The Regional Arctic System Model (RASM) for Studying High Resolution Climate Changes in the Arctic

Transcription

1 The Regional Arctic System Model (RASM) for Studying High Resolution Climate Changes in the Arctic Mark W. Seefeldt, John J. Cassano, Alice K. Duvivier, and Mimi H. Hughes University of Colorado Boulder Wieslaw Maslowski, Andrew Roberts Bart Nijssen, Joseph Hamann

2 RASM Participants RASM 1.0 Code Contributors Atmosphere WRF v3.2 University of Colorado Boulder: John Cassano, Mark Seefeldt, Alice DuVivier, Mimi Hughes, Matthew Higgins Land VIC University of Washington: Bart Nijssen, Joseph Hamman Ocean POP2 Naval Postgraduate School: Wieslaw Maslowski, Robert Osinski Sea Ice CICE5 Naval Postgraduate School: Andrew Roberts Coupler CPL7 Naval Postgraduate School: Anthony Craig Additional Contributors Iowa State University: William Gutowski, Brandon Fisel University of Arizona: Xubin Zeng, Michael Brunke

3 RASM Overview RASM 1.0 Code Configuration Atmosphere WRF v3.2 50km, 40 levels Land VIC 50km, 3 Soil Layers ESM Reanalysis Atmospher ic Boundary Conditions Ocean POP2 1/12 (~9km) & 1/48 (~2.4km), 45 levels (7 in the top 42 m), Atmosphere WRF Sea Ice CICE5 1/12 or 1/48, 5 thickness categories Anistropic(EAP)/Isotropic(EVP) rheology Land VIC Coupler CPL7 Sea Ice CICE Coupler CPL7x Flux exchange every 20 minutes, inertial resolving with minimized lags. Streamflow Routing RVIC Ocean POP

4 RASM Overview Model domain, components, and interactions of RASM

5 RASM Overarching Science Hypothesis Mesoscale processes and resulting feedbacks are critical to improved representation of the state of the Arctic Climate System and prediction of polar amplification of global climate change. Some Arctic processes/feedbacks poorly resolved/represented in ESMs - sea ice thickness distribution, deformation and export, fast ice, snow cover, melt ponds and surface albedo - oceanic eddies, tides, surface/bottom mixed layer, buoyancy-driven coastal and boundary currents, upper ocean heat content - atmospheric modes of circulation, clouds, aerosols, fronts - shelf-basin, wave-ice and air sea-ice interactions and coupling U.S. national strategies call for development of comprehensive Arctic System Models

6 CMIP5 - ERA40 Differences in winter MSLP (DJF ) BCC CSM1.1 CanESM2 NCAR CCSM4 The latest suite of global climate models has a large spread and a large magnitude of biases in simulating mean SLP in the Arctic. GISS-E2-H GISS-E2-R INM-CM4 NorESM1-M HadGEM2 ISPL-CM5A (Maslowski et al., 2012)

7 CMIP5 Mean September Sea Ice Thickness ( ) BCC CSM1.1 CanESM2 NCAR CCSM4 Atmospheric circulation biases (shown in the previous slide) affect the sea ice thickness distribution in the Arctic. GISS-E2-H GISS-E2-R NPS NAME (Ctrl) NorESM1-M HadGEM2-ES IPSL-CM5-LR (Maslowski et al., 2012)

8 Missing Heat Source in GCMs Arctic Ocean Heat Content (Haynes., 2010) Before freezing oceanic heat is removed only from the mixed layer!

9 Modeled changes between and means: heat content (TJ; m;left) & sea ice thickness (m;right) m Increasing heat content due to local insulation, advection of warm water from shelves, anticyclonic eddies, slope upwelling or advection (Maslowski et al, 2014)

10 RASM Deformation and Air-Sea Fluxes 3/2/93 3/2/93 Area & Drift Divergence 3/2/93 3/2/93 Upward heat flux (W/m 2 ) Frazil Ice Surface heat flux

11 RASM Simulated Mixed Layer Depth Matches Observations Reasonably Well : Average deepest mixed layer depth November 1 March 31 m (DuVivier et al., 2016)

12 SOM Identifies Range In Types of Wind Patterns Around Greenland Easterly tip jets Westerly tip jets Strong Barrier Flow Weak Barrier Flow (DuVivier et al., 2016)

13 Patterns with barrier flow and westerly tip jets have largest buoyancy fluxes. (DuVivier et al., 2016)

14 Westerly tip jets have strong positive correlation between frequency and seasonal deepest mixed layer depth (DuVivier et al., 2016)

15 RASM WRF v3.2 Configuration 50 km horizontal grid, 40 vertical levels Physics: Microphysics: Morrison Droplet concentration set to 200 cm -3 over land, 50 cm -3 over water / ice Radiation: RRTMG LW and SW Calculate radiative fluxes based on droplet size from Morrison microphysics PBL: YSU or MYNN PBL Surface layer: MM5 Cumulus: Grell-Devenyi or Kain-Fritsch Land: VIC Spectral nudging of wind and T (top 20 levels only)

16 RASM Domains and Evaluation Regions

17 RASM: Sea Level Pressure Sea-level pressure (atmospheric circulation) is well simulated (RASM 1.0 Atmosphere; Cassano et al., in review)

18 RASM: Surface Temperature DJF: A large cold bias over the Arctic Ocean and adjacent land areas, also a cold SST bias in North Pacific JJA: A cold SST bias and slightly warm bias (RASM 1.0 Atmosphere; Cassano et al., in review)

19 RASM: Precipitation Overall dry bias in both DJF and JJA Extended ocean domain indicates precip bias is tied to SST bias (RASM 1.0 Atmosphere; Cassano et al., in review)

20 RASM: Surface Downward SW Radiation There is a negative SWD across the N. Pacific and N. Atlantic, especially in JJA There is a positive SWD across the mid-latitude land (RASM 1.0 Atmosphere; Cassano et al., in review)

21 Addressing RASM Cold SST Bias Large SST cold bias in Pacific and Atlantic Oceans, especially in summer Due to excessive simulated cloud cover over these oceans Cold SSTs lead to reduced evaporation over oceans Results in negative precipitation bias across much of the domain To resolve these biases, an alternate version of RASM was run: Uses MYNN PBL instead of YSU PBL Uses Kain-Fritsch cumulus instead of Grell-Devenyi cumulus

22 RASM: Surface Downward SW Radiation

23 North Pacific and Lena TOA and Surface SW Radiation Pacific Lena (original) (revised)

24 RASM: Surface Temperature

25 RASM: Precipitation

26 RASM: Sea Ice Thickness Spring

27 RASM: Sea Ice Thickness Fall

28 RASM: WRF Standalone Simulations Running fully coupled RASM simulations require a large amount of computing and time resources Doing large numbers of physics sensitivity tests is unrealistic Stand-alone WRF simulations are run to understand the sensitivity of select physics parameterizations Intelligently choose evaluations of different physics paramterizations: Cumulus, Microphysics, Boundary Layer, Radiation Review the potential benefits for upgrading WRF in RASM to v3.7.1 to take advantage of new physics options and other model improvements (e.g. cu_rad_feedback)

29 RASM Atmosphere: Conclusions RASM climate shows significant sensitivity to WRF physical parameterization choices Changing PBL and cumulus parameterizations changes the sign of large cloud / radiation biases Improves ocean simulation but degrades land simulation These biases have a large impact on other aspects of the simulated climate including: Surface temperature (resulting impacts on evaporation) Precipitation Sea ice thickness and extent Next steps Identify WRF physics options that produce realistic cloud cover and radiative fluxes across entire model domain Run coupled RASM simulations with modified WRF physics

30 RASM Summary Arctic climate predictive models need to: Resolve critical processes (e.g. eddies, sea ice deformation, inertial motions, barrier winds, polar lows) and resulting feedbacks (air-ice-ocean coupling) Understand space dependence & optimize parameter space Expand validation data (e.g. fluxes and coupling across the airice-ocean interface) Reduce computational cost / guide requirements of future highresolution coupled climate simulations RASM a tool toward a climate model hierarchy to: Resolve / understand Arctic processes and feedbacks, Reduce uncertainty and Improve prediction

31 RASM Benefits to ESM Community The practices and lessons learned from RASM, the application of a regional coupled model, have benefits to the wider ESM community: Modifications and improvements in CICE based on RASM studies have been implemented by the wider community Improvements in coupling configuration of CESM as a result of studies done using RASM The ability to switch out atmospheric physics parameterizations provides greater flexibility and evaluation than provided in GCMs Developing a great understanding of the key physical processes that need to be represesented in ESMs.

32 RASM Future Plans Parameter space sensitivity studies in fully coupled RASM Updated model components: WRF v3.7.1 VIC5 CICE ColPkg Seasonal climate prediction: Alternative BCs for WRF, e.g. CFSv2 21st century global climate model scenarios (pending) Ensemble generation in RASM Initial perturbed conditions Varying BCs Parameter sensitivity

33 RASM Future Plans Higher resolution RASM component model configurations 25 & 10-km WRF / VIC 1/48º (~2.4 km) POP / CICE New components: ecosystem / marine biogeochemistry

34 RASM Mark Seefeldt John Cassano Acknowledgments: Research supported by the Department of Energy