This note describes the building and contents of several 1min and 5min resolution, solar wind magnetic field and plasma data sets timeshifted to the Earth's bow shock nose. Data from the ACE, Wind and IMP 8 spacecraft were processed in 20056, while Geotail data were added later, in 2007. Initially the data were for 1995 to nearcurrent. In 2009, the IMP 8 shifted data were extended back in time to 11/04/1973, shortly after launch. Also in 2009, we added GOES fluxes of protons above 10, 30 and 60 MeV to 5min OMNI. These products are primarily intended to support studies of the effects of solar wind variations on the magnetosphere and ionosphere. In addition, we address 19982000 1min ACE data sets shifted using various techniques to the Wind location.
Time shifting is based on the assumption that solar wind magnetic field values observed by a spacecraft at a given time and place lie on a planar surface (a "phase front") convecting with the solar wind, and that the same values will be seen at a different place at the time that the phase front sweeps over that location. A key element of the time shifting is use of the phase front normal (PFN) directions, which are to be determined individually for each input 1516 sec magnetic field observation by analysis of it and its near neighbors. We identify and compare results of two distinct PFN determination analysis techniques (minimum variance and cross products) and two separate combinings of these, for a total of four shift techniques.
The family of products introduced herein consist of
(a) 1min averaged 19982000 ACE magnetic field and plasma data
shifted to the Wind location by each of the four shift techniques,
along with 1min unshifted Wind averages, with which interested
persons can make independent judgements on the relative
effectiveness of the various shift techniques,
(b) 1min and 5min averaged ACE (1998present), Wind (1995present), IMP 8
(19732000) and Geotail (19952006) magnetic field and plasma data sets
shifted to the Earth's bow shock nose,
(c) a 1min spacecraftinterspersed data set at the bow shock nose
that we call the High Resolution OMNI (HRO) data set and
(d) a 5min averaged version of HRO having GOES proton fluxes appended.
Time tags in records of all these products are targetarrival times
and not observation times.
This note addresses in sequence: (a) the input data sets and their preparations, (b) the time shifting used, including discussion of the multiple PFN determination techniques available and including consideration and handling of "outofsequence" arrivals, (c) the building of 1min averages from the shifted 1516 sec IMF values and shifted 12 min plasma values, (d) discussion of the various data sets created (spacecraftspecific and the spacecraft interspersed HRO), including their record formats and meanings of each word in the records, (e) results of analysis of the 19982000 Wind data and ACE data shifted to Wind for predictability of IMF and plasma variations at one point, given observations elsewhere, as a function of the twopoint separation vector, of the solar wind state (variation level, fast or slow, etc.), and of the PFN determination technique. In addition, a series of Appendices address (f) interspacecraft comparisons of magnetic field and plasma parameter values for finding systematic differences and parameter crossnormalization used in interspersing data from three spacecraft, (g) selection criteria for which data to use in High Resolution OMNI when data from multiple spacecraft are available for a given interval.
The table below identifies the dates for which different parameters of high res. OMNI data are available.
YYYY DDD  YYYY DDD  1981 001  1994 365 IMF (from IMP8 only) 1981 001  1994 365 Plasma (from IMP8 only)  1995 001  2015 154 IMF and Plasma (from Wind, ACE, IMP8) 1981 001  1988 366] Final AE, AL, AU indexes 1989 001  2015 090 Provisional AE, AL, AU indexes 1981 001  2015 151 Provisional SYM/D, SYS/H, ASYM/D, ASYS/H indexes 1981 001  2014 365 (PCn index) 1986 001  2015 154 ( Fluxes from Goes, for 5min res. only)  Time span of the data shifted to bow shock nose Geotail: 19950315  20061231(365) IMP8: 19731104  20000609(366) ACE: 19980205  20150219(050) Wind: 19950101  20150603(154)Proton Fluxes from GOES (>10 MeV, >30 MeV, >60 MeV ) are taken from NGDC: http://goes.ngdc.noaa.gov/data/avg/ originally, and more recently, http://satdat.ngdc.noaa.gov/sem/goes/data/new_avg/. See near end of Section 2 below for further detail.
We have used publicly available ACE, Wind, IMP 8 and Geotail magnetic field and plasma in building the 1min and 5min data products described herein.
ACE (Advanced Composition Explorer) was launched August 25, 1997, and continues to provide magnetic field, plasma and energetic particle data from a ~180 day L1 orbit having X, Y, and Z (GSE) ranges of 220 to 250 Re, 40 to +40 Re, and 24 to +24 Re. The ACE home page is at http://www.srl.caltech.edu/ACE/.
Wind was launched November 1, 1994, as part of NASA's contribution to the International Solar Terrestrial Program. It continues to obtain magnetic field, plasma, energetic particle and plasma wave data. Since mid2004, it has been in an L1 orbit with excursions in Y(GSE) between +/ 100 Re. It had multiple earlier phases, including an interval spanning the last third of 2000 through mid 2002 with Y(GSE) excursions in excess of 200 Re and an interval in late 2003 and early 2004 in orbit about the Lagrange point on the antisunward side of Earth. The Wind home page is at http://pwg.gsfc.nasa.gov/wind.shtml.
IMP 8 was launched October 26, 1973, into a low eccentricity Earth orbit. Apogee and perigee distances have been in the ranges 3845Re and 2834 Re. On average IMP 8 is out of the solar wind for about 5 days of every 12.5 day orbit. The IMP 8 magnetometer failed June 10, 2000. Data from the MIT plasma instrument and from three energetic particle detectors were acquired until October, 2006. The IMP 8 web page is at http://spdf.gsfc.nasa.gov/imp8/project.html.
Geotail was launched July 24, 1992, into an eccentric orbit with apogee deep in the geotail. In early 1995, the Geotail orbit was adjusted to about 10 x 30 Re, and then to 9 x 30 Re in 1997 where it continues today (2008). In this orbit, Geotail has annual solar wind "seasons" with apogee local times on or near the Earth's dayside, and it has solar wind intervals during each ~5 day orbit of the solar wind seasons.
The web pages for the contributing investigations are: Magnetic field ACE: http://www.srl.caltech.edu/ACE/ASC/level2/index.html Wind: http://wind.gsfc.nasa.gov/mfi/ IMP 8: http://wind.gsfc.nasa.gov/imp8/ Geotail: http://www.stp.isas.jaxa.jp/geotail/ Plasma: ACE: http://swepam.lanl.gov/ Wind: http://web.mit.edu/space/www/wind/wind.html IMP 8: ftp://space.mit.edu/pub/plasma/imp/www/imp.html Geotail: http://wwwpi.physics.uiowa.edu/www/cpi/ The input data were pulled from: Magnetic field ACE: (ACE Science Center): http://www.srl.caltech.edu/ACE/ASC/ Wind: ftp://spdf.gsfc.nasa.gov/pub/data/wind/mag/3sec_ascii/ IMP 8: ftp://spdf.gsfc.nasa.gov/pub/data/imp/imp8/mag/15s_ascii_v3/ Geotail: http://cdaweb.gsfc.nasa.gov/ (GE_EDB3SEC_MGF) Plasma ACE: (ACE Science Center): http://www.srl.caltech.edu/ACE/ASC/ Wind: ftp://spdf.gsfc.nasa.gov/pub/data/wind/plasma_swe/swe_kp_unspike/ IMP 8: ftp://spdf.gsfc.nasa.gov/pub/data/imp/imp8/plasma_mit/sw_msheath_min/ Geotail: http://cdaweb.gsfc.nasa.gov/ (GE_H0_CPI) Key persons for these data sets include: Magnetic field: ACE: Chuck Smith (UNH), Norman Ness (Bartol) Wind: Ron Lepping (GSFC), Adam Szabo (GSFC), Norman Ness IMP 8: Adam Szabo, Ron Lepping, Norman Ness Geotail: Tsugunobu Nagai (Tokyo Inst. Tech.), S. Kokobun Plasma: ACE: Dave McComas (SWRI), Ruth Skoug (LANL) Wind: Al Lazarus (MIT), Justin Kasper (MIT), Keith Ogilvie (GSFC) IMP 8: Al Lazarus, John Richardson (MIT) Geotail: Bill Paterson (Hampton U.), L. Frank & K. Ackerson (U. Iowa)
ACE magnetic field and plasma data
"Level 2" 16s magnetic field data and 64s plasma data were pulled from the ACE Science Center. (Credit goes to Andrew Davis and the ASC team for a very effective data management and distribution facility). The field and plasma data there start on September 2, 1997, and February 5, 1998, respectively. Owing to the critical need for plasma flow speed data in time shifting magnetic field data to the bow shock nose or elsewhere, we limit the coverage of ACE data in our new data products to February 5, 1998, and later.
Wind magnetic field data.
The Wind magnetic field data are standardly produced by the instrument team at 3s, 1m and 1h resolutions. Because we apply phase front normal determination algorithms to 15.36s IMP magnetic field data and to 16s ACE data, we form 15s averages from the available 3s data to have similarly resolved Wind magnetic field data as input
The Wind magnetic field data are standardly available at 3sec resolution with no discrimination for orbit phase, in particular, for solar wind vs. nonsolar wind phases. We have filtered at hourly resolution the time continuous 3sec data against the Wind bow shock crossing identifications made by the Wind magnetometer team and available at http://wind.nasa.gov/mfi/bow_shock.html to give a solarwindonly input data set. We have made our own identifications of the few crossings that occurred after the October, 2003, end of the Wind team's list.
In October 2011, the Wind/MFI team finished the reprocessing of all MFI data. Among other things, welldetermined Bz offset values were used. The new MFI data were inserted into High Resolution OMNI when they became available, replacing the earlier MFI data. The new data were used to redetermine solar wind phase front normals used in shifting data.
There are rare spikes in the Wind magnetometer data. We have taken a simple approach to eliminating most of these by rejecting any 3sec record with a magnetic field magnitude or component in excess of 70 nT.
Wind/SWE plasma data
Wind/SWE plasma parameter data are available at ~92s resolution in three versions corresponding to three approaches to their production from underlying distribution functions. There are "key parameter" data, nonlinear fitsbased data (fits assumed convecting bimaxwellian distributions), and anisotropic momentsbased data. These are discussed at the MIT Wind/SWE web page cited above. The latter two are further discussed in Justin Kasper's dissertation whose most salient parts are webaccessible at ftp://spdf.gsfc.nasa.gov/pub/data/wind/swe/ascii/2min/thesis.pdf. Finally, "physicsbased" tests of the goodness of the nonlinear fits (NLF)based velocities (~0.16% in speed, ~3 deg in direction), densities (~3%)and temperatures (~8%) are discussed in Kasper et al. (2006).
The NLF data and the anisotropic momentsbased data are available to within several weeks of the current date, date of availability of Wind magnetic field version 4 data. The SWE KP data, on the other hand, are typically available to within several weeks of the current date. Given this and given the urging of the MIT plasma team to use the very good and more robust KP data, we have chosen to use the KP data in our high resolution OMNI data set.
But given that it was the NLF data for which the relatively small uncertainties cited above were determined, we shall normalize the KP density and temperature values to equivalent NLF values in the spacecraftinterspersed HRO data set. This point is further discussed and quantified in Appendices 1 and 2 addressing comparisons and crossnormalizations of the available multispacecraft data.
As for the Wind magnetic field data, the SWE KP data are available with no discrimination for orbit phase. We have extracted a solar windonly set of SWE KP data by again filtering at hourly resolution against the Wind bow shock crossing identifications cited above.
The SWE KP data are initially computed and loaded to CDAWeb. The SWE team at MIT improves this product by passing it through a despiking routine that compares a value with the median of three points (the point being tested and its immediate predecessor and follower). Some spikes elude detection. We have run a further despiking routine requiring (to be a nonspike) that the difference between a parameter value and the mean of the two preceding and two following values should be less than four times the standard deviation in that mean or that that difference relative to the mean should be less than some (parameterdependent) value. This is further discussed in Appendix 3.
IMP 8 magnetic field data
IMP 8 magnetic field data have long been available at 15.36 sec resolution (cf. ftp://spdf.gsfc.nasa.gov/pub/data/imp/imp8/mag/15s_ascii_v3/) in a data set that makes no distinction between the solar wind and nonsolar wind phases of the IMP orbit. We have used the IMP 8 bow shock crossing information at http://omniweb.sci.gsfc.nasa.gov/ftpbrowser/bowshock.html to separate, at 1minute resolution, the solar wind and non solar wind phases of the IMP orbit to ensure that only IMP 8 solar wind magnetic field data are included in the products described herein.
There are occasional data spikes in the 15.36 sec data. We have hopefully eliminated most if not all of these by applying the spike finder software discussed in Appendix 3.
As noted above, the IMP data in the products discussed in these notes run only to the June 10, 2000, failure of the IMP 8 magnetometer.
IMP 8 plasma data
Plasma parameters from the MIT Faraday cup are available at ~1 min resolution (cf. http://omniweb.sci.gsfc.nasa.gov/ftpbrowser/imp_mit_min.html). Parameter values as determined both by nonlinear fitting to assumed convecting Maxwellian distributions and moments are available. As in earlier work, we use the nonlinear fitbased data, as these are believed by the MIT team to be the more reliable. (Note that readers may compare the fitbased and momentbased parameter values using the interface at http://omniweb.sci.gsfc.nasa.gov/ftpbrowser/imp_mit_min_s.html
This data set has data from both the solar wind and magnetosheath phases of the IMP 8 orbit. However, each record has an MITassigned flag indicating whether the data definitely are, or are not, from the solar wind, or whether they may be from solar wind or magnetosheath. We have used this flag to eliminate from the products discussed in this documentation any data not tagged as being definitely in the solar wind.
There are some spikes in the IMP 8 plasma data. To eliminate most of these, we have applied the spike finder software discussed in Appendix 3 to the data. However, because the software assumes that the first two and last two data points of every interval not having a data gap in excess of one hour are good data, we have visually scanned plots of data after the application of the spike finder software, and have identified and eliminated a few extra points as being likely bad points.
The IMP 8 plasma flow elevation angle has long been recognized as having a ~2 deg offset. This is further discussed in Appendix 1. We have not taken this bias out of the data of the products discussed herein.
Geotail magnetic field and plasma data
First, we created 15s averaged magnetic field averages from 3sec values for input
compatibility with ACE, Wind and IMP IMF data used. Second, we determined the principal
time intervals during which Geotail was beyond the Earth's bow shock, in the solar wind.
This process, which does not distinguish foreshock intervals from nonforeshock solar wind
intervals, is extensively discussed at
ftp://spdf.gsfc.nasa.gov/pub/data/geotail/merged/sw_min_merged/00readme
Our despiking of Geotail magnetic field and plasma data is also described in this readme file.
The despiked, 15s, solarwindonly magnetic field data set is accessible from
http://omniweb.gsfc.nasa.gov/ftpbrowser/geotail_mag15s.html
We used CPI plasma data rather than Geotail LEP plasma data as the former seemed to
have cleaner solar wind parameter values and were more immediately accessible to us.
CPI despiked plasma data also available at http://omniweb.sci.gsfc.nasa.gov/ftpbrowser/geotail_pla_cpi.html
As we were doing this work, the magnetometer PI team was working to reprocess its data using more definitive Bz offset values. As of this date (February 5, 2008) we had not received reprocessed data. So we have done our Bz corrections using the expectation that, when averaged over a year, the Bz component in geocentric solar ecliptic coordinates should be within 0.1 of zero. It is possible that one day our new Geotail data sets and multispacecraft OMNI data sets will incorporate the notyetavailable reprocessed data of the PI team.
GOES energetic proton fluxes
Fluxes of protons above 10, 30 and 60 MeV, as measured by NOAA's geosynchronous GOES spacecraft are included in 5minute OMNI. Data from the following spacecraft were used for the indicated years: GOES 7, 1995; GOES 8, 19962002; GOES 10, 2003; GOES 11, 20042010; GOES 13, 2011 and later. Data are as taken from http://satdat.ngdc.noaa.gov/sem/goes/data/new_avg/ except that for GOES 13, where separate fluxes are given at NGDC for eastward and westwardlooking sensors. For GOES 13, we have averaged these two fluxes for inclusion in 5min OMNI. To view separate eastward and westwardlooking fluxes, and their ratios, see the FTPBrowser interface at http://omniweb.sci.gsfc.nasa.gov/ftpbrowser/goes13_flux_5m.html Principal Investigator for the GOES energetic particle instruments is currently T. Onsager, and key responsible NGDC person is D. Wilkinson.
Extra notes
Data providers may occasionally create replacement versions of their data. In such cases, we replace the superseded data in OMNI with the newer data values, and typically make note that this has happened at http://omniweb.gsfc.nasa.gov/html/ow_news.html. Such changes are relatively rare are typically involve only small parameter value changes.
We sometimes refer to "15s" input magnetic field data throughout these pages. Readers should appreciate this is a shorthand notation for 16s ACE data, 15s Wind and Geotail data and 15.36s IMP data.
To best support solar wind  magnetosphere coupling studies, it is desired to timeshift solar wind magnetic field and plasma data from their location of observation, which may be an hour upstream of the magnetosphere and several tens of Re or more removed from the Earthsun line, to a point close to the magnetosphere. We choose this point to be the bow shock nose. In addition, to assess the goodness of such shifts, we separately shift ACE data to Wind (by each of several shift techniques) and compare the shifted ACE data and in situ Wind data.
Given the availability of data on a specific solar wind magnetic field or plasma parameter P as a function of time t at the location Ro of an observing spacecraft, i.e., P(t, Ro), it is desired to infer values of this parameter at some displaced location Rd, i.e., P(t', Rd). The key underlying assumptions enabling estimation of the time shift, deltat = t't, between observation of the parameter at Ro and t and arrival of this value/variation at Rd at t', is that solar wind variations are organized in series of phase fronts (flat planes) that convect with the solar wind velocity V. Curvature of variation surfaces is ignored and propagation of these phase fronts relative to the solar wind flow is ignored. The unphysical interpenetration of these phase fronts is discussed later. Thus the time shift equation is deltat = n · (Rd – Ro) / n · V, where n is the variation phase front normal (PFN) and where “·” is the normal dot or scalar product of two vectors.
The target Rd to which we shall shift ACE, Wind, IMP 8 and Geotail data is the bow shock nose. This will best support future solar wind  magnetosphere coupling studies. We use the field and plasma parameters determined at a given time, and the bow shock model of Farris and Russell (1994) with the magnetopause model of Shue et al (1997), to determine where the bow shock will be when the phase front reaches it. See Appendix 4 for a discussion of these models. We include solar wind flow aberration associated with Earth's ~30 km/s orbital motion about the sun in bow shock nose location determination.
It is recognized that this is a very simplified approach, neglecting finite response times of the magnetosphere to solar wind variations, that may introduce some error. However, except for extreme excursions in solar wind parameters, the bow shock will not move enough to introduce significant uncertainty in the timing of arrival of solar wind structures observed upstream. (Uncertainties connected with other factors such as planarity of features and the interpenetration of variation phase planes are larger and affect the parameter profiles and not merely the timing of arriving plasma.)
The bow shock location to which the data are shifted is included in the output data records, among many other parameters.
In addition to shifting data to the bow shock nose, we shall also shift ACE data, by each of four techniques, to the location of the Wind spacecraft so that we can assess the predictability of solar wind variations as a function of the shift technique, the observertarget separation geometry, the variation level in the solar wind, and the nature of the flow (e.g., fast vs. slow).
There was no shifting of GOES energetic particle fluxes.
3a. Determination of the phase front normals (PFN)
Minimum variance analysis (MVA) has long been used to determine normals to discontinuity planes in the solar wind magnetic field. See for example Sonnerup and Cahill, 1968. In this approach, a 3x3 variance matrix
Mij = Sum (Bi(t)*Bj(t))/N – Sum Bi(t) * Sum Bj(t)/N**2 = <Bi*Bj>  <Bi>*<Bj>
is formed, with averages taken over a set of N points spanning the discontinuity and with i,j representing any two spatial directions. The matrix is diagonalized, and the eigenvector associated with the minimum eigenvalue gives the minimum variance direction (MVD). The number of points N to be used in the analysis, and the ratio of intermediate to minimum eigenvalues to take as a lower limit below which the MVD is considered not reliably determined, are part of the “art” of MVA.
3a.1. Technique 1, "Modified" MVA
Weimer et al (2003) applied the basic concepts of MVA to determine an MVD for each point of a continuous time series of interplanetary magnetic field data. In effect, they assumed each point lay on a planar phase front whose normal could be used, along with the solar wind flow velocity, in the determination of when that value (assumed constant everywhere on the plane) would be seen elsewhere in space.
After determining surprisingly good correspondence of timevarying time shifts thus determined with shifts determined by multispacecraft analysis (e.g., Weimer et al, 2002), an error was discovered in the Weimer et al. (2003) application of MVA. In particular, the 1/N**2 in the expression above was inadvertently replaced by 1/N. When the correct expression above was used, agreement with the multispacecraft time shift determinations deteriorated.
Shortly thereafter, Bargatze et al. (2005) demonstrated that the MVA equations used in Weimer et al (2003) corresponded approximately to an MVA constrained by the condition that the mean magnetic field vector over the analysis interval should lie in the plane of minimum variance, that is, that <B>·n (n is the MVD) ~ 0. The Weimer et al (2003) came to be known, at least on a limited basis, as Modified MVA.
Much of our early work in this twoyear effort utilized the Weimerprovided code used in his 2003 analysis. None of the final products made available from our effort are based on this technique, although some interim products, no longer available, were.
3a.2 Technique 2, MVAB0
A Comment by Haaland et al (2006) pointed out that MVA exactly constrained by the <B>·n = 0 condition was first used by Sonnerup and Cahill (1968) and has been discussed by Sonnerup and Scheible (1998). Such an MVA, called MVAB0 by Haaland et al, diagonalizes not the matrix M (see above), but the matrix P*M*P where the symmetric matrix P (P_{i}_{j} = delta_{i}_{j} – e_{i}e_{j}; delta_{i}_{j} is kronecker delta and e is the unit vector in the direction of the mean magnetic field) projects each vector B onto the plane perpendicular to e.
Weimer has developed and provided to us new code that correctly implements the MVAB0 approach.
We have used the MVAB0 code generously provided by Weimer in mid 2006. It is the only MVA code used in our final products.
Weimer spent significant effort determining parameters for the MVAB0 technique, by seeking parameter sets whose results gave best agreement with multispacecraft determinations of phase front normals. In particular, he found, and we have used, for the MVAB0 technique optimal results with 77 15s points in each analysis (~19 min spans for each MVD determination), eigenvalue ratio greater than or equal to 5.2 (for a reliable MVD determination), and angle between MVD and solar wind flow vector less than 73 deg. (Larger angles lead to excessively long predicted delays.)
To eliminate spurious PFN determinations associated with data gaps, we added the requirement that the interval between the first and last point involved in each PFN determination should be no more than 1.25 times what it would be in the absence of data gaps.
3a.3 Technique 3, Cross Product (CP)
A totally distinct approach to determining a phase front normal, that should be perfect for an ideal tangential discontinuity, is to take a cross product of magnetic field vectors just prior to, and following, a discontinuity. Weimer has developed code that determines phase front normals continuously using the cross product concept and has generously also provided this to us. In a private communication to us, Weimer cites the work of Knetter et al (2004)as the inspiration for developing this cross product (CP) code.
Weimer also spent significant effort determining parameters for the CP technique, by seeking parameter sets whose results gave best agreement with multispacecraft determinations of phase front normals. In particular, he found, and we have used, for the CP technique optimal results with the angle between the “before” and “after” vectors greater than 13 deg, that these vectors should be based on 17 points each, centered on the points 14 points before and after the point for which the PFN is sought (thus a span of 46 points, or ~12 mins, for the PFN determinatioin for each point), and that the component of the mean field vector normal to the phase front should be less than 0.035 nT. He also used a 73 deg limiting angle as for the MVAB0 technique. As for the MVAB0 technique, we added the requirement that the interval between the first and last point involved in each PFN determination should be no more than 1.25 times what it would be in the absence of data gaps.
3a.4 Technique 4, JK/NP Combination of Techniques 2 and 3.
Now, having two fundamentally different techniques for PFN determination, we are able to add combinations of these two. We devised one, called Technique 4, which is the one we in fact used for producing the bow shock noseshifted products discussed in these notes. The technique consists of first applying the CP method for a given point and its relevant neighbors, if an acceptable PFN is determined, this is used for this point. If CP does not produce an acceptable PFN (e.g., if the included angle between the “before” and “after” vectors is less than 13 deg), then the MVAB0 technique is applied and its resultant PFN, if acceptable, is used. If neither CP nor MVAB0 techniques produce an acceptable PFN, that point is marked for later interpolation, and a PFN is attempted for the next point in the time series.
3a.5 Technique 5, DW Combination of Techniques 2 and 3.
Weimer and King (2008) took an alternative approach and required that both the CP and MVAB0 techniques should produce the same PFN (to within some accuracy, arbitrarily set at 5 deg) in order to be acceptable, otherwise the point was marked for later interpolation. Weimer has provided the code implementing this technique, which we call Technique 5.
In all cases (Techniques 25), a PFN direction satisfying relevant tests may or may not be determined. Typically, such points are marked. Then, in a second pass, for each such point, a PFN is determined by linear interpolation between the last good and next good PFN. In our implementations, the span across which such interpolations are made can be no longer than 3 hours. Data belonging to such extended gaps are not shifted nor included in our new data products.
We hope to modify this in the future, as an IMF that was not varying over many hours would be highly predictable at the bow shock nose yet would not lead to acceptable PFN's and hence would not be "shifted" and included in our products. We have searched the interval MarchDecember, 1998, for such occurrences, and find 45 days with multihour data gaps in shifted data despite there being no gaps in the input ACE data. The average gap duration is 46 hours, so the fraction of data lost in our shifted data set is about 45*6 / 300*24 = 4%. Fortunately, this is when the IMF is most quiet and accurate bow shock nose predictions least critical.
Technique 5, the DW combination of 2 and 3, involves more interpolation of PFNs than the individual MVAB0 or CP technique, or than the JK/NP combination thereof, which is one of the main reasons we did our production work with Technique 4. In the same search of MarchDecember, 1998, data mentioned in the preceding paragraph, we found 60 days having intervals of 3 hours or more having Technique 4 data but not Technique 5 data (because no good PFN's were produced over such intervals by Technique 5.) Again assuming an average 6hour gap duration, the fraction of time for which we do not have Technique 5 data relative to the time for which we have Technique 4 data is 60 * 6 / 0.96 * 300 * 24 = 5%
Interestingly, which technique is used does not have a statistically significant affect on the profiles, as will be further discussed later.
3b. Mechanics of the time shifting
We introduced above the time shift equation as deltat = n·(Rd – Ro) / n · V. n is the phase plane normal, determined by analysis of magnetic field data only. V is the solar wind velocity, including the ~30 km/s in the Ygse direction associated with the Earth’s orbital motion about the sun. We initially shift 15sec magnetic field data using the vector velocity determined by interpolating velocity values most immediately preceding and following the time tag of the observed magnetic field value, as long as the interval of interpolation is less than one hour. Magnetic field data points whose most immediately preceding and following velocity data are separated by more than an hour are not carried forward into our output data products.
Shifting means changing the time tags of data records. There is no changing of observed parameter values in the process.
After shifting magnetic field data, plasma data are shifted by using the time shift duration associated with the magnetic field observation whose preshift time tag lies closest to the plasma record’s time tag, so long as two time tags lie within 2 minutes of each other.
Because the n and the V in the time shift equation vary at various time scales, it sometimes happens that, if phase front A is observed before phase front B, B may nevertheless be predicted to arrive at a remote location (e.g., the BSN location) before A arrives there. Such outofsequence arrivals may be due to “overtaking” associated with speed gradients or to “interpenetration” of variously oriented phase planes (especially given a significant separation of the locations of the BSN and of the observing spacecraft in the direction normal to the solar wind flow).
This “interpenetration” is clearly unphysical and is one of the primary shortcomings in our work. But there is no physically justified alternative yet. Two different alternatives have been considered. In Weimer’s earliest work, he imagined that, for any pair of outofsequence phase fronts, the latterarriving phase front would be precluded from arriving by the earlier arriving phase front and so could be dropped from further consideration. In more recent work, he imagined that the latter arriving phase front would displace the earlier arriving phase front, so that the earlier arriving phase front could be dropped from further consideration.
Our sense is that, while we cannot specify the physical processes that will occur and prevent interpenetration and outofsequence arrivals, there is no good a prior reason for favoring earlierarriving or laterarriving phase fronts in cases of outofsequence arrivals. As such, our approach is to accept all shifted data as belonging to the newly assigned time tags that each record acquires via our simple time shift equation, and to build 1min data products with averages over all points shifting into a given minute. We recognize this involves an unphysical mixing of plasma elements from differing domains. But in some sense it emulates our ignorance of the dynamical processes that happen in the real solar wind.
There is a parameter in the output 1min data records, the duration between observing times (DBOT), whose negative values indicate that outofsequence arrivals have occurred.
[Note added 01/22/2007. It should be recognized that occasionally our approach to averaging over all data shifting into a given minute leads to a series of minutes whose parameter values alternate between those characteristic of different plasma domains. That is, each minute average may not simply be an average of values from two domains, especially for Wind plasma data which starts at 92s resolution. For a recent example of this, see 2007/10/25 Wind SWE plasma data prior to shifting at http://omniweb.sci.gsfc.nasa.gov/ftpbrowser/wind_swe_kp.html, and see corresponding shifted data at http://omniweb.gsfc.nasa.gov/form/sc_merge_min1.html. There is a clean interplanetary shock in the unshifted data at 10:44 UT, while there is an interval of ~50 minutes duration, spanning 11:47  12:36, of shifting between preshock and postshock parameter values in the minute averages built from shifted data. Users must exercise care in using spacecraftspecific, bowshocknose shifted data, or High Resolution OMNI data created from them, in the presence of significant variability in field and plasma parameters and in derived phase front normal directions.]
4. Descriptions of the new data products.
This section describes the common format of (a) the 1min ACE, Wind, IMP 8 and Geotail spacecraftspecific data sets that have been created at the bow shock nose, (b) the ACE data sets shifted by various techniques to the location of Wind and (c) the unshifted Wind data. It also describes the shared format of the 1min and 5min spacecraftinterspersed OMNI data sets.
The 1min field and plasma averages are built from 15s magnetic field and ~1min plasma records whose shifted time tags indicate that any portion of the data underlying the parameter values (i.e., the higher resolution field values from which the 15 sec field averages were determined or the plasma spectra from which the bulk plasma parameters were determined) were observed during the relevant minute of interest. See Appendix 5 for a more discussion of the averaging, including the weighting used.
The 1min time tags are at the start (not midpoint) of the data used in the average.
The 5min OMNI averages are built from the five relevant 1min averages. The standard deviations in these averages correspond to the process of building the 5min averages and do not retain knowledge of the standard deviations in the 1min averages.
Identification of spacecraft and of shift technique, for the spacecraftspecific data sets, are captured in file names rather than in data records. To review, we use the following identifiers:
Spacecraft: ACE 71 Geotail 60 IMP 8 50 Wind 51 Shift technique: 2 MVAB0 (min variance constrained by <B> · n = 0) 3 Cross Product 4 Mixed  use PFN(3) if good, otherwise use PFN(2) 5 Mixed  use PFN(3) = PFN(2) only if they agree Identification of source spacecraft for field and plasma data in OMNI is contained in data records, using above spacecraft ID's.Only shift technique 4 is used in the spacecraftspecific data sets shifted to the bow shock nose and in the HRO data set created from them, while each shift technique is used in the ACE data sets shifted to Wind.
4a. Spacecraftspecific data sets
The common format for the spacecraftspecific data sets is as follows. Of the 37 words, words 815, 2328, and 3234 are 1min averages formed over nativetimeresolution data.
Word Format Comment Year I4 1995 ... 2006 Day I4 1 ... 365 or 366 Hour I3 0 ... 23 Minute I3 0 ... 59 at start of average # of points in IMF avgs I4 Percent interp. I4 See footnote A below CP/MV Flag F4.1 See footnote A below Timeshift, sec I6 Phase_frnt_nrml, X,GSE F6.2 GSE components of unit vector, Phase_frnt_nrml, Y,GSE F6.2 X comp. always > 0. Phase_frnt_nrml, Z,GSE F6.2 Scalar B, nT F8.2 Bx, nT (GSE, GSM) F8.2 By, nT (GSE) F8.2 Bz, nT (GSE) F8.2 By, nT (GSM) F8.2 Determined from postshift GSE components Bz, nT (GSM) F8.2 Determined from postshift GSE components RMS, timeshift, sec I7 RMS, Phase front normal F6.2 See footnote B below RMS, Scalar B, nT F8.2 RMS, Field vector, nT F8.2 See footnote B below # of points in plasma avgs I4 Flow speed, km/s F8.1 Vx Velocity, km/s, GSE F8.1 Vy Velocity, km/s, GSE F8.1 Vz Velocity, km/s, GSE F8.1 Proton Density, n/cc F7.2 Temperature, K F9.0 X(s/c), GSE, Re F8.2 Position of spacecraft Y(s/c), GSE, Re F8.2 Z(s/c), GSE, Re F8.2 X(target), GSE, Re F8.2 Position of bow shock nose or Wind Y(target), GSE, Re F8.2 Z(target), GSE, Re F8.2 RMS(target), Re F8.2 See footnote B below DBOT1, sec I7 See footnote C below DBOT2, sec I7 See footnote C below The data may be read with the format statement: (I4,I4,2I3,2I4,F4.1,I7,3F6.2,6F8.2,I7,F6.2,2F8.2,I4,4F8.1,F7.2,F9.0,3F8.2,4F8.2,2I7)Note that for missing data, fill values consisting of a blank followed by 9's which together constitute the Ix or Fx.y format are used.
Percent interp: The percent (0100) of the points contributing to the 1min magnetic field averages whose phase front normal (PFN) was interpolated because neither the MVAB0 nor Cross Product shift techniques yielded a PFN that satisfied its respective tests (see above for these).
CP/MV flag: The fraction (01) of the points, that contribute to the 1min magnetic field averages and that are not based on interpolated PFN's, whose PFN was based on the MVAB0 method.
If in a given 1min magnetic field average, there are n points with CPbased PFN's, p points with MVAB0 PFN's and q points with interpolated PFN's, then Percent interp = 100 * q/(n+p+q) and CP/MV flag = p/(p+n) (or = 9.9 if p+n = 0)
Note that standard deviations for the three vectors are given as the square roots of the sum of squares of the standard deviations in the component averages. The component averages are given in the records but not their individual standard deviations.
Footnote C: The DBOT (Duration Between Observing Times) words: For a given record, we take the 1min average time shift and estimate, using the solar wind velocity and the location of the observing spacecraft, the time at which the corresponding observation would have been made at the spacecraft. Then we take the difference between this time and the corresponding time of the preceding 1min record and define this as DBOT1. This difference would be one minute in the absence of PFN (phase front normal) and/or flow velocity variations. When this difference becomes negative, we have apparent outof sequence arrivals of phase planes. That is, if plane A is observed before plane B at the spacecraft, plane B is predicted to arrive at the target before plane A. Searching for negative DBOT enables finding of such cases.
DBOT2 is like DBOT1 except that the observation time for the current 1min record is compared to the latest (most timeadvanced) previous observation time and not to the observation time of the previous record. Use of DBOT2 helps to find extended intervals of outofsequence arrivals.
We do not capture outofsequencearrival information at 15s resolution but only at 1min resolution. The standard deviation in the 1min averaged time shifts may be used to help find cases of outofsequence 15s data.
End of footnotes for spacecraftspecific data format
4b. HighResolution OMNI data set
The common format for the 1min and 5min OMNI data sets is Year I4 1995 ... 2006 Day I4 1 ... 365 or 366 Hour I3 0 ... 23 Minute I3 0 ... 59 at start of average ID for IMF spacecraft I3 See footnote D below ID for SW Plasma spacecraft I3 See footnote D below # of points in IMF averages I4 # of points in Plasma averages I4 Percent interp I4 See footnote A above Timeshift, sec I7 RMS, Timeshift I7 RMS, Phase front normal F6.2 See Footnotes E, F below Time btwn observations, sec I7 DBOT1, See footnote C above Field magnitude average, nT F8.2 Bx, nT (GSE, GSM) F8.2 By, nT (GSE) F8.2 Bz, nT (GSE) F8.2 By, nT (GSM) F8.2 Determined from postshift GSE components Bz, nT (GSM) F8.2 Determined from postshift GSE components RMS SD B scalar, nT F8.2 RMS SD field vector, nT F8.2 See footnote E below Flow speed, km/s F8.1 Vx Velocity, km/s, GSE F8.1 Vy Velocity, km/s, GSE F8.1 Vz Velocity, km/s, GSE F8.1 Proton Density, n/cc F7.2 Temperature, K F9.0 Flow pressure, nPa F6.2 See footnote G below Electric field, mV/m F7.2 See footnote G below Plasma beta F7.2 See footnote G below Alfven mach number F6.1 See footnote G below X(s/c), GSE, Re F8.2 Y(s/c), GSE, Re F8.2 Z(s/c), GSE, Re F8.2 BSN location, Xgse, Re F8.2 BSN = bow shock nose BSN location, Ygse, Re F8.2 BSN location, Zgse, Re F8.2 AEindex, nT I6 See World Data Center for Geomagnetism, Kyoto ALindex, nT I6 See World Data Center for Geomagnetism, Kyoto AUindex, nT I6 See World Data Center for Geomagnetism, Kyoto SYM/D index, nT I6 See World Data Center for Geomagnetism, Kyoto SYM/H index, nT I6 See World Data Center for Geomagnetism, Kyoto ASY/D index, nT I6 See World Data Center for Geomagnetism, Kyoto ASY/H index, nT I6 See World Data Center for Geomagnetism, Kyoto PC(N) index, F7.2 See World Data Center for Geomagnetism, Copenhagen Magnetosonic mach number F5.1 See footnote G below  Proton flux (>10 MeV) F9.2 In 5min OMNI, but not in 1min OMNI Proton flux (>30 MeV) F9.2 In 5min OMNI, but not in 1min OMNI Proton flux (>60 MeV) F9.2 In 5min OMNI, but not in 1min OMNI The data may be read with the format statement 1min: (2I4,4I3,3I4,2I7,F6.2,I7, 8F8.2,4F8.1,F7.2,F9.0,F6.2,2F7.2,F6.1,6F8.2,7I6,F7.2,F5.1) 5min: (2I4,4I3,3I4,2I7,F6.2,I7, 8F8.2,4F8.1,F7.2,F9.0,F6.2,2F7.2,F6.1,6F8.2,7I6,F7.2,F5.1,3F9.2) Note that for missing data, fill values consisting of a blank followed by 9's which together constitute the Ix or Fx.y format are used. Footnote D: The following spacecraft ID's are used: ACE 71 Geotail 60 IMP 8 50 Wind 51 Footnote E: Note that standard deviations for the minuteaveraged phase front normal and magnetic field vectors are given as the square roots of the sum of squares of the standard deviations in the component averages. For the magnetic field vectors only, the component averages are given in the records but not their individual standard deviations. 1min averaged phase front normal directions are given in the spacecraftspecific data sets but not in the high resolution OMNI data set. Footnote F: There are no phase front normal standard deviations in the 5min records. This word has fill (99.99) for such records. Footnote G: Derived parameters are obtained from the following equations. Flow pressure = (2*10**6)*Np*Vp**2 nPa (Np in cm**3, Vp in km/s, subscript "p" for "proton") Electric field = V(km/s) * Bz (nT; GSM) * 10**3 Plasma beta = [(T*4.16/10**5) + 5.34] * Np / B**2 (B in nT) (Note that very low B values (<~ 0.3 nT) encountered rarely in high resolution data can drive plasma beta values to above 1000. In high resolution OMNI, there were about 20 such minutes encountered in ~12 years. We have assigned the value 998.0 to plasma beta in such cases. Correct values of T, Np and B are available in the records for recomputation of plasma beta values.) Alfven Mach number = (V * Np**0.5) / 20 * B Magnetosonic Mach Number = V/Magnetosonic_speed Magnetosonic speed = [(sound speed)**2 + (Alfv speed)**2]**0.5 The Alfven speed = 20. * B / N**0.5 The sound speed = 0.12 * [T + 1.28*10**5]**0.5 For details on these, see http://omniweb.gsfc.nasa.gov/ftpbrowser/bow_derivation.html
It is an important current research topic to determine under what conditions singlespacecraft observations of solar wind field and plasma variations upstream (and possibly off to the side of) the Earth's magnetosphere can lead to reliable predictions of the solar wind variations to occur at the Earth's bow shock. Goodness of predictability may depend on many variables, including the spacecraft tobow shock separation geometry, the level of variation in the solar wind, the nature of the solar wind (e.g., fast vs. slow flows) and the technique used to shift data from the observation point to the bow shock.
It is possible to assess predictability goodness by multiple techniques. One would be to compare singlespacecraft predictions with the results of multispacecraft analyses, as was done by Weimer et al (2003), but done over a statistically significant number of independent time intervals. Another would be to search out a statistically significant number of major solar wind field and/or plasma discontinuous or other variations, and to note agreement level between spacecraft A's observations and spacecraft B's observations as shifted to A (A  shifted_B cross correlation functions  ccf's). A third would be to simply compute A  shifted_B ccf's in a large number of fixedduration time intervals, each characterized by AB separation geometry, mean physical parameter values in the intervals, parameter variance levels and the shift technique.
We are taking the last approach of the above paragraph. We have built a database of ccf's for field and plasma parameters for ~6000 4hour intervals in 19982000, for ACE data shifted to the Wind spacecraft by each of the four shift techniques discussed in Section 3a of these notes. While a final and comprehensive assessment of goodness of predictability as a function of ACEWind separation, solar wind flow state, solar wind variation level and shift technique lies in the near future, we report herein some preliminary results. It is intended that the final assessment will be published and will be reproduced here when completed.
We focus here on predictability of Bz variations as the most geoeffective of the solar wind parameters. Imagine that computed 4hour ccf's are the dependent variable in an independent variable space consisting of WindACE separation geometry (along and across the flow direction), the means and standard deviations for each physical parameter in the 4hour intervals, and the shift technique. For any bin in independent variable space, we find a certain number of intervals whose ccf's make up a distribution itself having a mean, median, standard deviation, etc. We examine the medians of these distributions as indicating dependence of predictability on the independent variables.
With no selection of parameters but exercising each of the 4 shift techniques, we find four distributions with numbers of 4hour intervals ranging between 5109 and 5288 and with medians ranging between 0.691 and 0.706. Standard deviations in the (nonGaussian) distributions of medians are ~0.31 Thus, at least in the case of looking over all the data, the various shift techniques are giving statistically equivalent results. In fact this is also the case for virtually all the binned analyses we've done.
Except where noted, additional results in this section are for shifts by "technique4" that we have used in our production work.
To first assess dependence of predictability on the transverse separation of Wind and ACE, we do a series of runs binned only by ACEWind Impact Parameter (IP). We find that the median of the Bz ccf distributions increases through the values 0.35, 0.34, 0.54,0.63, 0.75, 0.85, 0.87 at the IP decreases through the bins >150, 120150,90120, 6090, 3060, 1530 and 015 Re. The numbers of 4hour intervals in these distributions range from 159 (120150 Re) to 2001 (3060 Re). It is interesting that the ccf is the same for the 120150 Re bin and the >150 Re bin, and that the ccf is the same for the 015 and 1530 Re bins. The latter may be due to the occurrence of rotational discontinuities which, because of their propagation relative to the ambient solar wind, are not well accommodated by the shift assumptions. If we define a Bz scale length as the distance over which the Bz ccf falls by 10% (cf. Richardson and Paularena, 2001), then the scale length is approximately (13515)/(0.850.35)*10 = 24 Re.
Interestingly, when we look at medians of Bz ccf distributions involving MVAB0determined and CPdetermined PFN's in the IP = 015 Re and 1530 Re bins, we find 0.87 (015 Re) and 0.84 (1530 Re) for both methods. That both methods give the same result may run counter to an expectation that the MVAB0 method may be good for PFN determination for both tangential and rotational discontinuities, while the CP method should be better for PFN determination for nonpropagating tangential discontinuities having no field component normal to the discontinuity plane.
To examine the dependence of predictability on the solar wind variability level, we did a series of runs for various values of the standard deviation in the 4hour Bz average (sigmaBz). Upon limiting the ACEWind Impact Parameter to be less than 60 Re, we found median values of the Bz ccf distributions of 0.66, 0.76, 0.82, 0.85, and 0.91 in the sigmaBz bins 01, 12, 23, 34, >4 nT. The numbers of intervals per distribution ranges between 277 (sigmaBz > 4 nT) and 1129 (1 < sigmaBz < 2 nT). Removing the constraint on the WindACE IP almost doubled the numbers of 4hour intervals per sigmaBz run, but decreased the median ccf's only by 7% (at largest sigmaBz) to 13% (at smallest sigmaBz). The conclusion here is that the higher the variation level in Bz, the more predictable are bow shock nose Bz variations, given upstream Bz observations.
To examine possible dependence of predictability on the X distance upstream, we define bins by X(ACE)  X(Wind). For a series of runs all having WindACE IP < 60 Re, we find medians in Bz ccf distributions of 0.78, 0.77, 0.81, 0.74 for bins of <50 Re, 50125 Re, 125200 Re, >200 Re respectively. The numbers of intervals in the distributions range between 345 (delta X > 200 re) and 1248 (125 Re < delta X < 200 Re). The conclusion here is that, while there's a hint of a downturn in the Bz ccf at delta X > 200 Re, there's no major dependence of predictability on delta X.
Finally, to assess predictability on flow speed, we do runs in flow speed bins <350, 350450, 450550 and >550 km/s for IP < 60 Re and for sigmaBz and for sigmaBz > 1 nT, we find medians in the Bz ccf distributions of 0.84, 0.83, 0.79, 0.72 as the speed increases through the four indicated bins. Numbers of 4hour intervals in the bins ranges from 346 (V > 550 km/s) to 1210 (350450 km/s). Predictability in Bz variations decreases modestly as the solar wind flow speed increases.
Appendix 1. Crossspacecraft Comparisons
While the key issue for our new products is the extent to which solar wind variations observed remote from the Earth's bow shock may be used to infer variations at the bow shock nose, it is also of interest to review whether there are systematic differences in parameter values between pairs of input data sets. This is largely because the spacecraftinterspersed data set (i.e., High Resolution OMNI  HRO) should not have excessive parameter changes due to transition between one source spacecraft and another, and so that the parameter values included in the new HRO are most likely "true" at least at the observation points.
This section discusses our search for systematic differences
among input data sets. We expect that any systematic
differences, while they may change slowly, will not change
on the scale of days or weeks. Thus we assess systematic
differences using hourly averaged physical parameter values
as built from higher resolution data shifted by the simple
technique used in preparing the hourly resolution OMNI 2
data set and discussed here
http://omniweb.gsfc.nasa.gov/html/ow_data.html.
Such data, and the tools for comparison, are available at
http://omniweb.sci.gsfc.nasa.gov/ftpbrowser/mag_iwa_s2.html (magnetic field data)
http://omniweb.sci.gsfc.nasa.gov/ftpbrowser/pla_iwa_s2.html (plasma data, linear)
http://omniweb.sci.gsfc.nasa.gov/ftpbrowser/pla_iwa_s3.html (logs of N and T)
These interfaces determine the slopes and intercepts in the linear regressions P1 = a + b*P2, where P represents any of the relevant physical parameters (or, as special cases, log N and log T). The interfaces also determine the uncertainties in the slope and intercept, cross correlation coefficients, and the rms deviations between the data points on the scatter plots and the best fit lines. The "1" and the "2" refer to the members of any spacecraft pair.
Our work uses linear regressions of logs of densities and temperatures rather than the values of N and T themselves because these parameters are more lognormally distributed than normally distributed.
Note that the documentation of our hourly resolution OMNI 2 data set at http://omniweb.gsfc.nasa.gov/html/ow_data.html extensively discusses intercomparisons of hourly ACE, Wind and IMP 8 magnetic field and plasma data. The rationale for the present discussion is to address the significantly extended time span over which data are now available for intercomparison.
For Wind/SWE, we would use the Key Parameter (KP) data, but would normalize them, if any normalizations were appropriate, to the nonlinear fit (NLF) data for which admirably small uncertainty estimates had been derived by Kasper et al. (2006). We have built a series of parameterspecific tables summarizing the results of the annual and multiyear cross correlations. For plasma comparisons, we used P(Wind/NLF) = a + b * P(2) where now P(2) might be ACE or IMP 8 or Wind/KP.
Magnetic field comparisons
If the Wind magnetic field data are right, then IMP field magnitude and components (absolute values) would need to be increased by 1.5 to 2 percent to match Wind. Thus there are systematic WindIMP magnetic field component differences of ~0.3 to ~0.4 nT at ± 20 nT. Averaged over 19962000, when Bz(Wind) = 0, Bz(IMP) = 0.06, indicating good IMP zero level determination. There is no clear evidence of any time dependence in the WindIMP relations in magnetic field data.
By contrast, Wind version 4 and ACE magnetic field data agree to within 1 percent for virtually all components and years, and to within 0.03 nT in Bz at Bz = 0.
The Geotail magnetic field data available as we were creating these new data sets were known to have preliminary and incorrect Bz offsets. See the discussions in Section 2 and in Appendix 2.
Flow speed comparisons
Flow speeds agree to within 1% or less. That is
V(Wind/NLF)  V(Z) / V(Wind/NLF) < 1%,
where Z = Wind/KP, ACE, or IMP. For the case of Z = ACE and IMP, V(Wind/NLF)
exceeds V(Z). V(Wind/KP) is virtually identical to V(Wind/NLF).
Flow direction angle comparisons
Flow azimuth angles between any source pair agree to within 1 degree over the ± 10 deg range. Flow elevation angle agreement level depends on the source pair. Wind/NLF and Wind/KP agree to within 1 degree over the ± 10 degree range. The same is true for Wind/NLF vs. ACE except that near +10 deg, Wind/NLF exceeds ACE by ~1.5 deg. The IMP elevation angle exceeds the Wind/NLF elevation angle by an amount ranging from ~1.2 deg at 10 deg to ~4 degrees at +10 degrees. An apparent IMP flow elevation angle offset of ~2 deg has been recognized for many years. The present analysis shows for the first time an elevation angle dependence in this offset. There are no evident time dependences in the relations between any source pair for flow speed or direction angles.
Density Comparisons and Temperature
Wind and ACE proton parameters. Previously we used Wind/SWE parameters based on anisotropic nonlinear fits to Wind/SWE plasma distributions through November 2004, and we used crossnormalized Wind/SWE Key Parameter data thereafter. Now, owing to their greater "robustness," the only Wind/SWE proton data we use for 1995current are the crossnormalized SWE KP data. (SWE KP crossnormalization is to the SWE nonlinear fit data.) All recent results are given in Appendix 2.
The full old OMNI documentation package made before February 15, 2013 user may find at http://omniweb.sci.gsfc.nasa.gov/html/HROdocum_old.html. (New upgrades for data crossnormalizations were made after February 15, 2013)
Appendix 2. CrossNormalizations
Using hourly averaged data, the previous section has revealed the mainly small systematic differences for each magnetic field parameter between Wind on the one hand and ACE and IMP 8 on the other hand. It has also revealed systematic differences for each plasma parameter between the nonlinear fitbased Wind/SWE data on the one hand and the Wind/SWE key parameter data, the ACE/ SWEPAM data and the MIT/IMP 8 data on the other hand.
The question is now whether and for which parameters we should cross normalize the data to be included in the spacecraftinterspersed high resolution OMNI data set. (Note that we do no such normalizations for our new spacecraftspecific data sets.) We choose to minimize crossnormalizations for multiple reasons. First, since we use 3hour swaths of samespacecraft data in 1min OMNI, there are at most only 0.55% of minutetominute transitions that would involve a change of source spacecraft. In fact, the actual fraction of transitions between sources is very much less than this. Second, we do not expect this data set to be used for long term solar wind variation studies; the hourly resolution OMNI data set is more appropriate for this.
So, as for the present hourly OMNI data set, we shall crossnormalize only plasma densities and temperatures.
For Wind/KP Density and Temperature data to Wind/NLF we use the same equations we used for hourly OMNI:
For Wind/SWE KP Np and Tp data For Np, for all V and time, LogN(Wind/KP, norm) = 0.055 + 1.037 * LogN(Wind/KP, obsvd) For Tp, for all V and for 19957, LogT(Wind/KP, norm) = 0.030 + 1.055 * LogT(Wind/KP, obsvd) For Tp, for all V and for >= 1998, LogT(Wind/KP, norm) = LogT(Wind/KP, obsvd) For ACE/SWEPAM Density and Temperature data to Wind/NLF we use the same equations we used for hourly OMNI: Let t be fractional years since 1998.0. (E.g., t = 1.5 on July 1, 1999.) Let V = solar wind speed N = ACE/SWEPAM proton density as observed Nn = value of N as normalized to equivalent Wind/SWE nonlinear fit proton densities For V < 395 km/s, Nn = [0.925 + 0.0039 * t] * N For V > 405 km/s, Nn = [0.761 + 0.0210 * t] * N For 395Appendix 3. Despike Algorithms405, Nn = [74.02  0.164*V  6.72*t + 0.0171*t*V] * N/10 For temperature (all V), LogT(norm) = 0.069 + 1.024 * LogT(obsvd) For IMP8/MIT we shall use the same timeinvariant equations we used for hourly OMNI. Density: V<350 km/s: LogN(norm) = 0.020 + 0.941 * LogN(obsvd) 350450 km/s: LogN(norm) = 0.033 + 0.919 * LogN(obsvd) V>450 km/s: LogN(norm) = 0.019 + 0.907 * LogN(obsvd) Temperature: V<350 km/s: LogT(norm) = 0.864 + 0.839 * LogT(obsvd) 350450 km/s: LogT(norm) = 0.491 + 0.920 * LogT(obsvd) V>450 km/s: LogT(norm) = 0.702 + 0.890 * LogT(obsvd) For Geotail, we use Density (all time and all V): LogN(norm) = 0.072 + 0.980 * LogN(obsvd) Temperature (19951998, all V): LogT(norm) = 0.166 + 0.925 * LogT(obsvd) Temperature (19992005, all V): LogT(norm) = 0.362 + 1.052 * LogT(obsvd) As explained in the Geotail data discussion of Section 2, The Geotail magnetic field data we worked with had preliminary and incorrect Bz offset values. Accordingly, we compared Geotail B data with B data from the other spacecraft and derived the following "normalizations" of the Geotail B data: Bx and By, all time, all Bz: Bx(norm) = 1.02 * Bx(obsvd) By(norm) = 1.02 * By(obsvd) Bz (depends on time) 1995010119951231: Bz(norm) = 0.490 + 1.004 * Bz(obsvd) 1996010119991231: Bz(norm) = 0.597 + 1.017 * Bz(obsvd) 2000010120040401: Bz(norm) = 0.149 + 1.019 * Bz(obsvd) 2004040220050401 : Bz(norm) = 0.461 + 1.020 * Bz(obsvd) 2005040220051231: Bz(norm) = 0.663 + 1.023 * Bz(obsvd) Bt (depends on time and on sign of Bz) 1995010119991231, Bz<0: Bt(norm) = 0.123 + 1.022 * Bt(obsvd) 1995010119991231, Bz>0: Bt(norm) = 0.180 + 1.012 * Bt(obsvd) 2000010120040401, Bz<0: Bt(norm) = 0.052 + 1.016 * Bt(obsvd) 2000010120040401, Bz>0: Bt(norm) = 0.021 + 1.014 * Bt(obsvd) 2004040220051231, Bz<0: Bt(norm) = 0.123 + 1.022 * Bt(obsvd) 2004040220051231, Bz>0: Bt(norm) = 0.180 + 1.012 * Bt(obsvd)
We have undertaken to eliminate spikes from the Wind and IMP 8 magnetic field and plasma data sets. Owing to their relatively clean state, we have judged it unnecessary to despike the ACE data. Wind magnetic field data were despiked with the simple approach of eliminating any record with a field magnitude or component absolute value in excess of 70 nT. Other data were despiked with the approach described as follows.
We test a point using its two predecessors and two followers. We require that the 1st and last of these 5 points be within 15 mins (for B data) or 60 mins (for plasma data). The first two and last two points in a data segment separated from its neighbors by intervals of >15 min (B) or >60 min (plasma) go untested by the algorithms discussed here. (We visually scanned output data looking for obvious spikes thereby missed, and deleted these.)
Any record having a declared spike in any of its physical parameters is rejected. For a parameter value to be declared a spike, it must satisfy two criteria.
Let P represent the value of the physical parameter being tested. Define <P> as the mean value of parameter P over the 1st, 2nd, 4th, and 5th points of the current set, and let sigma(P) be the RMS deviation in this average. The first test for a spike is to have P<P> > 4 * sigma(P).
The second tests 
IMP IMF data  For P = B, require P<P> > 0.2 * <P>. For P = Bx, By, Bz, require P<P> > 1.0 nT.
IMP plasma data  For P = V, N, W [W = thermal speed; T(deg) = 60.5 * W(km/s)**2], require P<P> > k * <P> where k = 0.1, 0.3, 0.3 for P = V, N, W respectively. For P = flow latitude and longitude angles, require P<P> > 4.0 deg. (We have also excluded all IMP plasma records having flow angle > 15 deg.)
Wind/SWE plasma data  For P = V, N, T, require P<P> > k * <P> where k = 0.1, 0.3, 0.3 for P = V, N, T respectively. For P = Vx, Vy, Vz, require P<P> > 0.1 * <V>.
For completeness, we note that the Wind/SWE plasma data came to us already having been run through MIT despike software that required that the relative difference between the point being tested and the median of that point and its immediate predecessor and immediate successor should be less than 0.1, 0.5 and 1.0 for flow speed, density and thermal speed, respectively. Some points accepted by the MIT software were rejected by ours.
Appendix 4. Determination of bow shock nose location
We assume the geocentric direction to the bow shock nose is parallel to the (aberrated) solar wind flow direction: Rt =  Rt * V/V. (V and V are determined from the aberrationcorrected V values provided in the input plasma data sets, but with 29.8 km/s, the mean orbital speed of the Earth about the sun, added to their Vy values.)
Rt is provided as a function of the geocentric magnetopause nose distance Rmp and the magnetosonic Mach number Mms by Farris and Russell (1994) as Rt = Rmp * [1.0 + 1.1 * ((2/3)*Mms**2 + 2) / ((8/3) * (Mms**2  1)] where Mms = Vsw / Vms Vms**2 = 0.5 * (Va**2 + Vs**2 + SQRT [(Va**2 + Vs**2)**2  4*(Va**2*Vs**2 * (cos th)**2])Va = B / SQRT (4pi * (4*Na + Np) * Mp) = 20.3 * B / SQRT (Np) (Alfven speed)
Vs = 0.12 * [Tp (deg K) + 1.28*10**5]**0.5 (sound speed) and where the magnetopause nose distance is given in terms of the solar wind pressure P and Bz, by Shue et al (1997) as
Rmp = (11.4 + K * Bz) * P**1/6.6 where P is the pressure defined as a function of Np and V by
P = (2*10**6)*Np*Vp**2 (N in cm**3, Vp in km/s; P in nPa) and where K = 0.013 if Bz > 0 and K = 0.140 if Bz < 0. Na and Np above refer to alpha particle and proton densities. The equation for P assumes a constant 4% alpha particle contribution.
Appendix 5. Computation of 1Minute and 5minute Averages
We have input records with (typically shifted) time tags T and parameter values P. The parameters are either ~15sec magnetic field or ~1min plasma parameters. Magnetic field parameters are typically averages of yet higher resolution magnetic field parameters that have been obtained between some first time Tf and some last time Tl. Plasma parameters are as derived from some distribution function accumulated between some first time Tf and some last time Tl. The relation between the input record time tag T and the first and last times (Tf & Tl) of the data on which the record's parameter values are based is datasetspecific. The duration TlTf varies between records for some data sets but not for others.
We want to create output records tagged at the start of every minute. The parameter values in the output records should be based, as much as possible, on observations made during that minute. This means that, for a given output minute, we want to do weighted averages over any input values whose underlying data were obtained, in whole or part, during the output record's minute of interest. One weighting factor is the extent to which the parameters of the input record cover the desired output interval. The other factor is the extent to which the parameters of the input record are determined by data taken outside the minute of interest. These weights may be written as follows.
Let Tf* = Tf or Tf* = the first instant of the output record, whichever is later. Let Tl* = Tl or Tl* = the last instant of the output record, whichever is earlier. Then Tl*  Tf* = the part of the duration of the input record which lies within the duration of the output record. Let S = Tl*  Tf*. The fraction of the input record which lies within the output record time span is (Tl*  Tf*)/(Tl  Tf). Let this fraction be F. Note that F = S/(Tl  Tf). For data sets having the same durations [i.e., (Tl  Tf) values] for all records, we have F = constant * S. ACE and Wind field records and plasma records each has a common TlTf, while both IMP8 field and plasma records have varying TlTf values.
To get parameter values <P> for the output records, find all input records whose parameters are based on observations taken within the output minute of interest. Define the weighted averages as <P> = SUM (Si * Fi * Pi)/SUM (Si * Fi), where i indexes the relevant input records and where the sums are over all the relevant input records. There is interest in defining variance measures of the P values. These may be attributed to variances within the contributing Pi values and to the spread of the Pi values about the mean <P> value. We consider below only the variability in our Pi values about <P>.
Since we build the mean using weighting, we do so also for the variance, using the expression
V = [SUM ((Si*Fi) * (Pi<P>)**2) / SUM (Si*Fi) = <P**2>  <P>**2
Fiveminute averages are computed from the 1min averages. The 5min averages tagged with minute = 0 are built from 1min averages tagged as being for minutes 0, 1, 2, 3 and 4. Likewise for 5min averages tagged with minutes 5, 10 ... 55.
Appendix 6. Prioritization of Sources for inclusion in OMNI
(This Appendix was originally written as we were creating HRO from ACE, Wind and IMP data. The variant used in adding Geotail data to HRO is described near the end of this Appendix.)
There will be many minutes when shifted data are available from multiple spacecraft. In building High Resolution OMNI (HRO), we shall follow the hourly OMNI practice of selecting data from one source when multiple sources are available. However, instead of following the hourly OMNI practice of selecting the source for each unit time increment, for our HRO products we shall select and intersperse 3hour data segments [both field and plasma data together] from among our multiple sources.
There are three criteria we shall use, namely, (a) the sourceEarth Impact Parameter (IP, separation transverse to the flow, with allowance for Earth's orbital motion), (b) the completeness of magnetic field data coverage in the 3hour interval, (c) source continuity. This latter means that if neither (a) nor (b) provides a strong discriminant between sources, we shall favor using the source used in the previous 3hour segment.
We make discrimination between spacecraft pairs algorithmically as follows. Let ScX and ScY represent the two spacecraft being compared.
Let A = IP (ScXEarth) B = IP (ScYEarth) C = fractional ScX coverage, this segment D = fractional ScY coverage, this segment E = +1 if ScX data used (i.e., if F>0) in prior segment E = 1 if ScY data used (i.e., if F<0) in prior segment Let's define F by F = a * (BA) + b*(CD)/(D+C) + c*E For the weights, a, b, c, we have experimented a bit and have chosen a = 1/30Re b = 2 c = 0.25 If F > 0, use ScX, otherwise use ScYFor 3hour intervals with some data available from each of three spacecraft (early 1998 through mid2000), we have determined the favored spacecraft for each of the three possible pairings of spacecraft and then determined by inspection which one spacecraft was preferable to both of the other two spacecraft.
When we added Geotail data to HRO, we treated the 3spacecraftbased HRO data set as a single data set and the Geotail data set as a second data set, and used the 2spacecraft algorithm described above for determining whether, for each 3hour interval, Geotail data should replace the data previously in HRO. We carefully used Impact Parameter appropriate to the spacecraft used in HRO for the interval. Further, if the spacecraft used in HRO for a given interval is different than the spacecraft used in HRO for the preceding interval, we ignore the "continuity factor" by setting E = 0 in the above algorithm.
Note that upon making extensions to HRO, we frequently have data from one source spacecraft reaching closer to current data than data from other source(s). In such cases, most current data will be used in HRO with no "F tests" relative to other spacecraft. But later, when data from other source(s) become available, interspacecraft tests will be performed and the originally included data may be replaced by data from the other source(s).
Bargatze, L.F., R.L. McPherron, J. Minamora and D.Weimer, A new interpretation of Weimer et al’s solar wind propagation delay technique, J. Geophys. Res., 110, A07105, doi:10.1029/2004JA010902, 2005.
Farris, M.H. and C.T. Russell, Determining the standoff distance of the bow shock: Mach number dependence and use of models, J. Geophys. Res., 99, 1768117689, 1994.
Haaland, S., G. Paschmann, and B. U. Ö. Sonnerup (2006), Comment on “A new interpretation of Weimer et al.'s solar wind propagation delay technique” by Bargatze et al., J. Geophys. Res., 111, A06102, doi:10.1029/2005JA011376.
Kasper J. C., A. J. Lazarus, J. T. Steinberg, K. W. Ogilvie, A. Szabo (2006), Physicsbased tests to identify the accuracy of solar wind ion measurements: A case study with the Wind Faraday Cups, J. Geophys. Res., 111, A03105, doi:10.1029/2005JA011442.
Knetter, T., F.M. Neubauer, T. Horbury and A. Balogh, Four point discontinuity observations using Cluster magnetic field data: A statistical survey, J. Geophys. Res., 109, A06102, doi:10.1029/2003JA010099, 2004
Richardson, J.D., and K.I. Paularena, Plasma and magnetic field correlations in the solar wind, J. Geophys. Res., 106, 239251, 2001.
Shue, J.H., J.K. Chao, H.C. Fu, C.T. Russell, P. Song, K.K. Khurana and H.J. Singer, A new functional form to study the solar wind control of the magnetopause size and shape, J. Geophys. Res., 102, 94979511, 1997.
Sonnerup, B.U.O and L.J. Cahill, Explorer 12 observations of the magnetopause current layer, J. Geophys. Res., 73, 1757, 1968.
Sonnerup, B.U.O. and M. Scheible, Minimum and maximum variance analysis, in Analysis Methods for MultiSpacecraft Data, edited by G. Paschmann and P.W. Daly, Int. Space Sci. Inst., Bern, 1998.
Weimer, D.R., and J.H. King, Improved calculations of interplanetary magnetic field phase front angles and propagation time delays, J. Geophys. Res., 113, A01105, doi:10.1029/2007JA012452, 2008
Weimer, D.R., D.M. Ober, N.C. Maynard, W.J. Burke, M.R. Collier, D.J. McComas, N.F. Ness and C.W. Smith, Variable time delays in the propagation of the interplanetary magnetic field, J. Geophys. Res., 107(A8), 10.1029/2001JA009102, 2002.
Weimer, D.R., D.M. Ober, N.C. Maynard, M.R. Collier, D.J. McComas, N.F. Ness, C.W. Smith and J. Watermann, Predicting interplanetary magnetic field (IMF) propagation delay times using the minimum variance delay technique, J. Geophys. Res., 108(A1), 1026, doi:10.1029/2002JA009405, 2003.

If you have any questions/comments about OMNIWEB system, contact:
Dr. Natalia Papitashvili, Mail Code 672,
NASA/Goddard Space Flight Center, Greenbelt, MD 20771 