Comment: Hierarchical statistical modeling for paleoclimate reconstruction
The article by Bo Li, Douglas W. Nychka, and Caspar M. Ammann (hereafter, LNA) has several goals. It considers the important problem of reconstruction of past (over a period of more than 1000 Years Before Present) climate from multiproxy data, and it directly recognizes the various uncertainties in this undertaking. These uncertainties are expressed through (conditional) probability distributions in a framework known to readers of this journal as hierarchical statistical modeling. LNA use a physical–statistical model that also includes climate forcings, and their statistical inference is Bayesian. Rather than using actual multiproxy data, LNA simulate their data. Then they design a computer-simulation experiment to assess the value of including the various (simulated) proxies and the forcings. The design of the experiment, its analysis, and the conclusions obtained from it, are intended to guide climate scientists towards more precise inferences when carrying out actual paleoclimate reconstructions. Our discussion of LNA in the sections that follow considers both the scientific and statistical goals summarized above.