Site Characterization of the Northern Site of the Cherenkov Telescope Array

We present the current knowledge and activities to assess the climatology of the “Observatorio del Roque de los Muchachos” (ORM), selected to host the Northern observatory of the CTA, with particular emphasis on molecular density profiles and aerosol extinction.


Introduction
Recently, the site selection process for the Cherenkov Telescope Array (CTA) [1] has been resolved.For its Southern part (CTA-S), a plateau of the Cerro Armazones in Northern Chile, close to Paranal was chosen 1 , while the "Observatorio del Roque de los Muchachos" (ORM) on La Palma, Canary Islands, has been selected for its Northern observatory (CTA-N) 2 .
The ORM looks also back on a history of more than 20 years of gamma-ray astronomy through Imaging Atmospheric Cherenkov Telescopes (IACTs), featuring first the HEGRA 3 and then the MAGIC Telescopes 4 .The latter dedicated considerable efforts to characterize the atmosphere above with respect to molecular and aerosol profiles, and to clouds [2][3][4].Moreover, existing optical telescopes are active in this task [5,6], and the very same host organization of the ORM, the "Instituto de Astrofísica de Canarias" (IAC), through their "Sky Quality Group" (SQG) 5  [7][8][9].In addition, also the "European Extremely Large Telescope" (E-ELT) have carried out site testing and characterization campaigns on La Palma [10].Moreover, the only about 150 km apart "Observatorio del Teide" (OT) 6 , and particularly the "Izaña Atmospheric Research Center" (IARC) 7 , or simply Izaña Observatory (IZO), both located on the neighboring island of Tenerife at about 2400 m a.s.l., have monitored atmospheric conditions over large time series.Atmospheric conditions between Tenerife and La Palma, at these altitudes, are highly correlated [11], and data from the IARC can hence be used, at least to better understand the aerosol characteristics, at the ORM.
The atmosphere in the sub-tropical region of the Canary Islands is characterized by great stability throughout all the year.This is due to the combination of two processes of large-scale atmospheric circulation [12]: the descending branch of the Hadley cell around 30 • N and the "Trade" or "Alisios" winds, coming from the Azores high area, that blow in the low troposphere, above an ocean which is relatively cold.As a result, a temperature inversion layer (TIL) appears above 1300 m a.s.l. in Winter (∼1000 m a.s.l. in Summer) 8 [13], which can be usually well identified by the "sea of stratocumulus" at its base.This layer separates two well-defined regimes: below it, the moist marine boundary layer (MBL) and above it, the dry free troposphere (FT).The TIL is present in > 84% of the time in the Western Canary Islands (> 90% observed at Guimar, Tenerife; the closest sounding station to La Palma), and shows moderate seasonal variation, with more frequent break-downs observed in February and November, probably modulated by the subsidence of the Hadley cell [13].
The ORM is located at an altitude between 2000 m and 2400 m a.s.l., hence well above the thermal inversion layer.This ensures that the installations are in the FT.At the ORM, clean air and clear sky conditions are hence prevailing all around the year.The reason is, by the way, twofold: apart from being located above the stable inversion layer, it is situated on an island far away from any significant industrial activity.Now that site selection is accomplished and expected average observational duty cycle and performance established [14], the CTA-N is primarily interested in assessing molecular density profiles up to about 25 km a.s.l. and their variations, in understanding aerosol profiles and their optical characteristics, in order to trim its atmospheric calibration strategy [15].

Molecular density profiles
We studied 3 years (2013-2015) of now-cast predictions of the "Global Data Assimilation System" (GDAS) 9 , available in time steps of 3 hours UTC for La Palma, more specifically for the grid point 29 • N, 18 • W, about 28 km northwards the ORM, above the Atlantic, smoothly joint with the NRLMSISE-00 data10 , for lat.28.77 • , lon.342.11 • above 20 km a.s.l.A spatially better resolved WRF model [16], as employed by J. Marín in these proceedings [17], has not yet been tried, however the European ECMWF 11 is currently being tested, yielding similar or even better results, while the IG2 model [19] does not agree well with instantaneous MAGIC data at these altitude ranges, at least what concerns the temperature profiles.Similarly, the NRLMSISE-00 model [20], although very accurate above 20 km a.s.l., is not able to resolve all tropospheric temporal variations and can show deviations from the GDAS model of up to 10%.
The GDAS data set is available in 24 pressure levels 12 , each associated with an individual geopotential height, a temperature and other data, such as wind components and humidity.Geopotential heights have been converted to local altitudes a.s.l.using the WGS-84 ellipsoid for the evaluation of the combined effects of gravitational and centrifugal forces on the local gravity at a given altitude and latitude, following the prescription of Mahoney [21].Values between these levels were interpolated.One has to keep in mind however, that fine-structures possibly present in the lower troposphere, and particularly in the tropopause (see e.g.Fig. 1 of [13] or Fig. 1 of [22]), cannot be reproduced by such an interpolated coarse grid.
For the purpose of ground validation, temperature and pressure values measured by the MAGIC weather station were compared with the predicted ones at the altitude of the weather station, which was assessed by a specialized company using a GPS survey in 2015.Fig. 1 shows all entries obtained for night-times.For the pressure comparison a bias of (−0.25 ± 0.01) mbar was obtained, compatible with no bias, once the accuracy of the MAGIC pressure sensor (of ±0.8 mbar) is taken into account.The temperature comparison yields a significant bias of (2.08±0.02)• C, which can however be understood by ground cooling effects, that are not present at the chosen GDAS grid point (lying above the Sea).Using only day-time data, an opposite effect is observed.Typically such ground effects are limited to low altitudes.Nevertheless, even in the exaggerated assumption of being constant throughout the troposphere, this temperature bias would alter the density profile by only 0.8%.Another direct validation was performed with the MAGIC LIDAR data, which shows pure molecular back-scattering and extinction signatures throughout most of the probed altitude ranges, and where contemporaneous GDAS profiles yield accurate χ 2 -fit distributions, in contrast to average molecular profiles [23].The inlets display the distribution of the respective differences.
Fig. 2 shows averaged density profiles, grouped according to the findings of [22]: an "average winter" (DEC,JAN,FEB,MAR,APR), characterized by multiple thermal tropopauses, which are several kilometers thick and can reach extremely high altitudes of up to 20 km a.s.l..This typical mid-latitude tropopause is influenced by the sub-tropical jet stream (STJ) centered at lower latitudes.The "average summer" (JUL,AUG,SEP) model is characterized by only one thermal tropopause, typical for a tropical upper troposphere, with the STJ having moved northwards of the Canary Islands.The intermediate months (MAY,JUN,OCT,NOV) show transition behavior between both scenarios.Tropical tropopauses occur nevertheless also occasionally during the rest of the year.The lower end of the multi-layered, thick mid-latitude tropopauses move slightly upwards during spring and de-scend again in autumn, while the single-layered tropical tropopauses show much less seasonality [22].From these average scenarios, and the extremes also seen in Fig. 2, the input models for detailed simulations have been created [24].
The most significant differences between both averages are found below 5 km a.s.l. and in the tropopause.The latter has also been observed in the South [17].At about 15 km, the strongest relative differences between Summer and Winter is observed, with an average seasonal variation amplitude of about 7% (see Fig. 2 bottom).Diurnal variations are more than an order of magnitude smaller at all altitudes above CTA-N.The dark blue, red and green points display an "average Winter", "average Spring/Autumn" and "average Summer" scenario, respectively.Error bars represent the standard deviation, while the light blue lines show the peak-to-peak differences for the entire data set.Bottom: seasonal variation of density evaluated at 15 km a.s.l.
Ozone intrusions into the upper troposphere, the "stratosphere-troposphere exchange" (STE) processes, have not been found for the Western Canary Islands, as assessed climatologies of the "Potential Vorticity" (PV) parameter [18].Nevertheless, a 20 yr ozone-sonde campaign revealed that the extra-tropical tropopause above Tenerife shows increased ozone concentration during late Winter and Spring in the range from 12 to 15 km a.s.l.[22], well below the stratospheric ozone layer above ∼18 km.

Aerosol Profiles
Aerosols have been characterized at the ORM with the help of elastic LIDARs [10,23], dust counters [6] and extinction measurements by optical telescopes [8].Moreover, several types of particulate analyzers and nephelometers are employed at the Izaña station on Tenerife [25][26][27] and on airplanes [28].Sun photometer inversion products are available for Izaña through the AERONET network 13 .Also satellite information has been investigated [11] and back-trajectories simulated, using the HYSPLIT code [10,29].
Ground-layer aerosol at the Western Canary Islands, of marine/oceanic origin with small admixture of anthropogenic aerosol [10] typically shows a bi-model size distribution during clear nights, with a size distribution peaking at 50 nm during night (see Fig. 7 of [25]), an exponential decay of density, with scale heights around 500 m [23] and aerosol optical thickness (AOT) of ∼0.02 at 532 nm [10,23].In this situation, only Aitken and Accumulation modes of aerosol are found while the Coarse mode is practically absent.This leads to typical Ångström exponents of 1.5-2.0[26,30] and a single scattering albedo of 0.91 [27] and LIDAR ratios around 30 sr [31].These findings are compatible with the median value for the V-band extinction of 0.13 mag airmass −1 found by [8], translating into an AOT of about 0.02.
The situation changes drastically during dust intrusion events from the Western Sahara (the so-called "calima"), when mean particle sizes reach 1 μm or greater [6].Such events occur during about 30% of the Summer nights [8], and rarely during the rest of the year.Typical LIDAR ratios are then ∼60 sr [31], and AOTs can reach values in excess of 0.5 [6].The calima-related dust can reach altitudes of and higher than 7 km a.s.l.[10], although lower altitudes are more frequent [23].
Fig. 3 shows the probability to observe a given value of AOT from different altitudes above the CTA-N, obtained with a 2-years statistics of the MAGIC LIDAR.

Conclusions
The atmosphere above the ORM has been characterized by a large variety of instruments in the past.For molecular and aerosol profiles, detailed climatological studies exist providing a deep understanding of their origins and characteristics.The CTA-N profits from these to design its atmospheric calibration strategy and reduce systematic uncertainties to hitherto unprecedented levels.) and the red line (emission height 6 km a.g.l.) are mainly due to calima, whereas clouds are frequently found between 4.5 km and 9 km a.g.l.

Figure 2 .
Figure2.Top: density profiles of 3-years GDAS nowcasts at 21h, 0h, 3h and 6h UTC, scaled to the U.S. Standard Atmosphere density n s = 2.55 • 10 19 cm −3 and multiplied with the exponential of an average scale height (H s = 9.5 km) for better visibility.The dark blue, red and green points display an "average Winter", "average Spring/Autumn" and "average Summer" scenario, respectively.Error bars represent the standard deviation, while the light blue lines show the peak-to-peak differences for the entire data set.Bottom: seasonal variation of density evaluated at 15 km a.s.l.

Figure 3 .
Figure 3. Aerosol Transmission probability for different emission altitude above the MAGIC LIDAR.Differences between the yellow line (emission height 3 km a.g.l.) and the red line (emission height 6 km a.g.l.) are mainly due to calima, whereas clouds are frequently found between 4.5 km and 9 km a.g.l.