O  record  in  a  low-&shy

Volltext

(1)36th Meeting of the AWI Scientific Advisory Board (SAB) Potsdam 6/7 May 2015. . Thomas Münch, Sepp Kipfstuhl Johannes Freitag, Hanno Meyer and Thomas Laepple. Signal vs. Noise: 18 Obtaining a representa4ve δ O record in a low-‐. accumula4on region 1 A 2D view of Antarctic δ18O. 1A Trench1. 15. ● ●. 0.4. 0.6. 0.8. profile correlation (after optimal shifting). year 2012. 2011. 2010. 2009. 2008. trench1 trench2 - not shifted trench2 - shifted by 3 cm. -40. δ18O [‰]. -42. -44. -46. -48. 2B. trench1 annual mean trench2 annual mean 0. 20. 40. 60. 80. 100. ●. ●. ●. ●. ●. ●. ●. ●. ●. ●. ●. ●. ●. ●. ●. ●. ●. ●. ●. 40 depth [cm]. 60. 20. 60. −50. 80. 80. profile positions:. 100 10. 20. 30. 0.3 m 10 m 29.8 m 40 m. 100. −55 40. -50. trench position [m]. -48. -46. 4. 3A 30. 25. 20. 15. 10. 5. 0. The signal and noise content of the trench record is expressed by the down-‐core variance, the lateral variance (Fig. 3A) and the spa4al covariance of the variability. A simple model of the stra2graphic noise structure (the spa2al covariance) is given by an AR(1) autoregressive process, as suggested by the isotopic seasonal layering and the inter-‐profile correla2on (Fig. 3B): 0. 20. 40. 80. 100. Xn (t) = S(t) + a"n. 1 (t) +. p. a2 "n (t) (1). 1. Assuming independence of signal and noise, the inter-‐profile correla2on as well as the correla2on between two sets of profile averages can then be given analy2cally for AR(1) noise: . rXY. upper bound of correlation with climate signal. rX̄ Ȳ. |x y|/ l. var(S) + a var(") = (2) var(S) + var(") ( ! !) ⇤ 2 ⇤2 var(") X̄ var(") Ȳ ⇡ 1+ 1+ 2 var(S) NX̄ var(S) NȲ2. 3C. inter-profile correlation. 1.0. 2 m profile spacing 20 m profile spacing AR(1) subannual noise dependent subannual noise seasonal signal for 20 m profile spacing. For annual δ18O 4me series, the ver2cal noise structure has to be known. Due to the limited ver2cal extent of the 1 5 10 15 20 trench record, it can only be approximated. In the best number of required profiles case, we assume the ver2cal noise to be also following an AR(1) process (dashed lines, Fig. 4), in the worst case we assume a complete dependence on the sub-‐annual 2me scale (dash-‐doeed lines, Fig. 4). . 60. depth below mean snow surface [cm]. 0.8 0.4. -38. parallel to horizon parallel to snow surface. 3B 0.2. -40. lateral trench variance. 0.6. correlation. Le`ng NY→∞ in Eq. (3) yields the correla2on with a hypothesised true climate signal (red doeed line in Fig. 4). . -42. 3 A model for stratigraphic noise. 1.0. The correla2on of single profiles in trench 1 (Fig. 3B) as well as that between the trenches (Fig. 3C) is well captured by our AR(1) model of the stra2graphic noise. . -44. δ18O [‰]. Ø We aim at a be)er understanding of noise in firn cores to allow a reliable es6mate of the climate signal recorded. . 4 Implications. 0.9 0.8. • the lateral noise structure is well-‐modelled by an AR(1) process • the ver2cal noise structure needs to be further inves2gated to constrain the repre-‐ senta2vity on annual and longer 2me scales . References Fisher, D. A., N. Reeh and H. B. Clausen (1985). Stratigraphic noise in time series derived from ice cores. Ann. Glaciol. 7, 76-83. Gfeller, G., H. Fischer, M. Bigler, S. Schüpbach, D. Leuenåberger and O. Mini (2014). Representativeness and seasonality of major ion records derived from NEEM firn cores. The Cryosphere 8, 1855-1870, doi: 10.5194/tc-8-1855-2014. Karlöf, L., D. P. Winebrenner and D.B. Percival (2006). How representative is a time series derived from a firn core? A study at a low-accumulation site on the Antarctic plateau. J. Geophys. Res. 111, doi: 10.1029/2006JF000552. Johnsen, S. J., H. B. Clausen, K. M. Cuffey, G. Hoffmann, J. Schwander and T. Creyts (2000). Diffusion of stable isotopes in polar firn and ice: the isotope effect in firn diffusion. In: Physics of Ice Core Records 159, ed. by T. Hondoh, 121-140. Münch, T., S. Kipfstuhl, J. Freitag, H. Meyer and T. Laepple (2015). Signal vs. Noise: A δ18O representativity study at a low-accumulation site on Dronning Maud Land, Antarctica. Manuscript in preparation. Persson, A., P. L. Langen, P. Ditlevsen and B. M. Vinther (2011). The influence of precipitation weighting on interannual variability of stable water isotopes in Greenland. J. Geophys. Res. 116, doi: 10.1029/2010JD015517.. 1/2. (3). correlation with trench2 profile set. 1.0. Trench1 Trench2 signal+AR(1) noise. trench results theoretical curves for AR(1) noise. 0.8. 0.7 0.6. 0.6. 0.5 0.4. 0.4 2m 5m 10 m 20 m. 0.3 0.2. Key findings: • Dune-‐induced stra2graphic noise is the most important noise component on spa2al scales of at least 500 m Ø Seasonal and annual δ18O 2me series from single firn cores show low representa2vity . 1B Trench2. −35. ●. −45. depth [cm]. . ●. 40. To obtain a representa2ve isotopic signal for our study site, the spa2al mean of all single isotopic profiles from the two firn trenches is calculated (Fig. 2B). The 38 trench 1 profiles are spaced between 0.1 and 2.5 m, the 4 trench 2 profiles at 10-‐20 m. The mean profiles are well correlated (R=0.75), indica2ng a common clima2c signal over a spa2al scale of 500 m. However, s2ll significant profile devia2ons are found. . . ●. −40. mean correlation. 2A. 0.2. ●. 20. variance [(‰)^2]. 0. ●. 0. 10. 5. ●. correlation. 20. ● ●. 0. Single isotope profiles of trench 1 show a broad distribu2on of correla2ons with the single profiles obtained from trench 2 (Fig. 2A). . 25. ● ● ●● ●●●. 0. 2 A representative firn record. frequency. ●●. δ18O [permil]. The trench approach (Fig. 1): • 2 firn trenches of 45 m length x 1.2 m depth • excavated at Kohnen sta4on, Dronning Maud Land (see picture on the top right) • spa4al separa4on 500 m • isotopic sampling of 38, respec2vely 4, profiles . depth [cm]. Single ice cores from low-‐accumula2on regions in Antarc2ca may not be representa2ve recorders of the Holocene climate evolu2on (Karlöf et al., 2006). Similarly, Gfeller et al. (2014) found ionic aerosol proxies from a single firn core at the Greenland NEEM site recording only 30 % of the inter-‐annual atmospheric variability. Both findings show that the climate signal in firn cores is obscured by noise processes on different spa2al and temporal scales (Fisher et al., 1985; Persson et al. 2011; Johnsen et al., 2000). . 0.2 0. 5. 10. 15. 20. 25. 30. 35. 5. inter-profile spacing [m]. Model Parameters: • noise variance var(ε); given by the lateral trench variance • signal variance var(S); given by the down-‐core variance reduced by var(ε) . 10. 15. 20. number of required profiles. • autocorrela2on parameter a; es2mated from autocorrela2on func2on of the lateral trench variance . POTSDAM Telegrafenberg A43 14473 Potsdam Telefon 0331 288-2228 Thomas.Muench@awi.de www.awi.de/en/research/young_investigators/helmholtz_university_young_investigators_groups/ecus/.

(2)

O &nbsp;record &nbsp;in &nbsp;a &nbsp;low-&shy

O record in a low-&shy