There are several proposed physical processes which may contribute to the cross-polar potential and thus drive ionospheric convection around the polar caps. It is generally believed that magnetic reconnection is the dominant process, however dynamos such as viscous interaction and impulsive penetration are other possible contributors. A comprehensive statistical study has been conducted using data from the DMSP F13 satellite for passages along the northern hemisphere dawn-dusk meridian, with focus on typical two-cell convection patterns during times of steady southward IMF conditions. The results show that the low-latitude dynamo (viscous interaction or reconnection in the LLBL) on average accounts for only 1–2 kV of the total potential drop, values lower than those previously predicted. At rare occasions this dynamo can be a significant source of energy, however, contributing to more than 20 kV of the cross-polar potential.
 As a result of the solar wind-magnetosphere interaction, part of the solar wind kinetic energy is transferred into the magnetosphere in the form of electrical energy. This interaction sets up electric fields in both polar caps together with an associated plasma convection in the ionosphere. The convection patterns can vary extensively with interplanetary conditions, but during steady southward IMF they are mostly variations of a two-cell convection pattern with anti-sunward flow over the polar cap and a sunward return flow on the dawn and dusk flanks, as thoroughly described by Heppner and Maynard .
 For the standard two-cell pattern, the cross-polar potential drop can be defined as the maximum potential difference in the polar region. There are several proposed dynamos that may contribute to this potential drop. Magnetic reconnection has long been regarded as the principal source, however, other processes such as viscous interaction [e.g., Axford and Hines, 1961; Axford, 1964; Sonnerup, 1980] and impulsive penetration [e.g., Heikkila, 1982; Lemaire and Roth, 1991] may also contribute to the cross-polar potential and drive plasma convection. Magnetic reconnection is believed to take place primarily at the dayside magnetopause and the equatorward borders of the cusps.
 The low-latitude boundary layer (LLBL) is a tangential discontinuity just inside the dayside magnetopause, as well as on the flanks of the magnetosphere. The LLBL plasma is believed to be a mix of magnetosheath and magnetosphere plasma, though its exact nature is still unclear. Some interpret it as being on closed field lines [e.g., Eastman and Hones, 1979; Ogilvie et al., 1984; Phan et al., 1997], partly closed [e.g., Roeder and Lyons, 1992], open field lines [e.g., Gosling et al., 1990; Fuselier et al., 1995] or recently closed field lines with plasma induced by reconnection during northward IMF [Song and Russell, 1992; Song et al., 1994]. In the closed field line theory, the LLBL potential contribution on the magnetopause flanks is primarily believed to be due to a viscous interaction between the magnetosheath and magnetopause plasma. This could for example be due to Kelvin-Helmholtz instabilities at the magnetopause, driven by the solar wind shear. Such instabilities have been observed on the magnetopause [Owen et al., 2004; Hasegawa et al., 2004], although there are still questions as to their effectiveness. An overview of the current controversies regarding the open or closed state of the LLBL can be found in the work by Lockwood et al. .
 The LLBL has generally been considered only to be a minor contributor to the cross-polar potential, however its importance has been disputed and not yet thoroughly investigated. Sonnerup  estimated the boundary layer contribution to be of the order of 10–15% of the total potential drop. Mozer  reached a similar conclusion with around 5 kV of the potential originating in the LLBL. Newell et al. [1991b] made a few observations of the LLBL potential that ranged between 3 and 15 kV, with the average around 5 kV. In a more recent study, Blomberg et al.  presented observational evidence that the low-latitude dynamo could sometimes generate as much as a third of the potential for southward IMF conditions. In this paper we present the first statistical study of the sources that contribute to the cross polar potential, as all previous estimates have either been analytical or based on a limited set of observations.
2. Methods and Data
 This study aims to determine the relative importance of the different dynamos for the cross-polar potential during times of steady southward IMF. For this, we have used six years of data (1996–2002) from the DMSP F13 satellite (about 30,000 orbits), an ionospheric satellite at approximately 850 km altitude. The reasons for this choice is the satellite's dawn-dusk orbit which provides data from a cross section of the convection pattern and the low altitude of the satellite which gives a short cross-polar transit time. The satellite should therefore be able to give valid estimations of the cross-polar potential. The data used cover half a sunspot cycle from a sunspot minimum in 1996 to a maximum around 2002. The F13 provides both precipitating ion and electron data from the SSJ4 instrument (measuring energies from 30 eV to 30 keV) and has an ion drift meter that gives the flow velocity of the plasma.
 The magnetospheric regions are usually well defined and easily identified at high altitudes, while the corresponding ionospheric map is more difficult to obtain. In a series of articles, Newell and coworkers [Newell and Meng, 1988, 1992; Newell et al., 1991a, 1991b] have defined a classification procedure for the ionospheric counterparts by comparing the field-aligned population at high altitudes to ionospheric particle characteristics. They also implemented a pattern recognition neural network together with an on-line database containing the classifications for each DMSP satellite pass [Newell et al., 1991c]. The precipitation regions of most interest here are those defined on the dayside of the dawn-dusk meridian: The low altitude cusp, the plasma mantle, open LLBL, LLBL, the dayside extensions of the boundary and central plasma sheet (BPS and CPS), polar rain and areas void of precipitation. CPS and BPS do not neccesarily map to the plasma sheet but they represent the auroral-like precipitation that often can be seen at low latitudes, see the work by Newell and Meng  for more information of the spatial distribution of the regions.
 In order to ensure that all analyzed events are comparable in terms of both interplanetary and ionospheric conditions, we have applied the following requirements on the data set:
 1. The IMF must be steadily southward during the event (Bz < 0), as well as for the preceding hour. The magnetic field data used are the five minute averaged OMNI data, which already take into account the solar wind transit time.
 2. The potential pattern should correspond to two dominant convection cells, typical for steady southward IMF. The cross-polar potential drop is thus easily interpreted as the difference between the calculated maximum and minimum potential. Only events from the northern hemisphere have been considered, since the satellite mostly passes on the nightside of the dusk-dawn meridian in the southern hemisphere in which the precipitation regions are defined differently.
 3. In order to ensure that values close to the true maximum and minimum potential are recorded only the orbits which closely follow the 6–18 MLT meridian have been selected, as they on average should give reliable estimates of the cross-polar potential. Another advantage with this is that we believe Newell's automated precipitation region classifications to be more reliable in this area, as the open field line portion is mostly void of particle precipitation or contains a soft polar rain. The transition from the field lines connected to the mantle and the lobe to those connected to the LLBL is then easily identified by a step in the measured flux, often seen simultaneously in both the ion and the electron data (see Figure 1 for a typical case). In the vicinity of the cusp however, the boundary between the LLBL and the cusp is often harder to distinguish, something that would give rise to a larger uncertainty in the final statistics if such events were included.
 The procedure for calculating the cross-polar potential is one commonly used when dealing with DMSP data [e.g., Rich and Hairston, 1994; Hairston et al., 1998]: The electric field along the spacecraft trajectory is calculated as the cross product of the measured ion drift velocity perpendicular to the spacecraft trajectory and the downward component of the modelled geomagnetic field. A reference line for zero potential has been defined at 50 degrees MLAT, and any offset between these endpoints is symmetrically removed from the data. This offset is believed to be primarily due to time variations in the electric field. A variation of less than 25% of the total potential drop is assumed acceptable, events with larger variations have been excluded from the analysis. The potential contribution for a specific precipitation region is defined as the integral of the electric field over the corresponding areas poleward of the convection reversals.
3. Results and Discussion
 Due to the choice of high-latitude orbits, we can assume that the precipitation regions of each event are defined with high accuracy. There are however certain cases when the classifications evidently are erroneous. For very intense precipitation of low-energy electrons over the polar cap, the classifications often switch over to BPS in the middle of clearly open field line regions. This results in a serious over-estimate of the low-latitude dynamo contribution for these events, in some cases values exceeding 40 kV. Due to this we have also done a manual inspection of the data set and a few events which contained a high percentage of clearly erroneous classifications have been removed from the statistics. Events which contain only minor deviations from the expected, as sometimes observed in the open/closed field line boundary region, have been kept and are regarded as acceptable classification deviations, this to avoid introducing too much subjectivity into the analysis. These questionable parts should only have a minor impact on the final result and will not affect the general trend. For orbits that do not pass through the global maximum and minimum potential areas the potential contribution from the low-latitude dynamo will be underestimated, as argued by Blomberg et al. . The total low-latitude contribution and the true cross-polar potential could thus be slightly higher than measured here, but in general we expect them to be representative.
 Another important issue to deal with are regions which are either labeled unclassified or parts for which we lack classification data all together (on average 3.3 kV (4.5% of the total potential drop) and 6.6 kV (9%) respectively). These regions can most often be found either as narrow regions in-between two different classifications, or as larger chunks mostly on the open field line portion of the events. As we cannot draw any general conclusion from this category, we remove all events in which the non-classified part of the potential exceeds 10% of the total cross-polar potential (about 38% of the remaining data set).
 After applying these additional restrictions, we have a list of 385 well defined events for the statistical analysis. The main regions observed are either void, polar rain or particles with BPS characteristics. A complete list of the contribution for each precipitation region can be seen in Table 1. Thanks to the hard restrictions on the data set, we believe that BPS, CPS and LLBL all can be grouped into a general low-latitude dynamo category, while void and polar rain relate to the high-latitude dynamo (driven by magnetic reconnection). Cusp, open LLBL and mantle are never observed in the selected events. For satellite passes in the vicinity of the cusp this would be a questionable procedure though, as it is less evident how the open/closed field line border relates to the observed LLBL, open LLBL, BPS and cusp at lower latitudes. By grouping them together into high- and low-latitude domains, we will also get more informative median values compared to treating each precipitation region separately. As there is a large difference between the mean and median values in Table 1, it is of interest to study the distribution of the potential contributions among the events. The histogram in Figure 2 shows us that the low-latitude contribution on average is close to zero, apart from a few occasions where it can generate larger parts of the cross-polar potential. Almost 45% of the events had less than 1 kV of the potential generated by a low-latitude dynamo, and for 70% of the events the contribution was less than 5% of the total potential drop. It is also worth studying the properties of the unclassified part of the potential drop. If we make the assumption that an unclassified area surrounded by high-latitude (low-latitude) dynamo regions on both sides can be regarded as a high-latitude (low-latitude) domain, then we can decrease the amount of unclassified potential drop and at the same time get an estimate of the uncertainty of the low to high-latitude border. This shows that slightly more than half of the previously unclassified potential drop belong to high-latitude regions, and barely nothing to low-latitude regions. The average potential in-between the regions is around 1–1.5 kV.
Table 1. Max, Mean, and Median Potential Contributions for the 385 Events Studied, Separated by Classifications. “Max” shows the maximum value observed among the events, and the “Mean Ratio” column shows the mean values as percentage of the total.
Mean Ratio, %
 We have also performed a simple regression analysis on the data set, to see how the regions vary with respect to the mean solar wind parameters for the duration of the event. For the IMF correlation, the part of the potential drop generated by the high-latitude dynamo shows a very clear relation to the magnitude of the southward magnetic field, while the low-latitude group shows a very weak IMF dependence (see Figure 3). This is a result which agrees with our expectations and thus lends credibility to the classification procedure and the presented results, as only the high-latitude dynamo is expected to have a strong dependence on the IMF direction and magnitude. The solar wind velocity on the other hand shows only a weak correlation with the cross-polar potential drop, and the low- and the high-latitude dynamo groups both give comparable results when treated separately. The values are slightly lower than expected, specifically for the low-latitude category as the occurrence of viscous interaction is expected to be closely related to the solar wind shear.
 For the specific conditions studied here: steady southward IMF and clear two-cell convection patterns, the low-latitude dynamo generally gives a very small contribution to the total cross-polar potential drop, not larger than 1–2 kV for the average case. As we selected only orbits close to the dawn-dusk meridian where the low-latitude dynamo should be at its peak, we expect this to be a valid estimate of both the total cross-polar potential and the low-latitude contribution to it. A regression analysis with solar wind lends credibility to the differentiation between the low and high-latitude dynamos. Even though the low-latitude part is insignificant at most times, we did also see a few unusual events where it can contribute up to 25 kV of the cross-polar potential. This result brings some clarification to how previous predictions and observations may be related, and the specific details of a few of these events will be addressed in a separate paper.
 Work at the Royal Institute of Technology was supported by the Swedish Research Council and by the Alfvén Laboratory Centre for Space and Fusion Plasma Physics. Work at the University of Texas at Dallas was supported by NSF grant ATM0536868. The authors thank the ACE, IMP-8, and Wind instrument teams for providing magnetometer and plasma data through the OMNIWeb data explorer. The DMSP particle detectors were designed by Dave Hardy of AFRL, and data obtained from JHU/APL. We thank Dave Hardy, Fred Rich, and Patrick Newell for its use. The DMSP thermal ion data were obtained from the Center for Space Sciences at the University of Texas at Dallas. We thank Rod Heelis, Marc Hairston, and Robin Coley for its use. We also thank N. Brenning for valuable insights.