Comparison of UV-B Broadband Brewer Measurements with Irradiances from Surface-Based and Satellite-Based Models

doi:10.4236/acs.2011.13016

Paper Menu >>

Journal Menu >>

Atmospheric and Climate Sciences, 2011, 1, 142-146

doi:10.4236/acs.2011.13016 Published Online July 2011 (http://www.SciRP.org/journal/acs)

Comparison of UV-B Broadband Brewer Measurements

with Irradiances from Surface-Based and Satellite-Based

Models

Jacqueline Binyamin1, John Davies2, Bruce McArthur3

1Department of Ge ography, University of Winnipeg, Winnipeg, Canada

2School of Geography and Earth Sciences, McMaster University, Hamilton, Canada

3Air Quality Research Branch, Meteorological Service of Canada, Toronto, Canada

E-mail: [email protected]

Received May 1, 2011; revised June 13, 2011; accepted June 28, 2011

Abstract

UV-B irradiance can be estimated from surface meteorological data or from satellite measurements. This pa-

per compares irradiance estimates from the Davies surface-based radiation model and the Canada Centre for

Remote Sensing (CCRS) satellite model with Brewer spectrophotometer measurements for all sky conditions

at six Canadian stations (Edmonton, Regina, Winnipeg, Montreal, Halifax and Toronto). The Davies model

is applied with both the discrete ordinate radiative transfer (DISORT) and the delta-Eddington algorithms to

solve the radiative transfer equation.

Both models’ estimates are compared with instantaneous Brewer measurements. Both perform similarly

with mean bias errors within 6% of the mean measured irradiance for the measurement period and root mean

square errors between 25% and 30%.

Keywords: UV-B Irradiance, Brewer Spectrophotometer, DISORT, Delta-Eddington, TOMS,

Satellite-Based Models, Surface-Based Models

1. Introduction

Stratospheric ozone depletion increases the amount of

harmful UV-B (290 - 325 nm) irradiance reaching the

earth’s surface [1]. Instruments such as the Brewer spec-

trophotometer measure spectral irradiance [2], but meas-

urements in Canada and internationally are spatially s-

-parse and are generally of short duration. Radiation md-

els can estimate irradiance for locations and times with-

out measurements and can predict irradiances for possi-

ble changes in ozone concentrations.

Radiation models use either surface meteorological

data or satellite measurements. Models, which use sur-

face data, apply algorithms, which vary from simple ap-

proximations [3] to rigorous solutions of the radiative

transfer equation. The two most widely used radiative

transfer solutions are the discrete ordinate radiative trans-

fer (DISORT) model [4] and the delta-Eddington model

[5]. Since these use local data, they should represent

point conditions more accurately than the large area es-

timates from satellite.

Satellites measure reflected radiances at the top of the

atmosphere which, combined with radiative transfer cal-

culations or inversion algorithms, provide estimates of

the irradiance at the surface [6-8]. Although satellite meas-

urements provide extensive spatial coverage they usually

provide, unless they are sun synchronous, one measure-

ment each day, which prohibits the calculation of daily

total irradiance.

Comparisons of model calculations with measurements

are mostly restricted to data for a few cloudless days

[9-12]. Few studies have validated surface-based models

for all sky conditions [13-17].

This paper compares broadband irradiances (290 - 320

nm) from a surface-based model [16,17] and from the

Canada Centre for Remote Sensing (CCRS) Meteor-3/

total ozone mapping spectrometer (TOMS) satellite ba-

sed-model [6-8] with Brewer spectrophotometer meas-

urements in Canada.

2. Irradiance Measurements

Brewer spectrophotometer measurements for six Can

J. BINYAMIN ET AL.143

Table 1. Canadian UV-B monitoring stations used in the

study.

Station Latitude,

˚N

Longitude,

˚W

Elevation,

Years of

data

Edmonton (Alta.) 53˚33′ 114˚06′ 766 1993-1994

Regina (Sask.) 50˚13′ 104˚40′ 592 1994

Winnipeg (Man.) 49˚55′ 97˚14′ 239 1993

Montreal (Que.) 45˚28′ 73˚45′ 24 1993-1994

Halifax (NS) 44˚44′ 63˚40′ 31 1993-1994

Toronto (Ont.) 43˚47′ 79˚23′ 198 1993-1994

dian locations, for which there were simultaneous satel-

lite-based irradiances, are used in this study (Table 1).

This instrument allows for the calculation of daily ozone

depth and measures spectral irradiance for wavelengths

between 290 and 325 nm at a resolution of 0.5 nm. Ra-

diation measurements are made once or twice each hour

from sunrise to sunset at irregular times in GMT [18].

Following Krotkov et al. [12] and Wang et al. [19] the

Brewer values were increased by 6% to compensate for

the cosine error.

3. The Davies Model

The model developed by Davies et al. [16] and modified

by Binyamin et al. [17] is used in this study. Surface

irradiance is calculated as a linear combination of

clear and overcast sky irradiances weighted

with cloud fraction:

GG



1GCGCG



 .



(1)

G and are calculated spectrally at wavelength

interval of 1 nm using either the DISORT [4] or the

delta-Eddington [5] solutions to the radiative transfer

equation for a vertically inhomogeneous 49-layer, 120

km, plane parallel atmosphere, with cloud inserted be-

tween 2 and 3 km heights. The model uses spectral val-

ues of the extraterrestrial irradiance from the Solar Ul-

traviolet Spectral Irradiance Monitor (SUSIM) ATLAS-3

space shuttle mission (D. Prinz, personal communication,

2002), ozone absorption coefficients from Paur and Bass

[20], Rayleigh scattering cross sections following Elter-

man [21], aerosol optical properties from Shettle and

Fenn [22], a fixed cloud optical depth of 27 [17,23], and

a fixed surface albedo of 0.05 [24]. Hourly total cloud

opacity observations were obtained from the Meteoro-

logical Service of Canada.

G

4. The CCRS Satellite Model

The CCRS satellite algorithm for retrieving surface ir-

radiance [6] is based on a linear relationship between

TOA albedo at 360 nm and surface absorbed irradiance.

The surface irradiance is given by:





0 03360

cos1 0.1960.798

(1 )

sus

ST a





 

 (2)

where 0 is the extraterrestrial irradiance;S



the solar

zenith angle ;03 the band-mean transmittance due to

ozone absorption; 360



the albedo for the earth-atmos-

phere system at 360 nm;



surface albedo; and d

and u represent aerosol absorption for the downwel-

ling and upwelling irradiance.

The extraterrestrial irradiance was taken from Fröhlich

and London [25]. Total ozone amount was taken from

the TOMS data set. Surface albedo was assumed to be

0.04 for Toronto and 0.03 for other stations. Aerosol

optical depth a



measurements were available for To-

ronto [19] and a value of 0.31 was substituted for miss-

ing days. For Winnipeg and Edmonton a



was assumed

to be 0.2, and 0.1 for the remaining stations. Aerosol

single scattering albedo was assumed to be 0.95 for all

stations. The model can only be used for days without

snow cover.

5. Comparison of the Davies Model and

CCRS Model Results

The Meteor 3 reflectance measurements were made at

different local times, with usually once per day, with a

maximum of two values per day that matched the times

of Brewer measurements (Pubu Ciren, CCRS, personal

communication, 2007). Each CCRS model estimate is

compared with a simultaneous Brewer measurement and

a calculation from the Davies model using both the

delta-Eddington and DISORT methods. There is a small

difference in the spectral integration range used by

CCRS and the Davies model. CCRS presents irradiances

integrated over the 290 - 320.5 nm wavebands in 0.5 nm

steps while the Davies irradiances were integrated in 1

nm steps over the 290 - 320 nm range. In this section,

both models are compared with Brewer measurements

integrated to the upper wavelength limit appropriate to

the model.

Comparison of the Davies model broadband irradian-

ces and the simultaneous satellite-based results with Bre-

wer measurements made at six stations (Edmonton, Re-

gina, Winnipeg, Montreal, Halifax and Toronto) in 1993

and 1994 (Table 1) are presented in Figure 1. They rep-

resent 10 station years of instantaneous data mainly be-

tween May and September, with 605 data points in total.

Generally, the agreement between the three is visually

good for all stations. The MBE (mean difference) values

correspond to less than 0.04 W·m-2 on average for indi-

vidual stations and pooled data for both surface-based

and satellite-based methods.

Table 2 provides performance statistics for both indi-

J. BINYAMIN ET AL.

144

Figure 1. Comparison between Davies model calculations, CCRS satellite-based model calculations and measured UV- B ir-

radiances for all sky conditions for Edmonton, Regina, Winnipeg, Montreal, Halifax and Toronto. The dotted lines represent

linear regressions constrained to pass through the origin. A different symbol represents data for each station.

vidual stations and pooled data. For the pooled data, the

Davies model with the delta-Eddington and DISORT

methods and the CCRS model underestimate Brewer

measurements by less than 6% of the mean measured

irradiance. For the individual stations the relative MBE

values range between ±6% for the delta-Eddington me-

thod, between –10% and 0.2% for the DISORT method

and between –11% and 0.3% for the satellite-based me-

thod. These irradiance differences are smaller than dif-

ferences between near simultaneous measurements made

with different ground instruments ±10% [26-29]. The re-

lative RMSE values for the pooled data are similar (26%

and 29%) for the two methods. For the individual sta-

tions, relative RMSE values range from 24 to 34% for

the Davies model and from 20 to 40% for the satellite-

based model. This agrees well with the findings of Fio-

letov et al. [30] and Wang et al. [31] who found that dif-

erences between TOMS-estimated UV-B irradiances

J. BINYAMIN ET AL.

145

Table 2. Summary of the Davies model and satellite-based method performance measures against Brewer measurements for

broadband irradiances (290 - 320 nm) for each satellite time for the period indicated for each station. is the number of

data points and N

M is the mean measured irradiance (W·m–2). MBE and RMSE are expressed as percentages (italic) of M.

Positive MBE values indicate model overestimation.

Irradiance results Edmonton

1993-1994 Regina 1994Winnipeg 1993Montreal

1993-1994

Halifax

1993-1994

Toronto

1993-1994 Pooled data

N 115 48 88 20 159 175 605

Davies model using delta-Eddington metho

1.16 1.20 1.54 0.66 1.19 1.13 1.21

MBE 6.22 –0.79 –4.97 2.13 –5.66 3.64 –0.29

RMSE 31.73 25.05 24.25 26.53 24.50 33.46 28.59

Davies model using DISORT method

MBE 0.15 –6.62 –10.06 –5.78 –10.28 –2.38 –5.80

RMSE 28.94 25.72 25.88 27.76 25.93 33.85 28.91

Satellite method

1.26 1.31 1.68 0.72 1.30 1.23 1.31

MBE 0.26 –5.14 –9.75 –10.86 –1.40 0.06 –2.73

RMSE 25.49 22.30 24.11 40.05 20.45 31.03 25.76

and Brewer observations range from 3 to 11% and can be

attributed to the Brewer angular response error.

6. Conclusions

This is the first study to compare the performance of

satellite-based model with ground-based model estimates

of UV-B irradiance. The two models perform almost i-

dentically. Differences in performance are within the

uncertainties in Brewer measurements. Different fields of

view of the TOMS instrument, the Brewer spectropho-

tometer and the Davies model seem to be inconsequen-

tial.

The satellite model can be applied virtually anywhere

and does not require other observations. However, it

does have two drawbacks. Firstly, because it cannot cal-

culate irradiance throughout the day, it cannot produce

daily totals. Secondly, it cannot be used over snow-cov-

ered surfaces because it is incapable of discriminating

differences between cloud and surface reflections. The

Davies model produces daily irradiance totals but its use

is currently restricted to stations with cloud observations.

Future work should examine the use of satellite cloud

data. The use of the delta-Eddington method is especially

appealing for calculating broadband irradiances since it

is computationally less demanding than DISORT

7. References

[1] S. Madronich, “Implication of Recent Total Atmospheric

Ozone Measurements for Biologically Active Ultraviolet

Radiation Reaching the Earth’s Surface,” Geophysical

Research Letters, Vol. 19, No. 1, 1992, pp. 37-40.

doi:10.1029/91GL02954

[2] J. B. Kerr, C. T. McElroy, D. I. Wardle, R. A. Olafson

and W. F. J. Evans, “The Automated Brewer Spectro-

photometer,” In: C. S. Zerefos and A. Ghazi, Eds., At-

mospheric Ozone. Proceedings of Quadrennial Ozone

Symposium, D. Reidel, Dordrecht, 1985, pp. 396-401.

[3] P. F. Schippnick and A. E. S. Green, “Analytical Charac-

terization of Spectral Actinc Flux and Spectral Irradiance

in the Middle Ultraviolet,” Phtochemistry and Phtobiol-

ogy, Vol. 35, No. 1, 1982, pp. 89-101.

doi:10.1111/j.1751-1097.1982.tb03815.x

[4] K. Stamnes, S.-C. Tsay, W. J. Wiscombe and K. Jayaw-

erra, “Numerically Stable Algorithm for Discrete Ordi-

nate Method Radiative Transfer in Multiple Scattering

and Emitting Layered Media,” Applied Optics, Vol. 27,

No. 2, 1988, pp. 2503-2509. doi:10.1364/AO.27.002502

[5] J. H. Joseph, W. J. Wiscombe and J. A. Weinman, “The

Delta-Eddington Approximation for Radiative Flux Trans-

fer,” Journal of the Atmospheric Sciences, Vol. 33, No.

12, 1976, pp. 2452-2459.

doi:10.1175/1520-0469(1976)033<2452:TDEAFR>2.0.C

O;2

[6] Z. Li, P. Wang and J. Cihlar, “A Simple and Efficient

Method for Retrieving Surface UV Radiation Dose Rate

from Satellite,” Journal of Geophysical Research, Vol.

105, No. 4, 2000, pp. 5027-5036.

doi:10.1029/1999JD900124

[7] C. Pubu and Z. Li, “Anisotropic Reflection of UV Radia-

tion at the Top of the Atmosphere: Characteristics and

Models Obtained from Meteor 3/TOMS,” Journal of

Geophysical Res ear ch, Vol. 106, No. D5, 2001, pp. 4741-

4755. doi:10.1029/2000JD900440

[8] C. Pubu and Z. Li, “Long-term Global Earth Surface

Ultraviolet Radiation Exposure Derived from ISCCP and

TOMS Satellite Measurements,” Agricultural and Forest

Meteorology, Vol. 120, No. 1-4, 2003, pp. 51-68.

doi:10.1016/j.agrformet.2003.08.033

[9] J. Zeng, R. McKenzie, K. Stamnes, M. Wineland and J.

Rosen, “Measured UV Spectra Compared with Discrete

Ordinate Method Simulations,” Journal of Geophysical

Research, Vol. 99, No. D11, 1994, pp. 23019-23030.

doi:10.1029/94JD02145

[10] P. Wang and J. Lenoble, “Comparison between Meas-

urements and Modeling of UV-B Irradiance for Clear Sky:

A Case Study,” Applied Optics, Vol. 33, No. 18, 1994, pp.

3964-3971. doi:10.1364/AO.33.003964

J. BINYAMIN ET AL.

146

[11] B. Mayer, G. Seckmeyer and A. Kyling, “Systematic

Long-Term Comparison of Spectral UV Measurements

and UVSPEC Modeling Results,” Journal of Geophysical

Research, Vol. 102, No. 7, 1997, pp. 8755-8767.

doi:10.1029/97JD00240

[12] N. A. Krotkov, P. K. Bhartia, J. R. Herman, V. Fioletov

and J. Kerr, “Satellite Estimation of Spectral Surface UV

Irradiance in the Presence of Tropospheric Aerosols 1.

Cloud-Free Case,” Journal of Geophysical Research, Vol.

103, No. D8, 1998, pp. 8779-8793.

doi:10.1029/98JD00233

[13] E. Leontyeva and K. Stamnes, “Estimations of Cloud

Optical Thickness from Ground-Based Measurements of

Incoming Solar Radiation in the Arctic,” Journal of Cli-

mate, Vol. 7, No. 4, 1994, pp. 566-578.

doi:10.1175/1520-0442(1994)007<0566:EOCOTF>2.0.C

O;2

[14] P. M. de F. Forster, K. P. Shine and A. R. Webb, “Mod-

elling Ultraviolet Radiation at the Earth’s Surface. Part II:

Model and Instrument Comparison,” Journal of Applied

Meteorology, Vol. 34, No. 7, 1995, pp. 2426-2439.

doi:10.1175/1520-0450(1995)034<2426:MURATE>2.0.

CO;2

[15] D. Lubin and E. H. Jenson, “Effects of Clouds and Strato-

spheric Ozone Depletion on Ultraviolet Radiation Trends,”

Nature, Vol. 377, 1995, pp. 710-713.

doi:10.1038/377710a0

[16] J. Davies, P. Kuhn, G. Duhamel, J. Binyamin and K.

Runnalls, “An Ultraviolet (290 - 325 nm) Irradiation Mo-

del for Southern Canadian Conditions,” Physical Geog-

raphy, Vol. 21, No. 4, 2000, pp. 327-344.

[17] J. Binyamin, J. Davies and B. McArthur, “Validation of

Spectral and Broadband UV-B (290 - 325 nm) Irradiance

for Canada,” Atmospheric and Climate Sciences, Vol. 1,

No. 3, 2011, In press.

[18] J. B. Kerr and C. T. McElory, “Evidence for Large Up-

ward Trends of Ultraviolet-B Radiation Linked to Ozone

Depletion,” Science, Vol. 262, No. 2136, 1993, pp. 1032-

1035. doi:10.1126/science.262.5136.1032

[19] P. Wang, Z. Li, J. Cihlar, D. I. Wardle and J. Kerr,” Vali-

dation of an UV Inversion Algorithm Using Satellite and

Surface Measurements,” Journal of Geophysical Re-

search, Vol. 105, No. D4, 2000, pp. 5037-5048.

doi:10.1029/1999JD900403

[20] R. J. Paur and A. M. Bass, “The Ultraviolet Cross-Sections

of Ozone: II. Results and Temperature Dependence” In:

C. Zerefos and A. Ghaz, Eds., Atmospheric Ozone Pro-

ceedings of the Quadrennial Ozone Symposium, D. Rei-

del, Dordrecht, 1985, pp. 606-616.

[21] L. Elterman, “UV, Visible and IR Attenuation for Alti-

tudes to 50 km,” Air Force Cambridge Research Labora-

tory, No. 285, Cambridge, 1986.

[22] E. P. Shettle and R. W. Fenn, “Models for the Aerosols of

the Lower Atmosphere and the Effects of Humidity

Variations on Their Optical Properties,” Air Force Geo-

physics Laboratory (AFGL), 1979.

[23] J. Binyamin, J. Davies and B. McArthur, “UV-B Cloud

Optical Properties for Canada,” International Journal of

Climatology, Vol. 30, No. 8, 2010, pp. 1246-1255.

doi:10.1002/joc.1966

[24] D. E. Bowker, R. E. Davis, D. I. Myrick, K. Stacy and W.

T. Jones, “Spectral Reflectance of Natural Targets for

Use in Remote Sensing Studies,” NASA reference publi-

cation 1139, NASA, Langley Research Center, Hampton,

1985.

[25] C. Fröhlich and J. London, “Revised Instruction Manual

on Radiation Instruments and Measurements,” WMO

Technology Note 149, World Meteorology Organization,

Geneva, 1986.

[26] R. L. McKenzie, M. Kotkamp, G. Seckmeyer, R. Erb, C.

R. Roy, H. P. Gies and S. T. Toomey, “First Southern

Hemisphere Intercomparison of Measured Solar UV

Spectra,” Geophysical Research Letters, Vol. 20, No. 20,

1993, pp. 2223-2226. doi:10.1029/93GL02359

[27] A. F. Bais, B. G. Gardiner, H. Slaper, M. Blumthaler, G.

Bernhard, R. McKenize, A. R. Wedd, G. Seckmeyer, B.

Kjeldstad, T. Koskela, P. J. Kirsch, J. Gröbner, J. B. Kerr,

S. Kazadzis, K. Leszczynski, D. Wardle, W. Josefsson, C.

Brogniez, D. Gillotay, H. Reinen, P. Weihs, T. Svenoe, P.

Eriksen, F. Kuik and A. Redondas, “SUSPEN Intercom-

parison of Ultraviolet Spectroradiometers,” Journal of

Geophysical Research, Vol. 106, No. D12, 2001, pp.

12509-12525. doi:10.1029/2000JD900561

[28] O. Meinander, S. Kazadzis, M. Blumthaler, L. Ylianttila,

B. Johnsen, K. Lakkala, T. Koskela and W. Josefsson,

“Diurnal Discrepancies in Spectral Solar UV Radiation

Measurements,” Applied Optics, Vol. 45, No. 21, 2006,

pp. 5346-5357. doi:10.1364/AO.45.005346

[29] S. Kazadzis, A. Bais, A. Arola, N. Krotkov, N. Koure-

meti and C. Meleti, “Ozone Monitoring Instrument Spec-

tral UV Irradiance Products: Comparison with Ground

Based Measurements at an Urban Environment”, Atmos-

pheric Chemistry and Physics, Vol. 9, 2009, pp. 585-594.

doi:10.5194/acp-9-585-2009

[30] V. E. Fioletov, J. B. Kerr and D. I. Wardle, “Comparison

of Brewer Ultraviolet Irradiance Measurements with To-

tal Ozone Mapping Spectrometer Satellite Retrievals,”

Optical Engineering, Vol. 41, No. 12, 2002, p. 3051.

doi:10.1117/1.1516818

[31] T. J. Wang, K. S. Lam, Q. Liu and T. K. Chan, “Assess-

ment of Long Term UV Radiation Measured by the

Brewer Spectrophotometer in Hong Kong during 1995-

2005,” Atmospheric and Climate Sciences, Vol. 1, No. 1,

2011, pp. 9-17. doi:10.4236/acs.2011.11002