ATM 298, Spring 2013 Lecture 15
Model Intercomparison and EvaluaAon
May 29, 2013
Paul A. Ullrich (HH 251) [email protected]
Sources of Uncertainty in GCMs
Structural Uncertainty Choice of dynamical core
Choice of physical parameterizaGons Model resoluGon (horizontal and verGcal)
…
Data Uncertainty IniGal data ObservaGonal error Boundary data (SSTs)
Parameter Uncertainty Physics tuning Physical constants Diffusion coefficients
Test Hierarchy
2D Shallow Water Test
Cases
3D Dry Dynamical
Core Test Cases
3D Dynamical
Core + Simplified Physics Test
Cases
3D Aqua-‐
Planet Experiments
(APE)
3D
Atmospheric Model Intercompar
ison (AMIP)
DeterminisAc Tests StaAsAcal Tests
Held-‐Suarez Jablonowski and
Williamson Williamson et al.
Increasing complexity
Deformational Flow
(Advection Test)
2D Shallow Water Test Cases
Deformational flow on the sphere (tests accuracy of the numerical method, preservation of monotonicity and functional relationships)
Source: Ullrich, Jablonowski and van Leer (2010) “High-order finite-volume methods for the shallow-water
equations on the sphere.” J. Comp.
Phys.
Steady-state geostrophically balanced flow which is not
aligned with the grid. Errors are measured after five days against the initial state.
Williamson Test Case 2
2D Shallow Water Test Cases
Galewsky et al. Shallow-Water Barotropic Instability
Galewsky et al.: Geostropically balanced shallow water jet which is perturbed and leads to the development of a vorGcal instability.
2D Shallow Water Test Cases
MCore FVcubed
HOMME EUL
3D Dry Dynamical Core Test Cases
Baroclinic Instability
Source: Ullrich, Jablonowski and van Leer (2010) “High-order finite-volume methods for the shallow-water equations on the sphere.” J. Comp. Phys.
Flow over Topography
3D Dry Dynamical Core Test CasesSource: Ullrich, Jablonowski and van Leer (2010) “High-order finite-volume methods for the shallow-water equations on the sphere.” J. Comp. Phys.
Conservation of Invariants
3D Dry Dynamical Core Test CasesNon-Hydrostatic Mountain Waves
3D Dry Dynamical Core Test CasesTests the response of atmospheric models to topography in the non-hydrostatic regime.
Held-Suarez Climatology
3D Dynamical Core +Simplified Physics Test Cases
Held-‐Suarez: HeaGng is prescribed plus a simple velocity relaxaGon scheme.
Aqua-Planet Experiments
Zonal mean 3-‐year mean zonal wind: Snapshots of 4 GCMs that parGcipated in the Aqua-‐Planet Experiment (APE).
3D Aqua-‐Planet Experiments (APE)
3D aqua planet tests evaluate the interacGon between the dynamical core and complex physical parameterizaGons using a simplified lower boundary (flat ocean-‐
covered Earth with analyGcally prescribed sea-‐surface temperatures (SSTs)
Source: Williamson et al., NCAR
Aqua-Planet Experiments
3D Aqua-‐Planet Experiments (APE) 3D aqua planet tests give insights into the characterisGcs of moisture processes.What drives these differences?
AMIP Simulations
3D Atmospheric Model Intercomparison (AMIP)3D AMIP tests evaluate the interacGon between the dynamical core and complex physical parameterizaGons (maybe even including chemistry packages) using a complex but prescribed lower boundary (orography, prescribed observaGon-‐based SSTs and sea-‐ice) over 25-‐year Gme frames
AMIP Simulations
Michael Wehner et al.:
Total column-‐integrated water vapor.
hip://www.youtube.com/watch?v=MrRpSzHkx40
1979 Hurricane Season:
Total column-‐integrated water vapor.
hip://www.youtube.com/watch?feature=endscreen&v=VKoZCzlBoDk&NR=1
3D Atmospheric Model Intercomparison (AMIP)
Fully Coupled
• The most complex GCM evaluaGons uGlize a fully coupled atmosphere (ocean – ice – land – chemistry – carbon-‐cycle – Earth system)
someGmes with prescribed greenhouse gas concentraGons are used (CLIVAR runs)
• Fully coupled simulaGons of past Gme periods are typically compared against observaGons, someGmes in the form of re-‐analysis data
• Differences between simulaGons are very hard to understand due to the complexity and non-‐linear interacGons
• Fully coupled GCMs are used for the assessment of future climate
scenarios (e.g. for the Intergovernmental Panel on Climate Change, IPCC, assessments)
Fully Coupled SimulaAons
Measures of ‘Truth’
• What is truth? How can we judge whether global atmospheric model simulaGons are robust, reliable and accurate?
• The higher we go up the test hierarchy the more difficult it is to understand the causes and effects, and to determine the accuracy of the simulaGon.
• Only very idealized test cases have analyGcal soluGons.
• In non-‐linear dry dynamical core test cases we rely on ensembles of high-‐
resoluGon reference soluGons to determine the perceived ‘truth’ and its uncertainty.
• Dry dynamical core tests converge within some uncertainty with increasing resoluGon.
Ensembles
• Ensembles are one way to assess the robustness of the simulaGons, and to gain insight into the uncertainty of the model simulaGons.
Perturbed Parameter Ensembles
• VariaGons of empirical tuning factors in the physical parameterizaGons
• Diffusion coefficients or physical constants in the dynamical core
IniAal Data and Boundary Value Ensembles
• Slight variaGons in the iniGal data
• Different topography datasets
• Different sea-‐surface temperatures MulA-‐Model Ensembles
• Different atmospheric models (or different versions of the same model)
Importance of Atmospheric Models
• Atmospheric models allow us to test our understanding of the physical system against observaGons.
• Atmospheric models are our primary tool for making predicGons on the future climate of Earth (10-‐100 year simulaGons).
• Atmospheric models can be thought of as scienGfic instruments that allow us to experiment with the Earth system (which would be impossible in pracGce).
Design of Earth-System Models
Earth-‐system models consist of dozens of interwoven parts, incorporaGng the vast base of knowledge we have developed around the Earth system:
• Different dynamical cores (solve the primiGve equaGons for quanGGes resolved on the grid scale)
• Sub-‐grid-‐scale dynamical parameterizaGons (turbulence closures, gravity wave drag for unresolved atmospheric moGons)
• Moist physics parameterizaGons (microphysics, macrophysics, precipitaGon, clouds)
• Chemistry (greenhouse gases, aerosols)
• Ocean, ice and land components
