1: % aa.dem
2: % AA vers. 6, LaTeX class for Astronomy & Astrophysics
3: % demonstration file
4: % (c) Springer-Verlag HD
5: % revised by EDP Sciences
6: %-----------------------------------------------------------------------
7: %
8: %\documentclass[referee]{aa} % for a referee version
9: %\documentclass[onecolumn]{aa} % for a paper on 1 column
10: %\documentclass[longauth]{aa} % for the long lists of affiliations
11: %\documentclass[rnote]{aa} % for the research notes
12: %
13: \documentclass{aa}
14: %
15: \usepackage{graphicx}
16: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
17: \usepackage{txfonts}
18: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
19: %
20:
21:
22: \def\deg{$^\circ$\hspace{-1.3mm}.\hspace{0.3mm}}
23: \def\as{\arcsec\hspace{-1.2mm}.\hspace{0.3mm}}
24: \def\am{\arcmin\hspace{-1.2mm}.\hspace{0.3mm}}
25: \def\e{e$^-$}
26:
27:
28: \begin{document}
29: %
30: \title{The Dancing Sky: 6 years of night sky observations at Cerro
31: Paranal \thanks{Based on observations made with ESO Telescopes at
32: Paranal Observatory.}}
33:
34: \subtitle{}
35:
36: \author{F. Patat\inst{1}}
37:
38: \offprints{F. Patat}
39:
40: \institute{European Southern Observatory (ESO), K. Schwarzschildstr. 2,
41: D-85748, Garching b. M\"unchen, Germany\\
42: \email{fpatat@eso.org}
43: }
44:
45: \date{Received ...; accepted ...}
46:
47: % \abstract{}{}{}{}{}
48: % 5 {} token are mandatory
49:
50: \abstract
51: % context heading (optional)
52: % {} leave it empty if necessary
53: {}
54: % aims heading (mandatory)
55: {The present work provides the results of the first six years of operation
56: of the systematic night-sky monitoring at ESO-Paranal (Chile).}
57: % methods heading (mandatory)
58: {The $UBVRI$ night-sky brightness was estimated on about 10,000
59: VLT-FORS1 archival images, obtained on more than 650 separate nights,
60: distributed over 6 years and covering the descent from maximum to minimum
61: of sunspot cycle n. 23. Additionally, a set of about 1,000 low resolution,
62: optical night-sky spectra have been extracted and analyzed.}
63: % results heading (mandatory)
64: {The unprecedented database discussed in this paper has led to the
65: detection of a clear seasonal variation of the broad band night sky
66: brightness in the $VRI$ passbands, similar to the well
67: known semi-annual oscillation of the Na~I~D doublet. The
68: spectroscopic data demonstrate that this seasonality is common to all
69: spectral features, with the remarkable exception of the OH
70: rotational-vibrational bands. A clear dependency on
71: the solar activity is detected in all passbands and it is particularly
72: pronounced in the $U$ band, where the sky brightness decreased by
73: $\sim$0.6 mag arcsec$^{-2}$ from maximum to minimum of solar cycle n.~23.
74: No correlation is found between solar activity and the intensity of the
75: Na~I~D doublet and the OH bands. A strong correlation between the
76: intensity of N~I 5200\AA\/ and [OI]6300,6364\AA\/ is reported here for the
77: first time. The paper addresses also the determination of the correlation
78: timescales with solar activity and the possible connection with the flux
79: of charged particles emitted by the Sun.}
80: % conclusions heading (optional), leave it empty if necessary
81: {}
82:
83: \keywords{atmospheric effects -- site testing -- techniques: photometric --
84: techniques: spectroscopic}
85:
86: \maketitle
87: %
88: %________________________________________________________________
89:
90: \section{\label{sec:intro}Introduction}
91:
92: Soon after the beginning of VLT science operations in Paranal, ESO
93: started an automatic $UBVRI$ sky brightness survey, with the aim of
94: both characterizing the site and studying the long term trend, in
95: order to detect any possible effects of human activity. The results
96: obtained during the first 18 months of operations (April 2000 -
97: September 2001) have been presented and discussed in Patat
98: (\cite{paperI}, hereafter Paper~I). This programme, which makes use of
99: all scientific images obtained with FORS1, is building one of the most
100: extensive, accurate and homogeneous optical sky brightness data sets
101: ever studied. As shown in Paper~I, these data allow a very detailed
102: analysis, including the study of correlations with other parameters,
103: and the investigation of short, medium and long term variations. The
104: interested reader can find exhaustive reviews on this subject in Roach
105: \& Gordon (\cite{roach}) and in Leinert et al. (\cite{leinert}), while
106: references to other published sky brightness surveys are given in
107: Paper I.
108:
109: Since the publication of the first results, obtained on 174 different
110: nights close to maximum of solar cycle n.~23, the data base has been
111: steadily growing and progressively extending towards the solar
112: minimum. In this paper I present a global analysis run on the whole
113: data set, which includes broadband observations taken on 668 separate
114: nights between 20 Apr. 2001 and 20 Jan. 2007. Additionally, I present
115: and discuss here a set of more than 1000 low-resolution long-slit
116: night sky spectra taken with FORS1 between May 1999 and
117: Feb. 2005. They were used to measure the fluxes of single lines or
118: integrated OH bands.
119:
120: The paper is organized as follows. In Sec.~\ref{sec:obs} I describe
121: the observations and the basic data reduction steps for photometry and
122: long slit spectroscopy. In Sec.~\ref{sec:dark} I review the general
123: results obtained during dark time, while the correlation with solar
124: activity and the seasonal variations are discussed in
125: Sects.~\ref{sec:sun} and \ref{sec:season}, respectively. The
126: spectroscopic analysis is presented in Sects.~\ref{sec:spectra},
127: \ref{sec:corr} and \ref{sec:short}. Finally, in Sec.~\ref{sec:disc} I
128: discuss the main results and summarize the conclusions in
129: Sec.~\ref{sec:concl}.
130:
131: \section{\label{sec:obs}Observations and Data Reduction}
132:
133: The data used in this work were obtained with the FOcal Reducer/low
134: dispersion Spectrograph (hereafter FORS1), mounted at the Cassegrain
135: focus of ESO--Antu/Melipal/Kueyen 8.2m telescopes (Szeifert 2002). The
136: instrument is equipped with a 2048$\times$2048 pixel (px) TK2048EB4-1
137: backside thinned CCD and has two remotely exchangeable collimators,
138: which give a projected scale of 0\as2 and 0\as1 per pixel (24$\mu$m
139: $\times$ 24$\mu$m). According to the collimator used, the sky area
140: covered by the detector is 6\am8$\times$6\am8 and 3\am4$\times$3\am4,
141: respectively.
142:
143: \subsection{\label{sec:phot}Photometry}
144:
145: The photometric data set includes 10,432 images obtained in the
146: $UBVRI$ passbands with both collimators. The reduction procedure is
147: described in Paper I, to which I refer the reader for a more detailed
148: description, while here I only recap the basic steps. All frames are
149: automatically processed by the FORS pipeline, which applies bias and
150: flat-field correction, the latter performed using twilight sky
151: flats. Once the instrument signatures are removed from the images, the
152: sky background is estimated using the robust algorithm described in
153: Patat (\cite{patatII}). The photometric calibration into the
154: Johnson-Cousins system is then achieved using zeropoints and colour
155: terms derived from the observation of standard star fields
156: (Landolt \cite{landolt}), regularly obtained as part of the FORS calibration
157: plan. Finally, the observed values are corrected to zenith using the
158: standard procedure (see for example Garstang \cite{garstang}) and
159: logged together with a number of relevant parameters.
160:
161:
162:
163: \subsection{\label{sec:spec}Spectroscopy}
164:
165: The spectroscopic data set includes a sub-sample of all long-slit
166: science data present in the ESO archive whose proprietary period had
167: expired by the time this paper has been written. For the sake of
168: simplicity, I have selected only the data obtained with the standard
169: resolution collimator and the single-port high-gain read-out mode,
170: since this combination is the most used for long slit spectroscopy
171: with FORS1. In order to accumulate a wide data sample, I have
172: retrieved from the VLT archive all public spectra taken with the 300V
173: grism coupled with the order-sorting filter GG435 and a long slit
174: 1\as0 wide, which is the most frequently used (676 frames). The
175: wavelength range was extended down to about 3600\AA\/ in the blue by
176: retrieving also all spectra taken with the same setup but with no
177: order-sorting filter, for a total of 163 frames. To increase the
178: sample, I have retrieved also all spectra obtained with the grisms
179: 600B (143 frames) and 600R (207 frames) coupled with the order-sorting
180: filters OG590 and GG435, respectively. Also in these two cases the
181: slit was 1\as0 wide. The main characteristics of each setup are
182: summarized in Tab.~\ref{tab:spec}. Exposure times range from a few
183: minutes to one hour.
184:
185: \tabcolsep 1.5mm
186: \begin{table}
187: \caption{\label{tab:spec} Main properties of the FORS1 spectral data set.
188: Only spectra taken with sun elevation below $-$18$^\circ$ were
189: included.}
190: \begin{tabular}{cccccc}
191: \hline \hline
192: Grism & Filter & Wav. Range & Resolution & Dispersion & N\\
193: & & (\AA) & (\AA\/ FWHM) & (\AA\/ px$^{-1}$) & \\
194: \hline
195: 300V & GG435 & 4300--8900 & 12 & 2.6 & 676\\
196: 300V & -- & 3615--8900(*) & 12 & 2.6 & 163\\
197: 600B & OG590 & 3650--6050 & 5.3 & 1.2 & 143 \\
198: 600R & GG435 & 5390--7530 & 4.5 & 1.0 & 207\\
199: \hline
200: \multicolumn{6}{l}{(*) Second order overlapping above 6000 \AA.}
201: \end{tabular}
202: \end{table}
203:
204:
205: All images were processed within the {\tt longslit} package of
206: IRAF\footnote{IRAF is distributed by the National Optical Astronomy
207: Observatories, which are operated by the Association of Universities
208: for Research in Astronomy, under contract with the National Science
209: Foundation.}. Due to the large amount of data and the purpose of this
210: work, the bias subtraction was performed using only a pre-scan
211: correction, while flat-fielding effects were neglected.
212:
213: Wavelength calibration was achieved using a 2D solution derived from a
214: set of reference arc exposures. Given the procedure adopted for the
215: spectrum extraction, this step is mandatory, since in FORS1 the line
216: curvature can reach a peak-to-peak value of about 10 px. If not
217: accounted for, this instrumental feature would produce an apparently
218: significant but artificial line broadening when collapsing the 2D
219: spectra perpendicularly to the dispersion direction (see next
220: section).
221:
222:
223: \begin{figure}
224: \centering
225: \includegraphics[width=9cm]{skyspectrum.jpg}
226: \caption{\label{fig:skyspectrum}Example of FORS1 flux calibrated night sky
227: spectrum obtained with the procedure outlined in the text, for the
228: 600R grism. Main lines and OH bands identifications are given. The
229: insert shows an example spectrum obtained with the 300V grism.}
230: \end{figure}
231:
232:
233: \subsubsection{\label{sec:extraction}Nightglow spectrum extraction and
234: calibration}
235:
236: After applying the appropriate 2D wavelength solution to all frames,
237: the night sky spectrum is extracted. For this purpose I have used a
238: robust algorithm to estimate the mode intensity in each column
239: perpendicular to the dispersion direction. This implicitly assumes
240: that most of the pixels are not ``contaminated'' by the contribution
241: of astrophysical objects, which is reasonable in the majority of the
242: cases, as verified by direct inspection of the whole two-dimensional data
243: sample. This is both a consequence of the relatively large
244: slit length featured by FORS1 (6\am8 on 2048 px) and the typical
245: targets observed with this instrument, which are very often faint and
246: star-like sources. After visual inspection, only a few spectra were
247: removed from the final data set.
248:
249: To allow for completely unsupervised line and continuum flux
250: measurements, the accuracy of wavelength calibration is a mandatory
251: requirement. Possible causes of rigid shifts in the dispersion
252: solution can be identified as instrument interventions, turning into
253: movements of the long slit on the focal plane, and flexures at large
254: zenith distances. To correct for these problems, I have produced a
255: reference night sky spectrum for each of the two resolutions I have
256: used, with a typical accuracy (estimated on isolated lines) better
257: than 1 \AA. Then, by means of cross-correlation, the zero point of the
258: wavelength scale of each spectrum is automatically corrected at the
259: end of the extraction procedure. This ensures that, at this stage, all
260: spectra have maximum wavelength errors that do not exceed 1\AA.
261:
262: For the absolute flux calibration I have used a set of
263: spectrophotometric standard stars to derive a reference sensitivity
264: function $s(\lambda)$, which I have applied to all spectra. Even
265: though this does not take into account the changes in sensitivity
266: which are mainly due to the aging of reflective surfaces (Patat
267: \cite{paperI}), at the wavelengths of interest they are of the order of
268: a few percent, and therefore can be safely neglected in this context.
269:
270: The flux calibration of the extracted spectrum $f(\lambda)$ to
271: physical units is finally computed as:
272:
273: \begin{displaymath}
274: F(\lambda) = \frac{f(\lambda)}{s(\lambda) \; t \; w \; p}\;
275: \;\mbox{erg} \; \mbox{s}^{-1} \;\mbox{cm}^{-2}
276: \;\mbox{\AA}^{-1} \;\mbox{arcsec}^{-2}
277: \end{displaymath}
278:
279: where $t$ is the exposure time (in seconds), $w$ is the slit width (in
280: arcsec) and $p$ is the projected pixel scale (arcsec px$^{-1}$).
281:
282: An example of flux calibrated spectra obtained with this procedure is
283: presented in Fig.~\ref{fig:skyspectrum}. The resulting signal-to-noise
284: ratio on the pseudo-continuum changes according to the exposure time
285: of the original frame, but it is always larger than 100.
286:
287:
288: \subsubsection{\label{sec:meas}Line/bands flux measurements}
289:
290: The line flux integration is carried out in a fully automated way
291: within a given wavelength window after subtracting the estimated
292: pseudo-continuum intensity. The integration boundaries have been set
293: according to the spectral resolution. For single lines, like the
294: [OI]5577, the optimal window semi-amplitude has been set to
295: 1.7$\times$FWHM which corresponds to a $\pm$4$\sigma$ from the line
296: center. The pseudo-continuum level is estimated in the adjacent
297: emission line-free regions. For more complex features (Na~ID doublet,
298: OH bands), the integration boundaries and the continuum region/s have
299: been set {\em ad hoc}, and are reported in Table~\ref{tab:flux}.
300: %An example of line integration is shown in Fig.~\ref{fig:fluxmeas}
301: %for the case of the Na~I D lines.
302:
303: Following Barbier (\cite{barbier}), I have also introduced four bands,
304: indicated as B1, B2, B3 and B4, in the blue spectral domain (see
305: Table~\ref{tab:flux}). The gap between bands B1 and B2 has been set in
306: order to avoid the strong Ca~II H\&K absorptions
307: (Fig.~\ref{fig:bluband}). Finally, a broad band ranging from
308: 5500\AA\/ to 7530\AA\/ has been introduced, with the main aim of
309: giving an overall flux estimate.
310:
311: \begin{figure}
312: \centering
313: \includegraphics[width=9cm]{bluband.jpg}
314: \caption{\label{fig:bluband}Definition of the blue bands B1, B2, B3 and B4.}
315: \end{figure}
316:
317: \begin{table}
318: \caption{\label{tab:flux}Integration boundaries and continuum regions used
319: for line flux measurements.}
320: \tabcolsep 1.7mm
321: \centerline{
322: \begin{tabular}{ccccc}
323: \hline \hline
324: Line/Band & Line Range & Cont. Range & & Setup \\
325: \hline
326: NI 5200 & $\pm 4\sigma$ & 5040-5120 &(C1) & 300V,600B\\
327: OI] 5577 & $\pm 4\sigma$ & 5480-5520 &(C2) & 300V,600B/R\\
328: Na~I~D & D$_2-4\sigma$-D$_1+4\sigma$ & 5800-5850 &(C3) & 300V,600B/R\\
329: OI] 6300 & $\pm 4\sigma$ & 6400-6450 &(C4) & 300V,600R\\
330: OI] 6364 & $\pm 4\sigma$ & 6400-6450 & & 300V,600R\\
331: \hline
332: OH(6-1) & 6435-6680 & 6750-6800 &(C5) & 300V,600R\\
333: OH(7-2) & 6810-7060 & 6750-6800 & & 300V,600R\\
334: OH(8-3) & 7200-7450 & 6750-6800 & & 300V,600R\\
335: OH(6-2) & 8250-8570 & 8160-8230 &(C6) & 300V+GG\\
336: O$_2$(0-1) & 8605-8695 & 8160-8230 & & 300V+GG\\
337: \hline
338: B1 & 3630-3890 & - & & 300V,600B\\
339: B2 & 3990-4330 & - & & 300V,600B\\
340: B3 & 4350-4850 & - & & 300V,600B\\
341: B4 & 4870-5170 & - & & 300V,600B\\
342: \hline
343: Broad Band & 5500-7530 & - & & 300V,600R \\
344: \hline
345: \end{tabular}
346: }
347: \end{table}
348:
349: As in the case of the broad-band measurements, the line fluxes need to be
350: corrected for airmass. Between the two cases, however, there is a
351: difference: in fact, while the integrated flux within a broad-band
352: filter is the result of extra-terrestrial sources (zodiacal light,
353: unresolved stars and galaxies) and emission within the atmosphere, in
354: the case of nightglow emission lines all the radiation is of
355: terrestrial origin. Practically this coincides with setting $f=1$ in
356: Eq.~C.3 of Paper~I, which I have used to correct the observed values.
357:
358: In general, the signal in the measured features is so high that the
359: uncertainty in the line fluxes is by far dominated by the
360: contamination by unresolved OH lines and, to a smaller extent, to the
361: uncertainty on the pseudo-continuum level.
362:
363:
364:
365: \section{\label{sec:dark}Dark time night sky brightness}
366:
367: Since the data set includes observations obtained under a wide variety
368: of conditions, in order to estimate the zenith sky brightness during
369: dark time it is necessary to apply some filtering. To this aim I have
370: adopted the same criteria described in Paper I: photometric
371: conditions, airmass $X\leq$1.4, galactic latitude $|b|>$10$^\circ$,
372: helio-eclitpic longitude $|\lambda-\lambda_\odot|\geq$90$^\circ$, time
373: distance from the closest twilight $\Delta t_{twi}>$ 1 hour and no
374: moon (fractional lunar illumination FLI=0 or moon elevation
375: $h_M\leq-$18$^\circ$). The results of this selection, which reduced
376: the number of suitable data points to 3736, are summarized in
377: Table~\ref{tab:dark}. As one can see, the average values are all
378: within 0.1 mag from those reported in Paper I (see Table~4). In all
379: filters there is a systematic shift towards darker values, with the
380: only exception of the $I$ band. Since the values reported in Paper I
381: were obtained during the sunspot maximum and, given the correlation
382: between solar activity and night sky brightness shown by Walker
383: (\cite{walker88}), Pilachowski et al. (\cite{pila}); Krisciunas
384: (\cite{krisc90}), Leinert et al. (\cite{leinert95}) and Mattila et
385: al. (\cite{attila}, Krisciunas (\cite{krisc97}), Krisciunas et
386: al. (\cite{krisc07}), this behavior was indeed expected. Given the time
387: distribution of data points (see Fig.~\ref{fig:hist}), the average
388: values reported in Table~\ref{tab:dark} are biased towards the sunspot
389: maximum phase. For solar activity corrected data, see next section.
390:
391: \begin{table}
392: \centering
393: \caption{\label{tab:dark} Zenith corrected average sky brightness
394: during dark time at Paranal. Values are expressed in mag
395: arcsec$^{-2}$. Columns 3 to 8 show the RMS deviation, minimum and
396: maximum brightness, number of dark-time data points, expected average
397: contribution from the zodiacal light, and total number of data points,
398: respectively.}
399: \begin{tabular}{cccccccccc}
400: \hline \hline
401: Filter & Sky Br. & $\sigma$ & Min & Max & $N_{dark}$ & $\Delta m_{ZL}$ & $N_{tot}$\\
402: \hline
403: U & 22.35 & 0.19 & 21.89 & 22.78 & 129 &0.20 & 264\\
404: B & 22.67 & 0.16 & 22.19 & 23.02 & 493 &0.28 & 1400\\
405: V & 21.71 & 0.24 & 21.02 & 22.30 & 692 &0.20 & 1836\\
406: R & 20.93 & 0.24 & 20.42 & 21.56 & 1285 &0.16 & 3931\\
407: I & 19.65 & 0.28 & 18.85 & 20.56 & 1137 &0.07 & 3001\\
408: \hline
409: Total & & & & & 3736 & & 10432\\
410: \hline
411: \end{tabular}
412: \end{table}
413:
414: \begin{figure}
415: \centering
416: \includegraphics[width=9cm]{hist.jpg}
417: \caption{\label{fig:hist} Image distribution along the time interval
418: covered by the present work for the global data set (thin line) and
419: for the dark-time set (shaded thick line). The vertical dotted line
420: marks the extent of the sample presented in Paper I, while the upper
421: scale marks the 1st of January of each year.}
422: \end{figure}
423:
424: Single measurements for the 5 passbands are presented in
425: Fig.~\ref{fig:dark}, which shows also surface brightness distributions
426: for the dark time (solid line) and global (dashed line) samples.
427:
428:
429: \begin{figure}
430: \centering
431: \includegraphics[width=9cm]{dark2.jpg}
432: \caption{\label{fig:dark} Zenith corrected sky brightness measured at
433: Paranal during dark time (thick dots) from April 1st, 2000 to April 8,
434: 2006. The selection criteria are: $|b|>$10$^\circ$,
435: $|\lambda-\lambda_\odot|\geq$90$^\circ$, $\Delta t_{twi}>$1 hour,
436: FLI=0 or $h_m\leq-$18$^\circ$. Thin dots indicate all observations
437: (corrected to zenith). The horizontal dotted lines are positioned at
438: the average values of the selected points, while the vertical dashed
439: line marks the end of the time interval discussed in Paper I. The
440: histograms trace the distribution of selected measurements (solid
441: line) and all measurements (dotted line), while the vertical dashed
442: lines are placed at the average sky brightness during dark time.}
443: \end{figure}
444:
445:
446: \begin{figure}
447: \centering
448: \includegraphics[width=9cm]{sun3.jpg}
449: \caption{\label{fig:sun2}Penticton-Ottawa Solar flux at 2800 MHz
450: (monthly average). The time range covered by the data presented in
451: this paper is indicated by the thick line, while the horizontal
452: segment indicates the time covered by the data presented in Paper
453: I. The upper insert traces the solar flux during the last six cycles.}
454: \end{figure}
455:
456: \begin{figure}
457: \centering
458: \includegraphics[width=9cm]{sunaveV.jpg}
459: \caption{\label{fig:sunave}Lower panel: nightly average dark time sky
460: brightness in the $V$ passband vs. solar flux density. The solid line
461: is a linear least squares fit to the data (empty symbols mark the data
462: points rejected by a kappa-sigma clipping). Upper panel:
463: Penticton-Ottawa solar flux at 2800 MHz during the time interval
464: discussed in this paper. The open circles indicate the monthly
465: averaged values corresponding to the $V$ nightly averages plotted in
466: the lower panel.}
467: \end{figure}
468:
469: \section{\label{sec:sun}Sky brightness vs. solar activity}
470:
471: As first pointed out by Rayleigh (\cite{rayleigh}) and confirmed later
472: on by several other authors (see for instance Rosenberg \& Zimmerman
473: \cite{rosenberg}; Walker \cite{walker88}; Krisciunas \cite{krisc90};
474: Leinert et al. \cite{leinert95}; Mattila et al. \cite{attila};
475: Krisciunas \cite{krisc97}; Krisciunas et al. \cite{krisc07}) many of
476: the emission features in the night sky spectrum show a clear
477: dependency on the sunspot cycle. In particular, $B$ and $V$ present a
478: peak-to-peak variation of $\sim$0.5 mag arcsec$^{-2}$ during a full
479: solar cycle. Less clear is the behavior at longer wavelengths, which
480: are dominated by the OH emissions, whose intensity is uncorrelated
481: with solar activity (see Sec.~\ref{sec:spectra}).
482:
483: The data presented here cover the descent from the maximum of sunspot
484: cycle n.~23 to the minimum phase, as shown in Fig.~\ref{fig:sun2},
485: that displays the monthly averaged Penticton-Ottawa solar flux at
486: 2800 MHz (Covington 1969)\footnote{The data are available in digital
487: form at the following web site: {\tt
488: http://www.drao.nrc.ca/icarus/www/archive.html}}. During this interval
489: the Solar Flux Density (hereafter $SFD$) spans from 0.8 to 2.4 MJy, a
490: range which is very close to that of a full cycle (the solar minimum
491: is expected for the end of 2007). Following what has been done by
492: other authors (see for instance Leinert et al. \cite{leinert95}), I
493: have studied the correlation between the sky brightness nightly
494: averages and the SFD monthly averages, computed during the 30
495: preceding days. The sky brightness measurements have been corrected
496: for the zodiacal light contribution computed for each data point as in
497: Paper~I (Sec.~4) and using the data by Levasseur-Regourd \& Dumont
498: (\cite{levasseur}).
499:
500: All passbands show very good linear correlations, an example of which
501: is presented in Fig.~\ref{fig:sunave} for the $V$ filter. In order to
502: give a quantitative representation of the effect, I have fitted a
503: relation of the type $m=m_0 + \gamma \;SFD$ to the data. The results
504: are shown in Table~\ref{tab:sunave} for all filters. Besides reporting
505: the zeropoint ($m_0$, mag arcsec$^{-2}$), the slope ($\gamma$, mag
506: arcsec$^{-2}$ MJy$^{-1}$) and their associated statistical errors, the
507: Table includes also the estimated full solar cycle variation ($\Delta
508: m=|(2.4-0.8)\gamma|$), the value attained at solar minimum, evaluated
509: for $SFD$=0.8 MJy ($m_{min}$), the value corresponding to the average
510: $SFD$ level $<SFD>$=1.6 MJy ($m_{ave}$), the RMS deviation from the
511: best fit relation ($\sigma$), the linear correlation factor ($r$) and
512: the number of nights used ($N$) for each filter.
513:
514:
515: As one can see, the values of $\Delta m$ are smaller than those
516: reported by other authors: with the only exception of $U$, which
517: reaches about 0.6 mag arcsec$^{-2}$, all the others show values that
518: are smaller than 0.4 mag arcsec$^{-2}$. Walker (\cite{walker88})
519: quoted maximum ranges of $\Delta V\simeq$1.0 and $\Delta B\simeq$0.8
520: mag arcsec$^{-2}$ for solar cycle n.~21, while Krisciunas
521: (\cite{krisc97}) reports $\Delta V$=0.6 for solar cycle n.~22, and
522: similar values are reported by Leinert et al. (\cite{leinert95}) and
523: Mattila et al. (\cite{attila}). On the other hand, Liu et
524: al. (\cite{liu}) quote an increase of the $V$ sky brightness of
525: $\sim$0.2 mag arcsec$^{-2}$ from 1995 to 2001. This value is
526: consistent with the measures discussed here, especially taking into
527: account that cycle n.~23 had a second maximum, which occurred after
528: the observations presented by Liu et al. (\cite{liu}). These facts
529: seem to suggest that not all solar cycles have identical effects on
530: the night glow. As a matter of fact, Walker (\cite{walker88}), while
531: revising the result of previous works, had suggested that the relation
532: between intensity of the [OI]5577
533: \AA\/ line and the solar activity might vary from cycle to cycle,
534: within a given cycle and possibly with geographical location.
535: Unfortunately, the number of sunspot cycles covered by the
536: observations is still too small to allow a firm conclusion, but the
537: very recent results discussed by Krisciunas et al. (\cite{krisc07}),
538: covering two full solar cycles, seem indeed to confirm this suspicion.
539:
540:
541:
542: \begin{figure}
543: \centering
544: \includegraphics[width=9cm]{sunanalV.jpg}
545: \caption{\label{fig:sunanal}Upper panel: linear correlation coefficient
546: (absolute value) as a function of time delay $\tau$ for the $V$
547: passband data, computed for three different values of $\Delta \tau$
548: (1, 10 and 30 days). Lower panel: example of SFD periodicity due to
549: solar rotation. The dotted curve is a replica of the original data
550: shifted by 27.3 days.}
551: \end{figure}
552:
553:
554: \subsection{\label{sec:timescale} Solar activity correlation time scales}
555:
556: Given the unprecedentedly large amount of data, one can investigate
557: the solar dependency in a bit more detail, for example trying to
558: deduce the typical timescales of night sky brightness fluctuations due
559: to short-term changes in the solar flux. Ideally, for doing this, one
560: would look for strong enhancements in the solar flux and try to seek
561: for a corresponding increases in the night sky
562: brightness. Nevertheless, due to the sparse time sampling, the only
563: viable approach is actually the opposite one, i.e. starting from the
564: available night sky measurements, one goes back to the solar data and
565: studies the correlation with the sun flux emitted as a function of
566: time lag. For doing this I have computed the Pearson's linear
567: correlation coefficient $r$ (Press et al. \cite{press}) between the
568: nightly average sky brightness $m(t)$ measured at any given time $t$
569: and the solar flux density $SFD(t-\tau)$, averaged within a time
570: window $\Delta \tau$, as a function of both $\tau$ and $\Delta
571: \tau$. While the first parameter gives an indication about the time
572: lag between a change in the solar flux and the consequent variation in
573: the night sky brightness, the second is related to the typical
574: timescales of the physical processes which govern the energy release
575: in the upper atmosphere.
576:
577: An example of this kind of analysis is shown in the upper panel of
578: Fig.~\ref{fig:sunanal}, which illustrates the behavior of the
579: correlation coefficient $r$ for the $V$ data, obtained on 170 separate
580: nights for three different values of $\Delta \tau$ (1, 10 and 30
581: days). The correlation peak is quite broad and it is centered at
582: $\tau\sim$30 days; moreover, the maximum correlation tends to increase
583: slightly for larger averaging windows. This plot also explains why the
584: average solar flux computed in the 30 days before the night sky
585: observations (empty circles) gives a better correlation than the sun
586: flux measured on the preceding day (filled circles), as already
587: pointed out by several authors (see for example Leinert et
588: al. \cite{leinert95}; Mattila et al. \cite{attila}). Finally,
589: Fig.~\ref{fig:sunanal} clearly shows that the correlation drops
590: significantly for $\tau>$40 days, a fact that is common to all
591: $UBVRI$ passbands.
592:
593: An interesting feature to be noticed in Fig.~\ref{fig:sunanal} (seen
594: also in the equivalent plots for the other filters), is the presence
595: of spurious correlation peaks at a constant separation of about 27
596: days from the main peak. This is due the a periodicity present in the
597: solar flux data (see Fig.~\ref{fig:sunanal}, lower panel), which is
598: related to the solar rotation, whose synodic period is $\sim$27.3 days
599: (Howard
600: \cite{howard}). This semi-regular recurrence in the solar data explains,
601: for example, the presence of the two bumps close to $\tau$=5 and
602: $\tau$=65 in Fig.~\ref{fig:sunanal}.
603:
604: The strongest correlation is shown by the $U$ passband data, which
605: presents a rather marked peak $r\sim$0.6 at $\tau\sim$15 days, while
606: $r\sim$0.15 for $\tau$=1 day. For this reason, averaging over the last
607: 30 days gives a strong increase in the correlation, much stronger than
608: in any other passband. A behaviour similar to that displayed in $V$ is
609: seen also in $B$ (correlation peak $r\sim$0.45 for $\tau\sim$25 days)
610: and $I$ ($r\sim$0.30 for $\tau\sim$20 days). Somewhat different is the
611: case of $R$ passband, for which the correlation peak ($r\sim$0.45) is
612: attained at $\tau\sim$2 days, suggesting that the sun-dependent
613: features that contribute to the flux in this filter react rather
614: rapidly to the solar flux fluctuations. In general, however, and with
615: the possible exception of the $U$ passband, the correlation peaks are
616: rather broad, indicating that different processes take place with
617: different timescales. The behavior of the $U$ band, where the
618: nightglow emission is dominated by the Herzberg and Chamberlain O$_2$
619: bands (Broadfoot \& Kendall
620: \cite{broadfoot}), indicates that the photo-chemical reactions that
621: are responsible for the emission in this region are more sensitive to
622: solar activity.
623:
624:
625: \begin{table}
626: \tabcolsep 0.7mm
627: \caption{\label{tab:sunave} Linear least squares fit parameters for the sky
628: brightness vs.solar activity relation. Values in parenthesis indicate
629: RMS uncertainties. Input data have been corrected for the differential
630: zodiacal light contribution (see text).}
631: \centering
632: \begin{tabular}{ccccccccc}
633: \hline \hline
634: Filter & $m_{min}$ &$m_{ave}$& $\Delta m$ & $m_0$ & $\gamma$ & $\sigma$ & $r$ & N \\
635: \hline
636: %U & 22.63 & 22.40 & 0.50 & 22.87 (0.10)& $-$0.29 (0.05)& 0.12 & 32 \\
637: %B & 22.81 & 22.68 & 0.29 & 22.95 (0.04)& $-$0.17 (0.02)& 0.12 & 123 \\
638: %V & 21.82 & 21.66 & 0.34 & 21.98 (0.05)& $-$0.20 (0.03)& 0.17 & 154 \\
639: %R & 21.14 & 20.96 & 0.40 & 21.33 (0.04)& $-$0.24 (0.03)& 0.17 & 202 \\
640: %I & 19.76 & 19.65 & 0.24 & 19.88 (0.06)& $-$0.14 (0.04)& 0.19 & 145 \\
641: U & 22.86 & 22.58 & 0.61 & 23.15 (0.12)& $-$0.36 (0.07)& 0.15 & 0.47 & 32 \\
642: B & 23.11 & 22.98 & 0.29 & 23.25 (0.04)& $-$0.17 (0.02)& 0.12 & 0.40 & 127 \\
643: V & 21.99 & 21.86 & 0.30 & 22.13 (0.05)& $-$0.17 (0.03)& 0.14 & 0.42 & 148 \\
644: R & 21.26 & 21.09 & 0.37 & 21.33 (0.04)& $-$0.22 (0.02)& 0.15 & 0.44 & 202 \\
645: I & 19.81 & 19.72 & 0.20 & 19.90 (0.06)& $-$0.11 (0.04)& 0.18 & 0.28 & 144 \\
646: \hline
647: \end{tabular}
648: \end{table}
649:
650: \section{\label{sec:season}Seasonal Variations}
651:
652: In the previous work I had attempted to detect night sky brightness
653: seasonal variations but, due to insufficient number of data points, I
654: could not draw any firm conclusion (see Fig.~14 in Paper I). Thanks to
655: the much larger sample now available, this analysis becomes feasible
656: and, as a matter of fact, traces of a periodic modulation in the
657: average sky brightness are visible already in
658: Fig.~\ref{fig:dark}. They become much clearer when each data point is
659: plotted against the number of days from the beginning of the
660: corresponding year. The result is shown in Fig.~\ref{fig:season}
661: where, besides reporting the single dark time measurements, I have
662: also plotted the monthly averages. The input data have been corrected
663: for differential zodiacal light contribution and the solar flux
664: dependency derived in the previous section has been removed using the
665: parameters presented in Table~\ref{tab:sunave}.
666:
667: This semi-annual oscillation (hereafter SAO) is definitely present in
668: $V$, $R$ and $I$, while its presence in $B$ is more questionable ($U$
669: data were not included since the sample in this passband is too poor
670: for this purpose). The modulation amplitude grows at longer
671: wavelengths, shows two maxima around April-May and October, and two
672: minima around July-August and December-January. In general, the
673: variation is more pronounced in Winter-Spring than in Summer-Fall. For
674: example, in $I$ it reaches a peak-to-peak value of about 0.5 mag
675: arcsec$^{-2}$.
676:
677: As pointed out by Benn \& Ellison (\cite{benn}), the variable
678: contribution of zodiacal light can mimic a seasonal variation. In
679: order to exclude a possible contribution by this source to the
680: observed behavior, I have analyzed the expected enhancement of
681: brightness due to the zodiacal light for each data point, using the
682: data presented by Levasseur-Regourd \& Dumont (\cite{levasseur}) and
683: the procedure discussed in Paper I (Sec.~4) for the Paranal
684: site. This does not show any significant regular pattern as a function
685: of day of the year. The conclusion is that the observed SAO is not due
686: to the periodic apparent variation of the ecliptic height above
687: Paranal's horizon.
688:
689: \begin{figure}
690: \centering
691: \includegraphics[width=9cm]{season.jpg}
692: \caption{\label{fig:season}Seasonal variation of dark time sky
693: brightness with respect to the average value. Small symbols are the
694: single measurements while the large symbols mark the monthly averages,
695: computed in bins of 30 days each. The error bars indicate the RMS
696: deviations from the average within each bin, while the vertical dotted
697: lines mark equinoxes and solstices. Data have been corrected for
698: differential zodiacal light contribution and solar flux dependency.}
699: \end{figure}
700:
701: An interesting thing to be noticed, is that the minima and maxima of
702: the SAO occur out of phase with respect to the Equinoxes and Solstices
703: (see Fig.~\ref{fig:season}, vertical dotted lines).
704:
705: While seasonal variations of emission lines and/or bands have been
706: studied by several authors in the past (see for example Chamberlain
707: \cite{chamberlain}, Roach \& Gordon \cite{roach} and references
708: therein), broad band measurements are much more scanty and the results
709: not always in agreement. For instance, Schneeberger, Worden \& Beckers
710: (\cite{schneeberger}) report particularly bright values obtained in
711: June at the Sacramento Peak Observatory and they find them to be {\em
712: marginally correlated with the strong seasonal trend evident in the
713: record of daytime sky brightness observations}. In their survey run at
714: the Lowell Observatory, Lockwood, Floyd \& Thompson (\cite{lockwood})
715: discuss the seasonal variation, concluding that {\em neither winter
716: enhancements [...] nor springtime rise [...] is indicated [...]}. Benn
717: \& Ellison (\cite{benn}) reach the same result from the analysis of
718: the data obtained on La Palma, concluding that {\em dark-of-moon sky
719: brightness does not vary significantly ($\leq$0.1 mag) with season
720: [...]}. Finally, Liu et al. (\cite{liu}), analyzing data taken at the
721: Xinglong Station between 1995 and 2001, find that {\em the sky is
722: darker in the fall and winter than in the spring and summer
723: [...]}. While part of the discrepancies can be due to latitude effects
724: (see for example Chamberlain \cite{chamberlain}), some of the negative
725: detections are probably to be ascribed to non sufficient time sampling
726: and coverage. In fact, the SAO amplitude is at most comparable with
727: the night-to-night fluctuations and hence large and well sampled data
728: sets are required.
729:
730: It is worth mentioning here that Garstang (\cite{garstang88}) has
731: produced some simplified models to predict seasonal variations in the
732: broad-band night sky brightness, based on periodic variations of
733: height and molecular density. The values predicted by these models are
734: far too small with respect to those presented here, suggesting that
735: other possible explanations must be investigated.
736:
737: The observed behavior might indicate that whatever the reason for the
738: periodic variation is, it is not directly related to the amount of sun
739: radiation received by a given patch of the atmosphere during the day.
740: In fact, one might think that since during the austral summer days are
741: much longer than nights, this could result into a brighter nightglow.
742: The data show actually the opposite behavior, since during austral
743: summer the night sky reaches its lowest average brightness. Moreover,
744: this appears to be in phase with what is observed in the northern
745: hemisphere, where the sky is darker in winter than is summer (see for
746: example Liu et al. \cite{liu}). This seems to indicate that the SAO
747: must be related to some other, non local mechanism (see the discussion
748: in Sec.~\ref{sec:disc}). Remarkably, but after all not surprisingly,
749: polar auroral activity shows a similar temporal fluctuation, with
750: maxima in spring and autumn (see for example Meinel, Neighed
751: \& Chamberlain \cite{meinel}).
752:
753: \section{\label{sec:spectra}Spectroscopic analysis}
754:
755: Since there are many distinct components that contribute to the global
756: nightglow emission (Leinert et al. \cite{leinert} and references
757: therein) that cannot be disentangled with broad-band photometry,
758: several attempts have been made in the past to increase the spectral
759: resolution using a set of narrow band filters. After the pioneering
760: work by Barbier (\cite{barbier}; see also Chamberlain
761: \cite{chamberlain} for a review on this subject), who used 8
762: intermediate-band photometry, several other researchers have
763: undertaken similar projects (see for example Leinert et
764: al. \cite{leinert95} and Mattila et al. \cite{attila} for two more
765: recent works). Nevertheless, due to the relatively large bandwidths
766: (typically broader than 100 \AA), the study of single features,
767: especially if not very intense, has always been hindered by the
768: presence of the pseudo-continuum and possible adjacent lines and/or
769: bands.
770:
771: \begin{figure}
772: \centering
773: \includegraphics[width=9cm]{oi5577.jpg}
774: \caption{\label{fig:oi5577}Lower panel: dark time, zenith corrected
775: [OI]5577 line flux as a function of $SFD$. The solid line traces
776: a linear least-squares fit to the data. Upper panel: line fluxes as a
777: function of time from the beginning of the year. The data have been
778: corrected to solar minimum ($SFD$=0.8 MJy) using the relation shown in
779: the lower panel. The large points mark monthly averages and the right
780: vertical scale is expressed in Rayleigh.}
781: \end{figure}
782:
783: \subsection{\label{sec:lines}Main atomic emission lines}
784:
785: In this section I will analyze the behavior of the main emission lines
786: in the optical domain. For convenience, I will express the line fluxes
787: in Rayleighs\footnote{1R=10$^6/4\pi$ photons s$^{-1}$ cm$^{-2}$
788: sr$^{-1}$ $\equiv$ 3.72$\times$10$^{-14}\; \lambda^{-1}$(\AA) erg
789: s$^{-1}$ cm$^{-2}$ arcsec$^{-2}$.}.
790:
791: \subsubsection{\label{sec:oi5577}The [OI] lines}
792:
793: The [OI]5577 is generally the most prominent feature in the
794: optical night sky spectrum. It falls right in the center of the $V$
795: passband, giving a typical contribution of 20\% to the global surface
796: brightness in this filter. It has a typical intensity of $\sim$250 R,
797: it arises in layers placed at about 90 km (Roach \& Gordon
798: \cite{roach}) and it displays a marked dependency on solar activity
799: (Rayleigh \cite{rayleigh}). This is clearly shown also by the data
800: presented here (Fig.~\ref{fig:oi5577}, lower panel), which indicate
801: also the presence of pronounced fluctuations (40 to 750 R
802: peak-to-peak) around the average level ($\sim$230 R). To quantify the
803: correlation with solar activity and following the procedure that has
804: been applied to the broad band data (see Sec.~\ref{sec:sun}), I have
805: fitted to the data a law of the type $\log F = \log F_0 + \gamma \;
806: SFD$. The results for this and other lines are presented in
807: Table~\ref{tab:lines}, which shows also the number of data points used
808: ($N$) and the linear correlation coefficient ($r$). What is
809: interesting to note is that, once the solar dependency is removed from
810: the data, the flux of this line displays a marked SAO (see
811: Fig.~\ref{fig:oi5577}, upper panel), similar to that seen in the broad
812: band data (Fig.~\ref{fig:season}). This fact has been already noticed
813: by Buriti et al. (\cite{buriti}), who found that this line and other
814: mesospheric features all show a SAO.
815:
816: The same behavior is, in fact, shown by the [OI]6300,6364\AA\/ doublet (see
817: Fig.~\ref{fig:oi6300}), which is produced at 250-300 km (Roach \&
818: Gordon \cite{roach}) and it is known to undergo abrupt intensity
819: changes on two active regions about 20$^\circ$ on either side of the
820: geomagnetic equator (Barbier \cite{barbier2}), hence marginally
821: including the Paranal site. Indeed, the [OI]6300 measured fluxes
822: show very strong variations (10 to 950 R) around the average level
823: ($\sim$150 R), with spikes reaching $\sim$1 kR, so that the line
824: fluxes span almost a factor 100 (to be compared with the factor
825: $\sim$25 measured for the [OI]5577. In this respect, the behavior
826: of the [OI]6300,6364\AA\/ doublet is different from that of the
827: [OI]5577 line, since in a significant number of cases its flux is
828: very small and the line is practically invisible, lost in the OH(9-3)
829: molecular band (see Fig.~\ref{fig:skyspectrum}). In fact, its flux is
830: less than 80 R for more than 50\% of the cases considered in this
831: work, the minimum recorded value being $\sim$10 R
832: (Fig.~\ref{fig:linehist}). The variation range of these two features
833: and their relative contributions to the $V$ and $R$ total flux are
834: summarized in Table~\ref{tab:oi}. From the values reported in this
835: table, one can see that [OI]5577 can produce a maximum variation
836: of $\sim$ 0.5 mag arcsec$^{-2}$ in the $V$ band, while a similar
837: effect is produced by the [OI] doublet in the $R$
838: passband\footnote{Since the [OI]6364/[OI]6300 lines ratio is 1/3,
839: the maximum combined contribution is about 60\%.}. Given the fact that
840: the RMS variation in the $V$ and $R$ passbands is about 0.25 mag
841: arcsec$^{-2}$ (see Table
842: \ref{tab:dark}), this implies that the fluctuations seen in these passband
843: are not completely accounted by the changes in the atomic O line
844: fluxes. The two lines appear to show a very weak correlation: the
845: average ratio F(6300)/F(5577) is 0.64, but in a significant number of
846: cases ($\sim$25\%) this ratio is larger than 1 (see also
847: Sec.~\ref{sec:corr}). In those circumstances, the [OI]6364\AA\/ line
848: becomes the most prominent nightglow atomic feature in the optical
849: domain.
850:
851:
852:
853:
854: \begin{table}
855: \tabcolsep 2mm
856: \caption{\label{tab:oi}Observed line intensities and relative flux
857: contribution to $V$ and $R$ passbands for [OI]5577, [OI]6300
858: and Na~I D. All values are corrected to zenith.}
859: \begin{tabular}{ccccc}
860: \hline
861: \hline
862: & \multicolumn{2}{c}{F} & V & R \\
863: {\bf [OI]5577} & (10$^{-16}$ erg s$^{-1}$ cm$^{-2}$ arcsec$^{-2})$ & ($R$) & & \\
864: \hline
865: min & 2.6 & 40 & 3\% & $<$1\%\\
866: ave & 16 & 230 & 20\% & 1\% \\
867: max & 50 & 750 & 62\% & 4\% \\
868: & & & & \\
869: \hline
870: {\bf [OI]6300} & & & & \\
871: min & 0.6 & 10 & $<$1\% & $<$1\%\\
872: ave & 9 & 150 & 2\% & 7\% \\
873: max & 56 & 950 & 12\% & 45\% \\
874: & & & & \\
875: \hline
876: {\bf Na~I D} & & & & \\
877: min & 0.7 & 10 & $<$1\% & $<$1\%\\
878: ave & 3 & 50 & 2\% & 3\% \\
879: max & 10 & 160 & 8\% & 9\% \\
880: \hline
881: \end{tabular}
882: \end{table}
883:
884: As far as the solar activity is concerned, it must be noticed that
885: even though a dependence from the $SFD$ is seen, it is less clear than
886: in the case of the [OI]5577 (see Fig.~\ref{fig:oi6300}, lower
887: panel). In fact, even though the slopes are very similar, the linear
888: correlation coefficient for [OI]6300 is 0.23, to be compared with
889: 0.49 measured for the [OI]5577 line (see also
890: Tab.~\ref{tab:lines}).
891:
892: It is well known that, during the first hours of the night, the
893: [OI]6300 line shows a steady decline in brightness which, at
894: tropical sites, is interrupted by abrupt emission bursts (see, for
895: instance, Roach \& Gordon \cite{roach}). A plot of the line fluxes as
896: a function of time elapsed after the evening twilight (see
897: Fig.~\ref{fig:decay}) clearly shows that these events occur during the
898: whole night, and not only before midnight, as sometimes stated in the
899: literature (see for instance Benn \& Ellison \cite{benn}). Actually,
900: one interesting fact that emerges from this analysis is that, when the
901: nights are short (filled symbols), i.e. during austral summer, most of
902: the line enhancements tend to take place in the second half of the
903: night. On the other hand, when the nights are long (i.e. lasting more
904: than 8.5 hours), flux changes tend to become smaller and smaller
905: during the last two hours of the night. No such dichotomy is observed
906: for the [OI]5577 line (see Fig.~\ref{fig:decay}, lower panel).
907:
908:
909:
910: \begin{table}
911: \caption{\label{tab:lines} Linear least squares fit parameters for the
912: line fluxes vs.solar activity relation. Values in parenthesis indicate
913: RMS uncertainties. Input data have been corrected to zenith.}
914: \tabcolsep 2.2mm
915: \begin{tabular}{cccccc}
916: \hline
917: Line & $\log F_0$ & $\gamma$ & $\sigma$ & $r$ & $N$ \\
918: \hline
919: OI~5577 & $-$15.36 (0.03) & 0.27 (0.02) & 0.22 & 0.49 & 876\\
920: OI~6300 & $-$15.69 (0.07) & 0.25 (0.05) & 0.41 & 0.23 & 635 \\
921: Na~I~D & $-$15.72 (0.03) & 0.04 (0.02) & 0.25 & 0.07 & 876\\
922: NI~5200 & $-$17.19 (0.05) & 0.37 (0.03) & 0.32 & 0.45 & 727\\
923: \hline
924: \end{tabular}
925: \end{table}
926:
927: \begin{figure}
928: \centering
929: \includegraphics[width=9cm]{oi6300.jpg}
930: \caption{\label{fig:oi6300}Same as Fig.~\ref{fig:oi5577} for the
931: [OI]6300 line.}
932: \end{figure}
933:
934: \begin{figure}
935: \centering
936: \includegraphics[width=9cm]{linehist.jpg}
937: \caption{\label{fig:linehist}Dark time line flux distributions for
938: [OI]5577\, [OI]6300,6364\AA, Na~I~D and N~I 5200. The vertical
939: lines trace the average (dashed) and the median (dotted-dashed) of the
940: distributions. The upper horizontal scale in each panel is expressed
941: in Rayleigh.}
942: \end{figure}
943:
944: \begin{figure}
945: \centering
946: \includegraphics[width=9cm]{decay.jpg}
947: \caption{\label{fig:decay}Line fluxes as a function of time elapsed from
948: evening twilight for [OI]6300 (upper panel) and [OI]5577
949: (lower panel). The vertical dashed lines mark the shortest and longest
950: nights (7.4 and 10.7 hours respectively), while dotted lines indicate
951: the midnight in the two cases. Filled symbols indicate data obtained
952: in nights whose length is between 7.4 and 8.4 hours.}
953: \end{figure}
954:
955: Finally, as for the broad band data (see Sec.~\ref{sec:timescale}), I
956: have calculated the correlation coefficient between line fluxes and
957: SFD for different values of the time delay $\tau$ and time window
958: $\Delta \tau$. An instructive example is presented in
959: Fig.~\ref{fig:scoi5577}, showing the case of the [OI]5577 line,
960: which displays the strongest dependency on solar activity. The
961: correlation function shows a peak at about $\tau$=15 days for $\Delta
962: \tau$=1 day. The maximum correlation increases for a $\Delta \tau$=10
963: days, peaking at $\tau\simeq$12.5 days. As for the photometric data
964: (see Fig.~\ref{fig:sunanal}), spurious correlation peaks due to the 27
965: days solar rotation are present. A similar analysis for the
966: [OI]6300 line shows two similar peaks at $\tau\simeq$13 days and
967: 27 days later.
968:
969: \begin{figure}
970: \centering
971: \includegraphics[width=9cm]{scoi5577.jpg}
972: \caption{\label{fig:scoi5577}Linear correlation coefficient (absolute value)
973: as a function of time delay $\tau$ for the [OI]5577 line,
974: computed for four different values of $\Delta \tau$ (1, 10, 30 days).}
975: \end{figure}
976:
977:
978: \subsubsection{\label{sec:naid}The Na~I D lines}
979:
980: The Na~I~D doublet is known to originate in a layer placed at about 92
981: km and to undergo strong seasonal variations around an average value
982: of 50 R (Roach \& Gordon \cite{roach}). For the Paranal site, the
983: doublet varies by a factor larger than 15, reaching a maximum value of
984: 160 R in April. The maximum contribution of the Na~I~D doublet to the
985: global night sky brightness reaches $\sim$0.1 mag arcsec$^{-2}$ both
986: in $V$ and $R$ passbands (see Table~\ref{tab:oi}).
987:
988: While the Na~I~D doublet intensity does not show any significant
989: correlation with solar activity (the correlation coefficient is only
990: 0.07; see Fig.~\ref{fig:naid} and Table~\ref{tab:lines}), it does show
991: a clear SAO, with a peak in April and a secondary peak in
992: October/November (see Fig.~\ref{fig:naid}, upper panel), as it is seen
993: in all atomic lines analyzed in this work. The SAO displayed by
994: mesospheric sodium column density is a well studied phenomenon
995: (Kirchoff \cite{kirkhoff}), also because of the importance of the
996: sodium layer for the laser guide star adaptive optics systems (see for
997: instance Ageorges \& Hubin \cite{ageorges}). The fact that different
998: species like Na~I and [OI] all show a SAO is interpreted in the light
999: of the role of O atoms in the source photochemical reactions (Slanger
1000: \cite{slanger}).
1001:
1002:
1003: \begin{figure}
1004: \centering
1005: \includegraphics[width=9cm]{naid.jpg}
1006: \caption{\label{fig:naid}Same as Fig.~\ref{fig:oi5577} for the
1007: Na~I~D doublet. In this case, no solar flux correction has been applied.}
1008: \end{figure}
1009:
1010: Since the resolution provided by the grism 600R is sufficient to
1011: resolve the D$_1$ and D$_2$ components (see Fig.~\ref{fig:uves}),
1012: following the work done by Slanger et al. (\cite{slanger}), I have
1013: estimated the intensity ratio D$_2$/D$_1$ during dark time, on a total
1014: of 147 spectra. The region of interest is contaminated by the presence
1015: of at two features belonging to the OH(8-2) band, which I will
1016: indicate as B and C (see Fig.~\ref{fig:uves} for their
1017: identification). Using a high resolution UVES spectrum of the night
1018: sky (Hanuschik \cite{hanuschik}), I have estimated the intensity ratio
1019: between these two features and another OH(8-2) line at 5932.9\AA\/
1020: ($P_{11}(3.5)$, Abrams et al. \cite{abrams}), which I will indicate as
1021: E. Since this feature is well measurable in the 600R spectra, assuming
1022: that these intensity ratios are constant (B/E=0.95, C/B=0.35), I have
1023: estimated the intensity of B and C from the measured intensity of
1024: component E. Then, after subtracting to the original data two Gaussian
1025: profiles centered at the laboratory wavelengths of B and C, I have
1026: fitted the residual with a double Gaussian profile, fixing
1027: the distance between the two components (5.98\AA) and their FWHM
1028: (4.5\AA). Finally, I have derived the D$_2$/D$_1$ simply
1029: computing the ratio between the central intensities of the two
1030: fitted Gaussians. The pseudo-continuum has been removed using a
1031: first order polynomial.
1032:
1033:
1034: \begin{figure}
1035: \centering
1036: \includegraphics[width=9cm]{uves.jpg}
1037: \caption{\label{fig:uves}Night sky spectrum taken with UVES
1038: (Hanuschik \cite{hanuschik}) in the region of the Na~I~D lines
1039: (resolution 0.15\AA\/ FWHM. The labels A, B and C indicate the OH
1040: Meinel 8-2 band lines $Q_{22}(0.5)$ 5887.1\AA, $Q_{11}(1.5)$
1041: 5888.2\AA\/ and $Q_{11}(2.5)$ 5894.5\AA\/ (Abrams et
1042: al. \cite{abrams}). The insert shows the corresponding spectrum
1043: obtained with FORS1 and the 600R grism (resolution 4.5\AA\/ FWHM).
1044: The same spectrum is shown also in the main panel (thin line).}
1045: \end{figure}
1046:
1047: The results, shown in Fig.~\ref{fig:naidratio} (upper panel), are
1048: perfectly in line with the findings published by Slanger et
1049: al. (\cite{slanger}): the intensity ratio, which should be 2.0 if the
1050: two transitions are produced according to their spin-orbit statistical
1051: weights, varies from 1.2 to 1.8, with most of the data lying between
1052: 1.5 and 1.7. The average value for the FORS1 sample is 1.64, with an
1053: RMS deviation of 0.08. Using a larger sample including more than 300
1054: high resolution spectra, Slanger et al. (\cite{slanger}) have found
1055: evidences for a SAO for the D$_2$/D$_1$ intensity ratio. This is not
1056: detected in the low-resolution data set presented here, but this is
1057: probably due to the fact that the present sample includes less than
1058: half as many data points.
1059:
1060:
1061: \begin{figure}
1062: \centering
1063: \includegraphics[width=9cm]{naidratio.jpg}
1064: \caption{\label{fig:naidratio}Upper panel: distribution of the Na~I D$_2$/D$_1$
1065: intensity ratio during dark time. Lower panel: Na~I D$_2$/D$_1$
1066: intensity ratio as a function of time from the beginning of the
1067: year. The horizontal dashed line marks the average value.}
1068: \end{figure}
1069:
1070:
1071: \subsubsection{\label{sec:ni}The N~I 5200 line}
1072:
1073: The N~I feature at $\sim$5200\AA\/ is actually a blend of several
1074: transitions. It is supposed to originate at about 260 km altitude and
1075: has a typical intensity of 1 R (Roach \& Gordon \cite{roach}). As I
1076: have shown in Paper I, this line shows abrupt changes, possibly
1077: following the behavior of the [OI]6300 line. The data discussed in
1078: this paper show that the flux of this line ranges from practically
1079: zero (the line is not detected) to about 30 R; moreover, the flux
1080: distribution is rather similar to that of [OI]6300 (see
1081: Fig.~\ref{fig:linehist}), strengthening the impression that these two
1082: lines are related. This line displays also a
1083: strong dependency on solar activity, with a correlation factor
1084: similar to that of the [OI]5577 line (see Table~\ref{tab:lines}).
1085: Finally, as for all other atomic lines discussed here, it shows a SAO
1086: (see Fig.~\ref{fig:ni5200}).
1087:
1088: \begin{figure}
1089: \centering
1090: \includegraphics[width=9cm]{ni5200.jpg}
1091: \caption{\label{fig:ni5200}Same as Fig.~\ref{fig:oi5577} for the
1092: N~I~5200\AA\/ line.}
1093: \end{figure}
1094:
1095: The maximum contribution of this line to the global flux in the
1096: $V$ band during dark time is about 2\%.
1097:
1098:
1099:
1100: \subsection{\label{sec:bands}Main molecular emission bands}
1101:
1102: \subsubsection{\label{sec:oh}The OH bands}
1103:
1104: The OH bands analyzed here (see Table~\ref{tab:flux}) show a very
1105: tight mutual correlation, in the sense that they appear to vary in
1106: unison. Moreover, they do not show any correlation with solar activity
1107: ($r\leq$0.1 for all bands), while they show a SAO, even though not as
1108: pronounced as in the case of the other features discussed so far (see
1109: Fig.~\ref{fig:oh83} for an example). The flux distribution appears to
1110: be much more symmetric around the average value than in the case of
1111: atomic lines. All bands shows the same range of variation, which is
1112: close to a factor 2 around the mean value. Given the intensity of
1113: these features (especially OH(8-3) and OH(6-2)), their variability is
1114: certainly the dominating source of sky brightness fluctuations in the
1115: I passband.
1116:
1117: \begin{figure}
1118: \centering
1119: \includegraphics[width=9cm]{oh83.jpg}
1120: \caption{\label{fig:oh83}Same as Fig.~\ref{fig:oi5577} for the
1121: OH(8-3) molecular band. No solar flux correction has been applied. }
1122: \end{figure}
1123:
1124:
1125: \subsubsection{\label{sec:02}The O$_2$(0-1) band}
1126:
1127: This band shows a clear correlation with solar activity ($r$=0.41) and
1128: the same SAO observed for all other features (see Fig.~\ref{fig:o2}).
1129: Its integrated flux varies by about a factor 15, between 80 R and 1.1
1130: kR.
1131:
1132: \begin{figure}
1133: \centering
1134: \includegraphics[width=9cm]{o2.jpg}
1135: \caption{\label{fig:o2}Same as Fig.~\ref{fig:oi5577} for the
1136: O$_2$(0-1) molecular band.}
1137: \end{figure}
1138:
1139: \subsection{\label{sec:blue} The blue bands and pseudo-continuum}
1140:
1141: None of the blue bands, which are supposed to trace the behavior of
1142: the pseudo-continuum generated by the O$_2$ and NO$_2$ molecular bands
1143: (Roach \& Gordon \cite{roach}), shows a clear dependency on
1144: solar activity ($r\leq$0.3). However, it must be noticed that, while
1145: the emission features are purely generated within the atmosphere, the
1146: continua are significantly influenced by the extra-terrestrial
1147: background (Roach \& Gordon \cite{roach}), which is difficult to
1148: remove. In fact, for the sake of simplicity, no differential zodiacal
1149: light contribution has been applied to the continuum
1150: measurements. This is certainly affecting the blue bands and the
1151: continuum regions, especially those close to the zodiacal
1152: light spectrum peak. This and the improper removal of airmass effects
1153: is most likely the cause for the lower correlations shown by these
1154: features and the solar activity, with respect to what is detected, for
1155: example, for the $U$ and $B$ passbands.
1156:
1157: The integrated flux varies of about a factor 2 around the average
1158: value, with a fairly symmetric distribution. As already indicated by
1159: the $B$ band data (see Sec.~\ref{sec:season}), there is no significant
1160: trace of a SAO for none of the blue bands (see Fig.~\ref{fig:b4},
1161: lower panel). Interestingly, while also the continuum range C1
1162: (5040-5120\AA, see Table~\ref{tab:flux}) does not display any evidence
1163: for seasonal fluctuations, the redder ranges C2 to C6 show a possible
1164: broad annual oscillation, with a peak in June (Fig.~\ref{fig:b4},
1165: lower panel).
1166:
1167: \begin{figure}
1168: \centering
1169: \includegraphics[width=9cm]{b4.jpg}
1170: \caption{\label{fig:b4}Seasonal variation of blue band B4 (upper panel) and
1171: continuum range C2 (lower panel).}
1172: \end{figure}
1173:
1174: While the region between 3100 and 3900\AA\/ (mostly not covered by the
1175: data set discussed in this paper), shows a number of emission features
1176: attributable to O$_2$ Herzberg and Chamberlain bands, the spectral
1177: interval 3900-4900 is almost a pure pseudo-continuum (see for instance
1178: Broadfoot \& Kendall \cite{broadfoot}, and Fig.~\ref{fig:bluband}
1179: here). Nevertheless, clear variations are seen in the FORS1 database,
1180: as it is illustrated in Fig.~\ref{fig:blufeat}, where I have compared
1181: to spectra obtained with the same instrumental setup and similar
1182: signal-to-noise ratio on two different nights. Whilst the two spectra
1183: show the same overall emission features, these are much more
1184: pronounced in the data obtained in November 1999 .
1185:
1186: \begin{figure}
1187: \centering
1188: \includegraphics[width=9cm]{blufeat.jpg}
1189: \caption{\label{fig:blufeat}The spectral region 3750-5400\AA\/ on
1190: two different epochs. Both spectra were obtained in dark time with the
1191: 600B grism. The two narrow absorption visible between 3900 and
1192: 4000\AA\/ are the CaII H\&K solar features.}
1193: \end{figure}
1194:
1195: For an explorative analysis, I have measured the fluxes of some of the
1196: most prominent features, which I have marked in Fig.~\ref{fig:blufeat}
1197: (lower panel) and I will indicate as bf1 (3720-3767\AA), bf2
1198: (3807-3860\AA), bf3 (4150-4200\AA), bf4 (4387-4460\AA) and bf5
1199: (4809-4859\AA). While bf1, bf2 and bf3 can be tentatively identified
1200: as N$_2$ and N$^+_2$ molecular bands (see Chamberlain
1201: \cite{chamberlain} and also the next section here), the identification of
1202: bf4 and bf4 is more difficult. For simplicity, I have used a common
1203: value for the blue pseudo-continuum (CB), which was estimated in the
1204: spectral range 4730-4760\AA, that does not show any trace of emission
1205: features (see for instance Hanuschik \cite{hanuschik}). The results
1206: are summarized in Table~\ref{tab:blufeat}, that reports the maximum
1207: and median values derived from the 114 suitable spectra used in the
1208: analysis, together with the percentage of non-detections.
1209:
1210: The largest range of variation is shown by bf1, which can reach a
1211: maximum flux of about 1.4$\times$10$^{-16}$ erg s$^{-1}$ cm$^{-2}$
1212: arcsec$^{-2}$, while bf2 to bf5 reach peak fluxes between 6 and
1213: 8$\times$10$^{-17}$ erg s$^{-1}$ cm$^{-2}$ arcsec$^{-2}$. Despite
1214: their similar peak values, these features display different
1215: behaviors. For instance, while bf5 is practically always detected, bf2
1216: is absent in more than 50\% of the cases. Since the median value of
1217: the pseudo-continuum is 7.2$\times$10$^{-18}$ erg s$^{-1}$ cm$^{-2}$
1218: arcsec$^{-1}$, each of these features contributes less than 2\% to the
1219: integrated flux between 3700 and 5000\AA. Therefore, their
1220: fluctuations can account only partially for the variations observed in
1221: the B passband, that must be related to the changes in the
1222: pseudo-continuum.
1223:
1224: \begin{table}
1225: \tabcolsep 2.5mm
1226: \caption{\label{tab:blufeat} Dark time maximum and median fluxes for the
1227: blue emission features indicated in Fig.~\ref{fig:blufeat}. The last
1228: column gives the frequency of non-detection (flux below 10$^{-17}$ erg
1229: s$^{-1}$ cm$^{-2}$ arcsec$^{-1}$).}
1230: \centerline{
1231: \begin{tabular}{ccccc}
1232: \hline
1233: bf & wavelength range & max & med & n.d. \\
1234: \# & (\AA) &
1235: \multicolumn{2}{c}{(10$^{-17}$ erg s$^{-1}$ cm$^{-2}$ arcsec$^{-2}$}) & \\
1236: \hline
1237: 1 & 3720-3767 & 14.4 & 5.4 & 26\% \\
1238: 2 & 3807-3860 & 8.4 & 0.0 & 54\% \\
1239: 3 & 4150-4200 & 6.4 & 1.3 & 38\% \\
1240: 4 & 4387-4460 & 7.2 & 2.8 & 17\% \\
1241: 5 & 4809-4859 & 8.0 & 3.9 & 1\% \\
1242: \hline
1243: \end{tabular}
1244: }
1245: \end{table}
1246:
1247:
1248: \subsubsection{The strange event of November 8, 2004}
1249:
1250: Normally, there are no conspicuous isolated emission features in the
1251: airglow bluewards of 5200\AA, which is dominated by the
1252: pseudo-continuum (Roach \& Gordon \cite{roach}. See also
1253: Fig.~\ref{fig:bluband} here). However, N. Castro and M. Garcia, while
1254: analyzing a set of low resolution spectroscopic data obtained with
1255: FORS2 on November 8, 2004, have noticed the presence of unexplained
1256: emission features between 3600 and 4400\AA. A more detailed analysis
1257: has shown that these emissions where present on all the FORS2 data of
1258: that night, i.e. a set of MXU exposures 2700 seconds each, obtained
1259: between 00:15 and 07:08 UT\footnote{On November 8, 2004 the evening
1260: twilight ended on 00:23 UT and the morning twilight started on 08:27
1261: UT.} using the 600B grism. As expected, no trace of these features was
1262: visible in a similar data set obtained three days later. An
1263: example is shown in Fig.~\ref{fig:thunder} where, for comparison, an
1264: analogous spectrum obtained on November 11 is also plotted. Clearly,
1265: the two spectra differ mainly for the presence of two prominent
1266: emission bands, peaking at 3194\AA\/ and 4278\AA, which are identified
1267: as N$_2^+$ first negative bands 1N(0-0) and 1N(0-1) (Chamberlain
1268: \cite{chamberlain}; Table~5.4). These features, which are normally
1269: very weak or even absent in the nightglow (Broadfoot \& Kendall
1270: \cite{broadfoot}) are on the contrary typical of the aurora spectrum
1271: (Chamberlain
1272: \cite{chamberlain}). Besides being an extremely strange phenomenon at
1273: the latitudes of Paranal, an aurora would certainly be accompanied by
1274: other spectral markers, like for instance a large increase in the
1275: emission of the [OI]5577 line, which can reach in fact an
1276: intensity of 100 kR during a IBC III aurora (Chamberlain
1277: \cite{chamberlain}). The flux carried by this line in the same spectrum
1278: presented in Fig.~\ref{fig:thunder} is $\sim$183 R, that is slightly
1279: below the average level measured for Paranal (230R, see
1280: Sec.~\ref{sec:oi5577}). This definitely rules out an exceptional
1281: auroral event as the responsible for the unusual spectrum observed on
1282: November 8, 2004.
1283:
1284: \begin{figure}
1285: \centering
1286: \includegraphics[width=9cm]{thunder.jpg}
1287: \caption{\label{fig:thunder}The weird night sky spectrum of November 8, 2004.
1288: The N$_2^+$ first negative band identifications are from Chamberlain
1289: (\cite{chamberlain}; Table~5.4). For presentation the Nov 11 spectrum
1290: has been scaled by a factor 1.1, in order to match the
1291: pseudo-continuum on Nov 8.}
1292: \end{figure}
1293:
1294: A plausible explanation, proposed by Castro \& Garcia (private
1295: communication), is the contamination by the reflection from clouds of
1296: a number of lightning strokes. In effects, in the spectral range
1297: covered by the FORS2 data (3600-6100\AA), the most prominent features
1298: in a lightning spectrum are the N$_2^+$ first negative bands 1N(0-0)
1299: and 1N(0-1) (see for instance Wallace \cite{wallace}). Additionally,
1300: on the night of Nov 8 2004 thick and thin cirrus were reported in the
1301: ESO-Paranal night logs, substantiating the hypothesis of scattered
1302: light from a rather far thunderstorm.
1303:
1304: This kind of events must be indeed very rare, since no other example
1305: could be found in the FORS1 spectral data base presented in this
1306: paper.
1307:
1308: \section{\label{sec:corr}Correlations between spectral features}
1309:
1310: For a first exploratory analysis I have computed the linear
1311: correlation coefficients in the $\log F$-$\log F$ plane between all
1312: measured features.. The results are presented in
1313: Table~\ref{tab:fluxcorr} and they basically confirm the correlations
1314: found by Barbier (\cite{barbier}), even though new interesting facts
1315: do appear.
1316:
1317: The pioneering optical, eight-color photometric studies by Barbier
1318: (\cite{barbier}) have shown the existence of the so-called {\it
1319: covariance groups}: the {\bf green-line group} ([OI]5577, O$_2$
1320: Herzberg bands, the blue bands, the green continuum and the O$_2$(0-1)
1321: band), the {\bf sodium group} (Na~I~D doublet and the OH bands) and the
1322: {\bf red-line group}, which includes only the [OI]6300,6364
1323: doublet). So far, the latter appeared to be completely independent
1324: from any other component of the airglow (see Chamberlain
1325: \cite{chamberlain}). Nevertheless, as it is shown in
1326: Fig.~\ref{fig:oinicorr}, [OI]6300 shows a very tight correlation
1327: with the N~I feature at 5200\AA. The linear correlation factor in the
1328: log-log plane is $r$=0.95 and this appears to be one of the strongest
1329: correlation between airglow features found in the data set presented
1330: here, surpassed only by that shown by the OH bands (see
1331: Table~\ref{tab:fluxcorr}). To my knowledge, this is the first time
1332: this finding is reported; most likely, it escaped the attention of
1333: previous investigations simply because the N~I feature is rather weak
1334: ($\leq$30 R) and hence practically impossible to measure with
1335: intermediate passband filters.
1336:
1337: Even though a correlation between N~I 5200 and [OI]5577 is
1338: found (see Fig.~\ref{fig:oinicorr}, upper panel), this is less marked
1339: ($r$=0.56) and the spread around the best fit relation is much larger
1340: ($\sigma$=0.29 vs. $\sigma$=0.11). Finally, the correlation between
1341: [OI]5577 and [OI]6300 is indeed weak ($r$=0.29,
1342: $\sigma$=0.39); nevertheless, the data presented here seem to indicate
1343: that, on average, the maximum value attained by the red line is
1344: related to the flux of the green-line through the simple relation
1345: F([OI]6300)$\leq$2$\times$F([OI]5577).
1346:
1347:
1348: \begin{figure}
1349: \centering
1350: \includegraphics[width=9cm]{oinicorr.jpg}
1351: \caption{\label{fig:oinicorr}Correlation between N~I 5200 and
1352: [OI]5577 (upper panel) and [OI]6300 (lower panel). Only
1353: airmass correction has been applied.}
1354: \end{figure}
1355:
1356:
1357:
1358: \begin{table*}
1359: \caption{\label{tab:fluxcorr}Linear correlation factors in
1360: the log F-log F plane. Only dark time data have been used. The last
1361: two columns and rows report the correlation with daily (SFD$_d$) and
1362: monthly (SFD$_m$) solar flux density averages in the semi-log plane.}
1363: \tabcolsep 0.8mm
1364: \begin{tabular}{l|cccccccccccccccccccc|cc}
1365: \hline
1366: &C1 &5200 &C2 & Na~ID & C3 &5577 &C4 &6300 &C5 &(6-1) &(7-2) &(8-3) &C6 &(6-2)
1367: & O$_2$ &B1 &B2 &B3 &B4 &Broad & SFD$_d$ &SFD$_m$ \\
1368: \hline
1369: C1 & &0.15 &0.79 &0.14 &0.94 &0.31 &0.82 &0.02 &0.88 &0.23 &0.14 &0.15 &0.67 &0.14 &0.16 &0.87 &0.95 &0.99 &1.00 &0.60 &0.25 &0.23\\
1370: 5200 &0.15 & &0.36 &0.43 &0.29 &0.56 &0.31 &0.95 &0.22 &0.34 &0.32 &0.32 &0.35 &0.29 &0.21 &0.31 &0.22 &0.16 &0.17 &0.60 &0.31 &0.45\\
1371: C2 &0.79 &0.36 & &0.58 &0.94 &0.41 &0.99 &0.28 &0.96 &0.40 &0.32 &0.37 &0.92 &0.46 &0.35 &0.75 &0.77 &0.76 &0.79 &0.90 &0.12 &0.14\\
1372: Na~ID &0.14 &0.43 &0.58 & &0.35 &0.41 &0.52 &0.32 &0.41 &0.70 &0.72 &0.75 &0.63 &0.61 &0.61 &0.19 &0.14 &0.15 &0.16 &0.72 &0.03 &0.07\\
1373: C3 &0.94 &0.29 &0.94 &0.35 & &0.43 &0.97 &0.17 &0.98 &0.21 &0.12 &0.16 &0.86 &0.33 &0.32 &0.85 &0.89 &0.92 &0.94 &0.80 &0.23 &0.22\\
1374: 5577 &0.31 &0.56 &0.41 &0.41 &0.43 & &0.38 &0.29 &0.31 &0.31 &0.29 &0.32 &0.26 &0.25 &0.85 &0.61 &0.55 &0.37 &0.36 &0.59 &0.30 &0.49\\
1375: C4 &0.82 &0.31 &0.99 &0.52 &0.97 &0.38 & &0.26 &0.99 &0.36 &0.28 &0.32 &0.93 &0.44 &0.33 & - & - &0.81 &0.83 &0.90 &0.13 &0.13\\
1376: 6300 &0.02 &0.95 &0.28 &0.32 &0.17 &0.29 &0.26 & &0.19 &0.15 &0.13 &0.13 &0.28 &0.15 &0.01 & - & - &0.02 &0.02 &0.49 &0.09 &0.23\\
1377: C5 &0.88 &0.22 &0.96 &0.41 &0.98 &0.31 &0.99 &0.19 & &0.27 &0.19 &0.23 &0.91 &0.37 &0.26 & - & - &0.87 &0.88 &0.83 &0.14 &0.10\\
1378: (6-1) &0.23 &0.34 &0.40 &0.70 &0.21 &0.31 &0.36 &0.15 &0.27 & &0.98 &0.97 &0.67 &0.97 &0.53 & - & - &0.25 &0.25 &0.61 &0.02 &0.07\\
1379: (7-2) &0.14 &0.32 &0.32 &0.72 &0.12 &0.29 &0.28 &0.13 &0.19 &0.98 & &1.00 &0.63 &0.99 &0.57 & - & - &0.17 &0.17 &0.55 &0.06 &0.04\\
1380: (8-3) &0.15 &0.32 &0.37 &0.75 &0.16 &0.32 &0.32 &0.13 &0.23 &0.97 &1.00 & &0.65 &0.98 &0.60 & - & - &0.18 &0.18 &0.59 &0.04 &0.06\\
1381: C6 &0.67 &0.35 &0.92 &0.63 &0.86 &0.26 &0.93 &0.28 &0.91 &0.67 &0.63 &0.65 & &0.64 &0.34 & - & - &0.68 &0.68 &0.90 &0.02 &0.05\\
1382: (6-2) &0.14 &0.29 &0.46 &0.61 &0.33 &0.25 &0.44 &0.15 &0.37 &0.97 &0.99 &0.98 &0.64 & &0.47 & - & - &0.16 &0.16 &0.63 &0.06 &0.10\\
1383: O$_2$ &0.16 &0.21 &0.35 &0.61 &0.32 &0.85 &0.33 &0.01 &0.26 &0.53 &0.57 &0.60 &0.34 &0.47 & & - & - &0.24 &0.22 &0.52 &0.22 &0.41\\
1384: B1 &0.87 &0.31 &0.75 &0.19 &0.85 &0.61 & - & - & - & - & - & - & - & - & - & &0.96 &0.91 &0.89 & - &0.28 &0.25\\
1385: B2 &0.95 &0.22 &0.77 &0.14 &0.89 &0.55 & - & - & - & - & - & - & - & - & - &0.96 & &0.99 &0.97 & - &0.28 &0.24\\
1386: B3 &0.99 &0.16 &0.76 &0.15 &0.92 &0.37 &0.81 &0.02 &0.87 &0.25 &0.17 &0.18 &0.68 &0.16 &0.24 &0.91 &0.99 & &0.99 &0.61 &0.24 &0.25\\
1387: B4 &1.00 &0.17 &0.79 &0.16 &0.94 &0.36 &0.83 &0.02 &0.88 &0.25 &0.17 &0.18 &0.68 &0.16 &0.22 &0.89 &0.97 &0.99 & &0.62 &0.27 &0.26\\
1388: Broad &0.60 &0.60 &0.90 &0.72 &0.80 &0.59 &0.90 &0.49 &0.83 &0.61 &0.55 &0.59 &0.90 &0.63 &0.52 & - & - &0.61 &0.62 & &0.14 &0.24\\
1389: \hline
1390: SFD$_d$&0.25 &0.31 &0.12 &0.03 &0.23 &0.30 &0.13 &0.09 &0.14 &0.02 &0.06 &0.04 &0.02 &0.06 &0.22 &0.28 &0.28 &0.24 &0.27 &0.14 & &0.71\\
1391: SFD$_m$&0.23 &0.45 &0.14 &0.07 &0.22 &0.49 &0.13 &0.23 &0.10 &0.07 &0.04 &0.06 &0.05 &0.10 &0.41 &0.25 &0.24 &0.25 &0.26 &0.24 &0.71 & \\
1392: \hline
1393: \end{tabular}
1394: \end{table*}
1395:
1396: The strongest correlation within the [OI]5577 covariance group is
1397: that with the O$_2$ band ($r$=0.85), followed by the blue bands B1
1398: ($r$=0.61), N~I 5200\AA\/ ($r$=0.56), B2 ($r$=0.55), C3 ($r$=0.43) and
1399: C2 ($r$=0.41). The correlation with the other continuum regions is
1400: weaker ($r\leq$0.4). As for the Na~I~D group, besides the very tight
1401: correlations existing between the OH bands ($r\geq$0.97), the
1402: strongest correlation is observed between Na~I~D and OH(8-3)
1403: ($r$=0.75), followed by the other OH bands. Na~I~D correlates rather
1404: well with the continuum ranges C2 ($r$=0.58) and C6 ($r$=0.63) and
1405: with the O$_2$ band ($r$=0.61). As anticipated in the previous
1406: section, the only meaningful correlation found in the third covariance
1407: group is between [OI]6300 and the N~I 5200 line
1408: ($r$=0.95). This is actually one of the tightest correlations found
1409: among all features. Finally, all continuum bands C1 to C6 are well
1410: correlated with each other and with the blue bands B1 to B4, which
1411: show as well a very strong mutual correlation.
1412:
1413: Even though the analysis is certainly hampered by the smaller sample,
1414: I have run a similar exploratory study for the blue emission features
1415: described in Sec.~\ref{sec:blue}. The results are presented in
1416: Table~\ref{tab:blucorr}. The first interesting fact is that the blue
1417: emission features bf1 to bf4 are uncorrelated with the CB continuum
1418: ($|r|\leq$0.15), while some correlation is seen for bf5
1419: ($r\simeq$0.4). On the other hand, all blue features show a
1420: significant mutual correlation, which ranges from a minimum
1421: ($r\simeq$0.4) for bf2 and bf5 to a maximum ($r\simeq$0.9) for bf3 and
1422: bf4. In general, bf5 is the feature that shows the weakest correlation
1423: with the remaining blue features. Additionally, it displays the
1424: strongest correlation with the [OI]~5577 feature ($r\simeq$0.8) and
1425: solar activity. Indeed, the correlation with [OI]~5577 is rather
1426: marked for all bf's ($r\geq$0.5), suggesting that these features might
1427: belong to the green-line group of Barbier
1428: (\cite{barbier}). Nevertheless, bf1 and bf2 show a similarly marked
1429: correlation to the N~I 5200 line ($r\geq$0.5), that belongs to the
1430: red-line group. This suggests a partial correlation between the abrupt
1431: micro-auroral events undergone by the [OI]6300,6364\AA\/ doublet and
1432: the blue features activity.
1433:
1434:
1435:
1436: \begin{table*}
1437: \caption{\label{tab:blucorr}Same as Table~\ref{tab:fluxcorr} for the blue features.}
1438: \tabcolsep 0.8mm
1439: \centerline{
1440: \begin{tabular}{l|cccccccccc|cc}
1441: \hline
1442: &C1 &5200 & C3 &5577 & CB & bf1 & bf2 & bf3 & bf4 & bf5 & SFD$_d$
1443: &SFD$_m$ \\
1444: \hline
1445: C1 & &0.21 &0.97 &0.40 &1.00 &0.05 &0.01 &0.15 &0.15 &0.40 &0.32 &0.30\\
1446: 5200 &0.21 & &0.32 &0.59 &0.20 &0.56 &0.51 &0.21 &0.12 &0.13 &0.49 &0.46\\
1447: C3 &0.97 &0.32 & &0.48 &0.96 &0.15 &0.19 &0.23 &0.07 &0.46 &0.38 &0.36\\
1448: 5577 &0.40 &0.59 &0.48 & &0.38 &0.65 &0.47 &0.54 &0.56 &0.80 &0.54 &0.45\\
1449: CB &1.00 &0.20 &0.96 &0.38 & &0.05 &0.00 &0.15 &0.14 &0.38 &0.30 &0.28\\
1450: bf1 &0.05 &0.56 &0.15 &0.65 &0.05 & &0.82 &0.76 &0.54 &0.50 &0.39 &0.30\\
1451: bf2 &0.01 &0.51 &0.19 &0.47 &0.00 &0.82 & &0.80 &0.69 &0.44 &0.32 &0.08\\
1452: bf3 &0.15 &0.21 &0.23 &0.54 &0.15 &0.76 &0.80 & &0.91 &0.62 &0.21 &0.02\\
1453: bf4 &0.15 &0.12 &0.07 &0.56 &0.14 &0.54 &0.69 &0.91 & &0.69 &0.26 &0.18\\
1454: bf5 &0.40 &0.13 &0.46 &0.80 &0.38 &0.50 &0.44 &0.62 &0.69 & &0.50 &0.32\\
1455: \hline
1456: SFD$_d$ &0.32 &0.49 &0.38 &0.54 &0.30 &0.39 &0.32 &0.21 &0.26 &0.50 & &0.79\\
1457: SFD$_m$ &0.30 &0.46 &0.36 &0.45 &0.28 &0.30 &0.08 &0.02 &0.18 &0.32 &0.79 & \\
1458: \hline
1459: \end{tabular}
1460: }
1461: \end{table*}
1462:
1463: As far as the correlation with solar activity is concerned, it is
1464: interesting to note that while the blue pseudo-continuum shows a very
1465: similar correlation factor with $SFD_d$ and $SFD_m$, for the blue
1466: features bf2 and bf3 this is significantly larger when the daily
1467: averages are used. On the contrary, bf5 shows a stronger correlation
1468: with the monthly averages. Because of the limited sample, though,
1469: these results have to be taken with some caution. In fact, for
1470: instance, the [OI]5577 line shows a suspiciously high correlation
1471: factor with $SFD_d$ (0.54), at variance with the value derived from
1472: the whole data set (0.30; see Table~\ref{tab:fluxcorr}).
1473:
1474: \section{\label{sec:short}Short timescale variations}
1475:
1476: In general, the spectroscopic time coverage of the present sample is
1477: sparse and this makes the study of short time scale variations
1478: (minutes to hours) rather difficult. Interestingly, the sample
1479: presented here includes some data sets for which the spectroscopic
1480: observations went on for a significant fraction of the night. The most
1481: complete case is illustrated in Fig.~\ref{fig:seq}, which presents the
1482: behavior of some selected features during the night of 2001-04-22 as a
1483: function of time elapsed from the end of evening twilight, with an
1484: average interval of thirty minutes. This sequence definitely shows that
1485: the variations are smooth, with typical timescales of the order of
1486: several hours. As expected, the most marked change is seen in the
1487: [OI]6300 feature that, during the first hours of the night, went
1488: through one of the intensity enhancements typical for sites close to
1489: the geomagnetic equator.
1490:
1491: The smooth time evolution seen in the emission features matches the
1492: behavior observed in broad band photometry when long time series are
1493: available (see Paper I, Sec.~6).
1494:
1495: \begin{figure}
1496: \centering
1497: \includegraphics[width=9cm]{seq.jpg}
1498: \caption{\label{fig:seq}Sequence of spectral measurements for some of
1499: the features studied in this work. The data were obtained with FORS1
1500: on 2001-04-22 (G300V+GG435). Exposure times ranged from 10 to 20
1501: minutes. Several sky targets at different airmasses were observed. The
1502: vertical arrows indicate the times of target change. Airmasses spanned
1503: from 1.2 to 1.9. The vertical dashed line marks the start of morning
1504: astronomical twilight. The horizontal dashed lines mark the average
1505: value, while the dotted lines indicate $\pm$20\% levels.}
1506: \end{figure}
1507:
1508: \section{\label{sec:disc}Discussion}
1509:
1510: The analysis presented in this paper gives a clear picture of the
1511: complexity shown by the nightglow fluctuations, most of which remain
1512: unexplained. If it is well established that several features show a
1513: definite correlation with solar activity, to which they react with
1514: timescales of the order of a couple of weeks, more thorough
1515: investigations need to be performed in order to better understand the
1516: link between space weather and the phenomena taking part in the upper
1517: layers of Earth's atmosphere.
1518:
1519: In all the studies of the nightglow in the astronomical context, the
1520: radio flux at 10.2 cm has been used as the only proxy for the solar
1521: activity. Nevertheless, other transient solar phenomena might have
1522: some impact on the night sky brightness, like Flares, Coronal Mass
1523: Ejections and Solar Proton Events (see Hanslmeier
1524: \cite{hanselmeier} for a very recent review on solar phenomena).
1525: During these events, large amounts of energetic charged particles are
1526: released and, in the case they interact with the Earth's magnetic
1527: field, they cause a series of geomagnetic effects, including aurorae.
1528: Therefore, charged particles are potential responsibles for at least
1529: some of the observed nightglow fluctuations, including the SAO.
1530:
1531: Even though this will require a dedicated analysis, I have run a
1532: preliminary study using the proton flux measured by the CELIAS/MTOF
1533: Proton Monitor (Ipavich et al. \cite{ipavich}) on board of
1534: SOHO\footnote{Data can be downloaded from
1535: http://umtof.umd.edu/pm/crn/}, orbiting on the Earth-Sun line at about
1536: 1.5$\times$10$^6$ km away from Earth. The proton number flux (PNF) has
1537: been derived multiplying the proton density by the proton velocity
1538: measured by CELIAS/MTOF and it is plotted in Fig.~\ref{fig:sohospot}
1539: for the relevant time interval (upper panel).
1540:
1541: The PNF average value during this time span is 2.6$\times$10$^8$
1542: cm$^{-2}$ s$^{-1}$ (corresponding to a mass loss rate of
1543: $\sim$2$\times$10$^{-14}$ M$_\odot$ yr$^{-1}$), while the yearly
1544: averages range from 1.9 to 3.5$\times$10$^8$ cm$^{-2}$ s$^{-1}$. The
1545: maximum average value was reached in July 2002, i.e. some time after
1546: the secondary maximum seen in the radio flux. The proton velocity
1547: ranges from 270 to 1000 km s$^{-1}$ (median value 440 km s$^{-1}$),
1548: while the number density ranges from 0.1 to 73 cm$^{-3}$ (median value
1549: 4.7 cm$^{-3}$). Looking in more detail at the SOHO data, one notices
1550: that a number of isolated and short duration peaks are present in the
1551: PNF. In fact, in some cases, values as high as 4$\times$10$^9$
1552: cm$^{-2}$ s$^{-1}$ are reached. Already a look at
1553: Fig.~\ref{fig:sohospot} (upper panel) shows that the frequency of
1554: these spikes is higher closer to the solar maximum. This can be seen
1555: more quantitatively in the bottom panel of Fig.~\ref{fig:sohospot},
1556: where I have plotted the rate of what I will indicate as proton events
1557: (PE)\footnote{Note that these events differ from the Solar Proton
1558: Events, which are bursts of relativistic protons, with energies larger
1559: than 10 MeV.}. In this context, a PE has been defined as a group of
1560: adjacent CELIAS/MTOF measurements with PNF$\geq$7$\times$10$^8$
1561: cm$^{-2}$ s$^{-1}$.
1562:
1563: \begin{figure}
1564: \centering
1565: \includegraphics[width=9cm]{sohospot.jpg}
1566: \caption{\label{fig:sohospot}Upper panel: CELIAS/MTOF proton number flux.
1567: Central panel: solar radio flux. Lower panel: CELIAS/MTOF proton
1568: events (PF$\geq$7$\times$10$^8$ cm$^{-2}$ s$^{-1}$ ).}
1569: \end{figure}
1570:
1571: In principle, since the solar rotation axis is inclined by about 7.1
1572: degrees with respect to the ecliptic (Balthasar, Stark \& W\"ohl
1573: \cite{balthasar}), and because the solar wind flux is known to be
1574: higher from the poles than the equator, a modulation of the solar wind
1575: as seen from Earth is indeed expected. More precisely, the proton flux
1576: should be maximum around March 5 and September 5 when Earth lies at
1577: its highest/lowest heliographic latitude, respectively. For the same
1578: reason, the proton flux should be minimum around January 5 and June 5.
1579: This prediction can be easily compared to the real data using the
1580: CELIAS/MTOF measurements and looking at their behavior as a function
1581: of time elapsed since the beginning of the year. The result is shown
1582: in Fig.~\ref{fig:seasonprot}, which was produced using data obtained
1583: towards solar minimum (January 2004 to January 2007), for a total of
1584: about 28,000 data points. Indeed the PNF shows a SAO, with maxima in
1585: April and November, i.e. significantly shifted in time with respect to
1586: the epochs of maximum/minimum heliographic Earth's latitude.
1587: Remarkably, the SAO observed in the PNF appears to be in phase with
1588: the SAO detected both in the broad band data (Sec.~\ref{sec:season})
1589: and in the emission features (Sec.~\ref{sec:spectra}). Interestingly,
1590: a similar plot for the 10.2 cm radio flux does not show any clear
1591: trace of a SAO. Also, the evidence of a SAO in the PNF for the years
1592: around the maximum of solar cycle n.~23 becomes weaker. A possible
1593: explanation is that the more frequent and probably more stochastic
1594: PEs tend to dominate over the smoother SAO when the sun is
1595: more active.
1596:
1597: \begin{figure}
1598: \centering
1599: \includegraphics[width=9cm]{seasonprot.jpg}
1600: \caption{\label{fig:seasonprot}Seasonal variation of the MTOF/CELIAS proton
1601: number flux. Only data from 2004-01-01 to 2007-01-01 are plotted. The
1602: filled circles trace the monthly averages. No correction for the
1603: 11-years solar cycle has been applied.}
1604: \end{figure}
1605:
1606: Besides the smooth, long-term variation possibly produced by the
1607: modulation of the average proton fluence, it is reasonable to think
1608: that isolated PEs might be the cause of sporadic nightglow
1609: enhancements. To explore this possibility, I have run a correlation
1610: analysis similar to the one described in Sec.~\ref{sec:timescale} for
1611: the solar radio flux. The results are in general rather noisy and show
1612: that the correlation with PNF is always low ($|r|\leq$0.3), both for
1613: short ($\Delta t$=0.5 days) and long ($\Delta t$=30 days) time
1614: windows. An example for the $V$ passband is presented in
1615: Fig.~\ref{fig:sohoanal} (upper panel) for an averaging window $\Delta
1616: t$=10 days. The correlation coefficient reaches a peak $\sim$0.3 for
1617: $\tau\simeq$10 days, it decreases and it suddenly drops for $\tau>$27
1618: days. Even though the result is not really convincing, it might
1619: indicate that the response of the night sky brightness to PEs takes
1620: place with some time delay and that what matters is the proton flux
1621: behavior during the last solar rotation. It must be noticed that the
1622: PNF shows a rather marked recurrence, with a period of 9.1 days, that
1623: coincides with one third of the solar rotation period (see
1624: Fig.~\ref{fig:sohoanal}, lower panel). Similarly to what happens for
1625: the radio data (see Sec.~\ref{sec:timescale}), this probably creates
1626: spurious correlation peaks. Interestingly, shorter-timescale periodic
1627: variations in the geomagnetic activity have been detected for 13.3,
1628: 9.1 and 6.9 days (see for instance Hauska, Abdel-Wahab \& Dyring
1629: \cite{hauska}).
1630:
1631: Since the typical speed of solar wind is $\sim$450 km s$^{-1}$, the
1632: swarms of particles released during the PEs reach the Earth about one
1633: hour after being detected by SOHO. Therefore, the delay in the
1634: reaction is completely due to processes taking place within the
1635: Earth's atmosphere.
1636:
1637: As a last check for intermediate timescale periodic variations, I have
1638: investigated the correlation with the Moon motion. In fact, it has
1639: been suggested that atmospheric tides might induce recurrent
1640: oscillations in the night sky brightness (see Chamberlain
1641: \cite{chamberlain} and references therein). For this purpose, I have
1642: investigated possible links between the broad band, dark time
1643: measurements and lunar phase or lunar hour angle. No significant
1644: correlation has been found, in agreement with a similar analysis run
1645: by Mattila et al. (\cite{attila}).
1646:
1647: \begin{figure}
1648: \centering
1649: \includegraphics[width=9cm]{sohoanal.jpg}
1650: \caption{\label{fig:sohoanal}Upper panel: Linear correlation coefficient
1651: between dark time $V$ sky brightness and PNF as a function of time
1652: delay $\tau$, computed for $\Delta \tau$=10 days. Lower panel: example
1653: of PNF periodicity. The dotted curve is a replica of the original data
1654: shifted by 9.1 days.}
1655: \end{figure}
1656:
1657: \section{\label{sec:concl}Conclusions}
1658:
1659: In this paper I have presented a photometric and spectroscopic
1660: analysis of the optical night sky emission at Cerro Paranal in the
1661: time interval April 2001 - January 2007. The main conclusions of this
1662: work can be summarized as follows;
1663:
1664: \begin{itemize}
1665:
1666: \item The UBVRI night sky brightness is well correlated with solar activity.
1667: The correlation is maximum in $U$ and minimum in $I$.
1668:
1669: \item The excursion between sunspot maximum and minimum of solar cycle n.~23
1670: is 0.6, 0.3, 0.3, 0.4 and 0.2 mag arcsec$^{-2}$ in $U, B, V, R$ and
1671: $I$ passbands, respectively.
1672:
1673: \item There are indications that the effects of solar activity on nightglow
1674: are not identical across different solar cycles.
1675:
1676: \item The reaction of the nightglow to the solar variations seems to take
1677: place with a time delay of the order of 2-3 weeks.
1678:
1679: \item The $R$ night sky brightness seems to react with a much shorter delay,
1680: of the order of a few days.
1681:
1682: \item $V, R$ and $I$ measurements show a clear semi-annual oscillation (SAO),
1683: with a typical peak-to-peak variation of $\sim$0.5 mag arcsec$^{-2}$.
1684: For the $B$ passband this oscillation is, if present, much smaller.
1685:
1686: \item Maxima and minima of the SAO are out of phase with respect to the
1687: Equinoxes and Solstices.
1688:
1689: \item All main emission features show a SAO, very similar to the
1690: well known seasonal oscillation of the Na~I D doublet.
1691:
1692: \item $[OI]$ 5577 and NI~5200\AA\/ show the strongest correlation with
1693: solar activity. For [OI]~5577, the maximum correlation is found for a
1694: time delay of 15 days.
1695:
1696: \item $[OI]$ 6300\AA and NI~5200\AA\/ show a very tight mutual correlation.
1697: Nevertheless, [OI]6300 displays a weaker correlation with solar
1698: activity.
1699:
1700: \item Flux variations in the OH bands are very strongly correlated with
1701: each other and do not show any correlation with solar activity.
1702:
1703: \item The main emission features, both atomic and molecular, show smooth
1704: flux variations on time scales of hours.
1705:
1706: \item A preliminary and exploratory analysis of the possible connection
1707: between nightglow and flux of charged particles from the Sun has shown
1708: that there is a weak correlation.
1709:
1710: \item The night sky emission seems to react, with a delay of about 10 days,
1711: to variations in the proton number flux.
1712:
1713: \item No correlation is found between the dark time, broad band night sky
1714: brightness and moon phase or moon hour angle.
1715:
1716: \end{itemize}
1717:
1718: Future investigations, with even larger databases, will have to
1719: address the possible relations with other solar phenomena, like
1720: Coronal Mass Ejections, Flares and Solar Proton Events, in an attempt
1721: to connect the observed short timescale variations of the nightglow
1722: with space weather.
1723:
1724:
1725:
1726: \begin{acknowledgements}
1727:
1728: I am grateful to K. Krisciunas, for suggesting me to investigate the
1729: time scales of the night sky brightness dependency on the solar
1730: activity. I also wish to thank R. Mignani, S. M\"ohler and the ESO
1731: Quality Control Group for the support received during this
1732: work. Finally, I am grateful to N. Castro and M. Garcia for reporting
1733: the weird case of November 8, 2004. This paper is based on archival
1734: data obtained with ESO Telescopes at Paranal Observatory.
1735:
1736: \end{acknowledgements}
1737:
1738:
1739: \begin{thebibliography}{}
1740: \bibitem[1994]{abrams} Abrams, M. C., Davis, S. P., Rao, M. L. P.,
1741: Engleman, J.R., \& Brault, J.W. 1994, ApJS, 93, 351
1742: \bibitem[2000]{ageorges} Ageorges, N. \& Hubin, N. 2000, A\&AS, 144, 533
1743: \bibitem[1987]{balthasar} Balthasar, H., Stark, D. \& W\"ohl, H., 1987,
1744: A\&A, 174, 359
1745: \bibitem[1956]{barbier} Barbier, D. 1956, The Airglow and the Aurorae,
1746: Special Suppl. No. 5 to the J. Atm. Terr. Phys., p. 38
1747: \bibitem[1957]{barbier2} Barbier, D. 1957, Compt. Rend., 244, 1945
1748: \bibitem[1998]{benn} Benn, C. R. \& Ellison, S. L. 1998,
1749: La Palma Technical Note n.115
1750: \bibitem[1968]{broadfoot} Broadfoot, A.L. \& Kendall, K.R. 1968,
1751: J. Geophys. Res., 73, 426
1752: \bibitem[2001]{buriti} Buriti, R.A., Takahashi, H. \& Gobbi, D. 2001,
1753: Braz. J. Geophys., 19, 169
1754: \bibitem[1961]{chamberlain} Chamberlain, J. W. 1961, Physics of the
1755: Aurora and Airglow, (New York, Academic Press)
1756: \bibitem[1969]{covington} Covington, A. E. 1969, JRASC, 63, 125
1757: \bibitem[1988]{garstang88} Garstang, R. H. 1988, Observatory, 108, 159
1758: \bibitem[1989]{garstang} Garstang, R. H. 1989, PASP, 101, 306
1759: \bibitem[2003]{hanuschik} Hanuschik, R. 2003, A\&A, 407, 1157
1760: \bibitem[2007]{hanselmeier} Hanselmeier, A., 2007, {\it The Sun and
1761: Space Weather}, Springer, 2nd edition
1762: \bibitem[1973]{hauska} Hauska, H., Abdel-Wahab, S. \& Dyring, E., 1973,
1763: Physica Scripta, Vol.7, 135
1764: \bibitem[1999]{howard} Howard, R. 1999, in Allen's Astrophysical Quantities,
1765: ed. A.N. Cox (New York: AIP Press; Springer), 4th edition
1766: \bibitem[1998]{ipavich} Ipavich, F.M., et al., 1998, J. Geophys. Res.,
1767: 103, 17205
1768: \bibitem[1990]{krisc90} Krisciunas, K. 1990, \pasp, 102, 1052
1769: \bibitem[1997]{krisc97} Krisciunas, K. 1997, \pasp, 109, 1181
1770: \bibitem[2007]{krisc07} Krisciunas, K., et al. 2007, PASP, 119, 687
1771: \bibitem[1992]{landolt} Landolt, A. U. 1992, AJ, 104, 340
1772: \bibitem[1995]{leinert95} Leinert, Ch., V\"aisanen, P.,
1773: Mattila, K. \& Lehtinen, K. 1995, \aaps, 112, 99
1774: \bibitem[1998]{leinert} Leinert, Ch., Bowyer, S., Haikala, L. K.,
1775: et al. 1998, A\&AS, 127,1
1776: \bibitem[1980]{levasseur} Levasseur-Regourd, A. C. \& Dumont, R. 1980,
1777: \aap, 84, 277
1778: \bibitem[2003]{liu} Liu, Y., Zou, X., Sun, W., Ma, J., Wu, H., Jiang,
1779: Z., Xue, S. \& Chen, J. 2003, PASP, 114, 495
1780: \bibitem[1990]{lockwood} Lockwood, G. W., Floyd, R. D. \& Thompson, D. T. 1990,
1781: PASP, 102, 481
1782: \bibitem[1996]{attila} Mattila, K., V\"ais\"anen, P. \& Appen-Schnurr,
1783: G. F. O. 1996, \aaps, 119, 153
1784: \bibitem[1986]{kirkhoff}Kirkhoff, V.W.J.H. 1986, Can. J. Phys., 64, 1664
1785: \bibitem[1954]{meinel} Meinel, A. B., Negaard, B. J., \& Chamberlain,
1786: J. W. 1954, J. Geophys. Res., 59, 407
1787: \bibitem[2003a]{paperI} Patat, F. 2003a, A\&A, 400, 1183, {\bf Paper I}
1788: \bibitem[2003b]{patatII} Patat, F. 2003b, A\&A, 401, 797
1789: \bibitem[1989]{pila} Pilachowski, C. A., Africano, J. L., Goodrich. B. D.
1790: \& Binkert, W. S. 1989, \pasp, 101, 707
1791: \bibitem[1992]{press} Press, W.H., Teukolski, S.A., Vetterling, W.T. \&
1792: Flannery, B.P. 1992, Numerical Recipes, 2nd edition,
1793: (Cambridge, Cambridge University Press), p. 630
1794: \bibitem[1928]{rayleigh} Rayleigh, L. 1928, Proc. Roy. Soc. London,
1795: Ser. A, 119, 11
1796: \bibitem[1973]{roach} Roach, F. E. \& Gordon J. L. 1973, The light of
1797: the night sky, (Boston, Dordrecht Reidel)
1798: \bibitem[1967]{rosenberg} Rosenberg, N., \& Zimmerman, S.P. 1967, Planet.
1799: Space Sci. 15, 863
1800: \bibitem[1979]{schneeberger} Schneeberger, T. J., Worden, S. P \&
1801: Beckers, J. M. 1979, PASP, 91, 530
1802: \bibitem[2002]{szeifert} Szeifert, T. 2002, FORS1+2 User's Manual,
1803: VLT-MAN-ESO-13100-1543, Issue 2.3
1804: \bibitem[2005]{slanger} Slanger, T. G. et al., 2005, J. Geophys. Res.,
1805: 110, D23302
1806: \bibitem[1964]{wallace}Wallace, L., 1964, ApJ, 139, 994
1807: \bibitem[1988]{walker88} Walker, M. F. 1988, \pasp, 100, 496
1808: \end{thebibliography}
1809:
1810: \end{document}
1811:
1812: %
1813: \end{document}
1814:
1815:
1816: