Introduction to proxy system modeling

Kevin AnchukaitisUniversity of Arizona

Suz Tolwinski-WardAIR Worldwide

Diane ThompsonBoston University

Michael EvansUniversity of Maryland

‘The best material model of a cat is another, or

preferably the same, cat.’Arturo Rosenblueth and Norbert Wiener

Under- vs. Over-specified Models

after Dean Urban, Duke University

Under- vs. Over-specified Models

after Dean Urban, Duke University

Trying to get here

‘A good forward model requires not only a good understanding of the key processes that are responsible for generating the signal (proxy

climate record) in the archive, but it also needs appropriate sources of estimates for the parameters and a reasonable understanding of

various levels of uncertainties related to the raw proxy data.’

Hughes and Ammann 2009

ENVIRONMENT SENSOR ARCHIVE OBSERVATION

SENSOR MODEL ARCHIVE MODEL OBSERVATION MODEL

RECONSTRUCTION

CALIBRATION

Evans, Tolwinski-Ward, Thompson, Anchukaitis, 2013based on SUPRAnet (UK) Working Group

ENVIRONMENT SENSOR ARCHIVE OBSERVATION

SENSOR MODEL ARCHIVE MODEL OBSERVATION MODEL

RECONSTRUCTION

CALIBRATION

TEMPERATUREPRECIPITATIONSOIL MOISTURE

SUNLIGHTSNOW WOOD

TREE

RING WIDTHMXD

EARLYWOODLATEWOOD

ISOTOPES

Evans, Tolwinski-W

ard, Thompson, Anchukaitis, 2013

�18Ocoral

= f(T, S)

Application: Observational network design

Comboul et al (2015), using Thompson et al (2011); Other examples: Bradley (1996), Evans et al (1998), Evans (2007)

�18Ocoral

= f(T, S)

Application: model/data comparison

Thompson et al, 2011, 2012, 2013 after Brown et al (2008) and Meehl et al (2007)

Application: model/data comparisonTRW = f(T, P,�)

Anchukaitis et al. 2012

Application: model/data comparisonTRW = f(T, P,�)

10−3

10−2

10−1

10−1

100

FREQUENCY (cycles/year)

NO

RM

ALIZ

ED

PO

WE

R

PSEUDOPROXYMODEL PDSI

Anchukaitis et al. in preparation

Application: paleoclimate inference & interpretation

Schmidt 1999

�18Oforaminifera

= f(T, S, �18Osw

)

Application: paleoclimate inference & interpretationTRW = f(T, P,�)

1955 1960 1965 1970 1975 1980−3

−2

−1

0

1

2

YEAR

WHITE MOUNTAINS (WM1), METHUSELAH WALK

A

r = 0.75, p = 0.00

SIMULATEDREAL

100 200 3000

0.2

0.4

0.6

0.8

1

DAY OF YEAR

GROWTH RATES

C

GrGrWGrT

100 200 3000

0.5

1

1.5

2

2.5

3

DAY OF YEAR

ENVIRONMENT

BSNOW/1000(mm)

TRANS(mm/d)SM(v/v)

Anchukaitis, Evans, Hughes

Application: paleoclimate inference & interpretation�18O

cellulose

= f(T, �18Orain

, �18Ovapor

,%RH)

60oE 80

oE 100oE 120

oE 140o

E 160oE

18oS

0o

18oN

36oN

54oN

r(δ18Ocellulose

,NINO3.4 SSTA

−1 0 1

Anchukaitis and LeGrande

Tolwinski-Ward et al. 2014

X

TRW = f(T, P,�)

Application: paleoclimate reconstruction constraint

Future work: more proxy systems models!

Proxy systems may be multivariate, nonlinear, spectral climate filters.

Use of “intermediate-class” proxy system models permits:

Paleo-observational network design/hypothesis testing.

Subtle interpretation of the environmental response.

Robust intercomparison of paleodata and climate model output.

Scientifically-defensible weak constraints for paleoreconstructions.

email: mnevans@geol.umd.edu