 |
| Titel |
Paleotemperature reconstruction from mammalian phosphate δ18O records - an alternative view on data processing |
| VerfasserIn |
Grzegorz Skrzypek, Rohan Sadler, Andrzej Wiśniewski |
| Konferenz |
EGU General Assembly 2017
|
| Medientyp |
Artikel
|
| Sprache |
en
|
| Digitales Dokument |
PDF |
| Erschienen |
In: GRA - Volume 19 (2017) |
| Datensatznummer |
250151747
|
| Publikation (Nr.) |
 EGU/EGU2017-16503.pdf |
|
|
|
|
|
| Zusammenfassung |
| The stable oxygen isotope composition of phosphates (δ18O) extracted from mammalian
bone and teeth material is commonly used as a proxy for paleotemperature. Historically,
several different analytical and statistical procedures for determining air paleotemperatures
from the measured δ18O of phosphates have been applied. This inconsistency in both stable
isotope data processing and the application of statistical procedures has led to large and
unwanted differences between calculated results.
This study presents the uncertainty associated with two of the most commonly used
regression methods: least squares inverted fit and transposed fit. We assessed the performance
of these methods by designing and applying calculation experiments to multiple
real-life data sets, calculating in reverse temperatures, and comparing them with true
recorded values. Our calculations clearly show that the mean absolute errors are always
substantially higher for the inverted fit (a causal model), with the transposed fit (a
predictive model) returning mean values closer to the measured values (Skrzypek et al.
2015).
The predictive models always performed better than causal models, with 12–65% lower
mean absolute errors. Moreover, the least-squares regression (LSM) model is more
appropriate than Reduced Major Axis (RMA) regression for calculating the environmental
water stable oxygen isotope composition from phosphate signatures, as well as for calculating
air temperature from the δ18O value of environmental water. The transposed fit introduces a
lower overall error than the inverted fit for both the δ18O of environmental water and Tair
calculations; therefore, the predictive models are more statistically efficient than the causal
models in this instance. The direct comparison of paleotemperature results from different
laboratories and studies may only be achieved if a single method of calculation is
applied.
Reference
Skrzypek G., Sadler R., Wiśniewski A., 2016. Reassessment of recommendations for
processing mammal phosphate δ18O data for paleotemperature reconstruction.
Palaeogeography, Palaeoclimatology, Palaeoecology 446, 162-167. |
|
|
|
|
|
|