Present and LGM permafrost from climate simulations: contribution of statistical downscaling 1Laboratoire des Sciences du Climat et de L'Environnement (LSCE), UMR8212, IPSL – CEA/CNRS-INSU/UVSQ, Centre d'étude de Saclay, Orme des Merisiers, 91191 Gif-sur-Yvette, France
16 Nov 2011
2Section Climate Change and Landscape Dynamics, Department of Earth Sciences, Faculty of Earth and Life Sciences, VU University Amsterdam, de Boelelaan 1085, 1081 HV Amsterdam, The Netherlands
Received: 04 May 2011 – Published in Clim. Past Discuss.: 25 May 2011 Abstract. We quantify the agreement between permafrost distributions from PMIP2
(Paleoclimate Modeling Intercomparison Project) climate models and permafrost
data. We evaluate the ability of several climate models to represent
permafrost and assess the variability between their results.
Revised: 07 October 2011 – Accepted: 07 October 2011 – Published: 16 November 2011
Studying a heterogeneous variable such as permafrost implies conducting
analysis at a smaller spatial scale compared with climate models resolution.
Our approach consists of applying statistical downscaling methods (SDMs) on
large- or regional-scale atmospheric variables provided by climate models,
leading to local-scale permafrost modelling. Among the SDMs, we first choose
a transfer function approach based on Generalized Additive Models (GAMs) to
produce high-resolution climatology of air temperature at the surface. Then
we define permafrost distribution over Eurasia by air temperature conditions.
In a first validation step on present climate (CTRL period), this method
shows some limitations with non-systematic improvements in comparison with
the large-scale fields.
So, we develop an alternative method of statistical downscaling based on a
Multinomial Logistic GAM (ML-GAM), which directly predicts the occurrence
probabilities of local-scale permafrost. The obtained permafrost
distributions appear in a better agreement with CTRL data. In average for the
nine PMIP2 models, we measure a global agreement with CTRL permafrost data
that is better when using ML-GAM than when applying the GAM method with air
temperature conditions. In both cases, the provided local information reduces
the variability between climate models results. This also confirms that a
simple relationship between permafrost and the air temperature only is not
always sufficient to represent local-scale permafrost.
Finally, we apply each method on a very different climate, the Last Glacial
Maximum (LGM) time period, in order to quantify the ability of climate models
to represent LGM permafrost. The prediction of the SDMs (GAM and ML-GAM) is
not significantly in better agreement with LGM permafrost data than
large-scale fields. At the LGM, both methods do not reduce the variability
between climate models results. We show that LGM permafrost distribution from
climate models strongly depends on large-scale air temperature at the
surface. LGM simulations from climate models lead to larger differences with
LGM data than in the CTRL period. These differences reduce the contribution
Citation: Levavasseur, G., Vrac, M., Roche, D. M., Paillard, D., Martin, A., and Vandenberghe, J.: Present and LGM permafrost from climate simulations: contribution of statistical downscaling, Clim. Past, 7, 1225-1246, doi:10.5194/cp-7-1225-2011, 2011.