Using high-precision and centennial-resolution ice core
information on atmospheric nitrous oxide concentrations and its stable
nitrogen and oxygen isotopic composition, we quantitatively reconstruct
changes in the terrestrial and marine
Nitrous oxide (
Reconstructions of past variations in terrestrial and marine
Using the different nitrogen isotopic signature of terrestrial and marine
Accordingly, the aim of this study is to improve our understanding of the
changes in the main
A total of 202 ice core samples were analyzed at the University of Bern for
The new data cover the period from about 27 to 0.3 ka BP and allow us to
reconstruct for the first time the
The analytical precision for the two systems used at the University of Bern
and for different sample batches is given in Table 1. Error bars of all
isotopic data in this publication represent one standard deviation (
Measurement precision (
Using measurements performed on various Greenland and Antarctic ice cores,
composite records for
Compilation of new and published
In order to compare
Methane synchronization of ice core records.
A well-known problem with
The most straightforward criterion to identify samples affected by in situ
production targets anomalies in the
To obtain a final data set compiled from all cores after the pre-screening
described above and to rigorously remove samples that may still be affected
by in situ processes, individual samples were finally excluded in our study
from the data set compiled from all ice cores if they were at least 20 ppb
higher than the average of our Monte Carlo smoothing spline (see details on
our Monte Carlo Average (MCA) below). For a group of EDML measurements (22–18 ka BP) this MCA method could not be applied, because precise
Greenhouse gas records over the last 21 kyr after removing
As none of the isotope and concentration records is continuous and as
isotope measurements are only available at a lower resolution, we derived a
common lower-frequency record using a spline fitting routine (Enting,
1987), which served as input into the box model inversion used to calculate
emissions. As the time interval younger than 16 ka BP was better resolved
than the glacial part of the record, we used a smaller cutoff period (higher frequency content) of the spline for ages
To derive the uncertainty of the spline approximation representative for the timescale resolved by the low-pass filter, the spline fitting was repeated 1000 times using a Monte Carlo approach, where the measurements were varied randomly within their analytical uncertainties. This Monte Carlo spline and its uncertainty was then used to detect outliers affected by in situ formation as described above. The procedure was iterated until no more outliers were found and the final Monte Carlo spline was calculated using the final data set (Fig. 3).
To assess the accuracy of ice core
Late Holocene evolution of
In case of the
The
Also, our
We deconvolved the evolution of tropospheric
The minimal two-box atmosphere model features a tropospheric and
stratospheric box as well as a land and ocean source and includes
A Monte Carlo approach is used in the deconvolution to estimate
uncertainties in
Parameter ranges used in the atmospheric two-box model for the
calculation of terrestrial and marine
In order to test whether the model is able to reliably reconstruct
terrestrial and marine emissions during rapid changes in the
Due to the slightly different atmospheric lifetimes of
We performed three test runs assuming (i) a ramp-up of land emissions only
over 50 years with a constant nitrogen isotopic signature of
Deconvolution of artificial ice core data for a hypothetical 50-year
emission ramp-up of (i) terrestrial emissions, (ii) marine emissions and
(iii) terrestrial and marine emissions in a fixed ratio.
The results of these performance tests are summarized in Fig. 5 with the
increase in
As can be seen in Fig. 5b and c, the
The long tail of old air in the gas age distribution together with the spline approximation is also the reason for the time needed in the emission reconstruction based on the splined artificial ice core data to reach its maximum, which takes several hundred years longer than the 50 years ramp-up in the originally assumed emission increase. Note that a log-normal age distribution is overestimating the width of the true age distribution in the ice and especially the long tail of old air is not realistic. The effect of this low-pass filtering on the synthetic data is, therefore, stronger than the effect of the true bubble enclosure in the ice and the effects illustrated in Fig. 5 can be regarded as an upper limit of the true delay and the response time. Note also that the reconstructed emission increases in Fig. 5d appear to start about 2 centuries earlier than the major increase in the input fluxes and the artificial ice core data. This lead is an artifact created by the use of the splined artificial ice core data as input to our model.
In view of the reasonable model performance described above, the comparison
of the artificial data runs with the true ice core reconstructions for the
BA warming in Fig. 5 suggests that for example the rapid
In summary, our model can reliably separate terrestrial and marine
emissions. The onset of rapid increases in total
In two sensitivity tests, we investigated the influence of the prescribed
initial range of the marine contributions to total
Results from sensitivity analyses to test the robustness of
Amundson et al. (2003) provide empirical relationships
between
As a measure of potential changes in
This yields two scenarios for isotopic source change. The temporal changes
in
Finally, we tested the sensitivity of our box model approach to temporal
changes in the atmospheric lifetime (yellow lines in Fig. 6c–e), which is
controlled by the troposphere/stratosphere exchange. The Brewer–Dobson
circulation and atmospheric lifetime of
Reconstructed tropospheric
During the deglaciation, the
During the early Holocene
The
Reconstructed global terrestrial
Reconstructed terrestrial and marine
Compared to the drastic emission changes over the deglaciation, the trends
in terrestrial
Our new ice core isotope data permit novel insights in the response timescales of terrestrial and marine
The low-pass filtering also leads to an underestimation of the true
atmospheric dynamics in
A similar picture in terms of the timing of the onset and timescale of
In summary, the terrestrial
We further test the robustness of the Monte Carlo deconvolution approach to
infer changes in global
Second, the influence of the prior assumption on the initial fraction of
marine emissions relative to total
Finally, we varied the atmospheric lifetime over the deglacial period in an
idealized scenario assuming an overall decrease in lifetime from 143 to 123 years from the glacial to the Holocene. This change in lifetime causes a
parallel increase in both land and ocean emissions by about 16 %. Late
Holocene emission anomalies relative to 21 ka BP increase by about 0.6 Tg yr
In conclusion, the main features of our standard reconstruction such as the
decrease in and recovery of global marine
We followed the approach by Schilt et al. (2014) and deconvolved the ice
core records of
The temporal resolution of reconstructed
Marine
Marine
Interestingly, both reconstructed marine
Paleo
While the centennial- to millennial-scale oscillations in terrestrial and
marine emissions over the Holocene cannot be interpreted with certainty due
to the still limited sampling resolution and precision of the Holocene
Generally, the difference in the timescales associated with the
reconstructed marine and terrestrial emission changes during the YD and over
the entire termination is consistent with the notion that terrestrial
Terrestrial emissions show a 40 % increase from the Last Glacial Maximum
to the late Holocene period. Most of the deglacial increase was realized in
two fast and large steps. These occurred at the onset of the BA and at the
end of the YD, when reconstructed terrestrial
Reconstructed terrestrial
The apparent decoupling of significant net terrestrial carbon storage from
We reconstructed the evolution of the stable isotopes
For the last 21 000 years, where (isotopic)
Our records show that both terrestrial and marine
Data presented in this study will be available on the NOAA paleoclimate database.
The concept of the study was developed by HF, JS, RS, FJ, AS and EJB. Analytical methods for
The authors declare that they have no conflict of interest.
Long-term financial support of this research by the Swiss National Science Foundation (grant no. 200020_172506 and 200020_172476) is gratefully acknowledged. Part of the research leading to these results has received funding from the European Research Council under the European Union's Seventh Framework Programme (FP/2007-2013)/ERC Advanced Grant Agreement no. 226172 (MATRICs) awarded to HF. This work is also a contribution to the European Project for Ice Coring in Antarctica (EPICA), the Talos Dome Ice Core Project (TALDICE), and the North Greenland Ice Core Project (NGRIP). EPICA is a joint European Science Foundation/European Commission (EC) scientific program, funded by the EC and by national contributions from Belgium, Denmark, France, Germany, Italy, the Netherlands, Norway, Sweden, Switzerland, and the UK. The main logistic support was provided by IPEV and PNRA (at Dome C) and AWI (at Dronning Maud Land). TALDICE is a joint European program led by Italy and is funded by national contributions from Italy, France, Germany, Switzerland, and the United Kingdom. The main logistical support was provided by PNRA at Talos Dome. NGRIP is directed and organized by the Department of Geophysics at the Niels Bohr Institute for Astronomy, Physics and Geophysics, University of Copenhagen. It is supported by funding agencies in Denmark, Belgium, France, Germany, Iceland, Japan, Sweden, Switzerland, and the USA. This is EPICA publication no. 312 and TALDICE publication no. 57.
This research has been supported by the Swiss National Science Foundation (grant nos. 200020_172506 and 200020_172476).
This paper was edited by Sönke Zaehle and reviewed by two anonymous referees.