Global-to-regional simulation of aerosols with 10 km grid spacing

Transcription

1 Workshop: Oct 17-19, Chi Minh City, Vietnam Global-to-regional simulation of aerosols with 10 km grid spacing Daisuke GOTO ( 五藤大輔 ) (goto.daisuke@nies.go.jp) National Institute for Environmental Studies (NIES), Tsukuba, Japan Collaborating with T. Dai, H. Yashiro, Y. Sato, K, Suzuki, T. Seiki, T. Takemura, K. Sudo, J. Uchida, S. Misawa, and T. Nakajima Supported by MEXT/RECCA/SALSA (FY ; PI: Prof. T. Nakajima) MOEJ/S-12 (FY ; PI: Prof. T. Nakajima) Supercomputer: K RIKEN (hp & hp & hp160004) 1

2 Motivation: Uncertainty of simulating air pollutions by global models Anthropogenic aerosol optical thickness large variability in AeroCom Myhre et al. (ACP 13) 2

3 Motivation: Uncertainty of simulating air pollutions by regional models Surface PM2.5 concentration in annual CMAQ EMEP Surface sulfate concentration in April LOTOS-EUROS SILAM [Prank et al., ACP2016] [modified & added to Goto et al., JGR2015] still large uncertainty 3

4 What is the reason of the gap between models? [Goto et al., JGR2015] Source region in annual mean The difference in clouds between NICAM-SPRINTARS (dx=220km) and MIROC-SPRINTARS (dx=2.8deg) strongly affects the difference in sulfate between them. Proper cloud simulation is important! 4

5 NICAM with 3.5km grid spacing : Cloud resolving model [Miura et al., science 07] MTSAT-IR NICAM 5

6 Motivations and contents in this study Motivation There are various uncertainties in various processes To minimize the uncertainty of clouds (i.e., investigate the impacts of cloud on aerosol distribution), we use global cloud-system resolving model (NICAM with O(10km) grid without cloud parameterization) Contents Model description Results Regional high-resolution (stretching grid system) Saving computational resources, before global calculation Check whether the model successfully run in the high-resolution Global high-resolution (uniform grid system) Application of NICAM (under MOEJ/S-12 project) 6

7 NICAM-Chem: Atmospheric aerosol-chemistry model coupled to NICAM Aerosol-Chemistry module Nonhydrostatic ICosahedral Atmospheric Model (NICAM) Scale-up Regional + RESOURCE INFORMATION Language: Fortran Program size: 100,000 lines Method: Finite volume method, Runge-Kutta 4 th Parallelization: MPI+Thread Effective parallelism: 99.99%(Strong) Parallel efficiency: 51.4% (Strong) Uniform grid (Global) Stretch/Diamond grid (Regional) Suzuki et al. (GRL 08), Dai et al. (AE 14; EP 14), Goto (EP2014), Goto et al. (GMD 15; JGR 15) Regional calculation (low-computer resource) Validation (East Asia) in regional scales Global calculation (high-computer resource) Aerosol Optical Depth (dx=14km) Extend K computer (RIKEN/AICS) 10km stretching: app node-hour(nh)/month 14 km (3.5km): app. 20,000 (820,000) NH/month 7

8 Regional simulations using NICAM (a) Usual regional simulation: Combination of multi-models by multiple nesting GCM/CTM (Spectral): dx>100km Regional CTM dx=180km (T63) dx=60km (Domain 1) dx=15km (Domain 2) (b) One model without nesting (New approach by our group) dx=5km (Domain 3) Courtesy of Dr. Ohara GCRM (Grid-spacing): dx~any scales Stretching grid Diamond grid Glevel 0 (dx=7141km) Glevel 6 (dx=112km) Glevel 6+Stretch100 (dx=11km) dx=3-5km 8

9 NICAM as host model Dynamic core GL05/GL06/GL07 Stretch-GL06 GL09 NICAM (Non-hydrostatic icosahedral atmospheric model) (Tomita & Satoh, 2004; Satoh et al., 2008, 2014) Grid configuration Uniform Stretch Uniform Transport Improved van Leer (1977) (Miura, 2007; Niwa et al., 2011a) Cloud Large-scale condensation & cumulus para. Auto-conversion Berry (1967) NSW6 by Tomita (2008) & no cumulus param. Boundary layer Meller & Yamada (1974), Nakanishi & Niino (2004, 2006) Radiation k-distribution by MSTRN (Sekiguchi and Nakajima, 2008) Khairoutdinov & Kogan (2000) Turbulence MY2-MOIST (Meller & Yamada, 1974; Nakanishi & Niino, 2004; 2009) Land MATSIRO (Takata et al., 2003) Aerosol SPRINTARS (Takemura et al., 2000, 2002, 2005, 2009) Horizontal res. 220km/110km/56km 11km (finest) 14km Vertical res. 40 layers 38 layers Timestep (sec.) 1200/600/ Nudging OFF/ON/OFF ON OFF Nudging data NCEP-FNL of winds above 2 km H (relaxing <2km) NO Period August of (3yr) 9

10 Aerosol module GL06 Stretch-GL06 GL09 Module SPRINTARS (Takemura et al., 2000; 2002; 2005; 2009), Goto et al. (2011) Species Dust, Seasalt, Carbonaceous (OC&BC), Sulfate Oxidants for S Offline-calculated {OH, H 2 O 2, O 3 } by CHASER (Sudo et al., 2002) Hygroscopitity On-line emission Inventory Sulfate, Carbonaceous, and Seasalt Dust & seasalt depending on wind speed at 10-m height EAGrid2000 (Kannari et al., 2007), REAS-v2 (Kurokawa et al., 2013) HTAP-v2 (Janssens-Maenhout et al., 2015) Reference Goto (2014), Goto et al. (GMD2015) Goto et al. (prep) NOTES SPRINTARS is originally developed and optimized in the framework of MIROC, which is a global climate model, developed by Japanese group: AORI/NIES/JAMSTEC. SPRINTARS is also implemented to MIROC-CHEM (fullchem) and MIROC-ESM. The module is focused on global simulation (with low-resolution ranging km & diagnostic precipitation & cloud parameterization & timestep>=10min) SPRINTARS in NICAM (high-resolution) is applied with prognostic precipitation & no cloud parameterization & timestep =<1min. 10

11 Regional-11km simulation (1) : Precipitation in August 2007 (a) High-resolution (b) Low-resolution [Goto et al., GMD2015] 11

12 Regional-11km simulation (2) : Aerosols in August 2007 (a) High-resolution (b) Low-resolution (c) High-resolution (d) Low-resolution [Goto et al., GMD2015] Good performance of high-resolved simulation with NICAM-Chem 12

13 Regional-11km simulation (3) : PM 2.5 around Tokyo in August 2007 ー : High-resolution, ー : Low-resolution Good correlation, but underestimation by twice 1) Underestimation of organic and other compounds, (which cannot yet be undefined even by chemical analysis of the measurement.) 2) The simulation is not included aerosol water. Only results with the 10 km resolution captured diurnal variations. 100 km grids are too coarse to simulate diurnal variation. [Goto et al., GMD2015] 13

14 Global-14km simulation (1) : Precipitation in July NICAM Very good correlation over tropics (especially, strong peak regions) However, underestimation over the other areas. Reanalysis (GPCP) Precipitation [mm/day] c) July ー : GPCP ー : NICAM

15 Global-14km simulation (2) : Black carbon in July O: CAWNET +: Asia (China) : United States : Europe NICAM No measurement R=0.52 NMB=-78% Observation O: IMPROVE O: EMEP R=0.24 NMB=-30% R=0.59 NMB=-30% 15

16 Global-14km simulation (3) : Sulfate in July O: EANET&CAWNET +: Asia (China) : Asia (Others) : United States : Europe NICAM No measurement R=0.34 NMB=-82% Observation O: IMPROVE O: EMEP R=0.74 NMB=-52% R=0.54 NMB=-87% 16

17 Global-14km simulation (4) : AOD in July NICAM NICAM O: AERONET, CARSNET, SKYNET MODIS-C6/Aqua Only AERONET NICAM vs AERONET Mean Bias = 0.78 R=0.59 Normalized MB=-19% GL09 17

18 Summary Conclusions Regional NICAM (Stretch NICAM) with the 10 km resolution (without cloud parameterization) as a regional model can capture diurnal and synoptic variations. It also shows that 100 km grids are too coarse to simulate diurnal variation. Global NICAM with the 14 km resolution can run during threeyears using the K computer. We continue to validate aerosols/clouds using multiple measurements. Future studies Further improvements of NICAM-Chem Investigating aerosol-cloud interactions using the results Thank you for your attention! 18