OMNI 2 Preparation
Joe King and Natalia Papitashvili, NASA/SPDF and ADNET Systems, Inc.
Abstract: This file describes the OMNI 2 data set. It identifies the data sources, edits and time shifts, cross regressions, normalizations, selection prioritization, etc. It identifies access paths to the OMNI 2 data set and to the multiple input data sets.It was most recently updated February, 2013
[Note: a complementary OMNI 2 documentation file, addressing time spans of various data types, a format description of the OMNI 2 data records, 1 and 27day averaged OMNI 2 versions, and other details is found at http://omniweb.gsfc.nasa.gov/html/ow_data.html]
In this section we address, in turn, the sources of IMF data, plasma data, energetic particle data, and solar and geomagnetic activity indices used in OMNI 2.
IMF data sources: Table 1 shows the sources of magnetic field data. The parentheses after the spacecraft names contain the value used in the IMF spacecraft ID field of the OMNI 2 records. Some parentheses also contain an "Exx." These are the corresponding Explorer numbers and may be useful in cross referencing. The PI column gives the Principal Investigator or other lead person. Key additional scientists for recent spacecraft are A. Szabo (IMP 8, Wind), R. Lepping (IMP 8) and C. Smith (ACE). The time span column of Table 1 gives the time range of data included in OMNI 2; it may be less than the time range over which the instrument produced useful data.
As of 2009, Wind and ACE data were being periodically added to OMNI. Information on current latest date of plasma data in OMNI is given at http://omniweb.gsfc.nasa.gov/html/ow_data.html.
All instruments were boommounted triaxial fluxgate magnetometers except those on IMP's 1 and 3 which were biaxial but flippable (spacecraft spin plus sensor orientation yielded three dimensional vectors) and that on ISEE 3 which was a vector helium magnetometer. Most older data were received as hourly averages of IMF intensities and GSE Cartesian components from the Principal Investigator (PI) teams. For old OMNI, GSM components transformed from, and the magnitude and GSE direction angles of, the IMF vector formed from the hourly GSE Cartesian component averages, were determined at NSSDC or at the data provider's site. For OMNI 2, GSM components were computed at NSSDC from hourly averaged GSE components using the GSMtoGSE transformations of the Tsyganenko Geopack software package and using time tags at the midpoints of the hours. (Only for 19634 IMP 1 data was this not the case since the Geopack code is only relevant to 1965 and forward.) For ISEE 3, Wind and ACE, NSSDC computed hourly averages from timeshifted higher resolution data as explained later in this text.
Standard deviations were provided to NSSDC in Cartesian components and field intensities, except that with the HEOS data standard deviations in the field magnitude and direction angles were provided. These were inserted into the words for standard deviations in Cartesian components in the records of OMNI to which HEOS contributed IMF data. For OMNI 2, we have computed from the HEOSprovided data values of standard deviations of Cartesian components and included these in OMNI 2 instead of the OMNI included standard deviations in angles.
Table 1 Spacecraft PI Time span Reference IMP 1 (18, E18) Ness 11/27/6302/15/64 Ness et al, 1964 IMP 3 (28, E28) Ness 06/01/6501/29/67 Ness et al, 1964 AIMP 1 (33,E33) Ness 07/04/6607/13/68 Behannon et al, 1968 IMP 4 (34, E34) Ness 05/26/6712/27/68 Fairfield, 1969 AIMP 2 (35,E35) Ness 07/26/6711/10/69 Ness et al, 1967 HEOS (1) Hedgecock 12/11/6810/28/75 Hedgecock, 1975 IMP 5 (41, E41) Ness 06/21/6910/26/72 Fairfield & Ness, 1972 IMP 6 (41, E43) Ness 03/14/7107/21/74 Fairfield, 1974 IMP 7 (47, E47) Ness 09/26/7204/03/73 Mish & Lepping, 1976 IMP 8 (50, E50) Ness 10/30/7305/12/00 Mish & Lepping, 1976 ISEE 3 (13, E59) E.Smith 08/14/7812/21/82 Frandsen et al, 1978 Prognoz10 (10) Yeroshenko 04/27/8511/04/85 Styazhkin et al, 1985 Wind (51) Lepping 11/21/94current Lepping et al, 1995 ACE (71, E71) Ness 02/06/98current Smith et al, 1998. Geotail (60) Nagai 1995/128  1997/272 For the past many years, the only IMF data included in OMNI have been from the IMP 8, Wind and ACE spacecraft, launched in 1973, 1994 and 1997, respectively. The IMP 8 magnetometer failed in June, 2000, after almost 27 years of operations. The Wind and ACE magnetometers remain operational in February 2013. The web pages of the IMP 8, Wind and ACE magnetometer teams are to be found at: IMP 8: http://wind.gsfc.nasa.gov/imp8/ Wind: http://wind.gsfc.nasa.gov/mfi/ ACE: http://www.srl.caltech.edu/ACE/ASC/level2/index.html Hourly and higher resolution versions of data from these magnetometers are available via multiple pathways, mostly cited on the foregoing web pages. In particular, NSSDC makes files ftpaccessible from ftp://nssdcftp.gsfc.nasa.gov/spacecraft_data and makes data accessible with graphical display and screen lists with subsetting by parameter and by time via CDAWeb and via FTPBrowser at http://cdaweb.gsfc.nasa.gov/cdaweb/sp_phys/ and at http://omniweb.sci.gsfc.nasa.gov/ftpbrowser/ Of special relevance to OMNI 2 preparation is an FTPBrowseraccessible merged hourly IMP8WindACE IMF data set at http://omniweb.sci.gsfc.nasa.gov/ftpbrowser/mag_iwa.html from which one may make overlapping timeseries plots. A second interface at http://omniweb.sci.gsfc.nasa.gov/ftpbrowser/mag_iwa_s2.html enables one to make scatter plots and linear regression fits of userselected parameter pairs and time span. Results of analyses of these data, with this latter tool, are reported below. A third interface at http://omniweb.sci.gsfc.nasa.gov/ftpbrowser/mag_iwa_d.html enables one to determine distributions, means and their standard deviations, and medians of any IMF parameter from any of the spacecraft named, for any time span. 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 OMNI 2 when they became available, replacing the earlier MFI data.
Plasma data sources: Most of the solar wind plasma data used in OMNI 2 are from the MIT Faraday Cups (cf. PI = Bridge in Table 2) or the Los Alamos National Laboratory (LANL) electrostatic analyzers (cf. PI = Bame in Table 2). Table 2 shows the data sources in some detail. The data were mainly provided to NSSDC and used in OMNI as 1hour averages. Exceptions are (1) the IMP 1, Vela and HEOS data which were provided as 3hour averages and were assigned to each of three successive onehour records in OMNI, (2) the ISEE 3, Wind and ACE data whose 1hour averages were computed at NSSDC from timeshifted higherresolution data and (3) the LANL IMP 6, 7 and 8 data whose 1hour averages were computed at NSSDC from higherresolution data.
As of 2009, Wind and ACE data were being periodically added to OMNI. Information on current latest date of plasma data in OMNI is given at http://omniweb.gsfc.nasa.gov/html/ow_data.html
It should be noted that A.J. Lazarus and J.C. Kasper of MIT contributed the Wind/SWE ion measurements and data processing and analysis thereof. The ACE/SWEPAM instrument was one of the series of LANL electrostatic analyzers to whose success R. Skoug contributed significantly. Additional key scientists contributing to IMP 8 plasma data are have been J. Gosling and J. Steinberg at LANL and A. Lazarus, K. Paularena and J. Richardson at MIT.
Table 2 Spacecraft PI Time span Comments, Reference IMP 1 (18) Bridge 11/27/6302/22/64 Bridge et al, 1965 Merged Vela (97) Bame 07/21/6403/18/71 Bame et al, 1971 Vela 3 (3) Bame 07/26/6511/13/67 Hundhausen et al, 1967 AIMP 1 (33) Bridge 07/06/6609/23/69 Lyon et al, 1968 IMP 4 (34) Ogilvie 06/03/6712/16/67 Ogolvie et al, 1968 AIMP 2 (35) Bridge 07/28/6707/03/68 Lyon et al, 1967 OGO 5 (5) Neugebauer 03/05/6804/29/71 Neugebauer, 1970 HEOS 1 (1) Bonetti 12/11/6804/15/70 Bonetti et al, 1969 IMP 6 (43) Bame 03/18/7107/21/74 Feldman et al, 1973 IMP 7 (44) Bame 10/06/7209/29/78 Asbridge et al, 1976 IMP 7 (47) Bridge 01/03/7509/20/78 Lazarus et al, 1998 IMP 8 (45) Bame 11/04/7307/16/00 Asbridge et al, 1976 IMP 8 (50) Bridge 12/05/7307/26/01 Lazarus et al, 1998 ISEE 1 (11) Bame 10/30/7712/19/79 Bame et al, 1978b ISEE 3 (13) Bame 08/16/7810/12/82 Bame et al, 1978a Wind (51) Ogilvie 01/01/95 Kasper, 2002 ACE (71) McComas 02/05/98 McComas et al, 1998 Geotail (60) L.Frank 1995/128  2004/341 [2008: As of 2008, the Wind/SWE nonlinear fitsbased proton parameters (cf. below) had not been determined past late November, 2004. Accordingly we have used in OMNI 2 a cleaned version of the Wind/SWE key parameter data set for later times. See the documentation of thisdata set in the High Resolution OMNI documentation file at http://omniweb.gsfc.nasa.gov/html/HROdocum.html.] The plasma parameters included in the earlyperiod OMNIinput data sets (i.e., the first 8 data sets of Table 2) are identified in the original OMNI documentation, available through OMNIWeb, and will not be repeated here. For the middle and later periods, we have used the following in OMNI 2 from the various input data sets: N V T phiV thetaV alpha/prot IMP 6 x x x x x IMP7 (LANL) x x x x x IMP7 (MIT) x IMP8 (LANL) x x x x x IMP8 (MIT) x x x x x ISEE 1 x ISEE3 (protons) x x x x ISEE3 (electrons) x x Wind/SWE x x x x x x ACE/SWEPAM x x x x x x
The parameters provided by the LANL plasma team were determined by taking moments over distribution functions. The parameters provided by the MIT team were determined by making nonlinear fits of convecting Maxwellian distributions (anisotropic biMaxwellians for Wind/SWE) to the observed distributions. The references cited above further describe these approaches. As a special situation, both momentsbased and fitsbased Wind/SWE parameters were provided by MIT. We chose to include in OMNI 2 the fitsbased SWE parameters rather than the momentsbased parameter values for consistency with using fitsbased parameters for other MITprovided data sets. That OMNI 2 involves the interspersal of LANLprovided momentsbased parameters with MIT provided fitsbased parameters will be at least partly compensated for by the cross normalizations of multisource data (discussed subsequently).
Readers interested in the differences between the two parameterdetermination approaches may access http://omniweb.sci.gsfc.nasa.gov/ftpbrowser/pla_iwa.html/ or pla_iwa_s2.html or pla_iwa_s3.html to compare Wind/SWE hourly averages based on time shifted momentsbased 92s parameters with equivalent fitsbased Wind/SWE hourly averages. Figure 1 (V), Figure 2 (N) and Figure 3 (T) show scatter plots and linear regression fits for Wind/SWE parameter pairs. In general, flow speeds are the same to within a percent or so independent of parameter determination approach, while for both densities and temperatures the momentsbased parameters are greater than or equal to the fitsbased parameter values.
The LANL plasma instruments on ISEE 3 measured ions and electrons separately. The ion instrument failed February 19, 1980. Electronbased flow speeds and densities were used in OMNI 2 after the ion instrument failure until late 1982, but neither electron based temperatures nor flow direction angles were included into OMNI 2. (This was also done for OMNI years ago.) It should be noted that on two days (July 4,1979, and July 31, 1979), ISEE 3 measured densities so low that for each of several hours, the hourly averaged density was less that 0.05/cc. Given OMNI's use of F6.1 format for densities, this yields an apparent density of 0.0 in OMNI for these few hours.
Only flow speeds are provided from IMP 7 (MIT) and from ISEE 1. In the former case, this limitation to flow speed was at the suggestion of A. Lazarus at MIT. In the case of ISEE 1, there were too few hours (<240) when ISEE 1 data were used in OMNI 2 (only when no other data were available) to prioritize doing new density and temperature normalizations. These may yet be done.
Web pages of the primary contributors of recent plasma data to OMNI 2 are at:
IMP8/MIT: ftp://space.mit.edu/pub/plasma/imp/www/imp.html
IMP8/LANL: http://imp8.lanl.gov/
Wind/SWE/MIT: http://web.mit.edu/space/www/wind/wind.html
ACE/SWEPAM: http://swepam.lanl.gov/
As with the magnetometer data addressed earlier, hourly and higher resolution plasma data are available via multiple pathways, mostly cited on the foregoing web pages. They are available via ftp from ftp://spdf.gsfc.nasa.gov/pub/data/ and with display and subsetting capabilities via CDAWeb and/or FTPBrowser at http://cdaweb.gsfc.nasa.gov/cdaweb/sp_phys and http://omniweb.sci.gsfc.nasa.gov/ftpbrowser/
Several multiplesource hourly data sets were created at NSSDC to aid in data checking and crossnormalization. These are discussed in the "Data cleaning" section to follow and at http://omniweb.sci.gsfc.nasa.gov/ftpbrowser/merged.html.
For completeness, we mention two additional adjustments we might have made to OMNI plasma densities but did not. First, for values shifted from ~200 Re (~0.01 AU) upstream to Earth (for ISEE 3, ACE and some Wind data as discussed in detail later), we might have multiplied by 0.98 to reflect the fact that densities are expected to fall as the inverse square of heliocentric distance, on average. Second, we might have normalized to 1.00 AU by allowing for the noncircularity of the Earth's annual orbit about the sun. Note that Earth is at about 1.017 AU midyear each year and at about 0.983 near the start and end of the year. This would contribute a yearend density at Earth ~7% higher than at midyear. (M. Collier, GSFC, private communication, reports higher longtermaveraged OMNI densities for OctoberJanuary than for the rest of the year.)
We neglect the first ~2% effect because it is small, and we neglect the second ~7% effect because the objective of OMNI is to reflect the state of the solar wind just outside the bow shock in support of solar wind  magnetosphere coupling studies more so than to support strictly solar wind studies.
Energetic particle data sources: Fluxes of protons above 6 energy thresholds (1, 2, 4, 10, 30, 60 Mev) from the IMP 7 and IMP 8 Charged Particle Measurement Experiment (CPME; Principal Investigator S.M. Krimigis, then R.B. Decker) are included in OMNI and OMNI 2 for the period January 1, 1973, through the end of 2005 shortly after which IMP 8 operations terminated. The data were prepared and provided by CPME CoInvestigator T.P. Armstrong and colleagues at U. Kansas and Fundamental Technologies, LLC. The instrument and data are further described at http://sdwww.jhuapl.edu/IMP/imp_index.html and at http://imp.ftecs.com/indexx.html.
Fluxes of protons above 1, 10, 30 and 60 MeV for mid1967 through the end of 1972, from the JHU/APL Solar Proton Monitoring Experiment (SPME) on IMP 4 (1967/150  1969/123) and IMP 5 (1969/172  1972/358) were added to OMNI 2 shortly after its creation. The fluxes were computed at NSSDC from count rates provided on tape to NSSDC decades earlier. The values are not reliable absolute measures of quiet time galactic fluxes, but are good for solar and shock accelerated particles. cf. Williams and Bostrom, J. Geophys. Res.,74, 3019, 1969.
Fluxes of protons above 10, 30 and 60 MeV, as measured by NOAA's geosynchronous GOES 11 spacecraft for 20062010 (cf. http://goes.ngdc.noaa.gov/data/avg/) and from GOES 13 for 2011 and later (cf. http://satdat.ngdc.noaa.gov/sem/goes/data/new_avg/), were added to OMNI 2. (The GOES 13 data added to OMNI 2 are actually averages over the fluxes given at NGDC for eastwardlooking and westward looking sensors.) Principal Investigator for the GOES energetic particle instruments is currently T. Onsager, and key responsible NGDC person is D. Wilkinson. Comparisons of IMP 8, GOES 10 and GOES 11 proton flux values obtained between 1999 and 2005 show reasonably good agreement during solar particle flux events; see http://omniweb.sci.gsfc.nasa.gov/ftpbrowser/flux_ogg.html.
Solar and geomagnetic activity indices: OMNI 2 contains daily sunspot numbers (Rz) assigned to each hour of the relevant day, hourly AE and Dst indices and 3hourly Kp values assigned to each hour of the relevant 3hour interval. Note that, in the OMNI toOMNI2 transition, AE is added and C9 is deleted. (For Kp whose normal values are of the form 1+, 2, 2, 2+..., we have used 13, 17, 20, 23,....)
The AE and Dst indices are computed at and obtained from the World Data Center for Geomagnetism, operated by the Data Analysis Center for Geomagnetism and Space Magnetism at Kyoto University, Japan. See http://swdcwww.kugi.kyotou.ac.jp/. Definitive time series of AE, Dst and other indices reach back to 1957 on the Kyoto web site. There are provisional values of AE and Dst also available from Kyoto for periods after the definitive time series that end in mid1988 (AE) and mid2002 (Dst). Definitive hourly values of AE and Dst are included in OMNI 2 from 1963 to their ends and are extended when possible. Provisional Dst values are included in OMNI 2 as a continuation of our practice started with OMNI years ago. Given their ready accessibility from Kyoto, we have not included the provisional AE values in OMNI 2. We thank Drs. T. Iyemori and T. Kamei for permission to include these indices in OMNI 2.
The Rz and Kp values are downloaded from NOAA's National Geophysical Data Center. See ftp://ftp.ngdc.noaa.gov/STP/GEOMAGNETIC_DATA/INDICES/
It is desirable that OMNI 2 data be as free of "bad data" as possible. Extensive checking of 1971current magnetic field and plasma data was carried out at NSSDC as part of creating the OMNI 2 data set. Several webbased tools were created and used to review input data sets individually and as compared to each other. Creation and use of such tools was not feasible in OMNI's infancy nearly 30 years ago.
There are two sources of "bad data" in OMNI 2. One is "noise points" in constituent data sets that may arise from transient instrument malfunction or, in the case of plasma parameters, from the time variation of plasma during the accumulation of one distribution function whose subsequent analysis for determining bulk parameters yields meaningless values. Such noise points typically yield singlepoint upward or downward spikes in parameter profiles. While instrument teams have removed most such points in the data they provided to NSSDC, it has been beneficial to seek out such points here.
The other main source of bad data in OMNI 2 is the inappropriate inclusion of magnetosheath field or plasma data in a data set intended as solarwindonly. This is more significant, the more times the source spacecraft crosses the Earth's bow shock. Thus it is insignificant for ACE which went into an upstream libration point orbit immediately after launch, very significant for IMP 8 in its ~12day geocentric orbit, and significant to an intermediate extent for Wind with its more complex and time varying orbit.
The data sets provided to NSSDC and used in OMNI and OMNI 2 were nominally for solar wind periods only. One exception was that, since about 1994 and by agreement between NSSDC and the IMP 8 magnetometer team, IMP 8 magnetic field hourly data were time continuous (exclusive of data gaps that were independent of orbit phase). These were filtered at NSSDC by MIT's plasmadatabased identification of solar wind intervals prior to adding to OMNI.
Magnetosheathcontaminated magnetic field data typically has higher field intensity and variance levels than concurrent data from a nearby spacecraft solely in the solar wind. Magnetosheathcontaminated plasma data typically shows lower flow speed, higher density and temperature and flow direction significantly further from helioradial than concurrent data from a nearby solarwindonly spacecraft. These signatures are not equally clear in all cases. Further, most of these changes seen upon crossing the bow shock may also be seen upon crossing solar wind structures, especially interplanetary shocks. Thus, they are not always unambiguous markers of bow shock crossings.
Many datasetspecific browse interfaces to OMNI 2contributing highresolution data sets from IMP 8, ISEE 3, Wind and ACE are available at http://omniweb.gsfc.nasa.gov/ftpbrowser/. These were typically used in dataset cleaning after suspicious hourly averages were found with some of the tools below.
Among the tools developed and used for concurrently screening multisource data and summarized at http://omniweb.sci.gsfc.nasa.gov/ftpbrowser/merged.html were: http://omniweb.sci.gsfc.nasa.gov/ftpbrowser/mag_iwa.html for plotting magnetic field intensity or components from IMP 8, Wind and ACE for 19942000; http://omniweb.sci.gsfc.nasa.gov/ftpbrowser/pla_iwa.html for plotting plasma parameters and/or variances from any pair of the 5 data sets IMP8/MIT, IMP8/LANL, Wind/SWE (fits based), Wind/SWE (momentsbased) or ACE/SWEPAM for 19952001; http://omniweb.sci.gsfc.nasa.gov/ftpbrowser/pla_lm.html for plotting plasma parameters and/or variances from IMP8/MIT and IMP8/LANL for 19732001; http://omniweb.sci.gsfc.nasa.gov/ftpbrowser/pla_ii.html for plotting plasma parameters and/or variances from IMP8/MIT, IMP8/LANL and ISEE 3 for 19781982; http://omniweb.sci.gsfc.nasa.gov/ftpbrowser/pla_imp_678.html for plotting plasma parameters and/or variances from IMP6/LANL, IMP7/LANL and IMP8/LANL for 19711978
These tools all work with timeshifted (see below) hourlyaveraged data. They yield plots with one panel per physical parameter selected, with colorcoded intensitytime profiles from each of the data sources. They make noise points visible and they make intervals visible when the apparent solar wind behavior at multiple sources differs significantly. Some cases of the latter correspond to one source being magnetosheath contaminated while other cases may correspond to real differences in the solar wind plasma domains seen at the two spacecraft.
Another family of tools was also developed that generates scatter plots of parameter values from various pairs of plasma data sources. (These tools also determine regression fits as will be further discussed in the later data comparison and cross normalization sections of this paper.) This set of tools makes outliers very visible and has led to identification of several magnetosheathcontaminated datahours. These tools include
http://omniweb.sci.gsfc.nasa.gov/ftpbrowser/mag_iwa_s2.html for 1994latest_available IMP 8, Wind, ACE
and Geotail magnetic field parameters
http://omniweb.sci.gsfc.nasa.gov/ftpbrowser/pla_iwa_s2.html for 19952001 IMP 8, Wind and
ACE plasma parameters
http://omniweb.sci.gsfc.nasa.gov/ftpbrowser/pla_lm_s2.html for 1995latest_available IMP 8, Wind, ACE
and Geotail plasma parameters
http://omniweb.sci.gsfc.nasa.gov/ftpbrowser/pla_ii_s2.html for 19781982 IMP8/MIT,
IMP8/LANL and ISEE 3 plasma parameters.
http://omniweb.sci.gsfc.nasa.gov/ftpbrowser/pla_imp_678_s2.html for 19971998 LANL IMP 6, 7
and 8 plasma parameters.
Replacing the "s2" in these url's with "s3" in the last four of these gives a variant of the pages for working with base10 logarithms of densities and temperatures rather than with densities and temperatures themselves. These tools allow filtering by values of any of the parameters in the relevant data records, including numbers of fine scale points in the houraverages. The second of them also allows filtering by the impact parameter (transverse separation distance, see below) between any pair of spacecraft.
For many years, the IMP 8 spacecraft was the only OMNI source, and is the dominant OMNI 2 data source from its late1973 launch to the mid1978 ISEE 3 launch and again from the late1982 departure of ISEE 3 from an L1 orbit until the late1994 Wind launch. In its 12day, 35Re nearcircular geocentric orbit, IMP makes at least 2 and frequently 1020 transitions into and out of the solar wind, across the Earth's timevarying bow shock. To enable a more reliable exclusion of magnetosheathcontaminated IMP 8 field or plasma data, and a more reliable inclusion of interesting solar wind intervals (that might once have been excluded by the magnetic field or plasma teams in their NSSDC submissions of hourly solar wind data as being magnetosheathcontaminated), a major effort was undertaken (with support from a NASA/AISRP grant) by the IMP 8 magnetic field team at GSFC and the plasma team at MIT to jointly study the field and plasma data and to identify and characterizeall IMP 8 bow shock crossings. The fruits of this effort is visible at http://omniweb.sci.gsfc.nasa.gov/ftpbrowser/bowshock.html. To date (August, 2003), the effort has been completed for 19772000.
The bow shock crossing database was used to create a file flagging IMP 8 hours as corresponding to (1) IMP wholly in the solar wind, (2) IMP making one or more bow shock crossings, (3) other hours. We have used this file along with an earlyversion OMNI 2 file to identify and delete magnetic field or plasma data when IMP 8 was not wholly in the solar wind. (The exception was that, for the case of LANL plasma data wherein hourly averages were created from ~2min data previously separated at LANL as being in the solar wind vs. magnetosheath, LANL hourly averages were retained in OMNI 2 for hours in which IMP encountered shock crossings and was therefore partly in the solar wind.)
Further, the (a) bow shockdatabasederived, regionflagging file was used with (b) OMNI 2 and with (c) a file of IMP 8 magnetic field hourly averages created at NSSDC from 15s, allorbitphases magnetic field data to find hours when IMP was in the solar wind (as per (a)), there were no IMF data in OMNI (as per (b)), and there were IMP IMF data (in (c)). For such hours, IMP 8 IMF data were added to OMNI 2 from (c). For 19772000, there were almost 2300 such hours. An equivalent effort was made to capture extra solar wind plasma data, but only 29 notpreviouslyincluded hours were found.
In this section we address why, when and how we time shift data of ~minute resolution before building hourly averages for inclusion in OMNI 2.
Why and when to shift: That most of the source spacecraft contributing to OMNI and OMNI 2 make IMF and plasma observations minutes upstream of the magnetosphere (e.g., <= 15 minutes for the moonorbiting, late1960's Explorer 35 spacecraft at ~60 Re) was not factored into the hourly averages interspersed into OMNI, nor is it for OMNI 2. However, the ISEE 3, Wind and ACE spacecraft are frequently or always about an hour upstream of the magnetosphere. As their data are to be interspersed with data from muchclosertoEarth spacecraft (e.g., IMP 8), it is appropriate to timeshift the hourupstream data at higher resolution and to compute hourly averages "at Earth" for inclusion in OMNI and OMNI 2. Such shifting has been done for the field and plasma data of these three spacecraft, as described herein.
How to shift: Several factors determine optimal shifts: the geometry of the Earth spacecraft separation vector; the Earth's orbital motion about the sun between observations upstream and at Earth; the geometry (shape, orientation) of the solar wind variation phase front; the solar wind flow direction; and local propagation of the phase front relative to the mean solar wind. (In the above, "Earth" can be replaced by "second spacecraft" for time shifts made for twospacecraft comparisons.) Weimer et al (2002) have shown optimal shifts differ significantly from interval to interval, and it is best to analyze the available field and plasma data for each interval to determine its optimal shift. However, for the purpose of shifting many years of upstream data for OMNI 2, we seek a statistically optimal approach. Relative to the above factors, we assume the variation phase fronts are planar, of arbitrarily large extent normal to the SunEarth line and normal to the ecliptic, and that they merely convect outward with a solar wind flow assumed radial. It remains to specify the angle between the SunEarth line and the intersection between the phase front and the ecliptic plane.
It is useful to introduce the concept of impact parameter (IP) as the distance by which a plasma element, flowing radially from the sun with speed V and observed by one spacecraft misses being seen by a downstream spacecraft (or Earth). Simple geometrical considerations show that for bodies indexed by i and j and located at (Xi, Yi, Zi) and (Xj, Yj, Zj),
IPij = SQRT {[(YiYj)+(XiXj)*Ve/V]**2 + (ZiZj)**2} Timeseries plots of IPij for various combinations of Earth, IMP 8, ISEE 3, Wind and ACE are available (given that Ve = 30 km/s and assuming V = 428 km/s) at http://omniweb.sci.gsfc.nasa.gov/ftpbrowser/impact_ii.html and at http://omniweb.sci.gsfc.nasa.gov/ftpbrowser/impact_iwag.html. IP values were used as filters in doing data set pair regressions, as will be discussed later. For solar wind variation phase fronts normal to the ecliptic (no ZiZj dependence), further geometric considerations say that the time delay equation for one spacecraft and Earth is Deltat = (X/V) * {[1 + (Y*W)/X]/[1  Ve*W/V]}, where Deltat is the time shift in seconds, X and Y are GSE X and Y components of the spacecraft position vector, in km, V is the observed solar wind speed in km/s (assumed radial), Ve is the speed of the Earth's orbital motion (30 km/s), The W parameter is related to the assumed orientation of the phase front relative to the Earthsun line. The orientations to be considered correspond to cases of convection, corotation, and "halfwayinbetween."
In first shifting ISEE 3 data for inclusion in OMNI many years ago, we showed that shifting by simple convection (assuming planes that were normal to the ecliptic plane and to the earthsun line) and by "corotation" (assuming not normality to the Earth sun line but alignment with the ideal IMF Parker spiral angle) were equally good statistically, and that both were significantly better than having no time shift. We somewhat arbitrarily chose to use the corotation shift. This is extensively discussed at http://omniweb.gsfc.nasa.gov/om_book/sup3/s3_main.html.
Subsequent to that early analysis, further study (Richardson and Paularena, 1998) revealed that best statistical correlation was obtained by assuming phase fronts were part way between the convection and corotation orientations considered in adding ISEE 3 data to OMNI. So in preparing ISEE 3, Wind and ACE data for OMNI 2, we have made the assumption that solar wind variation phase fronts are planar, normal to the ecliptic plane and intersecting the ecliptic plane along a line exactly half way between the ideal Parkerspiral and the normal to the SunEarth line (i.e., to the GSE Y axis).
For the three cases considered, we have for the W parameter introduced above: Convection: W=0 Corotation: W=V/428 "Halfway": W=tan [0.5 * atan (V/428)]where 428 (km/s) is the speed in the () GSE Y direction at Earth of an ideal IMF spiral line due to solar rotation.
The upstream orbits: Let us review the ISEE 3, Wind and ACE orbits briefly. From shortly after its 8/12/78 launch until August, 1982, when it was directed towards the Earth's deep magnetotail, ISEE 3 was in an L1 libration point orbit with X in the range ~200260 Re, Y in the range ~ +/ 100 Re, and Z in the range ~ 15 Re to + 20 Re. At its extremes (X ~ 220 Re, Y ~ +/ 100 Re, Z ~ 0), the impact parameter (IP, see above) values for ISEE 3 relative to Earth were ~ 83 Re and ~117 Re, where the asymmetry results from the Earth's motion towards Y during the ~hour that the solar wind moves from Xisee to Xearth. Note that the time shifts for ISEE 3 could range between ~25 min for high flow speed (700 km/s) and Ygse = 100 Re and ~ 80 min for low speed (350 km/s) and Ygse = +100 km/s.
Wind has been in a variable orbit since its 11/01/94 launch. Through 1998, Wind executed a series of ~30 geocentric orbits with nearnoon apogees of distances ranging between ~50 Re and ~250 Re and periods ranging between ~20 days and ~150 days. During these years, the Y component of the Wind position vector was typically in the range +/ 40 Re and almost never exceeded the range +/ 60 Re. For 1999 through the first half of 2000, Wind had three orbits reaching X values of 210, 180 and 100 Re, but otherwise many Wind apogees were of lower altitude and well away from the noon meridian. Starting in mid2000, Wind was put into an orbit reaching extreme values of +/ 250260 Re in the dawndusk meridian. After some time in this orbit, Wind was placed in an L1 orbit. Figure 4 shows the WindEarth impact parameter for 19942003.
ACE has been in a regular L1 orbit since shortly after its 08/25/97 launch, with X in the range ~218248 Re, Y in the +/40 Re range and Z in the +/ 24 Re range. The ACE project was assessing orbit adjustments in 2003.
Doing the time shifts: We have shifted 15 min ISEE 3, Wind and ACE IMF and plasma data using the above timeshift equation, using known locations of those spacecraft and using observed solar wind flow speeds in the data sets being shifted. We then created averages over all fine scale values whose shifted time tags placed them within a given hour "at Earth." Thus all the values with shifted time tags between 00:00 and 01:00 were averaged to give the first OMNI 2 average for a day.
The high resolution data sets that were fed into the time shift algorithm were:
ISEE 3 2min merged IMF/plasma data at
http://omniweb.sci.gsfc.nasa.gov/ftpbrowser/isee3_merged.html.
Wind/SWE 92s plasma data at
http://omniweb.sci.gsfc.nasa.gov/ftpbrowser/wind_swe_2m.html.
Wind 1m IMF data at
ftp://nssdcftp.gsfc.nasa.gov/spacecraft_data/wind/mag/1min_ascii/.
ACE 4min merged IMF/plasma data at
http://omniweb.sci.gsfc.nasa.gov/ftpbrowser/ace_merge.html.
For ISEE 3, a given fieldplasmamerged record was given a shifted time tag only if it had a flow speed value to use in the time shift equation above. If the record had no flow speed value, then its IMF data, if present, were not shifted nor otherwise carried forward for inclusion in OMNI 2. However, for both ACE and Wind, IMF data were shifted using a flow speed interpolated to the IMF record time tag with input from the closestbefore and closestafter good flow speed values, regardless of the duration between the two input points used. (The treatment of ACE data in this regard was changed from being ISEE3like to being Windlike in February, 2006.) Since plasma gaps may have been over many hours, users of shifted IMF data may want to assess the goodness of their time tags by examining whether there are concurrent plasma data and, if not, how long a plasma gap (during which flow speeds might have varied significantly) there was.
The above ISEE 3 and ACE merged data sets were created at NSSDC from data obtained from ISEE Principal Investigators and from the ACE Science Center.
A special case should be noted. After the February, 1980, failure of the ISEE 3 ion plasma instrument, only LANLprovided ISEE 3 hourlyaveraged electronbased parameters are available. For OMNI, years ago, we timeshifted these using the corotation delay algorithm, then built Earthtime "hourly averages" as weighted averages of any hourly averages falling in part into the Earthhour of interest. See http://omniweb.gsfc.nasa.gov/om_book/sup3/s3_main.html for further detail. For OMNI 2, we have taken the even simpler approach of shifting each ISEE 3 electron based density and flow speed in the LANLprovided hourly data set by one hour. No new "halfway" time shift of these electronbased parameters was done.
The hourly averages determined from the shifted field and plasma ISEE 3, Wind and ACE data are available, along with concurrent but unshifted IMP 8 data, at http://omniweb.sci.gsfc.nasa.gov/ftpbrowser/merged.html.
[2008: All comparisons of Geotail data with ACE, Wind, or IMP 8 data are addressed at http://omniweb.gsfc.nasa.gov/html/HROdocum.html. (HRO is High ResolutionOMNI.) Likewise all comparisons of Wind/SWE key parameter data with Wind/SWE nonlinearfits data are addressed there. Finally, the data set comparisons and cross normalizations section of that reference has results for years later than those addressed herein. Such recentyear cross normalizations have been applied to both HRO and to hourly OMNI 2.]
There are random and systematic differences between hourly averages of pairs of like parameters obtained by two spacecraft. Among the reasons for the random differences may be (1) the two averages have differently timelocated gaps in the averages, (2) spatial gradients in parameters being measured combined with offsets of the spacecraft locations relative to the flow direction, (3) incorrect (or no) time shifts used for one or another data set prior to hourlyaverage construction (see prior section), (4) etc. Among the reasons for systematic differences are differing processing approaches (e.g., taking fits vs. moments for deriving flow parameters from distribution functions), subtle calibration factors not adequately accounted for in data processing, etc.
As OMNI 2 involves the interspersal of IMF data and of plasma data from each of several spacecraft, it is desirable to understand and to compensate for the differences between pairs of sources. It is not feasible to decrease random differences between pairs of sources (except via identification and exclusion of "bad data" as discussed earlier), but it is valuable to understand their magnitude in order to understand the "accuracy" of the OMNI 2 data as representative of the nearby solar wind. It is feasible to find and compensate for systematic differences between pairs of data sets.
This section addresses differences found between pairs of IMF data sets and of plasma parameter data sets, while the next section addresses the compensation of these differences through normalizations of some plasma parameters to a fiducial plasma data set.
In preparing OMNI 2, we have done detailed new comparisons involving IMP 8, Wind and ACE IMF and plasma data, and we report the results in detail as these are the main sources of the past 20 years of data. We have also done new comparisons of plasma data from IMP 8, ISEE 3 and IMP 6 and 7 which we shall report in somewhat less detail. Recall that IMP 8 carries two independent plasma experiments, from LANL and from MIT, and we also did detailed comparisons of these.
Magnetic field comparisons: Using the scatter plot and regression fit interface at
http://omniweb.sci.gsfc.nasa.gov/ftpbrowser/mag_iwa_s2.html, we have done several runs of the
form Bi,j = a + b * Bi,k, where i designates a field component (GSE X, Y, Z) or field
magnitude ( <B>, for which i=M below), where j, k designate a spacecraft pair and
where a and b are intercept and slope of a linear regression fit determined by
minimizing sums of squares of perpendicular distances between (Bi,j, Bi,k) data points
and the best fit line. Use of the "perpdist" regression approach rather than the
"deltaY" regression approach is more appropriate to cases where the uncertainties or
errors in the Y and X variables are comparable. In all the runs, we required at least
half an hour's coverage in each hourly average by requiring at least 120 fine scale
(15s) IMP points, 30 1min Wind points and/or 10 4min ACE points. We also looked for
time dependence in the IMPWind regression. Units of all B values below are nanoTeslas.
[As of November, 2011, these results had not been updated to reflect the reprocessing
of IMP 8 or Wind magnetic field data. However, the data set intercomparison software at
the url's identified above presently address the reprocessed data, such that interested
parties can run on the latest data.]
For 6388 points between November 21, 1994 and December 31, 1996, we found B (Wind, X) = .018 (+/ .008) + 1.021 (+/ .004) * B (IMP, X) B (Wind, Y) = .002 (+/ .011) + 1.017 (+/ .004) * B (IMP, Y) B (Wind, Z) = .014 (+/ .013) + 1.016 (+/ .007) * B (IMP, Z) B (Wind, M) = .008 (+/ .016) + 1.016 (+/ .003) * B (IMP, M) For 7558 points between January 1, 1997 and July 27, 1999, we found B (Wind, X) = .045 (+/ .010) + 1.016 (+/ .003) * B (IMP, X) B (Wind, Y) = .012 (+/ .010) + 1.017 (+/ .003) * B (IMP, Y) B (Wind, Z) = .045 (+/ .012) + 1.023 (+/ .004) * B (IMP, Z) B (Wind, M) = .026 (+/ .015) + 1.020 (+/ .002) * B (IMP, M) For 10,090 points between February 5, 1998 and July 29, 1999, we found B (Wind, X) = .080 (+/ .007) + 0.985 (+/ .002) * B (ACE, X) B (Wind, Y) = .010 (+/ .008) + 0.997 (+/ .002) * B (ACE, Y) B (Wind, Z) = .053 (+/ .009) + 0.993 (+/ .003) * B (ACE, Z) B (Wind, M) = .020 (+/ .012) + 0.988 (+/ .001) * B (ACE, M) For 6864 points between February 5, 1998 and June 6, 2000, we found B (IMP, X) = .051 (+/ .011) + 0.972 (+/ .003) * B (ACE, X) B (IMP, Y) = .019 (+/ .012) + 0.983 (+/ .003) * B (ACE, Y) B (IMP, Z) = .020 (+/ .014) + 0.980 (+/ .004) * B (ACE, Z) B (IMP, M) = .062 (+/ .011) + 0.969 (+/ .001) * B (ACE, M)
In the range +/ 10 nT, where most IMF values lie, the difference between an observed value and any of the values computed from the above equations is almost always within 0.2 nT of zero. 0.2 nT is much less that the natural spread of data points about the best fit regression lines (see representative scatter plots at Figure 5 and Figure 6), so we shall not perform any crossnormalizations of magnetic field data.
Prior cross comparisons of multisource magnetic field data were given in OMNIWeb accessible documentation. It was determined that the only IMF normalization needed was for the sunpointing (spinaxis aligned) IMF component of the Prognoz 10 IMF vector. This was done for OMNI, and will be continued unchanged in OMNI 2.
Wind/SWE vs. ACE/SWEPAM(OLD version, we do not use these results scince Feb. 20013)
It is likely that comparisons among currently active
Wind/SWE and ACE/SWEPAM plasma data sources will be of most current interest as
we write this note in 2003. Accordingly, we have generated an extra detailed discussion
which, to minimize interruption to the flow of this note, we place as Appendix 1 to this
note. Summarizing the results of the analyses reported and discussed in Appendix 1,
we have:
V swe = 2.135 (+/ .387) + 1.010 (+/.001) * V ace V < 350 km/s: (log N) swe = .006 + (log N) ace 350 < V < 450: (log N) swe =.091 + 1.036 * (log N)ace V > 450 km/s: (log N) swe =.082 + (log N) ace V < 350 km/s: (log T) swe = 1.680 + 1.348 * (log T) ace 350 < V < 450 (log T) swe = 0.287 + 1.067 * (log T) ace V > 450 km/s (log T) swe = 0.417 + 1.091 * (log T) aceThe speed equation is close enough to Y = X to warrant no normalization. The density equations say that N swe/N ace = 1.01 at V < 350 km/s, is 0.83 at V > 450 km/s and ranges between 0.83 at N = 2 to 0.91 at N = 20 in the 350450 km/s flow speed range. The temperature equations say that T swe is typically larger than T ace by 525%.
IMP 8 (MIT & LANL) vs. Wind/SWE: For the Wind era, we have done many comparisons of IMP 8 MIT and LANL plasma parameters to each other and to Wind/SWE data. These are reported in Appendix 2.
We repeat here just the qualitative IMP/MITtoWind and IMP/LANLtoWind results:
In general V swe > V imp. The difference V swe  V imp increases from 1.0 to 10.5 km/s ("imp" = imp/mit) or from 4.6 to 7.6 km/s ("imp" = imp/lanl) as V imp increases from 300 km/s to 800 km/s.
SWE densities are less than IMP densities by varying extents up to 1520%, except that for slow flows (V < 350 km/s), SWE densities exceeded IMP/LANL densities by ~ 10%.
SWE temperatures are comparable to IMP/LANL temperatures except for slow flows for which Swe temperatures are less than LANL temperatures by 5% (at T = 25,000 deg K) to 30% (at T = 400,000 deg K). SWE temperatures are greater than IMP/MIT temperatures by up to a few tens of percent, except for relatively infrequent high temperature, slow speed cases.
IMP8/MIT vs. IMP8/LANL: IMP 8 plasma data span a 28year, 19732001 interval. We have looked for time variation in the regressions between the two data sets, and have found it useful and appropriate to do density and temperature regressions in four separate time intervals: 19731979, 19801981, 19821994 and 19952001. We note that the 19801981 interval was highlighted in the analysis of Petrinic and Russell (1993) as having anomalously low Nlanl/Nmit ratios.
ISEE 3 vs. IMP 8 Comparisons: ISEE 3 was a very significant nearEarth solar wind monitor from shortly after its 1978/08/12 launch until late 1982 when it was moved from its L1 orbit to probe the deep geomagnetic tail.
The LANL plasma experiment on ISEE 3 separately measured ions and electrons. Unfortunately the ion instrument failed on February 19, 1980. We include in OMNI 2 (as we did for OMNI) ionbased density, flow speed, temperature and flow azimuth information through 1980/02/19, but only electronbased density and flow speed thereafter.
Appendix 4 details the results of comparisons of IMP 8 data with ISEE 3 ionbased and electronbased parameters. We show there that ISEE ionbased flow speeds agree fairly well with IMP speeds, the largest difference in the 300800 km/s range being ~ 6 km/s at 300 km/s for ISEE 3 vs. IMP/MIT. The agreement of electronbased flow speeds with IMP is not as good. V isee is less than V imp by 515 km/s (IMP/LANL) or 710 km/s (IMP/MIT) over the 300800 km/s range.
ISEE ionbased densities are less than IMP/MIT densities , except for slow flows (V < 350 km/s), with the ratio of (N isee)/(N imp) falling significantly with flow speed. On the other hand, ISEE electronbased densities are consistently greater than IMP/MIT densities. The (N isee)/(N imp) ratio increases from a few percent at N = 2/cc to ~ 25% at N = 20/cc, independent of flow speed.
ISEE ionbased temperatures are higher than IMP/MIT temperatures, by several tens of percent at low temperature to only ~10% at high temperature. We do not compare ISEE electronbased temperatures with IMP ionbased temperatures, nor do we include any ISEE electronbased temperatures (or flow angles) in OMNI 2.
LANL IMP 6, 7, 8 comparisons: There were LANL plasma instruments on IMP 6 (19711974), IMP 7 (19721978) and IMP 8 (19732001), all of which produced good data. As noted earlier, we have abandoned the earlier LANLprovided hourly/3hourly resolution merged IMP 678 data set that contained only ion densities, flow speeds and temperatures, in favor of computing new hourly averaged spacecraftspecific data sets from the ~2min LANLprovided IMP 6, IMP 7 and IMP 8 data sets containing a greater range of parameters. LANL had originally built these 2min data sets to contain only solar wind data, so we were able to build LANL hour averages for hours with bow shock crossings.
Appendix 5 contains the results of comparisons across spacecraft pairs. The main points of the Appendix are that IMP 7 and IMP 8 flow speeds and temperatures agree extremely well, while IMP 7 densities are on the order of 20% higher than the IMP 8 densities. IMP 6 and IMP 8 densities agree with each other to within 10%, as do IMP 6 and IMP 8 temperatures. IMP 6 and IMP 8 flow speeds agree near the middle of the 300800 km/s speed range, but disagree near the ends of that range by about 12 km/s.
[2008: Crossnormalizations for years more recent than those addressed below, and for Wind/SWE/KP data relative to Wind/SWE nonlinearfits data are found in http://omniweb.gsfc.nasa.gov/html/HROdocum.html.]
Now, we must use all the twospacecraft flow speed, density and temperature regressions of the previous section to normalize data appropriately. This is so that the multi source OMNI 2 data set will have least impact of switching from using data from source x in hour n to using data from source y in hour n+1. Also, in this process, we would like to produce a data set whose parameters are most likely to be absolutely true and not useful merely for studies of relative changes and correlations.
In the foregoing section, we showed that, with rare exceptions, flow speed regression lines are typically within a few km/s of the Y = X line over the 300800 km/s range, so we have not normalized any flow speed data.
On the other hand, density and temperature data frequently diverge from Y = X by more than a few percent, so we shall normalize all densities and temperatures for the 19712003 period that we have examined in detail. We shall use the time dependence of the 28year IMP8/MIT  IMP8/LANL regression, but otherwise there will be no time dependence in the normalization of any given spacecraft. To be more explicit, we assume all time dependence in the IMP 8 LANL/MIT regression is due to IMP/LANL, so the IMP8/MIT normalization will itself be time invariant.
We shall also use the flow speed bins <350 km/s, 350450 km/s and >450 km/s in performing normalizations as in the previous comparisons.
The key question is which data set to use as the fiducial data set. After interactions with the MIT and LANL plasma teams, we have decided to use the Wind/SWE data set as the fiducial data set. MIT's Dr. Justin Kasper wrote an extensive Ph.D. dissertation on SWE data that included comparisons of (a) SWE proton and alpha particle densities derived from nonlinear bimaxwellian fits to (b) electron densities computed from the Wind/Waves/TNR radio instrument. Folding in theorybased contributions of electrons from Z>2 ions, he concluded that the SWE ion densities are good to within 2%. The relevant sections of his dissertation are available from ftp://spdf.gsfc.nasa.gov/pub/data/wind/swe/ascii/2min/thesis.pdf.
There is no equivalent compelling reason for choosing any specific temperature data set to which to normalize all other temperature data sets. So, given the choice of Wind/SWE as providing the fiducial density data set, we also choose Wind/SWE as providing the temperature data set to which we will normalize all others.
Another key assumption is that there has been no significant time variation in IMP8/MIT density and temperature not previously found and compensated for by the MIT team. This enables us to normalize IMP 8 and ACE/SWEPAM data to Wind/SWE data and then to normalize all other data sources (not contemporaneous with Wind/SWE) by chaining the regressions of each such data set to IMP 8 with the IMP8Wind/SWE regressions.
The normalization parameters will be specified as pairs (a b) where a and b are the intercepts and slopes in the equations
(logN)norm = a + b * (logN)obsvd
(logT)norm = a + b * (logT)obsvd
In most of the lines below, the three (a, b) pairs are for the flow speed bins < 350 km/s, 350450 km/s and >450 km/s.
In the other lines below, having 4 parameters in each of three parentheses, we show the ratios of normalizedtoobserved density values at 4 points (N = 2, 5, 10, 20 /cc) Again the three parentheses correspond to the three speed bins
The normalization parameters for ACE/SWEPAM(OLD version, we do not use these results scince Feb. 20013) are N: (.006 1.000) (.091 1.036) (.082 1.000) T: (1.680 1.348) (.287 1.067) (.417 1.091) N (1.01 1.01 1.01 1.01) (.83 .86 .88 .90) (.83 .83 .83 .83) The normalization parameters for IMP8/MIT are N: (.020 .941) (.033 .919) (.019 .907) T: (.864 .839) (.491 .920) (.702 .890) N (1.01 .95 .91 .88) (1.02 .95 .90 .85) (.98 .90 .84 .79) [Note added June, 2005: The analysis of Merka et al, Pl. Sp. Sci., 53, p 79, includes an independent comparison of Wind/SWE and IMP8/MIT plasma densities. Grouping data for all flow speeds, considering 19952000, and assuming a fit of the type N(Wind)/N(IMP8) = c + 1/(a + b*N(IMP)), they find N(Wind)/N(IMP8) = 0.73 + 1/(1.86 + 0.29*N(IMP)) At N(IMP8) = 2, 5, 10 and 20 cm**3, this ratio is 1.14, 1.03, 0.94 and 0.86, respectively. These ratios are directly comparable to the N(norm)/N(obsvd) values of the preceding line. They are clearly larger than our ratios above, except at N ~ 20 cm**3. Use of the Merka et al equation for normalization of IMP8 plasma data for OMNI 2 would yield higher IMP8based pressures in OMNI 2. These might be significant for the pre1995 era, when OMNI coverage is dominated by IMP8 data. We choose not to comment at this time on the likely correctness of their equation relative to ours, but simply offer it to OMNI users for their awareness.] The normalization parameters for IMP8/LANL are for densities 19731979 (.111 .943) (.064 .951) (.011 .958) 19801981 (.140 1.024) (.092 1.010) (.028 1.048) 19821994 (.064 .965) (.020 1.000) (.085 1.006) 19952001 (.040 1.007) (.023 1.013) (.093 1.022) 19731979 (1.24 1.18 1.13 1.09) (1.12 1.07 1.04 1.00) (.95 .91 .89 .86) 19801981 (1.40 1.43 1.46 1.48) (1.24 1.25 1.26 1.27) (.97 1.01 1.05 1.08) 19821994 (1.13 1.10 1.07 1.04) (.95 .95 .95 .95) (.83 .83 .83 .84) 19952001 (1.10 1.11 1.11 1.12) (.96 .97 .98 .99) (.82 .84 .85 .86) for temperatures 19731979 (.621 .879) (.290 .958) (.271 .969) 19801981 (.840 .811) (.482 1.091) (.207 .964) 19821994 (.497 .894) (+.044 .997) (.130 .982) 19952001 (.441 .895) (.044 1.008) (.165 1.036) The normalization parameters for ISEE3 protons are: N: (.059 .855) (.097 .911) (.076 .972) T: (.348 .926) (.007 1.000) (.053 1.016) N (1.04 .91 .82 .74) (1.18 1.08 1.02 .96) (1.17 1.14 1.12 1.10) The normalization parameters for ISEE3 electrons are N: (.019 .888) (.052 .842) (.004 .859) N (.97 .87 .81 .75) (.94 .81 .73 .65) (.92 .80 .73 .66) The normalization equations for IMP6/LANL are N: (.000 1.075) (.037 1.007) (.018 .952) T: (.509 .894) (.308 .950) (.560 .910) N (1.05 1.13 1.19 1.25) (1.09 1.10 1.11 1.11) (1.01 .96 .93 .90) The normalization equations for IMP7/LANL are N: (.050 .983) (.053 .967) (.099 .968) T: (.641 .872) (.359 .942) (.322 .957) N (.88 .87 .86 .85) (.87 .85 .83 .81) (.79 .77 .74 .72)To avoid discontinuities at 350 and 450 km/s in the normalization process, we have defined bands at 340360 km/s and 440460 km/s. Any normalizations of densities or temperatures for hours with speeds in these bands use weighted averages of the two sets of normalization parameters appropriate to either side of the band. For instance for a flow speed of 345 km/s, an averaged pair of regression parameters would be determined in which the <350 km/s regression parameters would be weighted by 75% and the 350450 km/s regression parameters would be weighted by 25%.
Since the original OMNI data set was normalized to the LANL IMP 7 data set, and since the OMNI 2 data set is being normalized to the Wind/SWE data set, the last row of numbers [(N norm)/(N obsvd)] should be close to (N swe)/(N imp7lanl) which should in turn be close to (N omni 2)/(N omni). That is, the renormalization of density data yields a decrease in OMNI 2 densities, relative to OMNI densities, of about 20%. Figure 7 in fact shows 27dayaveraged IMP7normalized OMNI densities and temperatures and Windnormalized OMNI 2 densities and temperatures. The ~20% density decrease is quite evident.
Because there is virtually no overlap between the OMNI and OMNI 2 data prior to the 1971 IMP 6 launch and after that launch, it is impossible to directly crossnormalize the data for those two eras. We have not attempted any new normalization of the pre1971 OMNI data, but have simply written the pre1971 data into OMNI 2.
Spacecraft prioritization for OMNI 2 inclusion
The final point to cover is the rules for selecting which source to use data from, when data from multiple sources are available for a given hour. In OMNI, we chose data from the closertoEarth spacecraft when the choice was between an hourupstream spacecraft and a minutesaway spacecraft. This meant that we chose data from IMP 8 rather than data from ISEE 3, Wind or ACE when both were available. This was in part to support solar windmagnetosphere coupling studies with the more reliable data from the closer spacecraft, that is, the data in which users could have more confidence that the spacecraftobserved data were representative of the field and plasma impinging on the magnetosphere. However, as all current eventoriented studies use higher resolution solar wind data, and as longerterm statistical studies probably do as well with L1 data as with closertoEarth data, we have chosen to prioritize for OMNI 2 the selection of more timecontinuous upstream data over closertoEarth gappier data (i.e., IMP 8 data) when both are available. The smaller number of transitions between sources realized by this prioritization , even given our attempts at uniformizing OMNI 2 through normalizations, may be statistically advantageous.
We choose Wind data when available in the 19941998 interval and ACE data thereafter. We use Wind data as the backup to ACE after 1998 and IMP 8 as the backup to those two for 19942001. (Recall that IMP 8 IMF data will only extend to June 10, 2000.) Note that we frequently add Wind data for most recent times, as these become available somewhat earlier than do ACE data, and then replace Wind data with ACE data when the latter become available. Likewise in the ISEE 3 era, we choose ISEE 3 in preference to IMP 8.
In June, 2007, we made the following change. Because ACE/SWEPAM densities are sometimes unreliable at times of slow solar wind flows, our input ACE data set, obtained from the ACE Science Center, sometimes had flow velocity values but no densities values. This led to gaps in OMNI 2 density and related parameters that could be filled with Wind values. In fact, we searched for 19992007 OMNI 2 records having ACE flow velocities but not densities, and replaced all data in such records with Wind/SWE parameter values if available.
For the frequent choice between IMP8/MIT and IMP8/LANL data, we choose the IMP8/MIT data in part because it has flow latitude data which the LANL data do not have. However for hours when data from both sources are available, we add alphato proton ratios to the OMNI records from the LANL data. There are far more IMP8/LANL data in OMNI 2 than there were in OMNI.
Plasma data from ISEE 1 and from ISEE 3 (for the period when only electronbased parameters were available) were chosen only when no other data were available.
Geotail data were added to OMNI 2 only for hours for which data from IMP 8, Wind and ACE were all unavailable.
Joe King and Natalia Papitashvili
August 11, 2003
REFERENCES Asbridge, J.R., S.J. Bame, W.C. Feldman and M.D. Montgomery, Helium and hydrogen velocity differences in the solar wind, J. Geophys. Res., 81, 2719, 1976. Bame, S.J., E.J. Hones, S.I. Akasofu, M.D. Montgomery and J.R. Asbridge, Geomagnetic storm particles in the highlatitude magnetotail, J. Geophys. Res., 76, 7566, 1971. Bame, S.J., J.R. Asbridge, H.E. Felthauser, J.P. Glore, H.L. Hawk and J. Chavez, ISEEC solar wind plasma experiment, IEEE Trans. Geosci. Elect., GE16, No. 3, 160, 1978a. Bame, S.J., J.R. Asbridge, H.E. Felthauser, J.P. Glore, G. Paschmann, P. Hemmerich, K. Lehmann and H. Rosenbauer, ISEE1 and ISEE 2 fast plasma experiment and the ISEE1 solar wind experiment, Trans. Geosci. Elect., GE16, No. 3, 216, 1978b. Behannon, K.W., Mapping of the Earth's bow shock and magnetic tail by Explorer 33, J. Geophys. Res., 73, 907, 1968. Bonetti, A., G. Moreno, S. Cantarano, A. Egidi, R. Marconero, F. Palutan and G. Pizella, Solar wind observations with satellite ESROHEOS 1 in December 1969, Nuovo Cimento, 46B, 307, 1969. Bridge, H.S., A. Egidi, A.J. Lazarus, and E. Lyon, Preliminary results of plasma measurements on IMPA, Space Research V, North Holland Publ. Co., 1965. Fairfield, D.H., Bow shock associated waves observed in the far upstream interplanetary medium, J. Geophys. Res., 74, 3541, 1969. Fairfield, D.H. and N.F. Ness, IMP 5 magnetic field measurements in the high latitude outer magnetosphere near the noon meridian, J. Geophys. Res., 77, 611, 1972. Fairfield, D.H., Whistler waves observed upstream from collisionless shocks, J. Geophys. Res., 79, 1368, 1974. Feldman, W.C., J.R. Asbridge, S.J. Bame and M.D. Montgomery, Double ion streams in the solar wind, J. Geophys. Res., 78, 2017, 1973. Frandsen, A. M. A., Connor, B. V., Van Amersfoort, J., Smith, E. J., ISEEC vector helium magnetometer, IEEE Trans. Geosci. Electron., GE16, No. 3, 195198, July 1978. Hedgecock, P.C., Magnetometer experiments in the ESROHEOS satellites, Space Science Instrumentation, 1, 74, 1975. Hundhausen, A.J., J.R. Asbridge, S.J. Bame, H.E. Gilbert and I.B. Strong, Vela 3 satellite observations of solar wind ions, J. Geophys. Res., 72, 87, 1967. Kasper, J.C., Ph.D. Dissertation, MIT, 2002. ftp://nssdcftp.gsfc.nasa.gov/spacecraft_data/wind/plasma_swe/2min/thesis.pdf Lazarus, A.J. and K.I. Paularena, A comparison of solar wind parameters from experiments on the IMP 8 and Wind spacecraft, in Measurement Techniques in Space Plasmas, Particles, AGU Monograph 102, Pfaff, Borovsky and Young, editors, pp8590, 1998. Lepping, R. P., M. Acuna, L. Burlaga, W. Farrell, J. Slavin, K. Schatten, F. Mariani, N. Ness, F. Neubauer, Y. C. Whang, J. Byrnes, R. Kennon, P. Panetta, J. Scheifele, and E. Worley, The WIND Magnetic Field Investigation, Space Sci. Rev., 71, 207, 1995. Lyon, E.F., H.S. Bridge and J.S. Binsack, Explorer 35 plasma measurements in the vicinity of the moon, J. Geophys., Res., 72, 6113, 1967. Lyon, E.F., A. Egidi, G. Pizella, H.S. Bridge, J.S. Binsack, R. Baker and R. Butler, Plasma measurements on Explorer 33 (I) interplanetary region, Space research VIII, 99, North Holland Publ. Co., 1968. McComas, D.J., S.J. Bame, P. Barker, W.C. Feldman, J.L. Phillips, P. Solar wind electron proton alpha monitor (SWEPAM) for the Advanced Composition Explorer, Space Sci. Rev., 86, 563, 1998. Mish, W.H., and R.P. Lepping, JGR 1976, IMP 7&8 magnetometer experiment description paper. Ness, N.F., C.S. Scearce and J.B. Seek, Initial results of the IMP 1 magnetic field experiment, J. Geophys. Res., 69, 3531, 1964. Ness, N.F., K.W. Behannon, C.S. Scearce and S.C. Cantarano, Early results from the magnetic field experiment on lunar Explorer 35, J. Geophys. Rtes., 72, 5769, 1967. Neugebauer, M., Initial deceleration of solar wind positive ions in the Earth's bow shock, J. Geophys. Res., 75, 717, 1970. Ogilvie, K.W., L.F. Burlaga and T.D. Wilkerson, Plasma observations on Explorer 34, J. Geophys. Res., 73, 6809, 1968. Petrinic, S.M. and C.T. Russell, Intercalibration of solar wind instruments during the International Magnetospheric Study, J. Geophys. Res., 98, 18,963, 1993. Richardson, J.D. and K.I. Paularena, The orientation of plasma structure in the solar wind, Geophys. Res. Lett., 25, 2097, 1998. Smith, C. W., et al, The ACE Magnetic Fields Experiment, Space Sci. Rev., 86, 613632, June 1998. Stayzhkin, V., E. Yeroshenko and P. Zakhanov, Magnetic field experiment in the Intershock project, in Intershock Project, ed. S. Fischer, p. 202, 1985. Weimer, D.R., et al, Variable time delays in the propagation of the interplanetary magnetic field, J. Geophys. Res., 107, SMP 291, August 2002.
Appendix 1. Comparisons between Wind/SWE and ACE SWEPAM parameters
Introduction: This Appendix is to report results of comparisons of Wind/SWE plasma parameters with ACE/SWEPAM parameters.
[2008: As of 2008, the Wind/SWE nonlinear fitsbased proton parameters (cf. below) had not been determined past late November, 2004. Accordingly we have used in OMNI 2 a cleaned version of the Wind/SWE key parameter data set for later times. See the documentation of thisdata set in the High Resolution OMNI documentation file at http://omniweb.gsfc.nasa.gov/html/HROdocum.html.]
We use hourly averages created at NSSDC from timeshifted 92s SWE parameters provided by MIT some months ago, and we use hourly averages NSSDCcomputed from similarly timeshifted LANL/ASCprovided 64s SWEPAM parameters newly computed by LANL in mid2003.
Flow Speed We get for flow speed regression 11,640 hours and: V swe = 2.135 (+/0.387) + 1.010 (+/.001) * V ace This says V swe  V ace = 1 km/s at 300 km/s and = 6 km/s at 800 km/s. We take this to be close enough to Y = X to not require flow speed normalization. Plasma (proton) densities: For densities, we obtain: All V: (log N)swe = .107 (+/.004) + 1.062 (+/.005) * (log N)ace, 10,894 hrs V<350: (log N)swe = .007 (+/.012) + .999 (+/.013) * (log N)ace, 1340 hrs 350450 (log N)swe = .091 (+/.005) + 1.036 (+/.006) * (log N)ace, 5759 hrs V>450: (log N)swe = .085 (+/.006) + 1.002 (+/.008) * (log N)ace, 3890 hrs It is nteresting that at low speed, we have effectively N swe = N ace. LANL deliberately excluded densities, but retained flow speeds, for many lowspeed hours in the latest version of its 64s data set on which the ACE plasma data in OMNI 2 are based. Density ratios corresponding to these equations are N ace 2 5 10 20 Nswe/Nace All V .816 .864 .902 .941 V<350 1.016 1.015 1.014 1.013 350450 .831 .859 .881 .903 V>450 .823 .825 .826 .827This says that at low speed, the SWE and SWEPAM densities are virtually the same while at high speed, N swe is only ~82% of N ace. At intermediate speed we get a transitional regression, akin at low N to the high speed result and half way between the high and low speed results at high N. Using only the speedindependent regression equation would lead to inappropriately decreasing N ace to match N swe at low speeds and would lead to insufficiently decreasing N ace to match N swe at high speeds (at all but the lowest N). So we shall use the Vdependent regression results in our OMNI2driven normalizations as we have for other data being prepared for OMNI 2.
We next look for evidence of time dependence in the Wind/SWEACE/SWEPAM density normalization by obtaining yearspecific results integrated over full flow speed ranges. We find:
1998: (log N)swe = .122 (+/.006) + 1.070 (+/.007) * (log N)ace, 4621 hrs 1999: (log N)swe = .107 (+/.006) + 1.067 (+/.007) * (log N)ace, 3550 hrs 2000: (log N)swe = .094 (+/.009) + 1.055 (+/.012) * (log N)ace, 1880 hrs 2001: (log N)swe = .109 (+/.013) + 1.098 (+/.018) * (log N)ace, 843 hrs It is risky to infer results from similarities of coefficients, especially from allV results which we showed above could be misleading. So we next look at Nswe/Nace ratios corresponding to these equations: Nace 2 5 10 20 Nswe/Nace 1998 .792 .845 .887 .931 1999 .819 .871 .912 .955 2000 .837 .880 .914 .950 2001 .833 .919 .975 1.044 At the lowest N, these ratios increase for the first 3 years, and the ratio for the 4th year matches that for the 3rd year. At the intermediate and higher N values, the ratio takes jumps from 1998 to 1999 and from 2000 to 2001, but the ratios for 1999 and 2000 are similar to each other. Let's next look for time dependence in the two higher V bins where there are enough points to give reasonable statistics. 350 < V < 450 km/s 1998 (log N)swe = .112 (+/.009) + 1.048 (+/.010) * (log N)ace, 2516 hrs 1999 (log N)swe = .093 (+/.009) + 1.048 (+/.011) * (log N)ace, 1924 hrs 2000 (log N)swe = .092 (+/.017) + 1.052 (+/.020) * (log N)ace, 897 hrs 2001 (log N)swe = .079 (+/.020) + 1.073 (+/.024) * (log N)ace, 422 hrs V > 450 1998 (log N)swe = .107 (+/.010) + 1.023 (+/.014) * (log N)ace, 1301 hrs 1999 (log N)swe = .077 (+/.009) + 0.989 (+/.013) * (log N)ace, 1350 hrs 2000 (log N)swe = .070 (+/.011) + 0.988 (+/.015) * (log N)ace, 877 hrs 2001 (log N)swe = .090 (+/.021) + 1.021 (+/.030) * (log N)ace, 362 hrs and the corresponding Nswe/Nace ratios at Nace = 2, 5, 10 and 20: 350 < V < 450 1998 .799 .835 .863 .892 1999 .835 .872 .902 .932 2000 .839 .880 .912 .945 2001 .877 .938 .986 1.037 V > 450 1998 .794 .811 .824 .837 1999 .831 .823 .817 .810 2000 .844 .835 .828 .821 2001 .825 .841 .853 .866 Consider first the lessspreadout V>450 Nswe/Nace ratios. Only the N=5 column shows a monotonic increase with time. Two years have regression equation slopes >1 and two <1. The range of ratios is limited to .794.866, with most values in the range .80.85. Recall that the 19982001 combined result for V>450 was (log N)swe = .085 (+/.006) + 1.002 (+/.008) * (log N)ace. We believe it would be reasonable to normalize SWEPAM densities to SWE densities in the V>450 range, for all time, using the equation (log N)norm = .082 + (log N)obsvd. This would give ACE Nnorm/Nobsvd ratios of .828 at all time and density values. It would introduce discrepancies of up to ~3% relative to what we'd get with timedependent normalizations at some parts of the relevant span of N values. Now consider the 350 < V < 450 km/s range. Here we see the same behavior in the ratios as we saw in the allV analysis earlier, that is, the ratios at each N jump from 1998 to 1999 and from 2000 to 2001, but are near each other for 19992000. An inclination here is to use one equation for 19992000 (log N)swe = .092 (+/.008) + 1.048 (+/.010) * (log N)ace, 2821 hrs, [ratios (2, 5, 10, 20) = .836 .870 .904 .934] and to use the above two equations for 1998 and 2001. However, let's look at scenarios for suppressing the time dependence. Let's gather the various 350450Vbin ratios for various times: 19982001 .833 .862 .885 .909 5759 hrs 1998 .799 .835 .863 .892 2516 hrs 19992000 .836 .870 .904 .934 2821 hrs 2001 .877 .938 .986 1.037 422 hrs
If we used the 19982001 equation, we'd have "errors" relative to using separate 1998 and 19992000 equations of mostly <2.7% but up to 3.4% for N=2, 1998. There are not many hours with N~2, and a 3.4% error in 2.0 gives 2.07 which is insignificantly different than 2.0.
Using the 19982001 equation for 2001, rather than the 2001specific equation, gives N values "too low" by 4.4%, 7.6%, 10.1% and 12.8% at N = 2, 5, 10, 20. These are getting to be large numbers. However, the equation for 350450, 2001 is based on only 422 hours (because Wind was mostly >100 Re to the side of the Earth from late August, 2000). We have plotted the 19982001 best fit line on the 2001 scatter plot. The 1998 2001 line lies in the main distribution of points, but visually removed from the center of the points.
This is an important judgement as to how to normalize the post2000 ACE data, as these will be OMNI's dominant data for 2000 and beyond. It will be important to get more Wind/SWE data for periods when Wind has finished its farsideofEarth orbits to see if the SWESWEPAM comparison looks like the 422pointbased 2001 result or is more typical of what we see for 19992000. The Wind/SWE team at MIT has made "key parameter" data available to mid2003, but not yet any definitive 20023 data; only definitive SWE data are used in OMNI 2.
We shall use for OMNI 2 the 19982001based NacetoNswe regression equation above for normalizing all 19982003 ACE/SWEPAM densities to SWE densities for hours with Vace in the 350450 range. In doing so, we are introducing ~<3% discrepancies relative to what we'd get with timespanspecific normalizations for 19982000, and somewhat larger discrepancies in 2001 where the NsweNace regression is based on relatively few points. We anticipate revisiting post2000 SWEACE normalizations in the future, when large numbers of data pairs are available for WindACE impact parameter < 60 Re.
Now let's go back to the V<350 case, where we only had 1340 hours for all of 1998 2001. These hours distribute as follows for 19982001 respectively: 851, 307, 119, 63.
The Nswe/Nace ratios for these years, for Nace = 2,5,10,20 are
1998 1.004 .998 .993 .988 1999 .929 .979 1.019 1.060 2000 .943 1.065 1.167 1.279 2001 1.219 1.189 1.167 1.145 Recall we had for V<350 19982001 1.016 1.015 1.014 1.013 If we use the 19982001 equation, we'd be changing results in 1998, where most of the points are, by <2.5% and likewise by <3.5% in 1999 at N=5,10 where again most of the points are. Since the later years and, for 1999, the N=2 and 20 limits, have only few hours with ACE and Wind/SWE data, we should use the 19982001 V<350 equation and plan on revisiting a different normalization for ACE/SWEPAM when later SWE data become available. For simplicity, we change the 19982001 equation from (log N)swe = .007 + .999 * (log N)ace to (log N)swe = .006 + (log N)ace which gives a constant Nswe/Nace [or (Nnorm/Nobsvd)ace] of 1.014 In summary, we will compute normalized ACE/SWEPAM densities, time independently, using: V<350: (log N)norm = .006 + (log N)ace 350<V<450 (log N)norm = .091 + 1.036 * (log N)ace V>450 (log N)norm = .082 + (log N)aceThese say that we increase observed ACE densities by 1.014 at V<350, decrease them by 0.828 at V>450 and decrease them in the 350<V<450 range by Ndependent amounts, ranging from 0.833 at N=2 to 0.909 at N=20.
Effect of fits vs. moments determination of densities
We expect fitsbased proton densities to be, on average, somewhat less than moments based densities. This presumably contributes significantly to the extent to which momentsbased Nace values need to be reduced at V>350 to match fitsbased Nswe values. Let's try to quantify this by comparing SWE fitsbased densities with SWE momentsbased densities (both of which were provided by Justin Kasper.MIT and both of which are addressible with the ftpbrowser tools identified earlier in this note.
With many thousands of points in each run, and with uncertainties in the slopes and intercepts in the .004.007 range, we get
V<350, 19982001: (log N)fit = .080 + 1.051 * (log N)mom 350450, 19981999: (log N)fit = .062 + 1.035 * (log N)mom 350450, 20002001: (log N)fit = .063 + 1.034 * (log N)mom V>450, 19981999: (log N)fit = .051 + 1.019 * (log N)mom V>450, 20002001: (log N)fit = .046 + 1.015 * (log N)mom The two higherspeed bins were split into two 2year spans because the analysis software only handles 12,000 points per run. However, it is clear that in each V bin there is no significant time dependence. In the following, we shall work with the 1st, 2nd and 4th of these equations. They give for Nfit/Nmom ratios at Nmom = 2, 5, 10, 20: V < 350 .86 .90 .94 .97 350 < V < 450 .89 .92 .94 .96 V > 450 .90 .92 .93 .94 The corresponding Nswefit/Nace ratios were V < 350 1.014 1.014 1.014 1.014 350 < V < 450 .833 .862 .885 .909 V > 450 .828 .828 .828 .828 There is a significantly larger difference, in each speed bin, between Nswefit and Nace than between Nswefit and Nswemom. This "extra difference" may be instrumental and/or the different ways the LANL and MIT groups took moments and/or ??? Temperatures Using the same data and tools as for densities above, we find 19982001: All V: (log T)swe = 0.508 (+/.049) + 1.110 (+/.011)*(log T)ace, 10974 hrs V<350: (log T)swe = 1.680 (+/.291) + 1.348 (+/.063)*(log T)ace, 1353 hrs 350450 (log T)swe = 0.287 (+/.081) + 1.067 (+/.016)*(log T)ace, 5785 hrs V>450 (log T)swe = 0.417 (+/.104) + 1.091 (+/.020)*(log T)ace, 3932 hrs 350<V<450 1998 (log T)swe = .204 (+/.113) + 1.049 (+/.024)*(log T)ace, 2517 hrs 1999 (log T)swe = .402 (+/.133) + 1.091 (+/.029)*(log T)ace, 1914 hrs 2000 (log T)swe = .339 (+/.339) + 1.078 (+/.055)*(log T)ace, 897 hrs 2001 (log T)swe = .204 (+/.310) + 1.054 (+/.064)*(log T)ace, 422 hrs We next show the ratios Tswe/Tace evaluated at Tace = 25, 63, 160 and 400 K deg corresponding to these equations: .946 1.047 1.160 1.283 .709 .978 1.352 1.860 1.018 1.083 1.153 1.226 .962 1.047 1.139 1.238 1.026 1.074 1.125 1.176 .996 1.083 1.179 1.282 1.009 1.085 1.167 1.253 1.080 1.135 1.191 1.255 Tswe is typically greater than Tace by 525% Time dependence in the 350450 V bin is not systematic. Tswe/Tace is steady for 19982000 and then jumps into 2001, for the lower T values. On the other hand, Tswe/Tace is steady for 19992001 for the higher T values, after having been lower in 1998. Using the 19982001 350450 Vbin equation rather than the yearspecific 350450 Vbin equations introduces "errors" of less than ~6%. We shall do likewise for the V<350 and V>450 bins. Thus, we shall use the following, time independently, for normalizing ACE temperatures for OMNI 2. V<350: (log T)norm = 1.680 + 1.348 * (log T)obsvd 350<v<450 (log T)norm = 0.287 + 1.067 * (log T)obsvd V>450 (log T)norm = 0.417 + 1.091 * (log T)obsvd The changes in T values upon normalizing are similar in our two higher speed bins; both increase from near zero at T=25,000 deg to ~25% at T=400,000 deg. At low speed, however, the change upon normalizing is a much more dramatic function of T, ranging from 29% at T=25,000 deg to +86% at T=400,000 deg. Alphatoproton density ratios ( OLD) Both Wind/SWE and ACE/SWEPAM provide alphatoproton density ratios. This parameter will be added to OMNI 2; it has not been in OMNI up to now. We do not propose to crossnormalize these ratios. However, we have done crossregression runs and find, for 19982001, All V (Na/Np)swe = .003 (+/.000) + 1.080 (+/.005) * (Na/Np)ace, 4503 hrs V<350 (Na/Np)swe = .003 (+/.000) + 0.868 (+/.011) * (Na/Np)ace, 701 hrs 350<V<450 (Na/Np)swe = .003 (+/.000) + 1.029 (+/.006) * (Na/Np)ace, 2700 hrs V>450 (Na/Np)swe = .008 (+/.000) + 1.173 (+/.012) * (Na/Np)ace, 1151 hrs These four equations give (Na/Np)swe values of .057, .046, .054 and .067 for (Na/Np)ace = .050. In the corresponding scatter plots, we see significantly more scatter than for the VV or logNlogN or logTlogT scatter plots.
Appendix 2. Comparisons of IMP 8 and Wind/SWE and ACE parameters
We report herein the results of comparisons of IMP8/MIT, IMP8/LANL and Wind/SWE data for 19952001. For flow speeds, we find: V lanl = 8.40 (+/ 0.49) + 1.016 (+/ .001) * V mit, 10758 hours V swe = 4.68 (+/ 0.59) + 1.019 (+/ .001) * V mit, 9660 hours V swe = 2.80 (+/ 0.36) + 1.006 (+/ .001) * V lanl, 5376 hours For the first of these three, we required >9 fine scale points in both the LANL and MIT averages, for the second, >19 points in the IMP/MIT average and > 29 points in the SWE average and, for the third, > 14 points in the LANL average and > 29 points in the SWE average. We also required that the IMPWind impact parameter should be less than 60 Re. These equations are reasonably consistent, given that "chaining" the first two yields an inferred Wind/SWE  IMP/LANL relation of V swe = 3.74 + 1.003 * V lanl, which is very close to the "directly observed" V swe  V lanl relation. As flow speeds increase from 300 km/s to 800 km/s, Vy  Vx vary across the range: y x Vy  Vx Vy  Vx @ Vx=300 @ Vx=800 lanl mit 3.6 +4.4 swe mit +1.0 +10.5 swe lanl +4.6 +7.6 Summarizing, the IMP/MIT and IMP/LANL flow speeds are within ~ 4 km/s of each other over the speed range of most solar wind flows, while the Wind/SWE flow speed tends to be greater either IMP speed, especially at higher speeds. For completeness, we note from Appendix 1 that V swe  V ace increases from 0.9 to 5.9 km/s as V ace increases from 300 to 800 km/s. Consider next plasma densities. Here we do separate comparisons/regressions in the flow speed bins <350 km/s, 350450 km/s (abbreviated 3545 below) and >450 km/s. With column headings consistent with (log N)y = a (+/ dela) + b (+/ delb) * (log N)x, with ny and nx being the minimum numbers of fine scale points required in the density averages from sources y and x and with #hrs being the number of hourly averages folded into the regression run, we summarize the results as Y x V ny nx #hrs a b dela delb Lanl mit <350 4 5 3648 .001 .922 .008 .008 Lanl mit 3545 4 4 6926 .064 .904 .006 .006 Lanl mit >450 4 5 3745 .117 .882 .007 .009 Swe mit <350 10 5 3344 .020 .941 .009 .008 Swe mit 3545 10 5 6224 .033 .919 .006 .007 Swe mit >450 10 5 4099 .019 .907 .007 .010 Swe lanl <350 10 4 2727 .040 1.007 .008 .009 Swe lanl 3545 10 4 5186 .023 1.013 .006 .007 Swe lanl >450 10 4 3043 .093 1.022 .008 .010 These regression equations yield Ny/Nx density ratios, in the same sequence as the 9 lines above, as Nx = 2 5 10 20 .95 .88 .84 .79 1.08 .99 .93 .87 1.21 1.08 1.00 .92 1.01 .95 .91 .88 1.02 .95 .90 .85 .98 .90 .84 .79 1.10 1.11 1.11 1.12 .96 .97 .98 .99 .82 .84 .85 .86 These equations say that on IMP 8, LANL densities exceed by MIT densities by up to 20% for fast, dilute flows whereas the opposite is true for slow, dense flows. Also, SWE densities are less than IMP densities by varying extents up to 1520% except that for slow flows, SWE densities exceed IMP/LANL densities by ~10%. For completeness, recall from Appendix 1 that SWE and ACE/SWEPAM densities were comparable for slow flows, but that SWE densities were less than ACE densities by 1018 % for moderate and fast flows (i.e., V > 350 km/s). Finally, we address the temperature comparisons among IMP/MIT, IMP/LANL and Wind/SWE. Using the same format as for densities above, and recognizing that logtemperatures (not temperatures themselves) are being regressed, we find: Y x V ny nx #hrs a b dela delb lanl mit <350 4 5 3649 .594 .909 .067 .016 lanl mit 3545 4 5 6927 .542 .910 .059 .012 lanl mit <450 4 5 3751 .774 .870 .097 .019 swe mit <350 10 5 3344 .864 .839 .063 .014 swe mit 3545 10 5 6224 .491 .920 .064 .013 swe mit >450 10 5 4101 .702 .890 .090 .018 swe lanl <350 10 4 2727 .441 .895 .050 .011 swe lanl 3545 10 4 5187 .044 1.008 .041 .009 swe lanl >450 10 4 3086 .165 1.036 .074 .014 These yield Ty/Tx temperature rations, in the same sequence as above (with "K" below referring to thousand, not deg Kelvin), as Tx= 25K 63K 160K 400K 1.56 1.44 1.32 1.21 1.40 1.29 1.18 1.09 1.59 1.41 1.25 1.11 1.43 1.23 1.06 0.92 1.38 1.28 1.19 1.10 1.65 1.49 1.35 1.22 0.95 0.87 0.78 0.71 0.98 0.99 0.99 1.00 0.98 1.02 1.05 1.09 These say that IMP/MIT temperatures are systematically less than IMP/LANL and SWE temperatures, while the latter two sources have temperature values quite close to each other at V > 350 km/s.
Appendix 3. Comparisons of IMP8/MIT and IMP8/LANL parameters
We have done separate regression runs for the two IMP 8 plasma data sets for the time intervals 19731979, 19801981, 19821994, 19952001. With the requirement that each MIT average have at least 30 fine scale points, and each LANL average have at least 18 points (except relaxing those numbers to 10 and 10 for 19801981, we find for flow speeds: 19731979 V lanl = 5.69 (+/ 0.98) + 1.010 (+/ .002) * V mit, 2549 hours 19801981 V lanl = 6.82 (+/ 1.17) + 1.012 (+/ .003) * V mit, 1585 hours 19821994 V lanl = 5.61 (+/ 0.51) + 1.009 (+/ .001) * V mit, 7917 hours 19952001 V lanl = 7.31 (+/ 0.97) + 1.013 (+/ .002) * V mit, 2728 hours These equations give V lanl  V mit differences between ~  3 km/s at 300 km/s and 23 km/s at 800 km/s. Next, consider densities in each of the four time spans and in each of the three flow speed bins <350 km/s, 350450 km/s and >450 km/s. All the runs reported require at least four LANL and five MIT fine scale points in the relevant hours. In all cases we report the results for (log N) lanl = a (+/ dela) + b (+/ delb) * (log N) mit: Time span speed a b dela delb #hrs 19731979 <350 .096 .998 .010 .009 2872 3545 .033 .966 .005 .005 7154 >450 .031 .947 .005 .007 6833 19801981 <350 .117 .919 .016 .017 818 3545 .058 .910 .016 .018 1066 >450 .045 .865 .015 .018 651 19821994 <350 .046 .975 .008 .007 4124 3545 .053 .919 .006 .006 7460 >450 .103 .902 .005 .006 8350 19952001 <350 .001 .922 .008 .008 3648 3545 .064 .904 .006 .006 6926 >450 .117 .822 .007 .009 3745 These regression equations yield (N lanl)/(N mit) ratios as follows (in the same sequence as the above 12 rows): N mit = 2 5 10 20 .80 .80 .80 .80 .91 .88 .86 .84 1.04 .99 .95 .92 .72 .67 .63 .60 .82 .76 .71 .67 1.01 .89 .81 .74 .88 .86 .85 .84 1.07 .99 .94 .88 1.18 1.08 1.01 .95 .95 .88 .84 .79 1.08 .99 .93 .87 1.21 1.08 1.00 .92 These ratios show the anomalously low (N lanl)/(N mit) ratio in the 19801981 interval. Neither the MIT nor LANL team had a good explanation for this when queried in late 2002. The ratios show largely similar values for the 19821994 and 19952001 intervals, with these ratios being a few percent higher than the corresponding rations of the 1973 1979 interval. The results also show that the ratio increases with speed (consistent with earlier findings of Lazarus and Paularena, 1998) at given density, and decreases with density at given speed. Next, we repeat the above for temperatures. Again, we require 4 LANL and 5 MIT fine scale points in each hourly average included. Again, we show parameters in the equation (log T) lanl = a (+/ dela) + b (+/ delb) * (log T) mit time span speed a b dela delb #hrs 19731979 <350 .276 .954 .067 .015 2709 3545 .210 .960 .046 .010 7249 >450 .460 .915 .059 .012 6904 19801981 <350 .029 1.034 .191 .043 805 3545 .892 .843 .136 .029 1084 >450 .518 .923 .190 .038 658 19821994 <350 .410 .938 .066 .016 4214 3545 .448 .923 .059 .012 7460 >450 .582 .906 .070 .013 7133 19952001 <350 .594 .909 .067 .016 3649 3545 .542 .910 .059 .012 6927 >450 .774 .870 .097 .019 3751 The corresponding (T lanl)/(T mit) ratios are T mit = 25K 63K 160K 400K 1.19 1.14 1.09 1.04 1.08 1.04 1.00 0.97 1.22 1.13 1.04 0.96 1.51 1.56 1.61 1.66 1.59 1.38 1.19 1.03 1.51 1.41 1.31 1.22 1.37 1.30 1.22 1.16 1.29 1.20 1.12 1.04 1.47 1.35 1.24 1.14 1.56 1.44 1.32 1.21 1.40 1.29 1.18 1.09 1.59 1.41 1.25 1.11 These ratios show that the momentsdetermined LANL temperatures are almost always greater that the fitsbased MIT temperatures. The ratios are smallest in the 1970's. The ratios typically decrease with temperature, except for the slow flow case (V <350 km/s) in the densityanomalous years of 19801981. The ratios of the 19821994 interval are intermediate between the 19801981 period and the 19952001 period.
Appendix 4. Comparisons of ISEE 3 and IMP 8 parameters
All runs below required that the IMP 8  ISEE 3 impact parameter be 60 Re or less. "lanl" and "mit" below refer to IMP 8 data. Consider first the 1978/08/16  1980/02/16 interval of availability of ionbased plasma parameters from ISEE 3. For these runs, we required at least 5 fine scale points in each IMP average and 10 fine scale points in each ISEE average. For flow speeds we find V isee = 6.468 (+/1.001) + 1.013 (+/ .003) * V lanl, 946 hours V isee = 10.667 (+/.988) + 1.016 )+/ .003) * V mit, 1733 hours These equations yield for V isee  V imp: V imp = 300 800 Imp=lanl 2.6 3.9 Imp=mit 5.9 2.1 For ionbased densities and for IMP/MIT only, we have V<350 (log N) isee = .046 (+/ .025) + 1.100 (+/ .023) * (log N) mit, 352 hrs 353450 (log N) isee = .070 (+/ .017) + 1.009 (+/ .019) * (log N) mit, 842 hrs V> 450 (log N) isee = .059 (+/ .016) + 0.933 (+/ .020) * (log N) mit, 551 hrs These yield (N isee)/(N imp) ratios of N imp = 2 5 10 20 V<350 .96 1.06 1.13 1.21 350450 .86 .86 .87 .87 V>450 .83 .78 .75 .71 For all but slow flows (V < 350 km/s), ISEE densities have to be increased significantly to match IMP/MIT densities. For ionbased temperatures, we find V < 350 (log T) isee = .557 (+/ .173) + .906 (+/ .038) * (log T) mit, 352 hrs 350451 (log T) isee = .484 (+/ .155) + .920 (+/ .033) * (log T) mit, 843 hrs >450 (log T) isee = .743 (+/ .194) + .876 (+/ .039) * (log T) mit, 557 hrs These yield (T isee)/(T imp) ratios of T imp = 25K 63K 160K 400K V<350 1.39 1.28 1.17 1.07 350450 1.35 1.26 1.17 1.09 >450 1.58 1.41 1.25 1.12 Independent of flow speed, T isee exceeds T mit by several tens of percent at low temperature, and by about 10 % at high temperature. Now let's compare the electronbased ISEE densities and flow speeds with IMP ionbased values. For these runs, we require at least 5 IMP points in each average, but it is meaningless to specify an equivalent number the electronbased parameters since they came to NSSDC originally as hourly averages with no indication of numbers of contributing fine scale values. For flow speed, we find V isee = 1.042 (+/ 1.534) + 0.980 (+/ .003) * V lanl, 1293 hours V isee = 5.786 (+/ 1.409) + .995 (+/ .003) * V mit, 2175 hours These equations yield for V isee  V imp: V imp = 300 800 Imp=lanl 5.0 15.0 Imp=mit 7.3 9.8 These deltaV's are large enough to warrant cross normalization, but this has not been done in OMNI 2. Perhaps a next version of OMNI will do so. For densities, we have V<350 (log N) isee = .001 (+/ .046) + 1.060 (+/ .043) * (log N) mit, 607 hrs 350451 (log N) isee = .022 (+/ .030) + 1.091 (+/ .032) * (log N) mit, 938 hrs >450 (log N) isee = .018 (+/ .025) + 1.056 (+/ .032) * (log N) mit, 649 hrs and these equations yield the (N isee)/(N imp) ratios N imp = 2 5 10 20 V<350 1.04 1.10 1.15 1.20 350450 1.07 1.16 1.23 1.32 >450 1.08 1.14 1.19 1.23 ISEE 3 electron based densities are greated than IMP/MIT ionbased densities by a few percent at low density and by ~25% at high densities, largely independent of flow speed.
Appendix 5: Comparisons of LANL data from IMPs 6, 7 and 8
These runs require at least 5 fine scale points in each average, and they are all run over the common IMP 678 interval of October, 1973, to the end of 1974. We use subscripts i6, i7, i8 to designate IMP 6, IMP 7 and IMP 8. For flow speeds we find V i8 = 27.776 (+/ 1.359) + 1.050 (+/ .003) * V i6 V i8 = 2.183 (+/ 1.263) + 0.997 (+/ .002) * V i7 These yield V i8  V ix differences of V ix = 300 800 V i6 12.8 12.3 V i7 1.3 0.2 The IMP 7 and 8 flow speeds agree very well, the IMP 6 flow speed disagrees by ~12 km/s at both ends of the 300800 km/s range, although it matches the IMP 8 speed at V ~ 550 km/s. For densities we find V<350 (log N) i8 = .118 (+/ .166) + 1.140 (+/ .173) * (log N) i6, 80 hrs 350450 (log N) i8 = .028 (+/ .031) + 1.059 (+/ .036) * (log N) i6, 450 hrs > 450 (log N) i8 = .030 (+/ .015) + 0.994 (+/ .022) * (log N) i6, 867 hrs V<350 (log N) i8 = .171 (+/ .103) + 1.042 (+/ .087) * (log N) i7, 126 hrs 350450 (log N) i8 = .123 (+/ .035) + 1.017 (+/ .035) * (log N) i7, 400 hrs >450 (log N) i8 = .092 (+/ .021) + 1.010 (+/ .026) * (log N) i7, 816 hrs These correspond to (N i8)/(N ix) ratios of N ix = 2 5 10 20 ix = i6 .84 .95 1.05 1.16 .98 1.03 1.07 1.12 1.07 1.06 1.06 1.05 ix = i7 .83 .87 .89 .92 .76 .77 .78 .79 .81 .82 .83 .83 These equations and ratios say that the IMP 6 and IMP 8 densities typically agree to within 10%, while the IMP 7 density is on the order of 20% greater than IMP 8 densities. In fact, it was this IMP 7 density "overage" that led the LANL plasma team to normalize its IMP 6 and IMP 8 densities upward in its creation of the merged hourly IMP 678 data set provided to NSSDC years ago and used by NSSDC in the 1970's as the fiducial data set for OMNI to which to normalize other densities. Finally, let us consider temperatures: V<350 (log T) i8 = .127 (+/ .363) + 1.017 (+/ .077) * (log T) i6, 80 hrs 350450 (log T) i8 = .019 (+/ .178) + .992 (+/ .035) * (log T) i6, 449 hrs >450 (log T) i8 = .298 (+/ .113) + .939 (+/ .021) * (log T) i6, 871 hrs V<350 (log T) i8 = .023 (+/ .232) + 0.992 (+/ .051) * (log T) i7, 126 hrs 350450 (log T) i8 = .072 (+/ .123) + 0.983 (+/ .025) * (log T) i7, 400 hrs >450 (log T) i8 = .053 (+/ .138) + .988 (+/ .026) * (log T) i7, 816 hrs These yield (T i8)/(T ix) ratios of T ix = 25K 63K 160K 400K ix = i6 .89 .90 .92 .93 .96 .96 .95 .94 1.07 1.01 .96 .90 ix = i7 .97 .97 .96 .95 .99 .98 .96 .95 1.00 .99 .98 .97 These say that IMP 7 and IMP 8 temperatures agree extremely well, and that IMP 6 and IMP 8 temperatures agree to within 10% of each other.

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