Choosing a site for a giant telescope is like picking a plot for a house: not only quiet nights matter, but also dry air, because water vapor absorbs infrared radiation and steals entire chunks of the spectrum from astronomers. Alongside this work, the archive holds studies of ice detectors listening to the radio voices of cosmic particles, and IceCube's neutrino confession of the Earth: all of them depend, in one way or another, on atmospheric transparency. For the future European Extremely Large Telescope on Cerro Armazones and the planned Cherenkov Telescope Array (CTA), it is crucial to know how much water vapor hangs above the site and how it is distributed across the sky. Earlier site-selection campaigns relied on ground-based weather stations and radiometers, but these provide only point or short-term data. Here, however, a six-year archive of spectra, accumulated by telescopes for entirely different scientific purposes, was available.
The archival spectra were obtained with the BESO instrument on two telescopes at Cerro Armazones Observatory: first on the 1.5-meter Hexapod, then on the 0.83-meter IRIS. Over six years, 996 nights of observations accumulated, and after quality filtering, 5879 spectra from 918 nights remained. For each spectrum, the molecfit program fitted a model of telluric absorption lines, built on the LBLRTM radiative transfer code and the HITRAN database, to two narrow windows in the near-infrared: blue (816–821 nm) and red (824–831 nm). The fitting was done independently in the two windows, allowing an estimate of the measurement error: the root-mean-square discrepancy was 7.95 percent. Three quality-control criteria were applied: a chi-square statistical test, the Kolmogorov–Smirnov test, and a threshold on the PWV difference between the two windows of no more than 30 percent. The derived PWV values were converted to zenith and compared with simultaneous data from X-shooter on the VLT, the LHATPRO microwave radiometer on Cerro Paranal, and the global GDAS model.
The median PWV above Cerro Murphy (the BESO site) was 2.43 mm, versus 2.96 mm at Cerro Paranal for the same time instants. The 22 percent difference is notably larger than the expected 8 percent from the 182-meter altitude difference, given a typical water vapor scale height of 1800 meters. This means the Armazones site is systematically drier than would be predicted from the altitude profile. The linear correlation between BESO and X-shooter gave a slope of 0.851 with a coefficient of determination R² = 0.89, and between BESO and the LHATPRO radiometer a slope of 0.838 with R² = 0.93, confirming the method's reliability. In contrast, the GDAS model systematically overestimated PWV by 52 percent, with a slope of only 0.561 and R² = 0.78. The sky distribution of vapor was homogeneous: within zenith distances up to 60 degrees, the mean values across all azimuthal sectors lie in the range 3–4 mm, with no noticeable asymmetries, including east-west. No influence of the 2009–2010 El Niño on humidity was detected, consistent with data on unusually frequent dry events at Paranal during those same years.
This work stands somewhat apart from its neighbors: instead of direct measurements with radiometers or weather balloons, it uses a byproduct of astronomical observations accumulated over six years. This yields not a point measurement but a spatially resolved picture of humidity above the site, something neither radiometers nor models provide. Unlike studies of ice detectors that listen to radio signals from particles, or IceCube's neutrino observations, here the atmosphere is studied through the absorption of starlight, meaning the method is passive and requires no additional equipment. The main takeaway for neighboring work is that global meteorological models in this coastal zone of the Atacama are unsuitable for precise predictions, so either local refinements or direct spectroscopic measurements are needed. For future CTA installations, this implies that ozone and water vapor monitoring can be carried out with a small telescope equipped with a fiber-fed spectrograph.
The method can be extended to measure ozone, particularly in the ultraviolet Huggins bands below 400 nm, which are not covered by the current BESO spectra. The proposed setup for CTA should record spectra down to 320 nm to capture these more distinctive ozone lines. Comparison with locally refined meteorological models, such as those by Lasko and Marin, would be a natural next step to test their accuracy in this zone. A transition to real-time monitoring is possible, as proposed for TMT on Mauna Kea, allowing observations to be adjusted promptly.
The results will directly affect observation planning for the European Extremely Large Telescope in the infrared and the calibration of Cherenkov Telescope Array data, where flux accuracy is specified at the 5–7 percent level.
The next step will be to compare the derived atmospheric profiles with results from high-resolution local meteorological models for Cerro Armazones. It is also necessary to perform a similar analysis for ozone, extending the spectral range into the ultraviolet.
For atmospheric physics, the key unresolved problem remains the vertical distribution of water vapor in the lower troposphere over mountain plateaus, where the classical exponential scale height breaks down. This work shows that even over a distance of 22 km between sites, the humidity difference does not reduce to a simple altitude scaling, indicating complex local circulation. For gamma-ray astronomy, the accuracy of accounting for stratospheric ozone absorption is critical, as it varies from night to night and is still poorly predicted by global models.
Over six years, the telescopes accumulated spectra from 996 nights, but only 63 percent of them passed strict quality filtering, yet the night coverage remained at 92 percent. This means that even rejecting nearly half of the spectra did not distort the picture of seasonal humidity variations.
In Kim Stanley Robinson's novel 'Aurora', colonists arriving at another star face the problem of atmospheric transparency, on which the operation of their telescopes depends. Here on Earth, the same problem is being solved for a site where sky transparency will be decisive for the giant eye of the E-ELT.