1. Design and evaluation of the National Institute for Environmental Studies (NIES) transport model Dmitry Belikov and Shamil Maksyutov National Institute for Environmental Studies, Tsukuba, Japan The 2010 Workshop on the Solution of Partial Differential Equations on the Sphere

2. Model formulation New version of the NIES TM (NIES-08) with flux-form advection algorithms have been designed. Just like in the predecessor model with semi-Lagrangian algorithms (Maksyutov et al., 2008), we presented the atmospheric constituent transport equation in the Lagrangian-style form (Willamson and Laprise, 2000): 8/27/2010PDEs 2010, Potsdam2

3. A reduced latitude-longitude grid scheme The offline global tracer transport model version uses a reduced latitude-longitude grid scheme (Peterson et al., JGR, 1998), in which the sizes of grids are doubled several times approaching the poles Advantages vs. icosahedral, cubic grids – easy to bring reanalysis data, and formulate 2 nd, 3 rd order approximations 8/27/2010PDEs 2010, Potsdam3

4. Horizontal mass flux correction method The horizontal mass fluxes, derived from the spectral data (the output of weather forecast models) are balanced with the surface pressure tendency by adding correction fluxes, which is necessary to be determined (Heimann and Keeling, Geophys. Mon., 1989) The correction flux is calculated by transforming Equation into a Poisson equation, which is solved with a discrete 2D Fourier transform for every level l 8/27/2010PDEs 2010, Potsdam4

5. NIES TM transport algorithm test Initial tracer field velocities SML – Semi-Lagrangian (Maksyutov et al. 2008); VL – 3-rd order van Leer scheme (van Leer, 1977); Pr – Second Moments (Prather, 1986) 8/27/2010PDEs 2010, Potsdam5

6. NIES TM transport algorithm test ResolutionNIES-08\SMLNIES-08\VLNIES-08\Pr 2.5º  2.5º CPU, sec6.0810.21292.90 emin–5.37E-036.22E-041.38E-04 emax2.16E-03–2.86E-03–3.48E-04 err16.09E-02–6.66E-03–5.96E-08 err25.08E-031.17E-031.02E-03 Memory, GB0.72 0.77 0.625º  0.625º CPU, sec82.20370.97512683.15 emin–1.04E-073.93E-05–5.11E-06 emax7.95E-08–2.44E-03–5.21E-03 err11.59E-02–1.75E-035.55E-03 err21.74E-026.15E-041.18E-04 Memory, GB1.94 2.46 SML – Semi-Lagrangian (Maksyutov et al. 2008), VL – 3-rd order van Leer scheme (van Leer, 1977), Pr – Second Moments (Prather, 1986) 8/27/2010PDEs 2010, Potsdam6

7. NIES TM meteorology data The Japan Meteorological Agency (JMA) Japan Climate Data Assimilation System (JCDAS) meteorological dataset (6-hourly time step, resolution of 1.25×1.25 deg, 40 hybrid vertical levels). Height of planetary boundary layer with time step of 3 hours are taken from ECMWF Interim Reanalysis. Global Point Value (GPV) - a special product prepared by the Japan Meteorological Agency Global Spectral Model (JMA-GSM) (3 hourly time step, resolution of 0.5   0.5 deg, 21 pressure levels). 8/27/2010PDEs 2010, Potsdam7

8. High resolution: 0.625 deg globally Simulated surface CO 2 concentration around Japan at 21:00UTC, March 26, 2008 using NIES-08 with resolution 0.625 deg 8/27/2010PDEs 2010, Potsdam8

9. Hybrid sigma-pressure and sigma-isoentropic vertical coordinate systems A hybrid sigma-pressure and a sigma-isentropic vertical coordinate systems with 32 levels up to 2 mb 8/27/2010PDEs 2010, Potsdam9

10. Hybrid sigma-pressure and sigma-isoentropic vertical coordinate systems Pressure, hPa Mean age of air (SF 6 ) simulated by the NIES-08 with sigma-pressure (left) and sigma-isoentropic (right) vertical coordinate systems 8/27/2010PDEs 2010, Potsdam10

11. NIES TM results (SF 6 ) Interhemispheric gradients of modeled and observed SF 6 concentrations Zonally averaged annual mean of SF6 concentration simulated by NIES-08 8/27/2010PDEs 2010, Potsdam11

12. NIES TM results (CO 2 ) Latitudinal distributions of CO 2 seasonal amplitude at 35 GLOBALVIEW-CO 2 (2008) sites. Seasonal amplitude is the difference between the maximum and the minimum of seasonal cycle. 8/27/2010PDEs 2010, Potsdam12

13. Convective parameterization scheme Kuo-type cumulus parameterization (Grell, 1994) including entrainment and detrainment processes on convective updrafts and downdrafts proposed by Tiedtke (1989); new method to determine cumulus convective updrafts where P conv denotes the convective precipitation rate at the surface [kg/m 2 /sec], q base is the absolute humidity at the cloud base [kg/kg]; 8/27/2010PDEs 2010, Potsdam13

14. Convective parameterization scheme Seasonally average convective mass flux (g/m 2 /sec) from the NIES TM and Modern Era Retrospective-analysis For Research And Applications (MERRA) data for summer 2006 NIES TMMERRA 8/27/2010PDEs 2010, Potsdam14

15. NIES TM results ( 222 Rn) 8/27/2010PDEs 2010, Potsdam Without convective parameterization With new convective parameterization 15

16. NIES TM results ( 222 Rn) The model results are compared with data from in situ observations (Kritz et al., JGR, 1998; Liu et al., JGR, 1984; Zaucker et al., JGR, 1996) and the results obtained from model GAMIL (Zhang et al., ACP, 2008) 8/27/2010PDEs 2010, Potsdam16

17. Conclusion Improvements in tracer transport simulation are achieved due to: Mass conservative numerical algorithm and horizontal mass flux correction method; A reduced latitude-longitude grid scheme; Hybrid sigma-pressure and a sigma-isentropic vertical coordinate systems; Convective parameterization scheme; Future: High-resolution meteorological data; 8/27/2010PDEs 2010, Potsdam17

18. Thank you! 8/27/2010PDEs 2010, Potsdam18