Quickstart Guide to using AAVSO data
This is a short guide for researchers interested in analyzing data from the AAVSO International Database (AID). It is intended to be an introduction to the data found in the AID, as well as a guide to help you begin utilizing these data in your research. Data in the AID are heterogeneous in nature, and all users of AAVSO data are encouraged to work with the AAVSO to ensure that the data are being interpreted correctly. All researchers are encouraged to contact the AAVSO at any stage of their research for assistance.
The visual observations found in the AID are obtained by individuals looking at a variable star field through a telescope, binoculars, or their naked eyes, and comparing the brightness of the variable to those of predetermined nearby comparison stars. You will find that the apparent scatter in visual data is larger than in most instrumental data that you may have worked with in the past. However, visual estimates made by large numbers of visual observers using a common set of comparison stars are normally distributed, meaning that their statistical interpretation is as straightforward as that for any other type of data. You will find that visual observations can be used to obtain periodic signals below 0.1 mag for many stars, and that visual data have very wide application to problems of astrophysical interest.
Basic statistical properties
Visual data are typically reported to the AAVSO in increments of 0.1 magnitudes, although some data are reported to 0.01 magnitudes. Ocular sensitivity varies from person to person, but the human eye is capable of distinguishing differences in brightness of a few hundredths of a magnitude when suitable comparison stars exist and the observer is highly skilled. When visual data are binned in time spanning 7-10 days, the resulting sample standard deviation is on the order of 0.1 magnitudes for most stars, and 0.2-0.3 magnitudes for particularly red stars (e.g. carbon-type Mira variables). When stars are well observed, multi-day averaged curves of long-period stars may have a standard error much lower than this -- 0.02 magnitudes or less. Times of observations are typically reported in Julian Date (without any heliocentric or barycentric correction) to a precision of a tenth of a day. Some observers report observation times accurate to the second; this is typically done for short period stars where the exact time is important (e.g. binaries or RR Lyrae).
It is important to note that the quality of light curves may vary from star to star, and is highly dependent upon a number of factors, including:
- Brightness, color, and variability range
- Number and suitability of near-field comparison stars
- Historical charts and the number of chart revisions
- Number and skill levels of observers
The AAVSO undertook a large-scale campaign to address some of the data quality issues for all visual data submitted prior to 2001. Most but not all obviously discrepant data points submitted prior to 2001 have been marked as such, and you may choose not to include such flagged data when you download a light curve. Data submitted since 2001 have not been fully validated, although some stars have been validated on a case-by-case basis. If you have any questions about an individual star's light curve, please contact the AAVSO.
There are a number of papers in the JAAVSO that discuss visual estimates and compare them to instrumental photometry such as photoelectric Johnson V. For a partial list of these papers, please see this bibliography.
For an overview of how visual estimates are made, and the various physiological phenomena that impact the statistical properties of visual data, we also recommend this article by Jean Gunther and Emile Schweitzer, both observers and former Councillors of L'Association Française des Observateurs d'Étoiles Variables (AFOEV).
When a large number of observers have contributed observations of a particular star, it is often useful to form mean curves by binning the visual observations in time. The temporal size of the bins must be a small fraction of the variable's period or timescale of variability. For long-term visual light curves of long-period variables (Miras and semiregulars), bins of 7 days are ideal; a week-long binsize will cancel out any potential statistical differences between observers who observe only on weekends and those who can observe any night of the week. We typically bin visual observations in magnitudes for the sake of simplicity, but the most rigorous method for binning data would be to convert them to flux units first. In reality, where the standard deviation of the magnitudes in each bin is small, the difference between magnitude averaging and flux averaging is also small.
When assessing the quality of visual mean curves, one should use the standard error of the mean, sigma/sqrt(N), as a metric; sigma is the standard deviation of the observations in the bin, and N is the number of observations in the bin. These will result in error bars on mean curves on the order of a few hundredths of a magnitude. Note that this carries the assumption that the visual observations are normally distributed, which may not be the case when different observers are using different sequences or have very different sets of equipment or experience. Such problems are often clearly apparent when raw visual data are inspected by eye, and you may be able to remove discrepant points prior to forming mean curves. When in doubt, please contact the AAVSO with any questions you may have. A good practice to follow is that your eyes examining the raw data will do a better job of detecting and understanding any inconsistencies in visual data than most data analysis packages will. You should visually examine any light curve -- visual or instrumental -- prior to analysis.
Long-term light curves
Long-term light curves are incredibly valuable for studying variability on timescales of several decades or longer. However, long-term light curves have a higher risk for being heterogeneous in nature, particularly due to improvements in comparison star sequences and calibration over time. AAVSO visual data are stored as magnitudes, rather than as "step magnitudes" (similar to differential magnitudes) relative to a comparison star, and so some of the original calibration data are lost. Researchers should be careful when examining long-term light curves to make sure that sudden jumps in behavior are real and not due to changes in sequences. Again, when in doubt, please contact the AAVSO prior to analysis and publication of any questionable segments of data.
The AID online database contains most of the AAVSO data holdings for variables in the early Harvard College Observatory program, with the notable exception of observations previously published in the Harvard Annals through 1920. The AAVSO is planning to digitize the observations in the Harvard Annals, which may extend light curves backwards in time by a few decades or more. We do not have specific plans to digitize other published data (e.g. Astrophysical Journal, PASP, or Astronomisches Nachrichten prior to 1900), but will often enter digitized visual data when submitted to the AAVSO by volunteers or other individuals or groups. Note that we are very interested in having older published data digitized and archived, and are eager to partner with groups or individual researchers in securing funding to do this.
Many hundreds of long-period and cataclysmic variables have visual light curves that start between 1900 and 1920. Another large group of stars have light curves that began in the 1960s. Depending upon the needs of your study you may find several stars of a given class with a given set of properties having well-sampled light curves between 50 and 100 years in length. It is entirely possible to do large-scale, systematic studies of variable star classes using such light curves; as an example, see Templeton, Mattei, and Willson (2005; AJ 130, 776). The AAVSO is particularly interested in seeing its long-term visual archives put to good use, and we may be able to facilitate the work of researchers interested in doing large "surveys" of AAVSO data. Please contact us if we can assist you.
You can read a broad overview of available AAVSO data at this page: Overview: Long-term visual light curves.
The AAVSO has CCD data collected from amateur telescopes starting around 1995. The increase in quality of consumer-grade CCD cameras and the decrease in their price has led to an enormous increase in the amount of CCD photometry produced by AAVSO observers. The AAVSO currently receives approximately one million CCD observations per year. A large fraction of these observations are time-series observations of specific stars. The vast majority (> 99%) are obtained via differential photometry rather than all-sky photometry. While a significant fraction of the data are taken using standard photometric filters (e.g. Bessell, Johnson-Cousins), only a small amount of the total are fully calibrated, transformed magnitudes in the standard system of the filters used. Most are filtered magnitudes without zero-point corrections or nightly extinction corrections. For that reason, caution must be exercised when using untransformed data, and when combining the observations of multiple observers. For many practical applications (e.g. analysis of differential time-series photometry) you will find that AAVSO data are largely identical in quality to photometry obtained from professional research observatories, and the reduction techniques applied to standard differential photometry are easily applicable to AAVSO data.
Data obtained by amateurs can be of very high quality with most observers capable of attaining photometric errors much less than 10 mmag on brighter stars. The limiting factor on absolute precision is typically calibration: transformation to a standard system and proper extinction correction. (Jump to calibration.)
Amateur data are as subject to errors as professional observations. An an example, many of the analysis tools used by the amateur community do not clearly flag pathological data (e.g. saturated stars) when these tools are used to automatically reduce large image sets from a given run. As a result, some AAVSO CCD data show much larger scatter than would be expected; this commonly occurs at the bright end. Many photometry packages also do not calculate photometric errors, and it is common to find data in the AID where each data point has been assigned the same rms error. As with any other data set, AAVSO data should be critically examined prior to use in any analysis.
The Julian Date (JD) is the time standard for all data in the AID. We have specified that the observers should be submitting data to the AAVSO using the mid-point of the exposure time for all CCD exposures. This specification was also given to all CCD and image processing software vendors whose software extracts photometry from CCD data and exports to the AAVSO Extended Format.
The data are always provided to the researcher with the Julian Date as the primary time stamp for the observation. When observers submit data in HJD, we store their HJD and also provide it to those who download the data, but we then populate the JD field by backing out the heliocentric correction after the fact. Depending upon the analysis you plan to do and the span of data you plan to use, we suggest using the JD and converting the observation times to HJD. Integration times for CCD observations are rarely shorter than one second, and so higher order corrections are rarely appropriate or necessary. Note that the HJD is irrelevant for observation times given to a precision of less than 0.01 days.
CCD photometry submitted to the AAVSO are a mixture of filtered and unfiltered photometry. In all cases we list filter information as provided by the observer. Filtered data will be noted as such, and our filter designations conform to standards (e.g. Johnson-Cousins, Sloan). When data are unfiltered, we encourage the observer to note whether their comparison star magnitudes are given in the V bandpass or the R. Such data will be marked as "CV" or "CR" (for "clear filter, V zero-point" or "clear filter, R zero-point"). More recently, observers have been submitting data using the RGB filters commonly sold with imaging cameras. Such filters do not conform to standards, but are designed to create true-color images. Such data is designated by the filter names "TR", "TG", and "TB" (tri-color R,G,B). Rarely we may have data for other filters, such as an orange photographic filter, older Johnson-perscription R and I, or Wing IR filters. If you have any questions about the filters with which data are taken, you should consult our list of available filters or contact us directly.
Most observers do not perform a full transformation and extinction correction of their data even when using standard photometric filters. When observers use comparison stars of similar color and restrict their observations to low airmasses, the data can be remarkably consistent from observer to observer, with the sole limiting factor being a simple zero-point correction. However, this is not a rule, and observers may be using inappropriate comparison stars over a larger range of airmass than is warranted. Researchers should be aware that most data are not transformed or extinction corrected, and should treat them as such. For many purposes (e.g. time-series analysis) zero-point corrections are not required, and the data may be used with a simple transformation to a common zero point.
The status of calibrations is given in the "transformed" field; if the data are transformed, they will be indicated as such. Fully transformed data is rare in the AID, making up only a few percent of the total. When using filtered data, you should be aware that even "V-filtered" data are not necessarily transformed to the standard Johnson-Cousins system, and that there may be other effects such as differential extinction present. If you have questions about this, please contact the AAVSO for assistance.