Concordia Station at Dome C. Our experiment is in the green and gold structure in the foreground. Photo E. Aristidi.

Exceptional astronomical seeing conditions above Dome C in Antarctica

by Jon S. Lawrence, Michael C. B. Ashley, Andrei Tokovinin, and Tony Travouillon

published in Nature, 16 September 2004.

Here is the PDF version of the Nature Letter.

A simulation (more details below):

Mid-latitude view

Dome C view
View with a 2.5x larger
mid-latitude telescope
simulation of star images

Frequently asked questions (to ask a question, send e-mail to

What is "seeing"?

"Seeing" is a term that astronomers use to quantify the turbulence in the atmosphere and how it affects observations from the ground. The stars appear to twinkle because of the effect of this turbulence. In conditions of bad seeing, the stars appear to twinkle vigorously, and the images that you take with your telescope are blurry. In conditions of good seeing, the stars appear more stable, and you can take very sharp images.

Just how significant is the good seeing at Dome C?

Extremely significant! The seeing is typically 2.5 times better at Dome C than at the best existing observatories. Star images taken through a telescope at Dome C would be 2.5 times sharper and 6 (i.e., 2.5 squared) times brighter. It is similar to comparing a photograph taken with a 1 megapixel camera in a mobile phone to a 6 megapixel semi-professional digital camera.

The image on the left is a simulation of a star field as observed from the best existing observatory sites; the image in the middle is the same star field as observed from Dome C. To see as many stars from a mid-latitude observatory, you would need to build a telescope 2.5 times bigger, which would cost ten times as much, and would give the image on the right, which makes the stars look brighter but doesn't improve the sharpness of the image.

Mid-latitude view

Dome C view
View with a 2.5x larger
mid-latitude telescope
simulation of star images

Can't adaptive optics improve the images from a mid-latitude observatory?

To some extent yes, but you will always win by starting with better natural seeing. Adaptive optics is a technique for cancelling out atmospheric turbulence by using deformable mirrors (i.e., mirrors that can change shape hundreds of times per second to compensate for the atmosphere). Adaptive optics allows you to extract the maximum performance from a given observing site. However, the technique has a number of problems: it only sharpens the image in the immediate vicinity of reference star(s) or laser beam(s), it is largely limited to infrared wavelengths, it leads to errors in measuring the brightness of stars, and it is very expensive. There are no realistic prospects for achieving significant adaptive optics correction at visible wavelengths at mid-latitude observatories.

How does the sharpness of images vary with the size of the telescope?

For a telescope in space, unaffected by the earth's atmosphere, if you double the size of the telescope, you will halve the diameter of the star images, which is a good thing. However, for ground-based telescopes at mid-latitude sites, once you go above an aperture of only 0.5 m, the atmosphere effectively stops any further improvements in the resolution of the images at visible wavelengths. At Dome C you could build a telescope 2.5 times larger before running into the atmospheric limit.

What is an "arcsecond"?

An "arcsecond" is a unit of angular measurement. There are 360 degrees in a circle, and 3600 arcseconds in a degree. The full moon is 1800 arcseconds across. A star observed from the best mid-latitude observatories appears to be between 0.5 and 1 arcsecond in diameter. A star observed from Dome C would on average appear to be 0.27 arcseconds in diameter.

How would a telescope at Dome C compare with the Hubble Space Telescope?

The Hubble Space Telescope (HST) has a 2.4 m mirror, and delivers 0.05 arcsecond resolution at visible wavelengths. The best seeing we measured at Dome C was 0.07 arcseconds, however, this figure becomes lower when corrected for the finite size of the outer scale of turbulence (see Tokovinin, PASP, 114, 1156-1166). We don't have enough information yet to accurately determine the correction. While a 2.4 m telescope on the ground can never equal HST's performance, a somewhat larger telescope, say 4m, at Dome C could well produce images of equivalent resolution to HST for about 10% of the time. And in the near-infrared (e.g., the K band at 2.4 microns), the percentage should go up to 50%.

What causes atmospheric turbulence?

Atmospheric turbulence is caused by heat from the ground rising through the atmosphere, and wind stirring the atmosphere up. At mid-latitude observatories there are numerous layers of strong winds (e.g., the jet stream) in the atmosphere which cause lots of turbulence; above Dome C the winds are low throughout the atmosphere, leading to very low values of turbulence.

For how long have people suspected that Antarctica might have good seeing?

The first published predictions of exceptional seeing in Antarctica appear to be by Peter Gillingham in 1991 in Publications of the Astronomical Society of Australia, volume 9, pages 55-56. However, the history goes back at least as far as 1970, when Arne Wyller wrote a letter to the National Science Foundation suggesting that the seeing might be good at the South Pole.

How good is the seeing at the South Pole?

When we talk about the "South Pole", we mean 90S latitude, where the US has built the Amunsden-Scott South Pole Station. The seeing there was measured in the 1990's and found to be poor (averaging 1.8 arcseconds) due to a 200-300 m layer of turbulent air near the ice. Our measurements show that this layer is almost absent at Dome C. Note that if you could build at telescope at South Pole on a 300 m tower, it would have similar performance to one at Dome C.

What about the North Pole?

The North Pole is quite different from the South Pole. The North Pole consists of floating ice at sea level. The South Pole consists of a 3 km thick ice-sheet resting on rock with an area greater than the continental US. The high altitude of the Antarctic plateau is one of the key reasons for its excellent astronomical performance.

Why has it taken so long to measure the seeing at Dome C?

No humans have yet witnessed a sunrise or sunset at Dome C. The French/Italian Concordia station at Dome C has so far only operated in the summertime, when the sun is always above the horizon. To make our observations we had to build a completely robotic telescope that was able to function with no one present. The telescope was placed inside the University of New South Wales' Automated Astronomical Site-Testing International Observatory (AASTINO) - which generates its own heat and electricity using engines powered by jet-fuel. The telescope made observations from March through May 2004, and communicated its results back to us using an Iridium satellite phone.

Another reason that measuring the seeing at Dome C has only recently become possible is the development of the Multi-Aperture Scintillation Sensor (MASS) by Andrei Tokovinin and his colleagues Victor Kornilov, Andrei Zaitsev, Ol'ga Vozyakova, Nicolai Shatsky and Sergei Potanin from the Sternberg Astronomical Institute of the Moscow University. This instrument makes it possible to measure the seeing accurately with a small telescope.

Where is Dome C?

Dome C is on the high plateau in the central region of Antarctica at an altitude of 3,260 m. It is 1,100km inland from the French station at Dumont D'Urville, 1,100 km inland from the Australia Casey station, and 1,200 km inland from the Italian Zucchelli station at Terra Nova Bay. The latitude of Dome C is 75° 06'S, the longitude 123° 23'E.

What is at Dome C?

The French (IPEV) and Italian (PNRA) Antarctic programmes are constructing a new station called Concordia at Dome C. The station is almost ready for year-round operation, and is likely to operate over winter in 2005. The French and Italians have given us very generous support that enabled our AASTINO observatory to be installed in January 2003. For lots more interesting information about Dome C, have a look at Paolo Calisse's Dome C website and the official Concordia website. There are also many good photographs accessible from Tony Travouillon's gallery of images.

The image below is a closeup of the AASTINO. You can see water vapour coming from a thin pipe - this is the exhaust from the WhisperGen engines that convert jet-fuel into heat and electricity.

The AASTINO at Dome C - photo G. Venturi

And inside the AASTINO you can see the telescope we used (black and teal coloured object near the upper left), the computer system (in the rack at lower left), and one of the two WhisperGen engines (at the bottom, to the right of centre).

Inside the AASTINO - photo J. Lawrence

How did you measure the seeing with such a small telescope?

Any amateur astronomers amongst you will realise that the 85 mm aperture Televue telescope we used has a theoretical resolution of only 1.5 arcseconds, so how can we measure seeing down to 0.07 arcseconds (the lowest value we have seen at Dome C)? The trick is not to measure the star diameter directly, but to measure the variations in the intensity of a star. This is analogous to measuring the waves on the surface of a swimming pool by looking at the patterns of dark and light on the bottom of the pool caused by sunlight. The instrument we used to make this measurement was a Multi-Aperture Scintillation Sensor.

How can you measure seeing from inside a warm enclosure that presumably has a plume of hot air rising from it?

Fortunately, the MASS instrument is very insensitive to any nearby turbulence. In fact, MASS only starts to become sensitive to turbulence 500 m above the telescope (we used a sonic radar, a SODAR, to measure the turbulence below 500 m; Travouillon et al 2003 have discussed the first year of data from the SODAR).

The telescope appears to have a fixed orientation, how did you track the stars?

We used a gimbal-mounted mirror to reflect starlight from a window in the AASTINO, to a 45 degree flat mirror, and then into our telescope. The gimbal was made by Newmark Systems Inc, and is the black object near the top and slightly to the right of centre in the image below. The MASS instrument is immediately below and to the right of the JPL sticker (we have collaborators from JPL and NASA).

To find a star we drove the gimbal to the expected location, and then performed a spiral search until the star was detected by MASS. A final "half-power peakup" centered the star precisely, and we then tracked using feedback from a CCD camera (visible at the upper left corner of the UNSW crest). All of this was fully-automated and controlled by a computer running the Linux operating system.

The instrument inside the AASTINO. Photo J. Lawrence.

How did you align your telescope on the stars given that the sun was up continuously when you installed it?

Yes, this was tricky! Fortunately, it was just possible to see stars during the daytime in the MASS eyepiece (it is similar in difficulty to seeing Venus during the day). By carefully aligning the gimbal, measuring the nine angles that determine the relationship to the sky, and setting the clock on the computer accurately, there was a star visible in the eyepiece on the first attempt (i.e., the blind pointing accuracy was 0.15°). Jon was able to observe several stars over a 24 hour period, and this enabled us to produce a computer pointing model which was accurate to within 0.03°. In early February 2004 the Concordia station closed for the winter, and our only contact with the experiment since has been through an Iridium phone.

How did you test the instrument before sending it to Antarctica?

We had a hard deadline of 1 Jan 2004 to ship the instrument to Antarctica - if we had missed the deadline by one day, we would have had to wait a year. The instrument concept had only been fleshed out in July 2003 at the Sydney meeting of the International Astronomical Union, so we worked long hours over the next five months to build it and write the software. With two weeks to go before the deadline we had a working system, at which point the clouds above our laboratory in Sydney refused to cooperate. Clear patches during the day allowed us to test the tracking software on Venus. It wasn't until 29 December, 2 days before shipping, that we got half a night of clear weather to align the MASS, test the spiral-search software, etc. Everything worked, so we installed the final dowell pins to fix the instrument alignment, and then packed it all for shipping.

What stars did you observe?

We needed to observe bright, single, stars that were close to the zenith. We choose Alpha Carina (Canopus), Beta Carina (Miaplacidus), Beta Crucis (Mimosa), and Alpha Triangulum Australae (Atria).

How was the instrument controlled and the data collected?

The AASTINO is controlled by a PC/104 computer (a 300MHz Celeron with 512MB of memory and 2GB of solid-state disk) running Linux (Redhat 9). This computer connects to the Internet via an Iridium phone, which basically acts like a 2400 baud modem. A series of Java, C, C++, perl, tcl, and bash programs controls everything.

Is it true that you could control the experiment from your mobile phone?
Yes. Michael has a Kyocera 7135 phone that can login to computers on the Internet. This was mostly used to display a webpage giving a summary of health and status information from the AASTINO, but we could also send commands to the AASTINO from the phone. This ability was important since if our automated systems needed human intervention (as they did on several occasions), then we needed to take prompt action to avoid, e.g., failure of the power system.

What do the authors look like?



What is an "isoplanatic angle"?

This is the angle over which the phase fluctuations in the atmosphere are correlated. In simpler terms, if you're using adaptive optics to try to correct for atmospheric turbulence, then you will get sharp images within the isoplanatic angle, and blurry images outside this angle. The isoplanatic angle is typically only a couple of arc seconds at a mid-latitude observatory. At Dome C, the isoplanatic angle is three times larger, so the area of sky that you can correct is 9 times larger.

What is the next step?

Having shown the exceptional possibilities of the Dome C site, we believe the next step is to build and deploy a 2 m optical telescope, which we have called PILOT. We are currently exploring avenues for funding such a telescope.

What do you mean by a "2 m" telescope?

The "m" refers to "metres". So a "2 m" telescope has a 2 metre diameter mirror. The mirror diameter is the most important parameter in a telescope, since it determines how much light the telescope can collect, and how sharp the images will be.

Why only a 2 m telescope, why not a 4 m or 8 m?

The cost of a telescope is a very steep function of the mirror diameter. You could build thirty 2 m telescopes for the cost of one 8 m. While an 8 m at Dome C would be superior to any other telescope on the planet, we think it is too great a step to take at this stage. We need a 2 m "pathfinder" telescope to develop the technologies and techniques required to successfully exploit the potential of the site. Interestingly, a 2 m telescope at Dome C is very competitive in the infrared with an 8 m at a mid-latitude site.

What is PILOT?

PILOT is a our name for a 2 m telescope we are aiming to build at Dome C. PILOT stands for "Pathfinder for an International Large Optical Telescope".

Wouldn't it be very difficult to build and operate a telescope at Dome C?

Not really. See the paper "Robotic telescopes on the Antarctic plateau" by Ashley et al 2004 for a detailed discussion.

How likely is it that a telescope will be built at Dome C?

It is inevitable, particularly in the light of our Nature results. An Italian group is already constructing a 0.8 m infrared telescope called IRAIT.

Would an Antarctic telescope be just for optical wavelengths, or would it take in the infrared?

There are compelling scientific reasons to build Antarctic telescopes from optical through the infrared and out to sub-millimetre wavelengths. In the infrared you gain factors of 10 to 100 reduction in the sky brightness, due to the cold atmosphere and cold telescope. A US group is building a 10 m sub-millimetre telescope called SPT at the South Pole (at these wavelengths the improved seeing at Dome C relative to the South Pole is not such an issue).

When could PILOT be operational?

In early 2008 if funded now.

How much would PILOT cost?

Between US$5million and 7million, not including an instrument package. The cost of transport to Dome C is about US$50k.

What are the funding options for PILOT?

The cost is too great for existing grant schemes in Australia, so we are exploring a range of options including an Australian Research Council Centre of Excellence proposal led by Professor John Storey, to be submitted in October 2004. We would like to raise US$2million in cash from non-government sources.

What happened to the Douglas Mawson Telescope proposal?

The Douglas Mawson Telescope was a proposal of ours in 2001, in collaboration with France and Italy, to build a telescope similar to PILOT at Dome C. It missed out on funding from the Australian government. At the time, the seeing conditions at Dome C were a matter of speculation, which is why we have put so much effort into obtaining the latest data. Our Nature paper now makes the case overwhelming.

Is there any independent verification of your results? Extraordinary claims demand extraordinary proof.

Until Concordia station opens for the winter, which may happen as early as 2005, no one will be able to duplicate our measurements. However, over the 2003/04 summer, the University of Nice group consisting of E. Aristidi, A. Agabi, J. Vernin, M. Azouit, F. Martin, A. Ziad and E. Fossat obtained summertime seeing data using an entirely independent technique. They saw periods of seeing as good as 0.2 arceconds in bright sunshine, which is unprecedented for an observatory site. Furthermore, balloon launches during summer have shown that the wind profiles as a function of height are consistent with superb seeing.

Would PILOT be an international collaboration?

Absolutely. One reason is that Concordia station is a French/Italian venture. John Storey is in Europe this week (11-18 September 2004) putting together a coordinated international program. We have already attracted collaborators and support for our Centre of Excellence proposal from some of the world's leading astronomical institutions, e.g., Arcetri, Cornell University, NASA Ames, University of Arizona, University of Nice, MPIA-Heidelberg, as well as Australian support from the University of New South Wales, the University of Sydney, the University of Tasmania, the Anglo-Australian Observatory, the Australian Antarctic Division, and industrial support from Electro-Optic Systems, Sinclair Knight Mertz, and Connell Wagner.

What unique science are you targeting with PILOT?

We aren't targeting any particular astronomical problem with PILOT. PILOT should be seen as a general-purpose telescope that simply has better performance by far than a similar telescope at a mid-latitude site. It is also a testbed for a future 8 m telescope. The science that we end up doing with PILOT will depend on the instruments that our collaborators contribute.

What are the main ways that an Antarctic telescope could benefit astronomy?

The main benefits are much greater performance at much lower cost. While there is an additional cost to "winterize" a telescope design, this is offset by far superior performance and savings in other areas (e.g., there is no need for a traditional dome structure). See Ashley et al 2004 "Robotic telescopes on the Antarctic plateau" for a comprehensive discussion of the advantages and disadvantages of Antarctica.

Another aspect is that Antarctica may be the only viable site for the next generation of Extremely Large Telescopes. The reasoning is that the low winds (yes! on the plateau it is very calm) and lack of seismic activity in Antarctica greatly eases the engineering challenges in constructing huge (30 m or greater) telescopes.

Are there any astronomical problems you could tackle from Dome C that you couldn't do elsewhere?

Yes. Wide-field, high resolution surveys in the optical and infrared would be a natural fit to the superb seeing conditions. However, interferometry is the field that has the largest potential gain. An interferometer at Dome C could be designed to directly image planets around other stars. Also, an interferometer designed for narrow-angle differential astrometry could directly determine the nature of the planetary systems around nearby stars; this is currently our best hope for detecting earth-like planets. The nature of the atmospheric turbulence at Dome C makes such interferometer projects many hundreds of times more sensitive than similar instruments at mid-latitude sites, which effectively makes the science feasible.

If we consider telescopes that operate at longer wavelengths, e.g., the sub-millimetre, there are whole new regions of parameter space that open up due to the increased transparency of the cold Antarctic atmosphere.

What is an interferometer?

An interferometer is an instrument consisting of two or more telescopes, separated by perhaps 100 m or more, that are optically linked to act as one.

Who are the members of the UNSW Antarctic astronomy group?

The Nature paper is brought to you by the University of New South Wales Antarctic astronomy group.

Can you tell us something about the computers you used?

To operate our experiments over winter, when there was no one at Dome C, we had several problems to address:
  1. For hardware reliability, we wanted to remove all moving parts in the computers, i.e., no disk drives, and no fans. So we used a small PC/104 form-factor computer system with solid-state disk drives.
  2. We had to generate our own electricity. We took two approaches to this:
    1. One experiment, ICECAM, relied entirely on a 5 kg pack of lithium thionyl chloride batteries. The batteries had to provide power for a year, so minimized the power consumption of the computer. The experiment only needed to take data every two hours, so we built a CMOS oscillator to power-up the computer for 30 seconds every two hours. We used MS-DOS 6.22 for the PC/104 computer since it boots quickly and was able to average 10 frames from the CCD camera and store them to CompactFlash disk.
    2. For the experiment that obtained the seeing results, the AASTINO, we needed much more power, up to 400W, and we had to operate continuously, so we used stirling engines running on jet fuel. For software reliability we chose Linux, Redhat 9 to be precise. Software and hardware watchdog timers helped to ensure that the system would recover from most failure modes.
  3. The ambient temperature at Dome C reaches a low of -85°C during winter. Computers, and electronics in general, are not designed to operate at these temperatures. We took two approaches:
    1. With ICECAM we had no reserves of power for heating, so we buried the computer in a crypt seven meters below the ice surface, at which point the temperature is stable at the yearly average of -57°C. This is still outside the computer's specfication. Fortunately, a test in a low-temperature fridge showed that the computer and solid-state disks worked reliably at these temperatures. ICECAM's camera, a Watec 902-HS, had to remain outside, and tests shows that it was able to operate flawlessly down to at least -80°C.
    2. With the AASTINO, the stirling engines produced up to 6 kW of waste heat, which we utilized to maintain a comfortable operating temperature of about -10°C.
  4. Internet connectivity was provided by an Iridium phone, which acts like a 2400 baud modem.
  5. The hardware and software had to be carefully designed so that we could recover from most problems remotely. There were no reset buttons to press, and no prompts to "click OK to continue".
The PC/104 computers we used were made by DSP Design, however it in unclear whether the company still exists, since all our attempts (using e-mail and filling out their laborious on-line enquiry form) over the past 6 months to have a simple technical question answered have received no response.

Can you describe how the fuel was topped up during the year?

We have insufficient space inside the AASTINO to store enough jet-fuel to last for a full year, so we built an external fuel tank. The problem is that jet-fuel turns to gel at temperatures much below -50°C, which is why we wanted the main fuel tanks to be inside the AASTINO where it is warm. We set up a crontab entry on the Linux computer to start a pump on two occasions during the year for a fixed number of seconds in order to transfer 1000 liters of fuel from the external to the internal tanks. If the pump hadn't turned off, the internal tanks would have overflowed. So there was considerable nervousness when the moment of the transfer approached. Fortunately, everything worked perfectly, and we were able to verify by the 20°C drop in the fuel tank temperature that the transfer was successful.

Can you tell us about the dramatic events of 17 May 2004?

By 17 May 2004 the AASTINO had worked remotely for 100 days in 2004, and then something went wrong...

The WhisperGen engine has a control panel connected to it using an RS-485 bus running on CAT-5 cable. The control panel contains a microprocessor, and the engine expects to communicate with it regularly (at least once a second). When this communication is interrupted, the engine shuts down and reboots its own microprocessor.

Unfortunately, this is what happened on 17 May. - the engine went into a cycle of rebooting every 40 seconds. Once the engine has stopped, we had a ten hour window in which to try to restart it before the 200AH lead-acid batteries would lose too much capacity and become too cold for a restart (which requires 15A at 24V for about 15 minutes).

During this period we worked feverishly to come up with a solution. Our first priority was to shut down all unnecessary power consumption in the AASTINO - we can do this via a series of Dallas one-wire switches which control power to all the subsystems. A call to the engine manufacturer came up with the suggestion that we wiggle the CAT-5 cable connection - we suspect they forgot that we were over 4000 km away from our engine!

The PC/104 computer was also on the RS-485 bus, and we reasoned that by rewriting the Linux device driver (which we had written in the first place, so we knew what we were doing) we could make the computer impersonate the control panel, and convince the engine that it should keep running. Fortunately, we had a snapshot of the communication traffic between the engine and the control panel from earlier testing in the lab with the manufacturer's MSDOS-based software. But with no hardware available to test our code, we had to modify the driver, send patches over the 2400 baud Iridium link, and rmmod/insmod the driver to try to restart the engine.

All the while, the internal temperature of the AASTINO was plumetting towards ambient, at about -60°C. We first modified the driver to allow the link traffic to be analysed, and this confirmed the communication problem with the control panel. After several attempts at generating fake packets from the control panel, punctuated by breaks in the Iridium link and agonizing waits for the system to redial (it is dialout only, controlled by a crontab entry), we were unable to prevent the engine from rebooting.

We watched helplessly as the battery temperatures sank below the minimum threshold for engine restart. Over the next 24 hours we received the occasional connection from the AASTINO computer, but that was all. We are now hoping that the solar panels will be able to recharge the batteries suffiently to re-establish communication before the Dome C station opens for the summer.

Do you have any nice photographic images of Dome C?

Yes, please look here for selected images. and here for some higher resolution images (including the one at the top of this page). You might also check Tony Travouillon's 2003/04 gallery, his 2002/03 gallery,
Paolo Calisse's Dome C website and the official Concordia website.

Do you have any other information on-line?

You might find the following links to published papers interesting.

Angel 2004 Buyer's guide to telescopes in the best sites: Dome A, L2 and Shackleton Rim (also discusses Dome C).
Ashley et al 2004 Robotic telescopes on the Antarctic plateau
Lawrence et al 2004 Infrared and sub-millimetre atmospheric characteristics of high Antarctic plateau sites
Lawrence et al 2004 A robotic instrument for measuring high altitude atmospheric turbulence from Dome C, Antarctica
Lawrence et al 2004 A remote, autonomous laboratory for Antarctica with hybrid power generation
Travoullion et al 2003 Low atmosphere turbulence at Dome C: perliminary results.
Travouillon et al 2004 Ground Layer Adaptive Optics performance in Antarctica

Michael Ashley / / last updated 16-Sep-2004