Holocene DA Validation Report

Validation of the custom Holocene Data Assimilation reconstruction against published Holocene global temperature reconstructions. Reference datasets are discovered at runtime from reference_data/ — any CSV with columns age_BP, median, q05, q95 or age_BP, anomaly, uncertainty_1sigma will be used.

GMST Validation Metrics

Pearson correlation (R) and Nash–Sutcliffe coefficient of efficiency (CE) of the ensemble-median GMST against each reference over their common age range. CE = 1 is perfect; CE = 0 equals climatology; CE < 0 is worse than climatology.

Spatial Anomaly at 6 ka

Mean temperature anomaly for 5500–6500 BP relative to 0–1000 BP baseline. Robinson projection.

Spatial Comparison at 6 ka

Side-by-side 6 ka temperature anomaly for the custom reconstruction and each spatial reference, plus their difference. Pattern correlation (R) and area-weighted RMSE are computed after nearest-neighbor regridding of the reference onto the reconstruction grid.

Proxy-Site Comparison (6 ka)

For each proxy site, the reconstruction’s 6 ka anomaly is sampled via nearest-neighbor on the model grid, then compared to the proxy’s own 6 ka anomaly (5500–6500 BP relative to 0–1000 BP baseline, falling back to the record mean if the modern baseline is absent). R, RMSE, and bias (recon − proxy) summarize the match.

Proxy Database Comparison vs Temp12k 1.0.2 (Erb et al. 2022) (shared 956, custom-only 59, reference-only 320)

Comparison of the proxy records in this run’s lipd_legacy.pkl against Temp12k 1.0.2 (Erb et al. 2022), the version used by the published Erb et al. 2022 reconstruction. Records are matched on paleoData_TSid after the same paleoData_inCompilation == Temp12k + paleoData_units == degC filter that da_load_proxies.py applies at runtime.

1,276

Temp12k 1.0.2 (Erb et al. 2022) records

1,015

Custom records

956

Shared

59

Custom-only

320

Temp12k 1.0.2 (Erb et al. 2022)-only

Side-by-side statistics

Statistic	Custom	Temp12k 1.0.2 (Erb et al. 2022)
Records	1,015	1,276
Distinct archive types	6	8
Earliest record start (yr BP)	1,277,346	1,277,346
Latest record end (yr BP)	-240	-275
Median record length (yr)	11690	11403
Median observations per record	73	71

Records by archive type

Archive	Shared	Custom-only	Temp12k 1.0.2 (Erb et al. 2022)-only	Total
LakeSediment	510	10	184	704
MarineSediment	262	42	95	399
Peat	163	4	28	195
GlacierIce	10	3	5	18
Midden	8	0	2	10
Speleothem	3	0	2	5
Wood	0	0	3	3
Ice-other	0	0	1	1
Total	956	59	320	1335

Spatial distribution

Temporal coverage

Records covering each 250-yr age bin, partitioned by which database they belong to.

Records exclusive to the custom run (59)

Records present in this reconstruction’s proxy database but absent from Temp12k 1.0.2 (Erb et al. 2022) — typically records added in later Temp12k versions or pulled from filtered queries.

TSID	Archive	Dataset	Age range (BP)	n_obs
GH2d963782	MarineSediment	BS79_38.Cacho.2001	867–24,517	71
GH6edd69de	MarineSediment	GeoB33131.Lamy.2002	283–6,990	75
LPD52d54dc4	MarineSediment	GEOFAR_KF16_MgCa.Repschlager.2016	524–11,258	98
LPD5d932ca5	MarineSediment	PP10-07.Mary.2016	422–10,150	206
LPD71287822	MarineSediment	TTR17_434G.Rodrigo-Gamiz.2014	27–13,728	92
LPD73260700	MarineSediment	PP10-07.Mary.2016	422–10,150	206
LPD74bc6584	LakeSediment	LittlePondRoyalston.Marsicek.2013	0–11,000	221
LPD779b21dc	LakeSediment	BloodPond.Marsicek.2013	0–11,000	221
LPD7abab671-dup	GlacierIce	LawDome.Dahl-Jensen.1999	-45–8,692	35
LPD92d845a8	MarineSediment	PP10-07.Mary.2016	422–10,150	206
LPD9c79485dxxxx	MarineSediment	ODP658C_UK37.Zhao.1995	137–22,589	203
LPDa0c55c63	MarineSediment	PP10-07.Mary.2016	422–10,150	206
LPDadaae75f	MarineSediment	SN-6.Tiwari.2015	154–4,772	36
LPDc82d344d	MarineSediment	GEOFAR_KF16_MgCa.Repschlager.2016	524–11,258	98
PYT917S9C8J	MarineSediment	GIK17748-2.Kim.2002	1,080–16,070	54
PYTDSG9BNO2	MarineSediment	GIK23258.Sarnthein.2003	698–13,987	285
PYTH91MXU0Z	MarineSediment	P1003.Sejrup.2011	-48–7,881	649
PYTRS33LNIK	MarineSediment	GIK23258.Sarnthein.2003	698–13,987	285
PYTUQPUNFJT	Peat	Corser.Nichols.2014	103–11,116	36
R60yDfufGWZ	Peat	StaroselskyMoch.Novenko.2018	-19–9,792	56
R7erF84XRXy	MarineSediment	DANA12_11_2_GC01.VanNieuwenhove.2018	-54–11,810	80
RCP5UgOcOkR	LakeSediment	Gunnarsfjorden.Allen.2007	-40–9,124	104
RGIi8rzDMmj	MarineSediment	GeoB10042_1.Setiawan.2015	4,200–39,420	106
RI2vCrHM2fw	MarineSediment	M35003_4.R_hlemann.1999	0–24,540	91
RJ64ej30dlD	MarineSediment	MD03_2607-Assemblage.LopesdosSantos.2013	1,100–134,800	172
RKz0xLW2LFh	MarineSediment	MD06_3067.Bolliet.2011	3,390–158,200	484
RMpGVpFWptO	LakeSediment	Duranunlak.EPD	368–11,501	36
RNAQJ20Sh7j	MarineSediment	RR1313_23PC.MoffaSanchez.2019	1,280–16,186	120
RPZj5YKrFr0	LakeSediment	PupukePollen.VandenBos.2018	1,001–16,138	62
RQ6N1Ktr5Ai	LakeSediment	Duranunlak.EPD	368–11,501	36

⬇ Download full CSV (59 rows)

Records exclusive to Temp12k 1.0.2 (Erb et al. 2022) (320)

Records used by the published reconstruction but missing from this custom run — typically records dropped by the user’s filter or removed from later Temp12k versions.

TSID	Archive	Dataset	Age range (BP)	n_obs
GH06e59526	MarineSediment	MD79-257.Levi.2007	1,554–15,695	47
GH194ddf83	Peat	HomesteadScarp.McGlone.2010	-39–12,247	170
GH73726cce	Peat	MountHoney.McGlone.2010	-18–12,563	109
LPD27f34021	Peat	Yakumo.Leipe.2013	40–5,506	38
LPD3267fe5d	MarineSediment	PP10-07.Mary.2016	422–10,150	206
LPD575ea390	LakeSediment	SihailongwanMaar.Stebich.2015	150–11,992	206
LPD5c8693b4	MarineSediment	MD01-2416.Sarnthien.2004	5,727–11,920	41
LPD5c89a1b5	LakeSediment	HAP0212.vanderBilt.2016	534–11,117	60
LPD6e0eacd1	LakeSediment	LagoDellAccesa.Finsinger.2010	4,479–11,552	86
LPD7abab671	GlacierIce	LawDome.Dahl-Jensen.1999	-45–8,692	35
LPD801dea7f	MarineSediment	MD01-2446.Salguero.2014	2,290–20,012	55
LPD8d84bf64	Wood	GreatBasin.Salzer.2014	-56–4,524	4,581
LPD91355717	LakeSediment	LagoVerdarolo.Samartin.2011	-59–13,720	50
LPDbf7cf949	Peat	SouthernCalifornia.Ohlwein.2012	408–11,446	28
LPDbfac52bb	LakeSediment	Qinghai.Hou.2016	0–15,629	184
LPDd2a984fe	LakeSediment	AMP112.vanderBilt.2016	1,144–12,353	66
LPDf97892fc	LakeSediment	XingyunLake.Wu.2018	-27–14,192	67
M2L1fcqp9rpP	LakeSediment	EC1.Gajewski.1993	58–6,610	32
M2L1r9snju2P	LakeSediment	Atocas.Gauthier.1981	0–11,977	41
M2L4ptir6l2P	LakeSediment	BanksIsland-12.Gajewski.2000	175–10,183	33
M2L68fmqq0lP	LakeSediment	Atocas.Gauthier.1981	0–11,977	41
M2L68iw0l2eP	LakeSediment	NorthPond.Whitehead.1978	-105–13,083	55
M2L6qbln98nP	LakeSediment	LoneFoxLake.MacDonald.1985	-32–12,638	37
M2L726ed2xjP	LakeSediment	CH2.Garralla.1992	1–8,332	89
M2L7thzxje2P	LakeSediment	Hams.Bennett.1987	-29–11,945	82
M2L8sc7gn3mP	LakeSediment	Hams.Bennett.1987	-29–11,945	82
M2LdjjxzkgtP	LakeSediment	Queens.Moser.1990	0–8,840	33
M2Lfut6kmdqP	LakeSediment	Queens.Moser.1990	0–8,840	33
M2Lg77nq75zP	LakeSediment	EC1.Gajewski.1993	58–6,610	32
M2Lh4fhttxhP	LakeSediment	CH2.Garralla.1992	1–8,332	89

⬇ Download full CSV (320 rows)

GMST Time Series

Custom reconstruction ensemble spread alongside reference medians. X-axis runs from the oldest age (left) to present (right).

GMST Ensemble Members (52 total)

Every ensemble member plotted individually (subsampled to 200 for readability), showing the full reconstruction spread.

GMST Difference (Custom − erb2022_gmst)

Year-by-year difference between the custom reconstruction median and the primary reference. Red = warmer, Blue = cooler.

← Back to results