Space Physics and Aeronomy, Ionosphere Dynamics and Applications. Группа авторов

Чтение книги онлайн.

СКАЧАТЬ the assumption of chemical equilibrium. As a result, the time step in GITM is 2–4 s. The model uses the same magnetospheric drivers as CTIPe and TIEGCM, viz. high‐latitude electric fields and energetic particle precipitation.

      As described in section 1.2 (energy entering the IT system), the primary coupling to magnetospheric energy input is via the empirical models of the high‐latitude electric field provided by the W05 or Cosgrove14 models, or an inversion of assimilated high‐latitude data from the AMIE model. More rarely, MHD models such as the Block‐Adaptive‐Tree‐Solarwind‐Roe‐Upwind‐Scheme (BATS‐R‐US) (Powell et. al., 1999), Open Geospace General Circulation Model (OpenGGCM) (Raeder et al., 1998), or Lyon‐Fedder‐Mobarry (LFM) model (Lyon et al., 2004) can be coupled directly to the IT models (CTIPe, TIEGCM, GITM, or others), which removes the need for separate high‐latitude electric field specification.

1.4 MODEL ASSESSMENT

      Rastaetter et al. (2016) carried out a comprehensive comparison of a number of empirical and physics‐based models for 6 events selected for the GEM‐CEDAR modeling challenges. The goal of the challenge was specification of Poynting flux. DMSP F15 observations were provided for validation purposes. The models used in the comparison included:

      1 Space Weather Modeling Framework (SWMF) (Toth et al., 2005, 2012), which couples BATS‐R‐US to the Rice Convection Model (Wolf et al., 1982; Toffoletto et al., 2003), which, in turn, is coupled to the Ridley Ionosphere model (Ridley et al., 2004);

      2 OpenGGCM coupled to the Coupled Thermosphere Ionosphere Model (CTIM) (Fuller‐Rowell et al., 1996), an earlier version of CTIPe (Two versions of OpenGGCM with different resolution were part of the study);

      3 Coupled Magnetosphere Ionosphere Thermosphere (CMIT) model (Wang et al., 2004; Wiltberger et al., 2004), which couples the LFM magnetosphere model to TIEGCM and the MIX ionosphere solver and coupler (Merkin & Lyon, 2010) (CMIT (LGM‐TIEGCM‐MIX) and LFM‐MIX were run as separate models);

      4 TIEGCM with W05 providing high‐latitude electric fields;

      5 CTIPe with W05 for high‐latitude electric fields;

      6 W05, three versions run with solar wind input provided by different models;

      7 Cosgrove14, two versions run with solar wind input provided by different models.

The models were divided into physics‐based (models 1–5 of the list above), or empirical (models 6 and 7 of the list above). Results of the comparison varied for the six selected events, which ranged from a relatively small storm event with the minimum in Dst of ‐40 nT, to the Halloween storm with Dst_min of ‐353 nT. Figures 1.7 and 1.8, reproduced from Figures 6 and 11 by Rastaetter et al. (2016), show the model versus data for the Halloween storm (29–30 October 2003) when Dst was ‐353 nT, and an interval from 9–12 July 2005 when Dst_min was ‐89 nT, respectively. We use these examples as illustrations of a strongly driven superstorm as the Halloween storm was, and an event of sustained low magnetic activity in July 2005 for contrast. In the figures, physics‐based models are at left (panels (a) through (d)), empirical models are at right (panels (e) through (g)). The Poynting flux or Joule heating integrated along the DMSP orbit track poleward of auroral latitudes are shown by different colored symbols in the upper panels. F15 Poynting flux observations are shown as solid black symbols connected with straight black lines. Vertical gray bars indicate the 25% uncertainty in the measurements. The comparison between integrated model and observed Poynting flux is shown in panels (a) and (e).

Figure 1.7 Summary of integrated values over auroral passes for event 1, on 29–30 October 2003: (a)–(d) Physics based and (e)–(h) empirical models. (a)–(c) Scores for physics‐based models: (a) Poynting flux or Joule heat integrated over full auroral passes (black symbols connected with black solid line are observations), ratio of model values to observed values constitute the Integrated value Yield (IYI). (b) Model Amplitude Yields YI (maximum Poynting flux or Joule heat divided by observed maximum Poynting flux) shown in base 2 logarithmic (ld) scaling for two pass segments (diamonds: evening side and crosses: morning side); in this scaling ld(1) = 0 is the perfect score. (c) Timing errors of maximum signal (time of model maximum minus time of observed maximum) for two segments of each pass of auroral region (symbols denote the same pass segments as Figure 1.6b). (d)Dst index; (e)–(g) Scores for empirical models: (e) Integrated Poynting flux or Joule heat, (f ) amplitude yields, (g) timing errors, and (h) AL index

      (figure and caption based on Fig. 6 by Rastaetter et al., 2016. Reproduced with permission of John Wiley & Sons).

Figure 1.8 Summary of integrated values over auroral passes for event 6, on 9–12 July 2005: (a)–(d) Physics‐based models and (e)–(h) empirical models with Dst and AL at (d) and (h). Results are in the same format as Figure 1.6. This event spans 3 days (72 hours)

      (figure and caption based on Fig. 11 by Rastaetter et al. (2016).Reproduced with permission of John Wiley & Sons).

The authors define the integrated value yield as the ratio of track‐integrated values for model to observations (IYI). In addition, they define the amplitude yield (YI) as the ratio of the model maximum to observed maximum values. The yields (YI) are plotted in panels (b) and (f) as base‐2 logarithms, denoted as Id. For a perfect prediction, Id (1) = 0. They use the term yield to refer to YI. The timing error (DT) is the difference in minutes between the time of maximum values in the model and the time of the observed maximum. This is shown in panels (c) and (g). Finally, the Dst and AL indices are shown for the time intervals in panels (d) and (h), respectively. Formats for Figures 1.7 and 1.8 are identical.

      The study showed that for each model tested, there was a large spread in yield in the six events. Overall, the physics‐based models predicted Poynting flux that could be larger or smaller than the observed values, while the empirical models tended to underpredict the EM power. There was little consistency in the range of predictions across the events. The skill scores for all the models tested for all six events are shown in Table 3 of the paper (not shown here).

      Apart from the generally wide spread in model results, the fluctuations in measured Poynting flux are not captured in any of the models (see Fig. 1.8, panels (a) and (e)). This is also apparent in the plots of the timing errors shown as DT in panels (c) and (g) in both figures.

1.5 СКАЧАТЬ