EUCLIPSE Toulouse meeting April 2012 Roel Neggers Process-level evaluation at selected grid-points: Constraining a system of interacting parameterizations through multiple parameter evaluation at Cabauw
Process-level evaluation We have built good experience at (idealized) single-case level (e.g. GCSS), and can demonstrate successes in model improvement. However, possible shortcomings: * Cases might not represent actual climate; parameterizations might get tuned to rare situations * Cases might not represent those situations that are most troublesome in GCMs; * The use of relevant observational data has been somewhat limited. These arguments motivate a move towards more comprehensive, statistically significant approach in model evaluation, in combination with a more efficient use of observational datasets.
The “testbed” idea 1) SCM and GCM evaluation for long periods of time at permanent meteorological sites (e.g. ARM, CloudNet) Emphasis: fast physics (boundary-layer, soil) Emphasis: fast physics (boundary-layer, soil) 2) Use a multiple-parameter approach in the evaluation (“CloudNet+”) Constrain the system of interacting parameterizations at multiple points with key measurements Constrain the system of interacting parameterizations at multiple points with key measurementsGoals: * Try and identify compensating errors in interaction between low clouds & radiative transfer (e.g. too bright – too few) * Try and identify compensating errors in interaction between low clouds & radiative transfer (e.g. too bright – too few) * Trace their impact through the coupled BL – soil system * Trace their impact through the coupled BL – soil system (heat & moisture budgets) (heat & moisture budgets) Neggers et al, BAMS, in press, 2012
Short example The motivation: problems with a new BL scheme in IFS Too little cloud cover at noon
Associated 2m T warm bias over land
Can long-term SCM evaluation at Cabauw provide some insight? 1. less PBL clouds 2. larger SW down 3. larger H 4. low level warming SW Hypothesis
Long-term SCM evaluation at Cabauw ( ) Obs vs Model scatterplots of monthly means RACMO SCM: Control (red) and new (blue) scheme RACMO SCM: Control (red) and new (blue) scheme RACMO 3D in forecast mode (grey) RACMO 3D in forecast mode (grey)
8-point check Expanding to multiple independently-measured parameters that reflect the impact mechanism as illustrated before
Model performance – Taylor diagram Reproduction of observed pattern in variation of monthly-mean
Correlation between model differences
Tracing impacts Following degrading correlations in coupled BL – soil system
Zooming in
Identifying regimes Cabauw cloud-scenes on 8 days with biggest model difference
Conditional sampling Make a list of shallow cu days Criterion: 1.positive surface buoyancy flux 2.LCL below BL top 3.Total cloud cover < 50%
Evaluation against LES New set of 4 parameters reflecting cloud vertical structure in the boundary-layer LES (x) vs SCM (y), daily values Evaluated for the time-range UTC to capture diurnal variation One month of DALES (June 2008) z base cf max z cf max overlap ratio controlnew
Bar-chart Monthly-mean bias for June 2008
Impact of including a new cloud overlap function Multi-year bias for Green: including SGS overlap for cumuliform clouds (Neggers et al, JGR, 2011)
Conclusions Multiple parameter evaluation was performed against multi-year Cabauw data RACMO physics was subjected to a 12-point check reflecting the cloud structure, radiative budget and heat budget of coupled boundary-layer soil system This revealed the existence of a compensating error between the representation of cloud vertical structure and cloud overlap in the cumuliform boundary layer