nm, A 254 , and spectrum slope, S ⋆ (computed from In A λ versus λ).These two descriptors are negatively correlated, and data points are arrayed along a hyperbola spanned between an oceanic pole (high S ⋆ , low A 254 ) and a confined pole (low 5 ⋆ , high A 254 ). Dissolved organic carbon (DOC) concentrations, [C], as assessed by HTCO, exhibit a narrow range (0.7–1.0 mg C.L −1 for most lagoons) contrasting with the wide diversity of optical characteristics. [C] and A 254 are positively correlated, with a significant intercept (0.5 mg C.L −1 ) representing non-chromophoric DOC. Carbon-specific absorption, e 254 increases (from 0.4 to 1.3 m 2 .g −1 ) with increasing [C], mainly according to the literature) owing to increased average molecular weight (MW) of the chromophoric DOC fraction, which also lowers S ⋆ . Our optical data thus illustrate a gradient of confinement (or residence time) that corresponds to a continuum in DOC nature, especially in MW and hence in bioavailability. Optical methods are confirmed as quick and effective means of assessing DOM distribution.

. This has a direct consequence on the perspectives of assessing DOM's bioavailability, which is at least as important as knowing total DOC concentration (see example in Benner et al., 1995). Describing "bulk characteristics" of marine DOM will thus remain a plausible aim, while it would be 916 useful to recognize some broad categories (Hobbie, 1992), for instance through nonchemical methods (Cable it al., 1990). Among these, optical characteristics have been widely used, perhaps more often in continental waters (partial review in Moore, 1987) than in open-sea waters. We applied spectrophotometrr to characterize the waters of several coral-reef lagoons, with the initial aim of a typological, semi-quantitative classification (Pages et al., 1997). We could thus describe the distribution ofqualitative characters of DOM (or at least of its "chromophoric" part) among different lagoons. Here, wc compare these results with those of the classical HTCO (high temperature catalytic osidation) assessment of DOC concentration, to determine whether i) there is a relationship between optical properties and DOC concentration, and ii) we can predict DOC from optics.

II. MATERIAL AND METHODS
Our observations were part of several programs studying the general biological produc-tion ofatoll lagoons, and more specifically the processes in the water column.
The main series of samples was taken during surveys (the "Typatoll" cruises) carried out in h'ovember 199.5 ("Typatoll 3") and March 1996 ("Typatoll 4")) during which we studied a total of ten atolls in the Tuamotu archipelago (1.518%. 141-146"u'). op en-sea samples were also collected. Some characteristics of the lagoons studied here are given in table I.
Sampling was performed within a few hours on each atoll, at 5-S stations (see details in Pages ~1 al., 1997).

Spectrophotometry
was performed on board during the surveys. We used a Milton Roy "Spectronic 1201" fitted with a 100 mm quartz cell. We measured absorption (Ah) at 12 discrete wavelengths between 2.50 and 400 nm against blanks of distilled de-ionized water. Reproducibility of absorption values is good. Mean coefficient of variation on 71 duplicates (either successive sub-samples or duplicate samples) is 0.019 (i.e. less than 2%).

Spectral parameters and qualitative aspects
Most of this section has been expounded elsewhere (Pages et al., 1997), but must be briefly recalled for comprehension of the following.
Absorption spectra show the classical linear relation between In A,and h (Bricaud ~1 al., 1981). Spectrum slope (S*) values range between 0.009 and 0,030 nm-' in atolls, while oceanic samples exhibit higher values, up to 0.040 nm-' (table I) Taiaro) A plot of A,,, against S* for all lagoon and seawater samples ( fig. 1) shows an inverse correlation between these two descriptors. The waters from Taiaro, a closed lagoon with abundant terrestrial vegetation, are conspi-CUOLIS outliers. Their abnormally high S* values stem from the presence of lignin degradation products and humic acids (ms in prep.). Excluding this lagoon, the best fit (r? = 0.60) is given by the regression of (l/S*) against (l/A,,,,) (see table II).
Data points are not randomly distributed along this hyperbolic regression curve. Spatially detailed observations in Tikehau lagoon (set details in Pages et al., 1997) show that data points are arranged between an "oceanic" pole, with high S* values (-0.04 nm-') and low A,,, (0.4 m-'), and a "eutrophic" pole, with low S* (< 0.01 nrn-') and high 4,,, (> 2 m-'). BY mixing different water types, we could see that the observed hyperbola corresponds to mixtures, in varying proportions, between two main types of water reflecting the above description of the two poles. We are thus dealing with a continuum of waters between which there is no basic difference, but only a continuous variation of proportions.

Quantitative aspects: DOC and carbonspecific absorbance (a) DOC concentrations
For most lagoon samples, DOC concentrations [C] exhibit a narrow range, from 0.7 to 1.0 mg CL-' (table I), against [C] values of -1.0 mg C.L-'in oceanic samples. Relatively high DOC concentrations are only found in Rekareka and Taiaro (averaging 1.11 and 1.82 mg C.L-' ). Intra-lagoon variability is very low in most cases. Coefficient of \iariation (CV) for a given lagoon on a given survey generally amounts to -5%, and this includes the (low) analytical variability. If we conside only "normal" lagoons and oceanic samples,  I).
Absorption, A,,, (in m-l), is correlated with DOC concentration ( fig. 2 and table II). The overall regression (r2 = 0.76) is slightly improved (r) = 0.84) when oceanic samples exhibiting "too much"DOC arc excluded. W'e can note that this correlation is acceptable only after including the "high DOC" lagoons (Taiaro and Rekareka); with "normal" lagoons only, the correlation is statistically significant (r = 0.521, n = 107, P > 0.001) but the regression is worthless in terms of predicting [C] from A,,,. The other salient point is the intercept, amounting to about 0.5 mg CL-' of DOC without optical activity.
Inspection of the results indicates that Ayn4 allows a better discrimination between lagoons than does [Cl. To quantify the discriminative potential of A,,, and [C] , we tested the null hypothesis of a homogeneous population comprising all stations in all lagoons and the ocean during the two surveys (22 sets of six data each), using the Kruskal-Wallis one-way analysis by ranks (two-by-two comparison on rank sums of adjacent sets; 21 comparisons). Significance of the differences between lagoons was also tested using the Mann-I#'hitney test (two-by-two comparison on all data; 213 (i.e. 21 -20 t 19 t...) comparisons). Results of the two tests confirm that optics are more able to discriminate between lagoons than chemical (HTCO) DOC measurements. Comparison between Ebb,, and spectrum slope S* shows a general inverse trend (high Q~.+ values for low Y), with Taiaro samples as prominent outliers (figure not shown). Excluding Taiaro, the best fit is given by a hyperbolic (l/E,,, versus l/S*) equation with significant correlation (r2 = 0.62, P < 0.001 for n = 130). Data points appear too widely scattered (figure not shown) for a satisfactory prediction of Q,, from S*.

(c) Prediction @DOC concentration
The various correlations seen above should allow determination of DOC conccntration, [Cl, from optical properties, in particular from the regression between A,,., and DOC concentration. We obtain a set of computed values, [C,,], which we compare with actual (measured) values [Cm] (see fig. 3).

Iv. DISCUSSION
The relations we found between optical properties and DOC concentration, [Cl, have two aspects: i) quantified prediction of [C] and ii) semi-quantified evaluation of the nature of the DOC.
Optics, and especially absorption at a given wavelength, have been widely used as a proxy estimator (or a "surrogate parameter" (Summers et al., 1987)) of DOM concentration, in the laboratory or in the field. Even if caution is necessary in some particular environmen ts, such as hard-water lakes (Stewart Excluding Rekareka and Taiaro, the "normal" lagoons show a small relative variation in total DOC concentration (range 0.74-1.05 mg C.L-'). This contrasts with the much wider variation in absorption (A,,, ranging between 0.43 and 0.93 rn-').
We have seen that A,,, discriminates better between lagoons than does bulk DOC concentration. The sensitivity of optical characteristics (both E,,, and S*) to minute variations in DOM nature and/or concentration has been shotvn in a previous paper (Pages rt al., 1997).
The DOC concentrations we found are quite normal for oligo-to mesotrophic waters (Martin and Fitwater, 1992;Guo it ul., 1993;Carlson and Ducklow, 199.5). In these "normal" lagoons, then, the purely quantitatij-e aspect of optical measurements is useful only as a preliminary estimation of bulk DOC concentration, especially when considering the cost/benefit ratio (see the analogous conclusion reached by Moore, 1987).
The semi-quantitative evaluation of DOC nature has higher potentials. Our argumentation is based on (i) the A,,,-versus-S* distribution, and (ii) the correlation between [C] and A,,,. This latter shows the existence of a "residual" colourless DOC fraction amounting to about 0.5 mg CL-'. Such an optically inactive fraction (appearing as the intercept in the [C] -versus-ALj,i4 -. regression) has been found in other environments, at concentrations ranging between 0.4 and 0.9 mg CL-' in oceanic samples (Amador et al., 1990;Vodacek et al., 1995) and between 2 and 6 mg C.L-' in coastal or continental waters (De Haan and De Boer, 1987).
The relative increase in the ratio between active and inactive Cat increasing [C] leads to the mathematical artifact of increasing E,,~.(, even if the "true" specific absorption (E*) were constant, which is not necessarily true (see below).
Among the identified factors of variation of E (at any wavelength) are ionic strength (Summers el al., 1987;De Haan el al., 1987) and pH (De Haan ~1 al., 1983), which are irrelevant for our measurement.5 on natural seawater. We deal here with marine DOM, which is an undetermined mixture of molecules (Amon and Benner, 1996). The chief factor controlling E in natural samples will be average molecular weight, MW, or "molecular size". Several studies show, more or less explicitly a positive correlation between E and log MW. and a negative correlation between spectrum slope S* (or the equivalent ratio E2/E3 (De Haan P/ al., 1988) and MW (Stewart and Wetzel, 1980;De Haan et al., 1983;Summers et al., 1987;Senesi d al., 1989).
The "confined" pole, with high .&, (and high [Cl) corresponds to gradually increased average MN' (as evidenced by the low S* values), as found in other environments (Tranvik, 1990;Guo et al., 1994). The increase in &2jI in this chromophoric portion may be also due to a parallel (and slight) increase in the proportion of high-e* molecules (Blough rf al., 1993), such as lignin derivatives, or LWprotecting compounds (Shick el al., 1992).
The continuity that we observe between lagoons suggests a continuum of MM:, such as is described by several authors (Moran et al., 1991;Guo pt al., 1993;Amon and Benner, 1996). This array of IVIES can result in part from bacterial heterotrophic activity, which leads to decreased S* (Blough et nl., 1993). Bacterial uptake induces very small variations in bulk propertics (Brophy and Carlson, 1989), but may alter the more sensitive optical properties of DOM.

v. CONCLUSIONS
Optical characteristics allow the prediction of DOC concentration with an accuracy of about + 0.1 mg CL' (average absolute difference between [Cm] and [C,]). The main potential of optics lies though in the ability to detect minute variations, or alterations. in the nature of DOC. or at least of its "chromophoric" portion.
We have noted that our data points (OUI lagoon samples) are arrayed. in the A,,.,versus-S* plane, between two poles. A distance along the regression curve (measured on the graph from th e d ata points for open-sea samples) would then also correspond to a factual distance (in space and/or in time) between the open sea and a water body. Optical characteristics would then estimate confinement. We could verify their good correlation with chlorophyll concentration Pages ~1 al., 1997), which has been shown to be related to water residence time (Furnas PI al., 1990;Delesalle and Sournia, 1992).