============
Data set revision
============
The dataset underlying this archive has been revised from previous versions to identify and correct errors, review the reliability of observations, enhance quality assurance when adding new observations, develop a comprehensive error model for astrometric results, incorporate codes for modern observing techniques, include information about fitting to asteroid shape models, and change the description of a MISS event to 'No occultation detected'. The revision involved the individual review of each of the more than 8000 events. 

The revision updated a number of data fields, and added an extra data item. Two fields have transitional issues:
* Signal-to-Noise. This is a new data element, with no data being available for most events before 2019 and a continuing non-availability for many new observations. The signal used in this measure is the depth of the occultation light drop compared to the noise in the full-light signal. The value of SNR is 0 [zero] whenever data is not available. It is generally not set when no occultation event is detected (a 'Miss' event).
* Method of observation. Prior to the revision there were several categories relating to visual observations - reflective of typical observation techniques prior to 2000. Those have been replaced by a single 'Visual' category. In the revision the majority of old visual flags did not get translated into the new Visual category; it may validly be assumed that any observation where the observing method is 'Unspecified' was made visually; this is especially the case if there is a value for Personal Equation.


============
Site coordinates
============
The great majority of observations have been made in Europe, Japan, North America, Australia and New Zealand. The dataset includes a field for the geodetic datum. The field includes tags for WGS84, ED1950, NAD1927, Tokyo and GB1936 - but not AGD1966 nor NZ1949. The transition from national coordinates to WGS84 coordinates occurred over quite a few years. For many events - but especially the older events - this field is not set.

Most site coordinates based on national geodetic datums occur before about the year 2000. After 2000 most coordinates are based on WGS84, or close equivalents. When a national geodetic datum has been used, the approximate maximum offsets from WGS84 are:
* Europe (ED1950) 150m
* Japan (Tokyo) 500m
* North America (NAD1927) 150m
* Great Britain (GB1936) 200m
* Australia  (AGD1966) 200m
* New Zealand (NZ1949)  200m

The effect of these offsets is two-fold:
a. when fitting chords derived using different datums, relative chord displacements up to the above maximum datum offset might occur. However (apart from very rare trans-continental observations) this can only occur over a transitional period when coordinates from a region were being reported against differing datums. In practice (especially having regard to observational uncertainties at those times) any such 'error' in the chord placements is generally too small to be noticeable. Many such instances will have been identified in the review of the data set, with coordinates from 'Google Earth' being used where the observing site was clearly identifiable (eg an observatory).
b. the derived astrometric position of the asteroid is potentially affected. For all but the Tokyo datum, reductions based on the WGS84 datum are fully adequate at the 0.1mas level; for the Tokyo datum 'errors' of up to about 0.3mas might arise. For this reason special effort has been made to ensure all site coordinates based on the Tokyo datum have that datum set in the dataset (all instances are before 2003).

The dataset includes an event observed from the Kuiper Airborne Observatory [KAO](1978 May 29), an event observed from an F18 fighter jet (2000 Jan 10), and an event observed from the Stratospheric Observatory for Infrared Astronomy [SOFIA] (2015 June 29). For these separate entries exist for the D and R events, as the site coordinates differ as a result of the aircraft movement. The datum of the site coordinates for these observations is unknown (presumably NAD1927). The position of the KAO was uncertain by about 1km [THE DIAMETER OF PALLAS FROM ITS OCCULTATION OF SAO 85009 (1979) AJ 84 at pg.262]. For the SOFIA observation, the times and aircraft location were deduced from a conference presentation [citation not known].

An issue that has caused problems in the past is ensuring the sign of Longitude is correct. Current processing regimes are good at ensuring the sign is correct. The greatest risk of having the wrong sign occurs when the longitude is within about 10 degrees of the Greenwich meridian. While every effort has been made to validate past observations, it remains possible that a very small number of observations before about 2005 might have an undetected incorrect sign for an observer's longitude.

An independent check of the validity of site coordinates has been undertaken by comparing the reported site altitude with the altitude of that site location from standard data sources. This identified a small but significant number of sites where the reported site coordinates were in error, sometimes by more than 1 degree. Where possible corrections were made - based on advice from the observer, or from secondary information in the observation record (in particular, the field OBS_NEARBY in the [AsteroidTimes] and [PlanetTimes] files.


============
Observer names and locations
============
The data set allows for the names of up to two observers, and a flag if there were more than two observers. There is a lack of overall consistency in observer names. The majority of names include a first name (or initial) together with a family name, whereas many old observations contain just the family name. In some cultures the family name is the first element of a name; such names are included with the family name placed last. Efforts have been made to identify and correct name errors, but undoubtedly errors remain. There an additional issue of whether a person has reported using their 'true' first name or a 'familiar' version [for example, Robert vs Bob, William vs Bill]. There is also the issue of cultural differences in how family names are constructed; generally, a multiple family name is limited to the principle family name component.

For a number of events the name of the actual observer is not known - just the name of the observatory or observing group.

The field for the observer location (OBS_NEARBY in the [AsteroidTimes] and [PlanetTimes] files) is frequently empty, as a result of variable data collection regimes over time and across the observing regions. Its prime purpose is to provide a low-level means of confirming the validity of the specified site coordinates. Similarly for the field OBS_COUNTRY, which gives the relevant International Standard 2- or 3- letter country code or (for Australia, Canada, and the USA) state or province codes.


============
Shape Model fits
============
Where a diameter is derived by fitting to a shape model, we have provided a 'volume-equivalent diameter'. That is, the diameter of a sphere having the same volume as the shape model. Also the fields SHAPE_SURFVOL_x in the files [Asteroid] and [AsteroidSummary] list, provides for each shape model the Surface-Diameter/Volume-Diameter ratio - the ratio of the Surface-equivalent diameter to the Volume-equivalent diameter of the shape model. The Surface Equivalent diameter is always greater than the volume-equivalent diameter.

The fit to shape models is readily available in the [AsteroidSummary] list. All fits have been made by way of a visual comparison of the occultation chords to the shape model. Shape models have been associated with all relevant events, irrespective of whether the occultation data was sufficient to derive a fit to the shape model. 

One category of shape model fit is 'Minimum diameter'. This typically involves an occultation with only one (or a small number of closely-spaced) chords - such that the correct location of the chord on the shape model is indeterminate. We have fitted the chord to the maximum extension of the asteroid to derive a minimum diameter of the asteroid consistent with that chord - provided the chord length is greater than about 60% of the expected diameter; otherwise the shape model is flagged as 'Unconstrained'.

In the file [Asteroid], the derived diameters of the asteroid from fitting to shape models has been made solely on the basis of that event. The file [AsteroidDiameters] provides a weighted mean diameter from all events where a diameter was determined from fitting to a shape model. Where there are more than one shape model for an asteroid, this is provided for each shape model.

Differences between observed chords and a shape model can be caused by any of:
* inaccuracies in the occultation data
* inadequacies in the shape model
* the shape model being incorrect. This issue has two aspects:
  1. When a shape model is determined there are generally two different solutions for the orientation of the axis of rotation, with corresponding different shapes. Occultation observations potentially/frequently identify the 'correct' model;
  2. It may be that all available shape models are incorrect.

Where the observation was sufficient to center the shape model on the observed chords, the uncertainty in the astrometric position is set to 5% of the asteroid's diameter - to reflect the accuracy provided by the shape model, the uncertainty in the detail of the shape model, and the uncertainty in fitting the shape model to the observed chords.

Ultimately occultations and shape models have an iterative relationship, with occultation chords providing a basis for improving shape models, and the subsequent fit of the occultation chords to improved shape models providing an improved diameter determination.


============
Asteroid satellites
============
The dataset includes a number of observations involving satellites of asteroids. In the reduction process the primary solution for such events is the Separation and Position Angle of the satellite from the primary body. The file [Satellites] provides a listing of asteroidal satellite events. If the satellite was discovered by an occultation observation, the Central Bureau Electronic Telegram (CBET) number announcing the discovery is included.

Where possible and appropriate, astrometry for the asteroid is for the center of mass of the system. That location is identified on the basis of the relative volumes of the two bodies as derived using circular or elliptic fits to the occultation chords for the two bodies. 

There are several instances where the observed occultation was of the satellite only. In such cases the dataset includes a nominal entry (usually, but not always, based on a prediction) for the main body of the asteroid, with the Separation and Position Angle of the satellite being referenced to that nominal entry. Importantly:
* no astrometry is reported for the 'artificial' main body location
* the astrometric position of the satellite is derived using the measured separation and position angle of the satellite from the 'artificial' main body location - which when combined gives the location of the satellite relative to the star independent of the location of the main body.

The Astrometry file for asteroids (including comets) includes astrometric position for asteroid satellites per se only if the satellite has a formal satellite designation. The file for astrometry of Planets includes the satellites of those planets


============
Accuracy/Reliability
============
Early occultation observations were hampered by low prediction accuracies associated with the accuracy limitations inherent in star catalogues and asteroid ephemerides prior to the Hipparcos mission, compounded by the consequential effect of visual observers needing to monitor the relevant star for perhaps 10's of minutes (with associated problems of reliability). These issues have improved over time with:
* Hipparcos {including Tycho2) - providing accurate positions for brighter stars, free of zonal errors
* UCAC catalogues (2, 3 and 4) - providing accurate star positions on the Hipparcos reference frame, for stars down to magnitude 16
* Astrometry of asteroids being reduced against the Hipparcos reference frame (typically via UCAC), removing zonal error problems from asteroid astrometry
* the move from visual to video observing techniques, which commenced in the late 1990's.
* improving time reliability by way of video time insertion based on GPS time signals, which commenced in the early 2000's
* Gaia astrometry, which essentially reduced prediction uncertainty to uncertainty in the asteroid's ephemeris

This history is apparent in the dataset by way of the increase in the number of successfully observed events each year.

To validate observations, the dataset generally includes a 'prediction' line with each event. This prediction line is not a basis for computation of O-C values; indeed, the location of the prediction line depends on the ephemeris used to generate that prediction - and this is variable throughout the dataset. Rather its purpose is merely to identify erroneous observations, through the level of (in)consistency with that prediction. In reviewing events prior 2008, prediction lines were generated using asteroid elements generated using the JPL Horizons system; comments in the records may refer to such prediction lines as a 'postdiction'. In later years prediction lines are generally derived from the actual prediction for that event.

The confidence level in the results of each event is indicated by way of the 'Quality Code for fit'. The lowest level is:
 
' 0 = 'No reliable position or size'
  The quality of the observations is insufficient to allow a reliable fit to either the asteroid's diameter or position '

About 4% of the events in the dataset have this Quality code. Reasons include:
* the observation is 'wrong' in the sense that it is greatly inconsistent with the prediction line. [Probable causes include long observing periods for visual observers, poor observing conditions, the difficulties with distinguishing between short events and atmospheric effects, and difficulties in detecting small magnitude changes (especially for visual observations).]
* inconsistent observations for an event
* single-chord events, where the observer only obtained a (reliable) time for one or other the D and R events.

In the years prior to 2001 'false' observations ran between 10% and 50% of events, with the major causative factor being the combination of low prediction accuracy and visual observing for a long duration. Since 2008 the rate has been less than 4%, with a common causative factor being a partial observation (eg one event obscured by cloud) rather than an erroneous observation.
