1: \documentclass[12pt,preprint]{aastex}

2: \newcommand{\vdag}{(v)^\dagger}

3: \def\vec#1{{\bf #1}}

4: \def\balpha {\mbox{\boldmath $\alpha$}}

5: \def\bbeta {\mbox{\boldmath $\beta$}}

6: \def\deg{\ifmmode^\circ\else$^\circ$\fi}

7: \def\Fscr{\ifmmode{\mathcal{F}}\else$\mathcal{F}$\fi}

8: \def\Ascr{\ifmmode{\mathcal{A}}\else$\mathcal{A}$\fi}

9: \def\spie{Proc. of the SPIE}

10: 

11: \makeatletter

12: %\usepackage{rotating}

13: 

14: \newcommand{\myemail}{cstubbs@fas.harvard.edu}

15: \shorttitle{Atmospheric Transmission}

16: \shortauthors{C.W. Stubbs et al}

17: \begin{document}

18: \title{Towards More Precise Survey Photometry for PanSTARRS and LSST:

19: Measuring Directly the Optical Transmission Spectrum of the Atmosphere}

20: \author{Christopher W. Stubbs, F. William High, Matthew R. George,\\

21: Kimberly L. DeRose, and St{\'e}phane Blondin} 

22: \affil{Department of Physics \\

23: and \\

24: Harvard-Smithsonian Center for Astrophysics \\

25: Harvard University, Cambridge MA 02138 USA}

26: %\email{cstubbs@fas.harvard.edu}

27: 

28: \author{John L. Tonry, Kenneth C. Chambers, Benjamin R. Granett}

29: \affil{Institute for Astronomy, University of Hawaii\\

30: 2680 Woodlawn Drive, Honolulu, HI, 96822 USA}

31: 

32: \author{David L. Burke}

33: \affil{Kavli Institute for Particle Astrophysics and Cosmology\\

34: Stanford Linear Accelerator Center\\

35: Sand Hill Road, Palo Alto, CA, 94025 USA}

36: 

37: \author{R. Chris Smith}

38: \affil{CTIO/NOAO, 950 North Cherry Ave\\

39: Tucson AZ, 85719 USA}

40: 

41: \begin{abstract}

42: 

43: Motivated by the recognition that variation in the optical transmission of the atmosphere

44: is probably the main limitation to the precision of ground-based CCD measurements of celestial fluxes, 

45: we review the physical processes that attenuate the passage of light

46: through the Earth's atmosphere. The next generation of astronomical surveys, such as

47: PanSTARRS and LSST, will greatly benefit from dedicated apparatus to obtain atmospheric 

48: transmission data that can be associated with each survey image. We review and compare  

49: various approaches to this measurement problem, including photometry, spectroscopy, and 

50: LIDAR.  

51: In conjunction with careful measurements of instrumental throughput, 

52: atmospheric transmission measurements should allow next-generation imaging 

53: surveys to produce photometry of unprecedented precision. 

54: Our primary concerns are the real-time determination of 

55: aerosol scattering and absorption by water along the line of sight, both of 

56: which can vary over the course of a night's observations. 

57: 

58: \end{abstract}

59: 

60: \keywords{Astronomical Phenomena and Seeing}

61: 

62: %\clearpage

63: 

64: \section{INTRODUCTION}

65: 

66: Careful CCD measurements of differential photometry, {\it i.e.} comparing fluxes from similar

67: stars in the same image, allow comparisons at the millimagnitude level (\cite{Hartman05}, \cite{EH01}). On the 

68: other hand, with considerable effort the SDSS survey achieved \citep{SDSS07} of 

69: order 1\% uncertainty in zeropoint uniformity

70: across the sky in the $g,r,i$ and $z$ bands, and twice that in $u$.  

71: 

72: What accounts for this apparent factor of ten difference 

73: in our ability to compare the fluxes of celestial objects? The SDSS team attributes \citep{SDSS07} 

74: the dominant contribution to their zeropoint uncertainty as arising from   ``...unmodelled atmospheric variations... .''  By combining 58 scans across one SDSS equatorial region, \cite{Ivezic07} achieved zeropoint scatter of just under 1\%. This falls short of the $\sqrt{58}~\sim~7$ fold improvement over single-scan calibration that 

75: one would expect from random errors. This implies that the SDSS photometry is bottoming out in 

76: some combination of flatfielding residuals and atmospheric variability. 

77: This fact does not diminish

78: the importance or the power of the SDSS data set, but future 

79: survey programs need to identify and overcome the factors

80: that limited the precision of the SDSS photometry, if they wish to do better. 

81: 

82: When comparing celestial fluxes from objects in a common image, only the {\em angular} variation in 

83: atmospheric transmission, across the angular separation between objects, can introduce

84: errors due to differences in atmospheric transmission. On the other hand, comparing fluxes from 

85: survey images taken at different times and through different airmasses is susceptible to both temporal and line-of-sight

86: changes in atmospheric transmission. In our view this difference accounts for the discrepancy in 

87: precision between differential photometry {\em within} one image and establishing a common 

88: photometric zeropoint {\em across} different survey images. 

89: 

90: This simple fact indicts variations in atmospheric transmission as 

91: imposing the major limitation to the 

92: precision of ground-based CCD photometry from all-sky surveys. Disentangling source brightness, 

93: instrumental response, and variable atmospheric transmission is challenging. Multiple visits

94: to each field, ideally with different instrument orientations,  

95: is certainly helpful, along with careful attention to flat-fielding and suppression of instrumental 

96: artifacts.  

97: 

98: High precision {\em differential} photometry does achieve the Poisson limit. 

99: Numerous examples of this have been

100: presented, and we summarize a few salient points from these studies. 

101: \cite{EH01} performed

102: aperture photometry on carefully flat-fielded images, and demonstrated that by comparing

103: fluxes to a robust ensemble average of flux from sources within 0.25 degrees, they 

104: could achieve Poisson-limited performance. \cite{Hartman05}  used image subtraction 

105: techniques and demonstrated differential photometry with scatter below 1 mmag. 

106: Enhanced dynamic range high precision photometry was carried out by \cite{JT05}, 

107: using an orthogonal transfer CCD. These successes all indicate that flatfielding can be done

108: at the millimagnitude level, and that Poisson-limited photometry is a worthwhile goal. 

109: Achieving uniform zeropoints across the surveyed region of the sky 

110: amounts to making a high precision set of measurements. The survey accuracy, 

111: {\em i.e.} knowing with certainty the corresponding fluxes in joules/sec/m$^2$, is far

112: less important (for nearly all astronomical applications) than attaining good precision.

113: 

114: The opportunity for achieving high-precision photometric measurements

115: in upcoming multiband sky surveys such as PanSTARRS \citep{PS02} and LSST \citep{LSST}

116: motivates a comprehensive assessment of the limitations of ground-based CCD photometry.

117:  \cite{ST06} provide a framework for this appraisal, and that paper suggests 

118: factoring the problem into (i) characterizing the wavelength-dependence of the 

119: response of the telescope, optics, and instrument, and (ii) determining the optical 

120: transmission properties of the atmosphere. This paper deals with the issue of

121: understanding and measuring the optical transmission of the atmosphere.

122:  

123: Our eventual goal is to produce the spectrum

124: of atmospheric optical transmission through which the observation

125: occurred, for each cataloged object flux from each image in the survey archive. In conjunction with detailed knowledge of instrumental sensitivity versus wavelength

126: we can then (at least in principle) perform synthetic photometry with trial Spectral Energy Distributions (SEDs) for 

127: all sources, and compare these spectral integrations with the number of counts 

128: actually detected. This ``forward modeling'' approach has been recently described for the 

129: infrared by \cite{Bailey07}. In essence we advocate using the same techniques

130: that are being applied to correcting for atmospheric absorption in spectrophotometry (e.g. \cite{Bessell99}, \cite{Stritzinger05}) and spectroscopy (e.g.~\cite{Hadrava06}) to broadband imaging data. By reporting the survey fluxes as measured in the ``natural 

131: system'' of each detector on the focal plane, along with the appropriate instrumental and 

132: atmospheric transmission profiles, we can construct an atmosphere-corrected

133: flux for each object of interest using one or more trial SEDs. This 

134: avoids having to specify color terms for the detectors, which are intrinsically

135: intermingled with the assumed spectrum of the source of interest. It also 

136: addresses the problem of second order ``color-airmass'' extinction corrections. 

137: Users who are satisfied with the more standard treatment are of course

138: still free to apply a global approximation to instrumental color terms and 

139: atmospheric extinction corrections. Knowledge of the actual optical transmission 

140: function for each measurement could also play a role in more precise K corrections for supernovae \citep{ESSENCE}.

141:  

142: It is interesting to note that in the era just before the advent of CCD's, photoelectric

143: photometrists were pushing towards millimagnitude photometry. In this context we

144: draw particular attention to the remarkable paper by Young and collaborators 

145: \citep{Young}, that raises many of the issues we address here.  

146: It would appear that the advent of CCD detectors stalled this initiative. Now that

147: CCD performance and flatfielding techniques have advanced, it is time to 

148: revisit the issues that stand between the current state of the art and the 

149: fundamental limit imposed by Poisson statistics. 

150: 

151: One important difference between our philosophy and that described in \cite{Young}

152: is that those authors considered a single photometric instrument for both atmospheric characterization 

153: and celestial flux measurements, whereas we intend to evaluate the optimal choice of 

154: {\em dedicated} instrumentation for atmospheric characterization. 

155: Given the choice between using their allocated telescope time to either fully characterize 

156: atmospheric extinction or to observe their program objects, most astronomers have opted for the latter. 

157: The SDSS project did 

158: include a separate ``Photometric Monitoring Telescope" that was equipped with nominally the same

159: filter set as the 2.5m survey telescope, but filter passband and detector QE differences between the

160: main survey system and the monitoring telescope gave rise to color terms that 

161: prevented this approach from reaching its full potential \citep{SDSSpt}.

162: 

163: The Pierre Auger Observatory measures \v{C}erenkov light from high energy cosmic ray showers in the atmosphere. Molecules and aerosols scatter and absorb this signal on the path to the detectors. Proper interpretation of the detected light requires a knowledge of the optical properties of the atmosphere, and this has led the team to establish a sophisticated set of instruments, including LIDAR and photometric monitors to measure the aerosol content and scattering properties of the local atmosphere (\cite{Auger1}, \cite{Auger2}). 

164: 

165: The photometric calibration plans for PanSTARRS \citep{MagnierPS} and LSST \citep{Burke06}

166: both include apparatus dedicated to the determination of atmospheric extinction, to allow

167: the wide field survey imagers to focus on science images of the sky without needing to 

168: allocate time to atmospheric characterization. 

169: 

170: The question then becomes: What is the best method to determine atmospheric extinction, 

171: using dedicated apparatus that runs in conjunction with a broadband all-sky survey, 

172: and what cadence of observations is needed to capture the temporal and directional

173: variations in extinction?

174: 

175: \subsection{Some Formalism} 

176:  

177: The total photon flux ($S(t)$, photons per sec)  we detect from some celestial source is an integral over wavelength,

178: 

179: \begin{equation}

180: S(t) =  \int \Fscr(\lambda, t)

181:    \times A \times T(\lambda, \hat{z}, t)\times  I(\lambda, t) ~ d\lambda,

182: \label{eq:psignal}

183: \end{equation}

184: 

185: \noindent

186: where the $\Fscr$ is the spectral photon distributions (SPDs, evaluated

187: above the atmosphere) of the sky and all sources present,

188: $I(\lambda,t)$ is the dimensionless instrumental transmission, 

189: including optics, filter, and detector, $A$ is the effective collecting area of the system, 

190: and   

191: $T(\lambda, \hat{z}, t)$ is the wavelength dependent atmospheric transmission along the direction $\hat{z}$ at the

192: time $t$ of the measurement, 

193: averaged over the exposure time.  (We adopt the convention here that the units of $\Fscr$

194: are photons/nm/sec/m$^2$, derived from the conventional spectral

195: energy distribution as $\Fscr = F_\lambda\;\lambda/hc$.)  

196: 

197: We have described elsewhere \citep{CTIOlaser} a program of measuring explicitly $I(\lambda,t)$, the

198: wavelength dependence of the instrumental response. Our focus in this paper is the 

199: determination of $T(\lambda, \hat{z}, t)$, the time-dependent and 

200: line-of-sight dependent optical transmission of the atmosphere.

201: 

202: The claim that observing conditions are ``photometric'' during a night amounts to assuming that 

203: $T(\lambda, \hat{z},t)=T(\lambda, \theta_{zenith})$, {\it i.e.~}time-independent and axisymmetric,  

204: depending only on zenith angle.  Furthermore, the typical corrections for 

205: atmospheric extinction assume a simple dependence in which all magnitudes

206: are corrected by a band-specific extinction coefficient $k$ that is multiplied by the secant of the 

207: zenith angle. The extinction coefficient for each passband is taken as a universal 

208: number, independent of the source's spectrum. As shown below, these standard photometric approximations will fail to produce 

209: the accuracy we propose to attain. Furthermore, the regression of all magnitudes to 

210: zero airmass (above the atmosphere) introduces a strong leverage on any uncertainty in the extinction coefficients,  as we are extrapolating into a regime that is not directly observable from 

211: the ground.  

212: 

213: \cite{Wade88} present a useful formulation of atmospheric 

214: transmission, which we will generalize to make clear the distinct

215: processes at work, and any potential time dependence.   

216: At any wavelength the optical transmission of the atmosphere can be represented as

217: 

218: \begin{equation}

219: T(\lambda,\hat{z},t)=exp\left [~\sum_i(-\tau_i(\lambda,t,\hat{z})(\chi^{\alpha_i}(\hat{z}))\right ],

220: \end{equation}

221: 

222: \noindent

223: where $\tau_i$ is the optical depth (at one airmass) for each attenuation process, $\chi$ is the airmass

224: along the line of sight with unit vector $\hat{z}$, and $\alpha_i$ is the airmass dependence

225: of the attenuation process.  For spectral regions where the attenuation processes are unsaturated 

226: we would expect $\alpha_{unsat}=1$ whereas for saturated lines we would expect the 

227: absorption from the wings of the line to scale with $\alpha_{sat}=0.5$. 

228: Atmospheric water has numerous narrow molecular absorption features; 

229: some are saturated and some are not. \cite{Wade88} suggest that this 

230: produces an effective $\alpha_{water}=0.6$.  We should expect similar 

231: behavior from other complex molecular absorption band structures. 

232:  

233: The direct determination of each component of atmospheric attenuation, ideally co-boresighted and 

234: simultaneous with the survey imager, would help address our currently inadequate correction for  

235: the atmosphere. Our goal in this paper is to explore alternative approaches to this 

236: measurement problem, and to suggest measurements that would help illuminate a 

237: shrewd choice of technology.    

238: 

239: Measurements of the optical transmission properties of the atmosphere can exploit either

240: natural celestial sources to backlight the atmosphere, or manmade illumination which is 

241: observed through backscatter. (\cite{Albert06} have undertaken a program to use

242: ground-based observations of artificial sources on satellites, but we will not explore

243: that option in this paper.) 

244: 

245: The resulting atmospheric 

246: transmission measurements can then be used in isolation, 

247: or in conjunction with detailed computer models such as MODTRAN \citep{2001SPIE.4381..455A}. We suggest that 

248: it makes sense to fully exploit the comprehensive knowledge of atmospheric physics that

249: has been incorporated into these sophisticated codes. 

250: 

251: In the sections that follow we review the physics of optical transmission through the 

252: atmosphere, and then we discuss in turn merits of spectroscopic, photometric and LIDAR

253: measurements for characterizing extinction. 

254: We then describe how the combination of atmospheric transmission 

255: functions and computer modeling could be exploited in a forward modeling approach to astronomical photometry, and 

256: we close with a suggested set of next steps. 

257: 

258: \section{ATTENUATION BY THE ATMOSPHERE AT OPTICAL WAVELENGTHS, AND ITS

259: VARIABILITY}

260: 

261: \subsection{The Final Four for a Photon: Rayleigh Scattering, Molecular Absorption, Aerosols and Clouds}

262: 

263: As light propagates to us from celestial sources it experiences

264: numerous opportunities to interact with material. Extinction at the source, 

265: through the Galaxy, and in other intervening regions can all play a role in 

266: distorting the spectrum. From the standpoint of this paper, these astrophysical

267: processes can be either sources of information or a nuisance, and we will

268: not address them here. Our concern is the astrophysically uninteresting 

269: attenuation from the final four hurdles faced by an astronomical photon:

270: scattering and absorption from atoms and molecules in the atmosphere, 

271: scattering from suspended aerosols, and extinction due to clouds.  

272: 

273: Figure ~\ref{fig:photon_transmission} shows the expected attenuation of flux due to the

274: atmosphere, alongside the filter passbands for the PanSTARRS survey, and

275: illustrates where the different components of extinction will afflict our measurements. 

276: 

277: Table~\ref{tab:smooth} provides an estimate of the integrated attenuation we expect 

278: within each survey passband, from the different components of 

279: atmospheric attenuation. This calculation presumes a source that has a 

280: photon spectrum $\Fscr(\lambda) \propto \lambda^{0}$, which for our spectral region

281: of interest approximately corresponds to a blackbody of 4500~K. We integrated

282: the different components of $T(\lambda)$ separately in order to 

283: gain some intuition about the relative importance of scattering vs.~absorption.

284: 

285: \begin{table}[htdp]

286: \caption{Estimated transmission at one airmass from the smoothly varying 

287: components (Rayleigh scattering and aerosols) and absorption from molecular lines, 

288: for the PanSTARRS system. 

289:  Note that the airmass dependence 

290: $\alpha$ of these can be quite different.

291: We have included the anticipated effects of 

292: two Al reflections, 

293: of filter transmission, and of detector QE in the system's response 

294: functions. These numbers presume

295: a source with a constant photon flux per nm. The effects of molecular 

296: absorption and aerosols are comparable for $i, z$ and $y$.}

297: \begin{center}

298: \begin{tabular}{cccc}

299: \hline

300: Band & T$_{smooth}$ & T$_{lines}$ & Total \\

301: \hline

302: PanSTARRS $g$ & 0.815 & 1.000 & 0.815 \\

303: PanSTARRS $r$ &0.894 & 0.996 & 0.890 \\

304: PanSTARRS $i$ & 0.949 & 0.961 & 0.912\\

305: PanSTARRS $z$ & 0.964 &0.970  & 0.935 \\

306: PanSTARRS $y$ & 0.961 & 0.947  & 0.910\\

307: \hline

308: \end{tabular}

309: \end{center}

310: \label{tab:smooth}

311: \end{table}%

312: 

313: The next step is to assess how different source spectra 

314: behave as they propagate through the atmosphere and the 

315: instrumental response function. 

316: Since it is awkward to obtain high precision spectrophotometric data 

317: in the attenuated spectral regions of interest we used 

318: theoretical models of stellar spectra for computing the effects of the atmosphere. 

319: We used the theoretical spectra from \cite{Kurucz} to compute the 

320: flux attenuation due to the atmosphere by performing synthetic 

321: photometry across the different $I(\lambda)$ system response 

322: functions we expect for PanSTARRS. We obtained Kurucz's modelled

323: stellar spectra from 

324: {\tt http://www.stsci.edu/hst/observatory/cdbs/astronomical\_catalogs.html}.

325: These STSCI stellar atlas data are 

326: $F_{\lambda}(\lambda)$ tabulations, which we converted into photon 

327: spectral distributions $\Fscr(\lambda)=F_{\lambda} \times \lambda$, with an 

328: arbitrary common multiplicative normalization.

329: 

330: Using a spectral resolution of $\Delta \lambda =~$0.1~nm we integrated

331: the source spectra shown in Table~\ref{tab:extinct_vs_spect} through the transmission function

332: that corresponds to $\chi=~$1 airmass.  

333:  The {\it error} that would be incurred by making the usual assumption that the 

334: extinction (in magnitudes) scales linearly in airmass for all attenuation processes 

335: is roughly 0.01, 0.02 and 0.03 magnitudes per airmass in the $r,i, $ and $z$

336: bands, respectively. 

337: 

338: \begin{table}[htdp]

339: \caption{Calculated extinction coefficients for PanSTARRS 

340: passbands, for different astronomical objects. The first column lists 

341: the template spectrum used, the subsequent columns list the computed

342: extinction coefficient, in magnitudes per airmass. The interplay

343: between color and extinction is most pronounced in the bluer

344: bands. The difference in spectral weighting accounts for the 

345: difference between these values and those of the previous

346: Table.}

347: \begin{center}

348: \begin{tabular}{lccccc}

349: \hline

350: Source & $k_g$ & $k_r$ & $k_i$ & $k_z$ & $k_y $\\

351: \hline

352: O5V & 0.23 & 0.13 & 0.08 & 0.06 & 0.07 \\

353: B5V  & 0.23 &0.13 & 0.08 & 0.06 & 0.07 \\

354: A3V  & 0.23 & 0.13 & 0.08 & 0.06 & 0.07\\

355: F5V  & 0.22 & 0.12 &0.08  & 0.06 & 0.07 \\

356: G3V  &0.21 & 0.12 & 0.08  & 0.06 & 0.07\\

357: K4V & 0.20 & 0.12 & 0.08 & 0.06 & 0.07 \\

358: M2V  & 0.20 & 0.12 & 0.08 & 0.06 & 0.07\\

359: \hline

360: \end{tabular}

361: \end{center}

362: \label{tab:extinct_vs_spect}

363: \end{table}%

364: 

365: We drew two conclusions from this exercise: 1) we cannot ignore the interplay

366: between extinction and the source spectrum, and 2) we need to properly account

367: for the different airmass scaling for line absorption vs. scattering.

368: 

369: A number of studies have explored the stability of extinction at astronomical sites, 

370: including \cite{KrisciunasMK}, \cite{Riemann92}, \cite{Frogel}, \cite{SanPedro}, 

371: \cite{DeVauc73} and \cite{SDSSpt}.

372: There are numerous reasons to expect variability in atmospheric extinction. 

373: We should expect a variation in aerosol and water content 

374: with the direction of prevailing winds and with local meterology.

375: The solar ``weather'' influences the ozone fraction in the upper atmosphere.  

376: Some of these dependencies are nontrivial. One such example is the apparent

377: dependence (\cite{Pakstiene}, \cite{Riemann92}, \cite{Roosen77}) 

378: of extinction on absolute humidity, even in bluer spectral regions that

379: are well away from H$_2$O line absorption, allegedly due to water absorption by 

380: aerosols changing their sizes and hence their scattering properties. 

381: The dependence of extinction on meteorology, and over time, has been explored by \cite{Riemann92}, 

382: \cite{Frogel}  and \cite{Pakstiene}. 

383: 

384: Table~\ref{tab:ctio} shows a summary of weather statistics for a 4 year period at CTIO. We will take these

385: values as being representative of weather variability at good astronomical sites around the world. 

386: The data were obtained from the online archive at\\ 

387: {\tt http://www.soartelescope.org/release/02about/eng$\_$about/weather/main$\_$weather.html}

388: 

389: \begin{table}[htdp]

390: \caption{Summary of Weather at CTIO, Jan 1 2001 to Jan 2 2005. The table lists 

391: percentiles for the distribution of the relevant meteorological parameter. We make the conservative

392: assumption that the time actually spent observing follows this overall pattern. }

393: \begin{center}

394: \begin{tabular}{cccccc}

395: \hline

396: ~&10th&25th&50th&75th&90th\\

397: \hline

398: Temperature (C)& 6.7& 11.9& 15.2& 17.4& 18.9\\

399: Pressure (mbar)& 781&782&784&785&786\\

400: Rel Humidity (\%)&8&13&23&38&55\\

401: Wind Direction (Degrees) & 32& 60& 107& 238& 313\\

402: Wind Speed (m/s)&  0.675& 1.62& 3.195& 5.85& 8.595\\

403: \hline

404: 

405: \end{tabular}

406: \end{center}

407: \label{tab:ctio}

408: \end{table}%

409: 

410: \subsection{Molecular  scattering}

411: 

412: Elastic scattering \citep{Rayleigh} from atoms and molecules in the air

413: has a cross section that varies as $\sigma_{Ray}(\lambda) \sim \lambda^{-4}$, 

414: with a slight correction that arises from wavelength dependence of the index of refraction.

415: \cite{Hansen} give the optical depth (at zenith) for Rayleigh scattering to be 

416: $$

417: \tau_{Rayleigh} =  0.008569 \lambda^{-4}\left (1+0.0113 \lambda^{-2} + 0.00013 \lambda ^ {-4}\right ) \times \left [\frac{P}{1013.25~mb} \right],

418: $$

419: \noindent

420: where $\lambda$ is in microns.  This process dominates over inelastic scattering, such that the 

421: inelastic contribution to attenuation is negligible.

422: 

423: The attenuation due to Rayleigh scattering thus depends on 

424: the pressure of the atmosphere along the line of sight. 

425: Barometric pressure varies with a typical timescale

426: of days, and will fluctuate as high and low

427: pressure systems pass over an observatory. This changes the effective airmass 

428: along a fixed line of sight, in proportion to the pressure fluctuation.  

429: A measurement taken at CTIO at an airmass $\chi$ under a pressure that is 

430: $\delta P$ different from the nominal $P_o=$784 nm will 

431: suffer a change in transmission of $\delta T_{Rayleigh} = 6 \times 10^{-3}~\chi~\left(\frac{\delta P}{P_o}\right)\left(\frac{1~\mu m}{\lambda}\right)^4$.

432: A variation in attenuation of less than one millimagnitude 

433: therefore requires $\delta T_{Rayleigh} < 0.001$ or $\chi~\left(\frac{\delta P}{P_o}\right)\left(\frac{1~\mu m}{\lambda}\right)^4~<~0.17$. Using the 90$^{th}$ percentile 

434: pressure excursion for CTIO, an {\it uncorrected} Rayleigh extinction perturbation under 1 millimagnitude at  $\lambda=$500 nm, in the $g$ band,  corresponds to 

435: restricting $\chi < $ 2.6 airmasses during a large pressure excursion.

436: 

437: It is however very tractable to calculate how changes in local barometric pressure 

438: would introduce  

439: an airmass-dependent shift in the Rayleigh transmission. 

440: Millimagnitude photometry will

441: require a pressure-dependent correction to attenuation from Rayleigh scattering 

442: only for very blue bands observed at airmasses around $\chi\sim3$. We are also fortunate that conditions

443: with very low barometric pressure are often accompanied by weather that precludes opening the dome. 

444: The Rayleigh scattering should 

445: be axisymmetric about the zenith with a time dependence driven only by pressure

446: variations. 

447: 

448: The characteristic spectral dependence and spatial symmetry of extinction by molecular scattering can be combined with independent measurements of barometric pressure and airmass to provide precise determination of extinction due this process, subject

449: to observational confirmation. 

450: 

451: \subsection{Aerosols}

452: 

453: Scattering from aerosols and particulates in the atmosphere is in the awkward 

454: ``Mie'' regime where the particulate size is comparable to the wavelength. This 

455: gives rise to a cross section $\sigma_{Mie}(\lambda)  \sim \lambda^{k}$ where $k$ 

456: depends on the size and shape distributions of the scattering particles. Since aerosol scattering is in the optically thin regime for all 

457: wavelengths, we expect an airmass exponent of $\alpha_{aerosol}=1$. \cite{Sinyuk07}

458: claim that most aerosols reside within $\sim $ 4 km of the Earth's surface.

459: 

460: There is ample evidence for volcanic eruptions producing long term change in the optical 

461: transmission of the atmosphere. These global events have 

462: more local counterparts, due to 

463: changes in wind direction, to regional forest fires, to lofted marine salts, etc. This implies 

464: that the transmission spectrum $T_{aerosol}(\lambda, \hat{z},t)$ will have both time dependence 

465: and azimuth dependence. 

466: 

467: \cite{Holben99} present data for both atmospheric water content and the optical depth due to aerosols, obtained from the AERONET system, 

468: at Mauna Loa. We strongly suspect this is a good proxy for the aerosol

469: characteristics we expect on Haleakala, where PanSTARRS-1 is situated.  

470: The AERONET instrument measures the scattering of solar

471: radiation as a function of angle away from the sun, in multiple bands. 

472: Although the observations in the different bands are not strictly simultaneous, 

473: the cycle time through the filter set is only 8 seconds. They

474: report values of optical depth $\tau$ for $\lambda=$ 340, 380, 440, 500, 675, 870, 940 and 1020 nm. 

475: They use the 940~nm channel for measuring water content, by comparing the 

476: solar flux seen there to that of adjacent bands. 

477: 

478: The AERONET Level 2 data have been selected to avoid days with obvious

479: clouds, and the data are processed in order to extract the various

480: components of attenuation, as described in \cite{Holben99}. 

481: In the descriptions that follow

482: we used the Level 2.0 AERONET data, obtained from {$\tt http://aeronet.gsfc.nasa.gov/new\_web/data.html$}. 

483: 

484: Figure~\ref{fig:t440_full} shows a 15 year record of aerosol optical depth at

485: Mauna Loa, at 440 nm. 

486: Figure~\ref{fig:depth_zoom} is an expanded view of a period spanning 0.2 years, and 

487: shows the aerosol optical depth fluctuations at 1020 nm, 870 nm, 675 nm, 

488: and 440 nm. Figures~\ref{fig:quad1} and~\ref{fig:quad2} show that different ``extinction spikes''

489: have different spectral character. Some extinction spikes are changes in 

490: transparency, with little wavelength dependence, while others clearly 

491: exhibit more attenuation at the shorter wavelengths. Figure~\ref{fig:cumulative_aerosol}

492: shows the cumulative aerosol optical depth distributions for 

493: the period of time shown in Figure~\ref{fig:t440_full}. The median aerosol optical 

494: depth values at 1020 nm, 440 nm and 340 nm are 0.007, 0.015, and 

495: 0.018 respectively. Figure~\ref{fig:cumulative_deriv} shows the cumulative

496: distribution of variation in the AERONET aerosol optical depth measurements.

497: 

498: Measuring the extinction due to aerosols is one of our main challenges.  

499: It may be that daytime measurements, of the kind reported by \cite{Holben99}, 

500: could be used to estimate night-time aerosol scattering, but we need to explore 

501: the extent to which this is true. 

502: Most aerosols are low in the atmosphere. This, and the observed

503: short-term variation in aerosol optical depth, implies that we should expect

504: azimuth dependence of the aerosol transmission across the sky. 

505: 

506: \subsection{Molecular Absorption}

507: 

508: Absorption features in the spectra of atmospheric constituents produce a 

509: complex set of absorption bands and features. Ozone, water, O$_2$ and OH molecules

510: all contribute to this absorption. In fact, the atmosphere is essentially opaque in numerous

511: narrow regions, 

512: especially for $\lambda >$ 740 nm. 

513: A high spectral resolution determination of the molecular

514: atmospheric absorption above KPNO is presented in \cite{KPNO}. 

515: 

516: \subsubsection{Oxygen}

517: 

518: The strong absorption features at 690~nm and 760~nm due to O$_2$, the ``B'' and ``A'' bands in the nomenclature of Fraunhofer, 

519: are stable over time since they depend on the integrated Oxygen

520: content along the line of sight. The same arguments given above about the pressure dependence of

521: Rayleigh scattering apply here as well.  We therefore expect this component of atmospheric transmission, $T_{O_2}$,  to be axisymmetric about the zenith and stable over time, again with a slight pressure dependence. 

522: 

523: This picture is borne out by inspecting the corrections applied to spectrophometric data by \cite{Bessell99}. Figures 

524: 1 and 2 of that paper clearly show much smaller residuals in the A and B bands as compared to water absorption regions.  

525: 

526: \subsubsection{Ozone}

527: 

528: The opacity of ozone is responsible for the total loss of atmospheric transmission below 300nm. The Chappuis band of ozone influences transmission for 500~nm~$< \lambda <$~700~nm, 

529: with an attenuation of a few percent at 600~nm. 

530: The measurement of ozone in the atmosphere is of great interest due to 

531: its important role in the Earth's radiation balance and climate change. 

532: This has led to the development and deployment of 

533: sophisticated space based instruments that are optimized for the 

534: determination of the ozone content of the atmosphere. These data

535: sets can be used to determine the ozone above observatories. 

536: Satellite measurements of the ozone content versus time over Hilo, HI

537: are shown in Figure~\ref{fig:ozone}. These illustrative data were obtained from the

538: TOMS instrument \citep{Jaross03} 

539: on board the EarthProbe satellite, and we obtained the 

540: data for Figure~\ref{fig:ozone} from {\tt http://toms.gsfc.nasa.gov/eptoms/ep\_ovplist\_a.html}.

541: 

542: The determination of ozone content in the atmosphere above the 

543: observatories can be obtained from satellite remote sensing data

544: with sufficient accuracy and temporal resolution for our purposes. 

545: Since the attenuation due to ozone is only a few percent, we need

546: only know the ozone content with a fractional precision of around ten percent.

547: \cite{Eck99} state that a 50\% change in ozone content would perturb

548: the optical transmission by 0.0036, 0.0045 and 0.0063 at 340 nm, 

549: 500 nm and 657 nm respectively. 

550:  

551: We expect therefore to be able to exploit data obtained from satellites to constrain nightly variations in attenuation due to ozone, 

552: verified by direct measurements of atmospheric throughput on and off the well-known spectral regions of ozone attenuation.

553: 

554: %\cite{Griffin05} used archival spectral to extract equivalent widths of 

555: %ozone absorption features for $\lambda < $350nm.  

556: 

557: \subsubsection{Water}

558: 

559: \cite{Frogel} presents evidence for variation of atmospheric attenuation due to water, measured in the IR, and shows month-to-month and longer term variability. \cite{Roosen77}  show that surface humidity

560: measurements only poorly correlate with the upper atmospheric water burden. Attempts \citep{Bessell99}

561: to make a statistical correction to spectra for the absorption due to water

562: have significant residuals, compared to oxygen A and B bands, which is evidence for 

563: significant variability. The PanSTARRS

564: filter set includes a $y$ band that takes advantage of the enhanced red 

565: response of high resistivity CCDs, and the break between the 

566: PanSTARRS $z$ and $y$ bands has been selected

567: to avoid the large water absorption feature at 950 nm. Nevertheless there are 

568: H$_2$O absorption features, as shown in Figure~\ref{fig:photon_transmission}, that impact other spectral regions. 

569: 

570: Figure~\ref{fig:cumulative_water} is a cumulative distribution of the measured water content

571: in the atmosphere over a ten year period, from the AERONET 

572: station on Mauna Loa. The units used are cm of precipitable water. 

573: Figure~\ref{fig:watervt} shows the temporal evolution of the water content over the 

574: same period shown for the aerosols in Figure~\ref{fig:t440_full}. 

575: 

576: \cite{Campanas07} present an interesting comparison of measurements of atmospheric water

577: content using both 255 GHz radiometery and high resolution optical spectroscopy. Their paper

578: contains a valuable review of the physics of optical absorption by water vapor, and 

579: demonstrates good agreement between the results obtained with these two techniques. 

580: They also show data with clear indications of short term variability in 

581: atmospheric water, over the course a night. The use of RF radiometry for the 

582: determination of water absorption in the optical clearly merits further consideration.

583: 

584: The evidence indicates that we should expect substantial temporal fluctuations in the optical 

585: absorption due to water, and this implies a potential azimuth dependence as well. 

586: We therefore conclude that optical attenuation due to water in the atmosphere is our second 

587: main challenge, with a complex behavior of both $\alpha_{water}(\lambda)$ and $T_{water}(\lambda, \hat{z}, t)$. 

588: 

589: \subsubsection{Clouds}

590: 

591: Water droplets and ice crystals also attenuate the transmission of

592: optical radiation through the atmosphere. The standard assumption is that these

593: objects that make up clouds are sufficiently larger than the wavelengths in question that

594: the scattering is wavelength independent, so that clouds are ``grey'' scatterers. 

595: \cite{EH01} achieved Poisson limited performance in differential photometry through high cirrus, lending

596: credence to the idea that clouds are grey absorbers. 

597: Clouds can be detected through their emission (e.~g.~with a camera operating in the 10 micron band), 

598: by optical attenuation (through their effect on survey photometry), or in reflection

599: (using LIDAR, see \cite{Auger1}). 

600: 

601: A thorough examination by \cite{Ivezic07} of drift scanned

602: SDSS data from repeated imaging of stripe 82 has also shown little dependence of 

603: photometric residuals with color, even through many magnitudes of extinction from clouds, and even when comparing $u$ and $z$ band fluxes. 

604: These authors place an upper limit of 0.02 mag of zeropoint uncertainty for photometry obtained

605: through one magnitude of extinction from clouds. It is important to distinguish

606: ``astronomical point source'' transmission through clouds, which we take to be the 

607: unscattered component, from the net radiation transfer of flux from the sun, 

608: as described in \cite{Kylling97}.

609: 

610: One might imagine that a survey would need at least one visit to each field under 

611: highly photometric conditions in order to achieve a uniform all-sky zeropoint in each 

612: passband. The SDSS experience suggests that the location of the stellar locus, in color-color

613: space, might be used to make zeropoint corrections for those bands that were imaged

614: through clouds. This of course assumes that the stellar locus has no dependence on Galactic

615: coordinates. 

616: 

617: PanSTARRS is planning to use a modest aperture photometric monitor that places

618: the entire survey field of view onto a single CCD array. This imager will be used to monitor the 

619: flux from Tycho catalog stars in the field \citep{MagnierPS}. Using these objects as flux references will 

620: allow for the detection of even small amounts of grey extinction, and scatter in the 

621: flux differences across the field will be indicative of variable cloud across the field of view. 

622: 

623: Contrails from aircraft are a nuisance that the astronomical community has not really faced

624: up to. These could in principle be detectable by searching for residuals in the stellar locus

625: that are linear across a frame, and excising or correcting the afflicted data.  While we are not aware

626: of any project that has done this, it should be straightforward. An approach akin to the

627: excision of satellite tracks could be implemented, for example, based upon

628: photometric residuals that lie along a common line in an image.

629: 

630: \section{THE ANGULAR AND TEMPORAL CORRELATION FUNCTIONS OF ATMOSPHERIC TRANSMISSION}

631: 

632: The different contributions to atmospheric optical attenuation have 

633: different expected azimuth, zenith angle and time dependencies.  

634: Short term (during the night)

635: time variability must necessarily be accompanied by variation in the azimuth and zenith

636: angle dependence. The converse is not necessarily true, as we can imagine

637: non-axisymmetric configurations of the atmosphere that are stable over a night. 

638: 

639: \cite{Anderson03} have published the two-point correlation functions in space and time

640: for aerosol scattering, but their data were not necessarily representative of an astronomical site. 

641: Nevertheless, they show evidence for the aerosol scattering component being highly 

642: correlated (correlation coefficient $r>0.8$) over time scales of many hours, and length scales of 

643: tens of km. 

644: 

645: We took the time series AERONET data and determined the rate of change

646: of aerosol optical depth at 1020 nm and 440 nm. Figure~\ref{fig:cumulative_deriv} shows the cumulative distribution of the rate of change of the measured attenuation, 

647: in units of (optical depth change) per hour. 

648: It would appear that hourly measurements would have optical depth 

649: changes of under 0.01/hr for more than 90\% of the time, assuming

650: nighttime variation is similar to the daytime AERONET data set. Taking 

651: 10 m/s as a typical wind

652: speed at the typical aerosol height of z=4 km, in one hour the atmosphere will have translated by

653: 36 km.  This implies that the aerosol burden is correlated over at least

654: this angular scale. We can therefore be optimistic that the aerosol attenuation

655: will be correlated across the sky, for an hour at a time. 

656: 

657: A similar analysis for the rate of change of water is shown in 

658: Figure~\ref{fig:cum_water_deriv}. This appears more demanding. 

659: The water column can change by 10\% in less than an hour, implying

660: an azimuth variation as well. Figure~\ref{fig:watervt} clearly shows

661: as much as a factor of 4 variation in water content of the atmosphere at Mauna Loa. 

662: Since we would expect the optical depth due to water to scale as the square root

663: of the water content in the air, this implies a factor of two fluctuation in optical

664: attenuation due to water. Referring back to Table 1 we note that water

665: absorption accounts for 3\% and 5\% attenuation in the $z$ and $y$ bands,  

666: respectively, at the zenith. This in turn implies that we should expect 1-2\% variations in 

667: zeropoint in these bands due to atmospheric  water content variation. This effect will 

668: of course increase with zenith angle.  

669:   

670: An important contribution by \cite{Ivezic07} is their measurement of the structure function of clouds. These authors state that the zeropoint offset between two regions

671: is proportional to both the angular separation $\theta$ and the cloud's 

672: extinction $A$, with $\Delta m \sim (0.02$ to $0.10) \times A \times \left(\theta/1^o\right)$. 

673: This implies that local relative photometry can be extracted through clouds

674: over surprisingly large areas, in agreement with the results 

675: from \cite{EH01}.  

676: 

677: \subsection{The implications of a wide field of view, and aperture considerations}

678: 

679: Much of the methodology currently used for correcting for atmospheric

680: extinction is carried forward from the era of single-pixel detectors, but

681: a distinguishing feature of the next generation of sky surveys is their 

682: wide field of view and Gigapixel CCD arrays. 

683: The PanSTARRS and LSST imagers will span 3.0 and 3.5

684: degrees, from corner to corner. Even at a modest zenith angle of 60 degrees, 

685: the span in airmass across the LSST imager will range from 1.9 to 2.1! 

686: Taking the extinction values from Table 2, the {\em differential} extinction across the field will be 0.10, 0.04, 0.02,  and 0.01 magnitudes

687: in $u,g,r,i $ and $z$, from this effect alone, at $\chi=$2 airmasses.

688: 

689: Our measurements are under the influence of attenuation processes 

690: that exhibit different airmass dependences, so we cannot simply assign 

691: a single effective atmospheric transmission curve to each survey image, 

692: but instead we should calculate what amounts to a wavelength dependent

693: illumination correction across the field of view. 

694: 

695: The atmospheric column traversed by the light from different objects depends

696: upon their separation and on the aperture of the telescope. 

697: Using the same arithmetic that applies to multi-conjugate

698: adaptive optics, the beams from two sources separated by an angle $\theta$ traverse distinct regions of the upper atmosphere only above a height $h>D/sin(\theta)$, 

699: where $D$ is the telescope aperture. This means that for LSST, with $D=8.5$~m, 

700: the atmospheric transmission below $h=15$ km is ``common mode'' for objects 

701: closer than 1 arcminute on the sky.  Aerosol attenuation arising from 

702: a static layer at z$\sim$4 km would be common over a focal plane region spanning 3 arcmin on LSST, and  0.6 arcmin for PanSTARRS.

703: Local atmospheric perturbations in the first few

704: hundred meters above the observatory are common mode across the entire LSST

705: field. The dynamics of the atmosphere greatly suppress possible variation 

706: across the field of view, however. A 10 m/s wind will drag hundreds of meters

707: of atmosphere across the telescope aperture over a typical survey exposure time of 

708: 15 or 30 seconds. This serves to homogenize the time-averaged atmospheric

709: profile through which the system is imaging, at least in one dimension, across the entire field of view.

710: 

711: \subsection{Measurement Priorities}

712: 

713: Based on the above considerations, we can expect the Rayleigh and 

714: O$_2$ components of atmospheric transmission to be well behaved, 

715: with minimal variation from night to night and little azimuth dependence. 

716: The spectral shape of these mechanisms is also essentially 

717: time independent. The aerosol term, on the other hand, clearly 

718: shows evidence for temporal variation and spectral evolution, and 

719: is difficult to measure. Water absorption has a fixed spectral profile but

720: its strength is highly time variable. Clouds are grey absorbers at the

721: 1\% level, with significant temporal variation. Ozone absorption is time

722: variable but can be measured from satellites. 

723: 

724: Our priorities are therefore the determination of aerosol scattering and

725: absorption due to water molecules. 

726:   

727: \section{SPECTROSCOPIC MEASUREMENTS OF ATMOSPHERIC TRANSMISSION}

728: 

729: There is a long tradition in astronomy of using spectrophotometry to

730: determine the spectral energy distribution of celestial sources of

731: interest. An integral part of the analysis of these data is to

732: compensate for atmospheric attenuation. This naturally provides an

733: opportunity to determine the wavelength dependence of atmospheric

734: transmission, within the spectral resolution limitations of the

735: apparatus. \cite{Bessell99} devised a technique to disentangle the

736: spectral structure of the atmosphere from that of the celestial

737: source. \cite{Stritzinger05} used a combination of their

738: spectroscopic observations and stellar models to produce a tabulation

739: of extinction values versus wavelength, derived from their measurements

740: of spectrophotometric standard stars. We caution, however, that their

741: tabulated extinction values do not include molecular absorption

742: effects.  We can consider the smooth extinction data of

743: \cite[][; their Fig.~2]{Stritzinger05} to be a good complement to the

744: high spectral resolution absorption atlas of \cite{KPNO}.

745:  

746: In principle, high precision spectrophotometry seems like the ideal

747: way to measure the atmospheric attenuation profile. Measuring the

748: Rayleigh, aerosol and ozone attenuation does not require much spectral

749: resolution (on the order of $\sim1$\,nm), but does benefit from a large

750: spectral range (300\,nm to 1\,$\mu$m) and requires high precision

751: spectrophotometry. The strength of the molecular absorption features

752: can be expressed in equivalent widths (as commonly done amongst

753: stellar spectroscopists), and since this is a spectrally local

754: differential measurement, broadband flux precision and knowledge

755: of the source spectrum are less of a

756: concern.

757: 

758: One major advantage of the spectroscopic approach is that the

759: measurements at all wavelengths are made simultaneously. This helps

760: distinguish spatial from temporal variation in transmission.  

761: 

762: There are two possible approaches to using spectroscopy for the

763: determination of atmospheric extinction. If the spectrum of the source

764: and the wavelength dependence of the instrumental sensitivity are well

765: known, a single measurement of the spectrum of the source suffices to

766: determine the atmospheric attenuation vs. wavelength. On the other

767: hand, measurements obtained at a diversity of airmasses can be used to 

768: distinguish between the airmass-independent aspects (source spectrum and

769: instrumental throughput) and the airmass-dependent atmospheric

770: attenuation, liberating us from having to know the source

771: spectrum. This second approach of course is susceptible to systematic

772: errors due to any time dependence of the atmospheric extinction. 

773: 

774: Drawbacks to the spectroscopic approach include the loss of signal to

775: noise, compared to the imaging techniques described below, due to

776: dispersion. There are also non-trivial instrumental

777: challenges. Grating dispersion elements have a large variation in

778: throughput with wavelength, and maintaining a high signal to noise ratio across a wide

779: spectral range is difficult due to limited CCD response at the limits

780: of the targeted wavelength range. Prisms have higher dispersion in the

781: blue than the red, which is the inverse of what we'd like to have, 

782: although this is in part compensated by the loose constraint on

783: spectral resolution.

784: 

785: A slitless spectrograph is essential for high precision measurements

786: in the blue at high airmass, but guiding errors can produce a

787: systematic offset between the wavelength solution from arc lamps and

788: the actual spectrum of the object. Nevertheless, the object

789: spectrum itself can be used to determine a wavelength solution

790: through identification of lines commonly found in stellar spectra

791: (such as the Balmer series of hydrogen).

792: 

793: Extracting information about the distinct attenuation processes

794: described above invariably involves comparing the fluxes in certain

795: spectral regions, and (for the absorption lines) determining

796: equivalent widths. This means the spectrometer is essentially being

797: used as a simultaneous narrowband imager. 

798: 

799: We have begun observing campaigns at Haleakala and CTIO 

800: designed to address these issues.

801: The first steps have focused on assessing our ability to detect the signatures of the elements

802: of atmospheric extinction in modest-resolution (R $\sim$ 400) stellar spectra.

803: Multiple stars are observed as they progress across the night sky.

804: Spectral features whose strengths vary with airmass are extracted by fitting to templates

805: of atmospheric extinction computed with models.

806: By use of appropriate patterns of stars on the sky one can attempt to disentangle the 

807: spatial and temporal characteristics of the major contributors to extinction.

808: Results from these studies will be subjects of future papers.

809: 

810: We carried out spectroscopy of a variety of standards stars of spectral class O-F

811: at the Hawaii 2.2~m on Mauna Kea, using the SNIFS spectrograph \citep{SNIFS04}, and on the CTIO 1.5~m with their

812: Ritchey-Chr{\'e}tien spectrograph. We will report on our results

813: in subsequent publications, but as described in section 7.1 our preliminary reductions illustrate the

814: diagnostic power of spectroscopy in conjunction with transmission models for determining the airmass-dependence

815: of atmospheric attenuation. 

816: 

817: \section{PHOTOMETRIC MEASUREMENTS OF ATMOSPHERIC TRANSMISSION}

818: 

819: To measure extinction using imaging systems, astronomers have traditionally used the same broadband imaging system used for their program targets to

820: observe standard stars.  This automatically avoids the ``color terms" required when transforming between photometry from different telescopes, cameras, or filters. If a set of comparison stars with spectra identical to the program objects can be found in the same field, this can potentially deliver high photometric precision. For PanSTARRS and LSST this approach is impractical, due to the time

821: needed to change filters. 

822: 

823: A number of programs have been undertaken to characterize the behavior

824: of atmospheric transmission. At the risk of oversimplifying, the astronomical community has turned their instrumentation to this task, typically using standard

825: astronomical passbands. The atmospheric sciences community, on the

826: other hand, has been 

827: acquiring data with filter sets that are optimized for understanding the 

828: properties of the atmosphere. Most of their imaging is done in the daytime, 

829: however.

830: 

831: Schuster et al (2002) describe a campaign of attenuation measurements using 13 narrowband filters than span 330 nm $< \lambda <$ 630 nm.  However, these filters were chosen for measuring the properties of stars, not the Earth's atmosphere.

832: 

833: \subsection {A Simultaneous Multiband Stellar Photometer to Measure Attenuation}

834: 

835: We have built and are now testing a dedicated, simultaneous multi-narrowband imaging photometer (High et al. in prep.).  The system uses a mask at the optical aperture, onto which wedge prisms and narrowband filters are mounted, in front of a fast camera lens.  We put these on a commercial deep depletion CCD camera (Pixis 1024BR, from Princeton Instruments) to achieve the same quantum efficiency we expect from PanSTARRS and LSST, and take pictures of bright stars.  A prism wedge offsets the angle of light rays from a given star just in front of the aperture, resulting in an offset but still localized stellar image at the focal plane.  By orienting each wedge differently and filtering their light independently, we produce an array of PSFs from a given star at different wavelengths in a single exposure.

836: 

837: The filter set is akin to that used for daytime solar photometers that are used

838: to characterize the atmosphere. We chose the central wavelength of one filter to coincide with the main water absorption feature at 950 nm, and another just off the feature at 1000 nm.  Another filter coincides with the narrow oxygen absorption band at 760 nm, and the other 3 are positioned where the Rayleigh/aerosol scattering dominates.  All our filters have FWHM of 10 nm.  This filter set allows us to probe the main, narrow absorption bands, and the Rayleigh/aerosol scattering components. Using a single

839: common shutter makes this a clean differential measurement. 

840: 

841: A test image from our simultaneous multiband imager is shown in Figure~\ref{fig:moon}, 

842:  and a representative profile of optical transmission versus airmass is presented in 

843:  Figure~\ref{fig:rainbow}. 

844: 

845: Such an apparatus is compact and largely made from commercially available parts.  During operation, it leaves the scientific program uninterrupted.  The data would later enter directly and quantitatively into the atmospheric attenuation model during analysis.  It is therefore an interesting potential alternative to the traditional sequential broadband imaging methods.

846: 

847: \section{ACTIVE INTERROGATION OF THE ATMOSPHERE: LIDAR}

848: 

849: The atmospheric science community has long used LIDAR (Light Detection and Ranging) to probe the properties of the atmosphere. The book by \cite{Measures84} is a comprehensive reference for this technique. A LIDAR system sends pulses of light up through the atmosphere, and measures the intensity of backscattered light as a function of time. The difference in light travel time to different altitudes provides a method for probing the vertical structure of the atmosphere.

850: 

851: Light pulses emitted into the atmosphere can be scattered elastically or inelastically. The latter process results in a wavelength shift in the scattered light after internal degrees of freedom are excited in the scattering particles. Inelastic scattering has a smaller cross section, so the detection of these events requires some combination of a more intense light source and a detector with a larger aperture than is necessary to measure elastic scattering. The benefit of measuring the inelastic channel is the ability to isolate the type of scattering particle by measuring only the light shifted by specific wavelengths corresponding to quantum mechanical transitions for a chosen molecule such as N$_2$ or O$_2$, for which the vertical density profiles are well known. These constraints on the probability of scattering from the inelastic channel allow for direct measurement of round-trip extinction. 

852: 

853: LIDAR systems exploiting both the elastic and inelastic channels have been used for decades to study atmospheric properties. \cite{Melfi72} gives an 

854: overview of inelastic (Raman) techniques, and  \cite{Vaughan93} describe using inelastic scattering to obtain precise temperature 

855: profiles in the atmosphere.   \cite{ALE06} and \cite{Dawsey06} have described their program of developing an eye-safe elastic-channel single wavelength LIDAR system for the determination of atmospheric extinction of astronomical sources, and we look forward to results from this system once it is deployed. The 

856: use of multiple elastic-scattering LIDAR systems to carry out what amounts to a tomographic measurements of aerosols above an observatory is described

857: in \cite{Auger1}. 

858: 

859: A high-power, tunable laser could be used to measure inelastic scattering and map atmospheric transmission over the wavelengths of interest. Measurements at multiple wavelengths can determine relative atmospheric transmission without any need for information about the molecular density profile, which is usually taken from models such as the US Standard Atmosphere (COSEA 1976) for absolute single-wavelength measurements. We note that the transmission of the atmosphere at 1.064 microns is very high, and this is a natural reference point since high intensity Nd:YAG lasers emit at this wavelength.

860: 

861: A simultaneous multiband LIDAR system could detect both the elastic and inelastic return signals. Using a tunable laser would allow us to measure on and off band regions near the water absorption feature, thereby measuring the strength of that feature. The LIDAR system can be aligned with the main survey system, and can take data during the readout and slew of the survey instrument.

862: 

863: The ability to map out the vertical profile of attenuation might provide a way to distinguish between attenuation processes that are spectrally coincident, such as aerosols (concentrated in the lower regions of the atmosphere) and ozone (concentrated in the lower stratosphere). Additionally, even thin clouds at high altitudes can be detected as an over-density of scattering particles. This added height dimension could contribute to atmospheric models discussed in the next section.

864: 

865: LIDAR is a mature technique that is commonly used in the atmospheric sciences, but has not yet been applied to the astronomical extinction problem. The use of high-power lasers at astronomical observatories does pose issues of safety as well as interference with photometric measurements at sites with multiple telescopes. However, adaptive optics systems already in place use similar devices which indicates that such problems are surmountable. We suspect that the narrowband laser light from

866: a nsec pulse tunable laser (with $\delta \lambda \sim $0.5 nm) is probably better

867: suited to measuring the aerosol scattering rather than water vapor absorption. A faster laser pulse would be correspondingly broader, of course.

868: 

869: \section{MODELING THE TRANSMISSION OF THE ATMOSPHERE: MODTRAN}

870: 

871: Sophisticated radiative transfer codes exist to model the transmission properties of the Earth's atmosphere.  

872: Here we discuss MODTRAN,  the \emph{MOD}erate spectral resolution atmospheric \emph{TRAN}smittance code 

873: developed by the US Air Force Research Lab \citep{2001SPIE.4381..455A}.   

874: MODTRAN produces a transmission function, from optical to infrared,  for arbitrary lines of sight through 

875: the atmosphere.  The code can be used with an input atmospheric profile, in full spherical refractive geometry, 

876: to decompose the integrated airmass into the altitude-dependent opacity along the line of sight.

877: The calculations are based upon spectroscopic band models and provide a maximum 

878: resolution of 2 cm$^{-1}$.  

879: %For high resolution applications, line-by-line radiative transfer codes, e.g.

880: %FASCODE \cite{1994SPIE.2266..613C}, are necessary.  

881: The band model approach is fast

882: but is limited to local thermodynamic equilibrium (LTE) conditions.  Non-LTE effects may be neglected in the lower

883: atmosphere, below 50km, but may become important for species found in the upper atmosphere.

884: The molecular species modeled by MODTRAN are derived from the HITRAN database 

885: \citep{2005JQSRT..96..139R} and include H$_{2}$O, CO$_{2}$, O$_{3}$, N$_{2}$O, CO, CH$_{4}$,

886: HNO$_{3}$, NO, NO$_{2}$, SO$_{2}$, O$_{2}$, N$_{2}$, NH$_{3}$. As noted by

887: \cite{Bailey07} and \cite{Adelman96}, the HITRAN database is not complete and it may be useful

888: to augment the line list with updated laboratory or observational results.

889: Aerosol scattering is modeled in discrete layers through the atmosphere.  Generic troposphere models 

890: (desert, rural, urban, maritime) are available as well as water vapor contributions from clouds.

891: User defined data may be substituted to extend the models to particular conditions of interest. 

892:   

893: MODTRAN is also capable of producing an emission calculation to model sky glow.  However, it is restricted

894: to the thermal contributions.  The important night sky emission lines are produced deep in the non-LTE regime

895: hundreds of kilometers up in the atmosphere and are not modeled.

896: 

897: \subsection{Blending measurements with models}

898: 

899: In our preliminary study, we have adapted MODTRAN for the atmospheric conditions of Mauna Kea Observatories.

900: This procedure will be extended to PS1 on Haleakala, and in principle, can be used to produce an atmospheric

901: model appropriate for any observatory.  

902: The Mauna Kea Weather Center provides real-time temperature, pressure and humidity measurements from the ground, 

903: as well as daily balloon radiosonde data to map the altitude profile.  We have used these measured pressure and 

904: temperature profiles to generate an input atmosphere definition for MODTRAN.  The Weather Center also provides the water vapor profile

905:  which can be incorporated into the model atmosphere directly.  To derive the concentrations of other molecular species, 

906:  we simply scale the concentrations  in the generic US Standard Atmosphere by the pressure and temperature.  

907: 

908: The atmospheric transmission models hide hundreds of free parameters.  We have 

909: explored the minimum set necessary to fit the measured atmospheric features in measured stellar spectra

910: with MODTRAN models and find that, with appropriate model atmosphere profiles, we can fit observations

911: by scaling only the water vapor content.

912: 

913: Figure~\ref{fig:spectrum} shows a measured O star spectrum from our Mauna Kea data set, overplotted with a MODTRAN model.

914: The spectrum has been flattened and normalized to isolate the line absorption features.

915: The primary absorption bands of interest in the optical are the oxygen line at 760 nm and the water band at

916: 950 nm.  We tune the MODTRAN output to these features by first fixing the line of sight to match the observation,

917: and then scaling the oxygen and water vapor concentrations to match the absorption depths.

918: The water vapor is the dominant absorber in the MODTRAN model and we can put strong constraints

919: on it through the fit.   It is less sensitive to the oxygen concentration. Thus, through joint spectroscopic 

920: measurements and modeling, the temporal and spatial variations in water vapor can be effectively tracked.

921: 

922: Atmospheric transmission models can be integrated into an observatory using measurements from a dedicated 

923: instrument to monitor key absorption features.  As discussed, a simple system could consist of a dedicated imaging 

924: camera with narrow-band filters centered on strong atmospheric absorption bands.  The camera would

925: collect atmospheric transmission data in synchrony with the survey telescope operations, and provide

926: constraints on oxygen and water vapor concentrations as well as aerosol scattering.

927: These measurements could then be fit by an atmospheric modeling code, such as MODTRAN, to

928: produce a full, high-resolution transmission function representing a particular line-of-sight and time period.

929: 

930: \section{Summary and Next Steps}

931: 

932: This paper addresses the goal of achieving precise relative photometry over the course of

933: next-generation ``all-sky'' imaging surveys such as PanSTARRS and LSST.

934: Calibrations of absolute scales may also be possible, but the scientific goals for these

935: future surveys stress uniformity of the photometry across large scales on the sky, 

936: and the identification of time-dependent celestial

937: phenomena.

938: 

939: Our thesis is that significant improvement in photometric measurement,

940: perhaps even to the millimag level of relative precision, can be enabled by direct

941: measurement of atmospheric throughput. 

942: Several techniques for making such measurements are suggested, but all rely on identification

943: of a relatively small number of contributors to atmospheric extinction - the final four.

944: Each of these contributors leaves characteristic signatures in transmitted spectra.

945: It is the challenge of our approach to show how to detect and quantify these signatures

946: with sufficient temporal and spatial resolution to allow precise relative photometric

947: measurements to be made.

948:  

949: Molecular absorption is significant for only a few species

950: (specifically ozone, oxygen, OH, and water vapor) 

951: and occurs at well-known and characteristic wavelengths.

952: Water vapor column heights vary substantially with atmospheric conditions, but optical 

953: depths of remaining molecular constituents are simply given by barometric pressure.

954: The theory of Rayleigh-Carbannes molecular scattering is well developed, and the process

955: presents a strong and stable $\lambda^4$ signature that is easily recognized.

956: Measurement of aerosol scattering is considerably more troublesome.

957: 

958: We anticipate that the variability of both the aerosol content of the

959: atmosphere and the water content will present the dominant atmospheric 

960: limitation to precision photometry from next-generation surveys. This suggests that instrumentation development 

961: focus on these two concerns. 

962: 

963: Our analysis leads to some thoughts on future research.

964: Firstly, we conclude that, while it may be possible to ``tune-up" techniques and algorithms

965: with parasitic use of existing data, it will be necessary to carry out dedicated

966: measurements to test the ultimate capability of any of these to meet our goals.

967: This conclusion is not surprising since what we are trying to do has not been done before,

968: {\it i.e.} no existing data set has met the goals of these next-generation surveys.

969: 

970: In follow-on campaigns it will be most useful to carry out tests

971: with simultaneous spectroscopic

972: measurements of stars with photometric measurements of a number of standard targets.

973: While these measurements need not be carried out over the scales of future surveys,

974: they will need to be made with care and dedication if it is to be shown

975: that survey-wide millimagnitude photometry is possible. 

976: 

977:  

978: \section{ACKNOWLEDGMENTS}

979: 

980: We thank Justin Albert, Tim Axelrod, James Battat, Yorke Brown, Kelly Chance, Chuck Claver,

981: Kem Cook, Doug Finkbeiner, Jim Gunn, Zjelko Ivezic, John McGraw, Gene Magnier, Eli Margalith, David Schlegel, 

982: Nick Suntzeff and Doug Welch for useful conversations and input. 

983: We are very grateful for the efforts of the AERONET 

984: consortium, and Brent Holben in particular, for both the establishment 

985: and operation of the 

986: aerosol monitoring site on Mauna Loa, and for making the AERONET data

987: readily available on the web. 

988: We have also used results from the Total Ozone Mapping Spectrometer (TOMS)

989: on the EarthProbe satellite, 

990: and we are grateful to the team that built and operated this instrument. 

991: We thank the LSST Corporation, Harvard University and the Department of 

992: Energy Office of Science and the US Air Force Office of Scientific Research for their support of this work.

993: The LSST design and development activity is supported

994: by the National Science Foundation under Scientific Program Order No. 9

995: (AST-0551161) through Cooperative Agreement AST-0132798. 

996: Additional support was provided through the NSF award to the

997: ESSENCE supernova cosmology project, through AST-0607485.

998: Portions of this

999: work were performed in part under Department of Energy contracts DE-AC02-

1000: 76SF00515, DE-AC02-98CH10886, DE-FG02-91ER40677 and W-7405-Eng-48.

1001: Additional funding comes from private donations, in-kind support at Department

1002: of Energy laboratories and other LSSTC Institutional Members.

1003: 

1004: \begin{thebibliography}

1005: 

1006: 

1007: \bibitem[Adelman et al.(1996)]{Adelman96} Adelman, Saul J., Gulliver, Austin F. and Holmgren, David E. 

1008:  \ 1996 in {\it Model Atmospheres and Spectrum Synthesis}, ASP Conference Series 108, 293.

1009: 

1010: \bibitem[Albert et al.(2006)] {Albert06} Albert, J et  al.~ \ 2006 astro-ph/0604339.

1011: %one-way from satellites

1012: 

1013: \bibitem[Anderson et al.(2001)]{2001SPIE.4381..455A} Anderson, G.~P., et al.~\ 2001, \spie, 4381, 455 

1014: 

1015: \bibitem[Anderson et al.(2003)]{Anderson03} Anderson, T.L. et al, \ 2003 Jour. of Atmos. Sci. 60, 119

1016: %2 point correlation function of aerosol transmission 

1017: 

1018: %\bibitem[Angione, Medeiros \& Roosen (1976)]{Angione76} Anginoe, Ronald J., Medeiros, Edward J., and

1019: %Roosen, Robert G., \ 1976, Nature, 261, 289. 

1020: %Stratospheric Ozone as viewed from the Chappius band, showing variability

1021: 

1022: \bibitem[Bailey, Simpson \& Crisp(2007)]{Bailey07} Bailey, J., Simpson, A. and Crisp, D. \ 2007

1023: \pasp, 119, 228

1024: %Correcting IR Spectra for Atmospheric Transmission

1025: 

1026: \bibitem[BenZvi et al.(2007a)]{Auger1} BenZvi, S.~Y.~ et al. \ 2007 Nucl. Instrum. Meth., A574 171 . 

1027: %The LIDAR System of the Pierre Auger Observatory.

1028: 

1029: \bibitem[BenZvi et al.(2007b)]{Auger2} BenZvi, Segev et al. \ 2007 in 30$^{th}$ International Cosmic Ray Conference, and arXiv astro-ph/0706.1710

1030: %new method for atmospheric calibration at the Pierre Auger Observatory using FRAM, a robotic astronomical telescope.   

1031: 

1032: \bibitem[Bessell(1999)]{Bessell99} Bessell, Michael \ 1999 \pasp,  111, 1426

1033: % Spectrophotometry: revised standards and techniques

1034: 

1035: \bibitem[Burke et al.(2006)]{Burke06} Burke, David et al.~ \ 2006 \spie, 6267, 626715

1036: %Calibrations of LSST camera and telescope systems

1037: 

1038: %\bibitem[Chetwynd et al.(1994)]{1994SPIE.2266..613C} Chetwynd, J.~H., Wang, J., \& Anderson, G.~P.\ 1994, \procspie, 2266, 613 

1039: 

1040: \bibitem[COSEA(1976)]{COSEA} Committee on Extension to the Standard Atmosphere, \ 1976 ``US Standard

1041: Atmosphere", US Govt Printing Office, Washington DC.

1042: 

1043: \bibitem[Dawsey et al.(2006)]{Dawsey06} Dawsey, M. et al.~\ 2006 \spie, 

1044: 6270 47. 

1045: %Lidar for measuring atmospheric extinction

1046: 

1047: \bibitem[De Vaucouleurs \& Angione(1973)]{DeVauc73}De Vaucouleurs, G. and Angione, R.J. \ 1973 

1048: \pasp, 86, 104

1049: % Atmospheric Extinction at McDonald Observatory 1960-1968

1050: 

1051: \bibitem[Eck et al.(1999)]{Eck99} Eck, T.F. et al.~\ 1999 Jour Geophy Res, 

1052: 103, 31. 

1053: % Wavelength dependence of the optical depth of biomass burning, urban, 

1054: % and desert dust aerosols. 

1055:  

1056: \bibitem[Everett \& Howell(2001)]{EH01} Everett, Mark E and Howell, Steve B., \ 2001 

1057: \pasp, 113, 1428 

1058: %A Technique for Ultrahigh-Precision CCD Photometry

1059:  

1060: \bibitem[Frogel (1998)] {Frogel} Frogel, J. \ 1998 \pasp, 110, 200 

1061: %A History of IR Extinction at CTIO and a Possible Connection with the El Nino Phenomenon.

1062: 

1063: %\bibitem[Griffin (2005)]{Griffin05} Griffin, E. \ 2005 \pasp, 117, 885

1064: %Telluric ozone from archival stellar spectra

1065: 

1066: \bibitem[Hansen \& Travis(1974)]{Hansen} Hansen, James, E. and Travis, Larry D., \ 1974, Space Science Reviews, 16, 527.  

1067: %Nice pedagogical overview of scattering in atmospheres. 

1068: 

1069: \bibitem[Hartman et al.(2005)] {Hartman05}  Hartman, J.D. et al.~\ 2005 \aj, 130, 2241 

1070: %Pushing the Limits of Ground-based Photometric Precision: Submillimagnitude Time-Series Photometry of the Open Cluster NGC 6791

1071: 

1072: \bibitem[Hadrava(2006)]{Hadrava06} Hadrava, P. \ 2006 A\&A 448, 1149

1073: % Disentangling telluric lines in stellar spectra

1074: 

1075: \bibitem[High(2007)]{High07} High, F.W. et al.~ \ 2007 (in prep).

1076: 

1077: \bibitem[Hinkle, Wallace and Livingston(2003)]{KPNO} Hinkle, K.H., Wallace, L., and 

1078: Livingston, W., \ 2003 \baas, 203, 3803 

1079: %Atmospheric Transmission Above Kitt Peak, 0.5 to 5.5 microns

1080:  

1081: \bibitem[Holben et al.(1999)]{Holben99}  Holben, B.N., et al. \ 2001 , J. Geophys. Res., 106, 67

1082: %An emerging ground-based aerosol climatology: Aerosol Optical Depth from AERONET

1083:  

1084:  %\bibitem[Howell et al.(2003)]{Howell03} Howell, Steve B., et al.~ \ 2003 \pasp,  11, 1340 

1085:  %Photometric Observations Using OT CCDs

1086:  

1087:  \bibitem[Ivezic et al.(2007)]{Ivezic07} Ivezic, Z., et al.~ \ 2007 astro-ph/0703157, submitted to \aj.

1088:  

1089:  \bibitem[Jaross et al.(2003)]{Jaross03} Jaross, G. et al.~\ 2003 AGU Fall 

1090:  Meeting abstract A21D-0993. 

1091:  %The Good, the Bad, and the Ugly: 25 Years of TOMS Optical Degradation

1092:  

1093:  \bibitem[Kaiser et al.(2002)]{PS02} Kaiser, Nicholas et al.~\ 2002 \spie, 4836, 154 

1094: % Pan-STARRS: A Large Synoptic Survey Telescope Array

1095: 

1096: %\bibitem[Kinney et al.(1996)]{Kinneyatlas96} Kinney, A. et al.~\ 1996, \apj 

1097: %Atlas of galaxy spectra

1098: 

1099:  \bibitem[Krisciunas et al.(1987)] {KrisciunasMK} Krisciunas, K. et al.~\ 1987 \pasp, 99, 887

1100:  %MK extinction and sky brightness

1101:  

1102:  \bibitem[Kurucz(1993)]{Kurucz} Kurucz, R., \ 1993, ATLAS9 Stellar Atmosphere Programs and 2 km/s grid. Kurucz CD-ROM No. 13. Cambridge, Mass.: Smithsonian Astrophysical Observatory.

1103:  

1104:  \bibitem[Kylling, Albold \& Seckmeyer(1997)]{Kylling97} Kylling, A., 

1105:  Albold, A. and Seckmeyer, G., \ 1997, Geophys. Res. Lett 24, 397.

1106:  %Transmittance of a cloud is wavelength-dependent in the UV-range: physical 

1107:  % interpretation  

1108:  

1109:  \bibitem[Lantz et al.(2004)]{SNIFS04} Lantz, B et al.~\ 2004 \spie 5249, 146

1110:  %SNIFS: A wideband integral field spectrograph with microlens arrays

1111:  

1112:  %\bibitem[Lejeune et al.(1997)]{Lejeune97} Lejeune, T., Cuisinier, F. and Busier, R. 

1113:  %\ 1997, A\&A Supp., 1215, 229. 

1114:  %A standard stellar library for evolutionary synthesis: I. Calibration of theoretical spectra 

1115:  

1116: \bibitem[LSST collaboration(2007)]{LSST} LSST collaboration, LSST Science Case.  {\tt http://www.lsst.org/Science/lsst\_baseline.shtml}

1117:  %lsst science case

1118:  

1119: \bibitem[Magnier(2007)]{MagnierPS} Magnier, E, in  

1120: The Future of Photometric, Spectrophotometric and Polarimetric Standardization, ASP Conference Series, Vol. 364, C. Sterken, ed, . San Francisco: Astronomical Society of the Pacific, 2007.

1121: 

1122: \bibitem[Measures(1984)]{Measures84} Measures, R. ''Laser Remote Sensing'', Wiley, 1984.  

1123: %primary reference for all things LIDAR

1124: 

1125: \bibitem[Melfi(1972)]{Melfi72} Melfi, S.H.~ \ 1972 Applied Optics 11, No. 7  1605.

1126: %Remote Measurements of the Atmosphere using Raman Scattering

1127: 	  	  

1128: \bibitem[Miknaitis et al.(2007)]{ESSENCE} Miknaitis, G. et al.~\ 2007 \apj,  in press, 

1129: astro-ph/0701043

1130: %The ESSENCE Supernova Survey: Survey Optimization, Observations, and Supernova Photometry

1131: 

1132: %\bibitem[Ochsenbein et al.~(2000)]{Vizier} Ochsenbein, F., Bauer, P. and Marcout, J. \ 2000, A \& A Supp 143, 221. 

1133: %Vizier preferred reference. 

1134: 

1135: \bibitem[Padmanabhan et al.(2007)]{SDSS07} Padmanabhan, N. et al.~\ 2007 astro-ph/0703454

1136: %SDSS all sky calibration paper. 

1137: 

1138: \bibitem[Pakstiene(2001)]{Pakstiene} Pakstiene, E. \ 2001 Baltic Astronomy, 10, 651

1139: % The Dependence of Atmospheric Extinction on Meteorological Conditions and Aerosol Size Distribution

1140: 

1141: %\bibitem[Plana-Fattori et al.(1998)]{Plana98} Plana-Fattori, A. et al.~\ 1998

1142: %Jour. of Appl. Meterology 37, 790

1143: %Estimating the Atmospheric Water Vapor Content from Sun Photometer Measurements

1144: 

1145: \bibitem[Rayleigh(1899)]{Rayleigh} Rayleigh, Lord, \ 1899 Phil. Mag. 47, 375Ð394

1146: %Need to find hardcopy of this paper! Original blue sky through molecular scattering calc. 

1147: 

1148: \bibitem[Reimann et al.(1992)] {Riemann92} Reimann, H.-G. et al.~ \ 1992 A\&A, 265, 360

1149: %Atmospheric Extinctino and meterological conditions, a long term study

1150: 

1151: \bibitem[Roosen \& Angione(1977)]{Roosen77} Roosen, Robert G. and Angione, Ronald J. \ 1977

1152: \pasp, 89, 814

1153: %Variations in atmospheric water vapor: baseline results from smithsonian obsevations

1154: 

1155: \bibitem[Rothman et al.(2005)]{2005JQSRT..96..139R} Rothman, L.~S., et al.~\ 2005, Journal of Quantitative Spectroscopy and Radiative Transfer, 96, 139 

1156: 

1157: \bibitem[Schuster, Parrao \& Guichard(2002)]{SanPedro} Schuster, W.J., Parrao, L and 

1158: Guichard, J. \ 2002, Journal of Astronomical Data 8, 2. 

1159: %The atmospheric extinction at San Pedro Martir, Mexico: Individual observations,

1160: %monthly and yearly averages 

1161: 

1162: \bibitem[Sinyuk et al.(2007)]{Sinyuk07} Sinyuk, A.~ et al.~ \ 2007 Remote Sensing of Environment 107, 90

1163: %simultaneous retrieval of aerosol and sirface properties from a combination

1164: % of AERONET and satellite data.

1165: 

1166: \bibitem[Stritzinger et al.(2005)]{Stritzinger05} Stritzinger, M et al.~ \ 2005 \pasp,  117, 810

1167: %Spectrophotometric Atlas and atmos extinction

1168: 

1169: \bibitem[Stubbs \& Tonry(2006)]{ST06} Stubbs, C. and Tonry, J. \ 2006 \apj,  646, 1436. 

1170: %Towards 1 percent photometry....

1171: 

1172: \bibitem[Stubbs et al.(2007)]{CTIOlaser} Stubbs, C.W. et al.~ in

1173: The Future of Photometric, Spectrophotometric and Polarimetric Standardization, ASP Conference Series, Vol. 364, C. Sterken, ed, . San Francisco: Astronomical Society of the Pacific, 2007., p.373

1174:  %Initial tunable laser throughput results from CTIO

1175:  

1176:  \bibitem[Tonry et al.(2005)]{JT05} Tonry, John L. et al.~\ 2005 \pasp, 117, 281 

1177:  %A Search for Variable Stars and Planetary Occultations in NGC 2301. I. Techniques

1178:  

1179:  \bibitem[Thomas-Osip et al.(2007)]{Campanas07} Thomas-Osip, J. et al.~ \ 2007 astro-ph/0706.2683 (accepted \pasp.)

1180: % Calibration of the Relationship between Precipitable Water Vapor and 225 GHz Atmospheric Opacity via Optical Echelle Spectroscopy at Las

1181: % Campanas Observatory

1182:  

1183:  \bibitem[Tucker et al.(2006)]{SDSSpt} Tucker, D. et al.~ \ 2005 Astron. Nachr. 327, 821

1184:  %SDSS Monitor telescope pipeline

1185:  

1186:  \bibitem[Vaughan et al.(1993)]{Vaughan93} Vaughan, G. et al.~ \ 1993 Applied Optics 32, 2758. 

1187:  %Atmospheric temperature measurements made by rotational Raman scattering

1188:  

1189:  \bibitem[Wade \& Horne(1988)]{Wade88} Wade, Richard A. and Horne, Keith \ 1988 \apj, 324, 411

1190:  %The Radial Velocity Curve and ... (actually we want part about airmass correction for telluric abs)

1191:  

1192: \bibitem[Young et al.(1991)] {Young} Young, Andrew T et al.~ \  1991 PASP 183, 221. 

1193: %Precise Automatic Differential Stellar Photometry

1194: 

1195: \bibitem[Zimmer et al.(2006)]{ALE06}  Zimmer, P., et al.~ \ 2007  BAAS 209 1540.

1196: %ALE elastic lidar system

1197: 

1198: \end{thebibliography}

1199: 

1200: \clearpage

1201: 

1202: %This is figure 1

1203: \begin{figure}

1204: \plotone{f1.eps}

1205: 

1206: \caption{Photon Transmission through One Airmass, with PanSTARRS Instrumental 

1207: Sensitivity. 

1208: This plot shows the 

1209: different contributions to attenuation of light in passing through the atmosphere, 

1210: along with filter bands times the expected detector QE, for PanSTARRS. The LSST filter set is similar, but also 

1211: includes a $u$ band. The atmospheric transmission was computed with MODTRAN for one airmass 

1212: at an elevation of 10,000 ft (305m), with an initial spectral resolution of 1 cm$^{-1}$, boxcar smoothed to

1213: 1 nm.} 

1214: \label{fig:photon_transmission}

1215: \end{figure}

1216: 

1217: \clearpage

1218: %This is Figure 2

1219: \begin{figure}

1220: \plotone{f2.eps}

1221: 

1222: \caption{Aerosol Optical Depth. The graph shows the daytime aerosol optical depth at 440 nm over time  

1223: as reported from solar flux measurements by the AERONET system on Mauna Loa, near the site of the 

1224: PanSTARRS-1 system. 

1225: The vertical axis is optical depth $\tau$ where a fraction $e^{-\tau}$ is 

1226: transmitted through one airmass. Time is in years after June 1994). 

1227: There is clear evidence for seasonal cycles, as well

1228: as considerable variation on short timescales. } 

1229: \label{fig:t440_full}

1230: 

1231: \end{figure}

1232: 

1233: \clearpage

1234: %THis is Figure 3

1235: \begin{figure}

1236: \plotone{f3.eps}

1237: 

1238: \caption{Aerosol Optical Depths at Different Wavelengths. This plot shows in more detail the AERONET reported daytime Mauna Loa

1239: aerosol optical depth at different wavelengths, vs. time. The plots span 

1240: a period of 72 days. Note the variation in spectral dependence of the 

1241: attenuation spikes.} 

1242: \label{fig:depth_zoom}

1243: 

1244: \end{figure}

1245: 

1246: \clearpage

1247: %This is Figure 4

1248: \begin{figure}

1249: \plotone{f4.eps}

1250: 

1251: \caption{An Extinction Excursion. 

1252: These panels show the Mauna Loa AERONET attenuation at 1020 nm (upper left), 

1253: 870 nm (upper right), 675 nm (lower left), and 340 nm (lower right) 

1254: plotted vs. the attenuation at 440 nm, for the period between 5.20 and 5.25

1255: years of Figure 4. A wavelength independent change in transparency would 

1256: generate a line with a slope of $m=+1$ in each of the panels. Aerosol scattering

1257: would have slopes less than unity in all except the lower right panel. The 

1258: data show evidence for both kinds of excursions in this time period.} 

1259: \label{fig:quad1}

1260: 

1261: \end{figure}

1262: 

1263: \clearpage

1264: %This is figure 5

1265: \begin{figure}

1266: \plotone{f5.eps}

1267: 

1268: \caption{An Extinction Excursion. Same panels as in Figure 4, but for the period between 

1269: 5.30 and 5.32 years. This attenuation spike appears to be predominantly

1270: due to aerosols.} 

1271: \label{fig:quad2}

1272: 

1273: \end{figure}

1274: \clearpage

1275: %This is Figure 6

1276: \begin{figure}

1277: \plotone{f6.eps}

1278: 

1279: \caption{Aerosol Optical Depth Statistics. Cumulative distributions of aerosol optical depth on Mauna Loa

1280: are shown, 

1281: from the AERONET observations over the period shown in Figure 2.} 

1282: \label{fig:cumulative_aerosol}

1283: 

1284: \end{figure}

1285: 

1286: \clearpage

1287: %This is Figure 7

1288: \begin{figure}

1289: \plotone{f7.eps}

1290: 

1291: \caption{Variation Statistics for Aerosol Optical Depth. Cumulative distributions of aerosol optical depth changes (per hour) 

1292: for the AERONET data on Mauna Loa are shown. The solid line corresponds to 1020 nm

1293: and the broken line to 440 nm.} 

1294: \label{fig:cumulative_deriv}

1295: 

1296: \end{figure}

1297: 

1298: 

1299: \clearpage

1300: %This is figure 8

1301: \begin{figure}

1302: \plotone{f8.eps}

1303: \caption{Ozone Attenuation Variability. This plot shows the evolution of the ozone content of the 

1304: atmosphere vs. time above Hilo, HI, as measured by the EarthProbe TOMS 

1305: satellite-borne instrument. The

1306: y axis is in units of Dobsons. Each Dobson unit is equivalent to a thickness of  

1307: 0.01 mm of ozone at STP. There is clear evidence of annual cyclic variation

1308: at the $\pm$25\% level about the mean value. Remote sensing data such as these

1309: can be used to determine the optical attenuation due to ozone without needing 

1310: any ground-based measurements.} 

1311: \label{fig:ozone}

1312: 

1313: \end{figure}

1314: 

1315: \clearpage

1316: %This is figure 9

1317: \begin{figure}

1318: \plotone{f9.eps}

1319: 

1320: \caption{Cumulative Distribution of precipitable water content (cm) above Mauna Loa, from the 12 year AERONET data set. } 

1321: \label{fig:cumulative_water}

1322: 

1323: \end{figure}

1324: 

1325: \clearpage

1326: %This is figure 10

1327: \begin{figure}

1328: \plotone{f10.eps}

1329: 

1330: \caption{Time evolution of water content of the atmosphere at Mauna Loa.

1331: This plot of water content (in cm) vs. time covers the same interval as

1332: the plots in Figure~{\protect \ref{fig:depth_zoom}}.} 

1333: \label{fig:watervt}

1334: 

1335: \end{figure}

1336: 

1337: 

1338: \clearpage

1339: %This is figure 11

1340: \begin{figure}

1341: \plotone{f11.eps}

1342: 

1343: \caption{Cumulative plot of rate of change of AERONET's derived 

1344: precipitable water content (cm per hour). For a typical value of 0.2 cm

1345: this implies that measurements more frequently than hourly are 

1346: required to track 10\% changes with confidence. } 

1347: \label{fig:cum_water_deriv}

1348: 

1349: \end{figure}

1350: 

1351: 

1352: \clearpage

1353: %This is figure 12

1354: \begin{figure}

1355: \plotone{f12.eps}

1356: 

1357: \caption{Image of the moon obtained simultaneously in four passbands, 

1358: as part of our development of an simultaneous multiband imaging instrument designed to measure

1359: atmospheric transmission in real time. Images of a point source would be

1360: well separated and the flux differences can be used to deduce attenuation.} 

1361: \label{fig:moon}

1362: 

1363: \end{figure}

1364: 

1365: 

1366: \clearpage

1367: %This is figure 13

1368: \begin{figure}

1369: 

1370: \plotone{f13.eps}

1371: 

1372: \caption{Aerosol Attenuation Data from Simultaneous Multiband Imaging. This plot shows preliminary relative attenuation results we obtained 

1373: using the multiband imager, from CTIO. The vertical axis is the ratio of transmission at

1374: 440 nm to that at 1000 nm, plotted vs. airmass. The two fluxes were obtained simultaneously

1375: using the instrument described in the text. We consider this a promising technique to 

1376: determine attenuation from both aerosol and molecular processes.} 

1377: \label{fig:rainbow}

1378: 

1379: \end{figure}

1380: 

1381: \clearpage

1382: %This is figure 14

1383: \begin{figure}

1384: \plotone{f14.eps}

1385: 

1386: \caption{This plot shows a spectrum we obtained on the UH 88 inch telescope, 

1387: normalized to show absorption features. The inset panels show the comparison

1388: between the observed spectrum and MODTRAN output, for the parameters

1389: shown. We intend to pursue the idea of combining observations and 

1390: models to determine the optical transmission function of the atmosphere.} 

1391: \label{fig:spectrum}

1392: 

1393: \end{figure}

1394: 

1395: 

1396: \end{document}

1397: 

1398: 

1399: 	 							

1400: 									

1401: