Constituents controlling UV attenuation in natural waters biooptical models

Bio-optical models have been developed to predict spectral attenuation as a function of conveniently measured parameters and are discussed extensively in Mobley [5] and Kirk [6]. Since the pioneering work of Smith and Baker [42], bio-optical models typically break down diffuse attenuation into optical constituents of natural waters. In an approach covering UV-B and UV-A wavelengths summarized by Baker and Smith [9,43], these components are represented as partial attenuation coefficients (A for each term not shown for simplicity):

to which Kd Tripton may be added for waters where attenuation is caused by nonliving particles.

These components can be individually computed from measurements at a site after proper "calibration" to establish "specific attenuation" factors. For example, from a series of sites that differ in DOC concentration and having low or constant levels of phytoplankton pigments, Kd totai and [DOC] are measured

Table 1. UV optical properties, ranked in order of UV transparency

Site WL(nm) c = * + b bm <*w (m1) 0^(01") ap (m1) ^(m1) Source

A. Ocean and coastal measurements

UV-A ocean data

Pure seawater (see Freshwater table for <0

Pure seawater

380

0.0094

Morel 1974 in [6]

Clearest natural waters

380

[0.031]

0.022

0.027

Smith & Baker [18]

Sargasso Sea

380

0.044-0.045

Tyler & Smith [110]

East Mediterranean

375

0.05

Jerlov [95]

Gulf of Mexico USA

380

0.06

Smith & Baker [8]

Eastern Pacific (Mexico)

380

0.075

Tyler & Smith [110]

West Mediterranean

375

0.02-0.08

Hojerslev [65]

Antarctica (61°S)

380

0.08-0.15

Helbling et al. [44]

Coastal Japan (3/98-3/99)

380

0.15a

0.16a

0.09-0.59,0.24a

Kuwahara et al. [108]

Gulf of California (Mexico)

380

0.11-0.28

Tyler & Smith, [110]

Arctic polynya

380

0.18-0.50

Belzile et al. [48]

Tropical, near coral reefs

380

0.18-0.76

Dunne & Brown [111]

Gulf of St. Lawrence

380

0.1-0.8

Kuhn et al. [57]

Gulf coast, Florida USA

380

0.28

Smith & Baker [8]

Continental Slope

365

0.10

0.02

Clark & James '39 in [4]

Bermuda

380

0.20

0.03

Ivanov et al. '61 in [4]

North Sea

375 375

0.09-0.15 0.16-0.53

Hejerslev [65] Stedmon et al. [77]

Kattegat

380 375 375

0.54

0.30-0.68 0.46-0.92

0.3

Jerlov '55a in [4] Hajerslev [65] Stedmon et al. [77]

South Baltic Sea

375

0.71-0.90

Stedmon et al. [77]

Baltic Sea

375

0.89-1.04

Hejerslev [65]

Delaware Bay mouth USA

380

0.6-1.2

Vodacek et al. [23]

Skagerrak

380 375 375

0.12-0.31 0.09-0.42

0.3

Malmberg '64 in [4] Hojerslev [65] Stedmon et al. [77]

UV-B ocean data Pure seawater Pure seawater Sargasso Sea

East Mediterranean Central Equatorial Pacific West Mediterranean Gulf of Mexico, USA Red Sea (Gulf of Aqaba) Western Greenland Orkney - Shetland Antarctica (61°S) Arctic polynya Coastal Japan (3/98-3/99) Tropical, near coral reefs Gulf of St. Lawrence Gulf coast, Florida USA German Bight, North Sea Skagerrak Kattegat Baltic

Delaware Bay mouth USA 320 Yellow Sea

320 320 320 310 310 310 310 310 310 310 310 320 320 320 320 310 310 310 310 310 310

0.0200

0.084

B. Lake measurements

UV-A lake data (Pure water) (Pure water) (Pure water) (Pure water) (Pure water)

375 380 380 380 380

0.045

0.0072

0.010 0.010

(Pure water)

Crater Lake, Oregon USA

380 380

0.022

Morel 1974, in [6]

0.094

Smith & Baker [18]

0.069

Hejerslev 1985 in Aas et al. [117]

0.15

H0jerslev [65]

0.15

Jerlov [95]

0.15

Smith & Baker [8]

0.16-0.43

Smith & Baker [8]

0.18

Smith & Baker [8]

0.19

11 uz '93 in [10]

0.19-0.21

Hcjerslev [65]

0.39

Hejerslev [65]

0.21-0.23

Helbling et al. [44]

0.39-0.83

Belzile et al. [48]

0.18-0.98,0.52a

Kuwahara et al. [108]

0.39-1.5

Dunne & Brown [111]

0.68-2.0

Kuhn et al. [57]

0.8

Smith & Baker [8]

0.53-5.0

Hajerslev [65]

0.59-1.20

Hojerslev [65]

1.2-2.4

Hojerslev [65]

3.0-3.5

Hojerslev [65]

Vodacek et al. [23]

32-37

Hojerslev '88 in [10]

Clark & James '39 in [4] Morel '74 in [3], see also [58] Sogandares & Fry [112] Pope & Fry [101] 0.018 Hargreaves (unpub. Crater Lake

0.004-0.009 0.022 Hargreaves (unpub. Crater Lake,

Site

B. Lake measurements (cont.) Crater Lake, Oregon USA

Crater Lake, Oregon USA

380

Lake Vanda, Antarctica

380

12 Lakes, S. Argentina

380

Lake Tahoe, USA

380

High lake, Austria

380

14 Lakes, NE USA

380

L. Giles, PA USA

380

4 Arctic lakes, Canada

380

L. Biwa, Japan

380

18 Subarctic lakes, Canada

380

7 Lakes, Canada

380

20 Lakes, Colorado USA

380

San Vincente Reservoir,

380

San Diego, CA USA

L. Lacawac, PA USA

380

13 Lakes, Alaska USA

380

UV-B lake data

(Pure water)

313

(Pure water)

320

(Pure water)

320

(Pure water)

320

(Pure water)

320

Crater Lake, OR, USA

0.0153

0.016-0.33

0.16-32

0.16

0.67-16

2.69

Hargreaves, Larsen, Girdner (unpub.) Crater Lake, 0-15 m, Jun-Jul '96-99 Tyler & Smith [110] Vincent et al. [68] Morris et al. [60] Smith et al. [99] Sommaruga & Psenner [63] Morris et al. [60] Morris et al. [60] Laurion et al. [61] Belzile et al. [49] Laurion et al. [61] Scully & Lean [33] Morris et al. [60] Tyler & Smith [110]

1.45

0.084

0.006-0.017

0.04

0.092 0.050

Boivin et al. [118] Morel 1974 in [6], see also [58] Quickenden & Irvin [113]b Hargreaves, unpub. (Crater L. 8/01)

Smith & Baker [18] Hargreaves, unpub. (0-20 m, 8/01, divided by D0=1.13)

Lake Vanda, Antarctica

320

0.055

Vincent et al. [68]

Crater Lake, OR USA

320

0.051-0.71

Hargreaves et al. (unpub.) 0-15

Jun-Jul '96-99

12 Lakes, S. Argentina

320

0.14-6.5*

0.14-7.7a

Morris et al. [60]

High elev. lake, Norway

320

0.17

Hessen [114]

High elev. lake, Austria

320

0.17-0.26

Sommaruga & Psenner [63],

seasonal

11 Lakes (rocky) Alps & Pyr 320

0.21-1.8a

0.17 2.5a

Laurion et al. [62]

Lake Giles, PA USA

320

0.25-1.6

Hargreaves & Moeller, unpub.c

320

0.23

0.32

Morris et al. [60]

320

0.1-0.7

0.55-1.3

Ayoub et al. [46]

14 Lakes, NE USA

320

0.23-165"

0.32-67a

Morris et al. [60]

10 Lakes (trees) Alps & Pyr

320

0.48-4.6

0.60-5.7

Laurion et al. [62]

4 Arctic lakes, Canada

0.75-7.9

Laurion et al. [61]

5 meadow lakes, Alps & Pyr

320

0.74-3.1

0.87-4.3

Laurion et al. [62]

L. Biwa, Japan

320

0.85-6.9

1.1-14

Belzile et al. [49]

7 Lakes, Canada

320

1.1-21.6

Scully & Lean [33]

18 Subarctic lakes, Canada

1.7-41

Laurion et al. [61]

20 Lakes, Colorado USA

320

2.0-17

2.8-37

Morris et al. [60]

13 Lakes, Alaska USA

320

5.8-27

7.1-48

Morris et al. [60]

Lake Lacawac, PA USA

320

6.0-19

Hargreaves & Moeller, unpub.c

320

5.8

7.8

Morris et al. [60]

320

1.2-10

10-16

Ayoub et al. [46]

a Error in either aCDOM, ap, or Kd likely because aCDOM + ap+ aw exceeds Kd.

bAfter conversion of decadic beam attenuation value into loge (value x 2.303), the spectral attenuation (280-320 nm) was exponentially regressed against wavelength to estimate cw320, then aw320 was computed by subtracting scattering coefficient for pure water (bj. c Seasonal range ofKd320, 1993-2001.

(the latter from water samples). Then a regression of Kd totai versus [DOC] is computed to yield the DOC-specific diffuse attenuation factor (K*DOc) from the regression slope. This approach can also be used to estimate other specific attenuation factors (X*Chl, X*Tripton) and a similar approach could estimate specific absorption factors as well. In early attempts to estimate K*cdom and Kw Smith and Baker [18] subtracted modeled phytoplankton attenuation spectra from Kd,total calculated from natural waters low in CDOM to estimate Kw. Baker and Smith [43] subtracted modeled phytoplankton attenuation spectra (developed as described above) and Kw from Xd,Total f°r coastal waters to estimate Kd,cDOM- Various non-linear models have been developed to relate absorption and attenuation of algal cultures and natural phytoplankton to chlorophyll a concentration [5]. There is some information on UV attenuation by algae [44-51] but numerous studies have focused only on visible wavelengths [30,38,52-56].

The use of partial attenuation coefficients in bio-optical models has been criticized because the measured value of Kd does not depend solely on the natural waters [31]. AOP's such as Kd vary with sun angle, sky conditions, and depth, although some [26] have argued that the effects are small if the sun elevation is reasonably high and thus Kd can be used as a quasi-inherent property. Gordon [17] has established through optical modeling that Kd can be converted into a quasi-inherent optical property, for near-surface conditions at least, if it is first adjusted to remove atmospheric effects using equation (11). The procedure is to measure Kd near the surface or average Kd over 0-10% attenuation depth, and then multiply Kd by /¡d0. The adjusted Kd can be expected to respond proportionally to changes in concentration of absorbing substances. Gordon's method for determining I/fid,o (for which he used the symbol D0) applies to calm (flat) surfaces of case 1 waters [17]:

where/direct and /diffuse are the fractions of incident irradiance in the direct rays from the sun and in the indirect skylight respectively. The value of cos(0w), the angle from the zenith of direct sunlight just beneath the surface after direct sunlight has been refracted from its incident angle (cos(0a)) by passing through the horizontal water surface, can be determined from Snell's Law [6]. For light passing from air to water, cos(0w) = cos(6a)/(nJnw), where njnw is the ratio of refractive indices for air and water (nominally 1.33 in the visible wavelengths but is within 0.3% of 1.345 over the wavelengths 300 nm to 400 nm). Gordon's equation is almost identical to a "mean pathlength" equation derived independently by Zepp and Cline [15] to determine the amount of light absorbed in a vertical metre of water column based on laboratory measurements of absorption coefficients and modeled or measured incident sunlight. Gordon [17] suggests simple adjustments of D0 for windy conditions that cause surface waves. He recommends using either Kdfi/D0 or Kd,o-io%/A) whenever the objective is to compare inherent optical properties of natural waters. This approach has been mentioned rarely in UVR studies (but see [57]). It will be used later in this chapter to estimate spectral attenuation by pure water.

The values for /djrect and ./diffuse vary with atmospheric conditions (e.g., aerosols), sun angle, and wavelength. They can be determined with simple field measurements [17]: a vertically oriented radiometer with cosine sensor records full sun and sky, then is partially shaded to block only the direct irradiance from the sun. The ratio £d,shaded/£d,fuii is /diffuse» while /direct is (1 -/diffuse)- Figure 4A shows summer values for /diffuse for extremely clear air at Crater Lake, Oregon (1882 m elevation) and the hazier air over Lake Lacawac, Pennsylvania (400 m elev.) on clear "blue sky" summer days. Under overcast conditions and low solar elevation the value of/djffUse approaches 100% and Dq (Figure 4B) approaches 1.2. At high solar elevation the longer UV wavelengths become more direct, more so

100%

100%

Irradiance Attenuation

Solar Zenith Angle (degrees)

1.30

PA 305 nm PA 380 nm CL 305 nm CL 380 nm

Solar Zenith Angle (degrees)

1.30

1.25

1.05

1.25

1.05

Solar Zenith Angle (degrees)

Figure 4. Examples of direct and diffuse solar irradiance and a correction factor for diffuse path length in Kd measurements (Hargreaves, unpublished). (A) Diffuse fraction of irradiance as a function of solar zenith angle during summer, 1996, L. Lacawac, Pennsylvania (41.3°N) and August 2001, Crater Lake, Oregon (42.9°N). (B) Calculated correction [17] to remove effects of irradiance field from near-surface diffuse attenuation (Kd) measurements, based on data in part (A).

PA 305 nm PA 380 nm CL 305 nm CL 380 nm

Solar Zenith Angle (degrees)

Figure 4. Examples of direct and diffuse solar irradiance and a correction factor for diffuse path length in Kd measurements (Hargreaves, unpublished). (A) Diffuse fraction of irradiance as a function of solar zenith angle during summer, 1996, L. Lacawac, Pennsylvania (41.3°N) and August 2001, Crater Lake, Oregon (42.9°N). (B) Calculated correction [17] to remove effects of irradiance field from near-surface diffuse attenuation (Kd) measurements, based on data in part (A).

in the clean dry air over Crater Lake than over L. Lacawac in the N.E. USA. For this range of conditions D0 varies from 1.09 to 1.26. The value of/diffuse at 320 nm during summer in the Gulf of St. Lawrence (latitude 47-50°N, solar zenith angle 24-54°) ranged from 48-72% [57]. If Kd measurements are made with SZA < 50°, the effect of incident diffuse and direct light on Kd will always be less than 20%. At the Latitude (41.3°N) of L. Lacawac, the SZA will be less than 50° if measurements are made within 3 hours of solar noon between the dates of 17 April and 26 August. Before 26 February and after 14 October the SZA will be greater than 50° even at solar noon.

The first spectral models, developed for marine systems by Prieur and Sathyendranath [39] and Baker and Smith [43], emphasized phytoplankton optics, although they included components for attenuation by CDOM, phytoplankton, and water. From these and subsequent studies (reviewed by Morel

[58]) has emerged the importance of variation in phytoplankton attenuation per unit of chlorophyll, hereafter referred to as "specific phytoplankton attenuation". While the Baker and Smith [43] model was optimized for UV-B wavelengths and included a CDOM component, this was based on scant data relating optical absorption to the concentration of DOM. When CDOM was detected in the open ocean it was assumed to covary with phytoplankton and often modeled without direct measurement. A recent summary of ocean bio-optical models in Morel [58] does not specifically address UV wavelengths. Yentsch and Phinney

[59] mention that the blue and UV regions of algal spectral absorption are the most variable, especially when algae produce UV-absorbing mycosporine-like amino acids (MAA's).

Freshwater optical models relevant to UV attenuation have been developed [33,60,61]. These generally find CDOM is the most important factor, although absorption by phytoplankton and detrital particles [46,49,62,63] or scattering by suspended solids [49] has also been important in some cases. From recent UV investigations in freshwaters has emerged the importance and variability of the CDOM absorption coefficient scaled per unit of DOC concentration, hereafter referred to the DOC-specific absorption factor (a*^Doc)- As is the case for marine systems, there are relatively few measurements of UV attenuation or absorption by freshwater phytoplankton [46,49].

Despite an emerging consensus on chlorophyll-specific absorption for visible wavelengths [58], current models appear inadequate to describe the highly variable UV attenuation exhibited by phytoplankton. With regard to DOC and CDOM, several studies (described later) have revealed regional patterns relating Kd and CDOM to [DOC] in relation to climate, precipitation, river discharge, and watershed properties. Integration of patterns and processes to explain UVR penetration into aquatic systems has been lacking.

3.3.1 Role of CDOM (Kirk type G waters)

CDOM was central to the earliest optical models for seawater (Jerlov, 1968 [4]). Interest in CDOM in freshwater goes back to the early 1900's (Shapiro, 1957

[64] cites a 1908 paper on humic substances precipitated from Finnish lakes). It has long been recognized that the penetration of ultraviolet radiation (UVR) depends on concentrations and optical qualities of dissolved organic matter (DOM) [4,65-67], In non-turbid waters where UV attenuation is high, absorption by CDOM easily surpasses attenuation by other components. Somewhat surprising is the recent finding that CDOM tends to be the most predictive optical component of UV attenuation even in low DOC systems. UV bio-optical models developed recently for lakes reflect a dominating effect of DOC and CDOM, with only a small or immeasurable contribution by phytoplankton in surface waters [33,60,62,68].

Lakes matching Kirk's "Type G" typically have moderate to high levels of DOC. Figure 5 shows an example the close relationship between aCdom380 measured in a spectrophotometer and Xd(38o calculated from field measurements of underwater irradiance at 380 nm in a moderately clear lake (L. Giles, PA, listed in Table 1). The two signals change together and increase below the mixed layer (9.5 m on this date in early September). Attenuation and acdom decrease substantially in the mixed layer of this lake during summer months when rates of UVR photobleaching of CDOM exceed the rates of CDOM production and import [22].

In the older literature CDOM has been called "yellow substance", "gelbstoff", and "gilvin'. It is considered to be a mixture of compounds chemically characterized as humic and fulvic acids [6,10,69,70]. Figure 6 shows typical absorption spectra for the CDOM in water from two mid-latitude lakes that has passed through a fine glass fiber filter. The samples are from the depth of the mixed layer of two lakes surrounded by mixed conifer-deciduous forest: L. Giles (watershed soils well-drained) and L. Lacawac (bordered 50% by a sphagnum bog). The values come from absorbance (Sample,!) recorded in a spectrophotometer using a quartz cuvette, corrected by subtracting (optically or numerically) the value for highly purified water, Awater^, and the cuvette to compute AcAom^. We assume negligible absorbance by inorganic dissolved matter such as ferrous iron, nitrite, or sulfate ions. An adjustment is often made to correct for instrument baseline drift and optical scattering within the cuvette that would otherwise cause errors in estimating Acdomj/l. A long reference wavelength (Abase) should be chosen for the correction of offset (e.g., > 650 nm) where absorption by CDOM is assumed nil; 700 nm or 775 nm are particularly useful in avoiding the strong temperature effects on the absorption of pure water that reach a peak at about 750 nm [71,72] but a longer wavelength (up to 900 nm) may be needed for highly concentrated CDOM. Suggestions for subtracting a spectral scattering term [67,73] from measured Ak are derived from empirical models showing scattering from small particles in nominally-filtered natural waters varies in proportion to a-1. Depending on the sample filtration and optical configuration of the instrument and cuvette, spectral scattering may affect the measurement; the suggested correction, >4cdom,A = ^A«xiom_xaw-^base(4ase/^)» has not been rigorously tested. Acdom^ is then converted into a (Napierian) absorption coefficient, aCdom, i (units m_1):

Acdom Figure
0 25 50 75 100

Figure 5. (A) UV-A CDOM absorption and diffuse attenuation coefficients (380 nm) for L. Giles (2 September 1999) determined using binned data from PUV-501 profiling radiometer and laboratory analysis of GF/F filtered water samples in 10 cm quartz cuvettes and Shimadzu UV160U spectrophotometer. (Hargreaves, unpublished). (B) Supplementary data from the PUV-501 profiling radiometer: water temperature and chlorophyll index (upwelling natural 685 nm fluorescence, NF, divided by down welling PAR). In (B) The thermally mixed zone above 9.5 m corresponds to the optically mixed zone in (A). Algal biomass increases with two peaks near 10 m and 17 m that correspond to optical changes in (A) (Hargreaves, unpublished).

Figure 5. (A) UV-A CDOM absorption and diffuse attenuation coefficients (380 nm) for L. Giles (2 September 1999) determined using binned data from PUV-501 profiling radiometer and laboratory analysis of GF/F filtered water samples in 10 cm quartz cuvettes and Shimadzu UV160U spectrophotometer. (Hargreaves, unpublished). (B) Supplementary data from the PUV-501 profiling radiometer: water temperature and chlorophyll index (upwelling natural 685 nm fluorescence, NF, divided by down welling PAR). In (B) The thermally mixed zone above 9.5 m corresponds to the optically mixed zone in (A). Algal biomass increases with two peaks near 10 m and 17 m that correspond to optical changes in (A) (Hargreaves, unpublished).

where i is the cuvette path length (in meters). In the chemical literature the decadic absorption coefficient {a = A/f) is sometimes reported. Although the units are identical (m_1), decadic units must be multiplied by 2.303 (the natural logarithm of 10) to be numerically equivalent to the standard (Napierian) exponential units from equation (12).

CDOM has an absorption spectrum that is nominally exponential in shape [4] and has been frequently characterized by the two exponential parameters,

Acdom Figure

Wavelength (nm)

Figure 6. Spectral slope of CDOM from two lakes (Hargreaves, unpublished). S (nm-1) is an exponential parameter from the relationship acdom ;: = ae ~SI. The value of S can be computed as the absolute value of the slope when Ln(acdoml) is plotted against wavelength over the UV and blue range. Such plots tend to be linear over UV wavelengths when DOC is high (upper curve) but can sometimes be separated into a steeper UV-B slope (280-320 nm) and shallower UV-A slope (320-380 nm) when substantial photobleaching has occurred (lower curve). These lake samples are from the upper mixed layer, June 2001 (particles removed with GF/F filter, Shimadzu UV-1601 spectrophotometer, 10 cm quartz cuvette, low DOC deionized water spectrum subtracted; small glitch at 345350 nm in lower curve is caused by spectrophotometer imperfection).

Wavelength (nm)

Figure 6. Spectral slope of CDOM from two lakes (Hargreaves, unpublished). S (nm-1) is an exponential parameter from the relationship acdom ;: = ae ~SI. The value of S can be computed as the absolute value of the slope when Ln(acdoml) is plotted against wavelength over the UV and blue range. Such plots tend to be linear over UV wavelengths when DOC is high (upper curve) but can sometimes be separated into a steeper UV-B slope (280-320 nm) and shallower UV-A slope (320-380 nm) when substantial photobleaching has occurred (lower curve). These lake samples are from the upper mixed layer, June 2001 (particles removed with GF/F filter, Shimadzu UV-1601 spectrophotometer, 10 cm quartz cuvette, low DOC deionized water spectrum subtracted; small glitch at 345350 nm in lower curve is caused by spectrophotometer imperfection).

"spectral slope" (S) and reference absorption (acdom) Aref):

where (X — Aref) is the difference between the desired wavelength and the reference wavelength over the range 350-700 nm [67]. The value of S (units, nm-1) is typically computed from a linear regression of Ln(acdom) versus wavelength. The waveband used in numerous published reports has varied but frequently covers the range from UV-B through 700 nm [74]. Bricaud et al. [67] likely chose 350 nm as their lower limit after observing nonlinear regions at shorter wavelengths in their published spectra of open ocean CDOM. When acdom is high, spectra tend to be exponential from below 300 nm well into the visible spectrum (upper curve in Figure 6). At lower levels of acdom, typically following substantial exposure to sunlight, the spectra become more irregular below 350 nm. Under these circumstances, separate values for S may be calculated for the UV-B (280-320 nm) and UV-A (320-340 nm) ranges of the spectrum ([75] and K. Mopper, personal communication), as shown in the lower curve of Figure 6.

Several authors [74,76,77] have recently suggested computing 5 using a nonlinear regression technique that gives less weight to longer (and noisier) wavelengths. They assume that S is uniform throughout the range of wavelengths included in the nonlinear regress (otherwise the nonlinear approach would bias S toward the slope at the shortest wavelengths where acdom has the greatest value) and they suggest that the nonlinear technique avoids a bias caused by log-transformation of instrument noise present at the longer wavelengths. This author strongly recommends the more conventional log-linear regression with a caveat to consider the following guidelines in order to compute S accurately:

• The spectrophotometer must be completely stable (e.g., warmed up for at least an hour at a stable room temperature; this is especially important with the diode array variety of instrument).

• A single carefully-cleaned quartz cuvette should be used for both blank and sample scans (referenced to air in the reference beam) with numerical correction for the blank during post processing.

• If ultrapure water is not available (stored water can develop substantial absorbance and many water purification methods leave a UV-absorbing residue) it may be preferable during post-processing to adjust the measured blanks recorded in the field with a file recorded earlier with the best quality water using the same instrument and cuvette.

• The initial selection of wavelength range for S should consider the shape of the spectrum (mentioned above).

• The baseline should be carefully adjusted to zero during post-processing at a non-absorbing waveband (e.g., 775-800 nm); this should be accompanied by visual inspection of a linear graph of aCdom versus wavelength (with scales expanded to show detail, e.g., ±0.05 m-1 for the range 600-800 nm).

• The longer wavelength should be revised if necessary so as to avoid wavelengths near the instrument limit of detection (typically A = ±0.001 after subtraction of the blank and ;4base) where noise can return negative acdom values. The eifect of the baseline adjustment (described above) is to ensure that the noise is symmetrical with respect to zero, but if half the noise values are negative (and thus automatically excluded from the regression), the value of S will be underestimated to an extent that depends on how many of these "noise" data are included.

A similar exponential treatment has also been applied to spectral modeling of UV diffuse attenuation coefficients for natural waters [57,61,76] but this seems ill-advised unless the absorption spectrum of phytoplankton or other particles is insignificant or has been observed to follow the same exponential pattern as CDOM that may be present.

DOM molecules are the chemical basis for CDOM optical absorption (acdom, m_1), but because of molecular variations in DOM and its chemical environment, DOC-specific absorption, (a*ooc) and DOC-specific attenuation (Kd*DOC) vary in natural waters. Both have units of m_1[g m-3]-1, typically simplified to m2 g The optical properties of CDOM are known to vary with the source, including type of watershed vegetation and in situ production [6,74,78-80] and modification in the water column [22,23,74,81].

Although variation in CDOM specific absorption has been recognized for some time there is disagreement among researchers on patterns of variation with DOC concentration. Currently the scaling factor of choice is DOC concentra tion, rather than DOM concentration, because of standardization in methods for measuring the carbon content of DOM [40]. A linear relationship between a*CDOM and DOC has been assumed to date [23,43] in marine systems but given the pattern of variation in DOC-specific attenuation (described later), a nonlinear model is proposed for DOC-specific CDOM absorption:

Values for a*Doc,320 computed from available data (Table 2) range from 0.3 to 3.2 m2 g-1 for both marine and freshwater sites (converted to 320 nm from other wavelengths as needed using reported S values and equation (13)). Data from a study of 61 lakes [60] reveal that DOC-specific absorption increases together with DOC concentration with an exponent of 1.12 and oc*doc,320= 1-2 m-1 (Table 2). CDOM from surface waters in the Gulf of Mexico was concentrated and separated into fulvic and humic fractions [69] to indicate their relative contributions to absorption. Fulvic acids have a much lower specific absorption than humic acids. While shifting proportions of the fulvic and humic fractions in

Table 2. Variations in DOC-specific absorption of CDOM (otj), using acdom320 = DOCx where units are m-1 for acdom320 and g m-3 for DOC concentration

*«1 '«5

t* r2

DOC

Region

Data from

Coastal marine DOC (extracted from Gulf of Mexico surface water)

0.06 0.3

(1)

extracted fulvic acids

Carder et al. [69]

0.5 2.5

(1)

extracted humics

Carder et al. [69]

Coastal marine CDOM

0.3 1.4

(1)

1.6

Japanese coastal, 13 months

Kuwahara et al. [108]

0.7 3.4

(1)

0.3-3.8

Danish coastal

Stedmon et al. [77]a

1.3 6.5

(1)

Danish coastal

Nyquist in Hajerslev [65]b

1.9 9.5

(1)

6.2

estuarine salt marsh

Miller & Moran [109]

Coastal water receiving Delaware River

discharge, comparing seasons

2.1 10.6

(1)

0.8-1.7

Spring, water column mixed

Vodacek et al. [23]

0.6 2.8

(1)

1.3-1.5

August, surface layer

Vodacek et al. [23]

Mid-latitude

lakes, comparing seasons

1.8 9.0

(1)

0.7

Spring, L. Giles

Morris & Hargreaves [22]

0.3 1.3

(1)

1.0

Summer surface, L. Giles

Morris & Hargreaves [22]

3.2 16.0

(1)

4.4

Spring, L. Lacawac

Morris & Hargreaves [22]

1.4 7.0

(1)

5.7

Summer surface,

Morris & Hargreaves [22]

L. Lacawac

Mid-latitude

lake surveys

1.2 7.4

1.12 0.79

1-24

61 Lakes, mid-latitudes

Morris et al. [60]

0.7* 10.6

1.70 0.91

4-22

30 Lakes, Northern USA

Reche et al. [115]c

0.8 10.8

1.58 0.90

0.1-15

85 Adirondack lakes

Bukaveckas &

Robbins-Forbes [24]c

* a, is equivalent to a*DOC320; a5 is aCdom32o computed for DOC = 5 g m~3 t(l) indicated for x where a proportional scaling pattern for DOC has been assumed. a TOC used to calculate specific absorption. b DOM g m~3 instead of DOC g m-3c a, and a2 adjusted from 440, 340 or 300 nm to 320 nm using equation (13) and 5 = 0.01565.

the CDOM source may cause some of the variation in specific absorption of natural CDOM, other factors include changes in pH, ionic composition, and photobleaching. While the absorption by humic acids is stable over a wide range of pH (6-11), that of fulvic acids is not [82]. Stewart and Wetzel [83] studied humic substances in experimental leachate of decaying plants and DOM from 55 lakes of southwestern Michigan. They concluded that calcium concentration affects average molecular size and this in turn affects DOC-specific absorption. Vodacek et al. [23] attributed the decline in specific absorption for CDOM in the coastal plume from the Delaware River to photobleaching in the surface mixed layer. Specific absorption at 320 nm was reduced from 2.1 m2g_1 during winter conditions of high river flow and low irradiance to 0.6 m2 g-1 in offshore stratified surface waters during high summer irradiance (August). Morris and Hargreaves [22] observed similar declines from spring to summer in Kd32o and acdom)320 for several lakes differing in their CDOM source (one surrounded by a sphagnum bog, the other by well-drained soil). They established a major causal role for photobleaching through experimental exposure of particle-free lake water to different wavebands of the solar spectrum (Tables 2 and 3 and Figure 5A). In some cases a sampling artifact appears to interfere with measurements of DOC-specific absorption. In the mountain lake study by Laurion et al. [62], surface acdom,320 was generally reduced compared to deeper in the water column, a pattern that may have been caused by photobleaching or surface inhibition of phytoplankton. However, measured acdom)320 was greater than Xd32o for 73% of lakes with low DOC and rocky watersheds and 21 % of lakes with higher DOC and forested or meadow-covered watersheds. The authors suggested that UV-screening pigments (MAAs discussed in the next section) known to be present in the phytoplankton may have leaked out of cells during filtration. This problem might partially explain a similar anomaly in several of the 61 lakes sampled by Morris et al. [60].

CDOM exhibits fluorescence by emitting blue light after absorbing UVR. The maximum fluorescence response per unit of absorbed energy occurs when coastal CDGM is excited at 380 nm [84]. Although CDOM fluorescence is sometimes well-correlated with UV attenuation [61] and CDOM absorption [23], it has also been a somewhat variable predictor of variations in UV attenuation or absorption in other cases when the CDOM source varies [62,83]. DOC-specific fluorescence appears to vary both among and within lakes. As in the case of CDOM absorption, variations in fluorescence properties of DOC are likely to reflect differences in source as wells as a history of photochemical and biological processing. CDOM from terrestrial and marine sources can be distinguished from each other using three-dimensional excitation-emission fluorescence spectra [85,86]. McKnight et al. [80] showed that for excitation at 370 nm the CDOM emission peak of an acidified filtered water sample would occur at 442-448 nm for microbially-derived fulvic acids and at 457-461 nm for plant-derived (terrestrial) fulvic acids. Their fluorescence index (Em450: Em500) based on these differences yielded 1.9 for microbial-derived DOM and 1.4 for terrestrial-derived DOM. This index is reported to be affected by environmental acidification, which changes DOC composition, but not by photobleaching [87].

Table 3. Relationship between KdjCDOm (m_1) and DOC (g m~3) using Kdm — Kw32o = k{ DOCx; Published Kd320 and DOC data were used with Kw320=0.04 to fit k{ and x (least squares regression) where k{ = KCDOM 320 at DOC = l g m~3 and k5 = KCDOM m at

Table 3. Relationship between KdjCDOm (m_1) and DOC (g m~3) using Kdm — Kw32o = k{ DOCx; Published Kd320 and DOC data were used with Kw320=0.04 to fit k{ and x (least squares regression) where k{ = KCDOM 320 at DOC = l g m~3 and k5 = KCDOM m at

A, ks x

r2

DOC

Region

Data from

Coastal & Marine

1.3 7 (1)

Ocean (based on DOM)

Hejerslev [65]

0.8a 4a (1)

2.5

St. Lawrence Estuary, Stn 24 Kuhn & Browman [57]

0.4 (1)

1.7

Coastal Japan, 8 m. w/rain

Kuwahara et al. [108]

0.2 (1)

1.5

Coastal Japan, 5 m. dry

Kuwahara et al. [108]

43 Canadian prairie lakes, ponds, wetlands, including saline systems

6.7* 18a 0.61

041

24- 80

52°N Wetlands, ponds

Arts et al. [34]b

1.4* 5 0.76

0.50

4-156

52°N Lakes

Arts et al. [34]b

Freshwater: photobleaching effects in

surface waters of lakes

2.0 (1)

0.7

41°N, L. Giles, spring

Morris & Hargreaves [22]

0.3 (1)

1

41°N, L. Giles, summer

Morris & Hargreaves [22]

19 (1)

4.4

41°N, L. Lacawac, spring

Morris & Hargreaves [22]

9 (1)

5.7

41 °N, L. Lacawac, summer

Morris & Hargreaves [22]

Freshwater: lakes differing in land cover and altitude

0.6 3a 0.89

0.54

0.2-1

Alps & Pyr., rocky

Laurion et al. [62]c

1.4 7a 0.81

0.68

0.4-4

Alps & Pyr., trees, meadows

Laurion et al. [62]c

1.0 8a 1.33

0.81

0.2-4

Alps & Pyrenees, combined

Laurion et al. [62]c

Freshwater: high latitude lakes

0.3 10 2.08

0.93

0.3-11

Arctic, subArctic, Antarctic

Vincent et al. [68]

0.3 13a 2.43

0.99

0.3-1

Antarctic

Vincent et al. [68]

0.4a 10 2.06

0.86

2-11

Sub-Arctic Canada

Laurion et al. [61]

0.7a 10 1.62

0.78

4-11

Alaska, USA

Morris et al. [60]d

0.8 6 1.28

0.97

1-5

Arctic Canada

Laurion et al. [61]

Freshwater: mid-latitude lakes

0.6 7 1.62

0.91

0.5-8

41-51°N. USA & Canada

Scully & Lean [33]d

0.6 7 1.57

0.90

1-24

41° N Pennsylvania,US A

Morris et al. [60]d

1.7 13a 1.24

0.78

0.4-3

40°S, Argentina

Morris et al. [60]d

3.1 12 0.83

0.60

0.8-10

Colorado, USA

Morris et al. [60]

1.5 10 1.20

0.84

0.4-24

Average, mid-latitudes

Morris et al. [60]d

a /c, or fc5 extrapolated beyond measured range of DOC.

b Waveband used for Kd was UV-B (280-320 nm) instead of narrow-band 320 nm. c Excluding lakes when C/L > 50 (C = catchment area, L=lake area). d Excluding data when DOC low relative to phytoplankton (DOC/chl < 1400, units, g m-3).

Figure 7 shows the emission spectrum of CDOM fluorescence for samples excited at 365 nm from two lakes having DOC in the range of 1-5 mg 1_1 (Hargreaves, unpublished). The small peak is Raman scattering by water molecules (centered at 417 nm, a shift in wavenumber of —3400 cm-1 from the excitation wavenumber, where wavenumber is 107 divided by wavelength in nm). The Raman water peak can be used to provide scale calibration of fluorescence emission spectra [88,89]. The broad peak in Figure 7 is contributed predominantly by the fulvic acid fraction of DOM [80]. The peak wavelength and

Emission WL (nm)

Figure 7. CDOM fluorescence of water from two lakes (Hargreaves, unpublished): emission scans for excitation at 370 nm (Shimadzu 551 fluorometer), before and after subtraction of water blank. Samples: deionized water (DIW), L. Giles water (ca. 1 g m-3 DOC), L. Lacawac water (ca. 5 g m~3 DOC) The Raman scattering peak at 417 nm represents a shift in wavenumber by 3400 cm-1 from the excitation wavenumber. The broad peak is contributed predominantly by the fulvic acid fraction of DOM. The peak wavelength and fluorescence index ratio for these samples (L. Giles, 452 nm peak and ratio = 1.5; L. Lacawac, 455 nm peak and ratio = 1.4) suggest a slight difference in CDOM

Emission WL (nm)

Figure 7. CDOM fluorescence of water from two lakes (Hargreaves, unpublished): emission scans for excitation at 370 nm (Shimadzu 551 fluorometer), before and after subtraction of water blank. Samples: deionized water (DIW), L. Giles water (ca. 1 g m-3 DOC), L. Lacawac water (ca. 5 g m~3 DOC) The Raman scattering peak at 417 nm represents a shift in wavenumber by 3400 cm-1 from the excitation wavenumber. The broad peak is contributed predominantly by the fulvic acid fraction of DOM. The peak wavelength and fluorescence index ratio for these samples (L. Giles, 452 nm peak and ratio = 1.5; L. Lacawac, 455 nm peak and ratio = 1.4) suggest a slight difference in CDOM

source [80], fluorescence index ratio for these samples (L. Giles, 452 nm peak and ratio = 1.5; L. Lacawac, 455 nm peak and ratio = 1.4) suggest a slight difference in CDOM source.

Values for i*doc,320 ranging from 0.3 to 3.8 (Table 3) have been calculated from published UV attenuation data for both marine and freshwater sites (converted into 320 nm as needed using reported S values and equation (14)). The attenuation of pure water was subtracted (K^32o discussed below) and sites with high chlorophyll relative to DOC (DOC/chl < 1400; units gm-3) were excluded where noted. Although some Kd s may be elevated by phytoplankton and other particles that attenuate underwater irradiance, the predominant source of variation in Kd*Doc,320 DOC quality. While Scully and Lean [33] reported no effect of phytoplankton in their optical model (chlorophyll ranged from 1.3-33 mgm~3), a reassessment shows that when lakes with a low ratio of DOC to algal chlorophyll (DOC/Chi <1400; units g m~3) were excluded, there was an improved r2 for the regression of Kd versus DOC and a reduced K*DOC. The regressions of Morris et al. [60] were also improved by reanalysis in which lakes with low DOC/Chi ratios were excluded, although only three lakes (out of 64 sampled) had chlorophyll levels exceeding 5 mg m~3.

A linear model for scaling Xd*DOc,320 to DOC concentration was adopted by Baker and Smith [43], who used a constant value of Xd*DOc,320 = 1-3 in their bio-optical model but cited a range of values from as low as K*DOc,320 = 0.75 for clear Sargasso Sea water to as high as K*doc,320 = 6.2 for a coastal site. A linear relationship between K*DOC and DOC has been assumed in several studies [61,62]. A power relationship between K*cdom and DOC (similar to that describe for CDOM absorption above) has been assumed by others:

where .K*doc,320 is the DOC-specific attenuation of CDOM at 320 nm.

The power model for relating UY attenuation to DOC was used by Scully and Lean [33], Morris et al. [60] and Vincent et al. [68]. Vincent et al. [68] found an unusually strong relationship between UV attenuation depths and DOC for high-latitude lakes (replotted in Figure 8 as Xd320 - -KW320 versus DOC), a relationship with an exponent much greater than one. Arts et al. [34] used both linear and power models (but preferred the power relationship) to relate wideband UV-B attenuation to DOC in prairie lakes of Canada. As reported by Arts et al. [34] a reasonable fit was obtained with a power model, but when all ponds and wetlands are grouped (including three with high salinity), they have exponents less than one, similar to the fresh and saline lakes (Table 3). While attenuation for the UV-B waveband (derived from detailed spectral irradiance) is too broad to serve as a rigorous attenuation coefficient, it is an index of Kdm that is largely a function of DOC concentration and quality. Although Arts et al. [34] concluded that UV-B irradiance penetrates more deeply into saline water bodies that it does into freshwater systems of similar DOC concentration, this reanalysis of their data supports a somewhat contrary conclusion (Figure 9). Saline systems tend to have higher [DOC] than freshwater systems (probably because evaporation causes DOC to become more concentrated). For similar [DOC], the greater penetration is actually observed in lakes (especially the large,

Irradiance Attenuation

Figure 8. Lake data computed from Vincent et al. [68] to show the power relationship between (Xd320 — Xw320) and DOC concentration for a range of high latitude lakes. The equation for all sites combined is (Kd320-Kw320) = 0.34DOC 208 (r2 = 0.93).

Figure 8. Lake data computed from Vincent et al. [68] to show the power relationship between (Xd320 — Xw320) and DOC concentration for a range of high latitude lakes. The equation for all sites combined is (Kd320-Kw320) = 0.34DOC 208 (r2 = 0.93).

1,000

□ freshwater lakes

▲ saline ponds & wetlands a fwponds & wetlands

1,000

Figure 9. Attenuation of UV-B irradiance in saline prairie lakes, ponds, and wetlands of Canada (52°N) from Arts et al. [34]. Saline systems tend to have higher [DOC] than freshwater systems (probably because evaporation causes DOC to become more concentrated). For similar [DOC], greater penetration of UVR is observed in lakes (especially the large, deep ones) compared to small and shallow ponds and wetlands. The equation for all ponds and wetlands (triangle symbols) is Kduv_B = 6.7 DOC061 (r2 = 0.41); for freshwater ponds and wetlands (open triangles) Kduv_B = 2.5 DOC091 (r2 = 0.56); for all lakes (squares), XdUV.B = 1.4 DOC076 (r2 = 0.50).

deep ones) compared to small and shallow ponds and wetlands. Arts et al. [34] noted this pattern and hypothesized that attenuation is lower per unit of DOC in deep lakes because their greater residence time allows for more complete photo-bleaching. Waiser and Robarts [90], studying one of these large lakes, found lower [DOC] but higher UY-B attenuation per unit of DOC in a major stream feeding the lake compared to the lake water column. Although salinity covaries with DOC in xeric regions such as this, it is a weak predictor of UV-B attenuation because DOC-specific attenuation also varies.

3.3.2 Role of phytoplankton and CDOM (Kirk type A and GA natural waters)

Phytoplankton can contribute significantly to UV attenuation in waters with moderate to low UV attenuation, especially when isolated from watershed sources of CDOM. In Case 1 oceanic waters where coastal discharge is not a source of CDOM [52] the levels of locally-produced DOC, algal pigments, and the associated microbial community are assumed to co-vary [43,58] but perhaps with a time delay between algal production and appearance of CDOM [67,69,85,91]. It can be difficult to establish the indirect contribution of phytoplankton to water column optics by their release of DOM which increases aCdom-For example, the spatial correlation between a peak in acdom just below the mixed layer of a lake (Figure 5A) and a peak in abundance of phytoplankton at the same depth (Figure 5B) are suggestive of a causal relationship, but not definitive.

Studies of optical properties of photosynthetic organisms are numerous, but relatively little has been published on the role of UV attenuation by phyto-plankton. Blough and Del Vecchio [74] summarize spatial and temporal relationships between CDOM, DOC and phytoplankton in coastal waters. Belzile et al. [49] report data from L. Biwa, Japan, in which both aCijom and [chl a] (over the range 1.5-7.5 mg m~3) are highly correlated with Kdm and Kd380. Twar-dowski and Donaghay [116] inferred from optical measurements at a coastal site the direct production of CDOM from a thin layer of phytoplankton in the water column.

A key parameter in bio-optical models is the chlorophyll-specific spectral absorption factor (a*Chu)- Chlorophyll concentration is most often measured optically after extraction from phytoplankton. In vivo methods involving measurement of fluorescence can provide a convenient index of biomass but are confounded with acclimation and species effects on calibration parameters. The Quantitative Filter Technique (QFT) is widely used to provide a measure of particulate absorption over the waveband (typically 400-700 nm) of photosyn-thetically active radiation (PAR). The method, pioneered by Yentsch [53], involves concentrating particles onto a filter and then measuring absorption on the filter in a spectrophotometer. Many

Was this article helpful?

0 0

Post a comment