Use of the Auroral Boundary Index for potential forecasting of ionospheric scintillation

Transcription

1 RADIO SCIENCE, VOL. 47,, doi: /2011rs004954, 2012 Use of the Auroral Boundary Index for potential forecasting of ionospheric scintillation James M. Griffin, 1 Thomas C. Connor, 1 and Hilary E. Snell 1 Received 8 December 2011; revised 5 April 2012; accepted 11 April 2012; published 19 May [1] The Hardy-Gussenhoven Auroral Dosing Model (HGADM) was developed to compute electron characteristic energy and energy flux values onto the global grid and is often used to generate the inputs for other phenomenological models. Forecasting auroral conditions is limited by rapid changes in the ionosphere due to variable solar conditions. However, through a statistical analysis of Auroral Boundary Index data we have developed a technique which allows us to forecast/predict the appropriate inputs to the HGADM, thereby providing a means of forecasting the characteristic energy and energy flux values. This paper will initially discuss the statistical analysis and the development of the forecast mode for the HGADM. We then discuss the possibility that aurora-based indices along with other environmental indicators can be correlated to ionospheric disturbances. Citation: Griffin, J. M., T. C. Connor, and H. E. Snell (2012), Use of the Auroral Boundary Index for potential forecasting of ionospheric scintillation, Radio Sci., 47,, doi: /2011rs Introduction [2] Nominal ionospheric scintillation is diurnal by nature because sunlight drives the population levels within various layers via photoionization. Elevated ionospheric scintillation is often caused by short-duration, energetic events. Among some of the more significant ones are the Sudden Ionospheric Disturbance (SID) and Polar Cap Absorption (PSA) by incident X-ray bursts resulting from solar flares, F2 layer reduction from geomagnetic storms due to coronal mass ejections (CME) from the solar surface, and D region modifications due to lightning strikes from below within the atmosphere. The mechanisms for these variations are varied, and extensive modeling efforts are underway to include all of the phenomena impacting the ionosphere [e.g., Belehaki and Stanisławska, 2004]. In order to properly validate these models, comparisons of the model results to measurements of archived events need to be differentiated by the type of sources of ionospheric disturbances. While it is desired to model all ionospheric variations, one of the most important factors in making such models operationally useful is the ability to forecast the occurrence, duration, strength, and extent of this scintillation beginning with the real-time detection of the onset of any external phenomenon. To best address that functionality, any methodology that has the forecast capability for a phenomena 1 Atmospheric and Environmental Research, Inc., Lexington, Massachusetts, USA. Corresponding author: J. M. Griffin, Atmospheric and Environmental Research, Inc., 131 Hartwell Ave., Lexington, MA 02421, USA. (jgriffin@aer.com) Copyright 2012 by the American Geophysical Union /12/2011RS should be utilized within the ionospheric scintillation models. Any phenomena indicator that is the most reliable should be coupled with the methodology. [3] While considerable literature are focused on the lowlatitude ionospheric scintillation [e.g., de La Beaujardière et al., 2004], some of the research also examined the issues associated with scintillation induced by the auroral oval formation and activity, for example the irregular features present in high-latitude scintillation which can impact the signals from satellites of a GNSS constellation [Spogli et al., 2009]. Because these features can impact satellites, the ability to forecast the onset of ionospheric scintillation due to changes in the atmospheric conditions is needed by operational users in the commercial and military sectors in order to take effective measures against potential outages of space-based communication links and GNSS functionality. [4] Based on the need to forecast the component of ionospheric scintillation associated with the geomagnetic storms, we investigated the modification and application of the forecast algorithm add-on that we had developed for use with the Hardy-Gussenhoven Auroral Dosing Model (HGADM) [Hardy et al., 1985, 1987]. The forecast algorithm was designed to provide an enhanced set of inputs to the HGADM, allowing the HGADM to be used in a forecast mode without any changes to the model itself. These inputs are computed from prior and nowcast values of geomagnetic activity indices and provided as the timeline of the most likely projected geomagnetic activity inputs to be used by HGADM. To best develop and drive the forecast algorithm, we have made use of the Auroral Boundary Index (ABI) as the preferred indicator. In this paper we discuss the background of both this forecasting algorithm developed for HGADM, and the ABI and how both could be useful in forecasting geomagnetic indices 1of8

2 Figure 1. Polar grids of the mean electron energy fluxes are shown for the seven Kp bins as computed by the Hardy-Gussenhoven Auroral Oval Model. The bin names are given in yellow inside the polar grid along with the individual Kp indices listed below the polar grid for which the DMSP SSJ particles were binned. The polar grid is a function of geomagnetic latitude and geomagnetic local time with the continental outlines shown for when the geomagnetic midnight is at 0 UT. The color scale is the linear scale of the order of magnitude for the energy flux values. for use with models that compute ionospheric scintillation when geomagnetic storms impact the ionosphere. 2. Auroral Oval Model [5] The HGADM [Hardy et al., 1985, 1987] is an empirical model used to relate positional auroral dosing parameters to a geomagnetic activity index such as Kp. The empirical coefficients were derived from the binning of measurements related to precipitating charged particles as detected by Special Sensor Precipitating Plasma Monitor (SSJ, formerly Special Sensor Precipitating Electron and Ion Spectrometer) onboard the Defense Meteorological Satellite Program (DMSP) [Hardy et al., 1984; Schumaker et al., 1988]. The DMSP satellites measure the meteorological, oceanographic, and solar-terrestrial physics environments with a myriad of instruments from within a 101 min, Sunsynchronous near-polar orbit at an altitude of 830 km above the surface of the earth. As one part of the space environment sensor package, SSJ/4 instruments initially record electron and ion fluxes with energies ranging between 30 ev to 30 kev over 16 channels; newer instruments (SSJ/5) measure 20 fixed energy channels between 50 ev and 30 kev. [6] As initially outlined by Hardy et al. [1984] and Schumaker et al. [1988] the SSJ electrons and ion counts are binned by geomagnetic latitude zones and geomagnetic local time. These groups were in terms of both the geomagnetic activity as measured by the 3 h planetary index Kp and the magnetic latitude of the equatorward edge of the oval at the midnight sector. The average results from the statistical analysis of the accumulated values were fitted with functional representations that allow for calculating over the global grid the electron and ion integral energy flux, number flux, average energy, Hall and Pedersen conductivities [Hardy et al., 1987]. The model can be driven by the use of seven Kp group values, geomagnetic latitudes of the auroral edge, or 30 combinations of solar wind speed times the IMF. To achieve this capability, tables of fitting coefficients for the functional representations are used as a lookup for geomagnetic activity; the functions themselves then provide the means to smoothly project the binned mean values onto the 2-D polar grid. [7] To demonstrate the utility of HGADM in reconstructing the mean aurora oval position and the corresponding electron precipitating dosing parameters, we present the model outputs projected onto a geomagnetic polar grid shown in Figure 1. In the figure, the global footprints of the participating electron energy flux over the northern polar region are plotted as a function of geomagnetic latitudes and geomagnetic local time for the seven Kp bins. Note that each of the grouped bins includes measurements associated with the three Kp values except for the last bin where measurements for all Kp values greater than 6 are collected together. Since the spatial positions of the model values from HGADM are based on geomagnetic local time and latitude, 2of8

3 Figure 2. Polar geographical grid of the mean electron energy flux values (Kp = 6 ) from the HGADM for 3 June 2008 when both the DMSP F15 and DMSP F16 passed over the northern auroral oval within 5 min of each other. The SSJ data provided the measurements for which the equatorward edge of the oval is extracted and used to determine the geomagnetic latitude of the edge at the midnight sector. That value is the Auroral Boundary Index (ABI). we can map the model results to geographic positions for any time of the day. 3. Auroral Boundary Index [8] Another derived quantity from the DMSP measurements is the Auroral Boundary Index (ABI). Originally referred to as the Midnight Equatorial Boundary (MEB), it is derived from determining the geomagnetic latitude of the equatorward boundary of the measured auroral oval and translating that positional value to the equivalent geomagnetic latitude of the edge located at the midnight sector of the oval as shown in Figure 2 [Gussenhoven et al., 1982]. During a quiet time when the solar wind from the Sun is flowing at its nominal steady state with no enhanced energetic plasma infusion from either a solar flare or coronal mass ejection, the auroral oval resides well within the high latitudes. However, under geomagnetic storm conditions the precipitating electrons deposit their energy over the poles causing the auroral oval to expand equatorward. It has been found that the oval will have a preferred shape and position for a given change in the geomagnetic field configuration due to a specific auroral dosing [Whalen et al., 1985]. Thus the magnetic latitude of the equatorward edge of the oval can be used as an index for level of auroral dosing by the precipitating electrons. The auroral boundary indices for years 1983 to current ( php/dmsp:ssj4_midnit) are derived from the SSJ data collected by DMSP satellites passing over the Polar Regions [Madden and Gussenhoven, 1990]. The time resolution of the indices is based on the number of flights actively making measurements and the spacing of their orbits. With only one satellite, the time step is about min between polar passes because both the northern and southern oval regions are sampled by the SSJ instruments. For that reason, ABI has a higher temporal resolution than the single Kp value with the latter assigned to each 3 h block as the weighted average of maximum fluctuations of horizontal components of the geomagnetic field observed relative to a quiet day from the data collected by the 13 ground magnetometer observation 3of8

4 Figure 3. Three hemispheric displays of the modeled mean electron energy flux values colored in two-tone with red being most intense and the yellow being the lower dosing region within the oval. The green means that there are little or no expected precipitating electrons. Note that this model is based on actual ABI values extracted for 12 September 2005 during a storm period with maximum Kp = 6 +. Note that the oval has considerably shifted its orientation and size whereas using a single Kp value would have not been able to replicate same dynamic measured responses. stations within that 3 h block (for more details go to the official NOAA Kp website). ABI also has an effective equivalent Kp value associated with it; this means that any geophysical model dependent on Kp can be driven by ABI indirectly simply by exploiting their linear relationship [Madden and Gussenhoven, 1990]. 4. Forecast Capability Using ABI [9] In order to construct a forecast algorithm, we made use of the ABI values to take advantage of its higher temporal resolution because the aurora oval can drastically change its size, strength, and position within a short period of time. This is evident in Figure 3 where the ABI-driven oval underwent considerable changes within just 45 min. This is consistent with the reports by ground observers located in the midlatitudes that auroral displays typically last only about a half hour within their field of view. This means that while a single Kp value is conventionally used to represent the average global state of the geomagnetic field within the 3 h period, the auroral dosing can be highly variable and can affect different regions during that same 3 h period. We felt that it would be best to account for all temporal variability of the aurora oval in order to have a better characterization of the timeline associated with the auroral dosing. [10] To construct the forecast algorithm we looked at the archived ABI values over 1.5 solar cycles. We tried to determine how often the oval changes, whether or not there are any preferred geomagnetic latitude orientations, and at what rate the aurora oval expands and contracts from any given position. Our approach was to statistically examine the expansion and contraction rates as a function of geomagnetic latitudes to determine the rate at which the aurora changes shape and whether or not there is a preferred rate for that aurora to shift from that position. The rationale behind the statistical study is that we already have information on how often the oval will be with any ABI value, and the issue is the behavior of the oval over time for a specific value of ABI. The goal is to examine whether the aurora will be consistent with its behavior or that it would have a generalized trend associated with it. If the results are that the aurora oval is chaotic by nature and that there is no preferred response to a particular geomagnetic storm s intensity and duration, then the forecasting skills are considerably more difficult to develop. Further, the continuous real-time monitoring of the geophysical environmental conditions will be required at a higher temporal cadence. [11] To that end, we performed numerous statistical analyses and found that we can indeed characterize auroral oval modifications to first order by knowing just its current ABI index and its previous five values. The statistics and the corresponding rule-based algorithm of describing the mean oval recovery used for making the forecasts were developed by examining the first 18 years of ABI measurements (see Figure 4 for some examples of the analysis done). The most frequently occurring value (i.e., the mode of the indices) for the ABI is about (geomagnetic latitude) with a standard deviation of about The average rate of change in the ABI deviating from the mode (i.e., depressing ABI) is about 0.18 /h, while the average rate for the ABI relaxing to the mode is about 0.16 /h. The rules-based algorithm starts by looking at the trend in the ABI measurement over the previous 5 h. If the trend is negative and the last measured aurora is within one sigma of the mode, then the aurora is allowed to continue its expansion phase at the average onset rate until it reaches the one sigma level, at which point it recovers back to the mode at the average rate of recovery. The same rules apply when the last measured ABI is between the one and the two sigma levels, the two and the three sigma level, etc. to a maximum sigma level of five (that is, the ABI is about 10 below the mode). If the trend is positive, then the ABI is simply allowed to continue at the average rate of recovery to the mode (or the forecast period expires). If the last measured/forecast ABI is greater than the mode, then the remaining forecast values are simply set to the mode since they are indicative of low auroral activity. This capability is limited to addressing the occurrences of a single geomagnetic storm within that time period, and we are still investigating the issue of multiple storms occurring closely together and we will report on the conclusion of that work. 4of8

5 Figure 4. Examples of the statistical analysis of the ABI temporal variability used to define a rules-based forecast model: (a) the average relaxation time for an oval to return to its quiescent state from a given expanded state and (b) the average rate of relaxation or expansion from the quiescent state. [12] So, the ABI values can be used to drive the forecast add-on to the Hardy-Gussenhoven Auroral Model. The next logical step is to determine whether the same forecast can be applied to determine the strength and duration of the ionospheric scintillations caused by geomagnetic storms impacting the upper atmosphere. 5. A Sample Case Study of Ionospheric Scintillations [13] Ionospheric scintillation calculations are typically done using models that can determine changes within the ionospheric composition due to short-term variability induced by onset of environmental events. For standard models, mean conditions of the ionosphere can be computed based on climatology and are modified with variations added to account for prior occurrences of geomagnetic storms. One such model is the International Reference Ionosphere (IRI) 2007 [Bilitza and Reinisch, 2008]. IRI is based on the monthly means of the electron density, electron temperature, ion temperature, and ion composition profiles in the altitude range from 50 km to 2000 km as a function of position and date. IRI has an option to read in a database of either the default or user-specified Ap values to construct the geomagnetic storm history to drive the calculations. In order to evaluate the utility of our HGADM/ 5of8

6 Figure 5. Depiction of the two sets of Ap values. The box symbols are the 3 h blocked planetary values, and the continuous line is the ABI-based Ap values at the full fidelity as determined by DMSP passes over the north and south oval regions. The red circled portion is the timeline of the Halloween storm of 2003 used for the case study. ABI preprocessing for forecasting, we first developed case studies for which we can compare outputs from IRI model runs using historical database consisting either the nominal Ap or the ABI-based Ap values. [14] In one of our case studies we examined the conditions for the Halloween Storm of 19 October to 5 November This storm is one of the stronger geoeffective events which occurred within the previous solar cycle [Weaver et al., 2004]. The power grids were severely affected within the northeastern region of the United States and outages lasted for several hours [Weaver et al., 2004]. For this investigation, we used the IRI 2007 model to provide electron density profiles utilizing the storm history based on the timeline of preceding Ap values. The immediate objective was to construct and incorporate an ABI-based Ap database in the format as required by the IRI 2007 model. The goal was to see immediately whether there was any variability between the model results based on two sources: the nominal Ap value set and those derived from the ABI archived database. Figure 5 shows comparison of the Ap sets in which the 3 h nominal Ap values were plotted as individual square symbols against the ABI-based Ap values at their fullest fidelity. The figure demonstrates that both sets of Ap values are similar in trend and amplitude but there are more excursions exhibited by the ABI-based values that will drive more variability in the models that used Ap indices. [15] Figure 6 shows a comparison of the electron density profiles computed for the nominal 3 h Ap values and the ABI-based Ap values at a position along the equator at 1400 local time. Note that the differences in the model responses to storm history manifested themselves at altitudes greater than 100 km and reached a maximum difference above 200 km. This suggests that the ionization conditions are similar at lower altitudes simply because the daytime photoionization processes dominate the ionosphere s interactions. For radio signals with higher frequencies that will penetrate at a higher altitude, increased scintillation due to strength of geomagnetic storms will be more of an issue. [16] Since the input Ap array to IRI model is read in as a 3 h blocked set, we selected the minimum ABI (i.e., the geomagnetic latitude for the maximum oval extension) for each of the eight daily 3 h blocks. This means that the highfidelity variations were not fully utilized within this model and more investigation is needed to best incorporate their dynamic properties. Our results suggest that the storm history governed the overall secular form but the short-term responses are apparently suppressed. [17] The exercise was to show that the ABI-based Ap values can be used with the IRI 2007 model with similar results. Since we have developed the forecast algorithm to project ahead the ABI values, an equivalent set of forecast 6of8

7 Figure 6. (left) The percent differences as function of altitude for electron number density profiles computed by the IRI 2007 model using both the nominal 3 h Ap values and the ABI-based derived Ap values. (right) The two distinct profiles from both model runs with the nominal Ap case shown as circles and the ABI-based Ap case shown as the solid line. Ap values can be constructed and be fed into the ionosphere model. This approach is in effect the ability to forecast the ionospheric conditions as suggested by the algorithm as to the likelihood of specific geomagnetic activities. Therefore, any ionospheric scintillation model that requires as inputs the geomagnetic activity indices can be driven by this forecast algorithm. Because this algorithm can be used with the HGADM, any ionospheric modeling that requires knowledge of the positions, durations, and levels of the auroral dosing changing over a period time can also be fed by HDAGM. 6. Conclusion [18] Our adaptation and use of the auroral model based on the ABI values has provided us an ionospheric forecast capability with a temporal responsitivity afforded by the use of the ABI values. As shown, the Ap values of both the measured and predicted sets are similar in trend and range, but the role of the ABI-based Ap data is that it can be extracted in near real time from DMSP passes, has a higher temporal resolution, has a greater sensitivity within the dynamic range of geomagnetic activity, and offers another way to be used with scintillation models to help better diagnose ongoing ionospheric disturbances. [19] Acknowledgments. Portions of this paper are from the results of the data products and work performed by Atmospheric and Environmental Research, Inc. (AER), for the following SBIR Phase II Efforts: MDA, W9113M-06-C-0152, Simulation of Stressing Optical Clutter for Scene Generation ; USAF, FA C-0023, Nowcasting/Forecasting the Battlespace Environment ; and USAF, FA C-0018, Real-Time Specification of the Battlespace Environment, under the supervision of the initiators, James H. Brown and Delia E. Donatelli, AFRL-Hanscom/ RVBYB, with additional or ongoing analysis done under AER Internal Research and Development efforts. The DMSP particle detectors were designed by Dave Hardy of the Air Force Research Laboratory, and the Auroral Boundary/Hemispheric Power Indices are provided with permission from the Space Vehicle Directorate, Air Force Research Laboratory, Hanscom AFB, Massachusetts 01731, via the Cedar Database at the National Center for Atmospheric Research, which is supported by the National Science Foundation. The authors would like to thank Pan Liang (AER) for her assistance with running the IRI 2007 model. References Belehaki, A., and I. Stanisławska (2004), Impact of variability of space environment on communications: Working Group 1 overview, Ann. Geophys., 47(2 3) suppl., , doi: /ag Bilitza, D., and B. Reinisch (2008), International Reference Ionosphere 2007: Improvements and new parameters, J. Adv. Space Res., 42(4), , doi: /j.asr de La Beaujardière, O., and the C/NOFS Science Definition Team (2004), C/NOFS: A mission to forecast scintillation, J. Atmos. Sol. Terr. Phys., 66(17), , doi: /j.jastp Gussenhoven, M. S., D. A. Hardy, N. Heinemann, and E. Holeman (1982), 1978 diffuse auroral boundaries and a derived Auroral Boundary Index, Rep. AFGL-TR , Air Force Geophys. Lab., Hanscom Air Force Base, Mass. Hardy, D. A., H. C. Yeh, L. K. Schmitt, T. L. Schumaker, M. S. Gussenhoven, A. Huber, F. J. Marshall, and J. Pantazis (1984), Precipitating ion and electron detectors SSJ/4) for the block 5D/flights 6 10 DMSP satellites, Rep. AFGL-TR , Air Force Geophys. Lab., Hanscom Air Force Base, Mass. Hardy, D. A., M. S. Gussenhoven, and E. Holeman (1985), A statistical model of auroral electron precipitation, J. Geophys. Res., 90(A5), Hardy, D. A., M. S. Gussenhoven, R. Raistrick, and W. J. McNeil (1987), Statistical and functional representations of the pattern of auroral energy flux, number flux, and conductivity, J. Geophys. Res., 92(A11), 12,275 12,294, doi: /ja092ia11p of8

8 Madden, D., and M. S. Gussenhoven (1990), Auroral Boundary Index from 1983 to 1990, Tech. Rep., GL-TR , Air Force Geophys. Lab., Hanscom Air Force Base, Mass. Schumaker, T. L., D. A. Hardy, S. Moran, A. Huber, J. McGarity, and J. Pantazis (1988), Precipitating ion and electron detectors SSJ/4 for the block 5D/flight 8 DMSP satellite, Rep. AFGL-TR , Air Force Geophys. Lab., Hanscom Air Force Base, Mass. Spogli, L., L. Alfonsi, G. De Franceschi, V. Romano, M. H. O. Aquino, and A. Dodson (2009), Climatology of GPS ionospheric scintillations over high and mid-latitude European regions, Ann. Geophys., 27, , doi: /angeo Weaver, M., W. Murtagh, C. Balch, D. Biesecker, L. Combs, M. Crown, K. Doggett, J. Kunches, H. Singer, and D. Zezula (2004), Halloween space weather storms of 2003, NOAA Tech. Memo., OAR SEC-88, Space Environ. Cent., Boulder, Colo. Whalen, J. A., R. R. O Neil, and R. H. Picard (1985), Aurora, in Handbook of Geophysics and the Space Environment, edited by A. S. Jursa, chapter 12, pp , Air Force Res. Lab., Hanscom Air Force Base, Mass. 8of8