Analyses of atmospheric extinction data obtained by astronomers—I. A time-trend analysis of data with internal accidental errors obtained at four observatories

Abstract

In this paper, we describe an analysis of astronomical atmospheric extinction data for longterm time trends. The data have been obtained by various observers, using either the broad-band UBV system or various narrow-band systems. We have used only data which permit either calculation bf internal accidental errors or reasonable guesses concerning these errors. We have analyzed data which fulfill these conditions for four observatories—Cerro Tololo Interamerican Observatory, Kitt Peak National Observatory, Lick Observatory, and McDonald Observatory. Using several methods, we have calculated monochromatic atmospheric extinction of the reference wavelengths of the B and V filters (4417 and 5504 Å, respectively) from the data. The broad-band methods all follow from the second-order extinction equation given by King. We have tested two of these broad-band methods observationally, and find that they yield results which are satisfactory for our purposes. We have calculated least-squares fits to the monochromatic extinction for each observatory, using constant and periodic terms and either step functions or terms linear with time. The step functions enable us to test for linearity of the long-term trends. We have used a weighting scheme which we call “histogram shaping” to make the random variation of the extinction essentially Gaussian. We find that we have insufficient data for meaningful conclusions about Cerro Tololo, but that this site probably has small periodic terms and is promising for future work. For the remaining three observatories, we find that the true night-to-night extinction variation is considerably greater than the scatter in the data due to their internal accidental errors, and that the true night-to-night variation therefore determines the precision of the results. Between 1961 and 1965, both blue and yellow extinction rose at McDonald, with the rise being greater for the yellow extinction; there was no detectable increase in either wavelength region between 1965 and 1968. There was a small rise of doubtful significance in the blue extinction at Kitt Peak between 1962 and 1972, with no detectable rise in the yellow extinction; there was no detectable rise in either wavelength region between 1955 and 1970 at Lick. If we assume that there have been no compensatory trends in local and background aerosol, our best values for extinction trends due to background aerosol changes during the years 1960 to 1972 are 0.006 ± 0.013 (rms) and 0.009 ± 0.009 (rms) stellar magnitudes per air mass per decade in the blue and yellow wavelength regions, respectively. (These values also apply if the trends are expressed as optical depth per air mass per decade.) The latter result is essentially unchanged if data originally analyzed by Lockwood and Hartmann are included in the analysis. The V extinction trend translates into a mass loading trend (to within a factor of 2) of ∼ 10 ± 10 μg m −3 decade −1. If typical background air has aerosol mass loadings of 10–20 μg m −3, the stated trend would imply an increase in the background loading of 50–100% decade. −1