Readme documentation for LMR_data_files Feb. 4, 2108

Horlick, Kaleb (Oregon State University College of Earth, Ocean, and Atmospheric Sciences, Corvallis, OR, kaleb.horlick@gmail.com, and David M. Anderson (unaffiliated, davidmanderson2@gmail.com)

Introduction This document describes the files available in this Github repository. The files were prepared between December 2014 through June 2016 for use by the Last Millennium Reconstruction Project. As described below, the data have been obtained from publically-accessible archives and can therefore be freely shared. The original references in each metadata record should be cited if a record is used. The primary purpose of this repository is to provide a staging of structured files that can be accessed using code developed by the last millennium project. These files have not been quality controlled. Quality control, including the identification of duplicate records and errors in data and metadata, occurs in the project workflow when this archive is read by the project code. This archive contains the raw, unprocessed data records.

Data Sources Data was primarily taken from the NOAA NCEI’s World Data Center for Paleoclimatology archive in its most raw form, quality controlled, and formatted into NOAA’s machine-readable ASCII text file format. But, through collaboration with the PAst Global changES team and Joerg Franke, data was incorporated from other data collection and standardization efforts such as PAGES2k (Oceans2kHR/LR), and Petra Breitenmoser’s 2014 ITRDB re-standardization. In the final months of the effort, data was solicited from multiple investigators such as Eric Steig, Stephanie Hayman, Sylke Draschba, Henning Kuhnert, and others amounting to approximately 30 new coral/ice core datasets. The files in this archive can be freely shared however

Data Selection Datasets were selected whose resolution was at least 25 years, temporal duration was greater than 40 years, and consisted of proxies that had established proxy system models (forward models). As compared to previous collection efforts, LMR records were not screened for a particular sensitivity to temperature variability, as the LMR will reconstruct a multitude of climate fields including –but not limited to- precipitation. Three paleoclimate proxy archives were targeted: corals, ice cores, and tree rings. Proxies associated with each that were targeted include: d18O and Sr/Ca (corals), d18O, dD, and accumulation rate (ice cores), maximum early/latewood density and tree ring width (tree rings). Although these were specifically collected, other proxies were collected (i.e. coral d13C, d14C, Mg/Ca, etc) when available in case of future development, even if not for present assimilation. Approximately 30 climate reconstructions were collected for independent verification of LMR results. Following the LMR’s Proxy System Modeling workshop in May of 2015, through collaboration with Andy Baker of the University of New South Wales, an expert selection of approximately 30 speleothem d18O records were added given the archive’s recently established preliminary proxy system model. A nominal number of arctic lake core varve thickness records (3) were targeted as well given their defensible linear relationship to summer temperature.

Database Contents by Archive (# of time series) This combination of the datasets quality controlled and collated from the WDC, along with datasets incorporated via international collaboration with investigators and data collection efforts, amounts to an inclusive multi-proxy database of almost 3500 time series. Some datasets hold multiple time series (i.e. coral d18O and Sr/Ca time series from the same core sample archived in two columns of the same datafile), and thus are counted multiple times according to the respective time series contained. Also approximately 223 of these time series are duplicates between PAGES2k and LMR efforts, but were archived nonetheless. These duplications are identified in the “Notes:” field of the duplicated PAGES version of the datasets. “Other” also includes datasets archived as the climate field they were used to reconstruct, i.e. most PAGES marine core records were archived as SST time series, rather than their raw proxy time series.

File Structure and Usage The files LMR database v2.0 are consistently formatted based on the WDC-P’s established ASCII text file format in order that they may be machine read. Metadata rows are denoted with the “#” symbol as each line’s beginning character. Rows of data have no such lead character. The time column header “age” is the ubiquitous denotation for the time column of the series. The “Time_Unit” field specifies the units that each age represents, i.e. “y_ad” for years AD, “yb_#” for years before a specified time. Age model, or “chronology”, metadata is sometimes tabulated when necessary as in the case of speleothem datasets. Each data column is tab delimited from the other columns, and “NAN” is generally the missing data value. The “Notes” field contains useful metadata regarding things like standardization/normalization details, data on how multi-core stacks were assembled, as well as metadata stored in JSON key:value format, such as: seasonality, climate sensitivity/dependence, filenames of matching files when duplicates exist, and the database the file is included in. All files, including datsets sourced from the PAGES2k database, are tagged with “{database”:”LMR”}, except for those files from PAGES2k that duplicate files already archived from NOAA WDC-P archives. These are tagged with {“duplicate”:….} tags denoting the corresponding duplicated file. All files sourced from PAGES2k contain a second value for the “database” key. See Supplemental File 1 for example data file.

Archives Corals A total of 121 unique coral d18O records, 65 unique Sr/Ca records and 90 other unique records are archived. Coral d18O and Sr/Ca are well established tropical temperature proxies, although d18O is also dependent upon d18O of ambient seawater and thus used as a sea surface salinity proxy. Although Sr/Ca does not have a formal physical PSM, it is widely accepted to vary linearly with temperature, thus the use of a univariate linear forward model will allow for their use in the LMR. Coral “other” proxies include d13C, d14C, extension rate, growth rate, Mg/Ca, Ba/Ca, U/Ca, Pb, Sr, d11B, pH, luminescence, and calcification rate. Although the climatological controls on these proxies are much more poorly established, or non-existent- they were archived simply in the case of the establishment of such relationships in the future. Ice Cores A total of 143 unique ice core d18O records, 22 unique dD records, and 10 “other” records were archived. Ice core d18O and dD are well established arctic temperature proxies. Although all of the influences are still not fully understood, our understanding of the physics behind the fractionation processes affecting the d18O of glacier and ice sheet ice have allowed for an established PSM for ice core d18O records. Similar to coral Sr/Ca, ice core dD is generally accepted as a temperature proxy, thus temperature can linearly be mapped to dD space and used in an assimilation. “Other” ice core proxy records include melt feature percentage and accumulation, which was collected with the idea that accumulation rate is needed for incorporation into the d18O PSM.

Tree Rings A total of 2285 restandardized tree ring width files were incorporated through collaboration with Joerg Franke, University of Bern, Switerland. The restsandardized ITRDB tree ring index files from Petra Breitenmoser’s 2014 publication, “Forward modelling of tree-ring width and comparison with a global network of tree-ring chronologies”, were machine-written into NOAA WDC-P’s text file format. Tree ring width has a well established PSM, VS-Lite, which allows for the incorporation and assimilation of thousands of tree ring records into the LMR. One strength of this collection is that all of the tree ring files are standardized in a consistent way: a linear detrending for almost all of the files. For those files whom ARSTAN would not linearly detrend (due to the trend slope approaching 0), a smoothing spline approach was applied, as outlined in Breitenmoser 2014. Another strength of this collection is that it is accompanied by unique metadata such as idealized VS-Lite parameters (i.e. M1/M2 and T1/T2 soil moisture and temperature thresholds), as well as VS-Lite-based determinations of each chronology’s sensitivity to climate (precipitation/temperature). General seasonalities were included in the metadata according to a general rule: tropical trees were denoted as year-round records, NH trees Boreal summer records, and SH trees Austral summer records. The use of the VS-Lite PSM in the LMR necessitated that only the records that had these corresponding VS-Lite metadata were collected for LMR purposes. Thus, only 2285 out of a total of approximately 2750 (~85%) from the Breitenmoser 2014 collection were included. In addition to the Breitenmoser 2014 collection, the PAGES2k tree ring database was included. A total of 348 unique tree ring width files were included by this source. Also, approximately 62 maximum late/earlywood, and other wood density records were included from this source. Most tree ring density proxies (I.e. maximum latewood density) are accepted to vary directly with temperature, thus can be modeled with a linear regression to temperature as in the case of lake varves, coral Sr/Ca, and ice core dD. “Other” tree ring records include d18O of cellulose, which is still poorly understood with respect to its climate dependence, but was archived anyways. Approximately 130 of the PAGES tree ring witdth records were archived dispite their duplication of records from the Breitenmoser 2014 collection. Metadata denoting the duplicated files is tabulated under the “Notes:” field in the text format.

Speleothems Speleothem records were not originally targeted by the LMR collection effort. After the LMR’s Proxy System Modeling Workshop in May of 2015, it was decided that that the speleothem d18O PSM was mature enough to justify the collection of at least a few of the most trusted published speleothem records. Thus, in collaboration with speleothem expert Andy Baker, University of New South Wales, Australia, 26 records were collected for LMR use. They are mostly speleothem d18O records as that is the only speleothem proxy with an established PSM. Some “other” speleothem records include 4 d13C time series that were included as a consequence of simply being archived as additional time series alongside d18O series. Four PAGES2k speleothem records are included as well. Lake/Marine Cores Virtually all of the lake and marine core records were included through collaboration with the phase two of the PAGES2k paleoclimate data collection effort (including Oceans2k HR/LR). These, thus, are all temperature sensitive records, and are mostly archived as SST or temperature reconstructions rather than the raw form of the proxy datasets (like foram d18O, chronomids, varve thickness, etc). This collection nevertheless totals 61 marine core records and 36 lake core records. These were also included because of emerging PSMs that could, in the future, be defensible enough for the inclusion of these datasets in future assimilation efforts (Kaustaub, foram d18O). Also lake varve sediment records are so widely accepted to covary closely with summertime temperature that they can be linearly fit with temperature (Overpeck) in an assimilation, much like coral Sr/Ca and ice core dD.

Others Other records include 4 borehole and 16 historical document records collected from the PAGES2k collection. These archives do not have viable PSM’s, but their usage could possibly, on a record-by-record basis, be justified by fitting linearly with temperature. Also, 28 climate reconstructions were collected for independent validation of LMR results

References: Breitenmoser, P., Brönnimann, S. and Frank, D. (2014) ‘Forward modelling of tree-ring width and comparison with a global network of tree-ring chronologies’, Clim. Past. Copernicus Publications, 10(2), pp. 437–449. doi: 10.5194/cp-10-437-2014.

Emil-Geay, J., et al. Last Millennium Reanalysis. https://www.researchgate.net/project/Last-Millennium-Reanalysis. Accessed March 3, 2018. https://www.researchgate.net/project/Last-Millennium-Reanalysis

Hakim, Gregory & Emile-Geay, Julien & Steig, Eric & Noone, David & Anderson, David & Tardif, Robert & Steiger, Nathan & Perkins, Walter. (2016). The last millennium climate reanalysis project: Framework and first results: LAST MILLENNIUM REANALYSIS. Journal of Geophysical Research: Atmospheres. 121. 10.1002/2016JD024751.