How we calculate internal variability of GCMs? What is the difference between internal variability and interannual variability?

please see the online pdf

Internal variability and model uncertainty components in

future hydrometeorological projections: The Alpine Durance

basin

M. Lafaysse1,2, B. Hingray1,3, A. Mezghani1,3,4, J. Gailhard5, and L. Terray6

University of Grenoble Alpes, LTHE UMR 5564, Grenoble, France, 2

Meteo-France/CNRS, CNRM-GAME, CEN, SaintMartin-d’He`res,

France, 3

CNRS, LTHE UMR 5564, Grenoble, France, 4

Now at Norwegian Meteorologisk Institutt, Oslo,

Norway, 5

EDF-DTG, Grenoble, France, 6

CERFACS/Sciences de l’Univers, CERFACS-CNRS URA 1875, Toulouse, France

Abstract A multireplicate multimodel ensemble of hydrological simulations covering the 1860–2099

period has been produced for the Upper Durance River basin (French Alps). An original quasi-ergodic analysis

of variance was applied to quantify uncertainties related to General Circulation Models (GCMs), Statistical

Downscaling Models (SDMs) and the internal variability of each GCM/SDM simulation chain. For temperature,

GCM uncertainty prevails and SDM uncertainty is nonnegligible. Significant warming and in turn significant

changes are predicted for evaporation, snow cover and seasonality of discharges. For precipitation,

GCM and SDM uncertainty components are of the same order. A high contribution of the large and smallscale

components of internal variability is also obtained, inherited, respectively, from the GCMs and the different

replicates of a given SDM. The same applies for annual discharge. The uncertainty in values that could

be experienced for any given future period is therefore very high. For both discharge and precipitation,

even the sign of future realizations is uncertain at a 90% confidence level. These findings have important

implications. Similarly to GCM uncertainty, SDM uncertainty cannot be neglected. The same applies for both

components of internal variability. Climate change impact studies based on a single SDM realization are

likely to be no more relevant than those based on a single GCM run. They may lead to poor decisions for climate

change adaptation.