Practical measures for combating communication system impairments caused by large magnetic storms

Authors


Abstract

[1] HF communication circuits, on an individual basis, may perform poorly during ionospheric disturbances associated with geomagnetic storms. This is especially true at middle and high latitudes. To achieve a greater degree of reliability, it is necessary to exploit various forms of diversity. The most powerful forms of diversity at HF must exploit multiple communication paths and frequencies. The most reliable communication with an individual entity (e.g., aircraft) will involve a star network involving several independent paths. Moreover, these (redundant) communication paths must have a wide range of communication bands at their disposal. Much of the paper deals with the process by which such diversity systems may provide acceptable communication reliability, even under adverse propagation conditions. A significant portion of the paper will deal with a particular system during a very realistic scenario. The specific system of interest is the GLOBALink™/HF system, a unique data communication network.

1. Introduction

[2] The GLOBALink™/HF system, developed and managed by Aeronautical Radio (ARINC), is a global high-frequency data link (HFDL) communications network providing service to commercial aviation worldwide. It consists of 14 ground stations located around the globe and a network control center located in Annapolis. The system was designed to provide reliable aircraft communications through the use of multistation accessibility, quasi-dynamic frequency management, and a robust time diversity modem with equalization. Although HF (i.e., 3–30 MHz) signaling can be problematic when considering individual circuits, it has been shown that near-real-time channel evaluation and/or adaptive frequency management can improve performance considerably. Moreover, multistation network operation provides an additional form of diversity, which is probably the most valuable design strategy. Our paper briefly describes the system, but a major segment of the discussion will be about performance metrics derived during geomagnetic storms and especially the Halloween storm period of 2003.

2. Motivation for Study

[3] The Halloween storm period of October–November 2003 was a period of significant ionospheric effects. Large geomagnetic storms were evidenced. We have examined the impact on HFDL of the various phenomena observed during this period. We have found some impact on HFDL performance for the 29–31 October period, but it is manageably small in amplitude. While HFDL is based upon HF propagation, a medium known for its vulnerability to ionospheric variability, the system performance metric does not reflect this vulnerability to an operationally significant degree. This is thought to be the result of the substantial amount of diversity built into the system, including a diverse network of ground stations, and an adaptive frequency management system, Dynacast®, a system developed by Radio Propagation Services (RPSI). The adaptive frequency management system involves the use of active frequency tables (AFTs) that are based upon space weather observables. For the stormy period of October–November 2003, ARINC issued seven changes to the AFTs used by every HFDL station. These changes helped the HFDL network maintain an average delivered message success rate of ∼97%. The paper indicates how this was accomplished, and we examine the attributes of diversity in mitigating the impact of large disturbances.

3. Systems Combating Space Weather Influence

[4] Outside the military arena, it is hard to find many telecommunication systems that take space weather into account as an integral component of the system. In the commercial world, designers typically use diversity to circumvent or mitigate against various forms of impairment. This process is not always successful, even for systems that use frequencies at L band and above. The GPS constellation is a case in point. As indicated above, scintillation can persist well into the GHz frequency regime. While GPS, like many satellite radiocommunication systems, can suffer scintillation in phase and amplitude, it has been designed to eliminate the impact of group path delay errors associated with the total electron content (TEC). Two-frequency receivers can eliminate the ionospheric effect since the GPS L1 and L2 channels suffer different amounts of signal delay for a fixed level of the total electron content. By measuring the time delay (or phase path) difference between the two channels, one can solve for the TEC and, using this information, subtract the excess path delay due to the ionosphere. Unfortunately, two-frequency GPS systems are expensive, and equipage is not widespread. However, engineers are not without imagination. For example, differential Global Positioning Systems (DGPS), used by the U.S. Coast Guard, among other organizations, exploit judiciously located (and fully equipped) reference stations that develop corrections for users within a certain correlation distance. The accuracy of DGPS is directly related to the separation between the reference station and user set. The Federal Aviation Administration (FAA) wide area augmentation system (WAAS) system uses a similar principle but is far more sophisticated.

[5] A number of forecasting schemes and systems exist. For the most part, the outputs (i.e., the forecasts) must be transmitted to system operators who use the data to modify system parameters or operational rules. In short, the forecasting systems are usually nonorganic. Nonorganic strategies predominate because the alternative methods imply increased cost and complexity, but sadly, lack of foresight is another reason. Of course, it makes no sense to invest in organic forecasting systems if the forecasts are not associated with a clear-cut mitigation strategy. In the following paragraph, we discuss systems that are designed to counter, or at least cope with, space weather effects.

[6] HF automatic link establishment (ALE) systems have been discussed in several other papers from the IES2005 conference [Lane, 2005]. ALE is an HF system process that automates many labor-intensive operator manipulations. It also has the provision to use organic sounding to exploit the most appropriate propagating band from among those available. While the system is superior to conventional HF radio, it is still vulnerable to ionospheric effects. ALE systems could be designed to exploit space weather information, including real-time ionospheric data, but ALE processes do not include this option at present. To cope with ionospheric effects, ALE systems exploit diversity countermeasures without the benefit of space weather data to “steer” the system parameters. Under moderate disturbances, ALE systems can perform quite well, and operators are generally well satisfied with HF radios that incorporate ALE, certainly in comparison with the performance of plain vanilla HF radios. However, under highly disturbed conditions, a first-generation ALE network can spend a significant amount of time reorganizing itself for optimal operation [Sutherland, 1993]. From the 1990s to the present, this potential problem has been addressed by second- and third-generation ALE systems. Nevertheless, with space weather nowcasting and forecasting information, it might still be possible to improve the efficiency of link establishment and link maintenance functions. It should be noted that such a suggestion is unlikely to gain much traction, since the ALE protocol is fairly efficient with its existing frequency management strategy, and nonorganic improvements are likely to be an unwelcome expense. In any case, by virtue of its design, HF-ALE can perform adequately in the face of modest space weather effects. The key to an efficient HF-ALE network is the way the scan lists are developed and exploited.

[7] There are two specific systems that exploit space weather directly. Like HF-ALE radio systems that are deeply rooted in the aviation community, these additional systems also support commercial and military aviation. They are (1) the ARINC GLOBALink™/HFDL system and (2) the FAA-WAAS system. In this paper we limit our discussion to the ARINC system. More details are found in a recent book by Goodman [2005].

4. GLOBALink/HF™

4.1. General Remarks

[8] It is generally recognized that the performance of HF communication systems has mixed reviews, at least in comparison with satellite systems. However, as a result of modern technological advance, including the elements of HF-ALE systems, the perception of HF has greatly improved. The HF radio band is extremely vulnerable to ionospheric effects under the best of circumstances. During disturbances caused by space weather conditions, individual circuits can be annihilated or rendered virtually useless. At other times, predicted coverage patterns may become distorted by magnetic storms, and sporadic E phenomena may introduce deleterious screening effects. In short, the situation can be quite unpleasant for a communicator unless steps are taken to cope with the environment. As odd as it may seem, some circuits may actually be improved with respect to climatological projections. The secret to making adaptive HF systems perform optimally is to track the channel conditions. Optimal system performance for a given circuit is achieved if one can successfully match the system parameters to channel conditions. This matching process is not always possible, but there are successful methods for approaching the ideal situation. One method is to employ sounding. This is usually achieved with an imbedded sounder to derive channel properties, but it can involve nonorganic sounders as well. Modern ALE systems, mentioned earlier, employ an imbedded channel probe to assist in organization of an optimal transmission frequency scan list. These methods have been described by Goodman [1991], by Goodman et al. [1996a], and in the ALE Handbook [Institute for Telecommunication Sciences, 1998].

[9] A comprehensive study of HF propagation conditions was carried out between 1994 and 1997 using Chirpsounder® assets. The SNRs for all frequency bands in the aeronautical spectrum were continuously monitored and archived for a substantial number of propagation paths in the Northern Hemisphere [Goodman et al., 1997]. From this database, it was possible to deduce the availability of communication for selected subnetworks. This experimental investigation was the basis for certain feasibility studies for HFDL during architecture and standards development. Design guidance for the HFDL system has been published [Aeronautical Radio Incorporated (ARINC), 1996].

[10] From the aforementioned 1994–1997 study, it was shown [Goodman et al., 1997] that HFDL communications can be as reliable as satellite systems given the diversity attributes than can be applied, albeit with significantly lower data rates. Frequency diversity is well established as a way to improve communications connectivity for point-to-point circuits. Since aircraft have multiple opportunities for connectivity (i.e., in terms of stations and frequencies), it should not be a surprise that HFDL can be successful. For example, if an aircraft has access to eight bands per station and four stations within the calling area, there are potentially 32 independent circuits to choose from. In general, there are fewer circuits than this, but the diversity is still substantial. By contrast, a satellite circuit, while advantaged in other ways, does not have the same diversity advantage (i.e., station and frequency diversity). It has been pointed out that a combination of satellite and HF data link can provide a very high level of connectivity [Goodman et al., 1996b]. Since the failure mechanisms of HF and Satcom are likely to be different, an HF availability of 0.9 and a Satcom availability of 0.99 imply a composite availability of 0.999 or an unavailability of ∼9 h/yr or ∼1.5 min/d. Given this high system availability, what can we say about the distribution of residual system outages? As one would expect, there is a tendency for one class of residual outages to cluster in the temporal neighborhood of space weather disturbances. Other outages are systematic and unrelated to space weather.

[11] In the final analysis, the best way for HF systems to cope with space weather events is to apply two principles of design and operation: diversity in all its relevant forms and adaptivity. Adaptivity implies a capability to sense the channel on the microscale and to assess or predict the propagation environment on the macroscale. The GLOBALink™/HF system, designed and managed by ARINC, employs these principles in some form. In the context of space weather, the GLOBALink™/HF system counters pathological changes in the environment by selecting frequencies that are optimal for use under current conditions. This is achieved by continuously monitoring the environment through a Dynacast® system that delivers AFTs to the network operations center in Annapolis, Maryland. In practice, the delivery cadence of these independent AFT listings is very slow, about once per week, if there are no storms involved. The AFTs in that instance simply reflect the normal daily variations of the global ionosphere. Storm times can drive sharp increases in the AFT delivery cadence. A brief commentary on the AFT system is found in section 4.3.

[12] The HFDL system is certified and has industry approvals based upon findings of the International Civil Aviation Organization (ICAO), the Radio Technical Commission for Aeronautics (RTCA), and the Airline Electronic Engineering Committee (AEEC). ARINC is the sole provider of HFDL service (namely, GLOBALink™/HF), which was inaugurated in 1995. The HFDL data transmission speed is governed adaptively by the prevailing radio propagation conditions. The rates are 300–1800 bps. These rates are relatively low but are acceptable for the mission involved. There are 14 ground stations as listed in Table 1 to satisfy global coverage requirements, including polar coverage.

Table 1. Ground Network for HFDL Network
StationStation DesignationLatitude, degLongitude, degGeomagnetic LatitudeGlobal ServicePolar Service
Dixon, CA, USAH0138.38 N121.76 W+44yesyes
Molokai, HI, USAH0221.18 N157.18 W+23yes 
Reykjavik, IcelandH0364.08 N21.85 W+65yesyes
Riverhead, NY, USAH0440.88 N72.64 W+52yesyes
Auckland, New ZealandH0537.02 S174.81 E−43yes 
Hat Yai, ThailandH066.94 N100.39 E−7yes 
Shannon, IrelandH0752.73 N8.93 W+51yesyes
Johannesburg, South AfricaH0826.13 S28.21 E−37yes 
Barrow, AK, USAH0971.30 N156.78 W+70yes 
Santa Cruz, BoliviaH1317.67 S63.16 W−9yes 
Krasnoyarsk, RussiaH1456.17 N92.51 E+52yesyes
Al Muharraq, BahrainH1526.27 N50.64 E+21yes 
Pulantant, GuamH1613.47 N144.40E+9yes 
Las Palmas, Canary IslandsH1728.12 N15.28 W+18yes 

[13] Of the 14 ground stations listed in Table 1, three of them are clearly within the equatorial region (namely, Hat Yai, Santa Cruz, and Pulantant); three are near the crest of the equatorial anomaly (namely, Molokai, Al Muharraq, and Las Palmas); two would be classified as high-latitude sites (namely, Reykjavik and Barrow), and the remainder would be considered midlatitude sites. Of the high-latitude sites, Barrow is always poleward of the auroral oval and would be expected to represent polar cap conditions, whereas Reykjavik is typically a site that straddles the oval. Because of the fact that the auroral oval, a primary geophysical marker, can move decidedly equatorward under magnetic storm conditions, a number of stations could be considered transient high-latitude sites (namely, Riverhead, Shannon, and Krasnoyarsk) when Kp indices are highly elevated.

4.2. Diversity Schemes Employed

[14] As indicated earlier, the ARINC HFDL system benefits from diversity to countersystem propagation impairments. Frequency diversity is mentioned in section 4.3. The other forms of diversity used are time diversity and path diversity. The time diversity that is used comes in two flavors, coding and message repetition. In general, ground to air traffic is routed to a single ground station. The ARINC “635” protocol [ARINC, 1998] is a connection-oriented system, and thus the aircraft has to log on to each HFDL station. If the initial attempt to transmit a block of data has failed, the system successively retries up to six more times spaced every 70 s. This is not a classical automatic transmission repeat system, but the goals are similar. The ground station knows to stop the retries when it receives a positive acknowledgment (i.e., AK) from the aircraft.

[15] Station (i.e., path) diversity is a significant component of the set of diversity measures utilized. Each HFDL-equipped aircraft has a system table loaded with the transmission properties of each HFDL ground station. As described in the “635” protocol [ARINC, 1997] and the “753” characteristic [ARINC, 1998], opportunities exist for connectivity with any of the ground stations. This enables a powerful diversity capability.

[16] Figure 1 is a map of the HFDL network of ground stations, and Figure 2 illustrates commercial air traffic for a given day in November 2005. It is obvious that the traffic pattern is a nonuniform distribution. Beyond this fact, the aircraft traffic has well-known diurnal patterns and seasonal tendencies. For example, the trans-Atlantic pattern is concentrated between 0100 and 0800 UTC (eastbound) and between 1130 and 1800 UTC (westbound). Other factors such as world conditions (i.e., economy, war, calamity, etc.) can also drive the patterns askew.

Figure 1.

Map of the HFDL network (GLOBALink™/HF) (from ARINC, used by permission).

Figure 2.

Map exhibiting aggregate commercial air traffic using HFDL for a single day in November 2005. There is no time information retained in this composite plot, but it shows the general traffic pattern. There is obviously more traffic in the Northern Hemisphere, and there are certain corridors that dominate the commercial traffic. From a space weather analysis perspective, the more dense traffic regions are the ones that demand the most attention (from ARINC, used by permission).

4.3. Frequency Management Subsystem

[17] The frequency management subsystem involves an appreciation of HF propagation (i.e., coverage patterns) for all propagating frequencies as well as airline traffic patterns. While knowledge of real-time ionospheric conditions is primary in an adaptive frequency management system, we need to derive a set of canonical coverage patterns over which frequency optimization is to be established. While purely dynamic considerations are possible in the pattern analysis, it was decided to convolve the seasonally averaged traffic patterns with the standard HF coverage associated with each ground station, taking the system parameters into account (i.e., antenna, transmitter power, etc.). To first order, this is a nontrivial modification of plain vanilla HF system performance models (i.e., Voice of America Communication Assessment Program (VOACAP) and/or Ionospheric Communication Enhanced Profile Analysis and Circuit Prediction Program (ICEPAC)). The U.S. government maintains HF models such as VOACAP (http://www.its.bldrdoc.gov/elbert/hf.html). However, in the HFDL application, a grid point population defining the desired coverage is weighted by the aircraft traffic patterns.

[18] At any given time, an HFDL ground station is designed to activate two distinct frequencies. The challenge is to activate the best two bands for the desired coverage area for each ground station from among a limited group of available frequencies. This generally requires a real-time adjustment in the ionospheric model that is used to derive the propagation parameters. The primary data sets used as input to the modified ionosphere come from vertical sounders and oblique incidence sounders. Other secondary options include the use of global total electron content (TEC) maps suitable to be analyzed to derive an estimate of the near-real-time foF2 values for insertion into the propagation model. In the TEC option, we invoke a quasi-static slab thickness model to convert the total electron content (in el/m2) to foF2, using the well-known relation

equation image

where the parameter τ is the ionospheric slab thickness in meters and foF2 is the ordinary ray critical frequency of the F2 maximum. The slab thickness is taken to be a slowly varying function of universal time, geomagnetic latitude, and storm time. In its simplest form we take τ = 400 km. The Dynacast® program manages this process and provides an optimal pair of frequencies for each station, taking potential interference and other factors into account.

[19] The frequency management product used by HFDL is the so-called AFT, mentioned in section 2. The AFT is a computer file that specifies the active frequencies for each ground station over a 24-hour period. Under benign conditions, the Dynacast® system submits weekly versions of the AFT, but emergency AFTs are submitted to net control as required by space weather conditions. Emergency AFTs are needed during certain pathological conditions, with ionospheric storms and polar cap absorption events being prime examples. Emergency AFTs are also needed if certain system elements are changed (i.e., new frequencies are added, etc.) Network control disseminates the AFT files to all ground stations for coordination and action. The GLOBALink™/HF system includes the features given in Table 2.

Table 2. Characteristics of the GLOBALink™/HF System
System AttributesBasic ElementFeatures
Organic features of HFDL radio systemHF data radio for data link communication (HFDL)1. digital modem (robust design using adaptive equalization - decision feedback)
2. Time-Division Multiple-Access (TDMA) protocol for message collision avoidance
3. unique “Squitter” message format (i.e., intelligent sounding probe) for protocol control/timing
Topological characteristicsHF ground station network1. two-three active transmitters per station for frequency diversity
2. state-of-the-art antennas for optimum coverage
3. interconnected network of ∼14 stations for station diversity
4. network control (Annapolis)
5. compatible aircraft equipage
Nonorganic processesManagement of resources (i.e., active stations and frequencies)1. frequency planning and coordination for avoidance of radio interference and intrasystem conflicts
2. near real-time frequency management process (Dynacast®) for development of AFTs
3. unique frequency management products for hindcasting, forecasting, and prediction of system performance

[20] The communication traffic for the HFDL system generally exceeds 400,000 messages per month, and the average network-wide message success rate is typically greater than 97%. This is comparable to satellite availability and is far better than typical HF voice circuits that employ predictions to guide frequency selection, even under benign conditions. This clearly shows the benefit of (1) a multinode network architecture (i.e., for path or station diversity), (2) an adaptive HF data radio (using, i.e., exploitation of time diversity and code diversity), and (3) adaptive frequency management (i.e., frequency diversity).

[21] Frequency selection is executed aboard the aircraft, and the airborne receivers are loaded with the 145 possible frequencies. On the basis of proprietary schemes, the receiver scans the list of frequencies and grades each received signal. This is like an organic sounding system. This enables the system to develop a table of ranked frequencies (and ground stations), and this is quite similar to the link quality analysis matrix (LQA) used within HF ALE protocols. When the AFTs are pushed out to the ground stations, the aircraft is forcibly logged off that frequency, and it relies upon its background scans to log on to an alternative ground station. The AFTs are not pushed out all at once. This prevents an aircraft HFDL system from being lost in the channel changes and ensures connectivity. The cutover of the AFT is done to several stations at a time, and care is taken not to cut over an entire region all at once (e.g., North Atlantic group of four (H04, H03, H07, and H17)).

5. Halloween Storm Impact on HFDL

[22] The Halloween storm period of October–November 2003 was a period of significant ionospheric effects. Large geomagnetic storms were evidenced. Patterson and Grogan [2004] of ARINC has examined the impact on HFDL of the various phenomena observed during this period and has provided certain data shown in Table 3. It is a listing from 19 October to 7 November of a daily metric that is proportional to the performance of the global HFDL system. We have added the magnetic activity index Ap for comparison. We see that some impact on HFDL performance on 29–31 October may be arguably present, but it is minimal in amplitude. While HFDL is based upon HF propagation, a medium known for its vulnerability to ionospheric variability, the system performance metric does not reflect this vulnerability to a significant degree. Some of these results have been presented at AGU [Goodman and Patterson, 2004].

Table 3. HFDL Performance During the Halloween Storm Period (19 October to 7 November 2003)
DateUBSR, %Estimated Ap
19 Oct5932
20 Oct6330
21 Oct5939
22 Oct5833
23 Oct5807
24 Oct5734
25 Oct5814
26 Oct5910
27 Oct5615
28 Oct6020
29 Oct51189
30 Oct55162
31 Oct5593
1 Nov5621
2 Nov5918
3 Nov5710
4 Nov5731
5 Nov5709
6 Nov6014
7 Nov5708

[23] The objective HFDL metric in Table 3 is the uplink block success rate (UBSR) rather than the uplink message success rate (UMSR). This is borne principally out of operational necessity. The uplink block metric is used as a feedback tool to give us a feeling of propagation conditions. This metric is only a measure of the first try success rate. This is the relevant statistic for an evaluation of propagation. The daily message success rate is used as well, but it is difficult to correlate a given block success rate with message success rate given the number of retries we use to deliver a message. On the assumption that no more than four tries are accommodated, we have defined an estimated message success rate. The UMSR, in percent, is approximated as follows:

equation image

where M is messages received successfully, Q is the successfully “heard” messages that are nevertheless rejected because of overloading, and I represents the dropped messages.

[24] In Table 3, the metric selected is the uplink block success rate (UBSR) in percent. The message success rate is much higher since there are typically several attempts (i.e., up to seven) made to send a block of data, and time diversity provides a gain in most instances. The diversity gain can be significant if the tries are independent and if multiple tries are attempted within the allotted time span. For example, if the block success rate is 60% and four tries are afforded, the maximum block delivery rate would be >97%, using statistical arguments. As indicated, more than four attempts can be made to attempt transmission of a given block. On the other hand, some messages may be larger than a block in length. The uplink metric data are courtesy of ARINC. The Ap data were obtained from the NOAA Space Environment Center. It is evident from Table 3 that the three lowest reliability days, in terms of the uplink block success rate, are 29–31 October. While this is interesting, a fully satisfactory interpretation of the UBSR for other than the storm days is elusive. Nevertheless, it is clear that the globally averaged UBSR (i.e., the specified metric) is clearly affected by magnetic activity for the period 29–31 October, when the superstorm activity occurred (namely, average Ap ∼ 148), since a suppressed average metric value of ∼54% was observed. In fact, a severely suppressed value of ∼51% occurred on the single day of largest Ap (i.e., 29 October with Ap = 189). These storm day excursions are statistically significant on the basis of almost any test, and most practicing ionospheric specialists would not be surprised at the reduced performance during such large magnetic storms. However, we must always recognize the difference between statistical significance and operational significance. From an operational perspective, an UBSR value of ∼51%, a seemingly low value, may not have any major operational significance. This is because of the block repeat strategy employed in the HFDL system. This strategy has been shown to return message success rates of between 94% and 99% for block success rates of ∼51% and higher. This estimate is based upon quasi-independence in the communication subchannel for the block repeats, and an average of at least four independent repeats out of seven tries. But it is possible for the time constant of the propagation channel to be several minutes at aircraft speeds, and this implies only partial independence of each message repeat. This can reduce the anticipated time diversity gain. Nevertheless, the 94% to 99% estimate for the delivered message success rate is not just academic. It appears to have been validated by an internal ARINC study cited below.

[25] Patterson and Grogan [2004, p. 4] remark that ARINC engineers monitor the solar data coming from the NOAA satellites and issue frequency changes to the ground stations that will be impacted by the solar event." They go on to say

this is the heart and soul of the adaptive frequency management system of HFDL. During the stormy weeks of October and November, ARINC issued over seven changes to the AFTs used by every HFDL station. These changes helped the HFDL network to maintain a delivered message success rate of 97%.

[26] The adaptive frequency management system referred to by Patterson and Grogan is the RPSI Dynacast® system, discussed in section 4.

6. Magnetic Activity and HFDL System Performance

[27] Even though the HFDL diversity schemes tend to reduce the impact of geomagnetic storms, there is still an effect that is measurable. This can be seen by looking at the relationship between the uplink block success rate and Ap. This has been shown time and time again. We illustrate the typical effect in Table 4.

Table 4. Relationship Between Ap and the Uplink Data (Block) Success Rate
Station SetCorrelation With ApNormalized Average Deviation in the Uplink Block Success Rate, %
All HFDL sites consolidated−0.2802.8
Polar (H09)−0.2916.7
Auroral (H03)−0.4607.8
Midlatitude (H01)−0.1010.7
Equatorial (H16)−0.2905.8
Krasnoyarsk (H14)+0.2610.1

[28] As indicated in Table 1, there are 14 operational HFDL stations, all with somewhat different levels of frequency diversity capability. To demonstrate the impact of geomagnetic activity, we have examined data for August 2005 and have looked at stations that represent the following geophysical regimes: (1) polar, H09; (2) auroral, H03; (3) midlatitude, H01; and (4) equatorial, H16. We have also examined a site in Krasnoyarsk, Russia, that for the period of observation was limited to a single frequency and thus was suboptimal in both space and time. A consolidated representation is also given (i.e., first line), whereby the daily average of the HFDL network (of 14 stations) was compared with Ap.

[29] The generally accepted benchmark for statistical significance is the 5% level. This means there is only one chance in 20 if a specified correlation is exceeded that the observed correlation is a fluke. For a month's worth of data (i.e., 31 days) the correlation required to achieve the 5% level is ∼0.30 for a directed analysis (R. Lowry, Concepts and applications of inferential statistics, 2005, Vassar College, Poughkeepsie, N. Y., available at http://faculty.vassar.edu/lowry/webtext.html). It is clear that the auroral correlation of block success rate is significant, while the midlatitude correlation is not. Taking some liberties with the lexicon of statistical mathematicians, the equatorial and polar correlations are borderline, but they are certainly significant at something close to the 5% level. Likewise, the consolidated correlation is statistically significant for all practical purposes.

[30] It is seen that with the exception of H14 (Krasnoyarsk), all correlations between Ap and the uplink block success rate are negative. This indicates that system performance does degrade with Ap in the aggregate and for most individual stations. It should be noted that all correlation coefficients are quite small, indicating that the relationship between system performance and magnetic activity does not overshadow system performance factors quite unrelated to the space environment. The fact that Krasnoyarsk appears to perform better as Ap increases is an HF propagation artifact and shows how a fixed suboptimal frequency can sometimes provide a better coverage match under disturbed conditions than under benign conditions (i.e., climatology).

7. Discussion

[31] We have indicated that diversity is the secret to reliable communication for a network that is challenged by ionospheric variability. While frequency diversity is likely a prerequisite, the power of station diversity is of major significance. We have found that the network-wide throughput is generally acceptable even in the face of the largest of magnetic storm episodes. This is largely the result of the way the HFDL system accommodates to regional variabilities. While conventional wisdom suggests that aircraft communication is best when the ground stations are within one hop (i.e., nominally 4000 km), it is also true that longer paths can sometimes provide the highest signal-to-noise ratio, especially if the shorter paths traverse pathological regions (i.e., auroral oval, trough, etc.). Moreover, higher frequencies, typically used at greater distances, will compete with a lower noise background, since ambient noise (namely, man-made and atmospheric) decreases with increasing frequency. The protocol for HFDL allows for any aircraft to communicate with any ground station in the global network. In principle, any aircraft has up to 30 opportunities to interact with the HFDL network, as there are 12 two-channel and 2 three-channel ground stations distributed globally, with 30 (out of ∼145) frequencies being active at any one time by virtue of the AFT specification. In most instances, it is unlikely that all propagation paths from a given aircraft would be totally impaired. Message throughput reliabilities are enhanced further by the time diversity measures that have been implemented at the data block level.

[32] We have examined how the magnetic activity index Ap impacts individual station performance as well as the aggregate HFDL network performance. At the network level, we see that the uplink block success rate (UBSR) deteriorates with Ap, and the effect is statistically significant, but is not operationally significant. The Halloween storm period of 2003 provided an excellent opportunity to benchmark system performance under the most adverse conditions.

[33] At the station level, we anticipated that most correlation coefficients would be negative, and this was illustrated for the August 2005 period discussed in section 6, with the results given in Table 4. Individual station relationships may vary, and this is especially true if the stations have a less than optimal set of frequencies to cover specified regions over the diurnal period. Krasnoyarsk, Russia, is a case in point.

[34] It was noted that communication traffic tends to “migrate” from one sector to another in response to regional propagation impairments. Given the remarks above, this discovery is reasonable. Even without comprehensive data examination, we would expect such a thing to happen. For example, we would expect high-latitude stations to carry less traffic during the early phases of a magnetic storm, leading to more traffic being supported by midlatitude stations. This, in turn, should lead to more competition for use of midlatitude stations and some migration toward equatorward stations. During X-ray flares, we would expect lower bands near the subsolar point to be preferentially disturbed. Thus longer paths (and higher frequencies) will be exploited, along with paths for which the D region intercept points are associated with higher solar zenith angles. The migration effect in this case would tend to be in the poleward direction and toward the hemisphere not illuminated by the Sun. In order to explore the migration effect more completely, the various stations have been grouped in accordance with the region they serve. Table 5 illustrates possible regional groupings using the station designation codes, as given in Table 1.

Table 5. Sample Station Groupings for Study of the Migration Effects Associated With Disturbances
AsiaAtlanticAustralasiaPacificPolar
H05H03H05H01H03
H06H04H08H02H09
H16H07H13H09H14
 H17   

[35] The traffic migration phenomenon, if borne out by more detailed analyses, is a major factor in the relatively healthy performance of HFDL during disturbances. It is our intent to explore this matter more extensively.

8. Conclusions

[36] It has been shown that diversity is a sufficient factor for achievement of robust performance in specified HF communication systems, specifically the HFDL system. The available data are not in a form to make a stronger statement, but we suspect that diversity may be an essential or necessary condition, especially during stormy periods. Because of the way the GLOBALink™/HF system is designed and operated, the potential diversity gain is significant. Adaptive frequency management enables the best two bands to be specified at each of 14 ground stations within their primary coverage areas. We have found that the HFDL network is remarkably resilient, even in the face of strong magnetic storms that affect the ionosphere adversely. For the Halloween storm period of 2003, the aggregate throughput efficiency was of the order of 97%, good by most standards. The relationship between the aggregate global performance metric and Ap index was not precise, but only the most elevated levels of Ap appeared to reduce the message delivery efficiency noticeably. Still, the overall performance was generally acceptable. There are other ways to improve the HFDL performance, through invocation of improved ionospheric specification as the driver for Dynacast® predictions, and we are actively exploring data-driven assimilation models such as the Global Assimilation of Ionospheric Measurement (GAIM) model for this purpose. Other improvements include access to additional frequencies (and transmitters) in areas where coverage is sparse. From a system perspective, the traffic migration effect is being examined as another factor in active frequency table revisions as a function of storm time. The bottom line is that while space weather perturbations are necessarily a negative influence on radio wave system performance under specified circumstances, the amount of performance degradation can be successfully managed (i.e., limited) if diversity countermeasures are employed, even for superstorms.

Acknowledgments

[37] The authors acknowledge the contributions of other ARINC staff, including S. Baqar, Michael Belt, Scott Beale, and Mike LaFond. Ed Goldberg and Dave Mansoir of RPSI are thanked for assistance in the development and implementation of Dynacast® software that is used in production of the active frequency tables.

Ancillary