The first ever comparison between seeing
data from RoboDIMM and the IAC DIMM (data graciously provided by the Sky Quality
Group at the IAC), using simultaneous samples from
October 2002, shows very good agreement. The median FWHM from
samples of several hours length were measured and converted to logarithms
(base 10) to produce a data couple, each representing one of 9 nights, from
which a correlation strength of up to 92% (see Figure) is obtained. The use
of a logarithmic scale to measure the correlation converts seeing FWHM into
an approximately normally-distributed variable, but it also happens that
log(FWHM) is proportional to the strength of the atmospheric turbulence, which
is what a DIMM measures.
The differences between the simultaneous samples of FWHM varies between
3 and 15% on any given night and on average is about 8%.
The average discrepancy is no larger than the internal discrepancy (between
the multiple seeing estimates) in both instruments. The good agreement is
undoubtedly helped by the length of the samples (several hours), which would
be expected to reduce the effects of local variations and the large distance
separating the two monitors (a couple of kilometers). However what is more
impresive is that such good agreement exists between two independently designed
instruments in spite of the factor 10 difference between the sample duty
cycles of the two instruments. Periods of rapidly fluctuating
seeing were generally excluded from the samples used in this comparison,
again to avoid possible local effects, and only long periods of steady seeing
conditions (judging from graphs) were used. One weakness in this test is
the lack of median values smaller than about 0.6". It should be possible
to improve upon this using data from summer 2003, as is intended.
Generally, the impression is that the average seeing published on the Weather
page agrees reasonably well with seeing being obtained at the William Herschel,
including NAOMI and the "Slodar" wave front sensors, and with that obtained
at other telescopes. It has shown sensitivity to all seeing conditions,
registering measurements as low as 0.31" and (during the passing of a recent
warm front) as high as 7 arcseconds!
Nevertheless on a few occasions this past summer, it has been observed that
the seeing obtained on the science instrument at the WHT has been significantly
larger than that measured by the DIMM. Since a ready explanation (such as
termperature differences in the dome) has not always been available, this
is a cause for some concern. The important point is that simultaneous and
long term averages need to be compared, not just single instances representing
a less than one minute sample. It is not a trivial problem and will require
careful logging of measurements to investigate.
Comparison of results from different DIMMs
One of the main controversies in site testing recently is the accuracy
in comparing results from different seeing monitors and therefore also comparing
the seeing quality at several sites. In October 2002, PASP published an
important article by Andrey Tokovinin about sources of systematic error
in measurements from DIMM type seeing monitors:
This article explains in detail how different factors introduce bias into
the FWHM estimates from the various DIMMs now being used around the world,
factors which unavoidably make comparison difficult between sites using
different monitor designs, to a precision better than some 20-30%. This
means that RoboDIMM statistics may not be comparable with previous seeing
campaigns at ING or those at other sites, to within such a margin.
Tokovinin's article suggests that the best optimistic
accuracy that can be achieved in comparison between DIMMs is about 10%.
In the test described in the previous section, the accuracy of
the measurements from each instrument, as suggested by the average discrepancy
between the two or four simultaneous estimates that each provides, is at its
greatest, about 10%. The experience of the above test suggests
that the "internal discrepancies" vary strongly (3-15%) with the instrinsic
variability of the seeing on a given night and that this factor may decide
the practical limitation in comparison tests between DIMMs rather than theoretical
derivations.
