* Submission of results by 1 December 2013

* Simulation start 1 October 2005
* Forcing by AMIP2 sea surface temperature and sea-ice extent
* Preferred: Nudge toward ECMWF reanalysis winds (not temperature) through 2010
* Acceptable: Nudge toward winds from one baseline simulation by your model
* Less acceptable: No nudging
* Greenhouse gas concentrations for year 2000
* Aerosol direct, semi-direct, and indirect effects taken into account.

* all_2000: simulation PD (present-day): year 2000 IPCC aerosol emissions
* all_1850: simulation PI (pre-industrial): year 1850 IPCC aerosol emissions (year 2000 GHG concentration)
* hom_2000: present day emissions no heterogeneous nucleation of ice
* hom_1850: pre-industrial emissions no heterogeneous nucleation of ice
* fix_2000: present day emissions fixed ice nucleation for T<-37 C using Cooper (1986) as a f(temperature)
* fix_1850: pre-industrial emissions fixed ice nucleation for T<-37 C using Cooper (1986) as f(temperature)

* All data except COSP diagnostics is to be collected at the AEROCOM server.
* Groups hold COSP diagnostics and contact Kenta Suzuki (Kentaro.Suzuki@jpl.nasa.gov) for analysis
* follow the aerocom data protocol (http://aerocom.met.no/protocol.html)
* Data in NetCDF format, one variable and year per file with CMOR variable names
* All data are 3-dimensional ( lon x lat x time )
* filenames aerocom_<ModelName>_<ExperimentName>_<VariableName>_<VerticalCoordinateType>_<Period>_<Frequency>.nc

  where <ModelName> can be chosen such that Model Name, Model version and possibly the institution can be identified. No underscores (_) are allowed in <ModelName>. Use (-) instead. Max 20 characters. 
  <ExperimentName> = all_2000, all_1850, hom_2000, hom_1850, fix_2000, or fix_1850 
  <VariableName> see list below 
  <VerticalCoordinateType> => "Surface", "Column", "ModelLevel" 
  <Period> => "2008", "2010", ...  
  <Frequency> => "timeinvariant","hourly", ,"3hourly", "daily", "monthly" 

* CFMIP COSP diagnostics provided by COSP do not need to be run through cmor because the names are the same,
* but please separate files for each variable

  In addition to the diagnostics below, it is highly recommended to store the AEROCOM standard and forcing diagnostics, 
  so that the simulations can be analysed for the direct forcing as well, and future more in-depth analyses are possible. 

(1) 2D diagnostics for evaluation with satellite data

5 years (years 2006-2010) of 3-hourly data from the PD run

(2) For forcing estimates: as in (1), but monthly-mean fields for both PD and PI simulations, plus a land-ocean mask (0 land, 1 ocean).

(3) 3D monthly mean diagnostics

(4) Optional CFMIP COSP diagnostics. Highly desirable but optional for now
3-hr snapshots and daily means for January-March 2008 PD simulation only.

The idea is to use the cloud overlap assumption (maximum, random, or maximum-random) to estimate which part of the cloud in a
layer can be seen from above.

Note: For the CCN, whether to sample it in the same way as CDNC, or use a similar apporach (going from bottom up)
to sample it at cloud base depends on your parameterization of the activation.

  let i=1,2,...,nx be the index for the horizontal grid-points
  let k=1,2,...,nz be the index for the vertial levels, with 1 being the uppermost level, and nz the surface level 

naming convention for the 3D input fields:

  iovl is the flag to select the overlap hypothesis
  cod3d(nx,nz) cloud optical thickness
  f3d(nx,nz) cloud fraction
  t3d(nx,nz) temperature
  phase3d(nx,nz) cloud thermodynamic phase (0: entire cloud consists of ice, 1: entire cloud consists of liquid water, between 0 and 1: mixed-phase)
  cdr3d(nx,nz) cloud droplet effective radius
  icr3d(nx,nz) ice crystal effective radius
  cdnc3d(nx,nz) cloud droplet number concentration 

thres_cld = 0.001
thres_cod = 0.3
IF ( iovl = random OR iovl = maximum-random ) THEN

clt(i) = 1.

ELSE

clt(:) = 0

ENDIF
icc(:) = 0
lcc(:) = 0
ttop(:) = 0
cdr(:) = 0
icr(:) = 0
cdnc(:) = 0

DO i=1,nx

DO k=2,nz ! assumption: uppermost layer is cloud-free (k=1)
	IF ( cod3d(i,k) > thres_cod and f3d(i,k) > thres_cld ) THEN ! visible, not-too-small cloud
		! flag_max is needed since the vertical integration for maximum overlap is different from the two others: for maximum, clt is the actual cloud cover in the level, for the two others, the actual cloud cover is 1 - clt
		! ftmp is total cloud cover seen from above down to the current level
		! clt is ftmp from the level just above
		! ftmp - clt is thus the additional cloud fraction seen from above in this level

		IF ( iovl = maximum ) THEN
			flag_max = -1.
			ftmp(i) = MAX( clt(i), f3d(i,k))  ! maximum overlap	
		ELSEIF ( iovl = random ) THEN
			flag_max = 1.
			ftmp(i) = clt(i) * ( 1 - f3d(i,k) ) ! random overlap	
		ELSEIF ( iovl = maximum-random ) THEN
			flag_max = 1.
			ftmp(i) = clt(i) * ( 1 - MAX( f3d(i,k), f3d(i,k-1) ) ) / &
 	            ( 1 - MIN( f3d(i,k-1), 1 - thres_cld ) )  ! maximum-random overlap	
		ENDIF
		ttop(i) = ttop(i) + t3d(i,k) * ( clt(i) - ftmp(i) )*flag_max 

		! ice clouds
		icr(i) = icr(i) + icr3d(i,k) * ( 1 - phase3d(i,k) ) * ( clt(i) - ftmp(i) )*flag_max 
		icc(i) = icc(i) + ( 1 - phase3d(i,k) ) * ( clt(i) - ftmp(i) )*flag_max 

		! liquid water clouds
		cdr(i) = cdr(i) + cdr3d(i,j) * phase3d(i,k) * ( clt(i) - ftmp(i) )*flag_max 
		cdnc(i) = cdnc(i) + cdnc3d(i,j) * phase3d(i,k) * ( clt(i) - ftmp(i) )*flag_max 
		lcc(i) = lcc(i) + phase3d(i,k) * ( clt(i) - ftmp(i) )*flag_max 
		
		clt(i) = ftmp(i)
	ENDIF ! is there a visible, not-too-small cloud?
ENDDO ! loop over k

IF ( iovl = random OR iovl = maximum-random ) THEN
	clt(i) = 1. - clt(i)
ENDIF

ENDDO ! loop over I

naming convention for the input variables:

  utctime current time of the day in UTC in seconds  
  time_step_len length of model time-step  
  lon(nx) longitude in degrees from 0 to 360   

* 2D cloud fields (lwp, iwp, cdr, cdnc, cod): Please compute them from grid-box mean values at each level but DO NOT divide by the total (2D) cloud cover, which will be done in analysis after averaging in time and space.

* The three months 1 October - 31 December 2005 are thought as spin-up, which can of course be longer. Please choose as overlap assumption the one you use in the radiation scheme.

* ATTENTION: clt(i) has to be initialized to 1 for random or maximum-random overlap assumptions in the “satellite simulator”

* CCN definition: Compute CCN using Kohler theory at 0.1 and 0.3 % supersaturation.