Volume 41, Issue 2
Free Access

Modeling polar ionospheric effects during the October–November 2003 solar proton events

Mark A. Clilverd

Mark A. Clilverd

Physical Sciences Division, British Antarctic Survey, Natural Environment Research Council, Cambridge, UK

Search for more papers by this author
Annika Seppälä

Annika Seppälä

Finnish Meteorological Institute, Helsinki, Finland

Search for more papers by this author
Craig J. Rodger

Craig J. Rodger

Physics Department, University of Otago, Dunedin, New Zealand

Search for more papers by this author
Neil R. Thomson

Neil R. Thomson

Physics Department, University of Otago, Dunedin, New Zealand

Search for more papers by this author
Pekka T. Verronen

Pekka T. Verronen

Finnish Meteorological Institute, Helsinki, Finland

Search for more papers by this author
Esa Turunen

Esa Turunen

Sodankylä Geophysical Observatory, University of Oulu, Sodankylä, Finland

Search for more papers by this author
Thomas Ulich

Thomas Ulich

Sodankylä Geophysical Observatory, University of Oulu, Sodankylä, Finland

Search for more papers by this author
János Lichtenberger

János Lichtenberger

Space Research Group, Eötvös University, Budapest, Hungary

Search for more papers by this author
Péter Steinbach

Péter Steinbach

Research Group for Geoinformatics and Space Sciences, Geological Research Group of the Hungarian Academy of Sciences, Budapest, Hungary

Search for more papers by this author
First published: 18 March 2006
Citations: 31

Abstract

[1] At Ny Ålesund, Svalbard (78°54′N, 11°53′E, L ∼ 18), a narrowband VLF receiver was used to monitor the behavior of the amplitude of several high-power transmitters located in the Northern Hemisphere under the influence of the solar proton events (SPE) of October/November 2003. We have used Sodankylä ion chemistry (SIC) atmospheric model profiles calculated at the midpoint location of the propagation paths in the northern wintertime polar region to investigate the radio propagation properties of several high-latitude paths. Different paths showed different responses to the proton precipitation, but propagation modeling was broadly able to account for most of the positive and negative responses observed. Using the SIC-based electron density profiles, we have been able to develop models of ionospheric effective height (h′) and sharpness (β) in order to describe the D region behavior as a function of proton flux, extending previous work which reported β and h′ values as functions of the X-ray flux from solar flares. As a result of these models, our understanding of VLF propagation influenced by SPEs is such that VLF observations might be used to predict changes in the ionospheric D region electron density profiles during other particle precipitation events.

1. Introduction

[2] Very low frequency (VLF) radio signals, or long-wave radio signals, are known to be severely affected during solar proton events (SPEs) [Potemra et al., 1967; Westerlund et al., 1969] as they propagate in the waveguide formed by the surface of the Earth and the lower boundary of the reflective ionosphere at around 50–90 km altitude. The Naval Ocean Systems Center (NOSC) have developed a computer program Long Wave Propagation Code (LWPC) to model VLF signal propagation from any point on Earth to any other point. The code models the variation of geophysical parameters along the path as a series of horizontally homogeneous segments. To do this, the program determines the ground conductivity, dielectric constant, orientation of the geomagnetic field with respect to the path and the solar zenith angle, at small fixed-distance intervals along the path. Given electron density profile parameters for the upper boundary conditions, LWPC calculates the expected amplitude and phase of the VLF signal at the reception point [Ferguson and Snyder, 1990]. Thus it can be used to investigate the modification of the ionosphere during SPEs, characterizing the electron density profile produced by the precipitating particle fluxes [Clilverd et al., 2005].

[3] During the series of SPEs of October/November 2003, high-energy proton precipitation penetrated deep into the Earth's polar atmosphere. The Sodankylä ion chemistry model (SIC) [see Verronen et al., 2002] models the major chemical reactions and interactions known to occur in the lower ionosphere. It has been used to calculate the effects of a series of SPEs that took place in October/November 2003. Model runs were produced at 70°N, 0°E (L = 7.5) to compare calculated ozone concentrations with measurements made by the instrument Global Ozone Monitoring by Occultation of Stars (GOMOS) in the northern latitude range 65°–75° on board the European Space Agency's ENVISAT satellite. GOMOS measured chemical constituents such as ozone at altitudes 10–100 km over the northern polar region during the SPEs.

[4] Here we extend the radio propagation analysis of the October/November 2003 storms made by Verronen et al. [2005]. In that study we used a technique described by Clilverd et al. [2005] in which the electron density profiles output from the SIC atmospheric model were used to model the southern polar radio propagation conditions, the results of which were compared in detail with a propagation path that passed close to the SIC modeling location. In this study we extend the comparison to several different propagation paths in the Northern Hemisphere using the SIC electron density profiles as an estimate of the average conditions on the path. In this way we test the validity of applying the SIC model results over the whole polar region. We focus on two separate areas of study. The first is where the propagation path is located within the polar region and the midpoint of the path is at a similar latitude to the atmospheric model calculation point. The second is when only some of the path is located within the polar region, and the midpoint of the path is at lower latitude than the modeling location.

2. Experimental Configuration and Event Conditions

[5] At Ny Ålesund, Svalbard, (78°54′N, 11°53′E, L ∼ 18) a narrowband VLF Omnipal receiver was used to monitor the signals from several high-power transmitters located in the Northern Hemisphere [Dowden et al., 1994; Clilverd et al., 2001]. We make use of the received amplitude of VLF signals to characterize the response of the upper atmosphere to SPE produced ionization changes. This receiver also logs the phase of the received transmissions. Previous studies have shown that variations in the received phase can be used to describe SPE-driven atmospheric modifications [e.g., Potemra et al., 1967], while the received amplitude often shows a more complex response depending on the length of the path and the transmission frequency [e.g., Westerlund et al., 1969; Cummer et al., 1997]. However, the drift of the frequency standard crystal at this location makes long-timescale phase comparisons essentially impossible. While Omnipals have been used to investigate short-timescale ionization changes (∼100 s) using received amplitude and phase [e.g., Clilverd et al., 2002], long-timescale comparisons require locking levels currently unavailable for this instrument at this location. Upgraded versions of the OmniPAL have been locked to GPS, creating the AbsPAL, which has proved successful for studies undertaken over hours to days [e.g., Thomson et al., 2004].

[6] The great circle paths (GCP) taken by the signals to reach the receiver at Ny Ålesund go through the polar zone as shown in Figure 1 as indicated by the region poleward of the L = 4 contour. Each path is expected to be significantly influenced during SPEs. Details of the location and frequency of the transmitters shown are given in Table 1. The SIC calculation position is marked with an asterisk in Figure 1. We outline the model details and setup for SPE analysis in the following section.

Details are in the caption following the image
Map of great circle paths taken by VLF signals to reach Ny Ålesund and Erd receivers. The L = 4 contour is shown by the dashed line, which defines the approximate northern polar region for SPE effects. The SIC modeling point (70°N, 0°E) is indicated by the asterisk.
Table 1. Details of the Propagation Paths Studied, Including the Transmitter Frequency, Transmitter to Receiver Path Length, and Local Midday SZA for the Midpoint of Each Path
Transmitter Call Sign Location Frequency, kHz Path Length, km Midpoint Coordinates SZA (1200 LT)
Latitude Longitude Latitude Longitude
Polar Paths Received at Ny Ålesund
JXN 66.4°N 13.1°E 16.4 1390 72.7°N 12.7°E 85.8°
NRK 64.2°N 21.9°W 37.5 1970 72.2°N 11.7°W 85.3°
NDK 46.4°N 98.3°W 25.2 5370 69.2°N 82.1°W 82.3°
NLK 48.2°N 121.9°W 24.8 5560 72.3°N 107.3°W 85.4°
SIC Modeling Location
SIC 70.0°N 0.0°E 83.0°
Partial Polar Path Received at Ny Ålesund
NPM 21.4°N 158.2°W 21.4 8840 61.1°N 155.5°W 74.3°
Partial Polar Path Received at Erd
NRK 64.2°N 21.9°W 37.5 3070 57.4°N 3.2°E 70.4°

[7] There are several notable differences between the illustrated propagation paths, particularly when they are under the influence of proton precipitation. Norway (call sign JXN, transmit frequency 16.4 kHz) and Iceland (NRK, 37.5 kHz) to Ny Ålesund are short paths completely within the polar region but with very different frequencies. Dakota (NDK, 25.2 kHz) and Seattle (NLK, 24.8 kHz) to Ny Ålesund have very similar path lengths, both almost entirely within the polar region, both very similar frequencies, but NDK passes over the Greenland ice cap and NLK does not. When a VLF propagation path crosses low-conductivity ground such as the Antarctic or Greenland ice caps the effect of the SPE is enhanced as a result of increased mode conversion to higher-order modes at the conductivity boundary which produces higher attenuation with distance [Westerlund et al., 1969]. Hawaii (NPM, 21.4 kHz) to Ny Ålesund is a very long path, with only the last portion of the path in the polar region. In this context the experimental setup is similar to the NPM-Halley study reported by Clilverd et al. [2005], except that no low-conductivity ice cap is encountered in the NPM to Ny Ålesund GCP. Finally the partial polar GCP from Iceland (NRK) to Erd near Budapest (47.4°N, 18.9°E, L = 1.9) is investigated to determine the ability of the modeling technique described by Clilverd et al. [2005] to reproduce the effect of the SPEs on this type of path. A detailed list of transmitters and receivers is given in Table 1, separated into polar paths and partial polar paths. The solar zenith angles (SZA) of the midpoints of each of the paths are given in order to compare the average path conditions with the SIC modeling location conditions. The aim is thus to use the SIC electron density profiles to model the average conditions on each of the paths. This should be appropriate given the uniform nature of the ionization generated by proton precipitation over the polar region. Limitations to this assumption are only likely to occur during the first few hours of an SPE when the precipitation is asymmetric around the geomagnetic pole [Krimgis and Van Allen, 1967].

[8] The solar proton events of October/November 2003 occurred on 26 and 28 October and 2 November. The top panel of Figure 2 shows the variation of the >50 MeV proton flux during the study period. Vertical dashed lines represent the start times of the three SPEs. The SPE period began on 26 October with an X1 class flare, followed by several large flare events throughout the period, culminating in the massive X45 flare on 4 November 2003 [Thomson et al., 2004, 2005]. The peak flux units for >10 MeV proton particles (pfu) were 466, 29500, and 1570 cm−2 sr−1 s−1 respectively. Prior to 26 October (day 299) the proton flux levels were at nondisturbed levels ∼0.1 cm−2 sr−1 s−1 for >10 MeV proton particles. After the SPE of 2 November (day 306) the proton flux levels slowly recovered to quiet levels, dropping below the 10 cm−2 sr−1 s−1 threshold at the beginning of 7 November (day 311) for the >10 MeV particles, equivalent to 0.5 cm−2 sr−1 s−1 at >50 MeV proton energies shown by Clilverd et al. [2005] to be associated with radio propagation effects.

Details are in the caption following the image
Amplitude change from their respective quiet day curves of signals received at Ny Ålesund during 20 days in October/November 2003 (22 October to 10 November 2003). The top panel shows the >50 MeV proton fluxes during this period. The remaining panels are plotted in longitude order, starting with the most easterly path and working west. The times of the onset of solar proton events are indicated by the vertical dashed lines.

3. Sodankylä Ion Chemistry (SIC) Model

[9] The Sodankylä ion chemistry (SIC) model is a one-dimensional chemical model designed for ionospheric D region studies, solving the concentrations of 63 ions, including 27 negative ions, and 13 neutral species between 20–150 km altitude at the time of this study in November 2004. An overview of the model was given by Clilverd et al. [2005], but we summarize in a similar way here to provide background for this study.

[10] In the SIC model several hundred reactions are implemented, plus additional external forcing due to solar radiation (1–422.5 nm), electron and proton precipitation, and galactic cosmic radiation. Initial descriptions of the model are provided by Turunen et al. [1996], with neutral species modifications described by Verronen et al. [2002]. Solar flux is calculated with the SOLAR2000 model (version 2.23) [Tobiska et al., 2000]. The scattered component of Lyman α solar flux is included using the empirical approximation given by Thomas and Bowman [1986]. The latest extension of SIC is the vertical transport code which takes into account molecular [Banks and Kockarts, 1973] and eddy diffusion [Chabrillat et al., 2002]. The background neutral atmosphere is calculated using the MSISE-90 model [Hedin, 1991]. Transport and chemistry are advanced in intervals of 5 or 15 min. However, within each interval exponentially increasing time steps are used because of the wide range of chemical time constants of the modeled species. Here we use the SIC model to investigate the electron density profile in the wintertime D region generated at 70°N, 0°E by the large solar proton events of October/November 2003. In all these solar proton events the proton spectra at the top of the atmosphere are assumed to be the same as those measured by GOES 11 at geosynchronous altitude. This assumption is valid for high magnetic latitudes. The angular distribution of the protons is assumed to be isotropic over the upper atmosphere, which is valid close to the Earth [Hargreaves, 1992]. The effects of the proton precipitation are contrasted against the same modeling runs made without any proton forcing, known as the control run.

4. Solar Proton Event Effects on Polar Paths

[11] All four mainly polar paths studied in this paper (JXN, NRK, NDK, and NLK to Ny Ålesund; see Table 1) are analyzed for SPE effects. In order to determine the changes due to the proton precipitation it is necessary to remove the experimentally observed quiet day curve (QDC) that is, the normal variability in VLF propagation along these paths. The QDCs were made by averaging several days together from before and after the SPE period. For each transmitter care was taken to use days that were geomagnetically quiet and which showed consistency in behavior from day to day; the QDCs will be discussed in more detail in the following section. Figure 2 shows the results of removing the QDC from the amplitude data of each transmitter during the study period. The amplitude data is shown with 1 min time resolution. Periods where a given transmitter is known to be off air for maintenance are omitted. Vertical dashed lines indicate the start times of the SPEs, while the horizontal dash-dotted line represents the QDC level (i.e., 0 dB). The panels are plotted in longitudinal path order starting from the most easterly (Norway to Ny Ålesund) and moving westward.

[12] Figure 2 shows several significant features. The Norway signals show virtually no offset due to absorption during the SPEs, although some increase in variability over the day can be seen. Toward the end of day 306 the transmitter goes into an ON/OFF transmission mode with only an hour of continuous transmissions at a time. The Iceland signal shows a significant absorption event only during the largest SPE (28 October, day 301). However, significant deviations from the QDC are seen, particularly during the “day” and this appears to increase after the last SPE with no recovery to QDC conditions observable by the end of the plot. The Dakota and Seattle signals undergo sudden absorption events of −10 to −30 dB at the start of each SPE. Initially the amount of absorption is linked to the severity of the SPE, with the first SPE being the smallest, and the second SPE being the largest. Throughout the data sets, mainly on Dakota and Seattle, significant noise spikes can be observed, often occurring once a day, and particularly during the periods of SPE absorption. These are likely to be interference effects from a Russian 25.0 kHz timing transmitter which have had less absorption than the transmitter being studied because of its shorter polar path. These spike events are omitted where possible in the following plots.

[13] As can be seen Figure 2 and the above discussion, we have several SPE effects that should be reproducible by a realistic atmospheric model under the influence of polar proton precipitation. The effects range from large, but different levels of absorption, increased diurnal variability, and in some cases little or no effect at all. In order to use VLF subionospheric propagation to detect and describe SPE effects on the middle atmosphere, significant understanding of all of these effects is required.

5. Ionospheric Parameters From SIC Modeling

[14] From 0000 UT on 26 October 2003 to 0000 UT 7 November 2003, the SIC model was run to investigate atmospheric changes in the northern polar region caused by the SPEs that occurred during this period. The model was used to generate hourly electron density profile curves as a function of altitude. In addition the SIC model run was repeated without any proton precipitation to create SIC-based QDCs from control run electron density profiles, that is, in the same way as Verronen et al. [2005]. In order to be able to apply the results to LWPC the SIC curves were approximated using equation (1) below, characterized by two parameters β and h′. This was done partly in order to be easily fed into the propagation code LWPC, but primarily because by parameterizing the electron density profiles in this way we are able to investigate the ionospheric changes more quantitatively. This type of parameterization extends the analysis of Clilverd et al. [2005] for SPEs, in a very similar way to that presented by Thomson et al. [2005] for solar flares. This parameterization technique has also been applied to the results presented by Verronen et al. [2005]. By using the SIC model to determine the values of β and h′ we effectively negate the issue of uniqueness that can occur when associating an observed change in amplitude with a single ionospheric profile. Figure 3 shows SIC electron density profiles with their respective β and h′ fits for three profiles taken from 1000 UT and 2200 UT on 28 October 2003 and 1000 UT on 29 October 2003. These times are prior to, and during, the largest solar proton event which started at 1215 UT on 28 October 2003. The fitted electron density profiles are calculated from
equation image
This is a standard formulation [see, e.g., Wait and Spies, 1964], and describes the electron density, in el cm−3, of the lower ionosphere through h′ (km), which represents the modified effective ‘reflective’ height of the ionosphere, and β (km−1), for the ‘steepness’ of the profile. The determination of the parameters was done by least squares fitting equation (1) to the SIC profiles below 3000 el cm−3, as at larger number densities VLF signals will not be able to propagate to higher altitudes when at a grazing incidence.
Details are in the caption following the image
Electron density profiles. The three curves represent SIC model-predicted values for 28 October 2003 (1000 and 2200 UT) and 29 October 2003 (1000 UT) prior to and during the largest SPE. The three straight lines represent an exponential profile approximation to the SIC model results.

[15] The results of parameterizing the SIC electron density profiles are shown in Figure 4. The plot shows how the lower part of the D region ionosphere changes in altitude and “sharpness” with time under the influence of proton precipitation during the SPEs. The hourly SIC-calculated quiet day curves (QDCs) and SPE-driven changes are represented by the dotted lines, while our attempts to empirically link h′ and β to proton fluxes are represented by solid lines in their respective plots. The formulation of the empirical models will be discussed in the following subsection. During the prestorm days there is a repeatable variation of h′ and β with values oscillating about 80 km and 0.35 km−1 respectively. During SPE days the SPE-driven behavior shown is imposed on top of the normal QDC behavior. Thus, when SPEs occur the altitude of the ionosphere is lowered, reaching about 55 km at the lowest point during our study days. At the same time the “sharpness” of the ionosphere increases in variability, sometimes reaching maximum values of 0.45–0.50 km−1, and sometimes dropping to minimum values of 0.20 km−1. These changes in the ionospheric parameters, β and h′, will influence radio propagation conditions and would be expected to be reflected in the data shown in Figure 2.

Details are in the caption following the image
Variation relative to QDC of the β and h′ (h-prime) parameters during 17 days in October/November 2003 (25 October to 10 November 2003) determined from the electron density profiles output from the SIC model at 70°N, 0°E. Both the QDC and the SPE-driven effects determined from SIC are plotted (dotted lines). The adopted h′ and β empirical models are shown as solid lines.

5.1. Relationship of β and h′ With Particle Fluxes

[16] Previously Clilverd et al. [2005] showed that the high-latitude southern ionosphere behavior under the influence of >50 MeV proton precipitation (f50) could be described by h′ and modeled as
equation image
where 71 km represented the altitude of the summer daylight ionosphere during nondisturbed times. In this study the winter ionosphere shows a significant QDC and thus we rewrite equation (2) as
equation image
In Figure 4 the solid line represents this model and a good fit to the data can clearly be seen, although small discrepancies occur at times during the highest flux conditions. Generally, as expected, the reflection altitude is lowered in response to increased proton fluxes.
[17] Clilverd et al. [2005] provided no model of the β behavior during a SPE, simply noting that it was a topic that should be investigated further. Here we present a model based on β changing both with solar zenith angle (SZA) and with >50MeV proton flux levels (f50). The model is formulated to reproduce the SIC electron density profile time-dependent behavior, following the suggestion for h′ used by Westerlund et al. [1969], and is shown as the solid line in the bottom plot of Figure 4:
equation image
The SPE-driven β behavior shown in Figure 4 is considerably more complex than the h′ behavior. Typically, increased proton fluxes increase the variability in β, with values occurring that are larger and smaller than the typical nondisturbed value. There is significantly less effect of the proton fluxes on β when SZA is >92° (i.e., during the nighttime) as the electrons produced by the precipitation are rapidly converted into negative ions. During the times when the Sun is above the horizon at >50 km altitudes the electrons are released through the action of solar photoionization and the ionosphere becomes increasingly “sharp” as proton fluxes increase. When proton fluxes become small and near normal at the end of our study period β continues to show a small diurnal variation with largest values when the Sun is highest above the horizon.

[18] When the SIC-calculated β behavior was modeled using equation (4) the third SPE in the series (starting on day 306) was not as well described as in the previous two SPEs. To improve the fit the sensitivity of β was increased from day 306 by changing the amplitude in equation (4) from 0.08 to 0.11, that is, an increase of 37.5%. This is included in Figure 4. However, despite a significant improvement in the fit to the SIC-calculated β values on days 307 and 309, the β behavior on day 308 is not well described by the model. Because this behavior is seen in the SIC-based calculations, which was not run with any energetic electron input parameters, the most likely cause of the poor modeling fit is due to changes in the energy spectra of the proton fluxes. Day 308 was the day with the softest proton spectra (ratio of >100MeV to >1MeV protons) during the period studied here. A softer proton spectrum would tend to increase the electron density at higher altitudes, decreasing it at lower altitudes, thus increasing β as observed. The effects of changing the energy spectra of the proton fluxes on the β behavior during SPEs will be studied in detail in a future paper.

5.2. Application of Ionospheric Profiles to Propagation Modeling

[19] To apply the SIC ionospheric profile parameters β and h′ to the LWPC runs we make the assumption that the whole of each of the polar paths investigated can be represented by a set of ionospheric profiles determined by SIC at the midpoint of the path. Given the influence of SZA on β this assumption would not necessarily be valid for midlatitude and low-latitude sections of the propagation paths, where the Sun would be significantly higher in the sky at midday than for the SIC modeling location (see Table 1), and there would be no influence from precipitating protons. However, for the four mainly polar paths to Ny Ålesund it is likely to be a reasonable first approximation, particularly when the particle precipitation zone is expanded as a result of the events under study here. Figure 5 shows the quiet day curves of each of the transmitters received at Ny Ålesund. The LWPC model results using β and h′ parameters determined from the SIC QDC calculations are shown by the asterisks. Local time differences between the propagation paths and the SIC modeling location have been accounted for by shifting the time axis appropriately, that is, assuming that the only difference between the SIC results and the propagation paths is due to local time.

Details are in the caption following the image
Quiet day curves of each of the transmitters received at Ny Ålesund. The LWPC model results using the SIC β and h′ parameters are shown by the asterisks. Local time differences between the propagation paths and the SIC modeling point have been accounted for by shifting the time axis appropriately.

[20] For the two shorter paths (JXN and NRK) the modeled QDC is generally one of constant nighttime amplitudes and lower daytime amplitudes (∼4–5 dB). This behavior is observed in the data, but with more structure, particularly during the daytime. The onset time of daytime conditions is well represented in the modeling for both paths once the SIC parameters have been phase shifted by +1 hour for Norway (JXN) and −1 hour for Iceland (NRK).

[21] The two medium length paths (NDK and NLK) show larger diurnal variation than the shorter paths. Both exhibit ∼8 dB ranges from day to night. The Dakota (NDK) to Ny Ålesund path is typically phase shifted by −5 or −6 hours in local time compared with the SIC modeling point and thus the daytime decrease in amplitude occurs around 1700 or 1800 UT. The propagation model runs show a daytime decrease in amplitude at the same time as the observed QDC, but only ∼2 dB. The nighttime variation (2200–1200 UT) is reasonably modeled. The Seattle (NLK) to Ny Ålesund path is typically phase shifted by −7 or −8 hours in local time and thus the daytime decrease in amplitude occurs around 1900 or 2000 UT. The observed QDC shows little effect of daytime, but a much larger nighttime decrease around midnight on the path (0700–0800 UT). This behavior is not well represented by the propagation modeling either in the nighttime decrease in amplitude or the lack of a daytime amplitude decrease.

[22] Clearly, the application of the SIC-based ionospheric profiles to the propagation modeling is reproducing some of the most significant QDC features in 2 out of 4 of the paths studied in this section. This is mainly based on reasonable modeling of the onset of daytime propagation conditions on the shortest paths, and the diurnal amplitude variation compared with the observations. The longer paths (Dakota and Seattle) are less well modeled, particularly poorly in the daytime, although reasonably well during the nighttime. Better modeling of the QDCs could be obtained by including background levels of particle precipitation; this is discussed in section 7.

5.3. Modeling the Largest SPE

[23] In Figure 2 it was apparent that a range of behavior occurred during the SPEs. The SIC-calculated ionospheric profiles show that β and h′ change significantly as the proton fluxes change during the SPEs (Figure 4), but as shown in the previous section the nondisturbed β and h′ behavior is able to broadly model the quiet time propagation conditions for mainly polar paths (Figure 5). In Figure 6 the effect of the largest SPE which started on 28 October 2003 (day 301), is shown in a similar format to that described in Figure 2. The propagation modeling is also included for each path (diamonds) applying only the SIC β and h′ variations for the period shown (from Figure 4) to the whole propagation path. The start of the SPE is indicated by the vertical dashed line.

Details are in the caption following the image
Comparison of the effect of the largest SPE (28 October 2003) of the series and the results from the propagation modeling runs using the SIC β and h′ parameters (diamonds). Periods where the transmitters went off air (usually 8 hours or so), or when there was some interference from competing transmissions (usually 10 min at a time), have been removed from the plot.

[24] The shortest paths (Norway and Iceland to Ny Ålesund) show SPE effects, although there is little absorption observed, either in the data or predicted by the model. In the case of Norway no absorption is seen, only an increase in daytime amplitudes to higher than QDC levels. The Iceland path exhibits about 6 dB of absorption after the SPE has started, but apparently delayed by several hours. This is a result of the amplitudes being higher than normal QDC levels before the start of the SPE and consequently a certain time is taken for the amplitudes to fall below QDC levels. During the SPE both Iceland and Norway show higher than normal QDC amplitudes during the daytime. Again, this is well reproduced by the propagation modeling through enhanced β variability. Figure 2 shows that this behavior continues after the proton fluxes have returned to normal values (about day 312, or 8 November). This suggests that both of these propagation paths are sensitive not only to proton precipitation but energetic electron precipitation that occasionally follows SPEs. The lack of absorption shown by the Norway to Ny Ålesund path is indicative of a relative insensitivity to the changes in h′ during the SPE.

[25] For the Seattle to Ny Ålesund path the start of the SPE coincides with the onset of significant absorption levels. This would probably also be true for the Dakota to Ny Ålesund path, except that the transmitter went off air 15 min before the SPE started and only came back on again about 8 hours later. However, in both cases the propagation modeling shows significant absorption is predicted from the SPE onset. The Dakota path experiences about 32 dB of absorption at maximum (0000 UT on day 302) with values slowly returning to normal by midday 304 (31 October 2003). The Seattle path only experiences about 20 dB of absorption. The additional absorption in the case of the Dakota to Ny Ålesund path is a result of the propagation of the signal across the low-conductivity ground of the Greenland ice cap [Westerlund et al., 1969] making the path more sensitive to the upper boundary conditions set by the proton precipitation. In both cases the absolute levels of absorption are well modeled.

[26] Both of the longer paths show oscillations in amplitude as the absorption levels slowly recover from the maximum SPE effect. In both cases the oscillations are large enough to reduce the SPE effect to near-zero levels at times. The propagation modeling agrees with these observations closely. The timing of the oscillations is consistent with the changes in β exhibited by the SIC model and at times both paths are close to normal QDC levels. Daytime conditions through enhanced β variability produce less absorption on the Dakota path, and more absorption on the Seattle path, as we expected.

6. Partial Polar Paths

[27] Two further paths are considered here. The paths either do not start or do not end in the polar region, and unlike the paths discussed above, have a significant proportion located at low latitudes and midlatitudes. SPE particles cannot access the entire global atmosphere as they are partially guided by the geomagnetic field, and are limited to the polar regions. Such paths with a significant proportion located outside of the polar regions must be treated as being inhomogeneous. In these cases the SIC ionosphere profiles can be applied only to the polar section of the propagation model as discussed by Clilverd et al. [2005]. The paths are Hawaii (NPM) to Ny Ålesund, and Iceland (NRK) to Erd near Budapest. Additional details of the two paths are given in Table 1. Hawaii to Ny Ålesund is similar in some respects to the Hawaii to Halley, Antarctica, path studied by Clilverd et al. [2005] in that they are mainly all sea, long (of the order of 10,000 km), and the same frequency (21.4 kHz). Where these equator-pole paths differ is that the NPM to Ny Ålesund path does not cross over any low-conductivity ice such as the Antarctic ice cap.

[28] The Iceland (NRK) to Erd path is relatively short (∼3,000 km) and mostly midlatitude. However, as the Iceland transmitter lies poleward of the L = 4 contour the first ∼500 km would be expected to be influenced by proton precipitation during SPEs, and this short section would be expected to be represented by the SIC ionospheric parameters during these times. In both these cases the nonpolar portion of the paths would be modeled by LWPC-only ionospheric profiles such as those in LWPC itself or the Thomson daytime model [McRae and Thomson, 2000].

[29] In Figure 7 we show the combined Figures 2, 5, and 6 equivalent plots for the long path Hawaii to Ny Ålesund and the short path Iceland to Erd. In the top plot only small (∼5 dB) deviations can be observed on the Hawaii signals at the start of the SPEs, which is about the same level as the variability from the QDC seen throughout the period studied. In the middle plot the Hawaii to Ny Ålesund QDC (solid line) shows little diurnal variation apart from at about 1700 UT where a sunrise modal minimum occurs. Following the technique used for the Hawaii to Halley path studied by Clilverd et al. [2005] we replaced the polar sections of the ionospheric profile model in LWPC with the SIC-based phase-shifted QDC profiles. The results from this are shown as asterisks. This time the sunrise modal minimum agrees but the model predicts slightly lower amplitudes during daytime on the path (1800–0400 UT). The nighttime model values agree with the observations.

Details are in the caption following the image
(top) Amplitude change from the quiet day curve of the Hawaii to Ny Ålesund signal and Iceland to Erd (near Budapest) signal during October/November 2003. The times of the onset of solar proton events are indicated by the vertical dash-dotted lines. (middle) Comparisons of quiet day curve (solid line) and the LWPC/SIC ionosphere model results (asterisks). (bottom) Comparisons of the effect of the largest SPE (28 October 2003) in the series with the results from the propagation modeling runs using the SIC β and h′ parameters (diamonds) on part of the paths only.

[30] The bottom plot shows that the propagation model (diamonds) confirms the observations (solid line) in that very little absorption occurs even during the largest SPE on this type of long path, particularly where no low-conductivity ground is included. This result is in stark contrast to the NPM-Halley path, which includes Antarctic ice shelf ground conductivity and shows significant absorption during SPEs [Clilverd et al., 2005].

[31] For the Iceland to Erd path the top plot has been extended to days 293–315 because of some data gaps between day 295 and 300. Although there is limited coverage for the two smaller SPEs, good coverage occurs during the largest SPE (days 301–304) and a significant increase in amplitude deviations from the QDC can be seen, both positive and negative. In the middle plot the Iceland to Erd QDC (solid line, with data gaps) shows large differences in nighttime amplitudes compared with the daytime (∼15 dB). A good match to the observed QDC occurs when SIC-based pre-SPE β and h′ values are applied to the first 500 km of the paths (asterisks), rather than the no proton forcing control run QDC (not shown). This suggests that background levels of proton ‘drizzle’ are important in defining the QDC on this path.

[32] In the bottom plot the Iceland to Erd observations during the large SPE show a delayed onset of absorption, with levels reaching −8 dB during the night. The recovery takes several days, although as with the other short paths, large positive amplitudes are observed during the daytime. The propagation model shows some agreement with this, particularly during the recovery phase of the SPE, despite the SIC profiles being only applied to the first 500 km of the path, close to the transmitter. Significant deviations of the modeling results from the observations can be seen during the first 24 hours of the event. This may be due to >500 km of the path being influenced by precipitation during some of this period because of geomagnetic disturbances [Smart and Shea, 2003], and asymmetry in particle precipitation around the geomagnetic pole at the start of the event [Krimgis and Van Allen, 1967].

7. Discussion

[33] The SIC control run electron density profiles for nondisturbed conditions are reasonably successful in modeling the quiet day curves of all four polar paths. Effectively the principal area of mismatch between the model QDCs and the observed QDCs is during the daytime. This is due to background levels of proton precipitation, which have not been included in the SIC model QDC or “control” run. The modeling of the Iceland to Erd QDC clearly suggests that the background levels of proton precipitation are a significant factor that should be taken into account and not set to zero in model background calculations for polar regions.

[34] Figure 8 shows the effect that differing levels of background proton precipitation (in cm−2 sr−1 s−1 at >50 MeV proton energies) have on the resulting SIC-based β and h′ during non-SPE conditions. The results are for (1) no flux (solid line), (2) daily average ∼0.05 cm−2 sr−1 s−1 which were the conditions just prior to the SPE on 26 October 2003 (dashed line), and (3) daily average ∼0.13 cm−2 sr−1 s−1 which were the conditions just prior to the SPE on 28 October 2003 (dotted line). As the background flux levels increase, h′ decreases by several km and β generally increases by 0.02–0.03 km−1, particularly at night. The effect of these changes are shown in Figure 9 where the Dakota (NDK) QDC is modeled using SIC-based β and h′ values determined under the influence of each of the background proton flux levels. The calculations show that as the background proton flux increases the modeled daytime amplitude levels are significantly changed. With no flux at all, the model output (solid line) is much less representative of the observed QDC for NDK (diamonds). The largest background flux levels cause the daytime amplitudes to significantly decrease (dotted line) and reasonably describe the observed QDC. In all cases the nighttime levels remain essentially unchanged. Not all paths respond in such a dramatic way to these relatively subtle changes, and because of this we have not recalculated the SIC control run profiles to include background proton “drizzle.” In this way we maintain consistency with the calculations made by Verronen et al. [2005].

Details are in the caption following the image
Effect on β and h′ during non-SPE conditions of various levels of background proton precipitation. The units are cm−2 sr−1 s−1 for >50 MeV fluxes. The solid line represents the non–proton forcing, or control, conditions.
Details are in the caption following the image
Effect on the Dakota (NDK) QDC of various levels of background proton precipitation. The units are cm−2 sr−1 s−1 for >50 MeV fluxes. The observed QDC is represented by diamonds, while LWPC/SIC calculations are given by solid, dashed, and dotted lines.

[35] As a result of these discussions we can say that even the background levels of proton precipitation into the polar region do influence the propagation conditions noticeably for some of the paths studied. In order to understand their behavior during SPEs it is important to get the prestorm QDCs right. This applies to setting up the SIC model runs. For calculations made at high latitudes in winter months the proton “drizzle” could maintain low, but continuous, levels of production of NOx in the mesosphere, potentially leading to higher levels of NOx [see Randall et al., 2005, and references therein].

8. Summary

[36] We have used the SIC atmospheric model profiles calculated at a single location in the northern wintertime polar region to investigate the radio propagation properties of four high-latitude paths. During the SPE series of October/November 2003 the different paths showed different amplitude behavior in response to the proton precipitation. Longer paths, mostly within the polar region, tend to show large absorption events with oscillations in amplitude during the daytime. Shorter polar paths showed less tendency to large absorption events, but similar oscillatory behavior in daytime amplitude during the recovery phase of the SPEs. Additional paths that were only partially inside the polar region showed no significant absorption effects when the transmitter was located outside, but significant changes when the transmitter was located inside the polar region. These results show that to monitor energetic precipitation from ground-based VLF receivers it is important to carefully select the transmitter-receiver path length and orientation, as well as the frequency used.

[37] Propagation modeling using time-varying ionospheric electron density profiles from the SIC model was broadly able to account for all of these types of behavior, despite the profiles being calculated only for a location in the polar region that represents the average path conditions. The same SIC profiles were also broadly able to model the SPE response on partial polar paths by being applied to only part of the path, leaving normal ionospheric conditions on the remaining path length.

[38] Using the SIC-based electron density profiles we have been able to develop models of ionospheric effective height (h′) and sharpness (β) in order to describe the D region behavior in response to proton fluxes. This is an extension of the work of McRae and Thomson [2004] and Thomson et al. [2005] who reported β and h′ values as functions of the X-ray flux from solar flares. The ionospheric effective height (h′) is primarily only sensitive to proton flux levels during SPEs, with lower altitudes being driven by higher fluxes. The ionospheric sharpness (β) shows much more oscillatory behavior through a combination of proton flux levels and SZA. As a result, this paper has shown that our understanding of VLF propagation influenced by SPEs is such that VLF amplitude observations might be used to accurately predict changes in the ionospheric D region during other particle precipitation events.

Acknowledgments

[39] M.A.C. would like to acknowledge the contribution of Pat Espy and Nick Rose in collecting the Ny Ålesund data used in this work. This work was supported in part by the LAPBIAT program of the European Commission, project HRPI-CT-2001-00132. The work of P.T.V. was supported in part by the Academy of Finland through the ANTARES space research program.