| Back to Bulletins index | Back to INAG-61 Contents page |
P. J. Wilkinson, IPS, P O Box 5606, West Chatswood NSW 2057, AUSTRALIA email: phil@ips.gov.au
E. Szuszczewicz, SAIC, McLean, Virginia, USA, email: szusz@mclapo.saic.com
A. Danilov, Institute of Applied Geophysics, Rostokinskaya 9, Moscow, 129226, RUSSIA; email: geophys@sovam.com
D. R. Lakshmi, National Physical Laboratory, New Delhi, INDIA; email: npl@sirnetd.ernet.in)
Tim Fuller-Rowell, SEC, Boulder, USA email: tjfr@sel.noaa.gov
T. Maruyama, Communications Laboratory, 4-2-1 Nukui-kita, Koganei, Tokyo 184 JAPAN, email: tmaru@crl.go.jp)
The STP Prediction Workshops are held roughly every four to five years. The last meeting was in Hitachi, Japan, January 1996. While this meeting does not deal only with ionosondes, its conclusions are of general interest to the ionosonde community. At this meeting, as the Draft Report indicates, the Ionospheric Working Group believes a larger, better connected ionosonde network is an important objective for improving our ability to forecast ionospheric change.
There has been a shift in emphasis for the Ionospheric Working Group since the first STP Workshop in 1979. Ionospheric processes are now better understood and physical models can reproduce reasonable, global synoptic and dynamic features. At Forecasting Centres, practical issues of relevance, and hence funding, have resulted in a shift towards applications and, in some respects, away from seeking greater physical insight. Meanwhile, systems affected by the ionosphere are becoming more complex (e.g. HF Automatic Link Establishment - HF ALE) and diverse (e.g. Global Positioning Satellites - GPS). The future challenge will be to develop our understanding of the ionosphere further, test this by making predictions that are compared against some metric, while demonstrating the relevance greater understanding has for modern systems able to capitalise on the information.
These elements are discussed in greater detail below. Tactical issues, that need to be addressed before the next Workshop, are identified together with strategic issues that will enjoy long term support and should form a theme for future ionospheric prediction developments.
What do ionospheric predictions offer? Generally, synoptic predictions for the key ionospheric parameters (e.g. F2 peak, F1 and E region peaks) are readily available. Ionospheric storms can be forecast based on geomagnetic forecasts, but there is little temporal or spatial information in these forecasts of the extent of a depression associated with an ionospheric storm and below some unknown threshold, storms merge with ionospheric variability. Thus the key dynamic process in the F2 region is currently poorly predicted. Other features, such as sporadic E and spread F are not usually predicted at all, although their importance has been appreciated for years.
Are predictions of ionospheric behaviour sufficiently accurate? Surprisingly, this question cannot be answered. No generally accepted measure of prediction accuracy exists. Since a prediction without a measure of accuracy is meaningless, an agreed metric for desirable prediction limits needs to be produced.
This section explores these issues further.
Long term predictions, based on simple empirical models of the ionosphere (e.g. the International Telecommunication Union, ITU; HF Prediction program; International Reference Ionosphere, IRI), have been most successful for planning HF systems and estimating F2 region conditions. These simple models supply information about the ionosphere for a particular epoch of solar activity and, once constructed, may make little use of new ionospheric information. To some extent, their success suggests there is nothing more of significance that needs to be done. Improvements in forecasting long term solar activity will improve these ionospheric predictions.
While the global F2 peak electron density model appears successful it has problems in detail (e.g. low latitude ionisation gradients) and there are also known problems (e.g. at low latitudes again) with the model for the peak height of the F2 region (usually expressed in terms of M(3000)F2 - the M factor). The F1 region peak electron density model also needs to be improved.
Short term: Real time information versus forecasting
Accurate forecasts with short lead times can be useful.
Historically, forecasts of significant ionospheric storm depressions contributed to the management of HF services. "Significant" in this context generally meant an ionospheric storm that accompanied a geomagnetic storm with an Ap exceeding roughly 30 units. While forecasts with experience-based thresholds will still be required, modern systems will use more detailed information.
In the most extreme form, systems can sense the current state on the ionosphere and digest this directly for system management, apparently removing the need for forecasts. Real time management - nowcasting - is likely to play a strong future role; data plus technology being emphasised. In some respects, these systems seek to eliminate the ionospheric effects by designing an ionospheric sensing step into the system at the outset. For instance, the HF ALE system sounds each of the operating frequencies available to it, selects the best and uses this for transmissions. Such a system may collect little information for diagnostic studies, making it difficult to assess where it has succeeded or failed. However, where data are available, it appears prediction models can improve system operation, although the role played is somewhat different from the historic setting.
There will be increased use of real time ionospheric data in the near future and the Working Group will need to take account of this.
Ionospheric variability places a natural limit on predictions.
What is understood by ionospheric variability and does this place a limit on the complexity of prediction models? Day-to-day variability, or ionospheric weather as we call it today, is variability in the prime characteristics of the ionosphere which cannot be linked directly to a single recognisable solar/geophysical event. It is due to the summation of a variety of energy inputs including solar EUV/X-ray, particle precipitation, Joule heating, gravity waves, electric fields generated both by magnetospheric processes and by winds driven by stratospheric and mesospheric heating. We still have an incomplete knowledge of how the wind fields generated by these heat sources modulate the ionosphere.
The sum of these effects is parameterised by the ITU decile factors. The factors are based on a rough spatial segmentation of the globe and temporal segmentation based on day/night and the seasons. Magnetic activity, and prior knowledge, are not used so the factors should form a worst case scenario for describing ionospheric variability.
Better estimates of global ionospheric variability will lead to more effective long term predictions. For instance, in the tropics, people using predicted MUFs bring their operational frequencies down much lower than the predicted MUF to allow for the predicted range of variability and thereby guarantee ionospheric support for their frequencies. This results in higher operational powers and greater radio environmental pollution. Even when real time data is available, the decile factors have not been constructed to respond to the additional information.
Improved dynamic estimates of ionospheric variability are needed. Real time data combined in more effective ways can produce narrower, dynamic bounds for ionospheric variability.
Poor predictions of ionospheric variability limit the value that can be obtained from more sensitive models. Because of the deficiencies in the current treatment of variability, it isn't possible to make quantitative comments on how they might limit models.
Ionosondes are the primary data source for ionospheric predictions. Consequently, the Working Group strongly supports the continuation and extension of the current, global ionosonde network.
The long term data continuity for basic ionospheric parameters must continue to be collected globally for studying long term ionospheric variability and potentially unknown periodicities in ionospheric behaviour.
Digital ionosondes are becoming more prevalent and can give a good variety of data. In addition, new methods are being developed to estimate neutral winds, using these data. This information has become useful for understanding a number of scientific issues. For example: the neutral temperature and wind anomalies in the equatorial ionosphere anomaly region, the night-time maintenance of the ionosphere and the winter anomaly. Although this new knowledge has yet to be applied to predictions, it demonstrates the wealth of information available and emphasises the importance of the world-wide ionosonde network. A further role for the network is in ground truthing space-based systems. Ground-based data will be needed for comparison, calibration and revalidation of satellite measurements.
For real time data to be useful, the current ionosonde network must be expanded and integrated with other data sources to fill in the ocean regions where no ionosonde data are available. Ionosonde networks need to show even greater cooperation; easier data exchange and better formats, both for the data and the archive media, are urgently required.
Summary
The key tactical issue that the Working Group must address is the development of a set of limits that predictions seek to meet and exceed. These limits will be formed from a benchmark that predictions will seek to exceed as well as a target that predictions should attain. Long term predictions, using current sunspot number or preferably ionospheric index predictions, could be the benchmark predictions for the lower bound of ionospheric predictions. Reasonable targets must now be set.
Maintaining the current network is an important tactical goal and expanding and integrating the network with other data sources is a vital strategic goal. To achieve this, freely and readily available real-time ionosonde data are needed.
All predictions, whether short or long term, should to be quantitative, providing an estimate of the most likely ionospheric conditions plus bounds for reasonable outcomes. Currently, empirical models give the estimate and bounds, if mentioned at all, are the ITU decile factors. Using real time data, dynamic decile factors can be constructed so that the bounds can shrink, and expand, according to ionospheric predictability.
Synoptic models can be improved and currently work is taking place in this area. These improvements should be extended to the ITU prediction model.
Finally, there will still be a role for medium term (several days to a few months) ionospheric predictions for at least the next decade.
Physical models aid understanding of the complex coupling web that forecasts seek to describe. Empirical models may well be the interface that puts the forecast into the system environment. Meanwhile, many poorly predicted ionospheric phenomena (e.g. ionisation irregularities, sporadic E, spread F) affect systems and statistical models of these phenomena may be the best support that can be offered at present.
This section looks in more detail at these models.
The advances in understanding of the ionospheric response to geomagnetic storms have evolved from analysis of results from numerical simulations using mature physically-based models of the upper atmosphere.
One of the most obvious manifestations of an ionospheric storm is the deep "negative phase" (decrease in NmF2), which develops in the main phase of a geomagnetic storm, and often persists well into the recovery phase. The realisation that thermospheric composition changes (increases in molecular species) are a likely cause, has been appreciated for many years, but the cause of the seasonal, local time, and longitude dependence was not understood. Numerical models have recently shed light on the physical processes involved, and provide the hope that an empirical description of ionospheric storms can be developed.
Numerical simulations confirm the theory that the "negative phase" is mainly caused by the changes in neutral composition. During a geomagnetic storm, the magnetospheric sources drive upwelling of molecular rich gas from lower altitudes to F region altitudes. The increase in molecular species forms what is termed a composition "bulge", which can cover a wide geographic region. Numerical simulations now indicate that the local time response is caused by the diurnal migration of this composition "bulge" by the background wind field, and the seasonal dependence is controlled by transport of the "bulge" to mid and low latitudes by the prevailing summer-to-winter circulation.
Modelling also shows that the changes of thermospheric composition are regional, i.e. the composition "bulge" can be preferentially located over one longitude sector, and it will be this sector that experiences the most pronounced decrease in NmF2. The strength of the negative phase, and the longitude sector most affected, will depend on the Universal Time of the main driven phase of the geomagnetic storm. The sector most affected will be that passing through the midnight sector during the times of the main energy input, and the effect will be biggest when this sector also coincides with the longitude containing the magnetic pole. In the latter case, the composition "bulge" is the most asymmetric, creating the greatest longitude dependence in the ionospheric response.
The increase in understanding from the physically-based models have given us hope that a relatively simple empirical expression could be formulated that would capture many of the seasonal, local time, and regional variations.
At the 1979 meeting, there were few comments on the ionospheric support for, or affects on, systems. Now, empirical models, sometimes used out of their original context (e.g. ray tracing with the IRI), are being used to support an increasing range of
applications. Where is this leading? Most probably towards tailored, empirical models embedded in systems. These models will be based on physical models and will be driven by real time data.
There are many areas where more sophisticated empirical models can be used. Adaptive models, based on empirical models and currently available data are a growing area of interest. For instance, Mikhailov has developed an adaptive system for foF2 for the SPARC model. Patterson, at IPS, has developed another adaptive model, based on ionospheric indices. These models use current data, making effective real time, empirical HF predictions available. In the future, adaptive models may have an even greater role, for instance, combining diverse data sets, such as electron density at one altitude, electron temperature at another, hmF2 from a different place. Such algorithms for finding the final result will obviously be based on theoretical relations between the various ionospheric parameters.
Regional models
Real time, regional data are being used more. Because of the regional nature of the ionosphere, global forecasting has limited meaning. Regional forecasts are valuable, although their value will depend on the successful interpretation of good magnetospheric forecasts (i.e. storm models need correct inputs). In the near term, real time data are likely to be communicated to systems using empirical models, rather than physical models, and the empirical models are likely to be simple and heavily tailored to suit the application.
High latitude ionospheric modelling may require a family of regional models; e.g. polar cap, auroral oval, trough, etc. For each region an empirical model can be developed, having its own set of input parameters. While magnetospheric-thermospheric-ionospheric progress may allow coupled models to deal simultaneously with these regions, for practical applications this time could be sufficiently far off to make empirical models an attractive strategic goal.
Possibly, the distinction between empirical and statistical models is artificial. Here, empirical models are assumed to have greater physical understanding attached to them and are convenient simplifications of reality. Statistical models may be based completely on data and are a reasonable attempt to summarise gross knowledge of a phenomenon.
Current empirical models of the ionosphere (e.g. ITU, IRI and many others) are based on relatively straightforward statistical models of the main ionospheric characteristics. Other features (e.g. sporadic E, spread F) are still not integrated into prediction systems. While sporadic E models have been available for many years, they are not used regularly and it is unclear how many sporadic E models exist nor which models should be used (e.g. models of foEs or fbEs). Scintillation models (e.g. WBMOD) have obvious value, but are not generally available for wider use. Over the next 5-10 years there could be over 1000 new communication satellites in low Earth orbit. Most will be using K or L band which presumably will be impacted by scintillations. Since physical and empirical models are not very good at predicting scintillation, particularly the changes in the global pattern during a storm, a lot of work will need to be done in this area over the next 5 to 10 years.
The single most important forecasting tool currently missing is a successful storm model. Any reasonable model in this area will be very welcome. Work is in progress to develop such an algorithm and is an important tactical goal for the group. Continued refinement of physical models, leading towards data assimilation models promises to be the most exciting long term prospect for ionospheric predictions.
In the near term, wider use of real time, data driven, empirical models is an important, potential development area that will probably grow over the next few years. Coupled with this is the likely increased importance of regional models.
Increased use of transionospheric systems will make the prediction of scintillations especially important and this will lead to a greater need for statistical models and possibly physical models of other ionospheric features not discussed at this meeting.
Bulk ionospheric ionisation supports many systems: e.g. HF broadcasting, HF direction finding, LF, VLF, ELF navigation, HF communications. Some of these systems are used less now (e.g. ELF, VLF) while others (e.g. broadcasting) will be in use for many years to come and some are being modernised (HF ALE).
For others systems, the ionosphere is, potentially, a system hazard: e.g. VHF through S-band and higher satellite communications, satellite radar, GPS, DMSP. Ionisation irregularities within the ionosphere may be the main limitation for transionospheric systems of the future.
Different systems react differently. Much more can be done to describe how the ionosphere, and forecasts, effect systems and improve system management. To ensure relevance we must find more customers, learn what their systems do and how ionospheric information can make them more efficient. We need to be sure about what ionospheric parameters have to be forecast. Forecasts have limited value without an application and often the value of a forecast will be measured by the support it gives to a system.
In recent years, in some areas, the ionospheric research community has been ignored by some customer groups. A number of reasons can be cited to explain this. Some of the big customers now regard HF as a back-up service and, faced with financial restraints, cut-back on ionospheric investments. Many groups feel there is nothing more of relevance they need to learn about the ionosphere - current F2 region empirical prediction models are reasonably faithful reproductions of past ionospheric behaviour. Also, the ground based, ionospheric community has lost some credibility because the advances in ionospheric research are not as spectacular as in computer and satellite communication systems.
Nevertheless, there are many applications that are sensitive to ionospheric change. Some examples of new service areas are offered here.
In many developing countries, with scattered, rural populations, HF is still attractive and more research is needed to optimise performance of HF digital packages. Furthermore, the tropics are populated by most of the developing countries that need to be educated about the advantages of HF. In an apparently developing mobile satellite communication world, cheaper HF systems have a role to play. There is much more fundamental work yet required to fully understand the low latitude ionosphere and its variations and to integrate this knowledge into modern HF systems.
The effects of irregularities on satellite based systems has been known for many years. Systems, such as GPS, are affected by the bulk ionisation and corrections for this are required for accurate measurements. While dual frequency GPS systems are available, these are still expensive, making ionospheric corrections important. Furthermore, GPS systems are likely to be affected by ionospheric scintillation in the equatorial ionosphere.
Currents flowing in the E region of the ionosphere, induced by magnetospheric changes, produce magnetic changes at the surface of the earth which can add errors to aeromagnetic surveys. While a magnetospheric effect, for logistic reasons this may fall into the category of an ionospheric problem because of the short time scales involved.
For systems to operate, signals must be useable within the ambient noise environment. For instance, HF field strength and atmospheric noise predictions may be region specific, e.g. in India, large departures from ITU predictions are found. While not directly an ionospheric issue, it is a customer issue that often attracts the attention of ionospheric groups.
Understanding the ionospheric contribution to systems covers a wide range of topics, including non-ionospheric effects.
It is difficult to separate the customers' service problems into simple packages labeled: ionosphere and "other". As funds become more difficult to obtain, and relevance takes on a broader meaning, the Ionospheric Working Group will have to become more familiar with the total service problem. Ideally, customers fully understand their systems, recognise the ionospheric needs and seek assistance. In practice, where systems are inherited rather than designed, people are often unfamiliar with their overall needs and appreciate informed assistance. Effort spent on understanding the effects of the ionosphere on systems should improve the application of predictions but it may not improve the predictions themselves. Experience suggests that if reliable predictions are available, customers will probably seek them. However, there is often little interest, aside from curiosity, in services offered unsolicited. For some groups, solving this conundrum is the key tactical goal.
The key tactical issue that the Working group must address is the development of a set of limits that predictions seek to meet and exceed. These limits will be formed from a benchmark that predictions will seek to exceed as well as a target that predictions seek to attain.
For real time data to be useful, the current ionosonde network must be expanded and integrated with other data sources to fill in the ocean regions where no ionosonde data are available. Maintaining the current network is an important tactical goal and expanding the network, and integrating it with other data sources, are vital strategic goals.
The single most important forecasting tool currently missing is a successful storm model. Any reasonable model in this area will be very welcome. Work is in progress to develop such an algorithm and is an important tactical goal for the group. Continued refinement of physical models, leading towards data assimilation models, promises to be the most exciting long term prospect for ionospheric predictions.
For practical reasons, much greater attention will need to be
paid to how the ionosphere affects systems and, in the short term,
it may be necessary to sell solutions to potential customers to
stay in business.
| Back to Bulletins index | Back to INAG-61 Contents page |