Portal de Periódicos da CAPES

Effect of estimations of ultraviolet absorption spectra of chromophoric dissolved organic matter on the uncertainty of photochemical production calculations

2011; American Geophysical Union; Volume: 116; Issue: C8 Linguagem: Inglês

10.1029/2010jc006823

2156-2202

Heather E. Reader, William L. Miller,

Air Quality Monitoring and Forecasting

Journal of Geophysical Research: OceansVolume 116, Issue C8 Free Access Effect of estimations of ultraviolet absorption spectra of chromophoric dissolved organic matter on the uncertainty of photochemical production calculations Heather E. Reader, Heather E. Reader Department of Marine Sciences, University of Georgia, Athens, Georgia, USASearch for more papers by this authorWilliam L. Miller, William L. Miller bmiller@uga.edu Department of Marine Sciences, University of Georgia, Athens, Georgia, USASearch for more papers by this author Heather E. Reader, Heather E. Reader Department of Marine Sciences, University of Georgia, Athens, Georgia, USASearch for more papers by this authorWilliam L. Miller, William L. Miller bmiller@uga.edu Department of Marine Sciences, University of Georgia, Athens, Georgia, USASearch for more papers by this author First published: 02 August 2011 https://doi.org/10.1029/2010JC006823Citations: 15AboutSectionsPDF 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 Share a linkShare onFacebookTwitterLinked InRedditWechat Abstract [1] The UV-visible absorption spectrum of chromophoric dissolved organic matter (CDOM) is crucial for accurate calculation of photochemical reaction rates in the ocean. The literature contains considerable variability for quantifying CDOM absorption spectra, and it is unclear how these different approaches affect subsequent photochemical calculations. Using 128 surface ocean samples collected during coastal transects from Texas to Maine, we examine the ability of four simple models to reconstruct the measured UV absorption spectra and examine the accuracy of photochemical production calculations made using the reconstructed spectra. Three exponential models are based on determination of a spectral slope coefficient (SSC) over distinct wavelength ranges (412–560 nm, 290–412 nm, and 290–650 nm) and one is based on absorption at only 412 nm. Including UV wavelengths to determine SSC resulted in the reconstruction of UV absorption spectra with high accuracy, underestimating measured absorption integrated over the UV by only −1.5% to −4.3% at worst for the 128 samples. The 412 nm model estimated UV-integrated absorption ranged between −4.3% to +6.5% of measured spectra for coastal stations. The 412–560 nm SSC model underestimated measured UV absorption at all wavelengths by up to 60%. A spectral correction factor based on the average percent underestimation for all samples was found to improve reconstruction of UV absorption and photochemical estimates. Without the correction factor, photoproduction from this model underestimates values calculated from measured UV spectra ranging from −13% to −20% for coastal stations. Corrected estimates improve this to between −1.4% and +6.8%. Key Points Accurate estimation of UV absorption spectrum is critical for photochemistry Spectral slopes determined over visible wavelengths underestimate UV absorption UV absorption can be corrected when visible is used to determine spectral slope 1. Introduction [2] Chromophoric dissolved organic matter (CDOM) absorbs UV and visible radiation and alters the optical properties of natural waters. CDOM is the fuel for photochemical transformations in natural waters. Photochemical transformations of CDOM are a dynamic part of DOM transformations in the marine environment. In coastal systems, where CDOM concentrations are high, photochemistry has potential to significantly impact the carbon cycle [Mopper and Kieber, 2002]. Most photochemical activity is largely driven by higher energy photons in the blue and UV wavelengths of the solar spectrum. In order to drive photochemical production models, it is essential to know the efficiency of the reaction of interest (i.e., apparent quantum yield (AQY)) as well as the energy and quantity of photons are absorbed by CDOM. The UV-visible absorption spectrum of the CDOM is thus a crucial piece of information required when attempting to implement photochemical models. [3] CDOM is a complex mixture of various compounds [Kujawinski et al., 2004; Mopper et al., 2007], and the shape of the UV-visible absorption spectrum has long been approximated by an exponential curve [Jerlov, 1976; Bricaud et al., 1981; Blough and Del Vecchio, 2002, and references therein]. Various mechanistic models have been postulated to explain the shape of the CDOM absorption spectrum from overlapping many individual single compound absorption spectra, which absorb fewer photons as the energy of the radiation decreases toward the visible [Blough and Del Vecchio, 2002; Stedmon et al., 2003] to an intramolecular electronic interaction model resulting in a relatively smooth decrease in absorption as the wavelength increases [Blough and Del Vecchio, 2004; Boyle et al., 2009]. In the literature, CDOM absorption is often reported using its "spectral slope" (S), which is the slope coefficient of an exponential decay function used to describe the spectral shape of absorption (e.g., S in equation (1)). Many methods have been used in the past to model spectral slope, and many of these depend on the instrumentation and wavelengths used to determine the absorption spectrum [Carder et al., 1989; Stedmon et al., 2000; Blough and Del Vecchio, 2002]. The addition of an offset, O, to an exponential decay function has been suggested as a better fit to the shape of CDOM absorption than the single exponential decay alone: where ag(λ) is the spectral absorption coefficient of CDOM, A is a fitting coefficient, S is the spectral slope, and λ is the wavelength. The primary cause of this offset has been ascribed to the difference in refractive index between pure water and salt water and arises from the common use of pure water as a blank in spectrophotometric measurements [Green and Blough, 1994]. Twardowski et al. [2004] did a comprehensive survey of various mathematical models used to determine spectral slope in the literature. They found that while the single exponential function with an offset described the shape of absorption better than the single exponential alone, a hyperbolic function performed statistically better than both. [4] Because many remote sensing applications retrieve CDOM at only one wavelength, typically 412 nm, Twardowski et al. [2004] suggested a single wavelength version of the hyperbolic model to describe spectral shape, derived from their extensive data set: [5] They found that for the visible wavelengths from 400 to 650 nm these two methods of describing spectral absorption by CDOM were able to model the absorption spectrum with less than 12% error. Because of the complexity of CDOM, the exponential shape is still only an approximation for the spectral shape of the absorption of light by CDOM, and deviations from the exponential shape can occur differently in different portions of the spectrum. More recent studies [Helms et al., 2008; Loiselle et al., 2009] have shown that the spectral slope varies considerably when applied to different portions of the absorption spectrum and that these variations occur differently in varying water types. These variations in concentration of CDOM as well as CDOM spectral slopes have been suggested as a semiconservative tracer for water masses [Nelson et al., 2007]. [6] Because the most critical wavelengths for CDOM absorption with respect to photochemical processes are the blue and UV wavelengths of light, it is crucial to know the absorption accurately at those wavelengths. Many literature values for spectral slope coefficients are not determined using absorption over UV wavelengths due to either instrument limitations or to the intentions of the authors in applying the data to nonphotochemical problems and thus not needing the information in the UV. While the fit of a particular model to optical data may be appropriate within certain statistical bounds and over certain wavelength ranges, owing to the spectral nature of all of the components involved in calculating photochemical production these statistical parameters can be misleading depending on the range of active wavelengths for different chemical species. In this paper we assess the usefulness of various approaches for the determination of spectral slope coefficients from visible data specifically for use in UV-dependent (i.e., photochemical) calculations by applying the single wavelength with offset model (equation (1)) calculated using various wavelength ranges, as well as Twardowski et al.'s [2004] generalized hyperbolic function (the single wavelength extrapolation) (equation (2)) to a large coastal water data set from the Gulf of Mexico and northwestern Atlantic Ocean. 2. Methods 2.1. Sample Collection [7] The data used to test the utility of various CDOM models were from 128 surface (between 2 and 4 m depth, avoiding the surface stagnant layer) samples collected from 10 July to 4 August, 2007, as part of the Gulf of Mexico East Coast Carbon Cruise (GOMECC; R/V Ronald H. Brown, NOAA) (Figure 1). This includes a wide variety of coastal and blue waters stations covering the eastern Gulf of Mexico and the East Coast of the United States from Texas to Maine. Samples were 0.2 μm filtered (Whatman Polycap AS75) into 150 mL acid washed amber glass bottles and stored at 0°–4°C until returned to the lab for processing. UV-visible absorption spectra were taken from 190 nm–700 nm at 1 nm resolution in 10 cm quartz cells using a Perkin Elmer Lambda-40 spectrophotometer. Pure water (Milli-Q, Millipore) was used as a blank. Figure 1Open in figure viewerPowerPoint CDOM absorption at 320 nm for the surface CDOM samples collected during the 2007 GOMECC cruise. Station colors show absorption coefficients as indicated with the color bar at the right. 2.2. Data Processing [8] Using the nonlinear fitting routine nlinfit, in the MATLAB ® Statistics Toolbox, the raw absorbance spectra were fit to a 3-part exponential equation (equation (1)) over the wavelength range 380–700 nm in order to determine the offset created by the differences in refractive index between seawater and the Milli-Q blank. The respective offsets were then subtracted from the individual spectra, and absorbance was converted to naperian (natural log scale) absorption coefficient using equation (3), where l is the path length of the spectrophotometer cell (m), A is unitless absorbance, resulting in absorption coefficients, ag with units of m−1. The absorption curves were modeled again using a single exponential curve model, with the offset removed, in order to determine the spectral slope [9] Spectral slope was calculated over three different wavelength ranges: 290–412 nm (Model A) encompassing the entire UV portion of the solar spectrum reaching the Earth's surface, 412–560 nm (Model B) in the visible [Twardowski et al., 1999], and 290–650 nm (Model C) covering the entire UV-visible spectrum. In addition to the single exponential models, the spectra were also reconstructed using the single point model (equation (2)) (Model D) as reported by Twardowski et al. [2004]. 3. Assessment 3.1. Spectral Slope Assessment [10] After determining the spectral slope coefficient over the given ranges, each UV spectrum was then reconstructed using the modeled coefficient. The spectral percent difference between the measured absorption spectrum and modeled absorption spectrum was calculated for each station and an average percent difference spectrum (±1 standard deviation) was created using the whole data set in order to assess the goodness of fit for each model in the UV (Figure 2). The average percent difference spectrum was then used as a potential UV correction factor for each model. A set of seven distinct stations (Table 1) was selected to provide a range of CDOM concentrations (ag(320) between 0.0958–3.6584) and geographic locations in order to examine how each model reconstructs different absorption spectra. The seven stations were also used to examine the effectiveness of using the average percent difference spectrum to correct for deviations from measured absorption arising from the different spectral models. Figure 2Open in figure viewerPowerPoint Average percent deviation of modeled CDOM absorption from measured CDOM absorption. Table 1. Sample Stations Chosen for Analysis of Models Station Latitude Longitude Basin ag(320) 43 30.8332 −79.4540 SAB 0.0958 28 25.7840 −86.3695 GOM 0.1050 35 31.4013 −80.8582 SAB 1.0197 18 27.5827 −90.0040 GOM 1.3097 10 28.8442 −90.3360 GOM 1.4885 71 40.8472 −69.0117 MAB 2.3637 60 39.3492 −74.0852 MAB 3.6584 [11] The range of spectral slope coefficients was larger for Models A and C than for Model B (Table 2). When using the spectral slope to then reconstruct the UV absorption curves over the wavelengths of interest (290–400 nm) the blue water stations (43 and 28; Figures 3a and 3b), which are in the Gulf Stream and Loop Current, respectively, had the largest deviations from the measured spectra for all models. For models A and C, the spectra were overestimated by between 13% and 18% at 320 nm, although when integrated over the entire UV spectrum (290–400 nm), the absorption was underestimated by −1.5% to −4.3% (Table 3). This highlights the difficulty inherent in using statistical values to measure goodness of fit along a spectrum such as this. Model D underestimated the UV spectrum for the blue water stations by −10% and −19%, as expected; Twardowski et al. [2004] do not recommend use of this model for blue waters. Figure 3Open in figure viewerPowerPoint UV-visible absorption spectra for the various models as fit to seven stations representing the transition from blue to dark water (Figures 3a–3g). In all cases, the difference between Model A and Model C was less than the thickness of the lines, and both are represented by the green line. Table 2. Spectral Slope Coefficients for the Various Fitting Models Used Spectral Fit Model A 290–412 (All Spectra) Model C 290–650 (All Spectra) Model B 412–560 (All Spectra) Model B 412–560 (ag(412) > 0.077) Minimum 0.0186 0.0174 0.0072 0.0131 Maximum 0.0316 0.0312 0.0272 0.0178 Mean 0.0235 0.0227 0.0146 0.0159 Median 0.0223 0.0217 0.0151 0.0159 Table 3. Percent Deviations of Modeled Spectra From the Measured Spectra at One Point (ag320) and Integrated Over the UV Station Model A (290–412) Model B (412–560) Model C (290–650) Model D (Single Wavelength) Percent Difference at 320 nm UV Integrated Percent Difference Percent Difference at 320 nm UV Integrated Percent Difference Percent Difference at 320 nm UV Integrated Percent Difference Percent Difference at 320 nm UV Integrated Percent Difference 43 13.1 −3.1 −59.4 −59.8 15.1 −1.5 0.6 −10.1 28 16.4 −4.3 −30.8 −39.4 17.9 −3.2 −5.5 −19.1 35 2.1 −0.2 −26.7 −25.3 2.6 0.6 −1.8 −1.9 10 1.5 −0.1 −20.6 −19.9 1.9 −0.4 −4.7 −4.3 18 2.3 −0.1 −19.7 −19.2 2.7 0.5 7.7 6.5 71 1.1 −0.1 −25.5 −24.1 1.4 0.3 −3.9 −3.3 60 1 −0.1 −17.1 −16.5 1.1 0.1 −2.6 −2.3 [12] Models A and C, which used UV wavelengths in the determination of spectral slope coefficient, showed higher accuracy in modeling the UV absorption spectra of the coastal stations (Figures 3c–3g). Percent differences between the measured spectra and the modeled spectra at 320 nm showed small overestimation by these two models on the order of 1.0%–2.7% (Table 3). When the absorption is integrated over the entire UV portion of the spectrum, these models deviated from the measured absorption between −0.4% and +0.6%. The ability of Model D to accurately reconstruct the UV spectrum was more variable than Models A and C with deviations between −4.7% and 7.7% at 320 nm and similarly, when integrated over the UV, 290–400 nm (−4.3% to 6.5%). [13] Model B, which uses the spectral slope coefficient determined over visible wavelengths only, showed underestimation of the true absorption spectrum for all stations. In this case, the deviation at 320 nm tracks closely to the deviation over the integrated spectrum (−17% to −59% at 320 nm, −17% to −60% integrated) (Table 3), because the spectrum is underestimated at all wavelengths for all stations. [14] In order to evaluate the possibility of correcting the underestimation of absorption in the UV for cases where visible absorption data is the only data available, the average percent difference between the measured and modeled spectra (Figure 2b) was used to create a correction factor (equation (5)) for Model B at each wavelength [15] Using the correction factor in conjunction with Model B, the modeled spectra were recalculated (Figure 4). While the correction factor did increase the modeled absorption for all stations, the absorption for blue water stations continued to be underestimated in the UV. Because the data set included a large number of blue water stations, which tend to deviate the most from exponential shape [Helms et al., 2008], the bluest stations (ag(412) < 0.077) were removed from the data set and Model B was reapplied. When the new correction factor was applied, the standard deviation of the corrected spectra showed a 50% improvement from ±21% to ±11% for the entire data set (Figure 5). Figure 4Open in figure viewerPowerPoint Model B (visible) spectral reconstructions with corrected spectra. The dashed lines are the corrected spectra calculated with ±1 standard deviation of the average spectral percent difference, illustrating the envelope of error. The seven stations are in order representing the transition from blue to dark water (Figures 4a–4g). Figure 5Open in figure viewerPowerPoint Standard deviation of the corrected fits (by percent difference) for the entire data set and for the data set with the clearest waters removed. 3.2. Photochemical Production Assessment [16] Because of the spectral nature of photochemical production, assessing the overall deviations of the absorption spectrum can overlook issues arising from the spectral distribution of the photochemically active wavelengths for a particular product. To assess the effect of different modeled UV-visible absorption spectra on photochemical production modeling, two distinct photochemical reactions were chosen: photoproduction of carbon monoxide and formaldehyde. Equation (6) [Johannessen and Miller, 2001] is used to calculate surface photochemical production rate, P(λ) (mol/m3/day), and is defined as the change in concentration of product over time: where E0 is the irradiance (mol photons/m2/day), ag is the absorption coefficient (m−1) of the CDOM, and Φ is the apparent quantum yield (or spectral efficiency of photochemical production, mol product/mol photons). Each term is calculated at each wavelength from 290 to 490 nm, and the product is summed over this range to give a spectrally integrated result. Surface daily photochemical production (integrated spectrally 290–490 nm) was calculated for each station and each model using a modeled solar spectrum for one day in August, at Sapelo Island, Georgia (latitude 31.464°, longitude −81.244°) (System for Transfer of Atmospheric Radiation (STAR) model) as a constant radiance spectrum for all calculations. The STAR model was developed at the University of Munich, is based on the work of Ruggaber et al. [1994], and was implemented here following the methodology of Fichot and Miller [2010]. Figure 6 shows the cross product of spectral irradiance, E0, and the apparent quantum yield for both reactions (from equation (6)) to show the active wavelength ranges where errors in the estimation of absorption coefficient will be most critical. While the active wavelengths for photochemical formaldehyde production are centered in the UV–B [Kieber et al., 1990] the photochemical production of carbon monoxide [Zafiriou et al., 2003; Ziolkowski and Miller, 2007; Fichot and Miller, 2010] extends much further into the visible, with a maximum of production ∼330–350 nm (Figure 6). Figure 6Open in figure viewerPowerPoint Cross product of the spectral irradiance and the apparent quantum yield for formaldehyde and carbon monoxide, revealing photochemically active wavelengths, to show where errors in absorption estimations will have the most impact. [17] Spectrally integrated formaldehyde (CHHO) photoproduction numbers for Models A and C agreed with the CHHO photoproduction numbers calculated from the measured CDOM spectra for the coastal stations, within ±2%. The blue water stations performed much worse, as expected, with a 9%–12% error. This is expected because over the active wavelengths for formaldehyde production (300–320 nm), Models A and C tend to overestimate absorption. Model D returned errors of −4% to –11% for the coastal stations. Similar to the results of Twardowski et al. [2004], the single wavelength extrapolation returned large errors for the blue water stations (−12% to −19%). Model B severely underestimated photoproduction of CHHO by up to 31% for the coastal stations and 70% for the blue water stations (Figures 7a and 7b). When adjusted with the correction factor for the entire data set, Model B overestimated the production by 11%–30% for the coastal waters and continued to underestimate the blue waters (−11% to −50%). After removing the blue water stations from the analysis, application of the new correction factor for Model B improved the estimates of coastal photoproduction of CHHO to between −2% and +14% error, performing as well as Model D. Figure 7Open in figure viewerPowerPoint Surface photoproduction estimates for (white left-hand bar) measured Model B, (gray middle bar) calculated Model B, and (black right-hand bar) corrected Model B. Figures 7a and 7c represent the Model B correction factor calculated with the entire data set, and Figures 7b and 7d are for the coastal application of Model B correction factor. Stations are in the order of bluest (most oligotrophic) to darkest (most coastal). [18] The active wavelengths for the photoproduction of carbon monoxide (CO) extend much further toward the visible than do those for CHHO. This serves to minimize the photoproduction error caused by misestimation of the absorption of CDOM in the UVB. The patterns of estimation of photoproduction for CO remained the same as for CHHO, however the errors were generally lower. For Models A and C the errors ranged from −0.37% to −0.02% and 0.48% to 1.1%, respectively, for the coastal stations. For the blue water stations the errors were ∼−4.5% and ∼−2.0%, for those models. The Model D errors ranged between −4.5% and 4.5% for the coastal stations and between −7% and 0% for the blue water stations. Using Model B applied to the full data set returned errors between −13% to −20% for the coastal stations, increasing to 0% to 11% after correction (Figure 7c). Photoproduction for the blue water stations (43 and 28) were underestimated by between −25% and −50%, with the correction factor improving the estimation to between −2.0% and −30%. After the removal of the blue water stations (Figure 7d), the new Model B correction factor improved the estimates to between −1.4% and 6.8%, again putting it on the order of Model D. 4. Discussion [19] While calculated predictions of photochemical reaction rates in surface waters can contain errors from AQY and irradiance uncertainties, it is also clear from equation (6) that they are equally susceptible to erroneous estimates for the absorption of solar radiation by CDOM. Published papers reporting the distribution and variability of CDOM absorption spectra continue to show considerable variability among the ways that CDOM absorption spectra are reported. It is clear from the results here that the methodology chosen to describe CDOM absorption can significantly affect resulting photochemical calculations. [20] Looking first at the fit of the exponential models, we note that Model B exhibits a strong spectral bias of increasing error toward the shorter wavelengths of the UVB portion of the spectrum (Figure 2). Models A and C also show a spectral bias in their error distribution, although this is much smaller (<20%). Taken together, this argues that the practice of fitting the CDOM absorption spectrum to a single exponential curve [e.g., Jerlov, 1976; Bricaud et al., 1981; Blough and Del Vecchio, 2002, and references therein] may be a fundamentally flawed approach. Using UV wavelengths in calculating the spectral slope coefficient results in a better fit, although none of the models we evaluated are able to accurately fit the shape of the absorption in the UVB for the blue water stations (Figure 3), suggesting that single exponentials are not appropriate in open ocean waters. Coastal waters, with higher concentrations of CDOM, have absorption spectra that tend to approximate an exponential curve better than blue waters. [21] Indeed, other studies have shown that the source of CDOM [Helms et al., 2008; Loiselle et al., 2009] as well as the concentration of CDOM [Bracchini et al., 2010] have a strong effect on the distribution of spectral slopes across the UV and visible wavelengths. Marine-derived CDOM tends to have higher spectral slope values over the UV as well as increased variability in these spectral slope values, potentially due to the lower overall concentrations of CDOM [Nelson et al., 2007; Swan et al., 2009]. Terrestrially derived CDOM and samples with higher concentrations of CDOM tend to have less variable spectral slope distributions as well as lower overall slope coefficients. Interestingly, Model D, which uses the absorption at 412 nm to extrapolate the absorption in the UV, shows no spectral bias (Figure 2) except for a slight decrease below 300 nm. This is particularly surprising because the equation was developed from visible absorption data only [Twardowski et al., 2004]. The sharper inflection of a hyperbolic curve relative to an exponential curve could explain the ability of this equation to fit the UV portion of the absorbance spectrum where the spectral slopes are generally higher than those in the visible [Helms et al., 2008]. Although the standard deviation of this model is higher than obtained with the spectral slope derived models, this lack of spectral bias is encouraging for use in photochemical calculations. [22] Obviously, the best option for calculating photochemical production for all cases is to use the true absorption coefficients measured over the photochemically active wavelengths in the UV and visible. When this is not possible, such as cases where only spectral slope coefficients are available, spectral slope coefficients that were determined over the UV allow for fairly accurate estimates of photoproduction in coastal waters. Because of the deviation of bluer waters from the exponential shape, the error for these depends highly upon the range of photochemically active wavelengths (i.e., the response spectrum of the product in question). A photochemical production reaction such as that seen for formaldehyde, with its narrow range of active wavelengths (Figure 6), has higher potential errors in photoproduction calculations than for a reaction like carbon monoxide, where the photochemically active wavelengths extend much farther into the visible. Because the entire photoproduction of formaldehyde is driven by absorbed radiation in the UVB (300 to 320 nm), it is crucial to know the absorption of CDOM accurately over these wavelengths. Carbon monoxide photoproduction is maximal between 330 and 350 nm with significant photoproduction driven by visible radiation, making it necessary to know the CDOM absorption accurately over a different and larger range than for formaldehyde. The lower errors found for the modeled photoproduction of carbon monoxide relative to that of formaldehyde are due to the more accurate estimation of UV-visible absorption as the wavelengths increase toward the visible. Despite the differences in the response spectra of different photochemical products, errors for photochemical production can range from −10% to +10% for blue waters, when the UV is used as part of the spectral slope determination. However, it is important to be aware of the active wavelengths of the photochemical product in any photochemical modeling. [23] When visible data for CDOM absorption is the only data available, using the single wavelength extrapolation reported by Twardowski et al. [2004] (equation (2)) is a good option to estimate the UV absorption spectrum in coastal waters, with relatively small errors for photochemical production and no spectral bias (Figure 2). Conversely, reconstructing the UV absorption spectrum from spectral slope coefficients determined using only visible wavelengths severely underestimates the absorption with the strong spectral bias of increasing error toward the shorter UVB wavelengths, and thus the photochemical production for coastal waters is also underestimated. The spectral correction factor for Model B, developed from our data set by fitting a quadratic equation to the spectral percent deviation and inserting it into the generalized equation (5) resulting in equation (7), may be successfully applied to the reconstructed absorption spectra for coastal waters from the Gulf of Mexico and the East Coast of the United States and results in significant improvement for photochemical production estimates. In cases where visible spectral slopes are the only data available (e.g.,