==============================================================================
SWICS 1.1 Level 2 Version 4.09 Data Release Notes

Jason A. Gilbert (to whom questions should be asked), George Gloeckler (SWICS
lead Co-I), Sue T. Lepri, Jim M. Raines, Paul Shearer, Patrick Tracy, Micah J.
Weberg, and Thomas H. Zurbuchen (Co-I)
==============================================================================

"SWICS 1.1 Version 4.09 contains a few incremental improvements that optimize
the ion identification model and reveal more counts of low-density species.
The process for removing accidental coincidence events has been improved,
leading to a small increase in identification of rare ions such as Fe6+, Fe7+,
and C4+. In addition, the forward model for ion identification now uses all
charge states of all 11 elements, making the ion count assignment as accurate
as possible. Statistical improvements were also made in the model so that ion
thermal speeds are calculated with a higher degree of accuracy, particularly
in the slow wind."


Overview
--------

The SWICS 1.1 dataset consists of a re-release of significantly improved time
series measurements by ACE/SWICS of the elemental abundance, charge state
composition, and kinetic properties of heavy ions in the solar wind. It is a
major new version produced with completely redesigned analysis methods to
account more rigorously for instrumental and statistical effects (Shearer et
al. 2014). Rare elements are now identified more reliably and estimates of
statistical error are provided. These release notes describe the data, the
methods used to determine them, and issues of data quality and measurement
uncertainty.

Table of Contents
-----------------

* Method
* Ions Analyzed
* Data Products
* Missing Data
* Systematic Error
* Cross-calibration and Systematic Error in Quantities Involving He
* Quality Flags
* Data Quality and Usage Caveats
* Computational Details
* References


Method
------

SWICS measures solar wind ions with a triple coincidence technique (Gloeckler
et al. 1998). An accumulation of ions can be represented as a three
dimensional histogram: ion counts versus energy per charge (E/q), time of
flight (TOF), and measured energy (E). At a fixed E/q value, each ion species
(He2+, O6+, etc.) appears as a distinct peak in the (TOF--E)
histogram. By assigning counts to the appropriate species peak, each species'
count rate can be estimated and its (reduced) phase space distribution
function inferred.

For the more abundant heavy ions, count assignment is relatively unambiguous
and was performed reliably in previous versions using the methods of von
Steiger et al. (2000). However, these methods can encounter robustness issues
with rare species, especially those overlapped with significantly more
abundant species (i.e., it is very hard to fit, at the same time, a molehill
and a mountain...). SWICS 1.1 addresses these issues with a new
forward model of the peak position and width, and a rigorous maximum
likelihood count assignment algorithm (Shearer et al. 2014). These advances
have increased the reliability of measured abundances for rare elements, such
as N, Ne, and S.

Almost all data products are calculated without cross-calibration to other
instruments, using only SWICS in-flight measurements and pre-flight
calibration data. The exceptions are products involving He and H, specifically
the He2+ density and the He/O ratio, as well as kinetic properties of H.

Ions Analyzed
-------------

SWICS detects heavy ions over a very wide range of mass, charge, and
abundance. However, data from some ion species are subject to issues requiring
expert instrumental knowledge and cannot be provided as ACE Level 2 data. This
dataset includes to the following ion species:

He  2+
C   4-6+
N   5-6+
O   5-8+
Ne  8-9+
Mg  6-12+
Si  6-12+
S   6-14+
Fe  6-20+

Under normal solar wind conditions, the vast majority of ions are found in
these charge states. However, charge states outside these bounds may occur
during extremely cold or hot ICMEs (Lepri and Zurbuchen, 2010).

Data Products
-------------

Time resolutions: 1 hour, 2 hour, 1 day. (Not all products are delivered at all time-resolutions.) 

* Density [particles per cm^3]: He2+
* Bulk velocity [km/s]: He2+, C5+, O6+, Fe10+
* Thermal velocity [km/s]: He2+, C5+, O6+, Fe10+
* Charge state ratios: C6+/C4+, C6+/C5+, O7+/O6+
* Elemental abundances: He/O, C/O, N/O, Ne/O, Mg/O, Si/O, S/O, Fe/O
* Average charge state: C, O, Mg, Si, Fe
* Charge state distributions: C, O, Ne, Mg, Si, Fe 

For all quantities except bulk and thermal velocity, we also provide estimates
of statistical uncertainty due to limited count statistics. Note that these
estimates do NOT include systematic error, which is discussed separately
below. Let us already state here, though, that systematic errors remain
important limitations of SWICS data, and similarly of all measurements of
heavy ions in space. These errors are most often caused by the fact that
- for both experimental limitations and time-reasons -
we are unable to do calibrations of all ion species delivered.

Each compositional value is provided with a statistical error mostly related
to the count-rate that is available. The data user will recognize that
- especially for highly resolved data - statistical
errors are big, often at 20% or higher. This also includes noise introduced by
the instrument electronics as well as the data-analysis and telemetry system,
since we do not have enough telemetry to send all data to Earth. This will be
discussed later as well.

Elemental charge states which are not in the "Ions Analyzed" list above are
not accounted for in any fashion. Thus the C/O ratio is in fact C[4-6+] /
O[5-8+], the average charge state of Mg is an average only over the charge
states 6-12+, etc. Thus, should there be a huge contribution of C2+ (which may
occur in prominence-associated plasmas), the C/O average will be flawed.


Missing Data
------------

Missing data is marked with a value of -1. Data is generally removed when

- a numerical error is encountered (e.g. a zero denominator in a density ratio)
- statistics are too low to reliably report a quantity
- anomalous instrument operation was identified by the operations or analysis team


Systematic Error
----------------

Systematic error of SWICS measurements is discussed extensively in the
Appendix of von Steiger and Zurbuchen (2011). It is primarily due to error in
count assignment and/or calibration factors measured pre-flight both related
to unavailable and insufficient calibration data for high charge-state ions.
Systematic errors can also occur due to changes of the instruments over time,
but that effect has been determined to be small by von Steiger and Zurbuchen
(2011).

Count assignment error occurs when one species is mistaken for another. This
error is greatest for rare species which overlap with more abundant species in
the (TOF--E) histogram and highly sensitive on the exact location
and distribution of measurements of a given ion in TOF-E space. For elemental
abundance ratios, assignment error due to overlap is generally between 5-10%.
However, in cases where rare elements overlap closely with abundant elements,
the error may reach as high as ~20%. In particular this higher estimate should
be applied to the N/O ratio, due to overlap of N5-6+ with highly abundant
O6-7+. In ACE/SWICS data, count assignment error tends to bias the abundance
of rare elements upwards.

The calibration factors that contribute most uncertainty are

- geometric factor: SWICS effective collection area, mostly effecting the
absolute density, expected to be approximately constant in time

- duty cycle: fraction of time the solar wind can enter the SWICS aperture,
affecting density and statistical accuracy as a function of time

- triple-coincidence efficiency: the probability of detecting a triple
coincidence given that a particle has passed the SWICS collimator. These tripe
coincidences are mostly model-based as detailed measurements cannot be made
pre-flight.

The geometric factor and duty cycle are well-approximated as
species-independent, so they do not generally affect quantities computed for
composition ratios. Every element in the current dataset is computed with
count ratios except for the H+ and He2+ densities, so they are not subject to
these errors. The He2+ densities and He/O ratio are subject to special
uncertainties described in the next section. H+ data will be discussed in a
later section.

The triple coincidence efficiency is different for each ion species and thus
does not completely cancel in count ratios. The absolute detector efficiencies
are thought to be accurate to within 10% for He and 10-15% for other elements,
while the effect of the detector efficiencies is an error of 10% for elemental
ratios and 5% for charge state ratios (von Steiger and Zurbuchen, 2011). The
effect for charge state ratios is smaller due to a partial cancellation of
correlated errors. For example, the error for C6+ and C5+ is nearly identical,
so it largely cancels out in the C6+/C5+ ratio.

Cross-calibration and Systematic Error in Quantities Involving He
-----------------------------------------------------------------

The He2+ density and He/O density ratio products have been subjected to
adjustments based on a cross-calibration of the He2+ density with WIND/SWE.
These adjustments change the long-term average values of these two data
products in a wind-speed dependent fashion. Due to the increased systematic
uncertainties associated with cross-calibration, the instrument team should be
consulted before drawing any scientific conclusions should from absolute
values of the He2+ density and He/O ratio, or their dependence on speed.

Quality Flags
----------------------------------------------------------------

A set of quality flags are provided for each data product based on analysis of
the velocity distribution functions (VDFs) for all ion species of the element
considered. The flags are encoded in a bit-field which may then have the
following base values:

Value	Description
----------------------------------------------------------------
0	Good quality data
1	All charge states have empty VDFs (data set to -1)
2	All charge states have single point or empty VDFs (data set to -1)
4	All charge states have two point, single point, or empty VDFs (data flagged
	but values are left as-is to allow scientists to get a feeling for marginally
	valid data. Should be used only with caution and in most cases is not
	publication grade).
8	O6+ has a quality flag of 1, 2, or 4 (data set to -1)
16	Abnormal VDF determined as part of our quality assurance process, which
	involves inspection by a data specialist (data set to -1)

Most quality flags are mutually exclusive. However, in the case of the
elemental X/O ratios the quality flag will have a value corresponding to the
union of the flags for oxygen and element considered. For example, if O has a
flag of 4 and Fe has a flag of 2, the Fe/O ratio would have a quality flag of
6 and thereby would be set to a value of -1. Likewise, if both elements have a
flag of 4, the corresponding ratio would then also be 4.


Data Quality and Usage Caveats
------------------------------

Instrument anomalies have been removed by extensive automatic screening and
our quality assurance process that involves inspection by a data specialist,
but it is likely that a few questionable data points remain. Single-point
outliers are especially suspect and may require consultation from the SWICS
instrument team to interpret.

Measurement uncertainty is sometimes quite high when measured counts are low.
It is not uncommon to observe 5 counts or fewer of a rare ion species, such as
Fe6+ or O8+, during a 2 hour period. When unusual volatility or outliers are
observed in the data, issues related to limited count statistics can be
diagnosed with the provided statistical error estimates.

Bulk and thermal velocity are calculated from the first and second moments of
the distribution (see Computational Details below) and this simple
moment-based method may be biased when the distribution is significantly
non-Gaussian. This tends to occur when the bulk speed is highly variable on a
timescale below the data time resolution, or when the distribution itself is
highly non-Maxwellian, e.g., it exhibits beaming or suprathermal tails.

Computational Details and Statistical Errors
---------------------

Count assignment is based on the principle of maximum likelihood for count
histograms, which are governed by Poisson statistics. Convex optimization or
Expectation-Maximization can be used to reliably determine the maximum
likelihood count assignment (Shearer et al. 2014).

The data products' statistical uncertainties are obtained by Gaussian error
propagation from the sqrt(N) errors associated with count statistics, where N
is the number of counts observed for a given interval. If zero counts are
observed we (conservatively) assume a statistical uncertainty of 1 count.
Largely speaking, our data are count-limited, especially for rate ions and/or
for high resolution.

The thermal velocity reported in this dataset only accounts for the solar
wind's thermal distribution along the direction of flow, as this is the only
direction where speed distribution is observable by SWICS. Formally, the
thermal velocity is computed from the 1-D reduced velocity distribution
function along the SWICS look direction, which is formed (to a good
approximation) by marginalizing out the two perpendicular dimensions from the
full 3-D velocity distribution. The thermal velocity we report is the
root-mean-square deviation of the 1-D reduced distribution from the bulk
speed.

References
----------

Gloeckler et al. (1998). Investigation of the composition of solar and
interstellar matter using solar wind and pickup ion measurements with SWICS
and SWIMS on the ACE spacecraft. Space Science Reviews, Volume 86, pp.
497-539.

Lepri, S. T., and T. H. Zurbuchen (2010). Direct observational evidence of
filament material within interplanetary coronal mass ejections. ApJL, 723.1,
L22.

Shearer et al. (2014). The solar wind neon abundance observed with ACE/SWICS
and Ulysses/SWICS. ApJ, Volume 789, Issue 1, p. 60.

von Steiger and Zurbuchen (2011). Polar coronal holes during the past solar
cycle: Ulysses observations. JGR Space Physics, Volume 116, Number A1.