Glacial Erosion Rates from Bayesian Inversion of Cosmogenic Nuclide Concentrations in a Bedrock Core, Streaked Mtn., ME

Friday, 19 December 2014
Zachary Thomas Ploskey and John O Stone, University of Washington Seattle Campus, Seattle, WA, United States
Glacial erosion is an important source of sediment and could be an important coupling to glacier and ice sheet models that track sediment. However, glacial erosion is difficult to quantify, and models of glacial erosion can benefit from independent erosion rate estimates. Here we present the results of a Bayesian Markov chain Monte Carlo (MCMC) inversion of a cosmogenic nuclide (CN) geomorphic model for glacial erosion rates on a bedrock landform formerly eroded beneath the Laurentide ice sheet. The CN 10Be was measured in quartz to 8 m depth in a bedrock core from the summit of Streaked Mountain, ME. The accumulation of 10Be was modeled over multiple glacial cycles of alternating exposure and glacial erosion. This model was invertedfor glacial erosion rates and burial history using MCMC algorithms implemented in PyMC (Patil et al., 2010). This Bayesian approach allows us to incorporate prior constraints on ice cover history, including oxygen isotope records and radiometric dates, which is otherwise difficult to differentiate from erosion in rapidly eroding areas. We compare these results to depth profile and surface CN measurements elsewhere in Maine (Ploskey and Stone, 2013).The forward model of CN production used in the inversion is part of Cosmogenic (github.com/cosmolab/cosmogenic), an open-source Python-based software library we developed for modeling the growth and decay of in-situ CN inventories in rock during geomorphic evolution. It includes calibrated production rates for 10Be and 26Al in quartz and 36Cl in K-feldspar by both neutrons and muons, with more isotopic production pathways and material targets to be added in the future. Production rates are scaled to the site altitude and latitude using modular scaling schemes. Cosmogenic includes a variety of functions representing common geomorphic histories, and can be used to model any arbitrary exposure, erosion and burial history that can be defined as Python function.References

Patil, A., D. Huard and C.J. Fonnesbeck (2010), PyMC: Bayesian Stochastic Modelling in Python. J. Statistical Software, 35(4), 1-81.

Ploskey, Z.T and J.O. Stone (2013), Measuring glacial erosion of bedrock landforms with cosmogenic nuclide depth profiles, Abstract #EP41A-0794 presented at 2013 Fall Meeting, AGU, San Francisco, Calif., 9-13 Dec.