Storm time polar cap expansion: IMF clock angle dependence

Beket Tulegenov1, Joachim Raeder1, William D. Cramer1, Banafsheh Ferdousi1, Timothy J. Fuller-Rowell2, Naomi Maruyama3, and Robert J. Strangeway4 1Space Science Center, University of New Hampshire, Durham, NH, USA 2Cooperative Institute for Research in Environmental Sciences, University of Colorado, Boulder, CO, USA 3NOAA Space Weather Prediction Center, Boulder, CO, USA 4University of California, Los Angeles, CA, USA Correspondence: J. Raeder (J.Raeder@unh.edu)

. List of the modeled storms with the mean and standard deviation of the differences in OCB latitude between the DMSP and OpenGGCM results. The second to last column is the mean model bias, where positive values indicate that the model on average predicts a lower latitude of the OCB, and the last column is the variance of the difference between the data and the model prediction. flux is used as a proxy for solar UV/EUV radiation in CTIM. The sunspot number is required for charge exchange calculations in the RCM model.

Event selection 95
In order to determine the behavior of the polar cap under dynamic solar storm conditions, four Coronal Mass Ejection (CME) events, and one Corotating Interaction Region (CIR) event were identified in the period from 2003 to 2019. The main selection criteria were dictated by the availability of solar wind and IMF data, as well as good coverage with DMSP data for the comparisons. All events have minimum Dst values below -35 nT. We also required the B y and B z components of the IMF to be larger than 10 nT in magnitude in the GSE coordinate system. These five solar storm periods are listed in Table 1. In 100 addition we considered three more cases, where we changed the signs of IMF B y , B z , or both, in event 1, to test a hypothesis which is described in detail in Section 3.3.

Comparison with DMSP data
Before we consider a detailed analysis of the simulation runs, we first assess the realism of the simulations by comparing the 105 model OCB latitude output with DMSP observations. The DMSP satellites are a series of polar-orbiting spacecraft with an altitude of ≈ 850 km in sun-synchronous orbits. The precipitating ion and electron data from the on-board instruments have been used to identify polar cap boundary crossings in a number of previous studies (Hardy et al., 1984;Sotirelis et al., 1998;Milan et al., 2003;Wing and Zhang, 2015;Wang et al., 2016Wang et al., , 2018. Spectrograms of ion and electron differential fluxes in a range from 30 eV to 30 keV were inspected to identify the polar cap boundary crossings of the satellites. Some previous 110 studies have suggested using the b6 boundary as an open-closed field boundary (Newell et al., 1996;Hubert et al., 2006). These boundary locations can also be identified using a set of quantitative algorithms developed by Newell et al. (1996). However, during geomagnetically active periods these algorithms tend to either fail to determine the boundary at all, or sometimes they misidentify the boundary.
The comparison of the OCB magnetic latitudes between DMSP and the OpenGGCM simulations is shown in Fig. 1 through 115 5. Figure 1 shows the results for the November 20, 2003 event. Figure 1d shows the IMF and solar wind for reference. For this event, data was available from four DMSP spacecraft (F13-F16). The blue markers in Fig. 1a indicate the polar cap crossing latitudes of the DMSP satellites, while the modeled OCB latitudes, determined along the same orbits, are shown in red. Figure   1b displays the Magnetic Local Time (MLT) coverage of the DMSP spacecraft during the event. The DMSP s/c are in sunsynchronous orbits that are close to the terminator. Thus, the MLT coverage is uneven and concentrated in the dawn and dusk 120 sectors. Near noon data are sparse, and there are no nightside crossings between 2200 and 0600 MLT. Figure 1a shows that the model follows the real OCB pattern geometry reasonably well. However, there is significant scatter, both in the data, as well as in the model results. In particular, the DMSP crossings often jump considerably from one orbit to the next, and the model shows a similar behavior. Obviously, the OCB is quite dynamic. Figure 1c shows a histogram of the differences between the model and the DMSP OCB determination. The visual method of 125 determining precipitation boundary from the DMSP spectrograms is fairly accurate, probably better than one degree. However, this method assumes that the precipitation boundary is also the OCB. During very active solar times that may not always be the case. There is a significant number of polar cap crossings where no clear precipitation boundary can be identified. Such cases are not included in the plots, but their existence indicates that the OCB and precipitation boundary may not always be the same. The simulation results shown in Fig. 1 through ?? also show that the OCB in the model is very dynamic and not always 130 a smooth curve, but rather corrugated. That would also explain some of the scatter, assuming that the OCB in nature behaves similarly. The comparison shown in Fig. 1c indicates that on average the difference between DMSP and model OCB latitudes is −0.30 ± 4.57 • . Thus, in spite of the scatter, there is no significant bias between the model and the data for this event case.
This, together with the fact that the data and the model pattern follow each other, this gives us confidence that the model results represent the true OCB to a high degree.

135
The other storm periods have data available from only three DMSP satellites (F16-F18). Figure   4.12 ± 2.20 • . However, this event also has by far the fewest data available.
To summarize the model-data comparisons, on average, the model overestimates the location of the polar cap boundary by    bias is not obvious. One might hope that there is a pattern in the solar wind or the IMF that might give a hint, but nothing stands out, although there is a trend that the larger storms seem to give more accurate model results than the weaker ones. However, the number of storms studied is too small to allow for any compelling conclusions.

150
As demonstrated in Fig. 1b   is typically an oval that is displaced towards midnight, during storms like these, that appears no longer to be true. In particular, the strong southward IMF driving shifts the maximum extension to the dayside and to the flanks, depending on the clock angle, i.e., depending on the relative strength of the IMF B y component versus the B z component. Figure 6b displays the OCB dynamics for the November 1, 2012 event. During this event, the IMF clock angle varies very 170 smoothly through a 270 rotation during an interval of almost one day. Again, the polar cap expansion starts with the southward turn of the IMF. However, this southward turn is accompanied by a strong IMF B y component. Therefore, the expansion does not start at noon, but rather near the dawn terminator. The expansion then rotates towards noon as the IMF become s more southward. At the end of the interval, when the IMF is essentially duskward, the maximum of the expansion is near the dusk terminator. We also note that during the CME sheath phase the IMF is mostly northward, and correspondingly, there is no 175 significant polar cap expansion.
The next event, shown in Fig. 6c in the same format, has a sheath that includes several significant southward IMF excursions.
During each of these excursions, the polar cap rapidly expands. Like in the other cases, the direction of the expansion is dictated by the clock angle. First, the IMF is dawnward, and the expansion is towards dawn, followed by a duskward turn that is matched by the excursions. After the initial part of the sheath with southward IMF the remainder of the sheath has northward field, and  south. This is different from the previous case, where the clock angle mostly changed because B y changed. However, just like in the previous case, the apex of the OCB follows the clock angle. This shows that the clock angle is the controlling factor, not the IMF B y component alone.

185
The next case, 10 March 2018, shown in Fig. 6d, is different in that the IMF slowly rotates from south to north, while IMF B y stays roughly constant and negative. The behavior of the OCB apex is as expected from the previous cases. As the clock angle slowly rotates southward the OCB first expands near noon and then the apex rotates towards dusk following the clock angle. A significant difference compared to the previous cases is that the rotation continues all the way to midnight. That shift of the apex to midnight occurs while the IMF is nearly northward. shows that in most cases the OCB is fairly smooth where it is most expanded. When the OCB contracts, it tends to become a more ragged line. The latter may be explained from the fact that tail reconnection is localized, and thus the poleward motion 195 of the OCB becomes localized as well.
Finally, Fig. 8a shows the strongest storm considered here, which occurred on 19-21 November 2003. This storm also has the prototypical full circle IMF rotation of a flux rope. As in the other cases, the OCB apex follows the clock angle. This occurs not only through the storm main phase, but also during the sheath rotation and the two other rotations that follow the storm main phase. We will later use this case for numerical experiments to rule out other causes than the clock angle for the polar cap By doing so, we try to exclude other possible influences on the OCB, for example, seasonal effects such as dipole tilt (Russell et al., 2003) and ionosphere conductance distribution (Lu et al., 1994). Also, the clock angle changes occur in different directions. Specifically, between cases 8a and 8c, and between 8b and 8d the clock angles are reversed. It is obvious that the clock angle alone is the controlling factor. In particular, Figures 8a and 8c, and 8b and 8d, respectively, are virtual mirror 210 images of each other. There is some net shift of the OCB in all of these cases, which is likely due to season, since the event date is close to winter solstice.
The experiment also tells us that the direction in which a clock angle change occurs is not very important. That, in turn, also means that the reaction to clock angle changes occur with only small time delays, at least on the time scales of these