Known Issues with the TMT Site Testing Data
This page lists known problems with the data in the TMT site testing database.
When we notice an issue, we document it here. It is nevertheless possible
(in fact, likely) that some problems exist that are not
listed here. We therefore ask that you contact us if you find something strange in the data.
We can then tell you if we have encountered this issue before and/or add it to the list below, so that other
users will be aware of it also.
- General Philosophy: Data on this website are not filtered for potentially invalid or corrupted points, as this would alter
some of the statistics of the data set. Data that are invalid for one analysis might have to be included for another, so as to not arrive at
misleading or incorrect results. Even clearly invalid data points might still carry information (if nothing else that the weather was good enough to
operate the respective instrument). We have therefore not excluded any data from the database even if there are known problems with them.
Instead, these problems are reported on this page.
- Invalid measurements #1, NaN's: If an instrument identifies a measurement as invalid (for example, because the extraction of a parameter
was not possible due to low signal-to-noise ratio), the respective data point is indicated by a series of **** symbols by the database.
- Invalid measurements #2, outliers, etc.: All data sets might contain periods when "something went wrong" and strange but theoretically possible values are recorded.
If this occurs the automated system does not know that something is wrong.
These cases might appear as non-sensical values, as outliers, as periods when a parameter has a constant value for a long time or the like. Such periods should generally be discarded, but there are exceptions, such as periods when a wind speed is so low for an extended amount of time that it is below the detection limit of a mechanical anemometer.
If you encounter questionable data of this kind, please check whether the problem is listed below or contact us, as we are likely aware of the problem and know how to deal with it.
- DIMM data quality criterion: The seeing measured by the DIMMs is larger than the real seeing if the optical system
is not sufficiently well aligned. Metrics for the quality of the alignment are the Strehl ratios of the DIMM images (note that it is not correct to use a constant Strehl cutoff, as the Strehl ratio also depends on the seeing), but other criteria
like the distance of the two star images on the detector can also be used. For the TMT data, only a very small fraction of the data taken at the candidate sites
are affected by this, but some of the early data taken at Tololo need to be taken with care. For details see:
L. Wang et al., "High-accuracy differential image motion monitor measurements for the Thirty Meter Telescope site testing program", Appl. Opt. 46
(25), pp. 6460-6468 (2007)
- DIMM Strehl value correction: The values of the DIMM Strehl ratios in the database are calculated for the effective wavelength of the DIMM CCD, which is ~625 nm.
They are the Strehls two-dimensional images would have estimated from our one-dimensional DIMM images.
However, due to a bug in the code (incorrect pixel scale), all values need to multiplied by 0.722.
- MASS data reprocessing:
The data in the MASS database have been reprocessed to take into account the effect of strong scintillation and correct analysis
parameters. The data file as well as the download button are therefore called 'mass_redo'. This is no cause for alarm.
Reprocessed MASS data in the data base have not been filtered for, e.g., Chi2 or flux.
MASS data were collected using turbina 2.047 and reprocessing was done using turbina version 2.052
- MASS/DIMM data - looking through 30-m tower: For some time in early 2007, the
telescope at Armazones (T2) was occasionally observing through the 30-m tower. This affects
MASS/DIMM photometry data. MASS seeing measurements might be affected by a small
amount. DIMM seeing measurements should be unaffected. The star list was changed after this problem was discovered.
List of affected data
- Values of invalid MASS data: The MASS software turbina reports some invalid data points as non-sensical values, such as 0 or 99999.
MASS data should therefore be screened for potential occurences of such values.
- AWS and sonic anemometer wind direction offsets: Some wind direction measurements of the automated weather
stations (AWS) contain artificial offsets,
some of them significant, because of problems with or misalignments of
the wind direction sensors. The sonic anemometer wind directions also contain offsets in multiples of 90 degrees. This is due to the orientation of the sonics (the measurement is with respect to an internal coordinate systems) and needs to be corrected.
- AWS wind speed and wind direction sometimes affected by frozen/stuck sensors:
The AWS wind sensors were sometimes affected by ice, which in extreme cases made them to
read zero wind speeds or get stuck and show steady wind directions.
Also some of the wind speed sensors were affected by dust, which causes them to underestimate
the actual wind speed.
Here is a report describing these effects and
how to identify such periods.
- AWS wind direction sensor at San Pedro Mártir (T4-SPM) gave a lot of zero readings.
This are mostly due to problems with the sensor.
Exact 0.00 deg readings from the T4-SPM wind direction sensor should be discarded.
- Early Tolar AWS data: In the early days of operation, invalid AWS data were indicated by a value of -99 instead of as NaN (****).
The points therefore appear as -99 in the database.
- Sonic anemometer readings are affected by surrounding structures:
If the wind comes out of a direction from which it first has to flow around a structure
which is close to the sensor head of the sonic anemometer, wind direction and wind speed
values will be unprecise.
The telescope dome affects the 7m sonic anemometer readings and the tower sonics are
affected by the 30-m tower. Also there is an effect of the sonic anemometer sensor
structure itself. Orientation of sonic anemometers
with respect to structures closeby.
- Sonic anemometer wind directions: The calculation of
sonic anemometer wind directions from the wind speed components vx and vy,
is described in this document.
Also: (example code for computation).
- Sonic anemometer wind speeds larger than about 50 m/s are invalid. Such wind speeds
come close to the sampling limit of the instruments and need to be discarded. This affects a negligibly small fraction of the data.
- Relative humidity sensor non-linear at around RH=25%: All relative humidity sensors
show a spike in the histograms of their readings in the region between 20 and 30%.
This is a known property of this particular sensor.
- Relative humidity sensor: The RH sensor at Mauna Kea 13 N (T6-Hawaii) reads values RH>100%.
When this occurs, one should set RH>100% -> RH=100%
- Dust sensor data larger than 3e6 are invalid. These are beyond the detection limit of the instrument.
Also, these sensors had the tendency to clog up, so a lot of zero readings
might also have to be discarded, although some of them do correspond to real periods of low dust activity.
- Ground heat flux sensor: The ground heat flux sensors were mounted
upside down at Armazones, San Pedro Mártir and Mauna Kea 13 N. These readings should be multiplied by -1 in order to be
compliant with the meteorological definitions. The sensor
at T3 had technical problems (lightning, cable chewed) and data before 21 September 2006
should be discarded.
- 30-m tower anemometer altitudes on T2-Armazones: Some of the altitudes of the tower wind measurements
are reported incorrectly. The height of the lowest anemometer is always 11 m, even though a part of the data set shows 10 m.
The height of the highest anemometer is always 28 m, even though a part of the data set shows 30 m.
Approximately 0.4% of the data also show non-sensical values between 2 and 3 m. These data points should be discarded.
- 30-m tower anemometer altitudes on T3-Tolonchar: The height of the lowest anemometer is at 11 m above the ground, even
though the values in the DB indicate 10 m.
- 30-m tower instrumentation on T2-Armazones and T3-Tolonchar:
Consult this file for the necessary information
in order to make use of the data collected on the 30-m towers.
This includes information about the height of the sensors on the tower and periods of invalid data.
Also consult the sensor calibration report.