Chapter 5

Chapter 5, Technical Studies

Section 5.2: Site Testing and Selection

NOAO Logo    Gemini Logo


A state-of-the-art and very expensive telescope such as the GSMT should clearly be placed on the best possible observing site to maximize the scientific return from the investment. However, such a choice is always a trade; apart from meteorological and climatological considerations, there are political, environmental, and cost criteria that need to be taken into account. These criteria interact in a complex way, and generally their relative importance will change with time, often unpredictably.

At this stage, it is essential to define and prioritize the site requirements based on the priorities flowing from the science requirements, combined with knowledge regarding the interaction of GSMT design with site properties. Because these will continue to be defined and refined through the early stages of the project, the site selection process needs to be staged, proceeding from a general collection of information in the first stage to detailed testing of a single site in the final stage. This final stage should continue throughout construction and interface seamlessly to the site measurements required in the commissioning and operations phases. A major difference between the GSMT and earlier telescopes is the critical reliance of the GSMT on adaptive optics (AO). As a consequence, atmospheric turbulence vs. height, as well as cirrus cover, must be well understood.


Earlier site surveys can provide valuable guidance. The most important lesson is that even well- funded and multi-year surveys (of which the best example is that conducted by ESO (European Southern Observatory) to identify and evaluate a site for the VLT (very large telescope))1 spend more effort on characterization than on identification, because the process of conducting even initial tests on multiple sites is very expensive and time-consuming. This argues for the use of remote sensing methods as much as possible in the initial selection process for the GSMT site. Secondly, long-term climatologic changes are not well understood. The well-known adage that observing sites can only get worse with time is an example, because the typical selection process of only a few years poorly samples the long-term conditions. Given a choice of sites of not too- dissimilar quality, the one chosen may only be ephemerally better than the others. Long-term changes, including "global warming," can significantly change site properties (cloud cover, wind direction, etc.) that are determined by a delicate balance between weather systems (see Figure 1). Island sites are generally more immune to such effects than are continental sites.

Figure 1 The difference in autumn precipitation for the continental United States between 1961-
1990 and 1931-1960. Note the large differences for southern Arizona and California. Although
the mechanisms for long-term changes such as this are known, the underlying causes are poorly
understood (see Huser and Stowe). (ref 2)

The best observing sites in the world, judged on the basis of a combination of atmospheric stability and transmission, are of two types: isolated high mountains in temperate oceans (e.g., Mauna Kea and La Palma), or coastal mountain ranges near a cold ocean with stable subtropical anti-cyclone conditions (e.g., coasts of Chile, Western USA-Mexico, and Namibia). Continental mountain sites (e.g., USA, Uzbekistan, China, India, and South Africa) appear not to be as good, although this general evaluation depends on the weighting of site characteristics. For example, Mt. Maidanak in Uzbekistan has good seeing and low wind.3 There are possibly other high-quality sites remaining to be fully evaluated (e.g., in North Africa and Northwest Argentina). Specialized sites (e.g., Antarctica, Chajnantor, both with very low water vapor column) are being developed, and even more exotic sites (e.g., a tethered balloon in the stratosphere, Greenland Ice-cap) have received some attention but are irrelevant for GSMT.

Most existing observatories monitor basic meteorological quantities such as the number of clear nights, wind speed and direction, temperature, and seeing. With few exceptions, the easily available records consist of mean quantities. Often there are differences in definition, particularly for quantities such as "photometric time." Thus, comparing existing observatory sites on the basis of published measurements is generally difficult. However, existing observatories do provide the location and infrastructure to ease the task of making new measurements. Remote sensing techniques, particularly from satellites, have greatly advanced in the past 25 years, as has hydrodynamic modeling of wind-flow over topography, at a variety of scales. Use of these techniques should allow the dispassionate identification and comparison of sites, and easy re-evaluation if the weighting of the comparison criteria changes. If sites can truly be pre-selected in this way, the expense of the time-consuming initial site identification phase should be reduced. Databases allowing the global evaluation of such important parameters as earthquake probability, volcanic activity, atmospheric dust content, and light pollution are available (e.g., Figure 2).The Meteosat and GOES (Geostationary Operational Environmental Satellite) series of satellite provide cloud cover and water vapor column measurements. Meteorological data on a 3-D global grid, used as input to operational meteorological models, have been collected for 45 years and in principle can be analyzed to determine wind speeds and direction, temperature changes through the night, and low resolution CN2 profiles. (CN2 is the refractive index structure constant, a measure of the strength of atmospheric turbulence.) We have funded remote sensing analyses to supplement existing databases, as detailed below. The use of Geographical Information Systems (GIS) as a tool for storing, combining, displaying, and evaluating large volumes of descriptive data has been proposed by Sarazin.4 Multiple data planes (e.g., longitude, latitude, value) are readily manipulated via use of a GIS.

Figure 2 Depiction of a typical global database (in this case, showing seismic hazard).


In addition to GSMT, there are several new telescope projects in the early stages. The Overwhelmingly Large (OWL) Telescope project (ESO) and the CELT (Caltech-University of California) project are for telescopes as large as or larger than GSMT, with roughly similar site requirements. Collaborative and complementary work is advantageous to all parties. The ALMA (Atacama Large Millimeter Array) and LMT (Large Millimeter Telescope) (mm wavelengths) projects are more advanced, with sites already chosen. ALMA will be sited in the Chajnantor Science Preserve, Chile, while the LMT is being built at Cerro La Negra near Puebla, Mexico. Both sites are high altitude, and were chosen for their extreme dryness during a large fraction of the year. Cornell University has been planning to build a 15-m telescope optimized for the mid-IR on a peak above the ALMA site (Giovanelli et al. 2001),5,6 but with a large group of possible partners is now considering a larger (TMT, twenty-meter telescope) instrument, possibly at lower altitude.  The University of Tokyo is also interested in installing a 6.5-m telescope above the ALMA site.  UNAM (Mexico) has developed concepts for a 6.5- m segmented mirror telescope, with its site nominally San Pedro Martir in Baja, California. All of these sites are worthy of consideration for GSMT.

The GSMT sites team has deliberately chosen to geographically restrict their efforts to the characterization of the Americas plus Hawaii. This decision is a consequence of the limited resources available, the already-existing superb sites in these regions, and the stated intent of ESO to consider sites in other continents for the OWL project.

Within the Americas plus Hawaii, there are three areas to be considered: the Southwestern USA and Northern Mexico, Northern Chile, and the summit of Mauna Kea (MK). The many observatories in these regions account for 80% of the world's telescopes with apertures larger than 3 m. All 12 presently operating telescopes with apertures larger than 6.5 m are in these regions. Relevant characterization work is proceeding at existing facilities. For example, the sodium layer measurements required as part of a new multi-conjugate adaptive optics (MCAO) facility for Gemini South are being carried out on Cerro Tololo, while the extensive weather and site monitoring by ESO at La Silla and Paranal are relevant also for any new facility contemplated at or near these sites.


The key measurements for the GSMT are listed in Table 1. Weighting will depend on the scientific priorities and telescope technical characteristics, together with any implied construction or operations restrictions. For example, a site at over 5000 m may have severe cost implications. The result of this weighting may be very different from existing telescopes. As an example, GSMT will likely have a primary resonance frequency closer to the maximum of the wind power spectrum than the present generation of large telescopes, and thus be more sensitive to wind- buffeting. Survivability issues - in particular, the frequency of strong earthquakes, the maximum tolerable wind speed, and the tolerable snow load - are likely to be important for GSMT.

It is clear that GSMT will be strongly reliant on AO, even if its scientific use is mostly spectroscopic rather than imaging. Thus, the photometric quality and the turbulence properties of the atmosphere will be extremely important. A long AO time constant is desirable for a large telescope, which implies low wind at high altitude.

Our strategy is to collect relevant data for candidate sites so that a reliable comparison can be made after the scientific and technical characteristics of the GSMT are known. In the early stages, the range of site properties of the "candidate sites" is quite large, reflecting the uncertainty in the telescope characteristics.

Table 1 GSMT site parameters.
Top LevelParameterRemote

AtmosphereCloudsatelliteconfirmRadiation (daytime)

PWVsatelliteconfirmHumidity, weather station

TurbulencemodelYesMASS Profilometer, GSM, Scidar, DIMM, Balloons

Windmodel?YesGround, near-ground, weather station

Sodium layer
YesGemini, for Cerro Tololo

Climate stabilitymodel

LocationAltitudemaps, etc.

General topographymaps, etc.

Local topographymaps, etc.confirm

YesWeather station

Seismic activityYes
contracted study

Volcanic activityYes
contracted study

Snow, rainsomeconfirm

Ground layer turbulencemodelYesSODAR?

Yescontracted study

AirsomeYesParticle counter

Ownershippublic record
contracted study

Mining rights

contracted study


Environmental issues

Cultural issues

Population predictions

Industrial development

Political issues

Construction phase issues

Operation phase issues


The above strategy translates into the following tasks:

The above program can be divided into these main activities:

We discuss these in turn in the following paragraphs.

5.2.6 REMOTE SENSING SURVEYS Archival Cloud Cover and PWV Analysis of Northern Chile, Topographical Analysis

An analysis of Meteosat and GOES-8 satellite (6 and 10 micron channels) data by Dr. D.A. Erasmus was completed in April 2001, covering five years spread over 1993-2000 and between latitudes 20.5 to 30.5 S in Chile. This includes all major Chilean observatories. This project has been facilitated by a data sharing collaboration with ESO, and a data and results sharing collaboration with U. Tokyo. The analysis provides an eight-year baseline, which is long enough to cover the most recent El Niño and La Niña phenomena, as well as several normal years. Within the large area covered by the study, specific sites were identified via a topographical study that evaluated mountains using as criteria: (1) distance from city lights and mining operations, (2) altitude, (3) relief, (4) isolation from the surrounding landscape, and (5) summit area size (see Figure 3a). The best sites, based on lack of cloud and on low precipitable water vapor (PWV), were identified and compared to the existing sites.7 A depiction of data that are incorporated into computing site statistics is shown in Figure 3b.

Figure 3a (left) Map showing sites identified as being worthy of further study, from the

 survey. With the exception of Cerros Tololo and Pachón, existing observatory sites

 not included.  Figure 3b (right) Contours of percentage clear weather as a function of latitude

 longitude for Northern Chile. Several such maps per day can be prepared and combined in various
ways. Archival Cloud Cover and PWV Analysis of Southwestern USA and Mexico

A second study has been initiated, funded by CELT and New Initiatives Office (NIO), that covers the whole of the Southwestern USA and Mexico, extending as far north as Pike's Peak in Colorado and south to the LMT site at Cerro La Negra. This study uses the same data set as the Chilean analysis, and extends an analysis carried out for the Rocky Mountains Observatory Consortium,8 which used a subset of the data set to compare sites in Colorado to Mount Graham and Mount Hopkins in Arizona. The results of the new analysis are expected in April 2002. Archival Cloud Cover and PWV Comparison Between Mauna Kea and Chile

NIO is financing a third study, which will compare cloud cover and PWV statistics for Mauna Kea and Chilean sites. This involves purchase of a new data set, because Hawaii is not included in the field of the images purchased for the western hemisphere analyses. The data set, which has also been purchased by NIO, will run from 1997-2002, and the analysis is expected to be complete by the end of 2003. Analysis of Cirrus Cloud Transparency at Candidate Telescope Sites

Many of the scientific programs for GSMT do not require completely clear conditions, and laser AO systems may well tolerate 10-20% absorption by cirrus cloud.9 A more refined calibration of the cloudy/transitional/clear boundaries determined by the satellite studies could be used in conjunction with the studies above to better quantify, in particular, the fraction of time that laser AO systems could be used. A ground-space comparison by Erasmus & Sarazin is due for completion in 2002, and will lead to an objectively defined and well-calibrated satellite-based indicator of atmospheric transparency at optical wavelengths. For sites measured in the above studies, this calibration could be retroactively applied to the data. This study is being considered for funding in 2003.


Portable differential image motion monitors (DIMMs) have been constructed, following the initial concept by Sarazin and Roddier10 and based on the hardware and software implementation by the Department of Astronomy at the University of Washington (Chris Stubbs, Eric Deutsch, and Armin Rest) at the Astronomy Research Consortium (ARC) telescope. They consist of a Meade 10-inch LX-200 telescope, SBIG CCD camera, custom wedge lens, and mount. One DIMM has been installed at Cerro Tololo, two are being used on Mauna Kea, and another is a fully robotic version for remote use and which initially will be used to characterize the remaining large telescope site on Cerro Pachón. Present status (February 2002) is that a software re-write is almost finished, and the robotic version is ready for deployment.

Several portable, solar-powered weather stations have been purchased. These contain sensors for wind speed and direction, humidity, temperature, and solar radiation. Data are stored in a data logger, which can be downloaded at intervals of a few months.

A turbulence profilometer (multi-aperture scintillation sensor (MASS)) that follows the design of Tokovinin and Kornikov11 has been delivered recently by the Sternberg Institute, Moscow (see Figures 4 and 5). From scintillation measurements of a single star, the characteristic strengths and altitudes of the dominant turbulence layers can be derived. Calibration and testing, together with a duplicate instrument being built for ESO, has taken place at Cerro Tololo and La Silla, and the instrument has been in regular use at Cerro Tololo for several months of late 2002.

Figure 4 Simplified optical diagram of the MASS detector box. Only two of the four detector

 are shown. The pupil image formed at P is reflected in different directions by annular

 and re-imaged onto four photomultiplier tubes. Figure 5 The four concentric apertures are depicted (left), together with associated weighting

 for normal (lower left) and differential (lower right) scintillation weighting functions.

Miscellaneous additional equipment includes an array of micro-thermal sensors and a particle counter.

As a follow-on from the topographical analysis and the Chile cloud-PWV survey discussed previously, we have performed the following activities:

Figures 7a (left) and 7b (right)   Histogram of the wind speed measured by the Cerro Honar

 station during the period July 12, 2001-November 12, 2001. Day 1 corresponds to 06-12

 LT, Day 2 corresponds to 12-18 hours, Night 1 corresponds to 18-24 hours, and Night 2

 to 24-06 hours. The information for Figure 7b is the same as that for Figure 7a

 that air temperature is plotted.



Turbulence of the atmosphere occurs on a wide variety of spatial and temporal scales. It can be induced by local topography in the telescope vicinity (ground layer); at the interface between topographically induced local wind flows and the trade winds (typically 1-3 km above the site); and at the boundaries of the jet stream, of which the lower boundary (~ 10 km) is the most significant. In order to characterize the turbulence, we need to measure wind profiles and the refractive index structure constant CN2 at each site. At minimum, the "classical" measurement of seeing FWHM can be made with a DIMM; however, good seeing is a necessary but not sufficient condition for a GSMT site. A variety of instruments has been used to characterize aspects of atmospheric turbulence, and all have advantages and disadvantages.

A complementary approach is to evaluate sites by numerical modeling. Micro-scale (~ 1-m resolution) modeling of the ground layer has been demonstrated for the Gemini telescope sites by De Young & Charles,12 while lower-resolution 100-1000-m meso-scale modeling (e.g., Masciadri et al.)13 provides turbulence profiles throughout the atmosphere. Erasmus has proposed using the National Center for Environmental Protection (NCEP) re-analysis data to model CN2 profiles for candidate sites.14

Figure 8 Digital elevation model (20-m resolution) for the Chajnantor science preserve. The
original radar images had substantial shadowing, and filling these in has been a major
undertaking. The volcano at the top of the picture is Licancabur (5916 m, The ALMA site is the flat area in the center. Just below this is
the Honar ridge with Cerro Honor towards the northeast end.

We have proceeded as follows:







  1. Sarazin, M. VLT Site Selection Working Group Final report, VLT Report No. 62 (1990)

  2. Husar, R.B.; Stowe, L.L. Tropospheric Aerosols Over the Oceans. From (1994)

  3. Ehgamberdiev, S., et al. "The astroclimate of Maidanak Observatory in Uzbekistan". A&AS 145, 293 (2000)

  4. Sarazin, M. Site-Testing for ELTs: Program and Methods. From (2001)

  5. Giovanelli, R., et al. "The Optical/Infrared Astronomical Quality of High Atacama Sites. I. Preliminary Results of Optical Seeing". PASP 113, 789 (2001)

  6. Giovanelli, R., et al. "The Optical/Infrared Astronomical Quality of High Atacama Sites. II. Infrared Characteristics". PASP 113, 803 (2001)

  7. Erasmus, D.A., & van Staden, C.A. "A Satellite Survey of Cloud Cover and Water Vapor in Northern Chile", a study for CTIO and U. Tokyo (2001)

  8. Erasmus, D.A. Pike's Peak Observatory. See (2000)

  9. Nelson, J. Site-Testing for ELTs: Program and Methods. From (2001)

  10. Sarazin, M., & Roddier, F. "The ESO differential image motion monitor". A&A 227, 294 (1990)

  11. Tokovinin, A., & Kornikov, V. "Measuring turbulence profile from scintillations of single stars". In Astronomical Site Evaluation in the Visible and Radio Range, Eds, Benkhaldoun Z., Munoz-Tunon C., Vernin, J., ASP Conference Ser. 266, 2001 (Marrakesh, November 13-17, 2000); also see

  12. De Young, D., & Charles, R. "Numerical Simulation of Airflow Over Potential Telescope Sites". AJ 110, 3107 (1995)

  13. Masciadri, E., Vernin, J., Bouseault, P. "3D mapping of optical turbulence using an atmospheric numerical model. I. A useful tool for the ground-based astronomy". A&AS 137, 185 (1999)

  14. Erasmus, D.A. NCEP re-analysis proposal to CTIO and CELT (2001)

March 2002