Matteo Marchetti, Davide Arenare, Filippo Concaro, Marco Pasian
IMAGE LICENSED BY INGRAM PUBLISHING
Ground stations working at high frequencies, such as the K-band, installed at polar latitudes and protected by radomes, are increasingly used to support modern satellites. The effect that snow accumulation may have on the radome is consequently important, as at these frequencies it may jeopardize the satellite link. This article analyses an operative case, referred to as SNOWBEAR (Svalbard grouND StatiOn for Wide Band Earth observation dAta Reception), where a 6.4-m antenna has been installed at Svalbard, Norway, to track an Earth observation (EO) satellite, NOAA-20, for a period of two years. We demonstrate, using experimental data and numerical models, that a chief effect can be described in terms of de-pointing of the main beam, with a particular focus on the difference of de-pointing between the sum and delta modes, and that a real-time, effective, solution is not yet available.
A general trend in the field of satellite communications embraces the possibility of providing downlink channels with new standards in terms of bit rate. This aspect is driving a frequency upscaling to bands higher than usual. For EO satellites, these are traditionally the S- and X-bands, with the K-band intended for future upgrades. Among the others, two projects exploiting the K-band are the Joint Polar Satellite System, developed by the National Aeronautics and Space Administration for the National Oceanic and Atmospheric Administration (NOAA), initially named JPSS-1 and now identified as NOAA-20, and the constellation MetOp-SG, developed by the European Space Agency (ESA) for the European Organisation for the Exploitation of Meteorological Satellites [1], [2].
These missions require ground stations located at polar latitudes to guarantee full coverage of all satellite passes. However, at these sites ground stations are often protected by radomes [3]. In general, studies on the effect of environmental conditions on the performance of reflector antennas, and more specifically on radomes, have been carried out over the years, with recent works addressing new and different calculation methods and perspectives. As an example, these include finite element method-based analyses to characterize thermal–structural interactions, interval analyses aimed at characterizing the effect of fabrication tolerances [4], [5], [6], [7], and analyses on the effect of rain precipitation, covering a broad spectrum of frequencies [8], [9], [10].
However, at such high latitudes, concern for precipitation events, such as rain, is relatively rare. Instead, snow accumulating on the radome surface can be an issue. In particular, the effects of snow accumulating on the radome surface (or directly on the reflector surface, for those cases where the radome is missing) have been investigated in the past [11], [12], [13], [14], with some preliminary works at frequencies slightly above the Ku-band [15], [16]. While at lower frequencies these effects are not as severe as they can be in the K-band, the migration toward higher frequencies has triggered further investigations [17], [18].
While some effects have been discussed (e.g., attenuation, de-pointing), a comprehensive analysis comprising numerical and experimental results, as well as extensive campaigns on a real ground station tracking an operative satellite according to different tracking systems (either program track or autotrack [19]), has been missing.
For these reasons, the purpose of the SNOWBEAR project was to de-risk the introduction of the K-band within ground stations at polar latitudes [20], [21]. In particular, SNOWBEAR consisted of two main phases: the deployment of a prototype ground station at Svalbard (completed in autumn 2018, shown in Figure 1) and a two-year operational trial using the satellite NOAA-20 (completed in autumn 2020) [22], [23], [24]. NOAA-20 presents an orbital cycle composed of 227 passes (i.e., NOAA-20 repeats the same orbit trajectory every 227 passes, roughly 16 days). The SNOWBEAR consortium comprises industries and research centers from all over Europe, mainly Belgium, France, Germany, Italy, and Norway.
Figure 1. The SNOWBEAR ground station.
In this article, the impact of snow accumulation on the radome is analyzed. In more detail, in the “Antenna, Radome, and Tracking Modes Description” section. a description of the ground station is provided, focusing on the parts relevant for the discussion presented in this article (antenna, radome, tracking systems). The “Description of the De-Pointing Caused by Snow Accumulation on the Radome” section describes the different de-pointing mechanisms that may affect the ground station when snow accumulation is present on the radome. A particular focus is on the de-pointing effect caused by dry snow, which is generally different for the sum and delta modes, thus jeopardizing the proper tracking operation of the ground station. Finally, the “Numerical Models” section provides a numerical validation of different cases, considering both a simplified setup and a more complex approach. The “Operational Results” section summarizes the experimental data collected during the two-year operation phase of SNOWBEAR.
The SNOWBEAR antenna, shown in Figure 2, is a ring-focus dual-reflector Cassegrain antenna composed of a 0.8-m hyperbolic subreflector, shaped to reduce the coupling effects, and a 6.4-m parabolic main reflector. The antenna is design to work in the S-band (2.2–2.3 GHz) and K-band (25.5–27 GHz) to support low Earth orbit satellites. To track these satellites, the antenna system is equipped with a three-axis control movement: the azimuth axis to rotate the antenna structure along a vertical axis, the elevation axis to rotate the reflector along a horizontal axis, and the cross-elevation axis to tilt the reflector perpendicularly to the elevation direction. Concerning the feed, this is positioned in the focus of the subreflector and it is characterized by two coaxial circular corrugated horns: the K-band feed inside the S-band one.
Figure 2. (a) SNOWBEAR antenna and radome partially covered by snow. (b) Enlarged view.
To protect the antenna from the harsh environment, it was covered by self-supporting multilayer radome, shown in Figures 1 and 2, characterized by a spherical truncated structure, with a diameter equal to 11.8 m and truncation of 87%, whose wall is composed by a random configuration of pentagonal and hexagonal shape multilayer panels jointed together. The panels are realized by alternately stacking several layers of high and low dielectric constant materials. Foam is used for the realization of the low dielectric constant layers, whereas fiberglass, due to its good compromise between mechanical and electromagnetic properties in addition to its reasonable costs, is adopted to realize the high dielectric constant layer. The outer layer is gelcoat, and it provides a contact angle for water drops in the order of 90°. The interconnections among panels are characterized by dielectric struts, which joint together the adjacent panels thanks to metallic bolts. Panel structure was designed to provide a compromise between mechanical stiffness and electromagnetic transparency for the frequencies of interest. In dry conditions, the radome provides a transmission loss lower than 1 dB in the K-band.
The SNOWBEAR system is provided with an open-loop pointing functionality, commonly called program track, which steers the antenna based on the predicted satellite orbit. This orbital prediction is provided in a format called two-line elements (TLE) that is a data format encoding a series of orbital elements of an object orbiting around Earth for a given time instant, called an epoch. However, with TLE information it is possible to know the position and velocity of that object accurately only for a limited amount of time. Typically, predictions are considered valid for a maximum period of three or four days. In the case of NOAA-20, TLE are provided to Kongsberg Satellite Services (KSAT), in charge of operating the ground station, from the NOAA flight dynamics team on a daily basis.
Besides the open-loop tracking mode, the SNOWBEAR system is also provided with a closed-loop mode, called autotrack. In more detail, this implements an autotrack mode called monopulse. This is a very well-known method that exploits the properties of the higher propagation modes (in this case TM01, called the “delta” signal), which presents a null in its central position and a specific phase distribution. Combining the delta signal with the fundamental mode (in this case TE11 mode, called the “sum” signal, used to normalize the delta signal in terms of magnitude and phase), it is possible to calculate the angular mispointing from the radio-frequency source direction, both in magnitude and direction. This is performed by the so-called tracking receiver, which generates the tracking errors, which are sent to the antenna control unit to correct in real time the antenna pointing. The tracking errors are thus a measure of the angular displacement between the spacecraft and the position of the of the antenna delta pattern null. In ideal conditions, the sum pattern peak and the delta null are supposed to be perfectly aligned, but this cannot be the case for all operating conditions, as will be described later. The autotrack mode is generally preferred to the program track, especially when the antenna beam is very narrow, as is the case for SNOWBEAR in the K-band, with a half-power beamwidth (HPBW) of approximately 100 mdeg, as shown in Figure 3.
Figure 3. Measured normalized radiation pattern for the copolar sum (black line) and delta (gray line) mode for the SNOWBEAR ground station without snow, for a generic exemplifying cut.
At the frequencies of interest, snow accumulation on a radome can provide different effects on the link, the most important being reflection, absorption, scattering, and de-pointing of the main beam.
Reflection occurs because of the impedance mismatch among the air, the snow layer on top of the radome, and the radome itself.
Even at those frequencies where, ideally, the radome is perfectly transparent when dry, snow introduces an extra condition not accounted for during the design phase of the radome itself, causing mismatch. Absorption is due to the fact that the signal travels through a lossy dielectric medium. For snow, this normally happens when the liquid water content (LWC) of the snow itself is larger than zero, i.e., when snow can be defined as wet snow. This can be the case in certain periods of the year because of increased temperatures and/or direct sunlight incidence. Scattering occurs when the dimension of the wavelength is comparable with the dimension of the snow crystals, and the incoming electromagnetic wave is scattered along multiple directions. Finally, de-pointing is induced when the wavefront is unevenly distorted because of the snow on top of the radome, for example because the snow accumulation shape and thickness vary along the radome.
All effects are more significant when the antenna is moving at elevations higher than around 30°, where the presence of snow is more probable because the radome above is less steep, and snow can accumulate more easily. While the final outcome of all of these effects is a reduction of the magnitude of the received signal, de-pointing is particularly critical because it is responsible for the generation of a misalignment between the antenna pointing and the satellite position. The modeling can be complex, as well as the possibility to correct for it in real time, as a number of parameters, for example the dielectric and geometrical properties of the snow accumulation, are difficult to predict and/or sense.
As an example, Figure 4 presents the effects caused on the antenna pointing by two snow layers, apparently similar, for an entire orbital cycle. In practice, for each of the 227 passes composing a cycle, the pointing error is graphically reported for each point of the pass itself, thus following the satellite trajectory along the sky. In more detail, Figure 4(a) and (b) show the magnitude of the pointing error, while Figure 4(c) and (d) show its direction, on the azimuthal plane. In particular, the pointing error (${\bf{p}}_{\text{ptg}}$) is a vector, calculated as the angular deviation between the antenna axes positions as recorded by the antenna encoders (${\bf{p}}_{\text{pass}}$) and as recorded during a reference case, identified considering the same pass recorded in autotrack during the summer and under ideal sky conditions (${\bf{p}}_{\text{ref}}$): \[{\bf{p}}_{\text{ptg}} = {\bf{p}}_{\text{pass}}{-}{\bf{p}}_{\text{ref}}{.} \tag{1} \]
Figure 4. Example of the pointing errors for two entire satellite cycles, with similar snow accumulation on the radome: (a) magnitude (mdeg) for the first cycle, (b) magnitude (mdeg) for the second cycle, (c) direction on azimuthal plane (°) for the first cycle, and (d) direction on azimuthal plane (°) for the second cycle. Each colored line corresponds to the trajectory of a single pass. The black silhouette along the external border represents the mask due to propagation obstacles (e.g., mountains, other radomes, and masts).
As one can see [Figure 4(a) and (b)], pointing-error distributions are similar in terms of magnitude (and it can be observed that magnitudes exceeding 100 mdeg are recorded, as large as the width of the antenna main beam, shown in Figure 3, thus potentially causing large impairments) but the direction distribution, corresponding to the de-pointing direction [Figure 4(c) and (d)], is considerably different between the two cases. Therefore, it is very difficult to predict the final pointing direction in winter conditions.
At the same time, even a real-time correction is very difficult. To better understand, Figures 5–8 show a set of key parameters recorded during the same pass, using both autotrack (Figures 5 and 6) or program track (Figures 7 and 8), in the presence of snow/ice on top of the radome. In more detail, Figures 5 and 7 show the recorded signal strength over time at the receiver. Both copolar (right-hand circular polarization, RHCP, Figures 5 and 7, blue line, ${\text{P}}_{\text{copolar}}$) and cross-polarization (left-hand circular polarization, LHCP, Figures 5 and 7, red line,${\text{P}}_{\text{crosspolar}}$) levels are shown and they are superimposed on a dashed line (Figures 5 and 7, black line, ${\text{P}}_{{\text{copolar}}{\_}{\text{th}}}$), which represents the calculated theoretical copolar signal strength in the ideal case (no snow and clear sky). Instead, Figures 6 and 8 show the so-called residual (Figures 6 and 8, black line,${\text{P}}_{\text{residue}}$), which is the ratio between the theoretical and recorded copolar signal strength (i.e., the difference, in decibels, between the black dashed line and the blue line in Figures 5 and 7): \[{\text{P}}_{\text{residue}} = {\text{P}}_{{\text{copolar}}{\_}{\text{th}}} / {\text{P}}_{\text{copolar}}{.} \tag{2} \]
Figure 5. Example of a winter pass recorded in autotrack. Measured copolar (blue) and cross-polar (red) signal strengths levels at the receiver are shown. In addition, the theoretical copolar (clear sky, no snow, dashed black line) signal strength is reported (label: Copolar th). Finally, the elevation angle (dashed gray line) of the antenna along the pass is shown.
Figure 6. Measured residual (black) and calculated pointing loss (Ptg. Loss, green) and tracking loss (Trk. Loss, orange) for the case reported in Figure 5. Again, the elevation angle (dashed gray line) of the antenna along the pass is shown.
Figure 7. Example of a winter pass recorded in program track. Measured copolar (blue) and cross-polar (red) signal strengths levels at the receiver are shown. In addition, the theoretical copolar (clear sky, no snow, dashed black line) signal strength is reported (label: Copolar th). Finally, the elevation angle (dashed gray line) of the antenna along the pass is shown.
Figure 8. Measured residual (black) and calculated pointing loss (Ptg. Loss, green) and tracking loss (Trk. Loss, orange) for the case reported in Figure 7. Again, the elevation angle (dashed gray line) of the antenna along the pass is shown.
Figures 6 and 8 also show the pointing loss (green line) and the tracking loss (orange line). These are not measured but calculated, and useful to understand the origin of the residual. In particular, pointing loss (${\text{P}}_{{\text{ptg}}{\_}{\text{loss}}}$) is simply calculated from the pointing errors (${\bf{p}}_{\text{ptg}}$), as introduced in Figure 4. That is, once the pointing error [as already stated in (1), calculated as the angular deviation between the antenna axes positions as recorded by the antenna encoders and as recorded during a reference pass in summer and under ideal sky conditions) is known, the pointing loss is calculated by mapping the pointing error on the radiation pattern of the antenna (basically using the information reported in Figure 4).
Tracking loss (${\text{P}}_{{\text{trk}}{\_}{\text{loss}}}$) is calculated from the tracking errors (${\bf{p}}_{\text{trk}}$), which in their turn are calculated as the angular deviation between the direction of the null of the delta signal (${\bf{p}}_{\text{null}}$) and the direction of the antenna axes positions as recorded by the antenna encoders (${\bf{p}}_{\text{pass}}$): \[{\bf{P}}_{\text{trk}} = {\bf{p}}_{\text{null}} - {\bf{p}}_{\text{pass}}{.} \tag{3} \]
Again, once the tracking errors are known, the tracking loss is calculated mapping the tracking error on the radiation pattern of the antenna. The instantaneous antenna elevation is also reported in all the graphs as reference (gray dashed line in Figures 5–8).
First, comparing the residuals, it is possible to see that similar losses occur for both operational procedures. This means that both program track and autotrack are not able to compensate for the de-pointing. In particular, for autotrack mode, it can be seen in Figure 7 that tracking losses are practically zero for almost the entire pass, confirming that the antenna followed the deflected delta null, as expected for the autotrack mode. Instead, pointing losses are large, indicating that the pointing of the antenna is not in line with the reference pass recorded during the summer with clear skies.
Conversely, in program track, the situation is practically the opposite, with minimal pointing losses (as expected, as in program track the predicted pointing is supposed to be in line with the reference summer pass) and large tracking losses, as the signal and delta beams are distorted by the presence of snow on top of the radome.
Overall, a key point emerged, regardless of the tracking mode, auto or program. The beam squint imposed by the snow affects differently the sum and the delta mode of the antenna. In autotrack, this can be inferred by the fact that the residual is present, despite the lack of tracking errors for the most of the pass. Indeed, if the squint of the sum mode was identical to the squint of the delta mode, the latter correctly compensated, as tracking errors are negligible, then residuals would be also negligible. In program track, an identical squint for sum and delta mode would have imposed practically identical values for pointing errors and residuals, but this is not the case. However, this assumption requires further investigation, presented in the next section.
To verify our measurements and confirm the assumptions made about the antenna de-pointing caused by the snow and, in particular what concerns the different deflections undergone by the sum and delta channels, a set of simulations were carried out with the aim of exemplifying notable cases.
In this first setup, to maintain a tolerable computational effort, while maintaining a general validity, instead of simulating the entire reflector antenna with its own radome, a cylindrical horn antenna was modeled in HFSS (a commercial full-wave 3D electromagnetic simulation software) and simulations were run at 27 GHz, considering the different scenarios represented in Figure 9. To keep the simulations as consistent as possible with the real SNOWBEAR case, the sum channel was generated exciting the two degenerate TE11 modes, 90° out of phase one to the other to obtain the circular polarization. Conversely, the delta channel was generated exciting the TM01 mode.
Figure 9. Simplified horn-based model: (a) case A, (b) case B, (c) case C, and (d) cases D and E.
First, the horn was simulated as a standalone to have a reference case to be compared against the others (case A). Computed sum and delta radiation patterns for this ideal case are shown in the UV-domain in Figures 10 and 11, respectively. All results are normalized to maximum directivity value. White and black stars in the graphs indicate the position of the maximum directivity and the position of the tracking null, respectively. Then, a layer of dry snow (density equal to 300 kg/m3, corresponding to a dielectric constant ${\varepsilon}_{r} = {1.549}$ [25]). was modeled and inserted in the horn immediately above the aperture (case B). The thickness of this layer is 1 cm and it is placed to cover half of the horn aperture. Computed sum and delta radiation patterns for this case are shown in the UV-domain in Figures 12 and 13, respectively. As a third case, a uniform layer of dry snow, modeled as a parallelepiped with a thickness of 1 cm, was placed in front of the feed at a distance of 15 cm (more than 10 free-space wavelengths) from the horn aperture (case C). The snow is displaced with respect to the antenna Z-axis so that the propagation direction is half covered by it. Computed sum and delta radiation patterns for this case are shown in the UV-domain in Figures 14 and 15, respectively. The fourth case (case D) is similar to the previous one except for the snow thickness, orientation, and shape. The thickness is 2 cm and the snow shape is not regular but designed to be conformal with a region of the radome where two or more panels are jointed together, thus resuming the profile of the radome panel junctions. This is because it was observed that sometimes the snow tends to stick on the radome following the joint profile, creating a region with an important discontinuity between the presence of snow at higher elevations and without snow at lower elevations, which may be considered a worst case. Computed sum and delta radiation patterns for this case are shown in the UV-domain in Figures 16 and 17, respectively. Finally, the setup of the last case (case E) is identical to the one of case D, but with wet snow (dry density equal to 400 kg/m3 and LWC = 3%, corresponding to ${\varepsilon}_{r} = {1.824}$, ${tg}\delta = {0.05094}$ [25]). Computed sum and delta radiation patterns for this case are shown in the UV-domain in Figures 18 and 19, respectively.
Figure 10. R-Channel, case A.
Figure 11. T-Channel, case A.
Figure 12. R-channel, case B.
Figure 13. T-Channel, case B.
Figure 14. R-Channel, case C.
Figure 15. T-Channel, case C.
Figure 16. R-Channel, case D.
Figure 17. T-Channel, case D.
Figure 18. R-Channel, case E.
Figure 19. T-Channel, case E.
For all relevant simulations (cases from B to E), the angular squint for the ∑-channel main lobe and Δ-channel null are summarized in Table 1, and some outcomes can be derived. Considering the horn half-power beamwidth, which is around 20.5°, in the presence of snow, it is possible to observe a significant misalignment between both channels and between each of them and the Z-axis direction. In particular, for cases D and E, the irregular contour of the simulated layer greatly changes their shape. The only significant change observed in the radiation patterns between case D (dry snow) and case E (wet snow) is a different position for both maximum directivity and tracking null caused by the different attenuation of that portion of the wave that travels inside the snow.
Table 1. Squint values [°] (elevation θ and azimuth φ) for cases shown in figure 10 for the sum (∑) and delta (Δ) mode.
Looking at the Δ-channel diagrams, it can also be observed that the identification of the null of the mode can be misinterpreted by the presence of local nulls, and in general more by the degradation of the pattern further worsening the performance in autotrack mode.
A second set of simulations were run, again in HFSS, to evaluate not only different plane layers of snow but also a more realistic case of snow deposition on the radome structure. This was done exploiting the shooting and bouncing rays (SBR+) tool of commercial software. Also, to maintain in this case a tolerable computational effort while maintaining a general validity, the radome dimension was kept smaller than the real one (2-m diameter with respect to 11.8 m). Following the same scheme as before, the simulations were run at 27 GHz, considering the different scenarios represented in Figure 20. Once again, to keep the simulations as consistent as possible with the real SNOWBEAR case, the sum channel was generated exciting the two degenerate TE11 modes, 90° out of phase one to the other to obtain the circular polarization, while the delta channel was generated exciting the TM01 mode.
Figure 20. Intermediate smaller-radome model: (a) case A, (b) case B, (c) case C, and (d) cases D and E
First, the horn covered by a radome of dry snow (1-cm thick) was simulated (case A). As done previously, the sum and delta radiation patterns are shown in the UV-domain in Figures 21 and 22, respectively, and all results normalized to maximum directivity value. White and black stars in the graphs indicate the position of the maximum directivity and the position of the tracking null, respectively. Then, a half radome covered by dry snow was considered (case B). Again, the thickness of this layer is 1 cm. Computed sum and delta radiation patterns for this case are shown in the UV-domain in Figures 23 and 24, respectively. As a third case, a quarter radome of 1 cm dry snow was modeled (case C). Computed sum and delta radiation patterns for this case are shown in the UV-domain in Figures 25 and 26, respectively. The fourth case (case D) present different snow thickness, orientation, and shape. The thickness is 2 cm and the snow shape is not regular. Computed sum and delta radiation patterns for this case are shown in the UV-domain in Figures 27 and 28, respectively. Finally, the setup of the last case (case E) is identical to the one of case D, but with wet snow. Computed sum and delta radiation patterns for this case are shown in the UV-domain in Figures 29 and 30, respectively. For all relevant simulations (cases from B to E), the angular squint for the ∑ -channel main lobe and Δ-channel null are summarized in Table 2.
Figure 21. R-Channel, case A.
Figure 22. T-Channel, case A.
Figure 23. R-Channel, case B.
Figure 24. T-channel, case B.
Figure 25. R-Channel, case C.
Figure 26. T-Channel, case C.
Figure 27. R-Channel, case D.
Figure 28. T-Channel, case D.
Figure 29. R-Channel, case E.
Figure 30. T-Channel, case E.
Table 2. Squint values [°] (elevation θ and azimuth φ) for cases shown in figure 20 for the sum (∑) and delta (Δ) mode.
It can be appreciated that also for this more complex model, the results are in line with those already presented for the previous numerical model.
Overall, despite Tables 1 and 2 referring to specific numerical examples, and for this reason, the actual de-pointing values are driven by the numerical parameters (e.g., snow properties and shape) and the numerical accuracy (particularly critical for large electromagnetic models); nevertheless, they clearly confirm that the de-pointing effect can be relevant, and in general not equal for the sum and delta modes. However, the particular case where the de-pointing effect is practically equal can happen, as demonstrated with the experimental passes reported in Figures 31–34, collected using autotrack. In particular, analyzing Figures 31 and 32, related to a pass recorded during the night of 9 December 2018, thus with conditions that strongly suggest the presence of dry snow (i.e., no sunlight and freezing temperatures), it is possible to see huge residuals, overlapped almost perfectly with the pointing loss estimation trend (apart from the central part when the autotrack was lost).
Figure 31. Orbit 5473, 9 December 2018, recorded in autotrack. Measured copolar (blue) and cross-polar (red) signal strengths levels at the receiver are shown. In addition, the theoretical copolar (clear sky, no snow, dashed black line) signal strength is reported (label: Copolar th). Finally, the elevation angle (dashed gray) of the antenna along the pass is shown.
Figure 32. Orbit 5473, 9 December 2018. Measured residual (black) and calculated pointing loss (Ptg. Loss, green) and tracking loss (Trk. Loss, orange) for the case reported in Figure 20. Again, the elevation angle (dashed gray line) of the antenna along the pass is shown.
Figure 33. Orbit 6968, 24 March 2019, recorded in autotrack. Measured copolar (blue) and cross-polar (red) signal strengths levels at the receiver are shown. In addition, the theoretical copolar (clear sky, no snow, dashed black line) signal strength is reported (label: Copolar th). Finally, the elevation angle (dashed gray line) of the antenna along the pass is shown.
Figure 34. Orbit 6968, 24 March 2019. Measured residual (black) and calculated pointing loss (Ptg. Loss, green) and tracking loss (Trk. Loss, orange) for the case reported in Figure 20. Again, the elevation angle (dashed gray line) of the antenna along the pass is shown.
This means that, even if the snow accumulating on the radome affected the antenna pointing, the snow distribution was such that the de-pointing effect is similar for the sum and delta mode. Another example where de-pointing is similar for the sum and delta mode is presented in Figures 33 and 34. This pass was recorded on 24 March 2019, during a clear-sky day, as shown in Figure 35 (the photo is unfortunately overexposed due to snow albedo). Thus, even if the external temperature was still below 0 °C, the snow was irradiated by direct sunlight, and this may cause an initial, mild, melting of the snow accumulated on the radome. In this case, not only the de-pointing effect is similar for the sum and the delta mode, but practically no pointing errors were recorded and consequently no loss contribution was estimated. This can be the case when the snow accumulating on the radome is distributed uniformly along the antenna axis. However, residuals are as high as 3 dB, thus suggesting that this estimated loss can reasonably be attributed to other effects than de-pointing, mainly to snow absorption.
Figure 35. Photo from the webcam taken on 24 March 2019 at 12 p.m.
The operational phase of SNOWBEAR provided the possibility to verify experimentally the aspects discussed in this article not only in terms of particular cases, useful to better understand the different conditions, but also deriving general statistics, useful to derive operational indications. In particular, the total number of passes collected during the measurement campaign, from 1 December 2018 to 30 November 2020, is roughly 10,300. After discarding passes not valid, the number of meaningful passes used for the statistics presented hereafter is roughly 8,200. The cumulative distribution (CD) of the residual power level received by the copolar channel is presented for different elevation steps in order to separate and distinguish the effect of snow—mainly above 30°—from the standard atmospheric propagation.
Figures 36 and 37 show the calculated CD considering the entire data set of usable passes, while Figures 38 and 39 show the CD just for the passes recorded with the presence of snow on the radome.
Figure 36. Residuals CD in the 5° to 30° range. Whole data set.
Figure 37. Residuals CD in the 30° to 90° range. Whole data set.
Figure 38. Residuals CD in the 5° to 30° range. Passes with snow.
Figure 39. Residuals CD in the 30° to 90° range. Passes with snow.
This latter data subset comprises approximately 1,300 passes, which is 15.8% of the whole data set. Results show a marked variability in residuals. In particular, the monthly variability is more pronounced as the elevation increases. At very low elevation, where there is little or no impact from the snow accumulation on the radome, the greatest attenuations were registered in summer, consistent with a general increase in humidity and more frequent atmospheric phenomena, such as clouds and rain, with respect to other months.
The highest residuals, recorded in December and January, are simply due to the New Year’s holidays weekend, when no one was on site to regularly clean the radome.
Looking at elevations above 30°, the estimated attenuation is generally greater in autumn and spring, when the simultaneous accumulation of wet snow and ice on the radome, which is difficult to remove with the installed rope, melts and freezes during the day. These months are indeed in the transition period between midnight sun and polar nights. Instead, from November to January the snow is reasonably dry and thus more volatile, which helps the cleaning process or makes it easier to be moved away by the wind.
These operational points are worth stressing. It is important to observe that the snow accumulation removal by dragging a rope, purposely installed at the apex of the radome, along the radome surface itself is a common procedure. However, as highlighted above in this same section, ropes, and even other methods, have serious problems in perfectly cleaning the radome surface, especially under certain weather conditions. Taking into account the numerical results in the “Numerical Models” section, where it is shown that even tiny layers of snow can significantly impair the tracking performance at the K-band, it can be understood that this is an important operational problem.
This article presented experimental and numerical results about a two-year operation campaign involving a ground station, developed in the framework of the project SNOWBEAR and installed at polar latitudes at Svalbard, Norway, used to track an EO satellite, NOAA-20. In particular, the effects at the K-band of snow accumulation on the radome were studied.
We demonstrated that, regardless of the operating mode of the ground station, either program track or autotrack, de-pointing is difficult to predict and to compensate for in real time, due to the unpredictable features of the snow accumulated on the radome (i.e., pattern, thickness, and dielectric properties), and that also the snow remaining on the radome after the traditional removal procedures can be significant enough to impair the antenna performance.
In addition, statistical analysis suggests that most of the critical cases are more likely to occur in spring and autumn, when transitory conditions between winter and summer conditions are in place.
We demonstrated that the effect can be severe, and that de-pointing of the main beam is a significant consequence that can jeopardize the satellite link. This is because at the K-band, the effect of the snow is more relevant, and at the same time is acting to narrower beam widths with respect to lower frequencies. For such ground stations, the importance of a cleaning procedure able to guarantee a perfect snow-free radome should be a key aspect to consider.
Matteo Marchetti (matteo.marchetti01@universitadipavia.it) was with the Department of Electrical, Computer and Biomedical Engineering, University of Pavia, 27100 Pavia, Italy.
Davide Arenare (davide.arenare01@universitadipavia.it) is with the Department of Electrical, Computer and Biomedical Engineering, University of Pavia, 27100 Pavia, Italy.
Filippo Concaro (filippo.concaro@esa.int) is with the European Space Agency, 64293 Darmstadt, Germany.
Marco Pasian (marco.pasian@unipv.it) is with the Department of Electrical, Computer and Biomedical Engineering, University of Pavia, 27100 Pavia, Italy. He is a Senior Member of IEEE.
[1] M. Goldberg and L. Zhou, “The joint polar satellite system – Overview, instruments, proving ground and risk reduction activities,” in Proc. IEEE Int. Geosci. Remote Sens. Symp. (IGARSS), Valencia, Spain, Jul. 2018, pp. 2776–2778, doi: 10.1109/IGARSS.2017.8127573.
[2] K. Holmlund, B. Bojkov, D. Klaes, and P. Schlüssel, “The joint polar system: Towards the second generation EUMETSAT polar system,” in Proc. IEEE Int. Geosci. Remote Sens. Symp. (IGARSS), Fort Worth, TX, USA, Jul. 2017, pp. 2779–2782, doi: 10.1109/IGARSS.2017.8127574.
[3] D. J. Kozakoff, Analysis of Radome-Enclosed Antennas, 2nd ed. Boston, MA, USA: Artech House, 2010.
[4] P. Li, B. Y. Duan, W. Wang, and F. Zheng, “Electromechanical coupling analysis of ground reflector antennas under solar radiation,” IEEE Antennas Propag. Mag., vol. 54, no. 5, pp. 40–57, Oct. 2012, doi: 10.1109/MAP.2012.6348117.
[5] P. Li, W. Pedrycz, W. Xu, and L. Song, “Far-field pattern tolerance analysis of the antenna-radome system with the material thickness error: An interval arithmetic approach,” IEEE Trans. Antennas Propag., vol. 65, no. 4, pp. 1934–1946, Apr. 2017, doi: 10.1109/TAP.2017.2670319.
[6] P. Li, W. Xu, and S. Li-Wei, “Power pattern tolerance analysis of radome with the material property error based on interval arithmetic,” IEEE Antennas Wireless Propag. Lett., vol. 16, pp. 1321–1324, 2017, doi: 10.1109/LAWP.2016.2633576.
[7] W. Xu, J. Zhang, P. Li, N. Hu, and Y. Qiu, “Amplitude-phase-based interval analysis method for radomes with thickness errors and its robust-design application,” IEEE Antennas Wireless Propag. Lett., vol. 19, no. 7, pp. 1103–1107, Jul. 2020, doi: 10.1109/LAWP.2020.2989818.
[8] E. S. Hong, S. Lane, D. Murrell, N. Tarasenko, and C. Christodoulou, “Mitigation of reflector dish wet antenna effect at 72 and 84 GHz,” IEEE Antennas Wireless Propag. Lett., vol. 16, pp. 3100–3103, Oct. 2017, doi: 10.1109/LAWP.2017.2762519.
[9] A. Mancini, R. M. Lebrón, and J. L. Salazar, “The impact of a wet S-band radome on dual-polarized phased-array radar system performance,” IEEE Trans. Antennas Propag., vol. 67, no. 1, pp. 207–220, Jan. 2019, doi: 10.1109/TAP.2018.2876733.
[10] C. Parry, A. J. Fenn, A. Morris, and H. Thomas, “Effect of a wet spherical radome on the reflected power for an S-band planar phased array radar antenna,” in Proc. Antenna Meas. Techn. Assoc. Symp. (AMTA), Daytona Beach, FL, USA, Oct. 2021, pp. 1–6, doi: 10.23919/AMTA52830.2021.9620656.
[11] H.-P. Ip and Y. Rahmat-Samiee, “Analysis and characterization of multilayered reflector antennas: Rain/snow accumulation and deployable membrane,” IEEE Trans. Antennas Propag., vol. 46, no. 11, pp. 1593–1605, Nov. 1998, doi: 10.1109/8.736606.
[12] B. Wang, D. Su, and J. Liu, “Analysis of snow covered radome with ray tracing method,” in Proc. Asia-Pacific Microw. Conf., Suzhou, China, Dec. 2005, p. 3, doi: 10.1109/APMC.2005.1606538.
[13] S. Nechitaylo, O. Sukharevsky, G. Khlopov, and O. Voitovich, “Method of pattern calculation for reflector antenna with surface partly covered by snow,” in Proc. Int. Conf. Math. Methods Electromagn. Theory, Kharkiv, Ukraine, Aug. 2012, pp. 308–311, doi: 10.1109/MMET.2012.6331252.
[14] O. I. Sukharevsky, S. V. Nechitaylo, G. I. Khlopov, and O. A. Voitovich, “Precipitations influence on reflector antenna radiation characteristics,” in Proc. 9th Int. Conf. Antenna Theory Techn., Odessa, Ukraine, Sep. 2013, pp. 447–449, doi: 10.1109/ICATT.2013.6650807.
[15] M. Shimba, T. Sato, and H. Koike, “The antenna pattern degradation by snow accretion on the reflector surface,” in Proc. IEEE AP-S. Int. Symp., Antennas Propag. Soc., Syracuse, NY, USA, Jun. 1988, vol. 2, pp. 871–874, doi: 10.1109/APS.1988.94219.
[16] E. Salonen and P. Jokela, “The effects of dry snow on reflector antennas,” in Proc. 7th Int. Conf. Antennas Propag., York, U.K., Apr. 1991, vol. 1, pp. 17–20.
[17] C. Amaya, J.-M. García-Rubia, P. Bouchard, and T. Nguyen, “Experimental assessment of snow-induced attenuation on an Earth-space link operating at Ka-band,” Radio Sci., vol. 49, no. 10, pp. 933–944, Oct. 2014, doi: 10.1002/2014RS005547.
[18] M. Rytir and J. Sander, “Attenuation due to snow accumulation on antenna at ka-band; effect and mitigation techniques,” in Proc. 25th Ka Broadband Commun. Conf., Sorrento, Italy, Sep./Oct. 2019, pp. 1–9.
[19] G. Maral and M. Bousquet, Satellite Communications Systems, 5th ed. Chichester, U.K.: Wiley, 2009.
[20] A. Martellosio et al., “High-frequency radomes for polar region ground stations: The state of the art and novel developments of radome technologies,” IEEE Antennas Propag. Mag., vol. 59, no. 6, pp. 88–101, Dec. 2017, doi: 10.1109/MAP.2017.2752694.
[21] F. Concaro et al., “The SNOWBEAR project: A Svalbard ground station for wide-band earth observation data reception,” in Proc. 15th Int. Conf. Space Oper., Marseille, France, May/Jun. 2018, doi: 10.2514/6.2018-2478.
[22] M. Marchetti, D. Lospalluto, F. Concaro, F. Romano, D. Cimini, and M. Pasian, “Performance trends at 26 GHz for a receiving ground station at polar latitudes: The SNOWBEAR project,” in Proc. 14th Eur. Conf. Antennas Propag. (EuCAP), Copenhagen, Denmark, Mar. 2020, pp. 1–5, doi: 10.23919/EuCAP48036.2020.9135624.
[23] M. Marchetti, D. Arenare, F. Concaro, and M. Pasian, “Snow on polar radomes and performance deterioration at 26 GHz for a receiving ground station: The SNOWBEAR project,” in Proc. 15th Eur. Conf. Antennas Propag. (EuCAP), Düsseldorf, Germany, Mar. 2021, pp. 1–5, doi: 10.23919/EuCAP51087.2021.9411054.
[24] F. Concaro, M. Marchetti, and M. Pasian, “The SNOWBEAR project: Performance and statistics analysis of the 26-GHz Earth-observation receive channel for a ground station at Svalbard,” in Proc. 16th Int. Conf. Space Oper., Cape Town, South Africa, May 2021, pp. 1–2.
[25] M. T. Hallikainen, F. T. Ulaby, and M. Abdelrazik, “Dielectric properties of snow in the 3 to 37 GHz range,” IEEE Trans. Antennas Propag., vol. 34, no. 11, pp. 1329–1340, Nov. 1986, doi: 10.1109/TAP.1986.1143757.
Digital Object Identifier 10.1109/MAP.2022.3229258