RBME coral temperature reconstruction: An evaluation, modifications, and recommendations

Testing of the RBME method reveals a number of problems. Firstly, RBME can produce many solutions to a given set of Sr, Mg and Ba data that are nearly identical in their minimization function (i.e., are equally good) but which result in temperatures that are very different. This makes the accuracy of temperature reconstruction by RBME uncertain. Secondly, tests with an artificial Sr, Mg and Ba dataset reveal that even small perturbations, such as analytical noise or vital effects, are enough to destroy the uniqueness of a true 'global' solution to the RBME equations. Such sensitivity means that simplifying assumptions made in the mathematical representation of coral biomineralization cannot be safely ignored. Finally, arbitrary choices in the formulation of the equations for the temperature dependence of Rayleigh Fractionation and minimization function can be critical in determining the outcome of the method. I conclude with a list of recommendations for use of the RBME method and reporting of results so that uncertainties in the reconstructed temperatures are accurately represented.