Study

This study (Wenny et al., 1998) focuses on the attenuation of UV-B that can be attributed to aerosol optical properties. The procedure is to measure UV-B transmission and some aerosol properties in a layer of atmosphere defined by a mountain site and valley site (about 1 km vertically, 6 km horizontally) and analyze the data for aerosol optical effects and aerosol optical properties. The relationship between air mass source region and aerosol optical depth at visible wavelengths and UV-B transmission is investigated. Empirical relations between aerosol optical depth and UV-B transmission are derived to improve predictions of solar noon UV-B transmission, such as done by the U.S. Weather Service (Long et al., 1996). A novel iterative procedure employing Mie scattering and UV-B radiative transfer calculations, in conjunction with actual measurements, is employed to retrieve aerosol optical properties (single scatter albedo (SSA) and asymmetry parameter (AP)) that physically describe the amount of aerosol scattering and absorption occurring, consistent with the observed UV-B transmissions.

The novel approach taken in the present investigation also requires analysis of radiation measurements using Mie theory and radiation transfer calculations; however, the distinction between the present approach and the diffuse/direct method is that the measurement is one of atmospheric irradiance transmission through a layer of atmosphere, determined by the ratio of the irradiance (i.e., diffuse + direct) at the bottom of the layer to the irradiance at the top of the layer, which accurately defines the radiative source illuminating the layer. Moreover, the wavelengths are in the UV region, while the previous investigations were in the visible region.

This study was a step-wise pilot experiment to develop a reliable methodology for gathering an extended set of aerosol information obtained by in situ collections and by confirming the results with measurements and analysis of the radiation field. Long-term measurements are highly desirable for aerosol work because aerosol properties are so variable due to their chemical and spatial inhomogeneous nature; thereby introducing large uncertainties in all radiative and in situ measurement methods of the free atmosphere aerosol.

The aerosol optical depth (r^) was calculated for the three operational wavelength channels of the mountaintop and the valley MFRSR, 415 nm, 500 nm, and 673 nm, using the Langley method. The Langley method assumes that atmospheric attenuation is generally constant throughout the measurement period (airmass length ranging from m = 2 to m = 6), which is generally true for a clear, stable atmosphere (Lenoble, 1993; Harrison and Michalsky, 1994). Langley plots on all days with clear sky intervals were derived for each instrument's wavelength channels and adjusted to the mean earth-sun distance. Greater variability of the valley instrument was observed and can be attributed to the variability of the atmosphere in the lower layer.

The optical depths for the total column above each respective site, tx, was calculated for each wavelength using an expression of Beer's law

V Vol J

where tx is the optical depth, V is the instrument output, Vo is the extraterrestrial constant adjusted for earth-sun distance on the day of interest, and d is the zenith angle. Each tx was calculated as a 12-minute average, centered around the zenith angle of interest.

Wavelength dependent Rayleigh scattering by air molecules was removed from the measurements, as it is a significant non-aerosol contributor to tx. Equation (11.3) was used to estimate the Rayleigh scattering (CTRayleigh) for each channel. The aerosol optical depth for the layer between the two sites was determined as the subtraction of the Rayleigh-corrected optical depths, rvalley - rmountain. The absorption optical depth of atmospheric gases, O3, NO2, and SO2 were estimated through calculations using a measured (O3) or assumed (NO2, SO2) column amounts. All three gases were found to have an extremely small contribution to tx, and were thus considered negligible.

The UV-B transmission (7W_B) for the layer between the two sites was calculated as the ratio of the UV-B measurement at the valley to the UV-B measurement at the mountain. The attenuation of the broadband UV-B irradiance during passage between the sites is assumed to be due to a combination of gaseous and aerosol absorption and scattering.

Aerosol size distributions were measured at the valley site using the Differential Mobility Particle Sizer (DMPS), on nine days coincident with significant clear sky periods. For each of the nine days, an average size distribution was determined from all the distributions collected during the coincident period. Lognormal parameters for the distribution were determined using the fitting software DISTFITTM provided with the instrument.

A flow chart outlining the method used to derive these aerosol optical properties using a Mie code, similar to the code of Dave (1968), in conjunction with a UV-B radiative transfer code (TUV) (Madronich, 1993) is shown below in Fig. 11.13. The resulting end product is the aerosol optical properties (at UV-B wavelengths) of an atmospheric layer between a mountain valley site and peak site in western North Carolina.

The lognormal parameters derived from the measured aerosol size distributions served as input to the Mie code using a first-guess for the complex index of refraction of 1.5 - 0.08i (1.5 is held constant). The Mie code calculations yield first-guess estimates of the asymmetry factor and single scatter albedo at 312 nm, the UVB-1's wavelength of maximum sensitivity to atmospheric UV. These values serve as input into the UV-B radiative transfer model of the UV-B transmission for the spectral response function of the Yankee UVB-1 radiometer. Additional inputs

Match modeled to measured U V-B transmission

Match single scatter albedo to value needed to match UV-B transmissions f Single scatter albedo

Resulting parameters -J asymmetry parameter

L index ofret'raction

Figure 11.13 Diagram outlining the modeling procedure used to determine single scatter albedo and the asymmetry parameter are the total ozone column as measured by the valley Brewer spectrophotometer and the aerosol optical depth at 312 nm, extrapolated from the spectral extinction determined from the Tx values of the MFRSR. This extrapolation of Tx to 312 nm is based on Angstrom's Formula for spectral extinction and has the following form:

where ra(2) is the aerosol optical depth at wavelength A, ¡3 is the turbidity coefficient, a is Angstrom's exponent which is related to the size distribution of the aerosol particles. Cachorro et al. (1989) showed that Angstrom's formula provides a good spectral representation of atmospheric aerosol attenuation. Iteration, varying the single scatter albedo, until the modeled transmission through the layer matched the observed transmission as measured by the UVB-1 instruments and then, varying the imaginary component of the complex index of refraction

Match modeled to measured U V-B transmission

Match single scatter albedo to value needed to match UV-B transmissions f Single scatter albedo

Resulting parameters -J asymmetry parameter

L index ofret'raction

Figure 11.13 Diagram outlining the modeling procedure used to determine single scatter albedo and the asymmetry parameter until the single scatter albedo matched the value resulting from the radiation code iteration was repeated until convergence was achieved. UV-B transmission is more sensitive to changes in S70 (compared to g), thus the decision to use it as the varying parameter. The aerosol size distribution acts as a constraint on g, which only changed slightly as the refractive index changed.

Optical depth values were obtained for all clear-sky periods throughout the six month observational period with a mean optical depth for the entire atmosphere as measured from the valley MFRSR at 500 nm of 0.356. The large seasonal variability inherent in optical depth measurements is evident by a standard deviation of 0.329. Past studies of annual turbidity cycles in nearby regions are consistent with this mean value (Flowers et al., 1969).

Clear-sky conditions were present over the two research sites for a large portion of the morning hours on 46 days. Using back trajectory analysis, 18 of these cases were classified as highly polluted air masses, 15 as polluted marine and 13 as polluted continental. Air masses arriving at the site from these sectors have been shown to exhibit differing physio-chemical characteristics (Deininger and Saxena, 1997; Ulman and Saxena, 1997). The MFRSR optical depth measurements for the intervening layer were categorized according to air mass source region. It was found that the r mean values differed for the three air mass types suggesting that aerosol optical properties vary in relation to air mass. The highest mean value of r for the highly polluted air mass classification is consistent with an assumption of higher sulfate aerosol concentrations, as this aerosol species has been shown to be an efficient attenuator at visible wavelengths (Whitby, 1978; Hegg et al., 1993; Yuen et al., 1994). The greater standard deviation of the highly polluted air mass cases is indicative of the greater variation in aerosol characteristics arising from differences in source strength, composition, size distribution and age of the aerosol. Two aerosol size distributions (highly polluted classification) displayed a multi-modal shape, which indicates the presence of an accumulation mode (particle diameter between 0.1 ^m and 1.0 ^m). Assuming that sulfate aerosols are the predominant species in the highly polluted air mass, the presence of a significant accumulation mode is not unexpected and can be explained through condensation and coagulation of sulfate aerosols. The nine clear sky days with DMPS aerosol size distribution measurements permitted a closer examination of the relationship of aerosol and radiative properties using the iterative procedure defined previously.

For the nine days investigated, m0 at 312 nm varied from 0.75 to 0.93 and showed no discernible dependence on air mass type; however, a dependence on relative humidity was observed. The three days possessing the highest relative humidities have S70 > 0.9, which is consistent with the characteristics of hygroscopic aerosols, which do not absorb strongly in the UV. The remaining six days exhibited S70 values of 0.75 to 0.82, comparable to the typical value given by Madronich (1993). This relatively narrow range of S70 for dryer conditions indicates that relative humidity is a controlling factor in the resulting value of S70.

The small sample size did not allow determination of any seasonal trend in The evidence from this study indicates that typically moist climates (such as in the Southeast or Northwest U.S.) will display larger variations of than arid climates (such as in the Midwest and Southwest U.S.). According to our results, the magnitude of m0 for the dryer atmospheric conditions implies that a fairly substantial portion of aerosol attenuation in the UV is due to absorption. Relative humidities are generally high in the eastern U.S. (typically > 80%), so it is probable that comparatively lower m0 values exist in arid climates and will contribute a significant absorption component to aerosol attenuation of UV-B. The m0 values found in this study are in general lower than those used in several UV modeling studies (Liu et al., 1991; Wang and Lenoble, 1994). The nine g values determined for our site fall within 0.63 to 0.76. No specific feature based upon air mass is discernible. Angstrom's exponent (a) was determined using the MFRSR spectral aerosol optical depth data and also the volume spectral extinction as calculated by the Mie code. The good agreement for five of the nine cases, evident from the small differences between the two values, gives increased confidence in the modeling procedure results. The larger differences for the remaining four cases can be attributed to greater uncertainty in the aerosol size distribution measurements. The effect of aerosol chemical composition on UV-B transmission is unquantified in this investigation. Future studies should incorporate coupled measurements of aerosol size distribution, aerosol chemical composition characteristics and in-situ aerosol optical properties. Such combined measurements are rarely obtained (Schwartz et al., 1995).

Extrapolation of the spectral extinction curve, determined from ln(r) versus ln(2) fit with linear regression permits an estimate of r312 for each day. A plot of ln(Solar Noon UV-B Transmission) versus 7312 exhibits a strong correlation (r2 = 0.904). The fit is surprisingly good considering the wide range of solar noon zenith angles, 20° to 55°,for the nine days used. The linear regression equation provides a simple expression for determining solar noon UV-B transmission if a value of r312 is known. The regression equation is as follows:

ln(UV-B Transmission at Solar Noon) = - 0.1422 x (r312) - 0.138 (11.10)

The j-intercept, i.e. no aerosol attenuation, converts to a transmission of 87.1%.

The average transmission for each clear morning was derived for the UVB-1 measurements. A morning average transmission was determined for the zenith range 65° - 40° as the transmission changed by less than 3% over this period for any given day. The highly polluted cases exhibited, on average, the lowest transmission, 77.8% + 4.6%. The polluted marine cases exhibited an average transmission of 81.9% ± 2.9%, and the polluted continental cases 83.5% ± 2.8%. The observed differences are due both aerosol differences and the abundance of absorbing gases. The mean ozone column for the layer was 12.6 DU (± 6 DU), and showed no discernible trend versus air mass type. A change in layer ozone column of 10 DU

resulted in an UV-B transmission change of only 1.5% (decreased transmission for increased ozone and vice versa). The effect of SO2 and NO2 absorption on UV-B transmission was modeled and showed a decrease in UV-B transmission of much less than 0.5% for the typically minute layer column amounts of each gas. Differences in gaseous absorption, although small, can account for a portion of the variability between air mass types and also the variability within a given air mass type. Therefore, the remainder of the variability is primarily due to differences in aerosol attenuation properties. Highly polluted air masses, which exhibited the highest aerosol optical depth, also exhibited the greatest attenuation of UV-B. In comparison, the "cleaner" marine and continental air masses had lower average tx values and higher UV-B transmission values. Our observed range of broadband UV-B attenuation for all the days was 14% - 31% for the 1 km layer. This range is quite consistent with the other reported values of an altitude effect, or the percentage decrease per 1,000 meters altitude in the lower troposphere.

An important aspect of this study is the development of an empirical relationship between optical depth and UV-B transmission, useful for calculation of the NOAA UV Index, UV climatology studies, and remote sensing applications. Specific knowledge of aerosol parameters and their effects on UV-B transmission allow for improved accuracy in UV climatology modeling. There is an inherent regionality to optical depth and aerosol optical properties which will in turn cause regional differences in UV-B transmission, and this needs to be considered in such applications as UV forecasts, UV climatology and long term trends. Remote sensing applications that rely upon UV attenuation measurements need to account for any change in aerosol UV attenuation due to changing aerosol properties.

To investigate the effect of aerosols alone, it is necessary to utilize an instrument that measures in wavelength bands where non-aerosol attenuation is minimized, such as the three channels from the MFRSR (415 nm, 500 nm, and 673 nm). Ideally, aerosol optical depths at wavelengths in the UV-B region would yield the best relation. However, optical depths at UV-B wavelengths are more problematic to obtain. Plots of natural logarithm of UV-B transmission versus aerosol optical depth were made for solar zenith angles of 65° to 40° in 2.5° intervals. Individual zenith angle plots were made as opposed to daily averaged plots because global UV-B transmission is slightly dependent upon zenith angle. Each plot was negatively correlated, i.e. increased aerosol optical depth results in decreased UV-B transmission. The correlation coefficient (r ) values for each plot show that the regressions fit very well, with only 4 of the 33 regressions having an r2 < 0.8. For each wavelength, the slope, or change in ln(UV-B transmission) for a given change in t, decreases as zenith angle decreases. This implies that as the sun nears solar noon and a shorter atmospheric pathlength, the UV-B transmission becomes less sensitive to rand also that the UV-B transmission approaches a maximum value. This is of importance in biological terms since the hours surrounding solar noon are the times of maximum irradiance.

For a given zenith angle, the regression slopes increase as the wavelength increases due to the fact that optical depth is proportional to A~l. The larger range of optical depths at the shorter wavelengths is reflected in a smaller percentage change in UV-B transmission (lesser slope) associated with a given change in aerosol optical depth. Since a given change in aerosol concentration produces a more significant variation in optical depth for shorter wavelengths, the sensitivity of UV-B transmission to aerosol optical depth must be less at these wavelengths. A percentage change in UV-B transmission for a given change in aerosol optical depth can be determined from the regressions. Using the 500 nm plot at 50° solar zenith angle as an example, a change in optical depth of 0.1 predicts a change in the UV-B transmission on the order of 4%. An additional feature of interest is the nearly constant value of the j-intercept values. Upon conversion to percent UV-B transmission, the averages of the j-intercepts over all zenith angles are 86.8% ± 0.7% for the 415 nm channel, 87.0% ± 0.8% for the 500 nm channel, and 87.0% ± 0.7% for the 673 nm channel. Thus, 87% is the UV-B transmission expected for an aerosol free layer and the 13% attenuation can be ascribed to Rayleigh scattering in addition to the slight absorption by atmospheric gases, i.e., O3, NO2, and SO2, in the layer. This value is consistent with the zero aerosol conditions derived for the nine day case study results discussed previously. Multiple regression analysis using the Statistical Analysis System (SAS) was performed to obtain the full model for prediction of the UV-B transmission versus zenith angle (in degrees) given a set of aerosol optical depth measurements, and the backward elimination procedure was used to identify the most significant variable. Aerosol optical depth at 415 nm (r415) was determined as the most significant variable for predicting UV-B transmission, with the ability to account for 72.8% of the data variability. Upon including the parameters r500 and zenith angle (the two next most significant parameters) the fit improves to account for 82.0% of the observed variability. All three variables are significant at the 0.0001 level. The empirical regression equation is of the form:

ln (UV-B transmission) = a0 + a1 xr415 + a2 xr500 + a3 x zenith angle (11.11)

where the regression coefficients are defined as a0 = — 0.0935, a1 = -0.8515, a2 = 0.7505, and a3 = — 0.0012. The addition of the two extra terms, resulting in a 10% improvement in fit, was deemed necessary as the inclusion of a zenith angle term enables prediction for UV-B transmission at solar noon, and the use of two optical depth values provides information on the spectral extinction shape. The r673 term was neglected, as inclusion did not significantly improve the quality of the fit (0.1% improvement). The 18% variability unaccounted for can be ascribed to the combined uncertainties of the measurements as well as factors, each individually small, not included in the regression (i.e., variations in ozone, aerosol absorption, chemical composition, and relative humidity). Equation (11.11) is specific to the region surrounding the experiment sites.

Was this article helpful?

0 0

Post a comment