Despite the extensive use of 10Be as the most significant information source on past solar activity, there has been only one record (Dye-3, Greenland) providing annual resolution over several centuries. Here we report a new annual resolution 10Be record spanning the period 1389–1994 AD, measured in an ice core from the NGRIP site in Greenland. NGRIP and Dye-3 10Be exhibits similar long-term variability, although occasional short term differences between the two sites indicate that at least two high resolution 10Be records are needed to assess local variations and to confidently reconstruct past solar activity. A comparison with sunspot and neutron records confirms that ice core 10Be reflects solar Schwabe cycle variations, and continued 10Be variability suggests cyclic solar activity throughout the Maunder and Spörer grand solar activity minima. Recent 10Be values are low; however, they do not indicate unusually high recent solar activity compared to the last 600 years.
 The cosmogenic isotope 10Be plays a central role in retrieving information about past cosmic ray intensity and solar activity, factors whose effect on climate is under debate [i.e., Carslaw et al., 2002; Lockwood and Fröhlich, 2007]. Deposition of atmospheric 10Be into polar ice sheets, through wet (associated with precipitation) and dry deposition, creates a unique archive with the possibility of annually resolved information on past solar activity and consequently for understanding possible connections to past climate change [Beer et al., 1990]. The isotope concentration in glacial ice is affected by changes in production rate and transport and deposition processes. Changes in production rate depend on solar activity and geomagnetic field strength. The transport and deposition is influenced by atmospheric mixing (e.g., stratosphere-troposphere exchange), scavenging and snow accumulation rate. Comparison with 14C and modeling [Heikkilä et al., 2008a] shows that production is the dominant global signal, while the transport or weather signal is weaker and of a local nature. Combined snow pit 10Be measurements and local weather data covering almost a year at Law Dome, Antarctica, indicate that ∼30% of short term 10Be variability is related to meteorological factors [Pedro et al., 2006]. The Law Dome site is, like Greenland, dominated by wet deposition.
 There are a few annual resolution 10Be records which span short periods [e.g., Aldahan et al., 1998; Heikkilä et al., 2008b; Moraal et al., 2005; Smith et al., 2000; Steig et al., 1996]. To our knowledge there is only one existing 10Be record with annual resolution that covers several centuries, a record from the Greenlandic Dye-3 site (65.18°N, 43.83°W, 2480 m a.s.l.) covering the period 1428–1985 AD [Beer et al., 1990, 1998]. A combination of two high resolution records will aid in isolating regional and local influences on 10Be distribution in glacier ice, and indicate how well measured 10Be in an ice core reflects the production signal. A new annual 10Be record covering the past 600 years will also be optimal for reconstructing variations in the 11-year Schwabe cycle which plays a crucial role in obtaining a better understanding of the long-term behavior of the solar dynamo and its potential relationship to climate changes on decadal and centennial time scales. A recent annual record also enables calibration between 10Be fluxes and high resolution neutron flux data, a cosmic ray parameter which has been monitored since 1953.
2. Method and Description
 To address some of the issues mentioned above, we have analyzed annual variability of 10Be during the period 1389–1994 AD in the NG 97-S2 shallow core from the NGRIP site (75.10°N, 42.32°W, 2917 m a.s.l.), Greenland (see Dahl-Jensen et al.  and Hvidberg et al.  for details on NGRIP). The timescale is based on annual layer counting in deconvoluted δ18O data, and has been synchronized to the GICC05 timescale [Vinther et al., 2006]. Synchronization has been accomplished by matching volcanic reference horizons in the NG 97-S2 shallow core to those observed in the NGRIP deep core. Net accumulation rates are calculated from annual layer thicknesses with a coupled density/ice-flow model. Densities are based on [Herron and Langway, 1980] as modified by [Johnsen et al., 2000] to better fit Greenland ice core density profiles. The ice-flow model is from [Johnsen and Dansgaard, 1992].
 The 10Be in the most recent 318 annual layers was extracted at Eawag and measured at the Ion Beam Physics laboratory, ETH, Zurich, Switzerland. The 10Be of the 288 older annual layers was extracted at the Department of Earth Sciences and measured at the Tandem Laboratory, Uppsala University, Sweden. Both laboratories used the same chemical separation procedure and carrier solution and inter-calibrated standards to minimize inter-laboratory discrepancies.
 The Dye-3 10Be record has annual resolution after 1780 AD, while in the older part of the core the sampling was made equidistantly, leading to samples covering 0.4–2.3 years. The timescale is based on annual layer counting in H2O2 data for the period 1777–1986 AD. From 1424 to 1777 AD the timescale has been synchronized to the GICC05 timescale [Vinther et al., 2006], by assuming identical accumulation rates in the Dye-3 core used for 10Be measurements and the Dye-3 deep core. Dye-3 accumulation rates are calculated as for NGRIP. We have adjusted the existing Dye-3 10Be data to the new re-calibrated timescale.
 The 10Be concentration and flux and snow accumulation rate are shown in Figure 1. NGRIP 10Be concentration ranges between 0.8 and 3.9 × 104, with an average of 1.8 × 104 atoms gice−1, and the flux is 0.003–0.023 with an average of 0.010 atoms cm−2 s−1. The flux is the 10Be deposition rate at the surface, and is calculated by multiplying each sample concentration with the snow accumulation of that specific year.
 NGRIP 10Be data show high inter-annual variations which are superposed on wider fluctuations of an irregular nature. To some extent periods of high 10Be values correspond to grand solar activity minima, most prominently during the Maunder (∼1645–1715 AD) and Dalton (∼1790–1830 AD) solar minima, while less distinctly during the Spörer minimum (∼1415–1535 AD). Since snow accumulation varies over time, the 10Be flux and concentration curves differ somewhat. The long term variations are similar in both parameters, except during parts of the Dalton and Maunder minima.
 The Dye-3 10Be concentration ranges from 0.4 to 5.0 × 104 atoms gice−1, while the flux range is 0.005–0.078 atoms cm−2 s−1 (Figure 1). 10Be in NGRIP and Dye-3 data have similar long term variability, although annual variations differ. Occasionally peaks occur simultaneously in both cores, e.g., at 1460 AD. This specific peak, which is especially large in Dye-3, has also been observed in South Pole [McCracken et al., 2004]. In the older part of the period, there is higher 10Be variability in Dye-3 than in NGRIP.
 During the period when data is available from both sites the average measured concentration at Dye-3 (1.0 × 104 atoms gice−1) is ∼60% of NGRIP. This is at least in part a result of a dilution effect, since snow accumulation at Dye-3 is 2–3 times higher than at NGRIP, as is clearly shown by the Dye-3 flux which is 1.6 larger (0.016 atoms cm−2 s−1) than the NGRIP average. The higher average 10Be flux at Dye-3 suggests a geographic dependency of beryllium deposition, as has also been modeled [Field et al., 2006; Heikkilä et al., 2008a].
4. Discussion and Interpretation
 With the two geographically separated annual resolution 10Be records we can examine to what extent flux and concentration represent regional deposition. A change in atmospheric circulation and transport patterns may bring air masses with varying 10Be content over longer time scales. The difference in geographic location of NGRIP and Dye-3, regarding not only latitude and altitude but also distance from the coast, is likely to affect transport to and deposition at the two sites differently [Heikkilä et al., 2008a]. Annual 10Be signals at the two sites are not directly comparable, due to local weather effects. Temporally extensive local differences in 10Be deposition are exposed by application of a low pass filter with a cut-off frequency of 1/22 years (Figure 1). Before the mid-16th century, 10Be expressions are different in NGRIP and Dye-3, which could possibly be caused by a complex ice flow regime at Dye-3, where ice from the oldest strata has experienced significantly different summer melt and accumulation conditions than observed presently at the Dye-3 drill site [Herron et al., 1981]. However, there are periods when long term 10Be variations are similar in the two cores, especially when the flux is considered. Dye-3 flux decreases in the 20th century, a fact that has previously been used to support a theory that the sun has been unusually active since the 1940s [Usoskin et al., 2003], although this conclusion has been disputed by [Raisbeck and Yiou, 2004], based on South Pole 10Be data. Our new data from NGRIP does involve low 10Be values during much of the 20th century, although by no means at an unusually low level. These different signals from the Greenland sites in the second half of the 20th century highlights the importance of using a combination of several high resolution datasets in conjunction to assess core uncertainties and local variability in 10Be deposition in order to make confident reconstructions of past solar variability. It is also clear that it is preferable to have 10Be records from sites with non-complex ice flow regimes, and no or infrequent summer melt, such as NGRIP and South Pole.
 With high resolution 10Be datasets, the response in 10Be to solar activity variability can be investigated in detail. Because the length of the Schwabe solar cycle during the investigated period varies between seven and 17 years, this is the first time where two datasets have sufficient resolution to capture this variability. Although 10Be data from the southern hemisphere would be a great contribution in this context, existing data is of too low resolution to accurately reflect the Schwabe cycle variations. Since we here focus on decadal variations, we exclude lower resolution data and refer interested readers to previous studies where 10Be from several cores has been compared [e.g., Muscheler et al., 2007].
 We apply a band pass filter around the length of the Schwabe solar cycle, with an 8-year cut-off at the lower end to remove weather noise, and 16 years at the upper end (Figure 2). In concentrations, we note a very good agreement between cores in the last two centuries, while further back concentrations in the two cores are periodically out of phase. Between fluxes in the two cores, the agreement is stronger, and differences mainly occur in part of the 17th century. At times, one core indicates two cycles when the other only indicates one. For instance, around 1780 AD there are two cycles in NGRIP concentration where Dye-3 has only one. Around 1810 AD there is one peak indicated in the NGRIP flux data, while there are two in Dye-3. In the NGRIP data, there are similar occasional differences between the parameters; while concentration has two periods around 1580 AD, there is only one in the flux, and the opposite is true around 1650 AD. Around 1780 AD there is some phase shift between the two parameters. In Dye-3, differences between concentration and flux exist mainly around 1490–1530 AD and partly during 1680–1770 AD.
 In order to investigate how these periods and differences correspond to solar activity, we compare NGRIP and Dye-3 concentration and flux with frequencies found in the sunspot number record (Figure 2). We note that NGRIP 10Be concentration and flux both have a negative relationship to solar activity, although there are occasional leads and lags before 1820 AD and during 1880–1910 AD in both 10Be parameters relative to the solar data. In Dye-3, there are phase differences until around 1780 AD, after which 10Be is well synchronized with the sunspots. This indicates that there is either some dating uncertainty in the older part of the cores, where Dye-3 dating was established by a different method than in the newer part, or there was a slower response in the 10Be deposition to changes in solar activity. It should be stressed that the good agreement between NGRIP and Dye-3 fluxes suggest that remaining dating inaccuracies are small. In the period around 1800 AD when NGRIP 10Be is slightly out of phase with the sunspot cycle, Dye-3 concentration is in phase, underlining the importance of having data from at least two high resolution cores for an accurate solar activity reconstruction. In addition, we reconfirm earlier findings [Beer et al., 1998] of a cyclically active sun during solar minima; a clear Schwabe cycle is present in both cores during the Maunder minimum, especially so in NGRIP.
 Finally, we consider the correlation between 10Be from the ice cores with neutron data (Figure 3). While sunspots reflect the closed solar magnetic field, which is not directly related to the open magnetic field that modulates the cosmic ray intensity, neutron monitors record the cosmic ray intensity with high resolution and high precision. They therefore provide the best possible parameter to compare 10Be with, except for the fact that 10Be production is slightly more sensitive to cosmic rays with lower energy [Beer, 2000]. We observe a high degree of agreement of 10Be concentration and flux in NGRIP and Dye-3 to the atmospheric neutron flux, which proves the usefulness of 10Be in ice cores to reconstruct past solar activity.
 In conclusion, our 10Be data from NGRIP provides a new archive that is strongly needed for identifying temporal and spatial variability of 10Be deposition over Greenland, thereby enabling more accurate solar activity reconstructions. Comparison over several centuries of our data with earlier data from Dye-3 indicates that despite the existence of local noise and some differences prior to the mid-16th century, the regional nature of the 10Be signal in ice cores is confirmed. The good long-term agreement between 10Be variations in both cores reflects a regional response to production and climate changes, but the disagreements in the earlier parts of the two records suggest that 10Be should be measured in ice cores from locations with non-complex ice flow regimes. 10Be deposition is anti-correlated to solar activity over the 11-year Schwabe solar cycle, and correlated to neutron monitor data. Periodicity in 10Be during the Maunder minimum reconfirms that the solar dynamo retains cyclic behavior even during grand solar minima. We observe that although recent 10Be flux in NGRIP is low, there is no indication of unusually high recent solar activity in relation to other parts of the investigated period.
 We acknowledge the financial support from the Swedish Research Council and the Swiss National Science Foundation. NGRIP is directed and organized by the Ice and Climate research group, Niels Bohr Institute, University of Copenhagen. It is supported by funding agencies in Denmark (FNU), Belgium (FNRS-CFB), France (IPEV and INSU/CNRS), Germany (AWI), Iceland (RannIs), Japan (MEXT), Sweden (SPRS), Switzerland (SNF) and the USA (NSF, Office of Polar Programs). We thank Irene Brunner for preparing the Zurich samples. Anonymous reviewers provided useful comments on earlier versions of the manuscript.