SPOOKIE: The Selected Process On/Off Klima Intercomparison Experiment

Transcription

1 SPOOKIE: The Selected Process On/Off Klima Intercomparison Experiment Mark Webb, Adrian Lock (Met Office), Sandrine Bony (IPSL), Chris Bretherton (UW), Tsuyoshi Koshiro, Hideaki Kawai (MRI), Thorsten Mauritsen (MPI), Tomoo Ogura (NIES), Romain Roehrig (CNRM), Steve Sherwood (UNSW), Masahiro Watanabe (AORI), Jessica Vial (IPSL), Ming Zhao (GFDL) International Workshop on Risk Information on Climate Change,Yokohama, November, 2014

2 Background The radiative feedbacks of clouds on climate change differ widely between climate models Our understanding of the physical mechanisms underlying the different model responses is still very limited, which makes it difficult to assess their credibility The Cloud Feedback Model Intercomparison Project (CFMIP) has in recent years demonstrated that much of the spread in cloud feedback in coupled atmosphere-ocean climate models can be reproduced in idealised atmosphere-only experiments forced with specified SSTs This provides a new opportunity to investigate and understand the physical mechanisms responsible for different cloud feedbacks in relatively computationally inexpensive experiments

3 Selected Process On/Off Klima Intercomparison Experiment (SPOOKIE) Aims Establish the relative contributions of different areas of model physics to inter-model spread in cloud feedbacks Approach Repeat CFMIP-2 AMIP/AMIP Uniform +4K SST perturbation experiments, switching off or simplifying different model schemes in turn Pilot Experiments Start by switching off convective parametrization

4 Convection-off (ConvOff) experiments: what do we expect to see? Zhang et al. (2013) argue that positive subtropical feedback is caused by enhanced entrainment of dry air into the boundary layer from the free troposphere in models with active shallow convection. - If this is the sole cause of positive subtropical feedback, then convoff feedbacks will be neutral or negative. Sherwood et al. (2014) argue that positive shallow cloud feedback (and much of its spread) is related to the strength of lower tropospheric mixing by convection and large scale circulations. Zhao (2014) also highlights sensitivity of feedbacks to deep convection in the GFDL model - Hence some reduction in magnitude and spread of cloud feedback would be expected in the absence of convective parametrization

5 Global Mean Cloud Feedback (Wm -2 K -1 ) (Including cloud masking)

6 How does convection affect inter-model spread in cloud feedback? Ensemble variance of amip4k Net Cloud Feedback Standard models Convection off

7 Angular LTS/Precipitation Index (ALPI)

8 All rd Models amip4k Cloud Feedbacks over 30 O N/S Oceans SPOOKIE Standard models SPOOKIE ConvOff models Net Cloud Feedback* Black diamonds mark significant correlations with the net cloud feedbacks in same regime Shortwave Cloud Feedback* Grey squares mark significant correlations with the net cloud feedback area averaged over tropical oceans Longwave Cloud Feedback* *Includes cloud masking

9 All rd Models amip4k Cloud Responses over 30 O N/S Oceans SPOOKIE Standard models SPOOKIE ConvOff models Total Cloud Fraction Response (%/K) Liquid Water Path Response (mm/k) Ice Water Path Response (mm/k)

10 All rd Models AMIP Cloud Properties over 30 O N/S Oceans SPOOKIE Standard models SPOOKIE ConvOff models Total Cloud Fraction Liquid Water Path (mm) Ice Water Path (mm)

11 All rd Models AMIP Low/Mid/High Cloud Fractions 30 O N/S Oceans SPOOKIE Standard models SPOOKIE ConvOff models Maximum Low Level Cloud Fraction (below 680 hpa) Maximum Mid Level Cloud Fraction ( hpa) Maximum High Level Cloud Fraction (above 440 hpa)

12 Summary of Results ConvOff experiments reproduce positive subtropical cloud feedback. They show strong convergence in tropical deep convective cloud feedbacks and some convergence in stratocumulus regimes but increased spread in the trades. Hence the range in global cloud feedback is not reduced, and convective parametrization differences cannot explain the spread. Convoff experiments generally have more low level cloud and larger ice and liquid water paths. Models with less low level cloud in heavily precipitating regions of the tropics and more cirrus in stable regimes have stronger tropical cloud feedbacks and vice versa.

13 Convective precipitation efficiency and cloud feedback Sherwood et al. Nature 2014 and Zhao 2014 argue that convective precipitation efficiency plays an important role in cloud feedback. Models with strong convective precipitation efficiency rain out condensate efficiently with little evaporation of condensate or precipitation. Models with weak precipitation efficiency need to transport more water vapour from the boundary layer to the free troposphere to produce a given amount of surface precipitation. Sherwood et al. argued that models with weaker convective precipitation efficiency and shallower overturning circulations transport more water vapour vertically out of the boundary layer, and that this effect strengthens in the warming climate, drying the boundary layer and reducing low cloud.

14 Could large-scale precipitation efficiency control overall spread in cloud feedback? Might instead large scale precipitation efficiency be the dominant control of the vertical flux of water vapour in deep convective regions? Large scale precipitation efficiency will depend on various aspects of cloud parametrization/cloud microphysics. Models that can form clouds and condensate more easily at low levels in ascending regions of the tropics may find it easier to rain out to the surface, and hence have stronger large scale precipitation efficiencies. Models which condense over a greater depth at higher levels where water contents are smaller might precipitate less easily and evaporate more. There will be more opportunity for precipitation to evaporate before reaching the surface, resulting in weaker precipitation efficiencies. Such models will dry the boundary layer more and moisten the free troposphere more, and this effect will strengthen with the hydrological cycle as the climate warms.

15 Could large-scale precipitation efficiency control overall spread in cloud feedback? Sherwood et al argued that enhanced drying of the tropical boundary layer will dry low cloud regions resulting in stronger low cloud reductions. Alternatively it is possible that the additional moistening of the free troposphere propagates to the subtropics and encourages the breakup of low level clouds by inhibiting radiative cooling in the boundary layer. Studies with LES have demonstrated that an enhanced free-tropospheric greenhouse effect can reduce turbulent mixing and cloudiness in the subtropical boundary layer - e.g. Bretherton et al. JAMES 2013

16 Does this idea fit with our results? Higher sensitivity models have less low/mid level cloud in strongly precipitating regions of the tropics while lower sensitivity models have more. Models with more low/mid level clouds in deep convective areas might be be more efficient at raining out and so have larger precipitation efficiency. Conversely models with less low/mid-level cloud in precipitating regions may be doing more of their condensation at higher levels. This might reduce precipitation efficiency. Higher sensitivity models have more ice cloud in the subtropics this could be a consequence of doing more condensation at upper levels.

17 Ideas For Future Work We can test these and other ideas with further sensitivity tests. These might form the basis for the next phase of SPOOKIE. For example: Reducing ice fall speeds would be predicted to reduce precipitation efficiency and make cloud feedback more positive. Modifying cloud and precipitation schemes to condense/rain out more easily at low levels in deep convective regions would be predicted to increase precipitation efficiency and make cloud feedback less positive. Adding an artificial humidity source to the subtropical free troposphere would be predicted to reduce low level cloudiness. Alternatively the radiative coupling between free tropospheric humidity and low level cloud could be explored by manipulating humidity inputs to the radiation code.