JSTOR

1. INTRODUCTION

The success of advanced interferometric and adaptive optics instruments for use at ground‐based observatories depends strongly on the quality of the astronomical seeing at the site (Woolf 1982; Racine et al. 1991; Hardy 1998). In particular, it is important during the design phase of such instruments to understand the characteristic behavior of the seeing and to have an estimate of the number of nights of subarcsecond seeing that can be expected each year at the site. Not taking this into account could result in developing a system that operates only under optimal conditions, greatly restricting the number of nights that it can be used, while designing for the worst conditions would reduce the instrument’s scientific leverage.

The results of site surveys typically provide a measure of the seeing over a relatively short time period, and, without repeated or long‐duration studies, little insight into the long‐term seeing can be achieved. Considerable effort has gone into studying and comparing the seeing at various astronomical sites (Walters & Bradford 1997) and investigating the short‐timescale fluctuations in the seeing (Racine 1996). Unfortunately, there are few studies of the seeing over extended time periods, so the long‐term seeing at most observatories has remained largely unexplored.

Shortly after the opening of the Carnegie Institution of Washington’s observatory on Mount Wilson, its director, George E. Hale, began recording the astronomical seeing at the telescopes. The observatory has remained active to the present day, in spite of the challenges of being located above the Los Angeles basin, and seeing measurements have been recorded nightly throughout its lifetime. In a recent paper by Teare et al. (2000, hereafter Paper I), this data set was used to examine the changes in the quality of the astronomical seeing at the observatory over the last eight decades.

The presence of such long‐period variations has several important implications. For instance, in site survey work, measurements made during a maximum or a minimum in the seeing cycle could mistakenly lead to the conclusion that the quality of a site is either better or worse than on average. Periodic variations would also make it more difficult to separate out the natural seeing variations from other short‐term events that may have influenced the seeing (R. B. Stothers 2000, private communication; Stothers 2001). In this paper we address the question of whether there are long‐period variations in the subarcsecond seeing, using the historical seeing record for the Mount Wilson Observatory.

2. ANALYSIS AND DISCUSSION

An overview of the Mount Wilson seeing data set is described in detail in Paper I. The objective in Paper I was to evaluate the trends in the seeing and to establish how the quality of the seeing had fared over the lifetime of the observatory. In that paper the seeing was evaluated on the timescale resolution of 1 yr and showed that in addition to several “low points” in the seeing that correspond to large‐scale construction and development activities, there has been an overall degradation in the measured seeing since the opening of the observatory. However, these changes are related to a degradation of the conditions in the vicinity of the telescopes rather than deterioration in the atmospheric and natural seeing conditions in the near‐coastal region of southern California.

In this paper, the timescale resolution of the seeing data has been increased by about 1 order of magnitude; that is, the time steps are approximately 1 month. The data set is composed of the number of days per month of subarcsecond seeing measurements, binned so that there are 12 seeing measurements per year over the period from 1919 to 1999. In the remainder of this paper the term “seeing measurement” will refer to the number of days per month of subarcsecond seeing to avoid using this long descriptor.

A typical sample of the data is shown in Figure 1. This figure illustrates the summer‐winter variation in the subarcsecond seeing; that is, the summer months can be identified by a peak in the seeing measurements. The 5 yr sample in Figure 1 also shows that while each year can be identified by its general shape, there are month‐to‐month and year‐to‐year variations.

(114 KB)

Fig. 1.— Subarcsecond seeing from 1940 through 1944. Each bar represents the number of days per month that the subarcsecond seeing was less than 1 .

Open New Window

The complete data set is shown in Figure 2. In general, the individual spikes in the data indicate the individual years. There are also several gaps that can be seen in the data that are due to either missing data or months where the seeing was poor so no subarcsecond values were recorded. Three separate time‐related features can be easily identified in this plot. These are individual years, an approximately decade‐length “waxing and waning,” and a gradual decrease in the intensities of the seeing measurements over the years. This latter point was investigated in Paper I, and the other two points are addressed now.

(112 KB)

Fig. 2.— Complete data set showing the subarcsecond seeing data for the years 1919–1999. In general, each year of the study can be identified by the summer maximum in the number of days of subarcsecond seeing. In addition to the yearly peaks, there is a secondary period of increase and decrease in the intensity of the seeing data.

Open New Window

The stability of the summer peak was evaluated by calculating the first moment of the seeing measurement distribution for each year. Figure 3a shows the centroid of the yearly seeing measurement distribution with time. In general, this centroid consistently occurs between the months of July and September. It should be noted that the first moment calculation looks at the seeing distribution over the whole year; however, in some of the years covered in this study the individual month with the highest number of subarcsecond days may be different. Figure 3b shows a histogram of the centroid data showing the distribution of the months where the year’s best seeing occurs, with the peak in the month of August. In the period up until the late 1970s there is little variation in the monthly position of the centroid of the seeing distribution; after this, the scatter in the data increases. Some possible causes for the increased scatter in the data are discussed in more detail in Paper I.

(1 of 2 images)

(73 KB)

Fig. 3.— (a) Plot of the first moment of the subarcsecond seeing data distribution for each year and (b) histogram plot showing the dominance of the summer months in having the best seeing in each year.

Open New Window

Figure 2 shows that there are variations in the seeing measurements that occur at periods longer than 1 yr. These variations were analyzed by calculating the power spectrum of the seeing measurements using both Fourier transform and periodogram methods. Figure 4 shows a plot of the power spectrum derived using a Fourier transform of the seeing measurements with a prominent peak in the seeing at 1 yr, consistent with the summer‐winter cycle in the seeing. The power spectrum also shows that there are other significant periodic variations in the seeing; however, it does not clearly resolve these longer periods.

(47 KB)

Fig. 4.— Power spectrum of the subarcsecond seeing. There are several peaks in the spectrum indicating periodicity in the data at greater than 1 yr. The peaks at the highest frequencies, that is, the shortest time periods, are due to aliasing.

Open New Window

A periodogram analysis, based on the approach of Scargle (1982) with improved normalization (Horne & Baliunas 1986), was used to determine the long‐term periods in the seeing data. Confining the results to those with low “false alarm” probability, that is, having a detection probability greater than 99% and approximately four complete cycles of data, the periods were identified as 1, 11.6, and 20.2 yr.

It is clear that there is a strong correlation between the 1 yr period in the seeing and the local weather variations over the summer‐winter cycle. It is tempting to look for other weather‐related effects to explain the other variations, such as the El Niño–La Niña effect of the Pacific Ocean that affects the California climate; however, this effect occurs irregularly over 2–7 yr intervals, making it a less likely candidate. One astronomical phenomenon with a similar period to the 11.6 yr seeing variation is that of the solar activity tracked by changes in the sunspot counts, which has a mean period of 11.04 yr (Allen 1975). Thus far, a convincing cause for the long‐term periodic seeing variations, environmental or otherwise, has not been demonstrated.

3. CONCLUSIONS

The presence of long‐term periodic variations in the subarcsecond seeing at the Mount Wilson Observatory has been demonstrated. Historically, the best seeing of the year at this site lies predominantly between the months of July and September. In the period up until about 1970 this pattern is fairly consistent; however, from the late 1970s to 1999 the scatter in the centroid of the distribution increases.