4 ESO-Paranal night sky brightness survey

The results we will discuss have been obtained between April 1 2000 and September 30 2001, corresponding to ESO Observing Periods 65, 66 and 67, and include data obtained on 174 different nights. During these eighteen months, 4439 images taken in the UBVRI passbands and processed by the FORS pipeline were analysed and 3883 of them ($\sim$88%) were judged to be suitable for sky brightness measurements, according to the criteria we have discussed in Patat (2003). The numbers for the different passbands are shown in Table 3, where we have reported the total number $N_{\rm t}$ of examined frames, the number $N_{\rm s}$ of frames which passed the tests and the percentage of success $f_{\rm s}$. As expected, this is particularly poor for the U filter, where the sky background level is usually very low. In fact, we have shown that practically all frames with a sky background level lower than 400 electrons are rejected (Patat 2003, Sect. 5). Since to reach this level in the U passband one needs to expose for more than 800 s (see Table 1), this explains the large fraction of unacceptable frames. We also note that the number of input frames in the various filters reflects the effective user's requests. As one can see from Table 3, the percentage of filter usage $f_{\rm t}$ steadily grows going from blue to red filters, with R and I used in almost 70% of the cases, while the U filter is extremely rarely used.

 

 

Table 3: Number of sky brightness measurements obtained with FORS1 in U,B,V,R,I passbands from April 1, 2000 to September 30, 2001.

Filter	$f_{\rm t}$ (%)	$N_{\rm t}$	$N_{\rm s}$	$f_{\rm s}$ (%)
U	1.8	204	68	33.3
B	11.3	479	434	90.6
V	17.3	845	673	79.6
R	27.1	1128	1055	93.5
I	42.5	1783	1653	92.7
		4439	3883	87.5

To allow for a thorough analysis of the data, the sky brightness measurements have been logged together with a large set of parameters, some of which are related to the target's position and others to the ambient conditions. The first set has been computed using routines adapted from those coded by J. Thorstensen and it includes average airmass, azimuth, galactic longitude and latitude, ecliptic latitude and helio-ecliptic longitude, target-moon angular distance, moon elevation, fractional lunar illumination (FLI), target-Sun angular distance, Sun elevation and time distance between observation and closest twilight. Additionally we have implemented two routines to compute the expected moon brightness and zodiacal light contribution at target's position. For the first task we have adopted the model by Krisciunas & Schaefer (1991) and its generalisation to UBVRI passbands (Schaefer 1998), while for the zodiacal light we have applied a bi-linear interpolation to the data presented by Levasseur-Regourd & Dumont (1980). The original data is converted from S₁₀(V) to cgs units and the UBRI brightness is computed from the V values reported by Levasseur-Regourd & Dumont assuming a solar spectrum, which is a good approximation in the wavelength range 0.2-2 $\mu$m (Leinert et al. 1998). For this purpose we have adopted the Sun colours reported by Livingston (2000), which turn into the following U, B, R and I zodiacal light intensities normalised to the V passband: 0.52, 0.94, 0.77 and 0.50. For a derivation of the conversion factor from S₁₀(V) to cgs units, see Appendix B.

The ambient conditions were retrieved from the VLT Astronomical Site Monitor (ASM, Sandrock et al. 2000). For our purposes we have included air temperature, relative humidity, air pressure, wind speed and wind direction, averaging the ASM entries across the exposure time. Finally, to allow for further quality selections, for each sky brightness entry we have logged the number of sub-windows which passed the $\Delta$-test (see Patat 2003) and the final number $n_{\rm g}$ of selected sub-windows effectively used for the background estimate.

Throughout this paper the sky brightness is expressed in mag arcsec^-2, following common astronomical practice. However, when one is to correct for other effects (like zodiacal light or scattered moon light), it is more practical to use a linear unit. For this purpose, when required, we have adopted the cgs system, where the sky brightness is expressed in erg s^-1 cm^-2 Å^-1 sr^-1. In these units the typical sky brightness varies in the range 10 ^-9-10^-6 (see also Appendix B). It is natural to introduce a surface brightness unit (sbu) defined as 1 sbu $\equiv$ 10^-9 erg s^-1 cm^-2 Å^-1 sr^-1. In the rest of the paper we will use this unit to express the sky brightness in a linear scale.

Figure 3: Distribution of telescope pointings in Alt-Az coordinates. The lower left insert shows the airmass distribution.

Due to the large number of measurements, the data give a good coverage of many relevant parameters. This is fundamental, if one is to investigate possible dependencies. In the next sub-sections we describe the statistical properties of our data set with respect to these parameters, whereas their correlations are discussed in Sects. 5-8.

  
4.1 Coordinate distribution

The telescope pointings are well distributed in azimuth and elevation, as it is shown in Fig. 3. In particular, they span a good range in airmass, with a few cases reaching zenith distances larger than 60$^\circ$. Due to the Alt-Az mount of the VLT, the region close to zenith is not observable, while for safety reasons the telescope does not point at zenith distances larger than 70$^\circ$. Apart from these two avoidance areas, the Alt-Az space is well sampled, at least down to zenith distances Z= 50$^\circ$. At higher airmasses the western side of the sky appears to be better sampled, due to the fact that the targets are sometimes followed well after the meridian, while the observations tend to start when they are on average at higher elevation.

Since sky brightness is expected to depend on the observed position with respect to the Galaxy and the Ecliptic (see Leinert et al. 1998 for an extensive review), it is interesting to see how our measurements are distributed in these two coordinate systems. Due to the kind of scientific programmes which are usually carried out with FORS1, we expect that most of the observations are performed far from the galactic plane. This is confirmed by the left panel of Fig. 4, where we have plotted the galactic coordinates distribution of the 3883 pointings included in our data set. As one can see, the large majority of the points lie at |b|> 10$^\circ$, and therefore the region close to the galactic plane is not well enough sampled to allow for a good study of the sky brightness behaviour in that area.

Figure 4: Distribution of telescope pointings in galactic (left panel) and helio-ecliptic (right panel) coordinates. The two histograms show the distribution of galactic and ecliptic latitudes.

The scenario is different if we consider the helio-ecliptic coordinate system (Fig. 4, right panel). The observations are well distributed across the ecliptic plane for $\vert\lambda -\lambda _\odot \vert>$ 60$^\circ$ and $\beta<$ +30$^\circ$, where the contribution of the zodiacal light to the global sky brightness can be significant. As a matter of fact, the large majority of the observations have been carried out in the range -30 $^\circ \leq \beta\leq+$30$^\circ$, where the zodiacal light is rather important at all helio-ecliptic longitudes. This is clearly visible in the upper panel of Fig. 5, where we have over imposed the telescope pointings on a contour plot of the zodiacal light V brightness, obtained from the data published by Levasseur-Regourd & Dumont (1980).

Figure 5: Upper panel: distribution of telescope pointings in helio-ecliptic coordinates. Over imposed is a contour plot of the zodiacal light V brightness at the indicated levels expressed in sbu (1 sbu = 10-9 erg s-1 cm-2 Å-1 sr-1). Original data are from Levasseur-Regourd & Dumont (1980). Lower panel: Zodiacal light V brightness profiles at four different ecliptic latitudes expressed in sbu (left scale) and mag arcsec-2 (right scale). The brightness increase seen at $\vert\lambda -\lambda _\odot \vert>$ 150$^\circ$ is the so-called Gegenschein. The horizontal dotted line is placed at a typical V global sky brightness during dark time (21.6 mag arcsec-2).

We note that the wavelength dependency of the zodiacal light contribution is significant even within the optical range. In particular it reaches its maximum contribution in the B passband, where the ratio between zodiacal light and typical dark time sky flux is always larger than 30%. On the opposite side we have the I passband, where for $\vert\lambda -\lambda _\odot \vert\geq$ 80$^\circ$ the contribution is always smaller than 30% (see also O'Connell 1987). Now, using the data from Levasseur-Regourd & Dumont (1980) and the typical dark time sky brightness measured on Paranal, we can estimate the sky brightness variations one expects on the basis of the pure effect of variable zodiacal light contribution. As we have already mentioned, the largest variation is expected in the B band, where already at $\vert\lambda-\lambda_\odot\vert=$ 90$^\circ$ the sky becomes inherently brighter by 0.4-0.5 mag as one goes from $\vert\beta\vert>$ 60$^\circ$ to $\vert\beta\vert=$ 0$^\circ$. This variation decreases to $\sim$0.15 mag in the I passband.

Due to the fact that the bulk of our data has been obtained at  $\vert\beta\vert\leq$ 30$^\circ$, our dark time sky brightness estimates are expected to be affected by systematic zodiacal light effects, which have to be taken into account when comparing our results with those obtained at high ecliptic latitudes for other astronomical sites (see Sect. 5).

  
4.2 Moon contribution

Another relevant aspect that one has to take into account when measuring the night sky brightness is the contribution produced by scattered moon light. Due to the scientific projects FORS1 was designed for, the large majority of observations are carried out in dark time, either when the fractional lunar illumination (FLI) is small or when the moon is below the horizon. Nevertheless, according to the user's requirements, some observations are performed when moon's contribution to the sky background is not negligible. To evaluate the amount of moon light contamination at a given position on the sky (which depends on several parameters, like target and moon elevation, angular distance, FLI and extinction coefficient in the given passband) we have used the model developed by Krisciunas & Schaefer (1991), with the double aim of selecting those measurements which are not influenced by moonlight and to test the model itself. We have forced the lunar contribution to be zero when moon elevation is $h_{\rm m}\leq -$18$^\circ$ and we have neglected any twilight effects. On the one hand this has certainly the effect of overestimating the moon contribution when -18 $^\circ \leq h_{\rm m} \leq$ 0$^\circ$, but on the other hand it puts us on the safe side when selecting dark time data.

As expected, a large fraction of the observations were obtained practically with no moon: in more than 50% of the cases moon's addition is from 10^-1 to 10^-3 the typical dark sky brightness. Nevertheless, there is a substantial tail of observations where the contamination is relevant (200-400 sbu) and a few extreme cases were the moon is the dominating source (>600 sbu). This offers us the possibility of exploring both regimes.

  
4.3 Solar activity

We conclude the description of the statistical properties of our data set by considering the solar activity during the relevant time interval. As it has been first pointed out by Lord Rayleigh (1928), the airglow emission is correlated with the sunspot number. As we have seen in Sect. 1, this has been confirmed by a number of studies and is now a widely accepted effect (see Walker 1988b and references therein). As a matter of fact, all measurements presented in this work were taken very close to the maximum of sunspot cycle No. 23, and thus we do not expect to see any clear trend. This is shown in Fig. 6, where we have plotted the monthly averaged Penticton-Ottawa solar flux at 2800 MHz (Covington 1969). We notice that the solar flux abruptly changed by a factor $\sim$2 between July and September 2001, leading to a second maximum which lasted roughly two months at the end of year 2001. This might have some effect on our data, which we will discuss later on.

Figure 6: Penticton-Ottawa Solar flux at 2800 MHz (monthly average). The time range covered by the data presented in this paper is indicated by the thick line. The upper left insert traces the solar flux during the last six cycles.