---------------------------------------------------- REPORT ON JANUARY 12-13, 1995 GGCM ASSEMBLY WORKSHOP ---------------------------------------------------- From: George Siscoe (siscoe@bu-ast.bu.edu) GGCM Assembly Working Group (WG 5) Report on January 12-13, 1995 Workshop Space Environment Laboratory Boulder, Colorado Local Coordinators: Ron Zwickl and Howard Singer Working Group Chairs: Joel Fedder and George Siscoe Sub-Working Group Chairs: Core Unit--Dick Wolf; Boundary Unit--Bill Lotko; Tail/Substorm Unit--Michael Hesse `There cannot be a greater mistake than that of looking superciliously upon practical applications of science. The life and sole of science is its practical application.' Lord Kelvin The winter meetings of GEM Working Group 5 emphasize applied aspects of the GEM program. Appropriately, this year it convened at NOAA's Space Environment Laboratory. There were over 50 participants. The program was too full to review comprehensively or in detail. This report tries to summarize the meeting in broad terms, with an emphasis on space weather issues, forecaster needs and concerns, agency roles in the National Space Weather Program, and GEM's role in the National Space Weather Program. To con- vey the scope of the meeting, we present the program more or less as it materialized (the times are approximate). Thursday, January 12 Morning Session Chair: G. Siscoe 8:30 Welcome to SEL... E. Hildner 8:35 Opening remarks... G. Siscoe 8:40 NSF's (and GEM's) role in the National Space Weather Program... O. de la Beaujardiere 8:50 NASA's role in the National Space Weather Program... R. Carovillano 9:00 Core Sub-Working Group Reports Introductory remarks...R. Wolf What needs to be done to improve the MSFM?...R. Wolf AMIE capabilities...B. Emery IZMEM capabilities...V. Papitashvili Capabilities of the generalized C-F code...J. Spreiter 10:00 Boundaries Sub-Working Group Introductory remarks...W. Lotko MSM Dst predictions.J. Freeman Dst predictions from solar wind data...G. Lindsay Magnetosheath specification capabilities...S. Stahara Operational boundary position algorithm...H. Singer Benchmarking boundary position algorithms...J. Lyon MHD storm prediction capabilities...J. Lyon Concluding remarks...W. Lotko 11:00 Substorm Working Group Introductory remarks...M. Hesse Microphysics for tail reconnection...M. Kuznetsova Micro-macro physics coupling: Particle and MHD code interaction...D. Winske Merging MHD tail and RCM models...J. Birn, M. Hesse, and R. Wolf Non-linear statistics applied to substorm predictions... D. Vassiliadis 12:00 Lunch and tour of the SEL Forecasting Center Afternoon Session Chair: M. Heinemann 1:30 NOAA's role in the National Space Weather Program... E. Hildner 1:40 DoD's role in the National Space Weather Program... W. Cliffswallow 1:50 USGS's role in the National Space Weather Program... W. Worthington 1:50 STEL's plan to contribute to the Space Weather Program... Y. Kamide 2:00 Forecaster perspectives, problems, and needs Introductory remarks...R. Zwickl Space environment services to the commercial GEO community...D. Speich Measuring performance of new space weather services...K. Dogget 3:00 DoD forecasting concerns...M. Heinemann 3:30 EPRI's Sunburst program...J. Kappenman (L. Zanetti) 4:00 Discussion Friday, January 13 Morning Session Chair: J. Fedder 8:30 Demonstration of MSFM capabilities...J. Freeman 9:00 Predicting geomagnetic indices...Y. Kamide 9:15 Operational GIC predictions with Freja data...L. Zanetti 9:30 Dst predictions at SESC with L1 data...T. Detman 9:45 Open discussion and presentations relating to research leading to understandings that have the potential to be transitioned into space environment specification and forecasting algorithms. Global MHD modeling with 3 D adaptive grids...T. Gomosi SAMPEX >1.0 MeV electron data...D Baker Simulating the 1991 2nd radiation belt event...X. Li Need for a Sun-to-Earth 3 D modeling effort...M. Dryer Status of the Toffoletto-Hill open tail model...F. Toffoletto Indices from global MHD modeling...J. Fedder Afternoon Session Chair: O. de la Beaujardiere 1:00 Open discussion on projects in the overlap between GEM and the Space Weather Program Opportunities for space weather research in NASA's Space Environment and Effects Program (SEE)...D. Evans The workshop covered four broad topic areas: 1. agency roles in the National Space Weather Program; 2. research needs from the view- point of operational space weather specification and forecasting; 3. ongoing research projects relevant to the National Space Weather Program; and 4. high-impact GEM research projects that can make a difference to operational space weather specification and forecast- ing. 1. Agency Roles in the National Space Weather Program: There is a sensitivity among the agencies that their roles be understood and respected by everyone involved in the space weather enterprise. These roles and sensitivities were explained during the Thursday session by agency representatives. The following summarizes these presentations. Operational space weather specification (nowcasting) and forecast- ing are the responsibilities of NOAA's Space Environment Services Center (SESC) and the Air Force (USAF) 50th Weather Squadron (50 WS). SESC supports the civilian sector and nonmilitary gov- ernment agencies. 50 WS supports military operations. NOAA and USAF provide the bulk of the data for space weather operations, but USGS provides essential ground magnetometer data. The forecast centers need an L1 monitor and fast data transfer and validation of most data. These requirements, especially the L1 monitor, involve interagency coordination. The USAF supports research from basic through applied through transition to operations of data-based and theory-based numerical specification and forecasting algorithms. The first theory-based numerical space weather algorithm (the Magnetospheric Specification and Forecast Model, MSFM) is scheduled for transition into operational deployment at 50 WS later this year. This is the first in a suite of specification and forecast algorithms that will cover space weather events from the sun to the ionosphere. Apropos one of the themes of this workshop, the Air Force acknowledges that transitioning algorithms into operational use presents special problems. SESC's parent organization, NOAA's Space Environment Laboratory (SEL), also supports research and technique development to guide forecasters, but it has a smaller budget for these activities than the Air Force. It is increasing its ability to absorb research and technique development, done in house and elsewhere, into operations. It plans to implement more forecast algorithms, including expert systems, empirical mod- els, kinematic models, and (eventually) physically based models. SEL will be involved in coordinating research and technique devel- opment intended for SESC applications though funded by other agencies. Regarding absorbing algorithms developed elsewhere, SEL is concerned that the `owners' be willing to work with SEL in testing their algorithms and transitioning them into operations. NASA's role in the Space Weather Program must be related to its charter. NASA is a mission agency, not a monitoring agency; moni- toring is NOAA's role. Still, there is a tradition established for weather satellites and solar x-ray instruments whereby NASA develops and flies prototypes which NOAA then adopts. This tradi- tion is at work in the space weather arena. For a certain period each day the Wind spacecraft delivers real-time solar wind data for NOAA and the Air Force to use to test their real-time algorithms. Beyond this, the Advanced Composition Explorer (ACE) space- craft, presently under construction, will deliver nearly continuous real-time solar wind data from L1 for operational use. If a talked-of joint Air Force--NASA mission to image solar mass ejections were to materialize, operational space weather forecasting could benefit immeasurably from what NASA would view as a science mission. Within NASA, space weather falls under the Division of Space Physics in the Office of Space Sciences (OSS). OSS and its divi- sions support basic research. In a manner of speaking, virtually all the basic research supported by the Division of Space Physics relates to space weather. The Division's Solar Connections Program can be thought of as providing the understanding and the data that will allow space environment specification and forecast algo- rithms to be built and tested. Moreover the Division supports the development of global numerical models intended as tools for inter- preting space data, but with the potential for transitioning into oper- ational space weather specification and forecasting algorithms. The Division has considered a Quantitative Magnetospheric Prediction Program (QMPP) to take an additional step in the direction of tran- sitioning into applications its expertise in modeling, its data archives, and the new the information coming from its GGS pro- gram. NASA represents a large and valuable potential resource for the National Space Weather Program. Utilizing this resource, how- ever, will require understanding and respecting the institutional perimeters that apply to it. Though it is not ostensibly a mission agency like NASA, in two respects NSF's role in the National Space Weather Program is simi- lar to NASA's: it supports basic research, and it has no responsibil- ity for operational specification or forecasting. There is also a difference. Within NSF, space weather falls under the Upper Atmo- sphere Research Section (UARS) of the Division of Atmospheric Sciences. Among UARS nearest neighbors in the NSF organization chart are the Lower Atmospheric Research Section and NCAR, both of which support research--and in the case of NCAR, conduct research--relevant to the National Weather Service. Consequently, UARS feels relatively comfortable in taking an active leadership role in the National Space Weather Program. UARS leadership along with the leadership of the space weather units of DoD and NOAA and other agencies have composed a draft of a strategic plan for the National Space Weather Program and have circulated it for comments to the space physics and aeronomy community. When implemented, the program will optimize the division of labor between the agencies by reducing duplications and, more impor- tantly, by coordinating interagency efforts in the space weather arena, thereby insuring that information, requests, products, and evaluations cross back and forth from agency to agency. The pro- gram will act to get relevant research identified, to get new funds for the research (everyone hopes), to get the results of that research transitioned into operations, and to get the evaluation of those results sent back to the researchers. Meanwhile, and this pertains to GEM, which is primarily an NSF program, it should be remembered that NSF supports basic research. Transitioning the results of that research into operational algorithms remains at present a problem. There was some discussion of how to deal with this problem, espe- cially within the GEM program. The problem can be restated as, `What is the role of basic research in the National Space Weather Program?' This invites the further question: `Where does basic research end and applied research begin?' The following example was raised: Is it within NSF's purview to support theoretical model- ing to extend the Sibeck-Roelof model of magnetopause position, or the Russell-Petrinec model, beyond its present empirically-based range of validity? This is a good theoretical problem with an immediate application. Putting the results of such research into operatonal forecasting gives a no-nonsense test of the field's under- standing of the physics of global magnetospheric dynamics. This example shows that there is no sharp division between pure and applied research and that the benefits go in both directions. Under- scoring the problem of identifying agency boundaries in terms of a distinction between pure and applied research, it was noted that NASA and NSF fund 50% of the US research published in the Jour- nal of Applied Meteorology. In summary, the agencies responsible for operational space weather specification and forecasting acknowledge that there are logistical, resource, and `owner involvement' problems in transitioning the results of pure or applied research into operational procedures. The agencies responsible purely for research insist that they do not fund the transitioning effort. New to this discussion is the realization that research can move considerably in the direction of applica- tions, even overlapping it, while retaining its pure science identity. As a quid pro quo, applications would provide a validating service, testing the science on which the research is based. We note that meteorologists have been exploiting the synergistic relation between research and forecasting for some time. 2. Research needs from the viewpoint of operational space weather specification and forecasting: Ron Zwickl introduced pre- sentations by two SESC forecasters. As a prelude to the first pre- sentation, we quote from a recent letter in Space News by Dean Olmstead, managing director of Hughes Asia Pacific Ltd., Hong Kong: `There are 145 commercial communication satellites already in orbit around the globe. Another two dozen are scheduled for launch this year, and more than 100 are on order. The international Telecommunication Union has received 900 filings for future sys- tems of all types--and that excludes hundreds of low earth orbit sat- ellites.' The services that SESC presently gives to this community of operators of commercial satellites in geosynchronous orbit (GEO) was the topic of the first presentation, by David Speich. He noted that GEO satellites experience several types of space weather problems: loss of orientation, surface charging, deep dielectric charging, transionospheric propagation, single event upsets, and solar cell damage. To address these problems SESC/SEL/NOAA provides continuous monitoring of GEO environment for retrospec- tive anomaly assessment (together with USAF 50 WS) and warn- ings and alerts to give satellite operators a `heads up.' Warnings and alerts are of several types: solar flare, solar particle event, geo- magnetic storm, energetic electron fluence, and (starting this year) magnetopause crossings. In the future SESC also plans to provide satellite operators daily descriptions of the space environment and education on indices, units, and solar cycle variations. At SEL's last Users' Conference, Speich polled 17 representatives from the satellite operations community about their use of space weather information. He asked them, `Do you want real time space environ- ment data?' They responded that even with such data they would do nothing until an anomaly had actually occurred. On the positive side, the response also implies that accurate forecasts could improve the operators' reaction time or readiness to take action before anomalies occur. He asked them, `Would you run models of the space environment?' They answered, not surprisingly, `No.' The National Weather Service doesn't ask the airline companies to run their weather codes either. And he asked them, `How do you envision the control center of the next century.' Here Speich asked a two-sided question, for besides simply requesting information, it tests this group of operators' foresight. For example, a group of planners would see how a decade hence changes in technology could increase their power to do their jobs; they would respond accordingly. Our user group responded instead with, `About the same as now.' This revealing answer told the audience to expect consumer resistance at the operator level when introducing improvements in space weather products--the well known `We don't need sliced bread' phenomenon. (Later, SEL's Tom Detman, who should know, warned the audience to expect resistance from forecasters, too.) Speich--and the following day, Dave Evans-- noted that from the operators viewpoint the best way to beat space weather problems is to engineer weatherproof satellites. Nonethe- less, Speich acknowledged a trend in satellite design and deploy- ment toward greater vulnerability to space storms. He mentioned another datum that could reduce an operator's confidence in an ulti- mate engineering fix to environmentally induced anomalies--the 1994 ANIK incident aside: Speich attributes more anomalies to non-environmental causes than to environmental causes. Non-envi- ronmental causes are, presumably, related to engineering. The prevalence of such anomalies points up the difficulty and expense of achieving flawless engineering in highly complex systems. The anomalies associated with space weather, while fewer according to Speich, tend to load the system during infrequent, short intervals of space storms. Then they can make the operator's life interesting. For example, Joe Allen reports that during the great storm of March 13/14, 1989, a series of 7 commercial geostaionary communication satellites required more operator interventions than is normal for a year. Despite these occasions when space weather gets the operators' attention, Speich's and Evan's presentations shook the audience out of its complacent view that if you build a better space weather forecast service, satellite operators will beat a path to your door. The operators are looking to the engineers to fix the problem, not the forecasters. This revelation exposed a need to open a dialog with members of an inherently conservative community to explore how their self interest could be served through dramatically increased forecast reliability, specificity, and comprehensiveness. For their part in the dialog, the forecasters must be convinced that new algorithms will increase the reliability, specificity, and compre- hensiveness of their forecasts, or in short, their quality. Apropos quality, SESC's Kent Doggett followed Speich with a presentation on the topic of measuring the performance of new space weather ser- vices. His purpose was mainly to educate the audience of primarily researchers on the five criteria forecasters use to rate forecast qual- ity: accuracy, reliability, resolution, discrimination, and skill. Beyond forecast quality is forecast value, meaning the forecast's use by an operator in making decisions. For a forecast to have value, an operator must be able to use it to help decide which action among alternatives to take. For example, if an anomaly happens when a high quality forecast states the presence at the satellite of strong anomaly-producing conditions, the operator might eliminate non-environmental causes and any actions pursuant to such causes. The requirement that, for a forecast to be of value, an operator must be able to take some action in response to it led Harry Petschek somewhat later to suggest including satellite design engineers in the dialog mentioned above. They might be able to build cost effective operator options into future satellites to take advantage of the increased quality of future space weather forecasts. Doggett con- cluded his presentation with a proposal to hold a verification work- shop that brings together forecasters, modelers, users, verification experts--and if we follow Petschek's suggestion, satellite design engineers--to explore methods and systems for forecast verification and to outline a strategy for space-weather model verification. The audience warmly endorsed the proposal. Compared to the commercial communications satellite industry, the situation regarding the availability of operator actions in response to high quality space weather forecasts is much better-defined in the electric power industry, another major customer for SESC. The actions are so well-defined that the North American Reliability Council (NERC)--which describes itself as `the principal organiza- tion for coordinating, promoting, and communicating about reli- ability for North America's electric utilities'-- following the Hydro- Quebec blackout in 1989, issued a position statement to NOAA which stipulated the following forecast requirement: `NERC believes that a forecasting procedure to provide at least one hour notice and an accuracy of at least 90% is required [to] allow suffi- cient time to implement special operating procedures.' This is a request for forecasts of the timing and amplitudes of ground induced currents (GICs)--the things that damage transformers which can lead to blackouts--at specific geographical sites, where transformers are located. The power industry itself, through the Electric Power Research Institute (EPRI) which is its research orga- nization, has initiated a project called SUNBURST to gather data on GICs. To learn about the SUNBURST Project, we invited its creator, John Kappenman from Minnesota Power, to address the workshop. Unfortunately terrestrial weather prevented his leaving Minnesota. But he faxed his viewgraphs, and Larry Zanetti, who is familiar with SUNBURST, gave the presentation. From it we learned that power systems are evolving to a condition of greater susceptibility to GICs. Engineering fixes are expensive and com- plex. All methods require reliable advanced warning. There are a variety of operator options to deal with GICs given sufficient warn- ing (essentially the NERC requirement applied to every power unit at risk together with an estimate of the beginning time and the dura- tion of the storm causing the GICs). SUNBURST has an archived data base of GICs and a near-real-time GIC-data gathering capabil- ity that might be useful for collaborative studies to test GIC predic- tion algorithms. 3. Ongoing research projects relevant to the National Space Weather Program--Selected highlights: In the same spirit that Bob Carovillano articulated NASA's position, GEM can also say that nearly all it does pertains to space weather. A geospace general cir- culation model (GGCM) is a tool that can be used for pure and applied research. Further, Working Group 5's central goal of creat- ing a GGCM has been expanded to include the creation and foster- ing of stand-alone modular algorithms for use in pure and applied research. A number of such modules were displayed at the work- shop. Work Relevant to GIC Predictions: Larry Zanetti demonstrated an algorithm that uses magnetometer data from the Freja satellite to locate the Iijima-Potemra field-aligned current pattern. Then it predicts GIC conditions on the basis of power facilities rotating under the pattern. He showed examples of successful predictions verified by SUNBURST data. An interesting display of this algorithm can be seen on www: URL address `http://sd-www.jhuapl.edu'. Other algorithms with the potential to predict GICs and ground and low- altitude magnetic perturbations generally were also presented. An algorithm with great flexibility and power is HAO's Assimilative Mapping of Ionospheric Electrodynamics code (AMIE), developed by Art Richmond. Barbara Emery summarized its potential for GGCM use and for GIC predictions. Its power lies in its ability to assimilate many types of data to achieve maximum specificity and reliability in predicting a suite of output parameters. It predicts ground magnetic fields at any geographical position (hence its value for GIC warnings), and it predicts energy input to the upper atmo- sphere from particle precipitation and Joule heating. It runs on a Sun in 1 to 2 minutes, much faster than the data availability rate, which can range from 10 to 60 minutes. If you can forecast its input parameters, you can run it in forecast mode. This is possible using input algorithms driven by IMF and solar wind data from an L1 station. One such input algorithm is the IZMIRAN Electrody- namic Model (IZMEM), which gives as its basic output proxy mag- netic field values from the same ground magnetometer network that was used to calibrate it. As Volodya Papitashvili explained, IZMEM is itself a stripped-down, forecast version of AMIE. It uses only IMF vectors as input; hence, its predictions ignore different magnetospheric and ionospheric states that can exist under the same IMF conditions. Using IZMEM to supply AMIE with data from a proxy ground magnetometer network would synergistically exploit the strengths of both techniques. Another technique with great potential for GIC predictions is nonlinear statistics. Dimitris Vassiliadis and several colleagues are developing an input-state space approach to predicting auroral geomagnetic activity from solar wind variables, especially IMF. Vassiliadis demonstrated the technique's ability to predict AL values that consistently maintain around a 90% correlation with observed AL values, including inter- vals with substorms. As reported later, this work is being taken up by a task group with the intent of developing an operational capa- bility to predict substorms and auroral electrojets. Work Relevant to Satellite Environment Predictions: Magnetically oriented satellites in geosynchronous orbit lose orientation when solar wind ram pressure rises to push the magnetopause close enough to Earth to plunge the satellites into the magnetosheath. A greater hazard to satellites than immersion in the magnetosheath is energetic ion and electron radiation. Referring to the problem of predicting the position of the magnetopause and specifying the radiation hazard as the applied side of the tasks of the boundary layer subgroup, Bill Lotko introduced several presentations on these topics. Howard Singer displayed a new magnetopause posi- tion algorithm based on the Sibeck-Roelof model. Plans are to implement it this year at SESC as an operational code. Being an empirical model, it is valid only for times when solar wind parame- ters fall within a certain not too extreme range. This is because cre- ating the model required sufficient data points to make it reliable. SESC wants also to forecast extreme positions, which affect satel- lites most. As a possible GEM research project having an immediate application, Singer suggested the extension of the Sibeck-Roelof model to extreme values of solar wind pressure, IMF strengths, and Dst values. As an approach to achieving the desired extension, John Lyon showed global MHD simulations of shock waves hitting and compressing the magnetosphere. These were simulations of actual events which revealed unexpected delays between when the model said the magnetopause should cross a satellite and when it did--an instance in which pursuing an application exposed an inter- esting science question. Steve Stahara gave evidence that showed good agreement between observed magnetopause positions and positions predicted by his and John Spreiter's solar wind transfer model (SWT), based on quasi-static pressure equilibrium between the solar wind ram pressure and the magnetic pressure of a vacuum magnetosphere. Regarding the radiation hazard, the highly variable outer radiation belt of energetic electrons is especially troublesome to geosynchronous satellites. Data on the outer belt are being accumulated by the SAMPEX satellite, as Dan Baker described. SAMPEX data are needed to guide and test theories on how outer belt electrons are generated. Baker is testing a version of Nishida's diffusion-recycling theory. Xinlin Li described an alternative theory in which the electrons are accelerated by gradient- curvature drifting in phase with the electric field of a solar wind-induced pressure wave transiting the magnetosphere. This resonant drift-surfing mechanism successfully accounts for the sudden creation of inner radiation belts of energetic ions and electrons simultaneous with SSCs. Baker's and Li's presentations evidenced a resurgence of interest in radiation belt properties and their understanding, an interest directly or indirectly traceable at least in part to the radiation hazard to satellites. The hazard increases dur- ing magnetic storms; thus, one way to forecast increased hazard is to forecast storms, which means to predict Dst. Gretchen Lindsay described her work at UCLA based on improving the 1974 Burton et al. Dst formula and applying it to solar wind data taken at differ- ent distances upstream from earth to see how much in advance it is possible to make predictions that show skill. The Lindsay-modified Burton et al. formula gives truly impressive predictions using data from L1, ~200 Re upstream from Earth. Predictions based on data from PVO, ~7000 Re upstream from Earth, still show definite skill. Predictions based on data from Helios, >14000 Re upstream from Earth, however show little skill in predicting Dst. The Burton et al. formula and its Lindsay modification represent physics-based fore- cast algorithms, which have the advantage of giving meaningful results even in extreme conditions. Statistics-based Dst forecast algorithms are also being developed, since sufficient data exist to make this approach viable for a useful range of conditions. Tom Detman showed results of his testing linear prediction filters and neural networks for possible operational Dst predictions at SESC. Dick Wolf described a neural network Dst prediction algorithm that is installed as part of the magnetospheric specification and forecast model (MSFM). It gives 1 hour forecasts that clearly beat persis- tence. Yoshi Kamide warned, however, that Dst is a global index that says nothing about local conditions: `The solar wind tells the magnetosphere just roughly what to do.: Internal magnetospheric and ionospheric effects add important details. To predict the details, we need something like the adaptive grid global MHD code described by Tamas Gombosi or the MSFM. MSFM--The First Numerical Space Weather Forecast Code: The MSFM is a comprehensive, physics-based algorithm intended for operational deployment. John Freeman gave a highly convincing demonstration of the MSFM's specification and forecast power. It was operating remotely, at Rice University, as close to real time as presently possible (4 hours behind real time because of data delays at several links). This demonstration can be seen at any time on Mosaic or Netscape from a Rice University home page. The URL address is http:rigel.rice.edu/~dmd/index.html. The demonstration is continuously updated to keep up with present space weather. Wolf reviewed the MSFM's attributes: It specifies the fluxes of plasma sheet and ring current energy ions and electrons throughout the magnetosphere (excluding the tail and the magnetopause), elec- tron precipitation fluxes, and electric fields. Given solar wind and IMF data from L1, it can operate in forecast mode. Wolf cited extensions and improvements that could be made to the MSFM. Those especially relevant to operational use include adding an algo- rithm to specify the radiation belts (Greg Ginet at Phillips Lab is working on this), adding an algorithm to use the MSFM's electric field to forecast GICs (a project Gary Erickson suggested at last summer's workshop), and adding an algorithm to predict the onset of substorms (perhaps using Vassiliadis's input-state space approach). Important improvements to the MSFM are also possible in the area of specifying and forecasting the transpolar potential, which drives the MSFM as an input boundary condition. The tran- spolar potential is not directly measured; it must be obtained from a model-dependent reduction of DMSP measurements, which too infrequently sample the polar cap, or from empirically-based for- mulas that use solar wind measurements, which, while in principle continuous, have large statistical errors. Joel Fedder suggested MHD simulations as a way to get more reliable continuous transpo- lar potentials from upstream data. Frank Toffoletto reviewed recent improvements in the Toffoletto-Hill model, which is expressly intended as an input algorithm for the MSFM and its parent code, the Rice Convection Model (RCM). It specifies the transpolar potential by mapping the electric field from the solar wind to the polar cap boundary through an open magnetopause on which the normal component of the magnetic field has been specified by a semi-empirical, quasi-MHD algorithm. Toffoletto announced that the 1993 version of the model is available via anomalous FTP from spacsun.rice.edu. There was an interest in seeing if greater verisi- militude might be achieved by combining the Toffoletto-Hill model with the Spreiter-Stahara SWT model, which gives remarkably good agreement with magnetosheath field observations. Even greater verisimilitude might be obtained if it were also combined with the Spreiter-Stahara generalized Chapman-Ferraro model, that is, a model that for the first time computes the size and shape of a magnetopause that includes the closure current of a cross tail cur- rent and that is in quasi-pressure equilibrium with the solar wind ram pressure. John Spreiter reported the near completion of his and Stahara's project to develop this model. Besides quasi-pressure equilibrium, it features a capability to specify an arbitrary normal component of the magnetic field. Predicting Substorms: The status of substorm research remains directed at determining the physical mechanism responsible for substorm onset. Until the mechanism is identified, a physics-based prediction algorithm is impossible. Michael Hesse introduced sev- eral talks addressing basic issues of tail reconnection as a compo- nent in the substorm life cycle. M. Kuznetsova explored alternatives to beat the problem of getting tearing against the notori- ous stabilizing power of hot, collisionless electrons. Interpreting the obduracy of the substorm onset problem to imply the operation of essential cross-scale coupling, Dan Winske stressed the need for a hierarchy of models from MHD to hybrid to full kinetic. Hesse, Wolf, and Joachim Birn gave a three-part progress report on their project to integrate the RCM with the Los Alamos MHD tail model. An approach aimed at achieving feedback coupling between the codes by alternately feeding the output of one code into the other looks feasible. Because of their necessarily basic research nature, the substorm onset problem and the RCM coupling project will be major topics for the summer Snowmass workshop. 4. High-impact GEM research projects that can make a difference to operational space weather specification and forecasting. During the last part of the workshop participants brainstormed in an effort to identify new GEM research topics with potential for early, signif- icant improvements in the quality of operational forecasting. The following possibilities were discussed: 1. a reference magneto- sphere; 2. a spatio-temporal model of the auroral electrojet; 3. time- dependent radiation belt models--extremes and predictions; 4. mag- netopause locator for extremes conditions; 5. L1-to-magnetopause project; 6. an inventory of models; and 7. a comprehensive mag- netic storm study to acquire a data base with which to guide and test specification and forecast codes. 1. The idea of constructing a reference magnetosphere was sug- gested before in connection with NASA's quantitative magneto- spheric prediction program. A precedent for the idea is the reference atmosphere, much used in meteorology. It would serve many uses, not least of which would be its use as a pedagogical aid for quantitatively describing the magnetosphere to students and the world at large. In the forecasting arena it would provide `cold start' parameters to initialize forecasting codes, but otherwise, as Jo Ann Joselyn remarked, forecasters would not be likely customers for it. Joel Fedder noted that using a reference model to cold start numerical codes requires it to be in dynamic equilibrium, which is an enormous task. John Freeman suggested that a reference magneto- sphere's radiation component could give engineers parameters against which to design satellites. Willow Cliffswallow agreed that the engineers keep asking for better radiation climatology. Harry Petschek thought that this project might make space weather more interesting to the satellite community. Ron Zwickl advised that there are already many radiation models; a new model would need to be put in context with what already exists. Still, he also pointed to the visibility this project might give the field, for example, through its pedagogical use. Freeman suggested that it be taken on by a subgroup of Working Group 5--a `standards committee.' Noting its likely interest to the international community, Baker suggested that a reference magnetosphere might be an appropriate topic to be taken up by the IAGA working group headed by Tujia Pulkkinen. Bill Lotko and Freeman were asked to pursue the reference magnetosphere project further within GEM. 2. The idea of developing a spatio-temporal model of the auroral electrojet for use in predicting GICs was quickly adopted. Vassilia- dis, Baker, and Singer volunteered to see how far the input-state space approach, which already predicts AL, can be extended to give spatial information. Erickson's approach of modifying the MSFM to give GIC predictions still needs to be implemented. 3. The idea of developing a time-dependent radiation belt model will be taken up starting with the summer Snowmass workshop. Mary Hudson has agreed to lead this effort as a subworking group leader. 4. Despite early interest in the idea of extending the Sibeck-Roelof or Russell-Petrinec magnetopause locator model to extremes condi- tions, this project is in limbo for the moment. Any volunteers? 5. The L1-to-magnetopause project is essential for accurate use of L1 data. Vassiliadis and company are tackling it as an element of the auroral electrojet prediction project. 6. An inventory of models will help keep track of the proliferation of models. David Stern has made a catalog of magnetospheric models. Tom Detman, Volodya Papitashvili, and Joachim Birn were asked to bring Stern's catalog up to date for the purposes of GEM activities. 7. Delores Knipp proposed that GEM participate in a comprehen- sive magnetic storm study to acquire a data base with which to guide and test specification and forecast codes. She has identified two storm candidates for which sufficient data exist to perform good AMIE runs. Solar wind data and data from a fair number of satellites in strategic parts of the magnetosphere are available. The storms have already been proposed and accepted for special study by CEDAR. Within GEM, this project could well fall into the subworking group headed by Mary Hudson. Final Words: The workshop succeeded in its usual role of providing a forum for the GEM modeling community to mark progress, exchange information, and plan for future work. It also succeeded in a new way by bringing together for the first time the modeling community and the operations and forecasting community. The first meeting of two things designed for different ends can be expected to have parts that don't match on contact. This is what made the meeting a valuable learning experience. The modeling community learned that it is proceeding under the false assumption that the community of commercial satellite operators is waiting for an improved forecasting service. There is work to be done in defining the positive role that increased forecast quality can play in a satellite operator's professional life. The modeling community must open a dialog with satellite operators, engineers, and forecasters to define that role. On the other side, the operations and forecasting community (we hope) gained an appre- ciation for the kinds of increased value the modeling community can bring to their forecasts. This kind of meeting should be repeated. It is clear that progress can be made through a continued dialog. All workshop participants, especially the working group chairs, greatly appreciate the work and effort of the SEL organizing team, notably Ron Zwickl and Howard Singer.