What do glacial moraine chronologies really tell us What do glacial moraine chronologies really tell us about climate?about climate?

Martin P. Kirkbride Martin P. Kirkbride

GeographyGeography School of the EnvironmentSchool of the Environment

University of DundeeUniversity of DundeeDundee Dundee DD1 4HNDD1 4HN

Scotland UK.Scotland UK.

European Geophysical Union General Assembly 2011European Geophysical Union General Assembly 2011

Session GM9.1/9.2 Session GM9.1/9.2

1. Introduction: moraine dating, and the principles of correlation1. Introduction: moraine dating, and the principles of correlation

Principles of correlation

physical similarity (classic stratigraphic correlation)

texture, structure, provenance, fossil assemblage

spatial continuity

tephra layers, Heinrich layers etc.

statistical matching

measures of central tendency and dispersion

Moraine are physically similar and lack spatial continuity, so we depend on statistical properties of proxy age samples

Five steps to linking climate records and moraines

Climate variations (various amplitudes, frequencies, and phenomena)

Glacier responses(variable reaction and response times re. glacier size and form)

Moraine deposition(variable debris supply)

Erosion censoring

Moraine dating

Correlation

(local, regional. interhemispheric scales, “lumping/splitting” decisions)

Climate variations (various amplitudes, frequencies, and phenomena)

Glacier responses(variable reaction and response times re. glacier size and form)

Moraine deposition(variable debris supply)

Erosion censoring

Moraine dating

Correlation

(local, regional. interhemispheric scales, “lumping/splitting” decisions)

Five steps to linking climate records and moraines

Timescales of glacier reaction and response

Cirque glacier = annual variationPotentially 6 moraines

Small valley glacier = decadal variationPotentially 3 moraines

Compound valley glacier = century scale variation0 moraines

Different glaciers respond to different frequencies of climatic fluctuation

Larger glaciers filter and smooth complex high-frequency fluctuations

(Haeberli and Hoelzle)

2. Relations between spatial correlation and temporal resolution2. Relations between spatial correlation and temporal resolution

How do different ranges of uncertainty in dating affect out ability to correlate distant moraine sequences?

Mt Cook, NZ: Schaefer et al (2009) Science

Uncertainty is c. 20% of individual 10Be age (ie. ± 10%)

Chronological resolution (uncertainty) and correlation

Our ability to correlate depends on the range of uncertainty (“error bars”) of the dated moraines.

“Error” Correlation Interpretation

± 500 M1 = M2 M3 1 glacier advance “event”

± 250 M1 = M2 M3 2 glacier advance “events”

± 75 M1 M2 M3 3 glacier advance “events”

“Lumping” is the tendency to correlate across wide areas because the temporal resolution of individual chronologies is too coarse to distinguish separate events

“Splitting” is the tendency to distinguish between closely-spaced events which are dated to high resolution.

The “Lumping/Splitting” Rule:

The spatial resolution of correlation is in inverse relation to the temporal resolution of the component chronologies.

The “Lumping/Splitting” Rule:

The spatial resolution of correlation is in inverse relation to the temporal resolution of the component chronologies.

“Lumped” chronologies may be appropriate for interhemispheric correlation where millenial-scale glacial fluctuations (eg LIA) are due to global climate changes.

“Split” chronologies may be appropriate for local to regional correlation where decadal scale glacial fluctuations (eg LIA maximum) are due to regional atmospheric circulation changes.

Response scale, chronological resolution, and correlation error

Type I error

False rejection of null hypothesis

False correlation of different climatic events.

Null hypothesis:

“there is no temporal or climatic association between moraines in different places”

Type II error

False acceptance of null hypothesis

False differentiation of the same climatic event.

Kaplan et al. (2010) NatureKaplan et al (2010) Nature 467

Response scale, chronological resolution, and correlation error

Ages spread over 2,500 years produce a statistical uncertainty of only 100 years (weighted mean).

Which is the most realistic estimate of uncertainty?

This affects our ability to correlate with confidence and risks errors of false differentiation.

Risk of a Type 2 error = false acceptance of null hypothesis: more likely if chronologies are unrealistically resolved.

Kirkbride & Brazier (1998) Quaternary Proceedings.

Age structure of moraine sequences

Age-difference curves: Graphs of moraine age against age-difference from next youngest moraine.

An increase in age difference with age is common – but is it due to:

Shift in recorded climate signal?

Loss of evidence?

“Correlation shift”?

Ag

e d

iffe

ren

ce

Age

3. Completeness of the moraine record: erosion censoring3. Completeness of the moraine record: erosion censoring

How do we know how complete our moraine records are?

Kirkbride & Brazier (1998 ) Quaternary Proceedings 6

Climate varies with frequencies of 2000, 200 and 20 years.

Age-difference line shows both the climate frequency recorded by the moraine field, and the degree of erosion censoring.

Erosion censoring of moraine assemblages

Mt Cook LIA

y = 0.1051x + 107.02R2 = 0.8143

0

50

100

150

200

0 100 200 300 400 500 600

Moraine age (Cal yr BP)

Ag

e d

iffe

ren

ce (

yr)

Blabreen LIA

y = 0.069x + 13.494R2 = 0.103

0

50

100

150

200

0 50 100 150 200 250

Moraine age (Cal yr BP)

Ag

e d

iffe

ren

ce (

yr)

Erosion censoring of moraine assemblagesLittle Ice Age chronologies

Mt Cook: Schaefer et al (2009)

Blåbreen: Bickerton & Matthews (1992)

If the age differences between moraines increase with age, older deposits may have been censored by erosion or aggradation.

Mt Cook: possible erosion censoring

Blåbreen: no erosion censoring

Smorstabbtinden 14C chronology (Matthews & Dresser 2008)

y = 401.68e0.0003x

R2 = 0.8528

0

500

1000

1500

2000

2500

3000

3500

4000

0 1000 2000 3000 4000 5000 6000 7000 8000

Moraine age (Cal yr BP)

Ag

e d

iffe

ren

ce (

yr)

Mount Cook 10Be chronology (Schaefer et al. 2009)

y = 7E-05x2 + 0.0092x + 129.08R2 = 0.9707

0

500

1000

1500

2000

2500

3000

3500

0 1000 2000 3000 4000 5000 6000 7000

Moraine age (Cal yr BP)

Ag

e d

iffe

ren

ce (

yr)

.

Erosion censoring of moraine assemblagesHolocene chronologies

Dec

reas

ing

qual

ity o

f pro

xy e

vide

nce

Iceland

(Kirkbride & Dugmore 2008 Quatern. Res.)

4. Conclusions and recommendations.4. Conclusions and recommendations.

1. We require a theoretical and practical basis for correlation. We have no systematic criteria at the moment.

2. Correlation decisions need to be informed by chronological resolution, timescale of glacier response, and presumed frequency of climatic variability.

3. We need better integration of modelling of glacier sensitivity into chronological investigations.

4. We need some way of assessing the completeness of moraine records as part of correlation attempts.

Thank you