The search for astronomical observatory sites in the tropics of northern South America has long been hampered by a lack of meteorological data: the mountainous terrain of the Andes creates microclimates, and the network of ground stations is sparse. Satellite-based cloud assessment methods were already in use nearby, for example, Cavazzani and Zitelli used monthly brightness thresholds for five observatories, while Hidayat and colleagues analyzed 15 years of GOES imagery over Indonesia. This work follows the same line but shifts the focus to the tropical Andes of Colombia, Venezuela, and Ecuador, where no systematic clear-sky map had existed before.
The authors took nighttime images from GOES-12 and GOES-13 in the 6.7 μm (upper troposphere, 8–10 km) and 10.7 μm (mid-troposphere, about 4 km) channels over five years. For each pixel, they calculated the brightness temperature and then converted it to actual temperature using calibration coefficients. Threshold temperatures for mid-level clouds were derived from long-term records of 63 IDEAM weather stations: they took the monthly mean temperature minus three standard deviations and applied a correction for nighttime observations. They then built a linear relationship between the threshold and terrain elevation for each month. Elevations in areas without stations were reconstructed using kriging, a geostatistical interpolation that accounts for the similarity of nearby points. For the upper troposphere, thresholds were computed from radiosonde data from Bogotá and the Caribbean coast: temperatures at 8, 9, and 10 km were measured twice daily, and the monthly mean minus three sigma was used. A night was considered spectroscopic if the sky was clear for more than one-third of the time, and photometric if clear for more than two-thirds.
Over five years, twelve sites stood out where the fraction of clear spectroscopic nights exceeds 30% during the dry season. The absolute record holder is a site in the Venezuelan Andes (70°28′48″ W, 9°5′60″ N) at an altitude of 3480 meters: 67.0 ± 10.7% clear nights in the dry season and 53.5 ± 12.2% in the wet season, yielding 220 ± 42 spectroscopic nights per year. For comparison, the operational Llano del Hato Observatory shows only 132 ± 25 such nights. In Colombia, the best regions are Sierra Nevada de Santa Marta (126 ± 34 nights) and Serranía del Perijá (111 ± 27 nights). In the Eastern Cordillera, sites with 107 ± 23 and 94 ± 13 nights were found. At the equator, in Nudo de los Pastos, two sites give 105 ± 36 and 101 ± 42 nights, with minimal seasonal differences of just 1–4%. Year-by-year analysis revealed a dip in clear nights during the wet seasons of 2010 and 2011, coinciding with positive values of the Southern Oscillation Index, that is, with La Niña.
This work differs from neighboring methods in that cloud thresholds are tied not to atmospheric models but to direct multi-year measurements of temperature near the surface and in the upper troposphere. This provides a more accurate picture for mountainous terrain, where standard models often miss the mark. For the first time, a continuous clear-sky map with a resolution of 28 × 18 km has been produced for the northern Andes, allowing comparison of remote peaks without costly expeditions. The method is easily transferable to other tropical regions with similar climate and topography.
The selected sites require ground-based verification: the astroclimate must be measured, including turbulence, humidity, wind speed, and sky brightness. The authors plan an observational campaign to characterize these sites. Further refinement is possible through longer satellite data records and the inclusion of imagery from other geostationary satellites, such as GOES-16 with its 1.6 μm channel for detecting thin cirrus clouds.
The results will directly influence the selection of sites for future optical and infrared observatories in Colombia, Venezuela, and Ecuador, where universities plan to build telescopes. The method is also useful for assessing the suitability of sites for solar power plants and for climate studies of cloudiness in the Andes.
Conduct year-long ground-based measurements of seeing (image jitter), humidity, and wind at three or four of the best sites. Compare the satellite-derived maps with data from automatic weather stations installed directly on the peaks.
An open question for astronomy remains: how to predict the fraction of clear sky decades ahead under a changing climate and El Niño events. This work shows that even a five-year record reveals a connection with the Southern Oscillation, but longer observations and accounting for local terrain effects are needed for reliable forecasting.
The threshold temperature for clouds at an altitude of 4000 meters in the Andes can drop to 0–2 °C, while in the Magdalena River valley it exceeds 20 °C. This means that the same infrared spot on a satellite image indicates a cloud over a peak but clear sky over a lowland.