Elk–testing climate–carbon cycle models: a case for pattern–oriented system analysis

Markus Reichstein, Miguel D. Mahecha, Philippe Ciais, Sonia I. Seneviratne, Eleanor Blyth, Nuno Carvalhais, Yiqi LuoNERC 2011

Abstract

Process–oriented models are a primary
\ntool being used to project future states of
\nclimate and ecosystems in the Earth system
\nin response to anthropogenic and other
\nforcing. Coupled climate–carbon cycle models
\nreceive tremendous attention, especially
\nin the context of the 5th assessment report
\nof the IPCC (International Panel on Climate
\nChange). However, intercomparison of
\nmodel scenarios indicate large uncertainties
\nregarding predictions of global interactions
\nbetween atmosphere and biosphere.
\nRigorous scientific testing of these
\nmodels is essential but very challenging,
\nlargely because it is neither technically nor
\nethically possible to perform global earth–
\nscale experiments—we do not have replicate
\nEarths for hypothesis testing. Hence,
\nmodel evaluations have to rely on monitoring
\ndata such as ecological observation
\nnetworks, global remote sensing, paleo
\nproxy data, or small–scale manipulative
\nexperiments.
\nHere, we critically examine strategies of
\nhow model evaluations should be performed. We put a particular emphasis on
\nthe representation of terrestrial ecosystems,
\nwhere the two key problems are:
\n1. weak (or inconclusive) ‘validations’ which
\ndo not take advantage of all the relevant
\ninformation in the observed data, and
\n2. apparent falsifications: “false alarms”
\nlikely to occur when individual system
\nprocesses (in the model) are compared
\nto the overall emergent system behaviour
\n(of the observed world).
BibTeX
@inproceedings{reichstein2011elktesting,
  author = {Markus Reichstein and Miguel D. Mahecha and Philippe Ciais and Sonia I. Seneviratne and Eleanor Blyth and Nuno Carvalhais and Yiqi Luo},
  title = {{Elk–testing climate–carbon cycle models: a case for pattern–oriented system analysis}},
  booktitle = {NERC},
  year = {2011},
}

Loading related content...

Figure