Home

The 1-way on-line coupled atmospheric chemistry model system

1. messy_init_memory DDEP JVAL LNOX MECCA ONEMIS PTRAC SCAV SEDI TNUDGE INITIALISATION aadu anit ling gt TRACER CVTRANS DDEP JVAL LNOX MECCA OFFEMIS ONEMIS SCAV messy_init_coupling SEDI TNUDGE TREXP CHANNEL messy_init_tracer gt TRACER MECCA messy_init_loop gt TIMER gt CHANNEL DATA TRACER IMPORT DRADON DDEP messy_global_start ONEMIS JVAL TNUDGE TREXP ap gt TRACER DATA DRADON OFFEMIS ONEMIS MECCA DDEP messy_vdiff messy convec gt TRACER DATA CVTRANS SCAV messy physc gt TRACER DATA DRADON LNOX JVAL SCAV SEDI MECCA TREXP Es TRACER DATA TNUDGE INTEGRATION PHASE messy_local_end messy global end gt TRACER CHANNEL QTIMER messy_write_output gt TRACER CHANNEL CO aa am gt CVTRANS DDEP JVAL MECCA OFFEMIS SCAV SEDI TNUDGE messy free_memory TREXP IMPORT TRACER CHANNEL QTIMER FINALISATION lt Fig 2 List of all entry points available in ECHAM5 MESSy The entry points highlighted with yellow boxes are currently imple mented in COSMO MESSy the hatched ones are not yet required as the dependent submodels are not used in COSMO MESSy so far Next to the boxes the generic blue and regular black submodels currently called in COSMO MESSy by the respective CONTROL subroutine are listed The green box surrounds those entry points located in the outer lo
2. Appendix B List of most important code changes COSMO Prognostic and diagnostic fields are transformed form ALLOCATABLE fields to POINTER in order to be allocated and used by the MESSy memory management This also includes the change of IF ALLOCATED field into IF ASSOCIATED field throughout the code www geosci model dev net 5 87 2012 107 Prognostic and diagnostic fields are de allocated and initialised by the MESSy memory management instead of inalloc_ meteofields The calculation of the diagnostic fields for the output is additionally called in those time steps when MESSy output is triggered Some local variables of COSMO are replaced by POINTER defined in the MESSy submodel DATA The POINTER are used in COSMO instead of the local vari ables in order to make the information available to the MESSy submodels The clock of COSMO is set and synchronised by with the MESSy submodel TIMER B2 MESSy B2 1 Infrastructure Naturally most extensions for the implementation of MESSy into COSMO are required for the infrastructure submodels because these form the adapter to the respective basemodel Here we name only the major changes CONTROL Only those submodels applicable to COSMO are called CHANNEL Representations are defined corresponding to the COSMO grid Restart checks depend on the basemodel DATA This code is naturally specific for COSMO i e base model depe
3. L6pez Puertas M Tsidu G M Milz M Steck T and Steil B Global distribu tions of HO2NO 2 as observed by the Michelson Interferometer for Passive Atmospheric Sounding MIPAS J Geophys Res 112 D09314 doi 10 1029 2006JD007212 2007 Tiedtke M A comprehensive mass flux scheme for parameteriza tion in large scale models Mon Weather Rev 117 1779 1800 1989 Tost H J ckel P and Lelieveld J Influence of different convec tion parameterisations in a GCM Atmos Chem Phys 6 5475 Geosci Model Dev 5 87 110 2012 110 5493 doi 10 5194 acp 6 5475 2006 2006 Tost H J ckel P and Lelieveld J Lightning and convec tion parameterisations uncertainties in global modelling At mos Chem Phys 7 4553 4568 doi 10 5 194 acp 7 4553 2007 2007 Tost H Lawrence M G Br hl C The GABRIEL Team and The SCOUT 03 DARWIN ACTIVE Team Uncertainties in atmospheric chemistry modelling due to convection parameterisations and subsequent scavenging Atmos Chem Phys 10 1931 1951 doi 10 5194 acp 10 1931 2010 2010 Jockel P Geosci Model Dev 5 87 110 2012 A Kerkweg and P Jockel COSMO MESSy Wetzel G Oelhaf H Kirner O Ruhnke R Friedl Vallon F Kleinert A Maucher G Fischer H Birk M Wagner G and Engel A First remote sensing measurements of CIOOCI along with ClO and CIONO in activated and deactivated Arctic vortex conditions using new
4. Baumgaertner A J G J ckel P Riede H Stiller G and Funke B Energetic particle precipitation in ECHAMS MESSy Part 2 Solar proton events Atmos Chem Phys 10 7285 7302 doi 10 5194 acp 10 7285 2010 2010 Br hl C Steil B Stiller G Funke B and Jockel P Nitrogen compounds and ozone in the stratosphere com parison of MIPAS satellite data with the chemistry climate model ECHAM5 MESSy1 Atmos Chem Phys 7 5585 5598 doi 10 5194 acp 7 5585 2007 2007 Geosci Model Dev 5 87 110 2012 A Kerkweg and P Jockel COSMO MESSy Burrows S M Butler T J ckel P Tost H Kerkweg A P schl U and Lawrence M G Bacteria in the global at mosphere Part 2 Modeling of emissions and transport be tween different ecosystems Atmos Chem Phys 9 9281 9297 doi 10 5194 acp 9 928 1 2009 2009 Butchart N Charlton Perez A J Cionni I Hardiman S C Haynes P H Krueger K Kushner P J Newman P A Os prey S M Perlwitz J Sigmond M Wang L Akiyoshi H Austin J Bekki S Baumgaertner A Braesicke P Br hl C Chipperfield M Dameris M Dhomse S Eyring V Gar cia R Garny H J ckel P Lamarque J F Marchand M Michou M Morgenstern O Nakamura T Pawson S Plum mer D Pyle J Rozanov E Scinocca J Shepherd T G Shibata K Smale D Teyssedre H Tian W Waugh D and Yamashita Y Multimodel climate a
5. composed grid similar to the ECHAMS grid decompo sition But each of these core regions is surrounded by a frame of ghost boundaries in order to more efficiently calculate physical processes which require information of the neighbouring grid points e g advection The ghost boundaries of the local domain overlap with the core regions of the neighbouring PEs At certain points within a model time step the variables on the bound aries are exchanged between neighbouring PEs Conse quently the ECHAMS global Eulerian grid consists of core regions only as no overlap exists Hence by con necting MESSy to the COSMO model the MESSy in frastructure was extended to take into account the differ ences between core region and ghost boundaries This is essential for instance for the calculation of domain integrated tracer masses see Sect 3 7 Next those interface or generic submodels are described which have undergone extensions or modifications while being connected to COSMO MESSy A list of all other www geosci model dev net 5 87 2012 A Kerkweg and P Jockel COSMO MESSy messy_setup gt TIMER SWITCH CHANNEL QTIMER gt TIMER DATA CHANNEL TRACER IMPORT CVTRANS DRADON DDEP JVAL messy_initialize LNOX MECCA OFFEMIS ONEMIS PTRAC SCAV SEDI TNUDGE TREXP gt TRACER MECCA DRADON PTRAC SCAV TREXP messy_new_tracer gt QTIMER CHANNEL TRACER DATA IMPORT CVTRANS DRADON
6. km pressure hPa 10 W 0 10 E 20 F 30 E 40 E 50 F 23 March 2010 200 Q altitude km altitude km 105 6 March 2010 11 March 2010 200 200 L10 H10 T a T 300 300 g ge E a 400 3 p 400 2 a 8 pr v Dp l D 500p Ls 500 Ls 600 600 700 700 800 800 900 900 1000 Lo 1000 Lo 10 W O 10 E 20 E 30 E 40 E S5SO0 E 10 W O 10 E 20 E 30 E 40 E S5SO E RN222 1029 mol mol RN222 1029 mol mol Fig 11 Vertical cross section of Rn 200 J pressure hPa TOPE 20 E 30 E 40 F 50 E 10 W 0 10 E 20 E 30 E 40 E 50 F 19 March 2010 23 March 2010 5 3 200 2 10 10 1 aa amp T 300 amp 0 5 o p v 0 3 3 2 400 3 8 2 0 2 QO a i L 5 500 5 0 1 600 0 075 700 0 05 800 900 0 025 8 Q 10 W 0 10 E 20 E 30 E 40 E 5O E TOPE 20 E 30 E 40 E 5O E RN222 1029 mol mol RN222 107 mol mol mixing ratios at 50 N Snapshots are taken at 6 11 19 23 March respectively Upper row result of the ECHAMS MESSy simulation middle row COSMO 40 MESSy simulation lower row COSMO 7 MESSy simulation the COSMO MESSy models the highest updrafts are lo cated between 10 W to O W and 20 E to 32 E the ECHAMS MESSy simulation shows a prominent updraft around 5 E This is most probably due to the different trigger mechanisms employed in the respective convection schemes On 11 March a bubble o
7. the same geometry in space and time From the rank 6 tracer field rank 1 2 and 4 span the spatial dimensions rank 3 is the tracer index 1 e the number of an in dividual tracer in the tracer set rank 5 is of length 1 and rank 6 contains the data instances e g the different time levels of a time integration scheme For a detailed overview of the submodel TRACER we refer to Jockel et al 2008 Here we provide details about the implementation of the TRACER interface into the COSMO model One of the striking differences between the COSMO model and ECHAMS is the treatment of the prognostic vari ables see Sect 3 4 In the COSMO model prognostic vari ables are allocated with an extra rank for the different time levels required by the time integration scheme The indices of the respective time levels are nnow nnew and nold which are rotated each model time step in order to avoid the copying of the data from the nnew to the nnow time level and so forth Therefore the data for one specific time level is not always located at the same memory space This is different in the MESSy Due to the association of the tracers to the channel memory management each of the instances rank 6 of the tracer field is permanently associated to one specific time level of the time integration scheme Table 1 shows the definition of the instances for COSMO MESSy as currently chosen The number of instances depends on
8. 2007 2010 the chemistry in and near ship tracks Lauer et al 2007 Franke et al 2009 the role of organic species Pozzer et al 2007 2010 the representation of the so lar cycle Nissen et al 2007 detailed aerosol and halogen chemistry Kerkweg et al 2007 2008a b emission of NOx from soil Steinkamp et al 2009 bio aerosols Burrows et al 2009 energetic particle precipitation Baumgaertner et al 2009 2010 and comparisons with satellite measure ments Stiller et al 2007 Br hl et al 2007 Wetzel et al 2010 Additionally EMAC took part in a number of impor tant model inter comparison studies e g Austin et al 2010 Butchart et al 2011 Gettelman et al 2010 Hegglin et al 2010 Morgenstern et al 2010 Son et al 2010 Here an overview of the functionality of COSMO MESSy is provided First the COSMO model and the MESSy inter face structure are briefly recapitulated in Sects 2 1 and 2 2 respectively Section 3 describes the connection of the MESSy infrastructure submodels to COSMO and Sect 4 de picts the required generalisation of the regular submodels A first application of some previously published sub models focusing on the evaluation of the functionality of the TRACER interface in COSMO MESSy and the tracer transport characteristics is presented in Sect 5 before the po tential of the newly built model system is shortly summarised in Sect 6 This is the first part of three c
9. COSMO 7 region For H and V1 all lines are on top of each other them locally to zero and to diagnose the corresponding mass conservation violation This has been applied here Figure 5 displays the integrated over the model domain negative tracer masses for the H V1 and V2 tracers which have been corrected by TRACER_PDEF The H and the V1 tracers do not become negative whereas for V2 advection produces small negative tracer mixing ratios at the beginning of the simulation after 10 days no more negative values are produced The fact that V2 is initialised with zero at the bot tom and that the tracer is well mixed in the lower layers after 10 days indicates that the negative values origin in the sur face layer but only 1f very small mixing ratios prevail Even then the negative mass produced for V2 is 7 orders of mag nitude smaller compared to the integrated over the domain mass of the tracer which is in the order of 10 kg Conservation of mass Since the air mass in the regional model domain is not con served e g when a low or high pressure system is moving across the model domain the tracer mass variation within the model domain must be related to the air mass variation For tracer mixing ratios in units of mol mol dry air the dry air mass is the appropriate measure Figure 6 left shows the dry air and tracer masses of H and V1 plus V2 in tegrated over the model domain normalised to their corre sponding time
10. COSMO arrays are replaced by the POINTER attribute The COSMO model arrays are allocated by creating the re spective channel objects in a channel called COSMO_ORT in the subroutine messy_COSMO_create_channel in stead of being allocated in the COSMO subroutine alloc_meteofields compare flow chart in Fig 3 For instance the 3 D field of the density of the reference atmo sphere rho0 is defined by CALL new_channel_object status amp i COSMO ORT A RHO amp ps rho0 reprid GP_3D_MID j CALL channel _halt callstr status RHOO is the name of the channel object The memory space required for the rank 3 POINTER rho0 is indicated by the representation ID GP__3D_MID which determines that the variable is defined at grid mid points The CHARAC TER formal parameter cal 1st r denotes the calling subrou tine Detailed information is provided in the CHANNEL user manual available in the supplement of J6ckel et al 2010 The channel COSMO_ORT 1s used for data management only 1 e its contents should normally not be subject to data output but the objects can be written if required for debugging pur poses CHANNEL provides its own output control Within the channel namelist channel nml defaults for each chan nel and channel object can be set to create output e g for instantaneous values averages standard deviations mini mum and maximum values etc see CHANNEL manual supplement of J ckel et al 2010 To keep the functio
11. PNT the emission point source of the first tracer is located at 20 00 W and 50 00 N in a pressure alti tude of 900 hPa The tracer is named PNT as abbrevi ation for PoiNT source This location is chosen to be outside but close to the COSMO 40 model domain Therefore the tracer needs to be transported into both COSMO model domains 2 Tracer VOL the second tracer is emitted at a pressure altitude of 800 hPa over Island at 19 60 W 63 63 N It is named VOL as its location coincides with the volcano Eyjafjallaj kull This point resides in the COSMO 40 but not in the COSMO 7 model domain Thus VOL is in contrast to PNT emitted within the www geosci model dev net 5 87 2012 larger COSMO model domain but is also advected into the COSMO 7 domain Figures 7 and 8 display snapshots of the 12th 15th and 18th simulation day at 12 00UTC and at the 900hPa or 800 hPa pressure level for the PNT and VOL tracer respec tively Pictures in one column correspond to the same simula tion day The first row displays the tracer distribution as sim ulated by ECHAMS MESSy the second and the third row show the tracer distributions in the larger i e 40 km and the smaller i e 7 km COSMO model domain respectively while the last row depicts a composite of all three simula tions To investigate the transport of the tracers into the re gional model domain the PNT tracer is emitted outside of both COSMO model domains Comparing the
12. appears for the tracers experiencing no transport at all V Int V2nt or only verti cal diffusion or convective transport V 1v V2v Vic V2c as shown in Fig 6 middle and left panel The amplitudes are smaller however since they are not or only vertically re distributed and the only contribution to the in and outflux is the Newtonian relaxation to the boundary condition at the domain boundary Although the analysis so far does not reveal any indica tion for potential mass conservation violations they cannot be entirely ruled out a mass conservation violation caused by mass wind inconsistencies of the advection operator de pends on the vertical tracer z axis gradients J6ckel et al 2001 Since in our case V1 V2 H throughout the do main at any time it follows dV 1 dz dV2 dz because dH dz 0O As a consequence spurious mass changes of V 1 and V2 due to mass wind inconsistencies would be of the same absolute value but of opposite sign As a consequence they would cancel out and in our analysis therefore be indistinguishable from in and outfluxes across the domain boundaries In summary convective transport and vertical diffusion are positive definite monotone mass conserving www geosci model dev net 5 87 2012 A Kerkweg and P Jockel COSMO MESSy dry oir H V1 V2 28 4 101 gt 10 15 20 25 30 MAR 29 TTT ier ir pr bod 95 Ree 5 10 15 20 25 30 5 10 15 25 30 MAR MAR
13. are see also Sect 3 among others the memory management and input output interface including a restart facility CHANNEL J ckel et al 2010 the tracer infras tructure for the handling of constituents TRACER J ckel et al 2008 or the time and event management TIMER J ckel et al 2010 For a more detailed overview of the cur rently available generic and regular submodels we refer to J ckel et al 2010 and the project web page The next sec tion Sect 3 describes the implementation details of the con nection of the MESSy infrastructure to the COSMO model Section 4 documents the required generalisation i e includ ing an adapter into the plug of the SMIL of the regular sub models The necessity for such a generalisation arose mainly from the fact that in the Eulerian grid point formulation of many processes the vertical direction needs to be particularly distinguished in other words plays a particular role from the horizontal directions As a consequence the rank of the ver tical index in multi dimensional array variables representing geo located data might depend on the basemodel and indeed it is different between ECHAMS and COSMO This depen dency has been bypassed 3 Implementation of the MESSy infrastructure An important principle for the implementation of the MESSy interface in the COSMO model is to keep the changes in the COSMO model code as small as possible 1 e minimally invasive Additionally
14. averages Due to this normalisation the mass variations of the tracers and of the dry air mass are directly comparable The lines for the variation of the dry air mass and the homogeneous tracer are on top of each other imply ing that the transport operators are strictly mass conserving for homogeneously distributed tracers Moreover the sum of masses of VI and V2 follows exactly the same varia tion Indeed the deviation of V 1 V2 from the initial con dition which is equivalent to H is negligibly small less than 0 02 c throughout the domain at any time during the Geosci Model Dev 5 87 110 2012 simulation not shown except for the very beginning where the negative masses occur see above As conclusion the transport operators are strictly linear The invariance of this linearity relation requires that each change of the mass per mass of dry air of V1 needs to be counterbalanced by a corresponding change of V2 with op posite sign This is indeed the case as is visible in the middle and right panel of Fig 6 the individual tracer masses nor malised to their time average and divided by the mass of dry air normalised to its time average of V 1 and V2 show a com plementary variation As important implication this must also comprise the in and outfluxes across the domain bound aries where in our setup the linearity relation V1 V2 H also holds throughout the simulation not shown This complementary variation also
15. box Blue text highlights the generic submodels whereas the regular submodels are written in black In those entry points located within the outer loop over the second hori zontal dimension in EMAC indicated by the green box 1 e messy_vdiff messy_convec messy_physc and messy_local_end the loop is mimicked with loop vari able jrow in agreement with the naming convention At first place in these loops some rank 2 POINTERs are asso ciated with the respective 2 D sub arrays of 3 D fields by calling the subroutine main_data_2D_set_jrow SUBROUTINE main_data_2D_set_jJrow jrow INTEGER INTENT IN Jrow ate gt qvtens l ie jrow END SUBROUTINE main_data_2D_set_jrow Figure 3 depicts a flow chart of the COSMO MESSy model It illustrates where the individual MESSy entry points are located in the COSMO basemodel flow The MESSy entry points are highlighted by yellow boxes whereas the COSMO model routines are coloured in blue COSMO model routines which are obsolete for COSMO MESSy are crossed out and substituted by the corresponding MESSy subroutine in the dark orange box directly below The light orange boxes indicate direct calls of generic submodel sub routines The COSMO MESSy Implementation Documen tation in the Supplement lists and explains all changes which became necessary in the COSMO model code for the implementation of the MESSy interface 3 2 CHANNEL the memory management output and restart control Th
16. in black The red and the blue lines show the maximum mixing ratios of the COSMO 40 and COSMO 7 domains respectively structure of the 7 Rn distribution is mainly driven by con vective transport Due to the trigger mechanisms of con vection parameterisations and due to the different grid box sizes the distribution is expected to differ in some places It should be noted that the ECHAMS5 MESSy and the COSMO MESSy models use convection schemes based on the Tiedtke scheme Tiedtke 1989 but includ ing different further developments On 6 March the con vective transport of the COSMO MESSy simulations is very similar with COSMO 7 MESSy showing some more individual updrafts than COSMO 40 MESSy While in www geosci model dev net 5 87 2012 A Kerkweg and P Jockel COSMO MESSy pressure hPa Q TOPE 20 E 30 E 40 F 50 E 10 w 0 6 March 2010 200 altitude km pressure hPa 10 W 0 10 E 20 F 30 E 40 E 50 F 11 March 2010 200 altitude km pressure hPa 300 300 E g gE E p 400 2 p 400 2 4 oe z 500 5 500 5 600 600 700 700 800 900 1000 0 10 W 0 10 E 20 F 30 E 40 F 50 E 800 900 1000 Q 10 W 0 10 E 20 E 30 E 40 E 5O E 10 W 0 10 E 20 E 30 E 40 E 5O E 19 March 2010 Gn Oo Oo 400 pressure hPa mom Oo G O O 700 800 900 1000 0 10 w 0 altitude km pressure hPa altitude
17. into an interface and a core layer part The MESSy infrastructure is coded as so called generic submodels which interfaces reside in the BMIL with COSMO members However all users are required to sign an agreement with a COSMO national meteorological service and to respect certain conditions and restrictions on code usage For questions concerning the request and the agreement please con tact the chairman of the COSMO Steering Committee In the case of a planned operational or commercial use of the COSMO Model package special regulations will apply cited from the COSMO User s Guide December 2009 http www cosmo model org content model documentation core cosmoUserGuide pdf 6 http www unidata ucar edu software netcdf 7 Appendix A contains a Glossary explaining some terms re peatedly used here The terms from the Glossary are written in italics throughout the article www geosci model dev net 5 87 2012 89 Base Model Layer power supply Base Model Interface Layer multiple socket outlet Submodel Interface Layer connector Submodel Core Layer the machinery Fig 1 Sketch of the four layer structure of MESSy whereas the interfaces for the regular submodels belong to the SMIL The core i e the basemodel independent parts of all submodels is located in the SMCL Examples of generic submodels
18. mass fluxes which have to be delivered by the applied convection scheme Currently in Geosci Model Dev 5 87 110 2012 A Kerkweg and P Jockel COSMO MESSy COSMO MESSy this data delivery is only implemented for the Tiedtke scheme Tiedtke 1989 therefore CVTRANS is presently only applicable in combination with the Tiedtke scheme 5 2 Tracer transport tests To verify the functionality of the TRACER interface in COSMO MESSy and to evaluate the tracer transport char acteristics different tests have been performed The model was used in the on line coupled mode as described in Kerk weg and J ckel 2012 to ensure consistent boundary condi tions for the tracers All tests use either artificial tracers or 22Rn Moisture in the COSMO model is transported by ad vection horizontal and vertical diffusion As the transport of tracers should be consistent with the moisture transport the COSMO transport routines are adopted for tracer transport in COSMO MESSy Even though horizontal diffusion of tracers is implemented in COSMO MESSy all tests showed that the numerical diffusion is larger than the calculated horizontal diffusion Thus we recommend to neglect the explicit calculation of horizontal diffusion and show no results for horizontal tracer diffusion here In addition to the transport processes of the basemodel the MESSy submodel CVTRANS Tost et al 2010 for the convective tracer transport is included The COSMO MESSy
19. meteorological boundary conditions The new de veloped model is as consistent as possible with respect to atmospheric chemistry and related processes with a previ ously developed global atmospheric chemistry general circu lation model the ECHAM MESSy Atmospheric Chemistry EMAC model The combined system constitutes a new re search tool bridging the global to the meso y scale for atmo spheric chemistry research MESSy provides the infrastruc ture and includes among others the process and diagnostic submodels for atmospheric chemistry simulations Further more MESSy is highly flexible allowing model setups with tailor made complexity depending on the scientific ques tion Here the connection of the MESSy infrastructure to the COSMO model is documented and also the code changes required for the generalisation of regular MESSy submod els Moreover previously published prototype submodels for simplified tracer studies are generalised to be plugged in and used in the global and the limited area model They are used to evaluate the TRACER interface implementation in the new COSMO MESSy model system and the tracer transport characteristics an important prerequisite for future atmospheric chemistry applications A supplementary doc ument with further details on the technical implementation of the MESSy interface into COSMO with a complete list of modifications to the COSMO code is provided 1 Introduction Since atmospheric chemi
20. model domain for the evaluation is located over Central Europe using a horizontal grid of ap proximately 40 km 0 36 mesh size and 40 vertical levels A second smaller model domain is located roughly over Ger many utilising a horizontal grid of approx 7 km 0 0625 and the same 40 vertical levels as in the 40km COSMO domain The simulations will be called COSMO 40 and COSMO 7 hereafter ECHAMS MESSy in T106L31 resolu tion provides the initial and boundary data for the COSMO 40 simulation and COSMO 40 the initial and boundary con ditions for the COSMO 7 simulation The simulated period starts at 1 March 2010 As these are simple tracer transport tests ECHAMS MESSy was not nudged to the analysed me teorology We first show transport tests for tracers which are ini tialised but have neither sources nor sinks Sect 5 2 1 con tinue with tracers which are initialised with zero and emitted from a point source Sect 5 2 2 and conclude with a simu lation for 7 Rn Radon For all these tests the Runge Kutta time integration and a Semi Lagrangian advection scheme are used 5 2 1 Tests with artificial passive tracers Tracer transport in COSMO MESSy is operator split into four different processes advection convective transport vertical and horizontal diffusion For the reason given above we do hereafter not consider horizontal diffusion For at mospheric chemistry simulations each of these transport www geosci model dev net
21. nothing in the functionality of the COSMO model should be changed for the usual COSMO only user Hence all changes have been introduced using the pre processor directive MESSY Shttp www messy interface org Geosci Model Dev 5 87 110 2012 90 ifdef MESSY new code tendif Or ifndef MESSY original COSMO code else modified code endif Thus the changes are only active if the model is config ured with enable MESSY default Otherwise with disable MESSY the original COSMO model code is compiled One of the basic ideas of MESSy is to keep as much code as possible applicable for all models employed In this way code doubling is avoided making it highly consistent and less error prone This is not always easily achieved as for 3 D applications MESSy currently still depends on a basemodel providing the model domain decomposition the dynamics and other distinct parts of the model physics The global climate model ECHAMS Roeckner et al 2006 was the basemodel chosen as basis for the development of MESSy Jockel et al 2006 Before COSMO MESSy was built ECHAMS5 was the only 3 D basemodel hosting the MESSy interface Therefore MESSy uses mostly the nam ing convention of ECHAMS For instance the names of variables used in the MESSy SMIL are in many cases the same as in ECHAMS As long as MESSy was connected to one 3 D basemodel only there was no need to take dif ferent realisations of 3 D model
22. results of the global and the COSMO 40 MESSy model simula tion PNT is advected correctly into the regional model domain Obviously details of the tracer distribution are much better resolved in the COSMO MESSy model than in the ECHAMS MESSy simulation These findings are also valid for the comparison of COSMO 40 MESSy to COSMO 7 MESSy Figure 9 displays the time series of maximum PNT mixing ratios within a rectangle ranging from 5 W to 25 E and 36 to 56 N at three pressure levels 850 900 and 950 hPa This region is part of all three model do mains A two peak structure is exhibited at all heights in all model domains The peaks coincide with the dynamical patterns when streamers of the PNT tracer are advected into the rectangular region The lines indicating the maxima as simulated by the COSMO MESSy models are very similar In contrast the maxima of PNT in ECHAMS MESSy coin cide with the other two models at 900 and 950 hPa while they are especially for the two peaks lower at the 850 hPa Geosci Model Dev 5 87 110 2012 102 TIME 12 MAR 2010 12 00 Z 900 hPa 40 w 20 W TIME 15 MAR 2010 12 00 Z 900 hPa A Kerkweg and P Jockel COSMO MESSy TIME 18 MAR 2010 12 00 Z 900 hPa fp amp MO T T T T e 40 E 60 E 40 W 20 W O 20 E 40 E GO E BPE 40 W 20 W 0 Aha T 40 E 60
23. specific order of dimensions in ECHAMS h1 v h2 has been introduced for code optimisa tion Within a loop over the second horizontal dimen sion outer loop arrays of rank 2 h1 v are processed This reduces memory consumption as the local vari ables are smaller by one rank and loops over the re maining horizontal dimension allow a configurable op timised memory access blocking on both vector and scalar architectures cache optimisation Most of the MESSy submodels are called within this outer loop in ECHAMS MESSy and make use of the reduced vari ables and the blocking feature To keep the MESSy sub model interfaces as general as possible and to be able to use the same interfaces for ECHAMS and COSMO we imitate the outer loop within COSMO MESSy which has in the present implementation no blocking capa bility This is discussed in further detail in Sect 3 1 3 The horizontal domain decomposition for parallel pro cessing In ECHAMS the global Eulerian grid used for physics except for advection which has its own parallel decom position is distributed bijectively in both horizontal di rections onto all parallel process entities PEs 1 e each grid point resides on one PE only This is different for the COSMO model grid The local horizontal grid on each PE is split into a core region and a ghost boundary or halo part Each grid point of the full model domain is bijectively attributed to one core region of the de
24. stratospheric chemistry climate models and associated external forcings J Geophys Res 115 www geosci model dev net 5 87 2012 109 DOOMO2 doi 10 1029 2009JD013728 2010 Nissen K M Matthes K Langematz U and Mayer B Towards a better representation of the solar cycle in general circulation models Atmos Chem Phys 7 5391 5400 doi 10 5194 acp 7 5391 2007 2007 Pozzer A Jockel P Tost H Sander R Ganzeveld L Kerk weg A and Lelieveld J Simulating organic species with the global atmospheric chemistry general circulation model ECHAM5 MESSy1 a comparison of model results with obser vations Atmos Chem Phys 7 2527 2550 doi 10 5194 acp 7 2527 2007 2007 Pozzer A Pollmann J Taraborrelli D J ckel P Helmig D Tans P Hueber J and Lelieveld J Observed and simulated global distribution and budget of atmospheric C2 C5 alkanes Atmos Chem Phys 10 4403 4422 doi 10 5 194 acp 10 4403 2010 2010 Pozzer A J ckel P Kern B and Haak H The Atmosphere Ocean General Circulation Model EMAC MPIOM Geosci Model Dev 4 771 784 doi 10 5194 gmd 4 771 2011 2011 Rockel B Will A and Hense A The Regional Climate Model COSMO CLM CCLM Meteorol Z 17 347 348 2008 Roeckner E Brokopf R Esch M Giorgetta M Hagemann S Kornblueh L Manzini E Schlese U and Schulzweida U Sensitivity of simulated climate to horizontal and vertica
25. subroutines named as in ECHAMS5 MESSy and struc tured where necessary either as the ECHAMS MESSy or the COSMO model routines respectively In particular it contains the following subroutines interfaces gather_gp this is a threefold overloaded interface for gathering 4 D 3 D or 2 D fields in grid point space The subroutines for 3 D or 4 D fields reduce the field by one rank and call the next lower sub routine thus gather_gp basically performs an ex change of parallel decomposed horizontal 2 D fields In the COSMO model this is accomplished by the subrou tine gather_field Consequently this subroutine is called from the subroutine gather_gp to actually perform the data transfer scatter_gp this subroutine provides the reverse action to gather_gp It distributes fields to a par allel decomposed grid Equivalently overloaded scatter_gp basically calls the COSMO model sub routine distribute_field p_bcast the subroutine p_bcast is twelvefold overloaded for the transfer of OD and 1D LOGI CAL CHARACTER LEN and different KINDs of www geosci model dev net 5 87 2012 95 INTEGER or REAL variables It is a renamed and expanded version of the COSMO model interface distribute_values which requires six parame ters the buffer to be sent received the rank of the task in the group communicator the group communicator the buffer length the MPI datatype and an error sta tus The corresponding MESSy r
26. 2 00 Z 800 hPa 103 TIME 18 MAR 2010 12 00 Z 800 hPa Fig 8 Horizontal distribution at 800 hPa of the artificial tracer VOL The location of the emission point is indicated by the light blue plus sign Results are shown for the 12th 15th and 18th simulation day at 12 00 UTC columns First row ECHAMS5 MESSy second row COSMO 40 MESSy third row COSMO 7 MESSy and last row composite of all three model domains of very different volumes leading to a higher degree of di lution and thus smaller mixing ratios in ECHAMS MESSy Thus the VOL distribution in COSMO 40 MESSy is more compact and shows higher peak values As the emission point is not located in the COSMO 7 domain the tracer VOL is only advected into this model domain Because www geosci model dev net 5 87 2012 the lateral boundary data are interpolated from the COSMO 40 MESSy simulation the maximum mixing ratios of the tracer VOL in COSMO 7 MESSy reflect mostly the maxima of the COSMO 40 MESSy simulation Geosci Model Dev 5 87 110 2012 104 Ho0 hPa 40 sy D nmel mgl ba cm 5 Mar 10 MAR 20 MAR 3G 900 hPa 40 SA D nmel mgl ba cm m Mar 10 MAR 20 MAR 3G Z oU hPa 40 a r nmel raol ba cm mo Mor 10 Mar 20 MAR SG Fig 9 Maximum mixing ratio nmol mol of tracer PNT The maximum is calculated in a rectangle ranging from 5 W to 25 E and 36 to 56 N The mixing ratios are disp
27. 20 Fig 6 Left integrated dry air and tracer masses normalised to their time averages see text for dry air the homogeneous tracers H and V1 V2 The lines are on top of each other Middle and right panel tracer masses normalised to their time average and divided by the mass of dry air normalised to its time average for the passive tracers V1 middle and V2 right All panels show results from the COSMO 7 simulation and linear The advection operator exhibits a small negli gible violation of the positive definiteness but only for very small mixing ratios in the lowest model layer Consequently the correction in the submodel TRACER _PDEF required for numerical stability results in these cases in a likewise negli gible mass conservation violation Other issues like poten tial mass wind inconsistencies could not be revealed with our tests Further studies including the budgeting of in and out flow across the domain boundaries will provide more details in future Mainly the linearity of the advection operators let us conclude that the tracer transport in COSMO MESSy is ready for atmospheric chemistry applications 5 2 2 Tracer tests utilising point sources As a second test case tracers have been initialised with zero and are emitted by a point source using the MESSy sub model TREXP see Sect 5 1 2 Two artificial tracers with out sinks emitted at different point sources have been cho sen for the experiment 1 Tracer
28. 4 acp 5 433 2005 2005 J ckel P Tost H Pozzer A Br hl C Buchholz J Ganzeveld L Hoor P Kerkweg A Lawrence M G Sander R Steil B Stiller G Tanarhte M Taraborrelli D van Aardenne J and Lelieveld J The atmospheric chem istry general circulation model ECHAM5 MESSy1 consistent simulation of ozone from the surface to the mesosphere At mos Chem Phys 6 5067 5104 doi 10 5194 acp 6 5067 2006 2006 J ckel P Kerkweg A Buchholz Dietsch J Tost H Sander R and Pozzer A Technical Note Coupling of chemical processes with the Modular Earth Submodel System MESSy submodel TRACER Atmos Chem Phys 8 1677 1687 doi 10 5194 acp 8 1677 2008 2008 J ckel P Kerkweg A Pozzer A Sander R Tost H Riede H Baumgaertner A Gromov S and Kern B Development cycle 2 of the Modular Earth Submodel System MESSy2 Geosci Model Dev 3 717 752 doi 10 5194 gmd 3 717 2010 2010 Kerkweg A and J ckel P The 1 way on line coupled atmospheric chemistry model system MECO n Part 2 On line coupling with the Multi Model Driver MMD Geosci Model Dev 5 111 128 doi 10 5194 gmd 5 111 2012 2012 Kerkweg A Sander R Tost H J6ckel P and Lelieveld J Technical Note Simulation of detailed aerosol chemistry on the global scale using MECCA AERO Atmos Chem Phys 7 2973 2985 doi 10 5194 acp 7 2973 2007 2007 Kerkweg A J ckel P Pozzer A T
29. 5 87 2012 A Kerkweg and P Jockel COSMO MESSy Table 2 Definition of initialisation patterns for passive tracers used in this study Name _ Initialisation pattern H Homogeneously 107 mol mol Vi Vertical gradient decreasing with height V2 Vertical gradient increasing with height 0 95 0 9 200 i 0 75 _ 0 7 O 0 65 300 ae v E 5 V n 400 610 45 s 0 4 500 0 35 0 3 600 0 25 700 800 01 900 0 05 1000 0 10 S 0 10 N 20 N Vi nmol mol Fig 4 Initialisation pattern of the passive tracer V 1 The horizontal axis shows rotated coordinates operators needs to be mass conserving positive definite and monotone Inside a regional model domain the mass of a specific tracer is not expected to be conserved With a per fect transport scheme however the mass budget of pas sive tracers 1 e without sources or sinks in the regional domain is expected to be closed implying that the tracer mass within the domain plus inflow minus outflow in out of the domain is conserved The latter inflow and outflow are determined at least implicitly by the boundary condi tions As there is no in and outflux budgeting routine in COSMO MESSy yet our analyses are somewhat limited but as a first step artificial passive tracers are used to diagnose to the extent possible the mass conservation positive definite ness and monotonicity of the transport processes as imple mented in COSMO MESSy Table 2 lis
30. CIOOCI IR absorption cross sections At mos Chem Phys 10 931 945 doi 10 5194 acp 10 93 1 2010 2010 www geosci model dev net 5 87 2012
31. E BPE 40 Ww 20 W 0 20 E 40 E 60E 80E Fig 7 Horizontal distribution at 900 hPa of the artificial tracer PNT The location of the emission point is indicated by the light blue plus sign Results are shown for the 12th 15th and 18th simulation day at 12 00 UTC columns First row ECHAM5 MESSy second row COSMO 40 MESSy third row COSMO 7 MESSy and last row composite of all three model domains level This is most probably due to less upward transport in the global model The VOL tracer is emitted at a point located in the COSMO 40 model domain As for PNT the more de tailed structures resolved on the finer grids are clearly visible in Fig 8 Here the discrepancies between the ECHAMS MESSy and the COSMO 40 MESSy simulations are larger as for PNT The VOL mixing ratio is nearly Geosci Model Dev 5 87 110 2012 always larger in the COSMO 40 MESSy compared to the ECHAMS MESSy simulation This finding is affirmed by Fig 10 which displays the maximum mixing ratios of the tracer VOL in a similar way as Fig 9 but for the 750 800 and 850 hPa pressure levels The main reason for these dif ferences is the parameterisation of the point source emission in the MESSy submodel TREXP As the emitted tracer mass is prescribed the same tracer mass is emitted into grid boxes www geosci model dev net 5 87 2012 A Kerkweg and P Jockel COSMO MESSy TIME 12 MAR 2010 12 00 Z 800 hPa TIME 15 MAR 2010 1
32. Geosci Model Dev 5 87 110 2012 www geosci model dev net 5 87 2012 doi 10 5194 gmd 5 87 2012 Author s 2012 CC Attribution 3 0 License C Geoscientific G Model Development The 1 way on line coupled atmospheric chemistry model system MECO n Part 1 Description of the limited area atmospheric chemistry model COSMO MESSy A Kerkweg and P J ckel Institute for Atmospheric Physics University of Mainz Germany 2Max Planck Institut fiir Chemie Abteilung Luftchemie 55128 Mainz Germany now at Deutsches Zentrum fiir Luft und Raumfahrt DLR Institut fiir Physik der Atmosph re 82234 Oberpfaffenhofen Germany Correspondence to A Kerkweg kerkweg uni mainz de Received 23 May 2011 Published in Geosci Model Dev Discuss 21 June 2011 Revised 31 October 2011 Accepted 15 November 2011 Published 19 January 2012 Abstract The numerical weather prediction model of the Consortium for Small Scale Modelling COSMO main tained by the German weather service DWD is connected with the Modular Earth Submodel System MESSy This ef fort is undertaken in preparation of a new limited area atmo spheric chemistry model Limited area models require lat eral boundary conditions for all prognostic variables There fore the quality of a regional chemistry model is expected to improve if boundary conditions for the chemical con stituents are provided by the driving model in consistence with the
33. ISA Distributed European Infrastructure for Supercomputer Applications Extreme Computing Initiative DECI we therefore thank the DEISA Consortium www deisa eu co funded through the EU FP6 project RI 031513 and the FP7 project RI 222919 We have used the Ferret program http www ferret noaa gov from NOAA s Pacific Marine Environmental Laboratory for creating some of the graphics in this paper We thank the Editor V Grewe and two ananymous referees for their constructive comments Edited by V Grewe References Austin J Struthers H Scinocca J Plummer D A Akiyoshi H Baumgaertner A J G Bekki S Bodeker G E Braesicke P Br hl C Butchart N Chipperfield M P Cugnet D Dameris M Dhomse S Frith S Garny H Gettelman A Hardiman S C J ckel P Kinnison D Kubin A Lamar que J F Langematz U Mancini E Marchand M Mi chou M Morgenstern O Nakamura T Nielsen J E Pitari G Pyle J Rozanov E Shepherd T G Shibata K Smale D Teyss dre H and Yamashita Y Chemistry climate model simulations of spring Antarctic ozone J Geophys Res 115D DOOM11 doi 10 1029 2009JD013577 2010 Baumgaertner A J G Jockel P and Br hl C Energetic particle precipitation in ECHAMS MESSy1 Part 1 Downward trans port of upper atmospheric NOx produced by low energy elec trons Atmos Chem Phys 9 2729 2740 doi 10 5194 acp 9 2729 2009 2009
34. P Jockel COSMO MESSy 2 Model description The model system COSMO MESSy consists of two model components the numerical weather prediction model of the Consortium for Small Scale Modelling COSMO model Doms and Schattler 1999 and the Modular Earth Submodel System MESSy J ckel et al 2005 Both components are briefly introduced Currently the model system is based on the COSMO model version cosmo_4 8_clm12 and the MESSy version 2 41 2 1 The COSMO model The COSMO model is a non hydrostatic limited area weather prediction model Steppeler et al 2003 Doms and Schattler 1999 It has been designed for both operational numerical weather prediction NWP and for research on the meso and meso y scale Additionally to the short range NWP application the COSMO model was set up for regional climate studies by the CLM community Rockel et al 2008 The COSMO model is based on the primi tive thermo hydrodynamical equations for compressible flow in a moist atmosphere and uses rotated geographical coordi nates and a generalised terrain following height coordinate The Arakawa C grid with Lorenz vertical grid staggering is used The COSMO model treats the following prognos tic variables horizontal and vertical Cartesian wind compo nents pressure perturbation temperature specific humidity and cloud water content There are additional optional prog nostic variables such as the cloud ice content the turbulent kinetic energy the
35. Submodel System MESSy J ckel et al 2005 2010 is a multi institutional project providing the in frastructure to expand state of the art geoscientific domain models e g of the atmosphere into comprehensive Earth System Models ESMs Although originally the focus was solely on the efficient and flexible implementation of pro cesses related to atmospheric chemistry into a circulation model of the atmosphere the methodology turned out to be much more powerful e g Pozzer et al 2011 In partic ular MESSy has been connected to the ECHAMS general circulation model of the atmosphere Roeckner et al 2006 expanding it into the global chemistry climate model EMAC J ckel et al 2006 with a wide variety of applications see also Sect 1 The basic idea behind MESSy is its four layer structure as visualised in Fig 1 consisting of the basemodel layer BML usually a legacy domain model e g of the atmosphere it can be regarded as power supply the basemodel interface layer BMIL which hosts the MESSy infrastructure and can be regarded as multiple socket outlet the submodel interface layer SMIL which provides the communication of the submodels with the MESSy infrastructure and therefore also with the basemodel it can be regarded as the plug the submodel core layer SMCL which hosts the actual process or diagnostic formulation it can be regarded as the machinery Each MESSy submodel is split
36. The transport of the tracers via advection and horizontal and vertical diffusion is included into the COSMO model in equivalence to the transport of water vapour qv Due to the meta information associated to each tracer each transport process can independently be switched on or off for each in dividual tracer Further information are provided within the comprehensive description of the generic MESSy submodel TRACER by J ckel et al 2008 4 Implementation of the MESSy submodels A MESSy submodel consists of two layers J ckel et al 2005 see also Sect 2 2 The submodel core layer SMCL hosts the code parts describing the physical chemical or diagnostic process dealt with by the submodel Because the SMCL con tains only the scientific content of a submodel it is com pletely independent of the basemodel global regional column or box model and therefore stays unchanged for the implementation of MESSy into the COSMO model www geosci model dev net 5 87 2012 A Kerkweg and P Jockel COSMO MESSy 97 Table 1 Definition of tracer field instances in COSMO MESSy The middle column lists the variable names of the respective fields in TRACER The abbreviations RK and LF denote the Runge Kutta and Leap frog scheme respectively Tracer field corresponding to time level nnew Tracer field corresponding to time level nnow for RK or nold for LF Two time levels of the tracer field boundary data Instance TRACER Meaning
37. by the MESSy sub models which are not directly supplied by the base model In DATA additional channel objects are defined and calculated during the integration phase if they vary with time or in the initialisation phase if they are con stant Examples for such variables are the geopotential at full and interface levels geopot and geopoti re spectively the mass contained in a grid box grmass or the volume of a grid box grvol1 3 5 MPI a high level interface to the MPI library As MESSy must be applicable in a parallel decomposed en vironment the generic MESSy submodel MPI builds a high level application interface API for the use of the MPI library Currently for each basemodel the access to the MPI library is mirrored in the MESSy submodel MPI In case of ECHAMS MESSy the ECHAMS routines are sim ply USEd into the generic submodel MPI Consequently for the COSMO model two aspects had to be taken into account Each basemodel uses the MPI library in its own way thus COSMO MESSy uses the MPI library for the MESSy submodels similar to the COSMO model The names of the high level routines organising the data exchange via the MPI library in the generic submodel MPI must be the same for different basemodels As the MESSy submodels call the MPI routines from their SMIL using the original ECHAMS names these names must also be used for COSMO MESSy To achieve this goal the generic MESSy submodel MPI con tains
38. e e g the specific hu midity in ECHAMS is determined by the two rank 3 variables qm1 and gte In the COSMO model the specific humidity field is the rank 4 variable qv The field after the last completed time step is accessed by qv nnow for the 2 time level scheme or by qv nold forthe leap frog scheme The cor responding specific humidity tendency field in COSMO is defined by qvtens To minimise the changes in the MESSy submodel in terface layer POINTERs named as the corresponding ECHAMS MESSy variables are associated to the cur rent time slice and or the tendency variable e g REAL DP POINTER PUBLIC amp DIMEN S LON Ss 5 454 gml gt NULL gm gt Oi 47 2434 MeEamMe qte_3d gt gtens where nt ime 1s nnow or nold depending on the time integration scheme Naturally the COSMO model uses different names for its loop and dimension variables and indices These names must be unique within the MESSy basemodel in terface layer and the submodel interface layer in order to make the submodel interface layer basemodel inde pendent Therefore DATA copies the COSMO model dimension variables to variables named as the respective MESSy variables The COSMO MESSy Implementa tion Documentation in the Supplement contains a list of the dimension variable names and their meaning www geosci model dev net 5 87 2012 A Kerkweg and P Jockel COSMO MESSy DATA also provides fields required
39. e MESSy code which were required for the implementation of the MESSy interface into the COSMO model are briefly listed in Appendix B while the details are provided in the Supplement The further developments and modifications are implemented for the full model system 1 e they either do not interfere with EMAC because they are not relevant for it or they are immediately available also for EMAC The code changes are entirely transparent for the COSMO model and EMAC users For instance the changes in the structure of MESSy required for the generalisation of the interface are all included in MESSy development cycle 2 as published by J ckel et al 2010 The changes in the COSMO model code are introduced via pre processor direc tives such that the standard COSMO code is compiled if the pre processor directive is not activated Due to the implemented generalisations it is possible to include the MESSy infrastructure into other models as well The feasibility to apply specific regular MESSy submodels depends however on the processes they describe and if they are applicable to the domain and scale of the basemodel For instance the global ocean model MPIOM implemented as a MESSy submodel by Pozzer et al 2011 is not straight forwardly applicable to the regional domain and thus not used in COSMO MESSy Similarly the application of the CVTRANS submodel for convective transport Tost et al 2010 is reasonable for COSMO MESSy studies
40. e generic MESSy submodel CHANNEL is described in detail by J6ckel et al 2010 It manages the memory usage the output the restart files and the model internal access to data Internally in CHANNEL all data are stored as concate nated list of structures each containing a description of the data meta information and a rank 4 POINTER for the data Geosci Model Dev 5 87 110 2012 92 A Kerkweg and P Jockel COSMO MESSy mog Section 1 model setup and namelist input morg Section 6 Time Stepping morg Section 2 Space allocation and constant fields morg Section 3 Input of first data sets morg Section 4 Initialisations sy INIT Coupling Fig 3 Simplified flow chart of COSMO MESSy Listed are the main routine calls of COSMO and the MESSy entry points Indented boxes are called by the non indented box above Blue boxes indicate COSMO routines yellow boxes highlight MESSy entry points CONTROL orange boxes point to direct calls of MESSy infrastructure submodels and dark orange boxes indicate direct calls of COSMO specific MESSy infrastructure submodel routines Geosci Model Dev 5 87 110 2012 www geosci model dev net 5 87 2012 A Kerkweg and P Jockel COSMO MESSy field itself To make the COSMO model data fields available to all MESSy submodels they have to be defined as channel objects As one prerequisite the TARGET ALLOCATABLE attribute of the
41. ev jrow define _RI_YVpl_ nlev 1 jrow endif ifdef COSMO define _RI YV jrow nlev define _RI_YVpl_ jrow nlev 1 endif where nlev is the number of vertical levels and 4 row is the loop variable of the second horizontal dimension indicated by the V and the Y in the rank identifier respectively The pre processor replaces the rank identifiers with the provided definition In the example above the code after running the pre processor reads dp l kproma amp pressi_3d 1 kproma nlevt l jrow amp pressi_3d 1 kproma nlev jrow for ECHAMS MESSy and dp l kproma amp pressi_3d 1 kproma jrow nlevtl amp pressi_3d 1 kproma jrow nlev for COSMO MESSy respectively Geosci Model Dev 5 87 110 2012 98 A detailed description of the naming convention of the rank identifiers and a list of the currently used rank iden tifiers is included in the COSMO MESSy Implementation Documentation in the Supplement 5 Tracer transport At the beginning of this section the previously published submodels for convective tracer transport CVTRANS simplified prognostic tracers PTRAC point sources and simplified chemistry TREXP and tracers of opportunity DRADON are shortly introduced before they are used for the evaluation of the functionality of the TRACER interface of COSMO MESSy and the tracer transport characteristics 5 1 Submodel description 5 1 1 PTRAC The MESSy submodel PTRAC Prognostic TRACer
42. f Radon formed in the free troposphere detached from the surface and the bound ary layer This bubble exists in all simulations but while the global model predicts an almost homogeneous distri bution within this bubble the COSMO MESSy simulations show much more structure inside On 19 March all sim ulations provide relatively similar well mixed Radon dis tributions despite the differences in the scales suggest ing that no major convective events took place during the last day A comparison of the panels for 23 March reveals www geosci model dev net 5 87 2012 a good agreement between the two COSMO MESSy model simulations with the COSMO 7 MESSy simulation show ing more distinct peaks than the COSMO 40 MESSy simula tion as expected The ECHAMS MESSy simulation gener ally shows a smoother Radon distribution and does not catch the peak at about 35 E evident in the COSMO 40 MESSy simulation Yet it overall compares well to the COSMO 40 MESSy simulation Finally we conclude that the vertical convective transport as evaluated with the 7 Rn tracer is realistically represented in COSMO MESSy Geosci Model Dev 5 87 110 2012 106 6 Summary and outlook The implementation of the MESSy interface and some reg ular submodels into the regional weather prediction and climate model COSMO is documented aiming at a new limited area chemistry model The most important code changes both in the COSMO model code and in th
43. grids and or representations into account During the course of connecting MESSy to the COSMO model MESSy was further generalised to allow for different realisations of model grids and representations Note that all the changes apply to the interface layers only as the core layer of the submodels is basemodel independent anyway The three most prominent extensions are 1 The order of the dimensions in multi dimensional array variables representing geo located information In principle the three dimensions in space can be arbi trarily defined in a three dimensional model But in me teorological models the vertical is a distinguished spa tial dimension Therefore the position of the vertical axis becomes important when using three dimensional fields In ECHAMS the order of the three dimensions in space is h1 v h2 with h1 h2 the two horizon tal dimensions and v the vertical dimension whereas the order in COSMO is h1 h2 v Before the con nection of MESSy to the COSMO model the MESSy interfaces of the submodels worked under the assump tion that the order of dimensions is h1 v h2 This Geosci Model Dev 5 87 110 2012 A Kerkweg and P Jockel COSMO MESSy restriction has been eliminated now the order of dimen sions 1s automatically switched for different basemod els This has been implemented using pre processor di rectives as explained in Sect 4 2 A big outer loop over the second horizontal dimension The
44. l reso lution in the ECHAMS atmosphere model J Climate 19 3771 3791 2006 Son S W Gerber E P Perlwitz J Polvani L M Gillett N P Seo K H Eyring V Shepherd T G Waugh D Akiyoshi H Austin J Baumgaertner A Bekki S Braesicke P Br hl C Butchart N Chipperfield M P Cugnet D Dameris M Dhomse S Frith S Garny H Garcia R Hardiman S C J ckel P Lamarque J F Mancini E Marchand M Mi chou M Nakamura T Morgenstern O Pitari G Plum mer D A Pyle J Rozanov E Scinocca J F Shibata K Smale D Teyss dre H Tian W and Yamashita Y Impact of stratospheric ozone on Southern Hemisphere circulation change a multimodel assessment J Geophys Res 115D DOOMO7 doi 10 1029 2010JD014271 2010 Steinkamp J Ganzeveld L N Wilcke W and Lawrence M G Influence of modelled soil biogenic NO emissions on re lated trace gases and the atmospheric oxidizing efficiency At mos Chem Phys 9 2663 2677 doi 10 5194 acp 9 2663 2009 2009 Steppeler J Doms G Sch ttler U Bitzer H W Gassmann A Damrath U and Gregoric G Meso gamma scale forecasts us ing the non hydrostatic model LM Meteorol Atmos Phys 82 75 96 doi 10 1007 s00703 001 0592 9 2003 Stiller G P von Clarmann T Br hl C Fischer H Funke B Glatthor N Grabowski U H pfner M J ckel P Kellmann S Kiefer M Linden A
45. layed for three differ ent pressure levels 850 900 and 950hPa The injection height is 900hPa The ECHAMS MESSy result is displayed in black The red and the blue lines show the maximum mixing ratios of the COSMO 40 and COSMO 7 domains respectively 5 2 3 Radon As a further test the MESSy submodel DRADON Sect 5 1 3 is used prescribing a constant emission flux of 1 atom cm s at the land surface not covered by snow or ice The constant emission flux was chosen in order to yield comparable results for all three model domains Figure 11 displays snapshots of vertical Radon distri butions taken at a cross section at 50 N at 00 00 UTC The upper middle and lower row display the results of the ECHAMS MESSy the COSMO 40 MESSy and the COSMO 7 MESSy simulation respectively The vertical Geosci Model Dev 5 87 110 2012 A Kerkweg and P Jockel COSMO MESSy 50 hPa nmel mal Se a r pai pai min D 5 Mak 10 Mar 20 MAR 5G I hFa nmel mol a ha in a o o mir min min 5 Mak 10 Hak 20 MAR SG 690 hPa nmel mal woof om pia aia min 5 Mak 10 MarR 20 MAR OG Fig 10 Maximum mixing ratio nmol mol of tracer VOL The maximum is calculated in a rectangle ranging from 5 W to 25 E and 36 to 56 N The mixing ratios are displayed for three differ ent pressure levels 750 800 and 850hPa The injection height is 800hPa The ECHAMS MESSy result is displayed
46. levels of the prognostic variables in both models As ECHAMS MESSy uses the leap frog time integration scheme it provides variables for the values after time Geosci Model Dev 5 87 110 2012 A Kerkweg and P Jockel COSMO MESSy integration i e at the end of the previous time step mostly indicated by m1 at the end of the vari able name Additionally the ECHAMS MESSy model utilises variables for the tendencies applied to the prog nostic variables during the current time step mostly in dicated by te at the end of the variable name In contrast to the ECHAMS MESSy model the COSMO model contains 2 and 3 time level integration schemes To simplify the treatment for these different schemes and to avoid copying one time level of a prognostic vari able to another time level of the same prognostic vari able the variable fields in the COSMO model are of rank 4 consisting of the three dimensions in space and an additional rank dimensioned by the number of time levels utilised in the integration scheme To access the different time levels index variables of type INTEGER nnew nnow and nold are defined Their values are rotated every time step thus the m1 or nnow val ues are not located at the same memory space all the time during the integration The tendency of a variable in the COSMO model is defined in a similar way as in ECHAMS MESSy It is mostly indicated by tens at the end of the variable nam
47. ll three processes Monotonicity For the analysis of the monotonicity we use the H tracers which are initially homogeneously distributed In an ideal model none of the transport processes must cause the tracer mixing ratio to deviate from its initial value at any time and any place within the domain i e minimum and maximum value in the entire domain are equal For a regional model this monotonicity test also requires the same constant mix ing ratio being prescribed at all domain boundaries which was the case in our setup The homogeneous tracers Hnt Hv and Hc stay indeed exactly homogeneous not shown indicating that vertical diffusion and convective transport are monotone The advected tracers exhibit a small but negligi ble violation of the monotonicity with a deviation from the initial mixing ratio of less than 0 03 o within 31 days Positive definiteness The generic MESSy sub submodel TRACER_PDEF J ckel et al 2008 provides the possibility to detect negative tracer mixing ratios for instance caused numerically to correct Geosci Model Dev 5 87 110 2012 100 019 010 Hnt Ha Hy He H 0090 0090 010 010 5 10 15 20 25 50 5 10 MAR Vint Vlo Viv Vie VI A Kerkweg and P Jockel COSMO MESSy V2nt V20 V2v V2e V2 io 20 2 20 5 10 15 20 25 30 MAR MAR Fig 5 Corrected negative tracer mass see text in kg for the passive tracers H left V1 middle and V2 right in the
48. ment of the Modular Earth Submodel System MESSy to achieve this goal has several advantages the same or sim ilar process formulations can be used in both the global ECHAMS MESSy and the limited area COSMO MESSy model Furthermore the required dynamical and chemical boundary conditions for the limited area model can be seam lessly taken from the global model In this way a highly con sistent scale bridging model chain is constructed As a pow erful research tool it can be applied for a wide variety of scientific questions it can serve as a zooming option for at mospheric chemistry related campaign and mission support for regional air pollution studies for the dynamical down scaling of the chemical state of the atmosphere on various time scales It can support the development of new param eterisations for global chemistry climate models etc Our ongoing developments therefore focus on the even closer i e on line coupling of the limited area model to the global model the evaluation of the coupled model system at first w r t the simulated meteorology and on the complete inte gration of all atmospheric chemistry related processes in the limited area model Consequently this is only the first in a series of three companion articles about the new modeling sys tem ECHAMS MESSy COSMO MESSy abbreviated MECO n While the basis for a regional chemistry model i e the implementation of the MESSy infrastructure and tracer t
49. mgaertner A Bekki S Braesicke P Br hl C Butchart N Chipperfield M Dameris M Dhomse S Frith S Garny H Hardiman S C J ckel P Kinnison D E Lamarque J F Mancini E Mi chou M Morgenstern O Nakamura T Olivi D Paw son S Pitari G Plummer D A Pyle J A Rozanov E Scinocca J F Shibata K Smale D Teyss dre H Tian W and Yamashita Y Multimodel assessment of the upper tropo sphere and lower stratosphere extratropics J Geophys Res 115 DOOMO9 doi 10 1029 2010JD013884 2010 Hofmann C Kerkweg A Wernli H and J ckel P The 1 way on line coupled atmospheric chemistry model system MECO n Part 3 Meteorological evaluation of the on line coupled sys tem Geosci Model Dev 5 129 147 doi 10 5194 gmd 5 129 2012 2012 J ckel P von Kuhlmann R Lawrence M G Steil B Bren ninkmeijer C A M Crutzen P J Rasch P J and Eaton B On a fundamental problem in implementing flux form advection www geosci model dev net 5 87 2012 A Kerkweg and P Jockel COSMO MESSy schemes for tracer transport in 3 dimensional general circulation and chemistry transport models Q J R Meteorol Soc 127 1035 1052 2001 J ckel P Sander R Kerkweg A Tost H and Lelieveld J Technical Note The Modular Earth Submodel System MESSy a new approach towards Earth System Modeling Atmos Chem Phys 5 433 444 doi 10 519
50. n ality of the GRIBOUT namelists of the COSMO namelist file INPUT_IO the requested output is redirected into MESSy channels The channels are named COSMOXXXy with XXX the number of the amp GRIBOUT namelist e g if three amp GRIBOUT namelists are specified in the INPUT_IO namelist file XXX in the respective channels is replaced by the numbers 001 002 and 003 respectively y indicates the respective variable group and is defined as in COSMO c for constants m for variables on the model grid and p and z for output interpolated on pressure or altitude levels re spectively In that way the MESSy CHANNEL output of COSMO model variables can still be managed by the origi nal COSMO model namelists if desired In addition the original COSMO output files can be writ ten if the LOGICAL switch L_BM ORIG OUPUT in the CHANNEL namelist is set TRUE Due to the fact that both COSMO and CHANNEL are us ing the same memory for each variable and that the on line statis tics of CHANNEL relays on the fact that the primary memory of a variable contains instantaneous values the accumulated fields www geosci model dev net 5 87 2012 93 The generic submodel CHANNEL also manages the model restarts For a model restart all variables re quired for the unambiguous continuation of the simula tion are dumped into so called restart files Unambigu ous in this context means that the results are binary iden tical compared with results f
51. n e g the 3 dimensional vorticity in spectral representation the ozone mixing ratio in Eule rian representation the pressure altitude of trajectories in Lagrangian representation dimensions they represent the basic geometry of one dimension e g the number of latitude points the num ber of trajectories etc event this is a data type provided by the generic sub model TIMER which is used to schedule processes at specific regular time intervals e g to trigger regular output or input during a simulation The event control is part of the MESSy generic submodel TIMER The supplement of J ckel et al 2010 comprises a manual for TIMER and details about the event definition representation it describes multidimensional geometric structures based on dimensions e g Eulerian or grid point spectral Lagrangian rerun_event it triggers the output of restart files restart a restart is performed if the computing time allowed by a job scheduler is limited and too short for the complete simulation In this case the simulation is interrupted in between and restarted in a new job To achieve binary identical results for simulations with and without interruption restart files are written of which the contents fully determine the state of a model sim ulation These files are read in the initialisation phase during a model restart tracer set a group of tracers in the same representation see J ckel et al 2008
52. nd variability of the stratosphere J Geophys Res 116 doi 10 1029 2010JD014995 2011 Doms G and Schattler U The Nonhydrostatic Limited Area Model LM of DWD Part 1 Scientific Documenta tion Deutscher Wetterdienst Offenbach available at www cosmo model org last access 20 June 2011 1999 Franke K Richter A Bovensmann H Eyring V J ckel P Hoor P and Burrows J P Ship emitted NO gt in the Indian Ocean comparison of model results with satellite data At mos Chem Phys 9 7289 7301 doi 10 5194 acp 9 7289 2009 2009 Gettelman A Hegglin M I Son S W Kim J Fujiwara M Birner T Kremser S Rex M A el J A Akiyoshi H Austin J Bekki S Braesike P Br hl C Butchart N Chipperfield M Dameris M Dhomse S Garny H Hardi man S C J ckel P Kinnison D E Lamarque J F Mancini E Marchand M Michou M Morgenstern O Paw son S Pitari G Plummer D Pyle J A Rozanov E Scinocca J Shepherd T G Shibata K Smale D Teyss dre H and Tian W Multimodel assessment of the upper troposphere and lower stratosphere tropics and global trends J Geophys Res 115 DOOMO08 doi 10 1029 2009JD013638 2010 Hegglin M I Gettelman A Hoor P Krichevsky R Man ney G L Pan L L Son S W Stiller G Tilmes S Walker K A Eyring V Shepherd T G Waugh D Akiyoshi H A el J A Austin J Bau
53. ndent see manual MPI Shared high level interfaces are used which internally call the corresponding COSMO routines B2 2 Regular submodels The most important change is the introduction of the rank identifier for flipping the ranks of the variables be tween the ECHAM and COSMO order of dimensions Consequently the definition of representations needs to be adapted to the dimension order of the respective basemodel Geosci Model Dev 5 87 110 2012 108 Supplementary material related to this article is available online at http www geosci model dev net 5 87 2012 gmd 5 87 2012 supplement zip Acknowledgements This work was funded by the German Sci ence Foundation DFG under the project MACCHIATO WE 2943 4 1 The authors are grateful to Holger Tost for the help in identifying the correct input fields for CVTRANS in the COSMO Tiedtke convection scheme We thank all MESSy developers for cooperation and support We thank the Max Planck Institute for Chemistry particularly J Lelieveld for supporting this model development and for providing computational resources The authors acknowledge the application support for high computer performance team of the MPG computer center Garching in particular I Weidl R Hatzky W Nagel and H Lederer Such a development project requires a lot of computing time for testing the new model The development of COSMO MESSy was supported within the project ChESS by the DE
54. number variable 1 Xt 2 XOLE Tendency of the tracer field 3 xtm1 4 5 xeba 6 x The submodel interface layer SMIL manages the com munication data flow between the basemodel inter face layer 1 e the MESSy infrastructure and the sub model core layer It allocates the memory for the sub model specific fields and organises the access to the fields defined by other MESSy submodels or the base model which are required in the respective submodel core Thus a SMIL file provides the connector to the basemodel So far for the implementation of the MESSy regular sub models the SMIL files of each submodel had to be changed for different basemodels Within the scope of the connec tion of MESSy to the COSMO model the SMIL files have been generalised in a way that for additional new base models no further changes of the SMIL files are required Most of the differences between the basemodels here the COSMO model and ECHAMS5 is accounted for by the generic MESSy submodels especially by DATA Sect 3 4 Only one big difference remains for all regular submodels which has to be taken care of in the SMIL files the order of the spatial dimensions Whereas ECHAM5 MESSy arranges the spatial dimensions of a grid point field h1 z h2 with h1 h2 the horizontal dimensions and z the vertical di mension the COSMO model applies the order h1 h2 z This difference in order needs to be taken into account in the SMIL when data fields a
55. of these representations are meaningless in a regional grid point model This leaves only one important subroutine namely locate_in_decomp This subroutine determines the corresponding horizontal indices and PE number in a par allel decomposed grid for a point given in geographical coor dinates or for COSMO MESSy optionally in coordinates of the rotated grid The subroutine locate_in_decomp is for instance required in the regular MESSy submodel TR EXP see Sect 5 1 2 and J ckel et al 2010 to correctly identify the geo location of point sources Geosci Model Dev 5 87 110 2012 96 3 7 TRACER the management of constituents The generic submodel TRACER J ckel et al 2008 pro vides the interface for the management of constituents e g chemical species in the model system The tracer definition consists of two parts the meta information defining the properties of the tracer e g the quantity amount fraction number density or concentration and the unit of the component the medium in which the tracer resides air aerosol cloud ocean etc switches for the processes the tracer should undergo within the simulation e g advection dry deposi tion scavenging etc and some tracer specific constants as the molar mass the henry coefficient etc The tracer data field itself It is always defined for one tracer Set i e for a group of tracers with the same rep resentation i e
56. ompanion articles It documents the implementation of the MESSy infrastruc ture plus a few diagnostic submodels to test mainly the TRACER infrastructure into the COSMO model The re sulting model is a regional model of the atmosphere en abled for now standardised further extensions into a re gional atmospheric chemistry model The second part Kerk weg and J ckel 2012 is about a different issue however re quires the COSMO MESSy model of the first part there we present a newly developed coupling technique that provides the boundary data required by the regional model directly on line from ECHAMS5S MESSy via the newly developed Multi Model Driver MMD library and two corresponding submodels This on line coupled system is called MECO n i e MESSy fied ECHAM and COSMO models nested n times The third article Hofmann et al 2012 provides a meteorological evaluation of the nested system described technically in Part 2 This evaluation focusing on distinct meteorological events on synoptic scale and on the question if and how they can be reproduced by MECO n is a pre requisite for further applications with chemistry like chemi cal weather air pollution forecasts measurement campaign analyses etc 3 convective tracer transport CVTRANS simplified prognos tic tracers PTRAC point sources and simplified chemistry TR EXP and tracers of opportunity DRADON Geosci Model Dev 5 87 110 2012 A Kerkweg and
57. on scales requiring convection parameterisations whereas it is inap propriate for convection resolving simulations The choice of appropriate submodels for a specific scientific purpose is at the end the responsibility of the user One prerequisite for the next step namely the inclusion of the submodels relevant for atmospheric chemistry appli cations with COSMO MESSy is the correct functionality of the TRACER infrastructure which handles the data and meta information of chemical constituents This function ality has been successfully demonstrated with three differ ent tests for artificial passive i e without sources and sinks tracers with point sources and with Rn with a standard source distribution The tests involved three regular MESSy submodels PTRAC TREXP and DRADON The results furthermore indicate that no severe violations of mass con servation monotonicity or positive definiteness of the tracer transport are apparent From that we conclude that the model is ready for atmospheric chemistry applications but further regular tests which also budget the in and outflow across the domain boundaries are desirable Geosci Model Dev 5 87 110 2012 A Kerkweg and P Jockel COSMO MESSy The implementation and the successful testing of the tracer transport forms an important first step and a pre requisite towards the development of a limited area atmo spheric chemistry model The application and further devel op
58. op over the second horizontal dimension in ECHAMS MESSy currently available generic submodels is provided in the COSMO MESSy Implementation Documentation in the Supplement Note that MESSy version 2 as documented by J ckel et al 2010 includes already the basis for the cou pling of MESSy to COSMO Therefore according to the MESSy concept the further developments discussed here either do not interfere with ECHAMS MESSy because they are not relevant for it or they are immediately applicable also for ECHAMS MESSy 3 1 SWITCH CONTROL switches and main entry points for individual submodels SWITCH CONTROL is the MESSy submodel managing the switching and calling of the individual submodels In the basemodels distinct entry points have been defined from which the MESSy submodels are called One entry www geosci model dev net 5 87 2012 91 point consists of one call to a subroutine of CONTROL This subroutine executes calls to the generic submodels and if switched on to the respective regular submod els Figure 2 lists all current entry points for MESSy As not all SMILs of all submodels have been adapted so far not all entry points available in ECHAMS MESSy are also used in COSMO MESSy Those marked by hatched boxes are presently not used in COSMO MESSy the yel low boxes in contrast denote the active entry points in COSMO MESSy The MESSy submodels currently called from the respective subroutine are listed next to the
59. ost H Sander R Schulz M Stier P Vignati E Wilson J and Lelieveld J Consistent simulation of bromine chemistry from the marine boundary layer to the stratosphere Part 1 Model description sea salt aerosols and pH Atmos Chem Phys 8 5899 5917 doi 10 5194 acp 8 5899 2008 2008a Kerkweg A J ckel P Warwick N Gebhardt S Brenninkmei jer C A M and Lelieveld J Consistent simulation of bromine chemistry from the marine boundary layer to the stratosphere Part 2 Bromocarbons Atmos Chem Phys 8 5919 5939 doi 10 5194 acp 8 5919 2008 2008b Lauer A Eyring V Hendricks J Jockel P and Lohmann U Global model simulations of the impact of ocean going ships on aerosols clouds and the radiation budget Atmos Chem Phys 7 5061 5079 doi 10 5194 acp 7 5061 2007 2007 Morgenstern O Giorgetta M A Shibata K Eyring V Waugh D W Shepherd T G Akiyoshi H Austin J Baumgaertner A J G Bekki S Braesicke P Br hl C Chipperfield M P Cugnet D Dameris M Dhomse S Frith S M Garny H Gettelman A Hardiman S C Hegglin M I J ckel P Kinnison D E Lamarque J F Mancini E Manzini E Marchand M Michou M Naka mura T Nielsen J E Olivi D Pitari G Plummer D A Rozanov E Scinocca J F Smale D Teyss dre H Toohey M Tian W and Yamashita Y Review of the for mulation of present generation
60. outine p_bcast as in ECHAMS in contrast requires only two parame ters buffer and rank of the sender others are optional To create a common interface the group communicator the buffer length and the MPI datatype have been made optional parameters in the COSMO MESSy section of the generic submodel MPI If they are not present they are deduced from other input the communicator is assumed to be the model s world communicator the buffer length is SIZE buffer and determined by the MPI data type is determined according to the type of the buffer messy_mpi_initialize this MESSy subrou tine is called directly from the basemodel to ini tialise the LOGICAL variables p_parallel_io and p_parallel indicating whether the actual PE is re sponsible for input and output and whether the model is working in a parallel environment respectively p_abort this subroutine manages the proper termina tion of a simulation in a parallel environment by calling the COSMO model routine model_abort Some further variables used in the MESSy submodels are defined accordingly The COSMO MESSy Implementation Documentation in the Supplement contains a description of these variables 3 6 TRANSFORM the interface for grid transformations The generic MESSy submodel TRANSFORM hosts utility subroutines to transform variable fields from one represen tation into another e g grid point spectral or Lagrangian Most
61. ransport into the COSMO model is presented here the second article Kerkweg and J ckel 2012 is about a dif ferent issue but based on the COSMO MESSy model of this first part It describes a newly developed coupling method ology for the on line nesting of COSMO MESSy instances into ECHAMS MESSy Last but not least the third arti cle Hofmann et al 2012 provides a first meteorological evaluation of the newly developed on line coupled model chain described technically in the second article This eval uation focusing on distinct meteorological events on synop tic scale and on the question if and how they can be repro duced by MECO n is a prerequisite for further applications with chemistry like chemical weather air pollution fore casts measurement campaign analyses etc with a focus on the representation of distinct meteorological situations Appendix A Glossary channel the generic submodel CHANNEL manages the memory and meta data and provides a data transfer and export interface J ckel et al 2010 A channel rep resents sets of related channel objects with additional www geosci model dev net 5 87 2012 A Kerkweg and P Jockel COSMO MESSy meta information The relation can be for instance the simple fact that the channel objects are defined by the same submodel channel object it represents a data field including its meta information and its underlying geometric struc ture representatio
62. re provided to the core layer subroutines where the vertical dimension is usually particu larly distinguished This is implemented by applying a rank flipping based on pre processor directives This approach proves useful to avoid unnecessary doubling of code imply ing a lower error proneness and a reduced risk of inconsis tencies as changes in the submodel interfaces are directly valid for all grid point basemodels The rank flipping is implemented via pre processor direc tives the so called rank identifiers Basically the ranks that need to be flipped dependent on the basemodel are re placed by a directive in the code As a lot of different combi nations of indices and colons is possible a naming conven tion for the Fortran95 index variables and the rank identifiers www geosci model dev net 5 87 2012 Tracer field corresponding to nnow LF only is required All rank identifiers are defined basemodel depen dent in the include file messy_main_ppd inc which is included in all SMIL files For instance with rank identifiers the calculation of the thickness of the lowest model layer in pressure units dp reads dp l kproma amp press 3d 1 skoroma Ri Vpl amp jo re ssa 30 koroma RA YY pressi_3d 1s the pressure at the layer interfaces and kproma the length of the first blocked horizontal dimen sion The rank identifiers _RI_YV_ and _RI_YVpl_ are defined as ifdef ECHAM5 define RI YV nl
63. restart facility is replaced by the CHANNEL restart interface in COSMO MESSy as mentioned above the subroutine messy_timer_COSMO_reinit_time is additionally called during a restart after reading the restart attributes from the restart files The generic submodel TIMER is described in more detail by J ckel et al 2010 and the corresponding Supplement 3 4 DATA the data transfer interface The generic submodel DATA is part of the memory manage ment and data transfer interface The basemodel data fields are made available and easily accessible by DATA One of the guidelines for the MESSy interface implementation into the COSMO model is to use as many parts as possible avail able for ECHAMS MESSy also for COSMO MESSy i e to minimise the code that needs to be maintained indepen dently for ECHAMS MESSy and COSMO MESSy There fore DATA is the model unifying the data structure and names of the basemodel variables Different aspects had to be taken into account When US E ing a basemodel variable provided by DATA the MESSy submodels access this variable al ways with the same variable name independent of the basemodel For instance the variable name of the sea land fraction in ECHAMS5S MESSy is s1f whereas the same variable in the COSMO model is fr_land Therefore fr_land is renamed in DATA to s1f by USE data_fields amp ONLY slf gt fr land PUBLIC 9 SLI An additional challenge is the treatment of the time
64. restart frequency are defined in the namelist file of TIMER timer nml the entries in the COSMO namelist RUNCTL in the namelist file INPUT_ORG for dt hstart hstop and ydate_ini are hence ignored as is the entry restart_hour of the namelist INPUT_IO The most important time variables are initialised via namelist in messy_timer_setup which is called from the organize_setup subroutine of the COSMO model via the CONTROL subroutine messy_setup see Fig 3 Sect 1 After this initialisation of the MESSy variables the COSMO model time variables are initialised or overwritten by TIMER Affected are the time step dt the calendar type it ype_calendar the initial date ydate_ini are not accumulated in the COSMO output files even if L_BM_ORIG_OUPUT TRUE Geosci Model Dev 5 87 110 2012 94 the indicator for model stop in hours hstop and in steps nstop the COSMO variables containing the distance from model start hst art or nstart in hours or steps re spectively and the date strings yakdat1 and yakdat2 All variables except for dt and itype_calendar are re set in the subroutine messy_timer_COSMO_reinit_time Addition ally the counters for the re initialisation of the maximum 10m wind velocity and the minimum and maximum temper ature are re set here After this re initialisation all COSMO time variables are consistent with the TIMER setup Because the original COSMO CLM
65. rom a continuous simula tion without interruption The restart files are read dur ing the initialisation phase of a restarted simulation The climate version of the COSMO CLM model already in cluded a restart facility Nevertheless as CHANNEL re places the complete COSMO model memory management the restarts are also managed by CHANNEL Thus the read procedure of the restart file in the COSMO subrou tine organize_input src_input f 90 is skipped in COSMO MESSy 3 3 TIMER the heart beat and event management The generic submodel TIMER became part of the MESSy system within the scope of the connection of MESSy to the COSMO model Until now ECHAMS5 MESSy used the time and event management of ECHAMS and all MESSy submodels utilising events depended on the ECHAMS event management routines Therefore the func tionality of ECHAMS events had to be made available within COSMO MESSy as well Consequently the generic MESSy submodel TIMER is based on the time and event managing routines of ECHAMS written by Ingo Kirchner Max Planck Institut for Meteorology Hamburg now at FU Berlin A user manual for the TIMER submodel can be found in the supplement of Jockel et al 2010 As MESSy provides the time management and be cause each model can apply only one time management the COSMO model time management is overwritten by the MESSy generic submodel TIMER Consequently all time variables as time step start date stop date and the
66. s pro vides a namelist interface for the definition of tracers Usu ally tracers are defined in the SMIL module of a submodel which requires a recompilation of the code if a new tracer is added To provide an easy test bed for tracer studies without the need of recompilation PTRAC enables the definition of tracers via namelist The submodel is described in detail by J ckel et al 2008 5 1 2 TREXP The regular submodel TREXP provides the possibility to de fine point sources of tracers and to define tracers experienc ing one degradation reaction either by zero order decay or by a first order reaction with another educt The tracers the reaction coefficients and the point sources are all specified in the amp CPL namelist of TREXP The point sources are also applicable to tracers defined elsewhere in MESSy submod els J ckel et al 2010 provide a detailed description of the MESSy submodel TREXP 5 1 33 DRADON The regular submodel DRADON Diagnostic RADON pro vides an ideal test bed for the diagnosis of tracer transport J ckel et al 2010 explain the concept of DRADON in de tail Different parameterisations of the radon emission fluxes can be applied For the test shown in Sect 5 2 3 we applied a constant emission of 1 atom cm s over ice and snow free surfaces 5 1 4 CVTRANS Convective tracer transport is managed by the MESSy sub model CVTRANS Tost et al 2010 The submodel re quires as input the convective
67. specific water contents of rain snow and graupel Different time integration schemes can be selected asecond order leapfrog horizontally explicit vertically implicit time split integration scheme including exten sions proposed by Skamarock and Klemp 1992 a three time level 3 D semi implicit scheme Thomas et al 2000 several options for two time level 2nd and 3rd order Runge Kutta split explicit schemes after Wicker and Skamarock 2002 and a TVD variant Total Variation Diminishing of a 3rd order Runge Kutta split explicit scheme A variety of physical processes are taken into account through parameterisations e g grid scale and subgrid scale clouds precipitation moist and shallow convection radia tion a soil model and so forth The COSMO model is available to the scientific commu nity To make the COSMO model applicable on the cli mate scale a number of extensions have been implemented 4 Climate Limited area Modelling community http www clm community eu gt The COSMO Model is available free of charge for scien tific and educational purposes especially for cooperational projects www geosci model dev net 5 87 2012 A Kerkweg and P Jockel COSMO MESSy by the CLM community for instance netCDF output and a restart facility Hereafter the COSMO model is named COSMO CLM when features are discussed which are only used by the climate community 2 2 MESSy The Modular Earth
68. stry related processes are often gov erned by local features e g emissions of a power plant or fire emissions which are not sufficiently resolved in global models smaller scale models for atmospheric chemistry are required In order to investigate those smaller scale ef fects a regional model including a chemistry setup consis tent with that of the global model is desirable This goal is achieved by connecting MESSy to the regional weather prediction and climate model of the Consortium for Small Scale Modelling COSMO model Steppeler et al 2003 Doms and Schiattler 1999 resulting in the model system COSMO MESSy The Modular Earth Submodel System MESSy Jockel et al 2005 was developed with a focus on atmospheric chemistry applications Currently MESSy depends on a basemodel which calculates the basic dynamic processes and provides the model domain and its parallel decomposi tion The interfaces for 3 dimensional basemodels in MESSy were initially developed for the global atmospheric gen eral circulation model GCM ECHAMS Roeckner et al 2006 A large variety of process studies has been performed l http www cosmo model org A http www messy interface org Published by Copernicus Publications on behalf of the European Geosciences Union 88 with the global ECHAMS5S MESSy atmospheric chemistry EMAC model J6ckel et al 2006 e g on the influence of different convection parameterisations Tost et al 2006
69. the 10 Note this specific order is not required of the TRACER sub model Geosci Model Dev 5 87 110 2012 A Kerkweg and P Jockel COSMO MESSy time integration scheme for the Runge Kutta scheme only 5 instances are needed whereas the leap frog scheme re quires 6 instances The different time levels of the tracer fields are copied at the beginning of the new time step 1 e in the subroutine initialize _loopin lmorg f90 Consequently the access to the respective time levels of the tracer field is not as simple as in the COSMO model it self To mimic the functionality of the access to the different time levels of a prognostic field by the indices nnow nnew and nold additionally the POINTER ARRAY xt_array is allocated to the number of time levels in the integra tion scheme In addition to copying the tracer time in stances in initialize_loop the individual POINTERs of xt_array are associated to the respective tracer fields xt_array nnew PTR gt xt L the cyclic changes in COSMO is applied Tor the tracer Ttield here IF l2tls THEN 2 time level scheme Sm ee At se ANS eS a a ee xXxt_array nnow PTR gt xtml EGOE 3 time level scheme nnow gt nold MEME ty age SEE aeit nnew gt nnow SOE Wiss Sey Sages aE reer E xt_array nnow PTR gt xtf xt_array nold PTR gt xtml ENDIF Thus different time levels of the tracer field are also accessi ble via the COSMO model time level indices
70. ts the initialisation patterns for the artificial tracers H is initialised homoge neously to a mixing ratio of 107 mol mol Figure 4 de picts the initialisation of the V1 tracer V2 is initialised re versely to V1 1 e its mixing ratio increases from bottom to www geosci model dev net 5 87 2012 99 Table 3 Table of passive tracers for the transport test with PTRAC X is aplaceholder for either H V1 or V2 and refers to the initiali sation patterns Tab 2 Transport process Tracer Advection Vertical Convection name diffusion Xnt Xa T Xv ae XC T X T a T top Thus V1 and V2 are defined in a way that V1 V2 is also homogeneously distributed Since the artificial trac ers do not have any internal sources or sinks their abun dance is only changed by transport processes including in flow outflow across the domain boundaries The MESSy TRACER interface nicely provides the possi bility to switch on or off the specific transport processes for each tracer individually Utilising this tracers with the three initial conditions Table 2 have been combined with the five different transport process switch combinations as listed in Table 3 Thus in total 15 tracers have been simulated for each initialisation pattern one tracer encounters no transport nt three tracers experience exactly one transport process advection a vertical diffusion v or convection c and one tracer undergoes a

The 1-way on-line coupled atmospheric chemistry model system

Contents

Download Pdf Manuals

Related Search

Related Contents