Ultraviolet Radiation and Bio-optics in Crater Lake, Oregon, 2005
DISCUSSION
Spectral diffuse attenuation, Validation, and Variations over Space and Time
In Crater Lake several intercomparisons have been carried out that have confirmed its ability to measure spectral diffuse attenuation accurately under extremely transparent conditions (e.g., Figure 4). The primary limitations of the LI-1800uw in Crater Lake are the inability to accurately record wavelengths shorter than 308 nm, and the time required to complete a scan when irradiance is changing rapidly (e.g. within 15 m of the surface). Measurements of diffuse attenuation would not benefit from higher spectral resolution than its 8 nm bandwidth, in contrast to measurements of solar irradiance spectra, except in the case of the shortest UV-B wavelengths. This is because spectra of typical attenuating substances tend to vary gradually with wavelength, and because there are instrument performance tradeoffs so that increasing spectral resolution leads to decreasing sensitivity to low light levels underwater (Kirk et al., 1994). While the effective center wavelength of a broad irradiance waveband can change as a function of depth (Hargreaves, 2003), the LI-1800uw has been tested previously in lakes with substantially more CDOM than Crater Lake and proved capable of generating accurate Kd values there in comparison with BSI radiometers and a scanning radiometer that had a 2 nm bandwidth (Kirk et al., 1994).
The surface waters of Crater Lake are remarkably transparent to UV-B radiation. The relative attenuation of solar UV-B (320 nm) at a depth of 20 m in Crater Lake was 27% of surface irradiance on average (Table 1). Both the average Kd,320 of 0.062 m-1 (1996–2002) and minimum Kd,320 (August 2001, 10–18 m average) of 0.057 m-1, for the surface of Crater Lake were more transparent to solar UV-B than attenuation estimated for pure water (Kdw,320 = 0.09 m- 1, Smith & Baker, 1981). On only a few other occasions have such low values Kd,UV-B been reported for natural waters. An ice covered lake in the dry valleys region of Antarctica (L. Vanda) had low levels of phytoplankton and DOC and Kd,320=0.055 m-1 near its surface (Vincent et al., 1998). An earlier study (Goldman et al., 1967) reported that the upper water column in L. Vanda was remarkably clear and similar to pure water based on blue photometer measurements (Kd,blue=0.031 on one occasion, identical to Kd,blue average over 0–40 m for Crater Lake in July 1940, Utterback et al., 1942). In a region of the Gulf of Mexico away from the influence of Mississippi River runoff (flowing at the lowest rate in 52 years during July 1988 when measurements were made) Højerslev and Aarup (2002) reported Kd,310=0.071 m-1. This value is equivalent to Kd,320=0.061 m-1 after wavelength conversion (using the exponential equation for CDOM absorption of Bricaud et al., 1981, but without the backscatter correction suggested by Markager & Vincent, 2000, using an exponent of -0.015 derived from regressing Crater Lake spectral Kd, averaged 0–40 m on 20 August 2001, against wavelengths from 305–380 nm). Morel and Maritorena (2001) also reported extremely low values (Kd,305=0.095, or Kd,320=0.076, converted as described above) for oligotrophic regions of the tropical Pacific Ocean where chlorophyll-a varied from 0.043–0.054 mg m-3over depths 0–72 m. In each of these published reports of extremely low UV-B attenuation the authors noted the discrepancy between their measurements and the Smith and Baker (1981) estimates for Kw for pure water. Hargreaves (2003) estimated the UV-B attenuation of pure water from Crater Lake measurements; the new estimate (Kdw,320=0.045) represented a 50% reduction from the 1981 estimate. Other cases of extreme transparency probably exist where water is isolated from organic soils and algal productivity or other sources of CDOM by temperature, altitude, water currents, and strong solar UVR and where photobleaching can reduce absorption by any CDOM that is present.
From Figures 4 and 5 one can see that the surface of Crater Lake typically has the greatest transparency but measurements here are problematic because of optical noise caused by waves and ripples reflecting and redirecting sunlight (Zaneveld et al., 2001). While a comparison of irradiance (Ed,0+) just above the surface with Ed,Z at a depth of 15 or 20 m should remove much of the noise attributed to surface waves, this approach introduces uncertainty about the wavelength-specific transmittance through the air-water interface. With a profiling radiometer it is easier to compensate for this noise by lowering the instrument slowly in the upper 15 meters so that multiple readings are available for averaging (of log-transformed irradiance) within 1 m, 2 m, or larger “depth bins”. Another method (we used this to generate the 10–18 m curve in Figure 3) involves fitting polynomial regressions to Ed,Z versus Z for each wavelength of interest, taking care to use a sufficient number of significant figures for fitted parameters and to avoid extrapolation beyond the bounds of the fitted data. From the regression equations for each wavelength the specific values of Ed,Z can then be calculated at several depths close to the surface and from these Kd,Z can be calculated.
