We present maps of ionospheric precipitation regions, based on 11 years of DMSP particle data, binned by the interplanetary magnetic field (IMF), with superposed SuperDARN convection streamlines gathered under similar conditions. The convection patterns are transformed into an inertial coordinate system. The maps, which include both the nightside and dayside, are created in a fully automated fashion, with, for example, the cusp centered at its centroid latitude for each half hour bin of MLT, with a latitudinal width equal to the statistical difference between the poleward and equatorward edges. The mantle asymmetry about noon does not fit the pattern expected from simple theoretical considerations (namely, that the mantle should be thicker postnoon for positive By in the Northern Hemisphere). The mantle is appreciably thicker prenoon than postnoon, especially for positive By but also even for negative By. This asymmetry matches the SuperDARN convection flows, in which, irrespective of the sign of By, most of the conversion of closed field lines to open occurs prenoon. Quantitatively expressed, for southward IMF, the potential encompassed by flux crossing the open-closed boundary prenoon (0600–1200 MLT) exceeds that for postnoon (1200–1800 MLT) by 30 kV to 15 kV for By > 3 nT and by 30 kV to 20 kV for By < −3 nT. The mantle shape thus matches convection pattern variations. Only ∼25–35% of the dayside open-closed field line conversion occurs within the particle cusp, with the lower number appropriate to northward IMF. Most closed-to-open field line conversion occurs away from noon. Merging is thus active throughout the frontside magnetosphere. Field lines that merge well away from noon do not experience enough particle inflow against the solar wind velocity to produce anything more than a weak, deenergized (mantle) precipitation. The boundary between the dusk cell and dawn cells consistently coincides with one edge of the cusp. IMF By also controls where most of the nightside reconnection occurs. For positive Bz and By > 3 nT, 31 kV reconnects from 1800 to 2400 MLT, but only 14 kV reconnects from 0000 to 06000 MLT. The convection reversal boundary (CRB) consistently coincides with the nightside open-closed particle boundary (OCB). On the dayside, the CRB lies equatorward (poleward) of the OCB in the postnoon (prenoon) sector for By < 0 (By > 0). This shift is consistent with the effects of an interhemispheric current produced by the partial penetration of the IMF By into the frontside magnetosphere.
 Twelve years ago we presented a map of the dayside precipitation regions according to the magnetospheric source region [Newell and Meng, 1992]. Here we considerably extend the work to include the nightside as well as the dayside, to include separate maps for separate IMF conditions, and to provide convection patterns compiled under the same IMF conditions with the same magnetic coordinate system. The combination of the particle precipitation regions with the convection patterns proves highly illustrative and allows for some cross-consistency checks. For example, the SuperDARN-based convection patterns, even transformed into an inertial reference frame, show a larger potential prenoon than postnoon. Likewise, the particle precipitation maps show that the dayside mantle, which corresponds to field lines with relatively recent merging, is also found preferentially prenoon. Either finding taken alone might raise questions, but taken together their consistency suggests that the result is not an artifact and needs a physical explanation. Indeed, several consistency checks are possible between these disparate data sets compiled with very different algorithmic approaches. The new maps also include enhancements such as average energy and energy flux for each region and MLT.
 Most of the time, the interplanetary magnetic field lies largely within the plane of the ecliptic so that typically ∣By∣ ≫ ∣Bz∣. Even with the large number of DMSP satellites used in this study, and even over an 11 year period, the data in some local time bins and some latitudes was not extensive enough to compile maps over less common IMF conditions. Thus the particle maps are all By dominated, in the sense that the typical magnitude of By is significantly larger than the typical magnitude of Bz. The convection patterns were recalculated to match. Thus the “northward” IMF conditions presented in this paper differ from cases often presented as northward IMF, in that our maps match the more typical IMF circumstances, with more merging and a wider polar cap.
2. Data and Techniques
2.1. DMSP Particle and SuperDARN Radar Measurements
2.1.1. DMSP Auroral Particle Data
 Data from the SSJ/4 electrostatic analyzers on the DMSP series satellites (F6 through F12) from 1 January 1984 through 31 December 1994 were used in constructing the particle maps. These 11 years, besides covering all phases of a solar cycle, in general contain higher-quality data than some later years, during which increased operational longevity produced deterioration of sensitivity and hence boundary identification reliability in several detectors. During recent years, some SSJ/4 detectors have been launched with the low-energy ion head performing suboptimally, which has also reduced the quality of automated boundary identifications. For these reasons, our studied concentrated on an even 11 years worth of good quality data.
 The DMSP satellites are in Sun-synchronous, nearly circular polar orbits at ∼835 km altitude, with orbital inclinations of 98.7 degrees. The orbits of the DMSP satellites are such that the least covered regions are postnoon and especially postmidnight, except at high magnetic latitudes. The SSJ/4 instrumental package included on all these flights uses curved plate electrostatic analyzers to measure electrons and ions with one complete spectrum each obtained per second [Hardy et al., 1984]. The satellites are three-axis stabilized, and the detector apertures always point toward local zenith. At the latitudes of interest in this paper, this means that only highly field-aligned particles well within the atmospheric loss cone are observed.
2.1.2. SuperDARN Radars and Convection Velocity Data
 The Super Dual Auroral Radar Network (SuperDARN) is an international collaboration involving the scientists and funding agencies of over a dozen countries [Greenwald et al., 1995]. In the Northern Hemisphere it consists of nine HF radars arranged along a longitudinal axis that extends from Scandinavia to Alaska. All nine contributed data to the development of the convection patterns used here. The radars observe the E × B convective drift of plasma in the high-latitude ionosphere. The measurements can be merged into a depiction of the instantaneous global convection pattern [Ruohoniemi and Baker, 1998]. The radars operate continuously and the resolutions available in the large-scale convection mapping are typically 1–2 min in time and 100 × 100 km2 in space. The convection product of SuperDARN has been utilized in many studies of convection dynamics, e.g., Greenwald et al. , Huang et al. , and Ruohoniemi et al. .
 The HF radar data can also be applied to the derivation of statistical convection patterns, which express the average condition of the convection for some stated conditions. An example is the convection model of Ruohoniemi and Greenwald , which specifies a range of convection patterns sorted by IMF magnitude and direction as determined from a bin-averaging analysis of Goose Bay radar data. Recently, the data set of the larger set of SuperDARN radars in the Northern Hemisphere has been analyzed for a finer examination of the factors that impact high-latitude convection (J. M. Ruohoniemi and R. A. Greenwald, Dependencies of high-latitude plasma convection, submitted to Journal of Geophysical Research, 2004, hereinafter referred to as Ruohoniemi and Greenwald, submitted manuscript, 2004). The period for capture of the data set is 1 January 1998 through 30 November 2002 or roughly half a solar cycle centered on solar maximum. For this study described here we have processed the SuperDARN data set using the specific IMF conditions that were applied to the processing of the DMSP particle data.
2.2. Particle Boundary Identifications
2.2.1. Nightside Boundaries
 Nightside boundary identifications and nomenclature were the subjects of a prolonged debate throughout the 1970s and 1980s [Feldstein and Galperin, 1985]. We incorporated the results of those years, along with a number of more recent findings, to produce a set of automated nightside boundaries with clear geophysical significance and quantitative computational definitions [Newell et al., 1996]. The algorithms proposed there are used here. The specific boundaries used on the nightside are as follows.
 The equatorward boundary is typically the “zero energy” (actually 32 eV) convection boundary (b1e), which usually corresponds to the plasmapause [Horwitz et al., 1986], since zero-energy particles have no curvature and gradient drifts. This low-energy electron boundary has long been used for Air Force space weather forecasting and generally moves equatorward with increasing geomagnetic activity [Gussenhoven et al., 1983; Nakai et al., 1986]. However, because ions with many keV energy in the dusk-midnight sector can precipitate equatorward of any electrons, our equatorward boundary is actually “b0” in the terminology of Newell et al. , i.e., the latitude at which significant electron or ion precipitation of any sort is observed. Outside of the dusk-to-midnight sector, however, it is overwhelmingly the same as b1e, the zero-energy boundary.
 Moving poleward, the next boundary used is b2i, the equatorward peak in ion precipitation in the 3000 eV to 30 keV range. Physically, this corresponds to the boundary of the isotropic plasma sheet at high altitude [Lyons and Speiser, 1982], which also corresponds to the point where the ion gyroradii is ∼1/8 the curvature of the magnetic field lines [Sergeev et al., 1983; Newell et al., 1998]. This boundary coincides with what Feldstein and Galperin  term the start of the main plasma sheet at high altitude.
 The next boundary is the structured/unstructured boundary (b4s), computed by a running autocorrelation between electron spectra. The “unstructured” region is the region where all electron spectra are very similar, having a high autocorrelation. This is a quantitative version of the traditional CPS/BPS boundary. Note that the terms “BPS” and “CPS” were applied by Winningham et al.  to describe precipitation regions 2 years before the terms “plasma sheet boundary layer” and “central plasma sheet” entered the literature of high-altitude observations. In fact, the low-altitude precipitation regions traditionally termed “BPS” and “CPS” do not correspond to the high-altitude regions with similar sounding but later-endowed names [Feldstein and Galperin, 1985]. Here “CPS” and “BPS” describe all precipitation respectively equatorward of, and poleward of, the structured/unstructured boundary (b4s) and thus preserves the long tradition of low-altitude satellite research.
 Finally, the b5e boundary is the poleward boundary of the main auroral oval, the point where the energy flux declines by an order of magnitude within a fraction of a degree, and b6 is the poleward boundary of the subvisual region, consisting (when present at all) of very weak electron and ion fluxes far below the intensity of the main oval but spectrally distinct from polar rain by, for instance, the presence of ions.
2.2.2. Dayside Boundaries
 A series of articles researching the quantitative differences between the various dayside precipitation regions has been previously presented. For example, cusp identification [Newell and Meng, 1988], mantle identification [Newell et al., 1991a], and LLBL identification [Newell et al., 1991b; Newell et al., 1998] have all been the subjects of dedicated research papers. Although the shocked solar wind is the ultimate source for much of the dayside precipitation at higher latitudes, clear quantitative differences have been documented in the literature.
 The cusp (a more accurate but much more cumbersome term is “low-altitude particle cusp”) is the region where energies and densities generally approximate that of the frontside magnetosheath. Thus for example, ions have spectral flux peaks somewhere around 1 keV, reflecting their original solar wind kinetic energy, with spectral flux peaks of ∼108 eV/cm2 s str eV (or number flux peak of 105/cm2 s str eV). Electrons have temperatures of a few tens of eV, and densities closely match the ions (usually in the 1–10/cm3 range). The particle cusp thus includes field lines merged from ∼1–3 min ago, long enough for the main ion population to reach the ionosphere, until ∼10 min old, when the high-altitude field line has convected sufficiently far downstream that ion entry is limited.
 Similarly, the low-altitude particle mantle consists of deaccelerated shocked solar wind. The mantle lies within the region of supersonic flow downtail, and hence only a weak flow of low-energy ions crosses the magnetosheath headed toward the Earth. Because j × E < 0 in the mantle, the ions are further decelerated in crossing the current sheet. Mantle precipitation is thus at energies below that of the cusp (below 1 keV), with electrons near the polar rain level. Also, because of the long time elapsed since merging, the low-energy ion cutoff observed in the cusp is not observed in the mantle. Densities are typically several times below magnetosheath levels.
 The LLBL can be either open or closed. More commonly, it is partially open and partially closed [Mitchell et al., 1987; Phan and Paschman, 1996], with the frontside of the magnetopause typically open, especially for southward IMF, and the flanks more typically closed. When the LLBL has a low-energy ion cutoff, as does the cusp, it can clearly be identified as open. In the absence of a low-energy cutoff, the LLBL is probably typically closed [e.g., Newell and Meng, 1998]. For the work done here, the LLBL was considered closed unless it was both immediately equatorward of the cusp and possessed a clear low-energy ion cutoff indicative of recent merging. These fairly strict conditions probably somewhat overestimate the fraction of the LLBL that is closed.
2.3. Constructing the Convection Patterns
 As discussed in an earlier section, a database of velocity measurements from the northern SuperDARN radars has been assembled and processed into a format suitable for the reduction to statistical convection maps. New findings on factors that influence convection are described by Ruohoniemi and Greenwald (submitted manuscript, 2004), who also present a new convection model with sorting by IMF magnitude and direction. This model is, in the main, consistent with that derived earlier on the basis of Goose Bay HF radar data alone [Ruohoniemi and Greenwald, 1996]. The two models are distinguished as RG04 and RG96, respectively. The methods of data reduction were substantially the same in both derivations, except for some minor improvements with RG04 in the specification of parameters such as the size and shape of the convection zone [Shepherd and Ruohoniemi, 2000].
 For comparison with DMSP particle data, we deemed it inappropriate to directly apply the patterns of the RG04 model, as the IMF sorting criteria were not the same as those chosen for processing the particle data, namely, (1) Bz < −1, By < −3; (2) Bz < −1, By > 3; (3) Bz > 1, By < −3; (4) Bz > 1, Bz > 3. It was, however, straightforward to rerun the RG04 processing code with these alternative IMF selections, resulting in a set of four convection patterns matched to the particle data sets. The convection patterns presented here are thus a variant of the RG04 model.
 We briefly outline the steps in processing the velocity data into convection maps. The radars scans were flagged according to the IMF criteria and the line-of-sight velocity data were bin-averaged within magnetic latitude/longitude cells measuring approximately 100 km on a side. The measurements in the solar wind were provided by the ACE satellite, and conditions on the constancy of the IMF at the magnetopause were applied. The result for each IMF sorting was a single composite map showing the variation in the line-of-sight velocity with azimuth within each cell. The velocity data were fit to an expansion of the electrostatic potential, Φ, in terms of spherical harmonic functions. The order and degree of the fitting were L = 8 and M = 8, respectively. Because the coverage of the high-latitude ionosphere over the full set of scans is rather complete from the pole to within a few degrees of 60° MLAT, there is no need to supplement the measurements with model predictions, as must generally be done for mapping the instantaneous convection pattern [Ruohoniemi and Baker, 1998]. The fitting produces a best-fit map of the spatial distribution of Φ, where the plasma flow is directed parallel to the equipotential contours and velocity magnitude is proportional to the contour density.
 We made one additional alteration to support this study. The radar data are best compared with the particle data in the inertial reference frame, for which an object at a fixed location in the Sun-Earth geometry has zero velocity. For this application we adjusted the line-of-sight velocities for the corotation of the radars to obtain convection patterns in the inertial frame. Generally, a mapping from corotating to inertial coordinates makes the convection pattern more symmetric by causing the dawn cell of the two-cell pattern to gain in size relative to the dusk cell and by more nearly centering the pattern on the noon-midnight meridian [Maynard et al., 1995].
2.4. Constructing the Particle Maps
 It is impossible for a precipitation map to accurately preserve all the various relationships present in the original data. In abstracting the data to produce the maps, our goal is to accurately represent the average centroid position, latitudinal width, and typical MLT extent of each region. Thus our approach emphasizes maintaining the statistical properties of each region individually over maintaining the relationships between regions (such as typical latitudinal separation between regions). One specific example (the most prominent we know about) may help illustrate. The cusp is longitudinally more extended for a larger magnitude southward IMF [e.g., Crooker et al., 1991], meaning that a satellite pass on the dayside is more likely to encounter the cusp when Bz is strongly negative. Thus passes for which the cusp is not observed have, on average, a smaller southward IMF component than passes for which it is observed. The net result is that even when the same overall IMF constraints are imposed, the average latitude at which the cusp is observed will tend to be lower than, say, the LLBL, which is actually more extended for northward IMF [Mitchell et al., 1987]. Thus even though the maps here preserve the observed latitudinal extent and location of both the cusp and the LLBL individually, an instantaneous picture observing both would probably find the cusp further poleward of the LLBL than our statistical techniques show.
 Our technique begins with determining the following basic property of precipitation observed in the ionosphere. For each half hour MLT and 0.5° MLAT bin, we calculate the probability, Pr(MLAT,MLT), of a given spectrum sampled being from precipitation region r (r = cusp, mantle, polar rain, etc.). This probability is simply defined as the number of spectra of type r observed (according to our identification algorithms) divided by the total number of spectra identified as anything at all within that bin. We also calculate Pr(MLT), the probability that region r is seen at all (at any latitude) at least once during a pass at the specified MLT. Thus Pr(MLT) is calculated on a pass-by-pass basis (using only passes that vary by less than 0.5 hours MLT), while Pr(MLT, MLAT) is calculated based on individual spectra. The resulting set of probability distribution functions is used to derive the numbers needed for the map.
 As a brief digression, the probability distributions, if plotted in their raw form, appear as clouds centered in the correct locations but much more diffuse than the final maps. The interested reader should examine Newell and Meng [1992, 1994] to see this. In a given MLT bin, for example, 1145–1215 MLT, the cusp may range over many degrees of latitude. However its average width is typically only about a degree, as observed on any given pass, and hence on our final maps.
 More concretely, we consider each 0.5 hours of MLT separately. Within that MLT bin, we identify the average position of each location through the centroid rule:
The latitudinal width at each 0.5 MLT local time for region r is simply:
It is not obvious, but neither is it particularly hard to show, that these equations yield the expected results for the time-averaged location. For example, we have (analytically) verified that if one calculates the latitudinal width for each region based on the difference between the poleward and equatorward boundary for each pass taken individually, then averages those individual widths together (taking a mean), one ends up with precisely the value calculated from equation (2). (A region may be encountered more than once on a pass. In that case, equation (2) gives the mean value of the sum of the various widths over single passes).
 The local time extent, or longitudinal width, of each region is similarly calculated from
The above formula gives the time-averaged local time width. For example, if the cusp spends half its time with a 4-hour longitudinal extent and half its time with a 2-hour extent, the formula would yield 3 hours. Thus our approach places each region at its correct centroid latitude, with its average latitudinal width at that local time, and distributes the region over its time-averaged longitudinal extent.
 Hourly average data from IMP-8 was used to bin the DMSP particle data by IMF condition. The broad range of IMF values permitted may raise questions for some readers. In fact, we have verified that our particle maps are insensitive to the upper limit of IMF permitted, probably because even a major storm event, with large magnitude IMF, contributes only a few boundaries a day per region and local time. The SuperDARN radars rarely see significant scatter under large magnitude IMF conditions and thus are equally unaffected by storms.
3. Particle Precipitation Maps and Convection Patterns
3.1. Morphological Maps
 Maps of the precipitation regions binned by IMF By and Bz, with superposed SuperDARN convection patterns are shown in Figures 1a–1d. A few notes about these figures are in order.
 First, the only data gap occurs in the in Bz < 0, By > 3 nT figure, at the equatorward edge of the CPS region, between 0030 and 0200 MLT. Everywhere else, and in all other figures, a white space means a lack of significant precipitation not a lack of data. Although the polar caps are shown empty, they do contain a polar rain signal, which can be extracted by careful averaging over many seconds of data. Since that would add little to the map information (the entire polar cap, at all times, has at least an extremely weak polar rain signal), only the region of intense polar rain (above ∼0.1 ergs/cm2 s) is plotted.
 There is an overall striking correspondence between the particle boundaries and the convection data. On the nightside, the convection reversal boundary fairly uniformly corresponds to the boundary between open and closed field lines inferred from the particle data, namely the poleward boundary of the subvisual drizzle. This coincidence on the nightside occurs for both IMF southward and northward and for both signs of By.
 The coincidence between the CRB and the particle data on the dayside is also good, albeit not quite as straightforward. For example, for southward IMF and positive By the CRB corresponds to a transition to antisunward flowing regions (all but the BPS and CPS) on the dawnside (crescent cell) but with a less clear correspondence on the duskside (round cell). For southward IMF and negative By, the dayside correspondence between the precipitation and convection is good on the duskside (crescent cell) and less clear on the dawnside (round cells). For either sign of By, the agreement is thus clearest in the crescent cell.
 Even in the rounder cell on the dayside, the agreement between the convection and particle maps is still substantial. The chief issue lies with the LLBL, which (whether open or closed) should be convecting antisunward, as indeed the DMSP drift meter studies have shown [Newell et al., 1991b]. For concreteness, consider the 0830–1030 MLT sector for negative By and southward IMF in Figure 1a. The convection pattern within the LLBL has an approximately sunward directed sense of general flow, contrary to expectations and previous findings. However the convection pattern is in fact rotating from primarily zonal to primarily meridional flow, and this rotation commences approximately along the LLBL equatorward boundary. One could argue that even here the CRB agrees with the particle data, although the remnant sunward component of the drift may still be somewhat puzzling. It seems likely that detailed examinations of several case instances using simultaneous DMSP drift meter, particle data, and SuperDARN instantaneous convection measurements are needed to fully understand this possible discrepancy and whether it is serious or insignificant.
 Early work typically assumed that the cusp and most dayside merging are associated with the dayside convection throat [e.g., Siscoe and Huang, 1985]. The present research indicates that the cusp lies entirely within the rounder convection cell. Indeed, the longitudinal boundary of the cusp, which is closer to noon, coincides with the final flow lines that connect with the rounder cell, at least for southward IMF. For northward IMF, the relationship may hold approximately also but with greater uncertainties.
 The mantle, along with the region of intense polar rain, consistently has a prenoon bias for either sign of By. Gussenhoven et al. , in perhaps the most comprehensive study of polar rain systematics, reported that the polar rain is most intense along the dayside prenoon region, which is consistent with what we find here. The mantle also proves to be most intense prenoon, although clear signatures of the mantle can be found widely distributed over the dayside. There does not seem to be any previous reports about this propensity of the mantle to prefer the prenoon sector. In section 4.2 below, we argue that there is a detailed correspondence between the convection flows across the open-closed boundary and the observed mantle thickness.
 The region of significant ionospheric flows, which is defined by the existence of sizeable gradients in the potential distribution, seems to begin at the same equatorward latitude as does auroral oval associated precipitation. With a few exceptions, the region of significant ionospheric flow velocities and the region of significant auroral particle precipitation overlap. This agrees with the individual examples studied by Greenwald et al. , although the result here is more global and of course purely statistical. One possible exception to the rule that regions of fast ionospheric flow and the region of auroral precipitation coincide occurs for northward IMF in the late morning sector. In that region and condition, the CPS precipitation can extend equatorward of the region of fast flows. This is because CPS precipitation in the morning sector for northward IMF is left over from past activity when the IMF was southward. The recovery time for the dayside CPS precipitation is several hours [e.g., Chen and Shulz, 2001], which greatly exceeds the dynamical scale for convection patterns.
3.2. Average Energy Maps
 To better characterize the structure of the nightside oval precipitation, the very broad categories used for the overall morphology of the auroral oval were subdivided when calculating energy flux and average energy. Specifically, these parameters were calculated independently from (1) b0-b2i, essentially, the first precipitation of any kind to the ion isotropy boundary, which is the start of the main plasma sheet; (2) b2i-b4s, the usually narrow and often vanishing region between the ion isotropy boundary and the start of unstructured electron precipitation (hence the equatorward boundary of the traditional low-altitude “BPS”); (3) b4s-b5e, the region of unstructured (poorly self-correlated) precipitation, containing discrete auroral arcs, extending up to the location of an the poleward boundary of the main oval (defined by an order of magnitude flux decrease over a short latitudinal distance); (4) b5e-b6, the sometimes present region of very weak ion and electron precipitation, easily distinguished from polar rain, at intensities far below the main oval.
Figures 2a–2d and Figures 3a–3d show the electron and ion average energy maps, respectively. Note that the ion energies average higher postmidnight than premidnight, while the opposite is true of electrons. This result has been previously reported from position-oriented precipitation models (those that bin by a fix MLAT/MLT grid, rather than by region) [Hardy et al., 1985, 1989]. The interpretation is usually that high-energy ions (those above ∼3 keV) curvature and gradient drift westward, while curvature and gradient drifts for energetic electrons add to the corotational drift eastward. The shift from primarily electron to primarily ion CPS precipitation occurs in the 1200–1500 MLT sector, as observed by the changing average energies of the ions.
 The dayside regions with magnetosheath origins (cusp, LLBL, and mantle) have, as expected, much lower average energies than the rest of the auroral oval. The region with the spectra most closely matching magnetosheath particles is, by our definition, the (particle) cusp. Hence it is no surprise to find that the cusp has the lowest average energies for electrons, as magnetosheath electrons are less energetic than are the boundary layers or polar rain. For the ions, the situation is slightly more complicated. The open LLBL contains only the highest-energy ions from the magnetosheath (which are further accelerated crossing the magnetopause current layer); hence the open LLBL has average ion energies of several keV.
 The high average ion energies in the polar rain simply reflect the lack of statistically significant ion precipitation in this region. Polar rain consists virtually of just the strahl or energetic portion of the solar wind population [e.g., Fairfield and Scudder, 1985]. These low ion count rates cause an elevation in average energy when computed on a moment basis, since the noise is randomly distributed across energy channels. Ions in other regions and electrons in all regions have high enough fluxes so that noise should be a much less significant consideration.
 Along the nightside CPS, especially in the dusk-to-midnight sector, a narrow band of high-energy ions can be seen near but not at the equatorward edge of precipitation. This effect is quite real and can be seen in most individual passes through the nightside oval. The ion average energies generally rise moving equatorward, just as plasma sheet ions are adiabatically heated as they move from regions of lower magnetic fields to stronger magnetic fields closer to Earth. However, the high-energy ion precipitation (above ∼3 keV) stops when the field lines cease being stretched enough for significant pitch angle scattering, and therefore the loss cone becomes depleted [Sergeev et al., 1983; Newell et al., 1998]. However, low-energy ions of ionospheric origin, generally flowing out of the auroral oval, especially from the conjugate hemisphere, are highly field-aligned. These low-energy ions can precipitate furthest equatorward [Sauvaud et al., 1981], at least from 1900 MLT through dawn. In the 1200–1800 MLT sector, the equatorward ion precipitation is mostly confined to ions with enough curvature and gradient drift to overcome corotation eastward and hence to ions above a few keV energy.
3.3. Energy Flux Maps
Figures 4a–4d and Figures 5a–5d show the electron and ion energy fluxes within the various precipitation regions, respectively. As one might expect, typical energy fluxes of both electrons and ions are higher for southward IMF than for northward IMF. This is true for the dayside as well as the nightside oval. Indeed, since the southward oval is both latitudinally broader and has the more intense precipitation, the total energy input into the ionosphere is much larger for southward IMF.
 Although IMF By does order such aspects of dayside morphology as the longitudinal position of the cusp, the broader impact of By upon the global precipitation pattern is less striking. Although one can obviously select a few features that differ between the By positive and negative maps, there does not seem to be much that is systematic (e.g., holds for both electrons and ions and for Bz positive and negative).
 In the more equatorward portion of the oval, where CPS and BPS precipitation dominate, there is a strong trend for electron and ion energy flux to anticorrelate. For example, on the dayside the CPS prenoon has a high electron energy flux and a low ion energy flux, while the opposite is true postnoon. As another example, in the premidnight region the electron energy flux is stronger in the poleward region, while the ion precipitation is stronger toward the equatorward part of the oval. Many such anticorrelations exist in the CPS and BPS energy flux values for electrons and ions. Note that the likely explanations for the two specific instances are quite different: the dayside CPS anticorrelates because energetic electrons and energetic ions have curvature and gradient drifts that take them to the dayside from opposite directions. The premidnight effect likely stems from the existence of intense electric fields accelerating electrons downward while retarding ions in the poleward portion of the oval. Nonetheless, the combined effects of opposing curvature and gradient drifts and opposite response to the sign of any parallel electric fields leads to repeated anticorrelations in electron and ion plasma sheet precipitation.
4.1. Agreement Between Particle and Convection Boundaries
 Although the particle maps and the convection patterns were compiled under the same IMF conditions, the original data sources and the techniques applied differ greatly. The SuperDARN radar data were bin-averaged and subsequently fit to a spherical harmonic expansion of the potential. The DMSP data had geophysically meaningful boundaries identified in the data and averages subsequently performed on these boundaries. It is therefore comforting to observe the many agreements between these disparate and disparately processed data sets.
 First, for both signs of Bz and both signs of By the nightside convection reversal boundary and the particle open-closed boundary (the poleward edge of the subvisual precipitation) agree consistently to within a few degrees and often even better. It is difficult to estimate the uncertainty in the boundary identification algorithm, but the scatter in the data strongly suggests that uncertainties are at any rate comparable to one to two degrees.
 On the dayside, the agreement between convection and particle boundaries is more approximate, with several complexities. We here consider the mantle, cusp, and empty polar cap or polar rain as all lying on open field lines, and thus the furthest equatorward encounter with any of these regions defines the open/closed boundary (OCB). For both signs of Bz the crescent-shaped cell has a CRB, which better agrees with the OCB. Even more tellingly, as the sign of By switches from positive to negative, and therefore the side of noon which contains the crescent cell flips, so too does the side of noon with the better agreement between the CRB and the OCB.
 However, in the rounder convection cell the agreement between the particle and convection maps is harder to interpret, especially within the interval from ∼1 hour MLT to 3 or 4 hours of MLT away from noon. For example, for negative Bz and negative By the convection patterns show more-or-less sunward convection within the LLBL in the 0830–1030 MLT sector. In contrast, instances of simultaneous DMSP drift meter and particle data consistently show that the LLBL is convecting antisunward in this local time sector, as one would expect [e.g., Newell et al., 1991b]. Nonetheless, the RG04 patterns do rotate from primarily zonal to primarily meridional, starting essentially at the equatorward boundary of the LLBL. Thus even within this area of less obvious agreement between the convection patterns and the particle maps, the difference may not be as serious as appearances first suggest.
 Despite the approximate agreement on the dayside, a systematic pattern of offsets exists between the CRB and the OCB. In the afternoon (morning) sector, the CRB lies equatorward (poleward) of the OCB for By < 0 (By > 0). Symmetrically, in the afternoon (morning) sector, the CRB lies poleward (equatorward) of the OCB for By > 0 (By < 0). Previously, Kozlovsky et al.  considered several individual cases of simultaneous SuperDARN and DMSP data in the afternoon sector and found the same pattern. The typical offset they reported (1–2°) is slightly smaller than what appears in the statistical maps.
Kozlovsky et al.  further considered the likely source of these systematic offsets. It is well established that the IMF By can partially penetrate into the Earth's magnetosphere on the dayside as well as the nightside [e.g., Wing et al., 1995]. This penetration necessarily implies an interhemispherical current [Stenbaek-Nielsen and Otto, 1997]. Recall that the convection across the open/closed boundary, the ionospheric conductivity, and the interhemispheric current must all be self-consistent [Atkinson and Hutchinson, 1978]. Kozlovsky et al.  produced a simple model to predict an IMF By driven shift of the CRB with respect to the OCB. That predicted shift agrees in sense and approximate magnitude with the observed shifts in the afternoon sector.
 Because of the overall general agreement between the particle and convection maps, it seems useful to carry the comparison further and calculate the amount of ionospheric potential crossing the particle open-closed boundary as a function of MLT. Table 1 shows the resulting estimates of closed-open (on the dayside) and open-closed (on the nightside) merging and reconnection voltages.
Table 1. Rate at Which Magnetic Flux, in the SuperDARN Model, Crosses the Open-Closed Boundary, Determined From the Particle Data, as a Function of MLT and IMF
Closed to Open Flux Rate, kV
Open to Closed Flux Rate, kV
Bz < −1; By < −3
Bz < −1; By > 3
Bz > 1; By < −3
Bz > 1; By > 3
Table 1 raises several additional issues. One point of comparison between the particle and boundary convection patterns lies in the behavior of the mantle, which the following section will show agrees nicely. Section 4.3 will discuss how the merging potential relates to the cusp.
4.2. Asymmetric Mantle
 For both signs of Bz, and, significantly, for both signs of By, the region over which significant mantle precipitation can be detected is latitudinally wider prenoon than postnoon. This behavior of the mantle differs from the cusp, which can be found prenoon (in the Northern Hemisphere, to which the data are normalized) for By < 0 and postnoon for By > 0. Since magnetic tension drives convection across noon, one might expect the mantle to be found prenoon for By > 0, and postnoon for By < 0 [cf. Cowley, 1981; Xu et al., 1995].
 The asymmetry of the particle region of mantle precipitation presumably arises from asymmetries in the convection pattern. Field lines in the polar cap which have been open for several minutes have only (typically weak) polar rain precipitation, since the high-altitude end convects far down the magnetotail, and the ions do not have enough thermal velocity to overcome the tailward directed bulk flow velocity. Therefore let us consider the abundance of relatively freshly merged field lines around noon. The rate at which magnetic flux crosses the open-closed boundary (dΦ/dt) equals the merging voltage. From Table 1 and for southward IMF and negative By, the merging potential prenoon totals 30 kV, while the potential postnoon totals 20 kV. Likewise, for southward IMF and positive By, the merging potential prenoon still totals 30 kV but the potential represent flux crossing postnoon is 15 kV. Thus there is consistently more freshly opened flux predicted from the combination of the SuperDARN patterns and the DMSP precipitation maps for prenoon than postnoon.
 It is historically typical to separate the potential voltages somewhat differently, as belonging either to the morning cell or afternoon cell. Although that traditional approach is doubtless the right one for certain problems, the correct way to understand the mantle asymmetry involves the actual local time at which the open-closed field line is crossed. Whether or not mantle precipitation exists at a given site depends simply on whether a field line at that spot recently crossed the open-closed boundary (and to a lesser extent, how far downstream the high-altitude end mapped at the time of the merging event). The asymmetry between mantle precipitation prenoon and postnoon arises precisely because all of the morning cell crosses the open/closed boundary prenoon but not all of the afternoon cell crosses the open/closed boundary postnoon.
 We regard the agreement between the asymmetry in the particle precipitation in the mantle and the convection patterns as particularly satisfying. The particle maps, taken alone, or the convection patterns, taken alone, would doubtless raise concerns. Together they provide powerful evidence that a basic aspect of solar wind-ionosphere coupling is correctly understood.
4.3. Merging Potential and the Cusp
 In the classic theoretical picture of the ionosphere, all dayside merging occurs within the cusp, which also coincides with the convection throat [e.g., Siscoe and Huang, 1985]. Indeed, the entire frontside magnetopause maps into the magnetic cusp so that even if merging is extensive over the dayside magnetopause, and it likely is [Crooker, 1979; Siscoe et al., 2001], the ionospheric footprint was traditionally expected to lie within the cusp. The combined particle and boundary maps differ markedly from these simplifying first approximations.
 Indeed, the cusp seems to have one longitudinal boundary coinciding with the middle of the convection throat. Thus for Bz and By, both negative, the cusp lies entirely within the dawn convection cell, with the eastward boundary of the cusp corresponding to the eastward boundary of the dawn cell. Likewise, for Bz < −1 nT and By > 3 nT, the cusp lies entirely within the westward cell, with the westward boundary of the cusp and the dusk cell coinciding. In general, the cusp lies within the rounder cell for both signs of By, certainly for southward IMF and less certainly for northward IMF also. The cusp longitudinal boundary closer to noon coincides with the boundary of the rounder cell.
 Recently, the region of weak magnetic field lines along the magnetopause has been shown to be much more extended than the high-altitude and high-latitude magnetic indentation originally identified as the magnetic cusp. This more extended region of weak magnetic fields, called the “sash,” has been shown by Siscoe et al.  to extend away from the magnetic cusp toward the nightside in a broad fanlike manner. Siscoe et al.  report that the sign of the IMF By determines whether the sash extends in the dawnward or duskward direction away from noon (with the sash extending duskward for positive By in the Northern Hemisphere). It may be that the observation that the particle cusp lies entirely within one convection cell (the rounder one, with most of the merging potential) owes to the behavior of the sash. It is not unreasonable to suppose that a ribbon of weak magnetic fields (with presumably greater merging rates) would correspond to the region of greater magnetosheath plasma entry and hence to the particle cusp. Some high-altitude observations support such a scenario [Maynard et al., 2001]. Nonetheless, the fact remains that the ionospheric results reported here imply merging occurs also away from the sash (i.e., on both sides of noon).
 The amount of merging potential associated with the cusp itself is fairly small, amounting to ∼17 kV (12 kV) for Bz < 0 and By < 0 (By > 0) and perhaps 4–5 kV for northward IMF. This amounts to a relatively small fraction of the statistical dayside merging potentials of around 45–50 kV for southward IMF and 22 kV for northward IMF. Thus only ∼25–35% of the total dayside merging is directly associated with the cusp, with the higher number appropriate to southward IMF and the lower to northward IMF.
 These results imply that merging is active over much of the dayside magnetopause. The particle cusp simply consists of that fraction of the merging wherein the high-altitude end maps to a region favoring the easy entry of magnetosheath particles, with access to the ionosphere. Such entry may be favored by weaker magnetic fields along the magnetopause (i.e., the “sash”). Certainly, the conditions required do include being within the region of slow flow on the frontside of the magnetopause, for which thermal velocities exceed bulk flow velocities and hence access to the ionosphere is relatively unimpeded [Reiff et al., 1977; Wing et al., 1996].
4.4. IMF Control of Nightside and Dayside Merging
 In calculating the merging rate as a function of MLT, only the net flux crossing the open-closed boundary as determined from the particle data is counted here. Thus the loops entirely in the polar cap in Figures 1c and 1d or entirely below the open-closed region as in Figure 1c do not contribute. Likewise, in Figure 1d, 3 kV of flux is both opened and closed while staying within the 0300–0600 MLT sector. This loop also makes no net contribution.
 It is well known that By controls whether the dawn or dusk cell is larger and which is rounder or more crescent-like. It is likewise known that the sign of By strongly influences nightside convection patterns [Ruohoniemi and Greenwald, 1995]. The behavior of merging as a function of MLT is not as well established, primarily because it requires knowing where flux crosses the open-closed boundary, which requires simultaneous particle and convection data. Table 1 shows that the effects of By on the MLT dependence of flux opening is actually stronger on the nightside than on the dayside. Specifically, for southward IMF and By < −3 nT, 19 kV are reconnected premidnight versus 31 kV postmidnight. Conversely, for southward IMF and By > 3 nT, 31 kV are reconnected premidnight and just 14 kV postmidnight, a reversal of the relative importance of the two MLT sectors. By contrast, on the dayside, for southward IMF and either sign of By, 33 kV are merged prenoon, dominating the postnoon merging.
 It is known that ionospheric conductivity places constraints on the convection patterns [Atkinson and Hutchinson, 1978], since the field-aligned currents into the ionosphere must be consistent with the electric fields across the open-closed boundary. We do not know of a model that incorporates realistic conductivities with realistic field-aligned currents to make realistic predictions for the convection patterns. Nonetheless, it is likely that the discrepancy between the strong effect of By on the nightside merging rate as a function of MLT, versus the relatively weak effect of By on the dayside merging rate as a function of MLT, owes to ionospheric conductivity effects.
5. Summary and Conclusions
 The convergence of the entire magnetosphere, with its various plasma populations, along magnetic field lines to thread just a few million square kilometers in the high-latitude ionosphere has long intrigued researchers. Potentially, the dynamics of the solar wind-magnetosphere-ionosphere interactions can be studied with economy. Establishing the manner in which plasma and magnetic flux flows between regions is one step in such a program of study, and the purely statistical relationship shown here is one aspect of that first step.
 When ionospheric convection maps and particle precipitation maps are compiled within the same coordinate system (in an inertial reference frame) and under the same set of IMF conditions, the comparison is revealing. Only ∼25% (35%) of the dayside merging is associated with the northward (southward) IMF cusp, which lies not in the convection throat but entirely within the rounder convection cell and which has a longitudinal boundary approximately coinciding with the separatrix between merging cells. Merging is thus present throughout the dayside magnetopause, not just in the region of slow magnetosheath flow that produces cusp precipitation.
 Merging is unequally distributed about noon, with 60–65% of the magnetic flux crossing the open-closed boundary before noon regardless of the sign of By. The low-altitude precipitation shares the convection skew so that mantle and intense polar rain precipitation is also preferentially found prenoon. The particle and convection maps thus in this aspect nicely corroborate one another.
 On the nightside, however, By controls whether reconnection occurs mainly premidnight or postmidnight. In the Northern Hemisphere, more magnetic flux is reconnected premidnight for positive By and more postmidnight for negative By.
 The overall agreement between the particle and convection maps is excellent. The equatorward boundary of significant ionospheric flows essentially matches the equatorward boundary of auroral precipitation. The significant exception is the CPS precipitation on the dayside for northward IMF, which reaches to latitudes below appreciable ionospheric flows. This dayside CPS precipitation is produced by drifting particles from the nightside over a timescale of several hours and thus has a much slower dynamic timescale than does the convection data.
 The poleward boundary of the nightside oval agrees with the CRB over the entire nightside. On the dayside, there is a general agreement between the region of magnetosheath origin precipitation and the region of antisunward flow but not as simple and complete as for the nightside. The CRB lies equatorward (poleward) of the OCB in the postnoon (prenoon) region for By < 0. For By > 0, the CRB lies poleward (equatorward) of the OCB in the postnoon (prenoon) region. Thus a systematic shift occurs around noon based on the sign of the IMF By component. The postnoon shift was previously noted in case studies by Kozlovsky et al. , who attributed the effect to an interhemispheric current originating from the partial penetration of the IMF By. This seems to us to be the most promising candidate explanation for the overall systematic shifts observed here.
 The boundary-oriented precipitation maps shown here accent more clearly a precipitation trend seen less sharply in position-oriented models (binned first by fixed MLAT/MLT grids). The electrons and ions are oppositely affected by curvature and gradient drifts, and by the presence of field-aligned potentials (i.e., a potential which accelerates magnetospheric electrons will deenergize magnetospheric ions). Thus a clear pattern of anticorrelation exists between regions of strong electron and ion precipitation, such as the tendency in the premidnight sector for electron precipitation to maximize in the poleward portion of the oval (where discrete arcs are common) and the ions to maximize in the equatorward portion (where curvature and gradient drifts preferentially transport ions but not electrons).
 This work was supported by NSF grant ATM-0227481 to the Johns Hopkins University Applied Physics Laboratory. F. Rich of AFRL at Hanscom AFB has been generous and instrumental in enabling our DMSP data analysis for many years. JMR wishes to acknowledge funding from NSF grant ATM-9819891 and NAG-10902. Operation of the SuperDARN radars in the Northern Hemisphere is supported by the national funding agencies of the US, Canada, UK, France, and Japan.
 Arthur Richmond thanks Alexandre Koustov and Frederick J. Rich for their assistance in evaluating this paper.