Can Convergent Cross Mapping Untangle Idiosyncratic Speleothem Proxy Records to Reveal the Structure of Shared Climate Forcing?

Abstract:

Rapid growth and development of speleothem paleoclimatology has generated diverse and important new terrestrial paleoenvironmental proxy records that increasingly illuminate both the enormous potential and great complexity of cave proxy systems and speleothem data. Speleothem records commonly exhibit complex covariation patterns between proxy variables (i.e. carbon and oxygen isotopes, various trace element concentrations and ratios, stratigraphic characteristics, growth rates, etc...). Such covariation patterns frequently change sign and magnitude over time, and often show periods without significant correlation that alternate with times with strongly coupled behavior. These patterns are evident when comparing records between sites and stalagmites, and even within a single stalagmite. Instability in covariation patterns and low long-term correlations both limit our confidence in applying speleothems proxy transfer functions over long time periods. Are these complex covariation patterns meaningful or merely mirages? When two speleothem records show the same result, replication is considered by the community to be evidence that both records are highly sensitive to a common climate signal and are thus reliable proxies for that climate signal. Signals derived from a single speleothem dataset could be noise, and thus of limited value until it is validated by the replication test. Are speleothems naturally idiosyncratic and noisy? Must all speleothem records be duplicated to establish reliability?

I consider whether Convergent Cross Mapping (CCM) may offer a fruitful approach to these problems. CCM is a powerful statistical tool developed in George Sugihara’s lab for complex dynamical systems that tests the direction of causality and strength of forcing among multiple time-series variables. I apply CCM to speleothem timeseries records to 1) reconstruct the underlying state climate variable of interest over time (in this case, precipitation), and 2) determine the directions of causality and strength of coupling, both among these proxy records and with respect to precipitation. The efficacy of CCM for unraveling the idiosyncracies of speleothems and their complex covariation patterns, as well as the implications of CCM for speleothem paleoclimatology is discussed.