• Introduction
• Instrumentation and methodology
• Results and Discussion
• Conclusions
• Acknowledgements
• References

FIGURES

• Figures 1a, 1b, 1c, 1d
• Figure 2
• Figure 3
• Figures 4a, 4b, 4c, 4d

A network of 21 Brewer spectroradiometers, operated by the U.S. Environmental Protection Agency/University of Georgia (EPA/UGA) is measuring UV spectral irradiances throughout the United States. Corrections to the raw data have now been implemented. These corrections include (1) the cosine errors associated with the full sky diffuser, (2) the temperature dependence of the response of the instruments and (3) the temporal variation in the instrument response due to optical changes in the characteristics of the instruments. While for many sites the total corrections amount to less than 10 %, for certain sites they are much larger, in some cases amounting to more than 25 %. Application of these corrections brings the errors of the absolute irradiance values to approximately ± 5 %.

Introduction

A network of 21 Brewer MKIV spectroradiometers (Kipp & Zonen Inc, Canada) is measuring UV spectral irradiances throughout the United States. The database from these sites extends for periods ranging from two to seven years.

Corrections to the raw data have now been implemented. These corrections include (1) the cosine errors associated with the full sky diffuser, (2) the temperature dependence of the response of the instruments and (3) the temporal variation in the instrument response due to optical changes in the characteristics of the instruments. The dark count, dead time and stray light corrections of the program RD_UX and the UVA correction and erythemal action spectrum used in the latest version of the program UVSUM, were also used for this paper (Sci-Tec 1999). When referring to the uncorrected UV data it is implied that dark-count, dead time and the UVA correction have been applied.

Corrected irradiances at four sites are compared with a clear sky UV model. These sites are Boulder (#101, 40.12 ° N, 105.24 ° W, 1689 m a.s.l.), Rocky Mountain (#146, 40.03 ° N, 105.53 ° W, 2891 m a.s.l.), Gaithersburg (#105, 39.08 ° N, 77.22 ° W, 20 m a.s.l.) and Research Triangle Park (RTP) (#087, 35.89 ° N, 78.88 ° W, 134 m a.s.l.). Details of these sites are available at the EPA web address: http://www.epa.gov/uvnet/

Intercomparisons of the daily-integrated damaging UV (DUV) measurements from the Brewers at these sites to DUV values inferred from satellite overpass data, corresponding to these sites, are made in order to validate the estimates from the satellite data. Available satellite data ranges from July 1996 until February 2000. Brewer and model DUV data will be presented from July 1996 until May 2000.

It should be noted that the erythemally-weighted irradiances for the Brewer, model and TOMS data all used the action spectrum of McKinlay and Diffey (1987).

Instrumentation and methodology

1. Brewer Spectroradiometer

   To measure UV irradiance the Brewer uses a quartz dome and Teflon diffuser with a hemispherical field of view. For the EPA/UGA network, a dynamic schedule is used by the Brewer allowing UV readings to be recorded approximately every 30 minutes throughout the day, ensuring that a UV scan coincides with solar noon. The Brewer has a UV spectral range of 286.5 to 363 nm in 0.5 nm steps. UV irradiance calibrations, using a secondary standard lamp traceable to a NIST 1000 W lamp, are performed at the sites by NUVMC. Resulting response functions were used to calculate irradiance from photon counts. Calibrations are targeted to occur once per year. In addition independent quality assurance audits of the instruments take place.

   The Brewer recorded total column ozone levels between UV measurements. The methodology of Sabburg et al. (2000) was used to obtain direct sun ozone values to less than ± 3 %.

2. UV Corrections

The data presented in this paper have been corrected for dark count, dead time, stray light, UVA, cosine response, temperature dependence and temporal response. Only limited quality assurance (QA) has been performed on the raw data, such as removing extreme outliers. The uncorrected data includes dark count, dead time, UVA correction and uses the latest response function.

1. Cosine correction

   Sabburg and Meltzer (2000) explain the cosine correction methodology associated with the full sky diffuser in detail. In summary, cosine response measurements were made on each Brewer using the irradiance of a standard 1000 W lamp. These measurements were performed in the laboratory. The final values were based on an average of measurements along the long and short sides of the Brewer, five wavelengths and two sets of zenith angles. The wavelengths were 306.3, 310.1, 313.5, 316.8 and 320.1 nm. The lamp rotated over the zenith angle range of ö 80 to 0 ° and 0 ° to 80 ° in 10 ° steps.

   The equations of Bais et al. (1998) were used to calculate the total cosine correction assuming a diffuse isotropic sky and the ratio of the direct / global irradiance was based on the clear-sky model of Rundel (1986). The model used corrected ozone amounts from the Brewer and a typical value of aerosol optical depth (AOD) was chosen for each site, namely, 0.2, 0.1, 0.8 and 0.5 for Boulder, Rocky, Gaithersburg and RTP respectively (Jeral Estupinan, Personal Communication, 2000). When ozone data was not available, the nominal value of 300 DU was used for that day.

2. Temperature Dependence

   The temperature dependence of two of the Brewers (#101 and #146), was measured by the NUVMC during a field campaign in July 2000. The methodology for measuring the temperature dependence is outlined by Meltzer et al. (2000). The Brewer temperature fluctuates with the ambient at the various locations from 0 to + 50 ° C. The temperature dependence from - 18 to + 42 ° C of the response of three MKIV Brewers (not the Brewers in this current paper), has been determined in the laboratory using a controlled variable temperature environment. The observed temperature coefficients of their responses at 306 nm were - 0.17 %, - 0.22 % and - 0.37 % per ° C. This results in a predicted seasonal variation of their sensitivity of up to approximately 20 %, well beyond the desired accuracy of ± 3 %. There is a significant wavelength dependence of the temperature coefficient below 325 nm. This is primarily due to the temperature dependence of the transmission of a nickel sulfate filter.

   The methodology also requires that local operators obtain the temperature coefficients of the response as a function of wavelength at each site. The method utilizes spectra of the 50 W Brewer calibration lamps recorded throughout the day during the diurnal temperature cycle. These measurements require more accurate and stable instrumentation than is supplied with the Brewer. Plots of the photon counts versus temperature at each wavelength were used to determine a temperature coefficient, D R/D T, which is the slope of the response versus temperature. The temperature corrections were achieved by normalizing photon counts to an equivalent photomultiplier tube (PMT) temperature of 20 ° C.

3. Temporal Variation

The temporal variation in the instrument response is due to optical changes in the characteristics of the instruments. This necessitates an annual UV irradiance calibration, using a secondary standard lamp traceable to the NIST 1000 W lamp, to be performed at the sites by NUVMC staff. Resulting response functions are used to calculate irradiance from photon counts. In addition, NOAA, using similar equipment to the NUVMC, conduct independent quality assurance audits of the instruments. Details of these procedures are available at the UGA web address: ftp://oz.physast.uga.edu/Outgoing/ by downloading the three documents entitled: SOP1_FEL-Lamp.doc, SOP for Field Calibration.doc and Irradiance Transfer of 1000 Watt lamps.doc.

C) Model UV data

In order to make a comparison of the corrected UV data (in addition to comparing to the uncorrected data), the data calculated by the clear sky model, Uvspec, was used (Kylling 1995). UVSPEC used the wavelength range of 280 to 400 nm with 1.0 nm resolution and used two-stream code to solve the radiative transfer equation. Ozone and oxygen absorption, as well as Rayleigh scattering was accounted for. UVSPEC used the same nominal AOD values as used for the cosine correction, scaled according to visibility. When ozone data was not available, the nominal value of 300 DU was used for that day.

D) Satellite Data

Finally, a comparison is made with the inferred DUV data from the NASA, Total Ozone Mapping Spectrometer (TOMS). TOMS data is indicative at the time of over-pass (approximately 11:15 AM local time) and the footprint for the TOMS instrument is approximately 40 km x 40 km. The TOMS erythemal exposures are calculated with a radiative transfer model (Herman et al. 1996; Udelhofen et al. 1999). The TOMS algorithm uses ozone data as well as reflectivity measurements at 380 nm to identify cloudy scenes. The daily integration is carried out assuming no diurnal variation in cloudiness.

Results and Discussion

1. Cosine, Temperature and Temporal response corrections

   Table 1 lists the percentage differences, (Uncorrected ö corrected)*100/corrected, for each UV correction, for the erythemally-weighted irradiance corresponding to the scan around local noon on the 30^th May 1999 for data at the four sites. In addition, Table 1 lists the range of percentage differences for all UV corrections, of daily-integrated DUV values for available data at the four sites. In this additional part of the table, the positive values for Boulder and Rocky are due to very low temperatures during the UV measurements. For example, at Boulder on the 12^th January 1997, the PMT temperature of the Brewer ranged from ö16 to ö10 ° C. The positive values for RTP were due to large values of stray light entering the PMT.

   TABLE 1

   Type of Correction	Boulder	Rocky	Gaithersburg	RTP
   Stray-light	2.2	2.5	5.4	2.1
   Temporal	-6.2	0.4	-9.3	-16.4
   Temporal & Temperature	-14.5	0.8	-	-
   Cosine	-5.7	-11.1	-11.6	-9.5
   All corrections	-18.4	-8.1	-16.0	-22.7
   % DUV range for all corrections	-22.2 to 14.8	-13.8 to 6.0	-23.5 to ö4.4	-26.8 to 20.8

2. Comparison between uncorrected, corrected and clear sky model data

   Figure 1 presents a graph of the uncorrected, corrected and clear-sky model data for (a) Boulder (b) Rocky (c) Gaithersburg and (d) RTP.

   

   (a)

   

   (b)

   

   (c)

   

   (d)

   Figure 1. Graphs showing the daily DUV levels for (a) Brewer # 101 (b) # 146 (c) # 105 and (d) # 087, compared to the corresponding clear-sky modeled data for the dates as shown on the x-axes.

   The corrections bring the measured DUV in closer agreement with the clear sky modeled calculations.

   Figure 2 shows an example correlation graph of clear-sky, uncorrected and corrected DUV data versus clear-sky modeled data for Boulder. Clear-skies are defined as days for which the corresponding TOMS reflectivity data was less than 5 %. Only data where a full day of UV scans and measured Brewer ozone data were available are presented and used for analysis purposes. A linear correlation is shown for both the corrected and uncorrected Brewer DUV data with respect to the modeled data.

   

   Figure 2. Graph showing daily, clear-sky Brewer DUV data versus clear-sky model data. The linear correlation and information box shown at the top of the graph corresponds to the corrected Brewer data.

   The clear-sky corrected data show improvement in the correlation coefficient and are in better agreement with model clear-sky calculations, but they still fall below the model DUV values by approximately 15 %.

   Figure 3 shows a time series graph corresponding to Figure 2, for corrected and model data only. The differences between the corrected and modeled data, (Corrected ö Model)*100/Model, is ö 44.1 to 5.7 %, with an average difference of ö 14.5 %. The larger values of the model calculation is in part due to the fact that some clear days, defined by TOMS overpass data at approximately 11:15 AM local time, may not remain clear throughout the whole day. In addition, the UVSPEC model may overestimate the irradiance because the optical depth and ozone were held constant throughout the day.

   

   Figure 3. Graph showing clear-sky daily DUV levels for Brewer # 101 compared to the clear-sky modeled data.

3. Comparison between corrected and satellite data

(a)

(b)

(c)

(d)

Figure 4. Graph of corrected and corresponding TOMS DUV data for all available data for (a) Boulder (b) Rocky (c) Gaithersburg and (d) RTP. In the case of Boulder, the right hand axis shows the percentage difference, (Corrected ö TOMS)*100/TOMS, for clear days.

Figure 4 shows time series graphs of corrected and TOMS data for the four sites. In the case of Figure 4 (a) for Boulder, the variation between the corrected and TOMS data, (Corrected ö TOMS)*100/TOMS, for clear days is also shown. The range of differences between TOMS-inferred and corrected DUV is from ö 40.1 to 18.5 %, with an average difference of ö 9.1 %. The larger values of the TOMS calculation is in part due to the fact that some clear days, defined by TOMS overpass data at approximately 11:15 AM local time, may not remain clear throughout the whole day.

Finally, Table 2 summarizes the correlation and percentage difference data for the clear days, for uncorrected, corrected, model and TOMS data for the four sites. The comparison with TOMS data for Rocky (showing a positive value of 3.4 in Table 2) is most likely a result of the lower altitude used by the TOMS model corresponding to this site (1389 m). The comparison with model data for Gaithersburg (showing a positive value of 7.9 in Table 2) may be due to overestimation of the AOD for this site.

TABLE 2 Comparison of clear day corrected daily-integrated DUV values at four sites with values predicted with the UVSPEC model and those values inferred from TOMS overpass measurements.

SITE	Correlation: uncorrected with model (r2)	Correlation: corrected with model (r2)	Average %: (Corr-model)*100/ model	Average %: (Corr- TOMS)*100/ TOMS
Boulder	0.84	0.87	- 15	- 9
Rocky	0.80	0.87	-13.9	3.4
Gaithersburg	0.97	0.98	7.9	-8.3
RTP	0.92	0.90	- 12.1	- 10.5

Conclusions

In order to bring the error of measured absolute irradiance values to within ± 5 %, correction for the temperature dependent response, temporal variation of response and the angular dependence of the collector are essential. The corrected DUV values for clear sky conditions are in much closer agreement with a theoretical model and generally show better correlation with that model compared to the uncorrected values.

Correction of the data at the four sites studied, fall in the range - 27 to 21 %. Clear-sky corrected DUV values fall in much closer agreement with TOMS-inferred DUV values. The average differences between clear-sky TOMS-inferred and corrected DUV values are in the range of - 11 to 3 %, with the smallest average value of 3 % corresponding to the Rocky site.

Acknowledgments

The authors would like to thank EPA project officers Jack Shreffler and Tom Lumpkin, the site operators at Boulder, Rocky Mountain and Gaithersburg, the NASA/Goddard Space Flight Center via Edward Celarier (Software Corporation of America) whom we would like to thank for the special TOMS UV overpass runs and Jeral Estupinan for advice on aerosol optical depth values.

References

Bais, A.F., Kazadzis, S., Balis, D., Zerefos, C.S. and Blumthaler, M. 1998 ÎCorrecting global solar ultraviolet spectra recorded by a Brewer spectroradiometer for its angular response errorâ, Applied Optics, vol. 37, no.27, pp.6339-6344.

Herman, J.R., Bhartia, P.K., Ziemke, J., Ahmad, Z. and Larko, D. 1996 :UV-B increases 1979-1992 from decreases in total ozoneâ, Geophys. Res. Lett., 23, pp.2117-2120.

Kylling, A., 1995: UVSPEC , a program for calculation of diffuse and direct UV and visible fluxes and intensities at any altitude', available by anonymous ftp to kaja.gi.alaska.edu, cd pub/arve.

McKinlay, A.F. and Diffey, B.L. 1987: A reference action spectrum for ultraviolet induced erythema in human skin, in Human Exposure to Ultraviolet Radiation: Risks and Regulations, edited by W.R. Passchier and B.F.M. Bosnjakovic, pp.83-87, Elsevier, New York.

Meltzer, R.S., Wilson, A., Kohn, B. and Rives, J.E. 2000: Temperature dependence of the spectral response for the MKIV Brewers in the UGA/USEPA networkâ, 6^th Brewer Workshop, Tokyo, Japan, 10-12^th July 2000.

Rundel, R. 1986: Computation of Spectral Distribution and Intensity of Solar UV-B Radiationâ: in 'Stratospheric Ozone Reduction, Solar Ultraviolet Radiation and Plant Lifeâ, ed. R.C. Worrest, M.M. Caldwell, Springer-Verlag, Berlin.

Sabburg, J. and Meltzer, R.S., 2000: Cosine corrections for ultraviolet radiation data from the USEPA/UGA Brewer Networkâ, Sixth WMO/GAW Brewer Users Group Meeting, Japan Meteorological Agency (JMA), Tokyo, JAPAN, July 10-12.

Sabburg, J., Meltzer, R.S. and Rives, J.E., 2000: Corrected Ozone and erythemal ultraviolet radiation data from the USEPA/NUVMC Brewer Networkâ, Quadrennial Ozone Symposium, Hokkaido University, Sapporo 2000, 3-8 July, Japan, pp.759-760.

SCI-TEC 1999: Brewer MKIV Spectrophotometer Operatorâs Manual, OM-BA-C231 REV B, August 15â, SCI-TEC Instruments Inc.

Udelhofen, P.M., Gies, P., Roy, C., Randel, W.J., 1999: Surface UV radiation over Australia, 1979-1992: Effects of ozone and cloud cover changes on variations of UV radiation, J. Geophys. Res., 104(D16), pp.19135-19159.