Near‐Earth heliospheric magnetic field intensity since 1750: 2. Cosmogenic radionuclide reconstructions

This is Part 2 of a study of the near‐Earth heliospheric magnetic field strength, B, since 1750. Part 1 produced composite estimates of B from geomagnetic and sunspot data over the period 1750–2013. Sunspot‐based reconstructions can be extended back to 1610, but the paleocosmic ray (PCR) record is the only data set capable of providing a record of solar activity on millennial timescales. The process for converting 10Be concentrations measured in ice cores to B is more complex than with geomagnetic and sunspot data, and the uncertainties in B derived from cosmogenic nuclides (~20% for any individual year) are much larger. Within this level of uncertainty, we find reasonable overall agreement between PCR‐based B and the geomagnetic‐ and sunspot number‐based series. This agreement was enhanced by excising low values in PCR‐based B attributed to high‐energy solar proton events. Other discordant intervals, with as yet unspecified causes remain included in our analysis. Comparison of 3 year averages centered on sunspot minimum yields reasonable agreement between the three estimates, providing a means to investigate the long‐term changes in the heliospheric magnetic field into the past even without a means to remove solar proton events from the records.


Introduction
Long-term solar variability has potentially important implications for studies of both space [Barnard et al., 2011;Usoskin, 2013] and terrestrial regional and global climate [Gray et al., 2010;Lockwood, 2012;Ineson et al., 2015].As such, the heliospheric magnetic field, HMF [e.g., Owens and Forsyth, 2013], is of direct interest to space weather science and forecasting and is a useful proxy for solar magnetism in general.This work provides new understanding and justification of the use of the cosmogenic radionuclide record in paleoclimate studies [Lockwood, 2006;Beer et al., 2012].Part 1 of this series of two papers [Owens et al., 2016] showed that B, the intensity of the HMF in near-Earth space, can be reconstructed back more than two centuries using both geomagnetic and sunspot number (SSN) observations with remarkably low uncertainties and considerable agreement between these two independent sources of information of past HMF variations.The composite geomagnetic reconstructions, B[GEO], provide the most accurate method for inferring B, both in terms of agreement with direct spacecraft observations and within the two methods adopted.B[GEO] can be reliably extended back to 1845.Sunspot number reconstructions, B [SSN], are obviously highly dependent on the underlying sunspot record, but also, to a lesser extent, on the methodology used to convert SSN to B. Nevertheless, a composite of records and techniques produces an extremely good match with B[GEO].This technique has been extended back to 1750.In principle, this method could be extended back to 1610, though that requires further assessment of the sunspot records prior to 1750, as discussed in Part 1.

PUBLICATIONS
as well as the inherent noise in the measurement itself.That indirectness arises from the fact that there are other influences on the cosmic ray flux at Earth and because the near-Earth HMF is a local measure of the heliosphere whereas the external source of cosmic rays means they are influenced by the heliosphere as a whole.Nevertheless, cosmogenic radionuclide reconstructions are invaluable for two reasons.First, they provide the potential to give independent verification of B reconstructions over the last 200-400 years.Second, and most importantly, they are the only tool available to study the evolution of the solar wind prior to 1610 and over the previous millennia.
Figure 1, adapted from Svalgaard and Cliver [2010] and Steinhilber et al. [2010], shows a comparison of the long-term variations in B from a geomagnetic reconstruction (solid red line) and cosmogenic radionuclide reconstructions, shown as various colored lines described in the figure caption.At the time of these studies, there was reasonable agreement between the different B reconstructions, though the geomagnetic reconstructions of B tended to show less long-term variation than the cosmogenic radionuclide reconstructions.
Similarly, estimates of B over the last century based on the sunspot number [Solanki et al., 2000;Wang et al., 2005;Vieira and Solanki, 2010] have tended to be lower before ~1950 than recent reconstructions based on geomagnetic data and cosmogenic radionuclide data [see, e.g., McCracken, 2007, Figure 5;Jiang et al., 2011, Figure 9].More recent sunspot-based estimates of B, however, show closer agreement with geomagnetic and cosmogenic estimates [Owens and Lockwood, 2012;Lockwood and Owens, 2014;Owens et al., 2016], primarily as a result of changes to the underlying SSN records.
It is important to emphasize that we should not expect the agreement between the various B reconstructions to be perfect.As mentioned above, the spacecraft and geomagnetic data are essentially point measurements of the solar wind in close proximity to Earth, whereas cosmic ray modulation involves the entire heliosphere [e.g., Jokipii and Wibberenz, 1998;Potgieter, 2013].Similarly, the sunspot number is a more global measure than just the HMF sampled near Earth.Nevertheless, the anticorrelation between annual averages of solar wind B near Earth and cosmic ray intensity [Cliver et al., 2013, Figure 10] is strong enough that we can reasonably expect to use radionuclides as a tool to infer B, at least to within some level of uncertainty.Extracting HMF data from these radionuclides, however, is a much more indirect process than inferring them from sunspot or geomagnetic data.The extracted cosmic ray record may be affected by such factors as terrestrial climate effects on the deposition into the reservoirs in which they are measured, geomagnetic field variability, variations in the local interstellar spectrum of cosmic rays, and high-energy solar energetic particle events (sometimes referred to as "solar cosmic rays") [Usoskin et al., 2006;Webber et al., 2007;Miyake et al., 2012;Bazilevskaya et al., 2014;Güttler et al., 2015;McCracken and Beer, 2015].
In order to understand the accuracy and limitations of obtaining solar wind parameters from cosmogenic radionuclides as we go back in time, it is necessary to make a detailed comparison between the B series derived from sunspots, geomagnetic data, and the cosmogenic radionuclide data for their period of overlap.measurements of the >4 GeV/nucleon particles began in 1935 [e.g., Hess and Graziadei, 1936;Forbush, 1937], later superseded by the worldwide neutron monitor (NM) network (sensitive to >1 GeV/nucleon particles) that commenced in 1951 [e.g., Simpson, 2000].These data, and more recent spacecraft data [e.g., Heber et al., 2009], have shown that the cosmic ray intensity at Earth varies strongly throughout the Schwabe cycle.

Comparison
Theoretical studies [Parker, 1958;Potgieter, 2013] have shown that these variations are a direct and quantifiable consequence of the varying structure and intensity of the heliospheric magnetic field.
Upon entering the atmosphere, galactic cosmic rays (GCR) initiate nuclear reactions that lead to production of the cosmogenic radionuclides, such as 10 Be and 14 C (half-lives 1.39 × 10 6 and 5730 years, respectively) [Beer et al., 2012].These radionuclides are sequestered in ice cores ( 10 Be) and tree rings ( 14 C).Continuous records of the concentrations of these radioisotopes in their host reservoirs exist for >10,000 years in the past.
Following the pioneering work of Masarik and Beer [1999], numerical models of cosmogenic radionuclide production, storage, and release have been developed [e.g., Webber and Higbie, 2003;Webber et al., 2007;Kovaltsov and Usoskin, 2010;Kovaltsov et al., 2012] making it possible to relate these observations quantitatively to the cosmic radiation intensity at Earth.Thus, the cosmogenic data, frequently referred to as paleocosmic ray (PCR) data, yield indirect measurements of the cosmic ray intensity, providing a detailed record of the variability of the GCR intensity at Earth, and solar activity, for >10,000 years into the past [McCracken et al., 2013].Due to storage and exchange in the oceans and biomass, 14 C production changes are strongly damped, which makes resolving individual solar cycles difficult.At the present time there are only two PCR records providing annual measurements covering the interval 1389-1994; 10 Be concentrations measured in ice cores from Dye 3 (Greenland) and the North Greenland Ice Core Project (North GRIP).Both are used in this study.It is important to note that the statistical variations in the annual cosmogenic data are much greater than those in the NM data in that the standard deviation of an individual yearly average of the 10 Be concentration is ~20%, compared to <0.1% for NM data [McCracken et al., 2004].While there is limited likelihood of the discovery/utilization of substantial amounts of additional geomagnetic and sunspot data as used in Part 1, there are several new prospective ice cores that will significantly improve the statistical accuracy of the PCR record and the estimates of the HMF derived therefrom.Accordingly, it can be expected that the quality of the PCR record and subsequent reconstructions of B will improve considerably over the next decade (see section 2.5).
Unfortunately, use of 14 C is made more difficult for the modern period (from about 1850) due to the increasing influence of burning of fossil fuel releasing of "old" carbon into the atmosphere (the so-called First Suess Effect) [Tans et al., 1979] and the atmospheric nuclear bomb tests commencing in the early 1950s.Both significantly alter the atmospheric 14 C/ 12 C and 13 C/ 12 C ratios, making it difficult to extract the cosmogenic production signal, though recent reconstructions are in broad agreement over the last 400 years [e.g., Roth and Joos, 2013;Muscheler et al., The revised sunspot number series in comparison to cosmogenic radionuclide based solar activity reconstructions, submitted to Solar Physics, 2016].

Estimation of the HMF From the Paleocosmic Ray Data
As outlined above, the PCR and instrumental records provide the output of a "heliospheric magnetometer."Using a three-dimensional model of the heliosphere that included a solar cycle-dependent current sheet, latitude dependent solar wind velocities, and the Hale cycle of solar magnetic fields, Caballero-Lopez et al. [2004] developed the means to use the annual PCR record to estimate the HMF intensity near Earth.Henceforth, we refer to this as B [GCR].McCracken [2007] modified this method, using the annual NM and Dye 3 records to calibrate the inversion process to the satellite observations of HMF intensity since 1965.Subsequently, a second annual record 1389-1994 (North GRIP) was published [Berggren et al., 2009] and detailed analysis made of the long-term changes in the cosmogenic data using all the available 10 Be and 14 C data [Steinhilber et al., 2012] Be events in the interval 1800-1942 followed major geomagnetic storms.As major storms are often also accompanied by solar energetic particle (SEP) events (e.g., 16 of the 18 Dst <À200 nT storms identified by Zhang et al. [2007] are associated with SEPs identified by Cane et al. [2010]), McCracken and Beer [2015] postulated that these seven impulsive enhancements in10 Be were also of solar origin.On that basis, they excised all >3 standard deviation impulsive enhancements from the 10 Be record and used the revised record to estimate the HMF intensity near Earth.We refer to these two different estimates B[GCR-MB1] and B[GCR-MB2]: the former including the solar energetic particle events and the latter after their removal.In the next section, we compare both estimates with B[GEO] and B[SSN] for the interval 1800-1983.
McCracken and Beer [2015] discussed the possibility that the 10 Be record may also contain impulsive anomalies associated with volcanic eruptions.Volcanic ejecta do not contain 10 Be; however, SO 2 and aerosols therein may influence the atmospheric transportation process.This may accelerate or decelerate the sequestration process, however, no Earth system process, other than GCR spallation of the atmosphere, is known to lead to a net increase in 10 Be at this time.Sigl et al. [2014] have provided detailed records of the sulfate production by volcanoes during the period of study herein, and further consideration of this putative enhancement processes may be warranted as our physical understanding develops.
An independent estimate of annual B[GCR] is produced using the same data set of 10 Be series from the Dye 3 and North GRIP Greenland ice cores but using a slightly different method for converting to B. It follows the lower time resolution open solar flux estimate used in Usoskin et al. [2015].First, using the 10 Be production model [Kovaltsov and Usoskin, 2010] and the most up-to-date geomagnetic field model (the International Geomagnetic Reference Field, IGRF, since 1900 and [Licht et al., 2013] before 1900), the modulation potential, ϕ, was evaluated for each 10 Be record.ϕ, effectively a measure of the average rigidity loss of GCR particles in the heliosphere [e.g., Usoskin, 2013], was then converted into the open solar magnetic flux [Usoskin et al., 2002;Alanko-Huotari et al., 2006]

Note the very low values of both B[GCR-U] and B[GCR-MB1
] centered on ~1860-1863, ~1884, ~1893, and ~1948, all four being close to sunspot maximum.These low values are not readily explicable in terms of our modern knowledge of the variation of the HMF during the Schwabe cycle but are consistent with the hypothesis that they are the consequence of GLEs produced by the active Sun; GLEs elevate 10 Be production, which is misinterpreted as enhanced GCR intensity resulting from a weaker HMF.Figures 3c and 4c of 1942, 1949, and 1956, and the nuclear weapon event of 1962.However there are no significant changes in the discrepancies with B[GEO] and B[SSN] for ~1860-1863 and the first two years of the 1948-1952discrepancy. McCracken and Beer [2015] examined the B[GCR-MB2] underestimate at 1948 and concluded that it was due to the production of 10 Be by the large, long duration GLE that was observed with ionization chambers on 25 July 1946 [Forbush, 1946].Their analysis shows that the resulting impulsive 10 Be enhancement merged with the rapidly falling onset of the 11 year modulation of the galactic cosmic radiation (1946)(1947)(1948), with the result that the impulsive enhancement was not statistically significant in the annual PCR record (Figure 2b).They concluded that large GLEs occurring during the steep rise in B during the ascending phase (typically the second or third year) of a Schwabe cycle may escape detection in this manner in the present-day PCR record.For example, the large discrepancy in the interval 1860-1865 in Figures 3c and 4c may be the consequence of solar energetic particle production during Schwabe cycle 10, 1857-1867 (the Carrington white light flare occurred in 1859).Thus, while Usoskin and Kovaltsov [2012] noted the absence of an impulsive 10 Be enhancement corresponding to the Carrington flare, McCracken and Beer [2015] speculate that this could be a consequence of the superposition effect described above.McCracken and Beer [2015] estimate that ~15% of all large GLEs may escape detection in the presentday PCR record in this manner.The inability to detect some large solar energetic particle enhancements appears to be a fundamental but temporary limitation to the estimation of B[GCR] (see section 2.5).
GLEs are rarely observed to occur within ±1 year of sunspot minimum.Figure 5 and Table 2 compare    which is reflected in the high correlation (r = 0.70) but high MSE (1.0 nT 2 ).We also note the qualitative agreement between an Antarctic ice core 10 Be reconstruction of solar modulation potential [Muscheler et al., 2016] and B[GCR-MB1] for this period.

Assessment of the Use of B[GCR]
Prior to 1750 While we have excellent measurements and theoretical knowledge of the active sun since 1850 primarily through geomagnetic observations, we have few direct observations of the solar "Grand Minima" (e.g., Spoerer 1420-1540; Maunder, 1645-1715; and Dalton 1790-1830).See, e.g., Riley et al. [2015] and Usoskin et al. [2015] for detailed discussions of the Maunder minimum.The geomagnetic record used to produce B[GEO] only commenced in 1845, although sporadic records are available during the eighteenth century.The cosmogenic estimates of B and the sunspot numbers (after 1609) are therefore the only known means to study the heliospheric and solar magnetic effects during the Maunder and earlier Grand Minima.On the basis of the comparisons made in this paper, we now summarize the uncertainties that will arise in the use of the estimates of B[GCR] in such studies.
1. Very large solar energetic particle events, similar to those of 26 July 1946 and23 February 1956, and especially extreme events like 775 AD [Miyake et al., 2012] which was the greatest documented SEP event over 11 millennia [Usoskin et al., 2013;Mekhaldi et al., 2015] will cause a significant reduction in the annual estimates of B based on the cosmogenic data.Provided annual 10 Be data are available from two or more ice cores, the largest solar contributions can be identified and removed from the cosmogenic data yielding B[GCR-MB2], except in the ascending phase of the Schwabe cycle.This may result in a <1.5 nT underestimate  10 Be enhancements is presently difficult prior to 1760.However, Figure 5 and Table 2

Future Improvements in B[GCR]
The statistical and measurement errors in the PCR data series set the limit on the accuracy of our estimates of B[GCR-MB2] and on our ability to identify solar energetic particle events during the ascending phase of the solar cycle.Improvements in both are possible and might include 1. Increasing the number of annual 10 Be records to a total of about five, with two of the new ones coming from the Antarctic.This is particularly important since both 10 Be series used here are from Greenland and may be potentially influenced by the regional climate variability [Usoskin et al., 2009;Beer et al., 2012].The overall goal would be a greater reduction in statistical and systematic noise, and minimization of long-term systematic changes that introduce errors into long-term comparisons.Five independent sets of data would yield a standard deviation of ~5.5% for the annual paleocosmic ray data, 0.3 nT for the annual estimates of the heliospheric magnetic field near Earth, and would allow smaller solar energetic particle events to be detected and eliminated from B[GCR-MB2].It would also permit the time profile of the 11 year cycle in the PCR to be determined for individual cycles ("sharp rising " or "flat topped") thereby identifying the polarity of the solar dipole into the past [Potgieter, 2013;Owens et al., 2015].2. Extending the annual PCR record back to 1000 before present (B.P.).( 950Common Era (C.E.)) to provide the ability to study the solar, cosmic radiation, and magnetic field effects with annual resolution through the whole cycle of Grand Minima from the Oort (~1040 C.E.) to the Gleissberg (1890 C.E.) Minima.3. Complementing the 10 Be records with records of 36 Cl and 14 C which are more sensitive to low-energy solar particles and differ in their geochemical properties.4. Improving the time assignment to ± 1 year and better.A relative error of greater than 1 year between two cores reduces the ability to identify the impulsive 10 Be events and thereby reduces the ability to detect and excise SEP events in the past.

Discussion and Conclusions
In this paper, we have compared the in situ observations of solar wind B with geomagnetic observations, sunspot time series, and the paleocosmic ray record to infer the solar wind magnetic field strength from 1750 to 2013.Our main results are summarized below where, for brevity, we refer to the estimates as B Ultimately, studies of the Grand Minima and millennia scale changes in the HMF will be based on the cosmogenicbased estimates, B [GCR].Their improvement and extension, as discussed in section 2.5, will be crucial for our understanding of such topics as the variability of the solar dynamo and terrestrial climate change.

Figure 1 .
Figure 1.Reconstructions of long-term variations in near-Earth magnetic field intensity, B, over the period 1700-2000, adapted from Svalgaard and Cliver [2010] and Steinhilber et al. [2010].The solid red line shows a 25 year running mean of the geomagnetic B reconstruction from Svalgaard and Cliver [2010].Other curves show cosmogenic radionuclide reconstructions of B. They are: 40 year running means (Φ St08 , blue and black lines) and 25 year running means (Φ PCA , purple line) from different heliospheric modulation potential estimates provided by Steinhilber et al. [2010]; 22 year running means of Caballero-Lopez et al. [2004] estimates (green and gold lines); 11 year running means of the McCracken [2007] estimate.
and ultimately into B[GCR] using the method ofLockwood et al. [2014].The mean of the two resulting B[GCR] estimates from the two ice cores yields the final data sequence used in this study, referred to as B[GCR-U].It is similar to B[GCR-MB1] in the sense that GLE events were not excluded, e.g., both exhibit very low values in the vicinity of 1893.However, while B[GCR-MB1] and B[GCR-MB2] use

Figure 2 .
Figure 2. 10 Be measurements from ice cores.Figure adapted from McCracken and Beer [2015].(a) Blue: The 10 Be concentration, normalized to the average for 1944-1987, after passing through a 1,4,6,4,1 binomial filter.Red: 11 year running means.(b) The impulsive increases in cosmogenic display a comparison of B[GCR-MB2] with B[GEO] and B[SSN], respectively.Examination shows that the two major discrepancies between B[GEO] and B[GCR-MB1] in ~1884 and ~1893 are substantially reduced in B[GCR-MB2],

Figure 3 .
Figure 3.Comparison of geomagnetic and cosmogenic radionuclide reconstructions of B over the period 1750-2013.(a-c) The green lines show the annual B[GEO] composite, passed through a 1,4,6,4,1 binomial filter.The green-shaded area shows the 90% confidence interval.The black line in Figure 3a shows B[GCR-U], the 10 Be reconstruction of B from Usoskin.The dark blue line in Figure 3b shows B[GCR-MB1], the 10 Be reconstruction of B from McCracken and Beer [2015].The yellow line in Figure 3c shows B [GCR-MB2], the McCracken and Beer [2015] estimate with impulsive 10 Be enhancements removed.(d) The 11 year running means of the annual data in the same format, with yellow/blue dashed lines indicating 11 year means of B[GCR-MB1] and B[GCR-MB2] when they have the same value.
B[GCR],B[GEO]  and B[SSN]  values for 3 year means centered on sunspot minimum, as defined inOwens et al. [2011].A 3 year mean is used to incorporate the uncertainty in both solar minimum timing and 10 Be deposition, though the results are similar for analysis of a single year.The 1845-1983 interval which is covered by B[GEO]

Figure 4 .
Figure 4. Comparison of the annual B[SSN] composite (red) with cosmogenic radionuclide reconstructions of B, in the same format as Figure 3.

Table 1 .
Comparison of the B[GEO] and B[SSN] Composites With B[GCR], Cosmogenic Radionuclide Reconstructions of B sufficient variability in the solar minimum B values to draw strong conclusions.Over the 1777-1983 interval covered by B[SSN], however, there is generally good agreement between B[GCR-MB1] and B[SSN] for the Schwabe minima (r = 0.69; MSE = 0.35 nT 2 ), even without the removal of 10 Be enhancements attributed to SEPs.B[GCR-U] shows the same trends as B[SSN] for Schwabe minima, though the long-term variability in B[GCR-U] is much greater than B[SSN], [OBS], B[GEO], B[SSN], and B[GCR], respectively.Journal of Geophysical Research: Space Physics 10.1002/2016JA022550 1.The original estimates of B[GCR] were based on paleocosmic ray data from a single experimental ice core.Using annual measurements from a second core, and after allowance for experimental uncertainties and long-term changes of atmospheric and geomagnetic origin, McCracken and Beer [2015] revised the earlier results upwards to obtain B[GCR-MB1].Section 2.3 examines the role of very large solar energetic particle events, such as that of 23 February 1956, in introducing significant reductions (~1.5 nT) into the estimates of B[GCR].The availability of data from two ice cores provides the ability to identify such solar energetic particle events in the past, except during the ascending phase of a solar cycle.McCracken and Beer [2015] excised eleven presumed solar energetic particle events from the PCR record, 1800-1980, leading to B[GCR-MB2].In the 11 year running mean data, there are still two ~1.5 nT excursions below in ~1860-1865 and ~1948, that we speculate are due to the production of solar cosmic rays during the second and third year of the solar cycle and consequently obscured by the rapidly decreasing cosmic ray intensities at those times.An independent estimate of annual B[GCR] based on the work of Kovaltsov and Usoskin [2010] and Usoskin et al. [2015], termed B[GCR-U], agrees reasonably well with B[GEO], B[SSN], B[GCR-MB1], and B[GCR-MB2] for the twentieth century but falls below those series prior to ~1900.This difference may be due to factors such as climate change or the procedure used to convert open solar flux to near-Earth B. 2. Analysis of the interval 1879-1940 shows that there is good agreement between all three estimates B[GEO], B[SSN], and B[GCR-MB2].The standard deviations of the statistical and measurement errors in annual 10 Be data are large (~20%), resulting in standard deviations of ~0.5 nT in the B[GCR] time series.The acquisition of annual 10 Be data from three new ice cores would provide enhanced ability to identify and eliminate solar energetic particle contributions in the PCR record and reduction of the standard deviation due to statistical fluctuations to ~0.3 nT.It would also provide the ability to determine the polarity of the solar dipole in the past and to provide insight into the differences between B[GCR] and the other estimates of B. 3.There are no direct or indirect measurements via B[GEO] or B[SSN] of B < 3.5 nT.Thus, the accuracy of B[GCR-MB2] estimates below this value is entirely dependent on the continued applicability of the cosmic ray propagation equation.Present day knowledge indicates that the estimates will be reliable for B > 2.5 nT, though it is important to undertake further study of the propagation equation for low values of B in the heliosphere.
. One goal of the ISSI workshop discussed in Part 1 was to use both of these recent advances to update the earlier annual estimates of B[GCR] which were based on an experimental ice core from Dye 3 alone and to compare them with B[GEO] and B[SSN].
Steinhilber et al. [2012] Beer [2015]combined the Dye 3 and North GRIP10Be data, reduced climatic and other terrestrial effects using the composite result ofSteinhilber et al. [2012], and estimated B[GCR] and expected equivalent neutron monitor count rates, dubbed the "pseudo-NM" data, for 1391-1983.Figure2a displays the 10 Be concentration for the interval 1400-1983; note it was substantially elevated during the Maunder (~1695) and Dalton (~1815) "Grand Minima" in the past.Tables of these annual data together with annual estimates of the HMF (see section 2.3 below) are given inMcCracken and Beer [2015].

Table 2 .
Comparison GCR-MB2] after passage through a 1,4,6,4,1 filter for ~15% of all Schwabe cycles.At present, there is no means to compensate for the occurrence of smaller solar events, and this may result in underestimation (<1 nT) of annual B[GCR] near sunspot maximum.2. There is reasonable agreement between the absolute values of B[GEO], B[SSN], and B[GCR-MB2] throughout the range 4-9 nT.The statistical errors in the cosmogenic data imply a standard error of 0.5 nT for annual estimates of B[GCR-MB2] based upon Dye 3 and North GRIP alone.The difference in the B[GCR-MB1] and B[GCR-U] series, particularly the lower values of B[GCR-U] prior to 1900, may reflect uncertainties related to the corrections for climate effects and the different models used (i.e., numerical Monte-Carlo models of cosmic ray induced cascades, modeling of 10 Be transport and deposition, and calibration of the models) and indicates that further work is warranted in this area.3. Identification of impulsive show that the sunspot minimum values of B[GCR-MB1] and B[GCR-U] are well correlated with B[SSN].Consequently, it will be possible to use Schwabe minimum estimates of B[GCR] to investigate long-term changes in the HMF far into the past.The 40% decreases in the PCR during the Schwabe cycle means that solar minima and maxima are identifiable without difficulty and without reference to a sunspot record.Note, however, that at periods of very low solar activity, it has been postulated on the basis of open solar flux continuity modeling[Owens and Lockwood, 2012]that the HMF variation and the solar cycle variation may shift into antiphase[Owens et al., 2012], which could complicate such analysis.4.There are no direct spacecraft measurements or indirect (B[GEO] or B[SSN]) estimates for annual B < 3.5 nT, and therefore, the accuracy of B[GCR-MB2] estimates below this value are dependent wholly upon the applicability of the cosmic ray propagation equation in this parameter regime.This requires further study.Present-day knowledge indicates that the estimates should be reliable for B[GCR-MB2] > 2.5 nT [McCracken and Beer, 2015].