A global climate system model of intermediate complexity is used to simulate the climate of the Holocene and to investigate the robustness of the model results with respect to changes in the experimental design and in the parameterization of the land surface.
The CLIMBER climate system model for the first time allows to run simulations of different model configurations (components: atmosphere, ocean, vegetation, the latter two either variable or fixed to a given state) and thus, to estimate consistently the impact of the individual climate subsystems and of their synergism. The respective effects are clearly demonstrated by the use of amplification factors. Also, it has not been possible to run the kind of transient simulations presented here with other models so far.
Similar to other models, equilibrium simulations of 6000 years before present reveal a warmer and wetter climate in the Northern Hemisphere in summer, a northward shift of the boreal forests and a widespread vegetation in the Saharan region. Synergism can be found mainly in the high northern latitudes due to the interaction of the vegetation-snow-albedo feedback and the sea-ice albedo feedback. It results in higher temperatures there nearly all the year round and resolves the so-called biome paradox in the geological data in this way. By a reduction of the northward heat transport in the Atlantic it also leads to a warming of the Antarctic region. In the subtropics, synergistic effects play a minor role.
Transient simulations through the past 9000 years show a gradual decrease of the high summer temperatures in the Northern Hemisphere and a corresponding slow retreat of the boreal forests. Only in North Africa, the strong positive feedback between precipitation and vegetation results in a rather abrupt retreat of the vegetation around about 5500 years before present.
When comparing two simulations, changes in the surface energy and water balance are traced back to changes in atmospheric variables as well as to certain parameters. The latter mainly determine differences between different surface types with their possibly variable fractions. For this reason, the effects not only of a modified experimental design but also of modified surface parameters are investigated in a series of sensitivity studies:
a) Changes in the initial or in the boundary conditions affect mainly the sea ice. A vegetation sensitivity in the model is seen only when applying unrealistic initial conditions. In this way, differences of up to 20% in the vegetation fraction can be found for the Sahara.
b) The response of vegetation in the Sahara and in the Sahel to changes in the ocean depends in different ways on the insolation forcing.
c) The transient simulations are robust against changes in the initial conditions.
d) With a modification of the boundary conditions, the timing of vegetation shifts can change in the range of hundred years.
e) A hysteresis effect with a magnitude of a few hundred years is found.
f) The vegetation shift in boreal latitudes and in the subtropics depends on both the ocean in high as well as in lower latitudes.
g) For the boreal latitudes, the model sensitivity is rather small and mainly to different but still reasonable values of the roughness length and to the parameterization of the snow fraction, and to a lesser degree to changes in the leaf area index. In this way, differences in the vegetation fraction stay smaller than 10%.
h) For North Africa, the different values of the roughness length can result in changes in the vegetation fraction of up to 35%. The albedo of the Sahara clearly influences the climate in the Sahel, the climate in the Sahel in turn is strongly influenced by the albedo of the central Asian deserts.
By modifying the surface parameters, it is possible to reduce differences between the model results and the geological data. Only such kind of (usually neglected) sensitivity studies reveal that quantitative differences, as for example in the vegetation cover, of several ten percent are possible, but that with respect to the land surface the model results are qualitatively stable.
|
--
