Space Weather Models Applied to Radiation Risk Reduction
A.1 SPACE WEATHER MODELS
To optimize our ability to avoid radiation risks to humans in space, a system should be developed that can provide "timely, accurate, and reliable space environment observations, specifications, and forecasts."1 At present, there are significant gaps in our understanding that diminish the ability of space weather models to perform this function. Nonetheless, the science of space physics has matured to the point where it is able to describe and model many aspects of the complex links between the Sun, the interplanetary medium, and Earth. These capabilities are essential for predicting space weather in general and the radiation environment at low Earth orbit in particular. Recent advances in numerical technology and computer architectures have meant a rapid growth in our ability to model space weather, in particular the propagation of solar eruptions through the heliosphere. The last few years have also seen rapid advances in our capabilities for representing the present state of the magnetosphere and providing short-term forecasts, and progress is likely to accelerate in the next few years.
The presently available empirical and semiempirical models that have the potential for providing predictions that will be useful during the ISS construction period should be adapted for operational use. CSSP/CSTR notes at the outset one criterion bearing on the likely usefulness of a model. From a flight director's perspective, false alarms are worse than missed events. Models with low false alarm rates are therefore the ones most likely to give predictions that flight directors would trust enough to act upon.
A.1.1 Solar and Inner-Heliospheric Models
CMEs and flares are the first links in the chain of efficient causes that connects eruptions on the Sun to space storms at Earth, including SPEs. To understand the solar origins of space weather, therefore, we must understand how CMEs and flares are initiated. The following discussion will focus primarily on fast CMEs, the predominant source of SPEs and geomagnetic storms. At present, we know the following:
The Sun's magnetic field is the most likely source of the substantial energy needed to launch and maintain a CME.
Most CMEs come from the streamer belt or from the boundary between the polar coronal holes and adjacent active regions.
CMEs are not caused or driven by flares; in some cases the causal relationship appears to be reversed, but in many cases the relationship appears more complex.
Fast CMEs and some slow CMEs drive shocks, which in turn accelerate particles.
CME initiation is still poorly understood. Existing theories of CME initiation generally can be categorized according to the underlying driver: gas-pressure-driven models, including buoyancy and flare-driven mechanisms; ideal magnetohydrodynamic (MHD) models, based on the emergence of subphotospheric flux ropes into the corona and their subsequent loss of equilibrium (for example, through prominence draining, which removes "ballast" holding down the flux rope); and resistive MHD models, which invoke either the formation and detachment of a flux rope through reconnection underneath (the so-called tether-cutting process) or the removal through coronal reconnection of magnetic flux above a stressed configuration, which then is free to erupt (the so-called breakout model).2-5 Most CME initiation models have only been studied analytically or through two-dimensional simulations. The use of solar observations to establish boundary conditions in such calculations and the move to three-dimensional simulations have both become feasible only with the advent of high-performance computers.
Modeling the propagation of fast CMEs and their associated shocks through the heliosphere, even with one-and two-dimensional MHD codes, has reached the point of demonstrating measurable skill in predicting shock speeds and arrival times at Earth.6 Recent improvements to these models include the incorporation of three dimensions, the ambient solar wind structure, and the heliospheric current and plasma sheets.7 Once an eruption has begun, it is a relatively straightforward matter to simulate its propagation as that of a pulse travelling through the heliosphere, as long as the plasma and magnetic-field conditions along the way are known. Unfortunately, this requirement becomes increasingly difficult to satisfy at and after a solar maximum, when close sequences of CMEs or eruptive flares and a highly distorted heliospheric current sheet prevent the establishment of a predictable, quiescent interplanetary medium. Real-time monitoring of the ambient interplanetary medium properties to establish background conditions would improve the accuracy of CME and shock propagation models. In addition, we do not know where these shocks first develop, which is a key determinant of the particle acceleration process addressed below.8 Better observations of, and insight into, the sources of type II solar radio emissions, which give information on shock speeds, would be valuable for initializing and testing models. Finally, the relationship between CMEs, which are identified as solar phenomena, and magnetic clouds, which are observed as interplanetary structures, is an unresolved puzzle that merits increased attention by both observers and theorists. Despite these caveats, the heliospheric phase of space weather development is reasonably well understood. Consequently, models of interplanetary propagation hold great promise for supporting NASA's efforts to keep astronauts from being exposed to harmful levels of radiation during ISS construction and beyond.
The next link in the Sun-Earth chain is the acceleration of particles by CME shocks. Several theories exist for this process,9 quantitative modeling of which provides the only feasible way to connect the MHD characteristics of the CME-shock system with the flux and spectrum of SPEs at Earth. This is a challenging task, as both large-scale MHD characteristics and small-scale particle properties must be considered. A related problem is our incomplete understanding of the flare-driven component of SPEs. Observations by SOHO, Wind, and other spacecraft are clarifying some of these complex phenomena.
Now and in the foreseeable future, only MHD models can span the enormous distances and range of scales from the Sun through the heliosphere into the magnetosphere. A principal long-term goal of the National Space Weather Program (see Section A.5) is to develop a three-dimensional, multiscale model of the heliosphere-magnetosphere-ionosphere system extending from the base of the solar corona to the base of the ionosphere. In its fully developed form, this ultimate space weather model will solve a set of multispecies, generalized, time-dependent MHD equations and will self-consistently describe the complicated interplay among the physical processes controlling the structure and dynamics of the heliosphere and the geospace environment, including the solar wind outflow, the generation and propagation of transient interplanetary structures, such as CMEs or corotating interaction regions (CIRs), and their interaction with the magnetosphere-ionosphere system.
A.2 NEAR-EARTH SPACE ENVIRONMENT MODELS
In the Sun-to-Earth flow of space weather influences, the Sun is the producer, the solar wind the deliverer, and the magnetosphere the receiver. The preceding section dealt with production and delivery. This section deals with reception.
A.2.1 Magnetospheric Conditions Influencing SPE Penetration
As shown in Chapter 2, for several orbits on most days ISS enters the high-latitude regions, which during SPEs are filled with solar energetic particles. Accurate forecasts of astronaut safety therefore depend critically on accurate estimation of the physical extent of solar particle penetration. Among the forecaster's tools are maps of the equatorward edge of the auroral oval. In addition, forecasters can calculate the ''solar particle cutoff." The ability of energetic particles to reach a specified point (latitude, longitude, and altitude) above the Earth can be characterized by the solar particle cutoff, which is controlled by a particle's energy (strictly speaking, by its rigidity, i.e., its momentum per charge). Solar particles with energies above the cutoff value can penetrate to the given location, whereas less energetic particles cannot. Although cutoffs traditionally have been calculated without considering the instantaneous state of the magnetosphere, we now know that the SPE zones grow when the magnetosphere is compressed by a CIR or a solar storm or inflated by a strong storm-time ring current.
A.2.2 Dynamic Physical Models of the Radiation Belts
Since their development in the 1960s and 1970s, NASA radiation belt models AP8 and AE810,11 have been very widely used to provide quantitative representations of the average observed particle fluxes. Only long-term space weather effects have been included, primarily by the creation of separate Solar Max and Solar Min NASA models. A new set of empirical models, including the electron model CRRESELE and the proton model CRRESPRO, was developed recently by the Air Force Research Laboratory based on measurements from the CRRES mission of 1990-1991.12,13 CRRESPRO is separated into quiet, active, and average models. The quiet model is taken as an average over the ~7 months preceding the great storm of March 1991, and the active model is an average over the following ~7 months, including the secondary proton belt formed by the March storm. The average model is the average over the entire 14-month mission. The CRRESELE model consists of six models ordered by the 15-day average of the geomagnetic activity index Ap (a relative of the Kp index mentioned in Section 2.3), plus mission-average and worst-case models. For certain radiation environment studies, direct dose measurements are preferred to flux measurements because of the uncertainty involved in transforming one to the other. CRRESRAD is a directly measured dose model constructed from CRRES data that are sorted by the same time intervals as CRRESPRO;14 APEXRAD is a low-altitude dose model constructed from APEX satellite data and sorted by geomagnetic activity in a manner similar to that used for CRRESELE.15 Although these empirical models take some account of various levels of geomagnetic effects, they obviously do not predict the detailed response of the radiation belts to variations in the magnetospheric configuration, as would a physics-driven model.
On the theoretical side, radiation-belt models traditionally assume adiabatic particle motion, with deviations characterized by diffusion coefficients chosen to fit the observed fluxes. This approach has been very effective in interpreting the average or quiet-time configurations of the radiation belts, but it cannot represent the dramatic flux increases due to solar storms, which pose the greatest risk to astronaut safety. Among the features that are not well described by the diffusion codes are the rapid changes that occur frequently in the outer-belt MeV electrons. For example, these particles typically disappear rather suddenly at the beginning of a geomagnetic storm caused by a CIR, remain at a low level from the main phase through the early recovery phase, and then return with flux levels sometimes orders of magnitude higher than the prestorm values. As discussed in Chapter 3, the decay rate of these enhanced electron fluxes can be estimated once the peak has been observed, but we cannot yet reliably estimate the timing or intensity of the peak. The physics governing the peak flux of these outer-belt electrons is not yet understood but is the subject of intense research activity (see Chapter 3 and D.N. Baker et al.16).
A dramatic case of nondiffusive behavior occurred on March 24, 1991: a remarkably strong interplanetary shock smashed into Earth's magnetosphere, briefly bringing the magnetopause to about 3 Re above the terrestrial surface and creating a new radiation belt between the inner and outer belts. In fact, the changes in the radiation environment wrought by this event were so substantial and long lasting that, as noted above, two sets of CRRES-based empirical radiation-belt models were developed to represent the averages before and after this event. The main features of these radiation-belt disruptions were quickly and convincingly explained in terms of a theoretical model that followed gradient-drifting particles in a time-varying magnetic field, first using an ad hoc quantitative representation and later a realistic, three-dimensional MHD simulation of the sudden compression (Plate 4).17 Note, however, that this simulation followed equatorially mirroring particles through only a few minutes of magnetosphere time.
A.2.3 Real-Time, Data-Driven Map of the Radiation Belts
The number of spaceborne instruments currently monitoring the radiation belts in real time is sufficient to provide the desired comprehensive picture of the belts if the data from these spacecraft can be effectively merged. Prime candidates include the Comprehensive Energetic Particle and Pitch Angle Distribution (CEPPAD) instrument on NASA's POLAR spacecraft, particle detectors on monitoring spacecraft operated by NOAA and Los Alamos National Laboratory, and the simple particle detectors on several Global Positioning System spacecraft. If these data can be optimally merged, making use of the known coherence characteristics of the radiation belts, a useful real-time map of the condition of the belts could be constructed.18
A.3 ADVANCED EMPIRICALLY BASED FORECAST MODELS OF RADIATION RISK PARAMETERS
Neural net and nonlinear dynamics models of radiation risk parameters may provide the quickest way to bring forecasting capability to radiation risk management.19 A large community of researchers is working in the area of applied empirical modeling. Neural nets have been trained to perform diverse forecasting tasks, including the following:
Make short-term flare predictions with a success rate reported to be around 80 percent;20
Predict the total dose of an SPE, reportedly to within 20 percent, from the fluxes observed earlier in the event;21
Predict up to an hour ahead geomagnetic disturbance indices from L1 solar wind and interplanetary magnetic field (IMF) data,22 which can then be turned into predictions of the size of the SPE zones (see Section 2.3); and
Predict the intensity of relativistic electrons at geostationary orbit a day ahead, based on a series of past Kp values, reportedly with an efficiency better than 90 percent.23
The evidence suggests that models based on neural nets and nonlinear dynamics can be developed that measurably reduce the radiation risk entailed in ISS construction and operations. A way is needed to focus and coordinate the development efforts on such a project.
A.4 OBSERVATIONAL SUPPLEMENTS TO MODEL-BASED FORECASTS
One approach to improving such forecasts lies in monitoring for "halo" events—CMEs directed at or away from Earth, which appear as expanding annuli or disks of enhanced density roughly centered on the Sun. Although these events were discovered by the SOLWIND coronagraph (SOLWIND) two solar cycles ago,24 the coronagraphs currently operating on SOHO's LASCO experiment are the first to observe a significant number of halo CMEs, thanks to their extended field of view and their improved sensitivity compared with earlier coronagraphs.
Because coronagraphs are constrained to observe limb events best, however, their data alone do not distinguish those CMEs aimed toward Earth from those originated on the far side of the Sun. The source region can be derived best from concurrent extreme ultraviolet observations of the lower corona using, for example, the Extreme Ultraviolet Imaging Telescope (EIT)/SOHO, which is sensitive to the high-temperature plasmas and transient dimmings characteristic of eruptive events. Vector magnetograph observations (and extrapolations into the corona) taken a few hours before CME onset would roughly indicate the orientation of the CME magnetic field at its origin, providing one factor that contributes to the extent of southward IMF at Earth. Closer to Earth, IMF monitors (e.g., ACE, SOHO, Wind, and IMP 8) can provide the missing link in magnetic topology; this would diminish considerably the lead time between a CME-based warning and the onset of geomagnetic effects and would also reduce the number of false alarms.
A striking illustration of the value of this approach is given by recent studies comparing LASCO halo events with large geomagnetic storms. In 1995, the accuracy of such forecasts was poor: only 27 percent of 173 observed storms were forecast correctly, while 63 percent of the 126 forecasts were false alarms.25 In contrast, the study of 25 front-side halo CMEs seen by LASCO and EIT during 1996 and 1997 revealed that over 85 percent portended geomagnetic storms with Kp greater than or equal to 6, and only 15 percent of such storms were not predicted (C. St. Cyr, personal communication, 1999). Note that no solar magnetograph or in situ IMF data were used in this study, so it is likely that fewer false alarms would have been reported if this context information had been available. Further work is needed to test and refine this procedure with a larger data set and to determine the cause of the geomagnetic storms, which are especially worrisome because they are apparently unaccompanied by detectable halo CMEs. In the meantime, this promising approach can and should be implemented by NASA to lessen radiation hazards during ISS construction.
Radio-frequency emissions associated with type II bursts give information on the outward propagation speed of the ejected plasma, in particular, providing important clues as to the origin and speed of shocks both close to the Sun and in interplanetary space. However, type II bursts that are first detected in the low corona are primarily associated with impulsive flares, whereas interplanetary type II emissions are clearly associated with CMEs. In addition, the mechanism whereby the shocked plasma produces type II emissions is poorly understood. Finally, some type II emissions have been detected during slow CMEs, which in theory should not be able to drive shocks. Until these uncertainties surrounding the nature of type IIs are clarified, their usefulness as a proxy or early warning of oncoming solar storms remains compromised.26
The majority of SPEs are associated with CME-driven shocks, which take a day or two to reach us, thus in principle allowing plans for astronaut activities to be altered as needed. The problem is that unless measurements are made of the source region and the magnetic field geometry of the approaching shock/CME system, as well as of the white-light signature, many false positives will be predicted, canceling out the advantage of the long lead time. Similarly, if forecasters do not detect most of the CMEs aimed at Earth, the relativistic shock-accelerated particles unfortunately will be the sole "warning" of the oncoming geomagnetic storm. Continuous monitoring of the Sun using SOHO-type instruments could provide early warning for many if not the majority of SPEs and geomagnetic storms.
A.5 NATIONAL SPACE WEATHER PROGRAM
The National Space Weather Program, which is managed by the National Science Foundation (NSF) but which also integrates and coordinates the space weather interests of five other agencies—NOAA, USAF, NASA, DOI, and DOE—provides an institutional structure that could help implement the actions suggested below. In broad terms, the goal of the NSWP is to improve the ability of the nation's providers of space weather services to nowcast and forecast the space environment accurately. As the following quote from NSWP's Strategic Plan makes clear, many development areas discussed in this section fall within its purview: "The National Space Weather Program encompasses all activities necessary for the timely specification and forecast of natural conditions in the space environment that may impact technical systems. The domain of primary interest to the program includes the sun and solar wind, the magnetosphere, the ionosphere, and the thermosphere. Because of the vastness
Table A.1 Space Weather Parameters and Goals
and complexity of the region of interest, all traditional areas of space sciences can contribute to achieving the program goals."27
The approach the NSWP is taking to achieve its goal is also compatible with the development of models that can be applied to reducing radiation risk to ISS construction and operating teams. It calls for developing both data-based models for the near term and deterministic models for the middle and long term, consistent with this passage from the Strategic Plan: "In order to substantially improve forecasting abilities, several advancements must take place. We can make some, albeit limited, progress without improving our understanding of the physical mechanisms which generate space weather. This will require examining and applying data to develop or improve statistical or empirical models. However, in parallel, as our understanding of the space environment improves, physics-based research models will be developed and modified as part of the process of improving our understanding."28
The Strategic Plan specifically calls for developing nowcast and forecast models in areas of direct importance for managing radiation risk for ISS construction and operating teams. Table A.1, which is an excerpt from Table II.2 in the NSWP Strategic Plan,29 shows that the models needed for ISS radiation risk management form a subset of models whose development is called for by the NSWP. The priorities advocated in this Appendix are, therefore, addressed to the affiliated agencies of the National Space Weather Program.
A.6 SUMMARY AND FINDINGS
Space weather modeling aims to develop models that use information from places where instruments happen to be to specify and forecast conditions at places where the information is wanted. Instruments are deployed in space and on the ground in arrangements intended to optimize their usefulness. Nonetheless, there are strategic holes in instrument coverage that planned missions will help fill, as discussed in Chapter 4. On the modeling side, the situation can be described as encouraging. Physics-based, MHD-type models have progressed to advanced stages in all the links that connect solar storms to terrestrial effects. Data-based, neural net, nonlinear filtering models appear to be close to producing operational-quality forecasts of radiation risk parameters.
The CSSP/CSTR findings address the specific elements of an effort that would lead most directly to reducing radiation risk by providing high-quality information on the parameters most crucial to assessing radiation risk. These findings identify and prioritize two kinds of project. The first kind includes relatively mature projects of central importance to radiation risk management that will provide results soon. It also includes projects that will ultimately provide comprehensive (end-to-end) information relevant to radiation risk management and that therefore merit early institutional emphasis. The second kind includes projects that are less mature: although they will ultimately be of great value, it is likely to take longer to realize the results of their operational use in radiation risk management.
A.6.1 Projects Deserving Earliest Possible Attention
Because the following projects could, if started now, be completed in time to help reduce radiation risks during ISS construction and operations, they deserve the earliest possible emphasis by one or more of the affiliated agencies of the National Space Weather Program:
Critically evaluate and develop the best of (at least) the following methods that have been suggested for mapping latitudinal cutoffs for SPE particles at the altitude of ISS:
Numerical integration of particle trajectories, using semiempirical models of the magnetosphere (e.g., Tsyganenko30). Results should be compared with observations to assess the degree of improvement offered by this approach.
Direct inference from real-time monitoring of magnetospheric boundary locations. For example, magnetometers on board the Iridium spacecraft should be capable of mapping the equatorward edge of the auroral oval. Solar protons are observed to penetrate to lower latitudes when auroras reach lower latitudes, but more research is needed to define the quantitative relationship between the location of the edge of the auroral oval and solar-particle cutoffs.
Numerical integration of particle trajectories, using three-dimensional global MHD models of the magnetosphere driven by real-time solar-wind data upstream from Earth. In the near future, it will be technically feasible to run a magnetospheric model of this type in real time at a forecast center, thus improving short-term (up to about an hour ahead) forecasting. Such global MHD models provide the only means for obtaining a realistic representation of the magnetosphere under extreme conditions, e.g., in response to the impact of an interplanetary shock followed by a CME.
Several existing advanced, data-based space weather nowcast and forecast codes could be developed relatively quickly into operational codes that can give SEC and SRAG the ability to forecast at least some radiation-risk parameters during most of the ISS construction period.
A prototype of a comprehensive space weather simulation tool could be developed, tested, and made available to a forecast center in time for the peak radiation hazard during ISS construction, given sufficient investment of resources in the near future. Cooperation among the concerned agencies—NASA, NOAA, DOD, and DOE—would be essential for cost-effective and timely progress. By using measurements of the solar magnetic field obtained by ground-based observatories and spacecraft and by running it faster than real time on a high-performance computer, this model would be able to make 4- to 7-day predictions of the near-Earth space environment. Utilizing real-time inputs from upstream spacecraft, it should be able to predict many components of the near-Earth space environment several days to several hours beforehand.
Dynamic radiation-belt models (i.e., models that respond over time to changing input conditions) are technically feasible and will aid the monitoring and short-term forecasting of conditions near manned and unmanned spacecraft that can be endangered by high particle fluxes. As mentioned, the ability to run a global magnetospheric model at a forecast center is anticipated within the next couple of years. Once observational tests of the radiation-belt model (with input from the global MHD simulation) show that it has sufficient accuracy, the model should be put into real-time operation and made available for use in the ISS program. Sufficient resources should therefore be devoted to developing a dynamic theoretical model of the evolution of the radiation belts in time-dependent magnetospheric electromagnetic fields calculated by a global three-dimensional MHD model of the magnetosphere driven by measurements of the solar wind.
If observational tests show merit, a scheme should be developed and implemented for calculating a real-time, data-driven map of the radiation belts that uses as input observations by available monitoring spacecraft.
A.6.2 Projects Deserving Early Attention
The following projects, which should eventually contribute to reducing radiation risk in ISS operations, merit early attention from one or more of the affiliated agencies of the National Space Weather Program:
Quantitative, theoretical attacks on the particle acceleration problem are sorely needed. Numerical simulations alone can illuminate the details of the acceleration process under realistic conditions, but such efforts have only just begun. Aspects of shock acceleration in SPEs still elude quantitative understanding. Among the less mature areas of understanding in the field of Sun-Earth connections, this area is particularly critical. Until we understand how particles are accelerated at shocks, there can be no first-principles numerical code for predicting SPE parameters from solar inputs.
Projects aimed at understanding the underlying physics of CMEs and solar flares are needed for developing an end-to-end capability for forecasting the hazards of radiation for humans in space.
It is time to move beyond existing one- and two-dimensional models to three-dimensional MHD simulations in order to correctly model the initiation of solar storms (and the associated shocks), their propagation through the interplanetary medium, and their impact on the near-Earth environment. Observational inputs, preferably real-time, to initialize and drive MHD simulations are now feasible and can impart much-needed realism to the modeling results. The combination of heliospheric modeling with real-time solar data could be used to predict arrival times and magnetic-field characteristics of CMEs directed toward Earth with a 1- to 2-day lead time.
Most of the impulsive SPEs reach Earth nearly as rapidly as any electromagnetic signatures (~8 to 80 minutes), leaving insufficient time to make appropriate changes in most EVAs. As a result, flare monitoring with present and upcoming spacecraft will provide only short-term warning, at best, of oncoming energetic particles from such events. To lengthen the lead time available for incorporating information on flare occurrence into EVA scheduling, flare prediction models should be given increased attention.