Generalizing cloud overlap treatment to include solar ... - Cosmo

Statement of the Problem Proposed new overlap scheme Scheme validation in a 1D framework Validation in 3D simulation ConclusionsGeneralizing cloud overlap treatment toinclude solar zenith angle effects on cloudgeometryFrancesca Di Giuseppe and Adrian M. TompkinsICTP, Trieste Italy and ARPA-SIM, Bologna,ITCOSMO, Cracow 2008All material available at:http : //users.ictp.it/ tompkins/Tompkins A. M. and F. Di Giuseppe. Generalizing Cloud Overlap Treatment to Include Solar Zenith Angle Effects on Cloud Geometry.Journal of the Atmospheric Sciences. 64, 2116 − 2125. (2007)

Statement of the Problem Proposed new overlap scheme Scheme validation in a 1D framework Validation in 3D simulation ConclusionsOutlineStatement of the ProblemProposed new overlap schemeScheme validation in a 1D frameworkDataResultsValidation in 3D simulationECMWFCOSMOConclusions

Statement of the Problem Proposed new overlap scheme Scheme validation in a 1D framework Validation in 3D simulation ConclusionsSW Radiative Transfer in cloudsProblem: modelling the photon interaction with a 3D cloud field.Several “geometrical” aspects need to be considered:1spatial arrangements of clouds element2vertical overlap3sun position

Statement of the Problem Proposed new overlap scheme Scheme validation in a 1D framework Validation in 3D simulation ConclusionsRepresentation of clouds in GCMGCMs represent the previous two scenes in the same way regardlessof the sun positionAt each vertical layer the cloud is represented in terms of a cloud fraction.The vertical arrangement is established following overlap rules. The sunposition is only accounted for by rescaling the photon path length.2 effects are neglected: the horizontal transport of photons and the effectiveincrease of cloud cover at lower sun angles.% bias correction of PP by TICAcompared to benchmark 3D MCcalculation: sun low=error fromgeometry, sun high error from 3DscatteringDi Giuseppe, F., and A.M. Tompkins, 2005: Impact of Cloud Cover on SolarRadiative Biases in Deep Convective Regimes. , 62,1989 − 2000

Statement of the Problem Proposed new overlap scheme Scheme validation in a 1D framework Validation in 3D simulation ConclusionsOverlap rules1 maximum overlap: C MAXi,j2 random overlap: C RANi,j= max{C i , C j } i.e. min Cloud Cover (CC)= C i + C j − C i ∗ C j3 max-ran overlap: adjacent layer=max overlap, non-adjacent layer=ran

Statement of the Problem Proposed new overlap scheme Scheme validation in a 1D framework Validation in 3D simulation ConclusionsHogan approachEXP-RAN overlap: Using as proxy RADAR data (Hogan and Illingworth(2000)) proposed a decorrelation length L 0 (approx 2 km) for continuousclouds.C EXP−RANi,j= αC MAX−RANi,jwhere: α = exp(− ∆zL 0)+ (1 − α)C RANi,j

Statement of the Problem Proposed new overlap scheme Scheme validation in a 1D framework Validation in 3D simulation ConclusionsParameterising solar zenith angle effectsThe new parametrisation represents the solar zenith SZA (θ 0 ) effect onapparent cloud geometry by extending the schemes EXP-RAN of Hogan, sothat:L = K (θ 0 )L 0The new scheme is called EXP-SZA-RAN.Assuming L 0 the decorrelation length-scale for the sun overhead case. Weneed a parametrisation for K (θ 0 )

Statement of the Problem Proposed new overlap scheme Scheme validation in a 1D framework Validation in 3D simulation ConclusionsDATARadar retrievals from the CloudNet EU project which provided almostcontinuous observations between the April to November 2003 at Chilbolton,England and between March 2003 to September 2004 at Palaiseau, Franceare used.Radar classification field from typical day (27/04/2003) of the Chilbolton radar retrieval. These scenes are split into 2.25 hour segments,which are equivalent to an approximate dimension of 40km assuming a uniform horizontal advection velocity of 5ms −1

Statement of the Problem Proposed new overlap scheme Scheme validation in a 1D framework Validation in 3D simulation ConclusionsMethodologyMissing data points corrected if possible, or scenediscardedOnly scenes with 5% < CloudCover < 80% areconsideredHorizontal sub-cloud inhomogeneities are removed(plane parallel)Scenes are shifted using the SZA for the TICAbenchmark simulation.L 0 chosen from data to given correct total cloudcover when sun overhead (L 0 =4.0km)Radiative calculations made with scheme ofMorcrette (1991) as used in ECMWF operationalmodel in 2005.

Statement of the Problem Proposed new overlap scheme Scheme validation in a 1D framework Validation in 3D simulation ConclusionsTotal Cloud Cover (TCC)TCC vs SZA averaged for over 150 cloud scenes at Chilbolton site. The trueTICA TCC reference is shown by the dotted line on the left.1 Random overlap is good approximation for low sun angles2 The linear parametrization fits the cloud cover the best

Statement of the Problem Proposed new overlap scheme Scheme validation in a 1D framework Validation in 3D simulation ConclusionsImpact on radiative biasesThe new scheme greatly reduces the net bias (wrt TICA) integratedover all sun angles compared to the standard overlap schemes

Statement of the Problem Proposed new overlap scheme Scheme validation in a 1D framework Validation in 3D simulation ConclusionsECMWF’s implementation3D calculations repeated with RapidRadiative Transfer Model (RRTM) radiationcode (Mlawer and Clough 1997) which isnow operational at ECMWF (Scheme choicedid not affect the conclusions)Panels show effect on TOA SW fluxes in 3hour forecast of changing progressivelyMAX-RAN − > EXP-RAN − >EXP-SZA-RAN (new) − > RANnew scheme has half impact of EXP-RAN(depends on L 0 )dynamical feedback limited in year-longintegrations (with imposed SSTs)

Statement of the Problem Proposed new overlap scheme Scheme validation in a 1D framework Validation in 3D simulation ConclusionsCOSMO’s implementationPreliminary results from the COSMO implementation.Panel shows effect on TOA SW fluxes in 3 hourforecast of changing EXP-RAN − > EXP-SZA-RAN (new)new scheme has major impact where sunlower above the horizon ( thus it depends on latitude)to put into contest:in March at 12:00 UTCSZA = 45deg at 35N latitudeSZA = 60deg at 50N latitudeLocally, effects as large as 40 Wm-2.

Statement of the Problem Proposed new overlap scheme Scheme validation in a 1D framework Validation in 3D simulation ConclusionsConclusionsHogan generalised the MAX-RAN to the EXP-RAN overlap toaccount for shearing of continuous cloudsWe have further generalised EXP-RAN to the EXP-SZA-RAN toaccount for the sun positionThe new scheme improves the total cloud cover errors asfunction of sun angleThe scheme improves radiative flux/heating rate calculationrelative to benchmark TICA calculationsParametrisation extremely simple to implementThe dynamical impact in 3D model was limited - but this is alsothe case changing from MAX-RAN to Hogan’s EXP-RAN. Largerin coupled model?