-1 0 1 2 3 Peat C Acc. anomaly 10 20 30 40 50 60 Number of sites -6 -4 -2 0 2 4 6 -6 -4 -2 0 2 4 6 -4 -2 0 2 4 6 8 -4 -2 0 2 4 6 8 -6 -4 -2 0 2 4 -6 -4 -2 0 2 4 0 1000 2000 3000 4000 5000 6000 7000 8000 9000 year BP 3-Methodology (a suit of functions) Charly Massa 1 , Zicheng Yu 1 , Maarten Blaauw 2 , Julie Loisel 1 Author for correspondence : chm313@lehigh.edu Lake Igaliku Uwitec coring platform at lake Igaliku Bayesian MC-EOF analysis of PCAR with three modes (MCEOF1, 2 and 3) Upper panel - Scaled time series (grey lines, median of 1000 iterations), median value (green line) and number of site per 500 yrs bin (red line) Lower panels - Temporal patterns (MCEOFs, blue lines) with 95% confidence intervals based on 10,000 Monté-Carlo iterations (blue shades) Right panel - Spatial modes (maps of loadings) References Anchukaitis, K. et al. (2012) Climate Dynamics, 41: 1291-1306. Loisel, J. et al. (2014) The Holocene (in press). Sundqvist, H.S. et al. (2014) Climate of the Past Discussions, 10 : 1-63. 1. Department of Earth and Environmental Sciences, Lehigh University, Bethlehem, PA, USA 2. School of Geography, Archaeology and Palaeoecology, Queen’s University, Belfast, United Kingdom. 1/ Spatiotemporal patterns of peat carbon accumulation rate (Artic Peatland Database, 66 records) 2/ Spatiotemporal patterns of Arctic temperature proxies (Arctic Holocene Transition Database, 167 records) Exploring Spatiotemporal Patterns of Holocene Carbon Dynamics in Northern Peatlands by Incorporating Bayesian Age-Depth Modeling into Monte-Carlo EOF Analysis X = Step 1: Bayesian age depth modelling 1000 Peat accumulation rate Step 2: Calculation of Peat C accumulation rates Step 3: Smoothing & binning C bulk densities ±10% of assumed analytical error 1000 C acc. rate time series/site Focus on multi-millennial variability Step 4: Random sampling Buid 10,000 datasets combinations by randomly sampling 1 PCAR iteration out of 1000 for each site 2-The Circum Arctic Database 4-Results 1-Introduction The database (Loisel et al., 2014): - 268 cores (215 sites) with peat properties - 151 cores with age information (127 sites) Core selection criteria for B-MCEOF: - Bulk density and C data available - Cover at least the last 4500 years - Length/ N of 14 C dates < 2500 cal. years - Less than 5000 cal. years between 2 dates 63 high quality records EOF analysis of temperature proxies with three modes (EOF1, 2 and 3) Upper panel - Scaled time series (grey lines), median value (green line) and number of site per 500 yrs bin (red line) Lower panels - Temporal patterns (EOFs, blue lines) Right panel - Spatial modes (maps of loadings) The first three components explain 47.6% of the overall peat C accumulation rate (PCAR) varia- billity, suggesting that the MC-EOF analysis catch significant large scale climate-driven pat- terns. The lower percentage of variance captured by the analysis of peatland records (9.6% less than for the AHT records) is likely due to the more noisy nature of PCARs and/or the sensitivity of peatlands to local and ecological factors. MCEOF-1 shows robust multi-millennial scale change, consistent with the main temperature trend of the Holocene: the Neoglacial cooling. This result illustrate the potential of PCAR as a paleoclimatic proxy for the Circum Arctic analy- sis. The temporal patterns of PCAR and tempera- tures share strong similarities (based on their 2 leading EOFs), suggesting a dominent tempera- ture control. But the peat records are too spar- sely distributed to allow for the comparison of spatial pattern (sites loadings) and thus, go fur- ther into this interpretation. EOF (Empirical Orthogonal Functions): Common tool for exploring the spatiotemporal modes of ins- trumental climate data, but rarely applied to paleo proxy records. Monté-Carlo EOF (Anchukaitis et al. , 2012): Adap- ted to time-uncertain paleoclimate proxy records. Problem: Flux-based proxies (e.g. peat C accumula- tion rates) are strongly dependent to age-depth modeling (small uncertainties in ages may lead to large differences in accumulation rates). Investigate the spatiotemporal patterns and climate controls of peat C accumulation from Northern peatlands. Compare the spatiotemporal patterns of peat C accumulation with other climate proxy datasets on the basis of their leading EOF modes (common variance) 1000 Peat accumulation rate Regional scale analysis (e.g. North America) with a higher temporal resolution. Compare with EOFs of globally gridded cli- mate data (e.g. Climatic Research Unit (CRU) Time-Series). Reduce spacial heterogeneity and regional overweighting by averaging records over a 0.5° latitude x 0.5° longitude grid. Combined analysis of the Circum Arctic PCAR Database with other paleoclimatic data- bases, such as AHT (Sundqvist et al., 2014). 86-Kvartal Aero Altay BearBog ....... Zoige 500 2.2 12.9 9.4 30.2 ....... 54.7 1000 6.3 16.7 4.3 25.0 ....... 27.6 1500 7.0 15.5 15.0 28.1 ....... 44.7 2000 2.8 13.0 20.0 32.5 ....... 29.8 2500 6.7 12.0 18.1 20.4 ....... 12.0 3000 9.2 7.0 5.4 21.2 ....... 20.5 3500 11.4 11.3 11.0 9.7 ....... 31.2 4000 15.1 18.6 25.5 11.1 ....... 21.3 4500 17.0 13.8 28.1 23.3 ....... 22.0 5000 15.2 12.8 26.0 23.9 ....... 12.9 5500 7.2 NA 22.3 22.6 ....... 18.6 6000 22.0 NA 22.9 17.1 ....... 29.6 6500 16.8 NA 28.8 9.4 ....... 38.8 7000 33.3 NA 15.2 23.1 ....... 29.7 7500 29.9 NA 32.0 10.2 ....... 20.5 8000 39.7 NA 4.9 9.7 ....... 12.0 8500 53.8 NA 15.8 25.5 ....... 26.5 9000 NA NA 30.2 25.6 ....... 26.7 9500 NA NA 23.9 25.1 ....... 41.3 10000 NA NA 23.2 26.1 ....... 103.4 86-Kvartal Aero Altay BearBog ....... Zoige 500 6.6 11.5 9.1 23.4 ....... 48.2 1000 5.5 14.8 5.9 32.4 ....... 26.8 1500 7.0 16.0 9.8 27.1 ....... 34.2 2000 4.9 13.4 21.2 29.8 ....... 23.6 2500 13.7 11.5 11.1 21.9 ....... 23.5 3000 9.2 7.0 11.2 15.6 ....... 20.1 3500 11.4 11.3 11.0 9.7 ....... 31.2 4000 15.1 18.6 25.5 11.1 ....... 21.3 4500 17.0 13.8 28.1 23.3 ....... 22.0 5000 15.2 12.8 26.0 23.9 ....... 12.9 5500 7.2 NA 22.3 22.6 ....... 18.6 6000 22.0 NA 22.9 17.1 ....... 29.6 6500 16.8 NA 28.8 9.4 ....... 38.8 7000 33.3 NA 15.2 23.1 ....... 29.7 7500 29.9 NA 32.0 10.2 ....... 20.5 8000 39.7 NA 4.9 9.7 ....... 12.0 8500 53.8 NA 15.8 25.5 ....... 26.5 9000 NA NA 30.2 25.6 ....... 26.7 9500 NA NA 23.9 25.1 ....... 41.3 10000 NA NA 23.2 26.1 ....... 103.4 86-Kvartal Aero Altay BearBog ....... Zoige 500 6.6 11.5 9.1 23.4 ....... 48.2 1000 5.5 14.8 5.9 32.4 ....... 26.8 1500 7.0 16.0 9.8 27.1 ....... 34.2 2000 4.9 13.4 21.2 29.8 ....... 23.6 2500 13.7 11.5 11.1 21.9 ....... 23.5 3000 9.2 7.0 11.2 15.6 ....... 20.1 3500 11.4 11.3 11.0 9.7 ....... 31.2 4000 15.1 18.6 25.5 11.1 ....... 21.3 4500 17.0 13.8 28.1 23.3 ....... 22.0 5000 15.2 12.8 26.0 23.9 ....... 12.9 5500 7.2 NA 22.3 22.6 ....... 18.6 6000 22.0 NA 22.9 17.1 ....... 29.6 6500 16.8 NA 28.8 9.4 ....... 38.8 7000 33.3 NA 15.2 23.1 ....... 29.7 7500 29.9 NA 32.0 10.2 ....... 20.5 8000 39.7 NA 4.9 9.7 ....... 12.0 8500 53.8 NA 15.8 25.5 ....... 26.5 9000 NA NA 30.2 25.6 ....... 26.7 9500 NA NA 23.9 25.1 ....... 41.3 10000 NA NA 23.2 26.1 ....... 103.4 86-Kvartal Aero Altay BearBog ....... Zoige 500 6.6 11.5 9.1 23.4 ....... 48.2 1000 5.5 14.8 5.9 32.4 ....... 26.8 1500 7.0 16.0 9.8 27.1 ....... 34.2 2000 4.9 13.4 21.2 29.8 ....... 23.6 2500 13.7 11.5 11.1 21.9 ....... 23.5 3000 9.2 7.0 11.2 15.6 ....... 20.1 3500 11.4 11.3 11.0 9.7 ....... 31.2 4000 15.1 18.6 25.5 11.1 ....... 21.3 4500 17.0 13.8 28.1 23.3 ....... 22.0 5000 15.2 12.8 26.0 23.9 ....... 12.9 5500 7.2 NA 22.3 22.6 ....... 18.6 6000 22.0 NA 22.9 17.1 ....... 29.6 6500 16.8 NA 28.8 9.4 ....... 38.8 7000 33.3 NA 15.2 23.1 ....... 29.7 7500 29.9 NA 32.0 10.2 ....... 20.5 8000 39.7 NA 4.9 9.7 ....... 12.0 8500 53.8 NA 15.8 25.5 ....... 26.5 9000 NA NA 30.2 25.6 ....... 26.7 9500 NA NA 23.9 25.1 ....... 41.3 10000 NA NA 23.2 26.1 ....... 103.4 iteration 10,000 time bins sites Step 5: Monté-Carlo EOF 0 2000 4000 6000 8000 10000 -5 0 5 age MCEOF1 Standardisation and EOF analysis of each of the 10,000 Monté-Carlo dataset iterations -1.0 -0.5 0.0 0.5 1.0 -1.0 -0.5 0.0 0.5 1.0 Peat C acc. MCEOFs loadings Temp. proxy EOFs loadings Here we present a new approach that combines Baye- sian age modeling and Monte-Carlo EOF to analyze flux-based paleo-datasets by systematically addres- sing both chronological and flux estimate uncertain- ties Aims and scopes: -3 -2 -1 0 1 2 3 4 T emp. proxies (scaled) 80 100 120 140 160 number of sites -10 -5 0 5 10 EOF 1 (32.2% var.) -10 -5 0 5 10 -5 0 5 10 EOF 2 (17.1% var.) -5 0 5 10 -10 -5 0 5 EOF 3 (7.9% var.) -10 -5 0 5 0 1000 2000 3000 4000 5000 6000 7000 8000 9000 year BP Future works: Main results: iteration 1 MCEOF 1 (20.5% var.) MCEOF 2 (15.2% var.) MCEOF 3 (11.9% var.)