Ground-level solar spectral irradiance in Glasgow: an inter-comparison

of two sites

H. Moseley1,2, I. Clark3, A. Pearson3, J. Smyth2, H. Oliver1,2, R. M. Mackie4

1The Photobiology Unit, University of Dundee, Ninewells Hospital & Medical School, Dundee, UK, 2Department of Medical Physics, Ninewells

Hospital & Medical School, Dundee, UK, 3National Radiological Protection Board, Hardgate Road, Glasgow, UK and 4Department of Public Health,

University of Glasgow, Lilybank Gardens, Glasgow, UK

Background: Solar spectral radiometry presents

significant challenges to produce accurate and repro-

ducible data. To investigate the reliability of the

measurements, several inter-comparisons have been set

up. Although these are useful, their main drawback is

that equipment must be dismantled and transported to

a common site and re-calibrated.

Methods: In this study, an inter-comparison has

been performed of two spectroradiometers that are

located 3 miles apart some 30m above sea level. These

two systems have operated using different calibration

techniques. Data were compared on clear days, to

minimise actual differences in ultraviolet irradiation.

Results: There were substantial differences at some

individual wavelength points, but overall the mean

difference of results at 5 nm intervals on an individual

scan from the two systems agreed to within 11%. If

the data were used to compute the erythemal

irradiance, the differences were reduced to 4%.

Conclusion: This study demonstrates both the lim-

itations and the level of reliability that might be

expected from these systems operating under careful

scientific supervision.

Key words: inter-comparison; spectroradiometer; sun-

light.

There is strong evidence that ultraviolet (UV)

radiation from sunlight is a major causal factor

for squamous cell carcinoma (SCC), basal cell

carcinoma (BCC) and malignant melanoma (MM)

(1, 2). However, the relationship between exposure

and incidence is not the same in each case. For SCC, it

appears that the risk increases in proportion to

cumulative exposure to UV radiation. For BCC,

however, risk increases to a certain point beyond

which greater cumulative exposure does not appear to

increase the risk any further; and as regards MM, it

seems that intermittent, intense exposure is more

important than cumulative exposure (2).

Incidence of skin cancer has been rising in recent

decades. In Scotland, the age-standardised incidence

of cutaneous MM over the period 1979–1994 showed

a significant increase from 3.5 to 7.8 per 100 000 per

year for males and from 6.8 to 12.3 per 100 000 for

females (3). During the same period, world-standar-

dised rates increased from 2.7 to 6.0 per 100 000 per

year in men and from 4.6 to 8.5 in women (3). The

increased incidence of all types of skin cancer is

thought to be linked to increased UV exposure,

mainly from recreational exposure. This is largely due

to increased leisure time allied to the social desirability

for having a tan (4–6).

Another factor which has caused concern recently is

the depletion of stratospheric ozone which has a

major role in filtering out short-wavelength UV

radiation from sunlight. Whilst originally observed

in Antarctica, ozone depletion has also been reported

in the Northern Hemisphere (7). UVB radiation (280–

315 nm) has also been shown to have increased at

ground level during periods of ozone depletion (8, 9).

Findings such as these and the continual release of

ozone depleting substances into the atmosphere

provide justification for real concern about the UK’s

most prevalent cancer, skin cancer.

As part of an area of ongoing research, two

independent centres have been measuring solar UV

radiation using rooftop spectroradiometers in Glas-

gow. One system has been in place since January 1994

on the roof of one of the buildings at the University of

Glasgow, a currently collaborative project with the

Photodermatol Photoimmunol Photomed 2004; 20: 138–143Blackwell Munksgaard

CopyrightrBlackwellMunksgaard 2004

138

University of Dundee. This system, situated in

Glasgow, provided the first recorded spectral data in

the UK that UVB radiation was increased during a

period of ozone depletion (9). Because of the

significance of this finding, it is extremely important

that there should be some corroboration of the data

obtained from this device. The other system used in

the current inter-comparison was located at a distance

of approximately 3 miles on the roof of the National

Radiological Protection Board (NRPB) in Scotland.

The independent operation and proximally close

citing of these two instruments has provided Glasgow

with a unique opportunity for corroboration of results

in the United Kingdom. Inter-comparison and

verification of spectroradiometric data is paramount

(10) due to the technical difficulties involved in

making accurate measurements of the solar spectrum.

The purpose of the present study was to carry out

an inter-comparison of results from the two systems.

The systems were operated completely independently,

using different procedures for collecting and calibrat-

ing data. Scanning protocols were significantly

different. Because of their close geographical proxi-

mity, it was anticipated that UV levels would be

similar in both places. Notwithstanding, it has been

recognised that local variations may well exist

between the two locations. Previous inter-compari-

sons have suffered from problems associated with the

need to move equipment to a common location. In

this new piece of work, two systems are compared

in situ.

Materials and methodsGlasgow University system

Light enters via a 4 in integrating sphere mounted

outdoors and is transmitted to a double-grating

spectroradiometer (Spex 1680, Jobin Yvon Ltd,

Stanmore, UK incorporating an R928 photomulti-

plier tube) by an optical fibre. Grating dispersion was

0.9 nm/mm. Slit width was 1.25mm and bandwidth

was approximately 1.2 nm. Each hour, three scans are

performed at 0.1 nm intervals from 280 to 400 nm,

and the mean raw data value at each wavelength is

recorded. Dark current is measured prior to each scan

and subtracted from the sunlight measurements.

Scans begin on the hour and take a total of 25min

to complete. Stray light checks confirmed that this

was below the limit of detection. Monthly checks are

carried out using a mercury lamp (for wavelength)

and a tungsten lamp (for gain) in situ and every 6

months absolute calibrations are performed using

both a tungsten filament lamp and a deuterium lamp.

These lamps have been calibrated at the National

Physical Laboratory (NPL, Teddington, UK). To

carry out the absolute calibrations, the integrating

sphere and fibre optic bundle are removed from the

outdoor location and brought inside. Calibration files

are archived, along with the raw data from the hourly

scans. Wavelength checks are carried out using

mercury lines and solar absorption lines.

NRPB system

Light enters via a 15mm diffuser, made of a

polytetrafluoroethylene (PTFE)-type material and

mounted outdoors, and is transmitted via an optical

fibre to a double-grating monochromator (Spex 1680

incorporating gratings ruled at 1200 g/mm an R928

photomultiplier tube). Slit width was 0.5mm and

bandwidth was approximately 1.1 nm. Dark current is

measured prior to each scan and subtracted from the

sunlight measurements. Scans are performed from 280

to 400 nm at 1 nm intervals every 15min and each

scan is recorded. Scans take 101 s to complete. Stray

light checks confirmed that this was below the limits

of detection. Calibration is carried out using a

mercury lamp and a deuterium lamp calibrated at

NPL in situ. To achieve this, the deuterium lamp is

located indoors and coupled to the outdoor spectro-

radiometer via an optical fibre. The calibration factor

is applied to the incoming data, which is saved as

calibrated data. In addition, the wavelength is

checked regularly using solar absorption lines.

Inter-comparison

Two clear days were used for the inter-comparison.

These were 29 May 1997 and 21 July 1997. Clear days

were used to minimise the discrepancy at the two sites

caused by variation in cloud cover. This was an

essential requirement to ensure that any variations

noted were attributable primarily to the two measure-

ment systems, and not to variable climatic conditions.

The scan results were recorded as paired data,

comprising a number which corresponds to wave-

length and another which corresponds to UV data,

calibrated in the case of the NRPB site and raw

(uncalibrated) data from the Glasgow University.

Collection of raw data allows retrospective correction

and calibration. Inspection of the raw data showed

that wavelength drift had occurred in the Glasgow

University system. The time dependency of the

wavelength shift was easily accounted for by correct-

ing using the most recent wavelength checks to the

raw data. This resulted in wavelength accuracy to

within 0.1 nm using mercury lines.

139

Ground-level solar spectral irradiance in Glasgow

Two sets of calibrated data were produced from the

Glasgow University spectroradiometer since it was

calibrated using a 30W deuterium lamp and a 100W

tungsten filament lamp, both calibrated at NPL. This

was done in order to provide a comparison between

the two calibration methods. The deuterium lamp is a

point source, which is an attractive feature, but the

output decreases at longer wavelengths. On the other

hand, the tungsten filament lamp has increasing

output at longer wavelengths, thus more resembling

sunlight, but it is a large area source. The NRPB

system is calibrated using a 30W deuterium lamp and

so produces a single data set at each time point.

Since each scan generates over 1000 data points,

analysis of the data was extremely time consuming.

For practical purposes, the inter-comparison was

limited to data collected on four different occasions.

ResultsThe uncertainty budget is given in Table 1. Expanded

uncertainty at 95% confidence level was 5.8% for the

Glasgow University system and 8.4 % for the NRPB

system.

Results of the solar scans performed at 10:00 hours

on 29 May 1997 are plotted in Fig. 1. This includes

NRPB data using deuterium lamp calibration and

Glasgow University data based on both the tungsten

and deuterium calibrations, i.e. all three sets of data).

Figure 2 shows data from the 29 May 1997 scan at

12:00 hours. Then 21 July 1997 at 10:00 hours (Fig. 3)

and 21 July 1997 at 12:00 hours (Fig. 4) are also

shown. These data demonstrate a substantial degree of

consistency, albeit closer inspection reveals significant

discrepancy at particular data points. For example,

310 nm on 29 May at 10:00 hours gave a value of only

0.058 at NRPB compared with 0.071–0.073 at the

University. This pattern was repeated at the other

scans, although less marked. This may be due to slight

wavelength discrepancy between the two systems as

this wavelength corresponds to part of the spectrum

where there is some environmental absorption and

irradiance is changing rapidly (in percentage terms).

Table 1. Spreadsheet model showing the uncertainty budget

Source of uncertainty Value (%) Probability distribution Divisor Standard uncertainty (%)

(a) Glasgow University system

Spectroradiometer current 2.0 Normal 1 2.0

Positioning distance 1.0 Rectangularp3 0.58

Positioning angle 0.5 Rectangularp3 0.29

Standard lamp 2.0 Normal 1 2.0

Combined uncertainty Normal 2.9

Expanded uncertainty Normal (k5 2) 5.8

(b) NRPB system

Spectroradiometer current 3.7 Normal 1 3.7

Positioning distance 1.2 Rectangularp3 0.7

Positioning angle 1.2 Rectangularp3 0.7

Standard lamp 1.8 Normal 1 1.8

Combined uncertainty Normal 4.2

Expanded uncertainty Normal (k5 2) 8.4

0.0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

290 310 330 350 370 390

Wavelength (nm)

Spe

ctra

l Irr

adia

nce

(W/m

2 /nm

)

NRPBGlasgow University (tungsten halogen calibration)Glasgow University (deuterium calibration)

Fig. 1. Solar scans on 29 May 1997 at 10:00 hours.

NRPB

Glasgow University (tungsten halogen calibration)

Glasgow University (deuterium calibration)

0.0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

290 310 330 350 370 390

Wavelength (nm)

Spe

ctra

l Irr

adia

nce

(W/m

2 /nm

)

Fig. 2. Solar scans on 29 May 1997 at 12:00 hours.

140

Moseley et al.

It is instructive to consider the ratio of values at

each wavelength. This was also calculated and is

shown in Table 2. Mean ratios ranged between 0.89

and 1.02. In other words, the overall mean differences

at the two sites differed by no more than about 10%.

These differences were significant at the Po0.01 level

at the 12:00 hour scans on both dates.

In addition, erythemal-weighted irradiances were

calculated using the standard Commission Interna-

tionale de l’Eclairage (CIE) erythemal effectiveness

spectrum (11). Results are shown in Table 3. On the

basis of the erythemal-weighted irradiances, overall,

taking the average of all four readings the mean from

the Glasgow University spectroradiometer (tungsten

calibrated) was 1.6% higher than that from the NRPB

spectroradiometer. Using the Glasgow University

(deuterium lamp-calibrated) values, the Glasgow

University system was 4.0% less than the NRPB.

DiscussionDifficulties in measurement of solar UV radiation

have been well recognised (12) and in recent years

several inter-comparison exercises have been under-

taken in order to verify participants’ instruments and

calibration methodologies (13, 14). These inter-

comparisons have contributed significantly to im-

provements in the accuracy of measurements and

standardisation of methodologies for solar UV

monitoring (15). To date, the best results are that

spectroradiometers can agree with each other and a

reference to within � 6% over a wavelength range of

300–393 nm (14). Even so, the quest for accuracy in

solar UV monitoring is still a major concern.

The purpose of the present study was to compare

the two spectroradiometers which were in location in

Glasgow under clear sky conditions. One of the main

difficulties in the inter-comparison related to the time

and bandwidths employed. As regards the timing, this

will inevitably introduce variation in the two values.

The Glasgow University system took three scans in

the hour, this procedure took 25min and the mean

was recorded. This contrasts with the NRPB metho-

dology where scans were taken every 15min and

scanning times were only 101 s. Nonetheless, this has

provided a unique opportunity to compare these two

sets of data. Clearly, there are some significant

discrepancies at individual wavelengths; particularly

where the signal level is low (i.e. short wavelengths)

and is changing rapidly. One should therefore be

cautious in interpreting values at individual wave-

lengths. Examining the ratio of spectral irradiances

from the two systems revealed that the mean ratio

ranged between 0.89 and 1.03. The ratios differed

significantly (Po0.01) from 1.00 at the 12:00 hours

scans. Further investigation is required to determine

the reason for this disagreement, which may be due to

the time difference of the scans, or different angular

response of the input optics. Overall, the mean

difference was approximately 10%.

There is considerable interest in monitoring erythe-

mal-weighted irradiance because of its biological

significance. However, most observers use filtered

broad-band detectors which are relatively low-cost,

easy-to-use devices. When used under optimal condi-

tions, errors may be less than 8% or 20% depending

on the device, provided they are calibrated against a

spectroradiometer in the field (16). Otherwise, errors

may be of the order of 40%. Using the spectro-

radiometers to obtain the erythemal-weighted irra-

diances in the present study showed a remarkable

agreement between the two systems. Differences of

between 2% and 4% in the mean values show very

good agreement in measuring this parameter.

The fact that both instruments used the same type

of monochromator (Spex 1680) would be expected to

NRPB

Glasgow University (tungsten halogen calibration)

Glasgow University (deuterium calibration)

0.0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

290 310 330 350 370 390

Wavelength (nm)

Spe

ctra

l Irr

adia

nce

(W/m

2 /nm

)

Fig. 3. Solar scans on 21 July 1997 at 10:00 hours.

NRPB

Glasgow University (tunsten halogen calibration)

Glasgow University (deuterium calibration)

0.0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

290 310 330 350 370 390

Wavelength (nm)

Spe

ctra

l Irr

adia

nce

(W/m

2 /nm

)

Fig. 4. Solar scans on 21 July 1997 at 12:00 hours.

141

Ground-level solar spectral irradiance in Glasgow

contribute to the level of agreement obtained

although there were significant differences in the

calibration methodologies. Whereas NRPB per-

formed the calibration in situ using an optical fibre

to transmit the light from the calibration lamp, the

University moved the integrating sphere and fibre

(without any dismantling) indoors where the calibra-

tion was carried out. The path of traceability of the

calibrations is probably a significant factor in the

agreement of these results. If there were calibrations

made according to the spectral irradiance scales held

at other National Standards Laboratories, then

further uncertainty could be introduced due to the

differences in the spectral scales (17). The more

accurate the metrology itself becomes, the greater

the uncertainty contribution from the standard lamps

becomes (18). This study goes some way to demon-

strating that lamps from the same standards lab can

contribute to reducing variability.

The fact that these instruments remained in situ

throughout this study is the other major factor that

can be said to validate these results. Moving any

spectroradiometric equipment creates uncertainty

because instruments are very sensitive to mechanical

forces caused by transporting these instruments.

Spectroradiometers have been described as ‘notor-

iously unstable [and] complex’ (19) and should be

handled with care. In order to allow the instruments

to be transported safely, they must be carefully

dismantled and packed, and at the site chosen for

the inter-comparison, they are then unpacked and re-

assembled. This means that the spectroradiometers

must be re-calibrated. In the present study, it was

possible to compare instruments in their normal

Table 2. Ratio of spectral irradiances from Glasgow University site (two separate calibration methods) and NRPB site, on two separate days with

clear sky conditions

Wavelength

(nm)

Ratio of spectral irradiances

29 May 1997, 10:00 hours 29 May 1997, 12:00 hours 21 July 1997, 10:00 hours 21 July 1997, 12:00 hours

University

(tungsten

halogen):

NRPB

University

(deuterium):

NRPB

University

(tungsten

halogen):

NRPB

University

(deuterium):

NRPB

University

(tungsten

halogen):

NRPB

University

(deuterium):

NRPB

University

(tungsten

halogen):

NRPB

University

(deuterium):

NRPB

290 1.110 1.035 0.881 0.823 1.944 1.815 0.881 0.823

295 0.802 0.769 0.929 0.890 0.740 0.710 0.929 0.890

300 1.094 0.983 1.124 1.014 1.000 0.899 1.124 1.014

305 1.075 1.000 1.078 1.004 0.395 0.368 1.078 1.004

310 1.243 1.223 1.176 1.156 1.156 1.137 1.176 1.156

315 0.986 0.937 0.949 0.899 0.932 0.879 0.949 0.899

320 0.955 0.890 1.013 0.941 1.020 0.950 1.013 0.941

325 0.955 0.959 0.931 0.934 0.946 0.950 0.931 0.934

330 0.945 0.864 0.861 0.787 0.878 0.800 0.861 0.787

335 0.995 0.937 0.896 0.845 0.919 0.865 0.896 0.845

340 1.000 0.934 0.902 0.821 0.923 0.842 0.902 0.821

345 1.026 0.893 1.010 0.878 1.043 0.906 1.010 0.878

350 1.067 0.874 0.949 0.779 0.978 0.800 0.949 0.779

355 0.959 0.787 0.847 0.696 0.886 0.726 0.847 0.696

360 0.988 0.842 0.956 0.816 1.003 0.858 0.956 0.816

365 1.025 1.000 0.875 0.855 0.927 0.906 0.875 0.855

370 0.989 0.977 0.925 0.914 0.971 0.959 0.925 0.914

375 1.021 1.058 0.874 0.904 0.916 0.947 0.874 0.904

380 1.002 0.998 0.862 0.859 0.903 0.899 0.862 0.859

385 1.088 1.094 0.941 0.946 0.994 0.997 0.941 0.946

390 0.998 1.008 0.956 0.963 1.013 1.021 0.956 0.963

395 1.334 1.338 0.926 0.929 0.983 0.986 0.926 0.929

Mean 1.030 0.973 0.948 0.893 0.976 0.919 0.948 0.893

SD 0.107 0.130 0.086 0.096 0.259 0.249 0.086 0.096

Significance NS NS Po0.01 Po0.01 NS NS Po0.01 Po0.01

NRPB, National Radiological Protection Board; NS, non-significant.

Table 3. Erythemally weighted spectral irradiances (mW/m2) from

Glasgow University site (two separate calibration methods) and

NRPB site, on two separate days with clear sky conditions

NRPB

University

(tungsten

halogen)

University

(deuterium)

29 May 1997, 10:00 hours 105.8 108.9 103.0

29 May 1997, 12:00 hours 118.9 130.6 123.5

21 July 1997, 10:00 hours 99.2 91.3 86.4

21 July 1997, 12:00 hours 116.8 118.0 111.4

Mean 110.2 112.2 106.1

NRPB, National Radiological Protection Board.

142

Moseley et al.

operating condition. The actual transportation of

many instruments in traditional inter-comparisons

can be levelled as a major criticism in the reliability

of the results from these collective gatherings once

the instruments are returned to their home sites.

The proximity of monitoring sites such as those in

Glasgow gives an ideal opportunity for cross veri-

fication of results without uncertainty from trans-

portation.

Although individual data points showed significant

differences, overall there was a relatively small degree

of disagreement between the two systems. This study

may also provide a useful model for further verifica-

tion of measurements made at sites which are

independent. It has long been acknowledged that this

is a challenging area of meteorology with many

potential pit-falls (20). This study has shown that

there can be reasonable agreement between measure-

ment centres, for example erythemal-weighted irra-

diance differences of no more than 4% in the present

series. Finally, this inter-comparison has provided

corroboration of the reliability of the system that was

used to record the increased UVB spectrum in the UK

at the same time stratospheric ozone levels were

reduced (8).

References1. Swerdlow AJ. Epidemiology of chronic disease risks in relation

to ultraviolet radiation exposure. Rad Prot Dos 2000; 91: 19–23.

2. MacKie RM. Effects of ultraviolet radiation on human health.

Rad Prot Dos 2000; 91: 15–18.

3. MacKie RM, Hole D, Hunter JA, et al. Cutaneous malignantmelanoma in Scotland: incidence, survival, and mortality,

1979–94. The Scottish Melanoma Group. Br Med J 1997; 315:

1117–1121.

4. Mawn VB, Fleischer AB. A survey of attitudes, beliefs, andbehaviour regarding tanning bed use, sunbathing and sunsc-

reen use. J Am Acad Dermatol 1993; 29: 959–962.

5. Shuttleworth D. Sunbeds and the pursuit of the year-roundtan. BMJ 1993; 307: 1508–1509.

6. Monfrecola G, Fabbrocini G, Posteraro G, Pini D. What do

young people think about the dangers of sunbathing, skin

cancer and sunbeds? A questionnaire survey among Italians.Photodermatol Photoimmunol Photomed 2000; 16: 15–18.

7. Niu XF, Frederick JE, Stein ML, Tiao GC. Trends in column

ozone based on TOMS data – dependence on month, latitude,

and longitude. J Geophys Res Atmos 1992; 97: 14661–14669.

8. Kerr JB, McElroy CT. Evidence for large upward trends of

ultraviolet-B radiation linked to ozone depletion. Science 1993;

262: 1032–1034.

9. Moseley H, MacKie RM. Ultraviolet B radiation wasincreased at ground level in Scotland during a period of ozone

depletion. Br J Dermatol 1997; 137: 101–102.

10. Herman JR, Hudson RD, Serafino G. Analysis of the eight-year trend in ozone depletion from empirical models of solar

backscattered ultraviolet instrument degradation. J Geophys

Res Atmos 1990; 95: 7403–7416.

11. CIE Standard erythema reference action spectrum andstandard erythema dose. CIE S007/E-1998, Commission

Internationale de l’Eclairage, Vienna, 1998.

12. Steinmetz M. Uncertainty Evaluation, UV monitoring of solar

radiation. Appendix 3.4, Working Group 2, Final Report, BA41-BA 48, UV News, 2000.

13. Koskela T. The Nordic intercomparison of ultraviolet and

total ozone instruments at Izana from 24 October to 5

November 1993. Final Report. Meteorological Publications,Vol. 27, Finnish Meteorological Institute, Helsinki, 1994.

14. Kjeldstad B, Johnson B, Koskela T. The Nordic intercompar-

ison of ultraviolet and total ozone instruments at Izana,October 1996. Final Report. Meteorological Publications, Vol.

36, Finnish Meteorological Institute, Helsinki, 1997.

15. Gardiner BG, Kirsch PJ. Setting standards for European

ultraviolet spectroradiometers. Final Report of ContractSTEP-CT 90076, 138, Brussels, Commission of the European

Communitites, 1995.

16. Wong CF, Toomey S, Fleming RA, Thomas BW. UV-B

radiometry and dosimetry for solar measurements. HealthPhys 1995; 68: 175–184.

17. Lambe RP, Saunders R, Gibson C, Hollandt J, Tegeler E. A

CCPR international comparison of spectral radiance measure-ments in the air-ultraviolet. Metrologia 2000; 37: 51–54.

18. Jokela K, Ylianttila L, Visuri R, Leszczynski K, Karha P,

Ikonen E. Intercomparison of lamp and detector-based UV-

irradiance scales for solar UV radiometry. J Geophys ResAtmos 2000; 105: 4821–4827.

19. Moore J. Sources of error in spectroradiometry. Lighting Res

Technol 1980; 12: 213–220.

20. Bernhard G, Seckmeyer G. Uncertainty of measurements ofspectral solar UV irradiance. J Geophys Res Atmos 1999; 104:

14321–14325.

Accepted for publication 25 February 2004

Corresponding author:

Harry Moseley

The Photobiology Unit

University of Dundee

Ninewells Hospital & Medical School

Dundee DD1 9SY

UK

e-mail: H.Moseley@dundee.ac.uk

143

Ground-level solar spectral irradiance in Glasgow