Pub. Astron. Soc. Pacific, Volume 86, August 1974 ABSOLUTE SPECTRAL ENERGY DISTRIBUTION OF THE NIGHT SKY AT PALOMAR AND MOUNT WILSON OBSERVATORIES BARRY E. TURNROSE Hale Observatories, Carnegie Institution of Washington, California Institute of Technology Received 25 March 1974 The absolute spectral energy distributions of the night sky at Palomar and Mount Wilson Observatories have been obtained by means of photoelectric spectrophotometry. Measurements made on a number of dates are discussed and compared. The average sky brightness at Palomar, still a relatively dark site, shows no very significant long-term changes over the period 1969-73. The average brightness in the continuum at 5400 Â is AB ~ 22 magnitudes per square arc second. The overall sky brightness at Mount Wilson is typically close to two magnitudes brighter than at Palomar, and the artificial mercury emission-line intensities are up to 30 times stronger than at Palomar. Key words: sky brightness absolute spectral energy distribution I. Introduction An increased awareness in the astronomical community of the significance of light pollution for optical observations has prompted a number of recent discussions of the sky brightness at observatory sites: for example, see Walker (1970, 1973), Riegel (1973), Weymann (1973), and Hoag, Schoening, and Coucke (1973). In general, calibrated quantitative measures of the night sky in the literature are of a broadband nature. The high-resolution airglow spectrum of Broadfoot and Kendall (1968) is a notable exception, but the segmented scanning procedure employed hampers the calibration over long baselines in wavelength. For a number of reasons, including, for example, the selection of sky-blocking filter sets, spectrophotometric measurements of the night sky over a large spectral range are desirable. In this paper a representative set of such measurements made at Palomar and Mount Wilson Observatories is presented. II. Palomar Night Sky A. Observations The backlog of observations made with the Palomar multichannel spectrophotometer (MCSP, Oke 1969) over the past several years provides an excellent source of data on the spectral energy distribution of the light from the night sky. Because of the star/sky chopping system employed, the routine observational procedure generates a 545 large set of sky measurements. J. B. Oke very kindly made his MCSP observations available to the author for the purpose of extracting these data. From the complete MCSP data, a sample of 37 long integrations (generally 600 sec) using circular entrance apertures larger than 7" in diameter, obtained on clear, moonless nights from 1969 to 1973, was extracted for processing. The spectral resolution is 80 Â for A ^ 5800 A and 160 Â for A > 5800 Â; the entire observed range is AÀ.3200-11,000. The average dark count for a given observing run was removed from the raw sky counts in each channel, and the data were reduced on the absolute flux system of Oke and Schild (1970), assuming no atmospheric extinction corrections for the sky. The absolute flux of energy f v (ergs sec -1 cm -2 Hz -1 ) entering the telescope from each square arc second of the night sky is thereby obtained. The statistical accuracy of the flux measurements, allowing for uncertainty in the dark count, is typically several percent or less. B. Discussion In Figure 1 the spectral energy distributions for several representative nights throughout the sampling period are shown in temporal order, spectrum A being the earliest. Data identifying each of the spectra in the figure are presented in Table I: successive columns list the spectrum designation, date of observation, declination, hour angle at start and at end of observation,
546 BARRY E. TURNROSE X(A) 11000 10000 9000 8000 7000 6000 5000 4000 3000 LOG 1/ Fig. 1 Absolute spectral energy distributions of the night sky at Palomar Observatory, plotted as functions of log of the frequency in Hz. A wavelength scale is also provided. The AB-magnitude ordinate scale is described in the text; each tick represents a unit of one magnitude. The level AB = 22.0 per square arc second is indicated for each spectrum, to provide the normalization. In general, standard deviation bars are too small for display; in regions where they become significantly large, however, typical error bars are shown. The positions of the following prominent natural and artificial emission lines are indicated: [O i] 6364 and 6300, Nai 5892, [Oi] 5577, and Hg 5791, 5770, 5461, 4358, 4047, 3663, 3655, and 3650. The small tickmarks show the positions of the principal airglow OH emission bands observed by Broadfoot and Kendall (1968). Bandpasses are 80 À for A ^ 5800 Â, 160 À for A > 5800 À. See Table I for identification of each spectrum.
ENERGY DISTRIBUTION OF THE NIGHT SKY 547 Spectrum Fig. 1 A Fig. 1 B Fig. 1 C Fig. 1 D Fig. 1 E Fig. 2 A Fig. 2 B Date 13 Aug. 6 Jul. 20 Mar. 28 Nov. 1 Dec. 5 Jan. 5 Jan. TABLE I Identification of Spectra in Figures 1 and 2 Declination H. A. Start H. A. Finish n (h:m) (h:m) 1969 1970 1971 1972 1972 1973 1973-5 + 38 + 46 + 9 + 9-8 + 22 W 1:38 W 1:46 W 0:16 W 0:17 W 0:25 E 0:06 W 1:51 W 2:31 W 2:31 W 1:25 W 1:20 W 1:25 W 1:15 W 3:09 (sec 2) 1.52 1.12 1.04 1.13 1.13 1.36 1.24 and the average value of the secant of the zenith angle. The ordinate scale in Figure 1 is AB magnitude (AB = 2.5 log^ 48.60) per square arc second. Oke and Schild (1970) have discussed the relationship between the standard V magnitude and the zero-air-mass flux density f v at 5480 Â; the constant in the definition of the AB magnitude is chosen so that AB 5480 ~ V. (The reader is reminded that the energy fluxes and the associated AB magnitudes for the sky given in this paper refer to radiation incident upon the telescope and thus are not referred to zero air mass.) The data indicate that there is a characteristic sky spectrum, dominated by airglow, which can in an overall sense usefully represent the true energy distribution at any given time. Notable features of the typical spectrum are a large and highly structured energy flux in the red and infrared region due to numerous strong atmospheric OH emission bands, a strong [Oi] 5577 Â atmospheric emission line, and a comparatively smooth energy distribution in the blue and violet region (primarily due to overlapping 0 2 and other atmospheric bands see Chamberlain 1961) punctuated with mercury emission lines from artificial sources. The long-term average behavior of the sky spectra in the sample (extending over a threeand-one-half-year period) is characterized by relative stability at low light levels. Flux-vs.- time graphs of the sky brightness at 10,000 Â, in the continuum at 5400 Â, and in the "continuum underlying the 4358 Â Hg line reveal no significant secular trends: nightly and monthly variations are much more outstanding (these shortterm variations are discussed below). The aver- age flux level in the continuum at 5400 Â is about 22 AB magnitudes per square arc second. A flux-vs.-time graph of the sky brightness at about 3570 Â reveals a marginal secular decrease in the average level, although, as in the other spectral regions, large short-term variations are dominant. (The lack of clear-cut secular trends over the whole spectrum despite the fact that the start of the sampling period was near sunspot maximum is perhaps not surprising in view of the difficulties that have been encountered historically in attempting systematically to correlate long-term airglow trends with possible causes, such as sunspot activity see Chamberlain 1961, p. 509.) Despite the overall long-term average similarity of the sky spectra, significant short-term variations in particular features are quite possible. The spectra in Figure 1 illustrate the variability of the 5577 Â line and the blue-violet "continuum, for example; of particular note in this regard are spectra D and E, taken only three days apart. Such short-term variations in the airglow are well known: Chamberlain (1961) presents a discussion of this point, and the reader is referred to his work for a thorough analysis. We remark here simply that the character of the variations evident in our sky spectra seems consonant with what is known about the systematics of airglow variations, viz., the "covariance group concept (Chamberlain 1961, p. 517). The Palomar sky spectra always exhibit artificial contamination by mercury emission lines. The absolute strengths of these lines show substantial changes: for example, an intensity-vs.- time plot for the 4358 Â Hg line exhibits largeamplitude short-term variability about a mean intensity of approximately 3.5 X 10 16 ergs
548 BARRY E. TURNROSE sec -1 cm -2 (arc second) -2. The changes in mercury-line strength illustrated by the spectra of Figure 1 are indicative of this short-term variability, which can easily result in light levels differing from the mean in either direction by factors of 2 or 3. (Local meteorological conditions at the source, such as low fog, are the principal cause of the variability in the artificial component of the sky brightness. Additionally, at any given time the mercury-line intensities show a positional variation in the sky, being greatest west of the meridian, i.e., towards San Diego and Los Angeles.) There is some evidence that the mercury-line intensities have on the average increased during the last few years; the evidence, however, is marginal in view of the large short-term variations which are always present. While evidenced principally by the mercury lines, artificial contamination of the night sky at Palomar exists, if in but small amounts, in other spectral regions as well. This is inferred from a slight azimuthal dependence of the sky brightness similar to that exhibited by the mercury radiation. Gunn (private communication) has found from direct photographic work that on a given night, in a bandpass from 6000 Â to 7000 Â, the sky brightness west of the meridian is typically about 25% higher than in the east (where there are no cities). Nevertheless, the absolute level of the sky brightness indicates that Palomar is still a relatively dark observing site. (One may compare, for example, the flux levels with Walker s (1970) value of V ~ 22.0 per square arc second for a completely dark site. Integration of the f k in Table II (see below) over the V-band response curve yields a computed approximate V magnitude of 21.5 per square arc second on 28 November 1972. The spectrum used, D in Figure 1, represents one of the brighter scans and shows a rather strong [ O i] 5577 Â line; it was made at an average zenith angle of 28 to the SW. This computed V is several tenths of a magnitude brighter than Walkers (1970) measurement to the NW at Palomar in 1966.) In Table II we present a compilation of f k, incident absolute energy flux per unit wavelength (ergs sec 1 cm 2 Â 1 ), per square arc second, as a function of wavelength for the sky on the night of 28 November 1972. This table is TABLE II Absolute Energy Flux per Unit Wavelength per Square Arc Second for Sky at Palomar Observatory 28 November 1972 80-A Bandpasses 16Q-A Bandpasses Wavelength (A) 3180 3260 3340 3420 3500 3580 3660 3740 3820 3900 3980 4060 4140 4220 4300 4380 4460 4540 4620 4700 4780 4860 4940 5020 5100 5180 5260 5340 5420 5500 5580 5660 5740 V 4.30 5.18 6.13 4.75 4.86 5.29 7.24 4.75 4.43 3.45 4.31 8. 58 6.09 5.83 5.39 11.40 6.25 6.38 6.16 6.27 6.14 6.45 6.24 5.60 5.80 6.37 6.26 6.56 7.85 11.00 25.40 7.78 9.70 Wavelength (A) 5760 5920 6080 6240 6400 6560 6720 6880 7040 7200 7360 7520 7680 7840 8000 8160 8320 8480 8640 8800 8960 9120 9280 9440 9600 9760 9920 10080 10240 10400 10560 10720 10880 V 9.43 11.40 7.89 13.00 9.60 8.36 6.67 9.73 7.11 9.55 13.80 10.70 13.20 23.60 16.60 5.54 22.70 19.30 20.10 36.10 28.30 8.22 21.40 32.40 15.80 26.30 66.00 68.30 99.60 87.10 25.80 64.30 134.00 *f k measured in units of 10 18 ergs sec -1 cm -2 Â -1 (a second) -2. included to facilitate numerical calculations based on the night-sky spectral energy distribution. The relation between f k at 5480 Â and V magnitude is implicit in the description of the AB magnitude scale given earlier. For the
ENERGY DISTRIBUTION OF THE NIGHT SKY 549 reader s convenience, however, we cite explicitly the Oke and Schild (1970) relation: for a star of V magnitude 0.00 the zero-air-mass flux at 5480 Â is f k = 3.64 X 10-9 ergs sec -1 cm -2 Â -1. III. Mount Wilson Night Sky A. Observations Data on the night sky at Mount Wilson were obtained by the author with the Cassegrain single-channel scanner on the 100-inch (254-cm) telescope in the course of a spectrophotometric galaxy-observing program between February 1972 and January 1973. The observations consist of a series of 60-sec integrations through a 10" entrance aperture at 26 discrete wavelengths between 3580 Â and 7400 Â. Except for 40 Â bandpasses centered near 3580 Â and 3630 Â, all bandpasses for A. = 4600 Â are 20 Â; those for A. > 4600 Â are 30 Â. The centering of the bandpasses was chosen for the galaxy program and varies from object to object; features of interest in the night sky are thus included only coincidentally [ O i] 5577 Â is a notable exclusion. The average dark count was subtracted from all measurements on a nightly basis, and reduction was carried out as for the Palomar data, on the Oke-Schild absolute flux system with no atmospheric extinction corrections for the sky. Because of the relatively short integration times and small bandpasses used, the statistical accuracy of individual measurements is typically ~ 10% except in the violet region where somewhat larger uncertainties are encountered. B. Discussion The Mount Wilson sky-brightness data are rather sparse in comparison with the complete spectral coverage provided by the MCSP observations. No information on the far red or infrared portion of the spectrum is available, and gaps exist in the region observed. The data which are available, however, are useful in a comparative sense. Figure 2 presents the measured absolute energy distribution of the night sky at Mount Wilson. Identifying data are given in Table I. Only two spectra, both obtained on the same night, are shown since they illustrate well the range of characteristics exhibited by the full Mount Wilson sample. Dominated as it is by artificial illumination from the Los Angeles basin, the sky at Mount Wilson undergoes changes in overall flux level which are in general more pronounced than the variations observed at Palomar. The difference in sky brightness between the two spectra in Figure 2 illustrates the considerable dependence upon position in the sky at Mount Wilson; this exceeds the directional dependence of the sky brightness observed at Palomar over the normal working range of small to moderate zenith distances. Night-to-night changes, influenced largely by local meteorological conditions in the Los Angeles basin, are of the same order of magnitude as the differences in Figure 2, with average flux levels similar to that of Spectrum A. Comparison with Figure 1 then shows that the overall sky brightness at Mount Wilson typically exceeds that at Palomar by nearly two magnitudes. No data on long-term trends at Mount Wilson are available due to the short sampling time. The strong mercury lines are a striking feature of the Mount Wilson sky spectra; some caution is required, however, in interpreting Figure 2 in this respect. In the first place, it should be noted that the absence of the 4047 Â line in Spectrum B of Figure 2 is due entirely to the bandpass centering employed: the line is completely excluded from the bandpass which measures it in Spectrum A. Also, the violet group of mercury lines is not completely included in Spectrum B. In the second place, the bandpasses containing mercury lines in the Mount Wilson spectra are 40 Â and 20 Â wide; corresponding bandpasses in the Palomar spectra are 80 Â wide. Consequently, comparison of emission-line strengths at the two sites cannot be made by casual visual inspection of Figures 1 and 2, which are based on flux densities averaged over different bandpasses. It is of course a simple matter to calculate the absolute line strengths using the proper bandpass: the intensity of the 4358 Â line in Spectrum A of Figure 2 is 1.2 X 10 14 ergs sec - l cm -2 (arc second) -2, and in Spectrum B, 5.7 X 10-15 ergs sec -1 cm -2 (arc second) -2. Typically, therefore, the average mercury-line intensities at Mount Wilson exceed those at Palomar by factors on the order of 10-30. I wish to thank Drs. A. R. Sandage and W. L. W. Sargent for suggesting this work, Dr. J. E. Gunn for communicating his photographic data, and an anonymous referee for several useful comments. I wish to thank especially Dr. J. B. Oke
550 BARRY E. TURNROSE \(A) 1000 10000 9000 8000 7000 6000 5000 4000 1 1 T T T 3000 TI MOUNT WILSON SKY AB MAGNITUDE PER SQUARE ARCSECOND A i 22.0- B i * 22.0- i * % ï. i x 'TI 1.0 14.40 14.50 14.60 Nal 14.70 LOG v r Hg Hg Hg 14.80 14.90 15.00 Fig. 2 Absolute spectral energy distributions of the night sky at Mount Wilson Observatory. The abscissa and ordinate scales are as in Figure 1. The marks at AB = 22.0 per square arc second provide the normalization and allow comparison with Figure 1. Only selected representative error bars are shown. The positions of Na i 5892 and Hg 4358, 4047, 3663, 3655, and 3650 are indicated; these are the only lines measured in the Mount Wilson scans. Bandpasses are 40 Â for A < 3700 Â, 20 Â for 3700 À ^ A ^ 4600 Â, and 30 Â for A > 4600 Â; exact bandpass centering is slightly different for the two spectra. See text for discussion of the mercury-line strengths and Table I for identification of each spectrum.
ENERGY DISTRIBUTION OF THE NIGHT SKY 551 for suggesting the use of MCSP observations and making them available, and for many very helpful discussions and suggestions regarding both the course of this work and its presentation in this paper. I wish also to thank the Hale Observatories for observing time at both Mount Wilson and Palomar. REFERENCES Broadfoot, A. L., and Kendall, K. R. 1968, J. Geophys. Res. 73, 426. Chamberlain, J. W. 1961, Physics of the Aurora and Airglow (New York: Academic Press). Hoag, A. A., Schoening, W. E., and Coucke, M. 1973, Pub. A.S.P. 85, 503. Oke, J. B. 1969, Pub. A.S.P. 81, 11. Oke, J. B., and Schild, R. E. 1970, Ap. J. 161, 1015. Riegel, K. W. 1973, Science 179,1285. Walker, M. F. 1970, Pub. A.S.P. 82,672. 1973, ibid. 85, 508. Weymann, R. J. 1973, Mercury 2, 2 (No. 2).