Final Report Summary - ITOP (Integrated Theory and Observations of the Pleistocene)
The ERC-funded project Integrating Observations over the Pleistocene (ITOP) had a double objective : improve our knowledge of past climate dynamics, and develop the methodological background to articulate theory and observations at time scales of several hundreds of thousands of years.
Concretely, the project was organised around three main packages:
(a) Dynamical system analysis
We analysed a series of conceptual models of ice ages by reference to the concepts of 'synchronisation' and 'attractors'. We discovered that many of these models feature a type of attractor known as "strange non-chaotic attractor" and, as such, they display a behaviour that is mid-way between stability and chaos. Specifically, such systems are episodically sensitive to stochastic perturbations. This opens the prospect that there exists critical times in the climate history during which the long term evolution of the climate system becomes unpredictable. This critical behaviour would have developped after the 'Mid-Pleistocene Revolution', about 1 million years ago.
(b) Global sensitivity analysis of general circulation models
We used a statistical methodology based on a type of model called 'emulator'. This approach allows us to study in depth the sensitivity of climate models on the basis of a scheme of interpolation between a set of well-chosen experiments. We successfully developed emulators of two climate models: LOVECLIM, and HadCM3. In two already-available publications we develop the methodology, show scientific results associated with the response to world-wide temperature and tropical monsoon dynamics. Crucially, we are able to predict the phase of the response of climate record with respect to the precession signal, and also predict the timing of accelerated episodes of the climate change to the smooth astronomical forcing. This generalises earlier work on the Sahara desertification. Now that the methodology is validated, another series of articles is in preparation, which will document the response phase of ocean temperature and circulation to the astronomical, obliquity and land-ice forcings. These will constitute important model tests with respect to the available palaeoclimate observations.
(c) Bayesian calibration :
We have established that benthic formanifera do actually contain enough information to partially discriminate between simple dynamical system models, though results remain too sensitive to the age model applied to the data. Monte Carlo methodology and computer power are now sufficiently advanced that we can tackle the joint reconstruction, age model, and model selection problems in a fully Bayesian manner, but it remains computationnally expensive. This work is detailed in publications in preparation.
Selected references
M. Crucifix, Traditional and novel approaches to palaeoclimate modelling, Quaternary Science Reviews, 57, 1--16 2012
M. Crucifix, Why could ice ages be unpredictable?, Climate of the Past, 9, 2253--2267 2013
T. Mitsui, M. Crucifix, K. Aihara, Bifurcations and strange nonchaotic attractors in a phase oscillator model of glacial-interglacial cycles, in revision in Physica D
P. A. Araya-Melo, M. Crucifix, and N. Bounceur, Global sensitivity analysis of Indian Monsoon during the Pleistocene, Climate of the Past Discussions, 10, 1609--1651 2014 (in press in Climate of the Past)
Software pakcages :
Package palinsol : http://cran.r-project.org/package=palinsol
Package GP : Gaussian process for emulation of GCMs : https://bitbucket.org/mcrucifix/gp
Package iceages : https://bitbucket.org/mcrucifix/iceages
Concretely, the project was organised around three main packages:
(a) Dynamical system analysis
We analysed a series of conceptual models of ice ages by reference to the concepts of 'synchronisation' and 'attractors'. We discovered that many of these models feature a type of attractor known as "strange non-chaotic attractor" and, as such, they display a behaviour that is mid-way between stability and chaos. Specifically, such systems are episodically sensitive to stochastic perturbations. This opens the prospect that there exists critical times in the climate history during which the long term evolution of the climate system becomes unpredictable. This critical behaviour would have developped after the 'Mid-Pleistocene Revolution', about 1 million years ago.
(b) Global sensitivity analysis of general circulation models
We used a statistical methodology based on a type of model called 'emulator'. This approach allows us to study in depth the sensitivity of climate models on the basis of a scheme of interpolation between a set of well-chosen experiments. We successfully developed emulators of two climate models: LOVECLIM, and HadCM3. In two already-available publications we develop the methodology, show scientific results associated with the response to world-wide temperature and tropical monsoon dynamics. Crucially, we are able to predict the phase of the response of climate record with respect to the precession signal, and also predict the timing of accelerated episodes of the climate change to the smooth astronomical forcing. This generalises earlier work on the Sahara desertification. Now that the methodology is validated, another series of articles is in preparation, which will document the response phase of ocean temperature and circulation to the astronomical, obliquity and land-ice forcings. These will constitute important model tests with respect to the available palaeoclimate observations.
(c) Bayesian calibration :
We have established that benthic formanifera do actually contain enough information to partially discriminate between simple dynamical system models, though results remain too sensitive to the age model applied to the data. Monte Carlo methodology and computer power are now sufficiently advanced that we can tackle the joint reconstruction, age model, and model selection problems in a fully Bayesian manner, but it remains computationnally expensive. This work is detailed in publications in preparation.
Selected references
M. Crucifix, Traditional and novel approaches to palaeoclimate modelling, Quaternary Science Reviews, 57, 1--16 2012
M. Crucifix, Why could ice ages be unpredictable?, Climate of the Past, 9, 2253--2267 2013
T. Mitsui, M. Crucifix, K. Aihara, Bifurcations and strange nonchaotic attractors in a phase oscillator model of glacial-interglacial cycles, in revision in Physica D
P. A. Araya-Melo, M. Crucifix, and N. Bounceur, Global sensitivity analysis of Indian Monsoon during the Pleistocene, Climate of the Past Discussions, 10, 1609--1651 2014 (in press in Climate of the Past)
Software pakcages :
Package palinsol : http://cran.r-project.org/package=palinsol
Package GP : Gaussian process for emulation of GCMs : https://bitbucket.org/mcrucifix/gp
Package iceages : https://bitbucket.org/mcrucifix/iceages