CLMCUBE - A perturbed land surface parameter experiment · 2010. 7. 17. · CESM workshop, June...

CESM workshop, June 2010

Erich Fischererich.fischer@env.ethz.ch

CLMCUBE - A perturbed landsurface parameter experiment

Erich Fischer1,2

1 Institute for Atmospheric and Climate Science, ETH Zurich, Switzerland2 Climate & Global Dynamics, NCAR, Boulder, CO

Thanks to: Dave Lawrence, Ben Sanderson, Keith Oleson and Jerry Meehl

Introduction and Motivation



Motivation

• Land surface processes/feedbacks act on subgridscale and need to be parameterized

• Important source of uncertainty in climate projections

Objective

• Quantifying uncertainties in temperature and precipitation extremes induced by land surface parameterizations

Experimental setup



Experimental setup

Experimental setup



Experimental setup• CCSM 3.5 with mixed-layer

ocean

• Perturbation of 5 poorly constrained CLM parameters

• Different combinations of perturbed parameters -> 108 ensemble members

• Simulations with 1xCO2 (30 years) and 2xCO2 (20 years)

Results



Response to 2xCO2

Results



Global land temperature response

ΔTempLand = 2.65K

Results



Global land temperature response

ΔTempLand = 2.65K

Stainforth et al. 2005

climateprediction.net

Range smaller than in PPEs focusing on atmospheric parameter uncertainties

Uncertainties in CAMCUBE are only 2.5 times larger than CLMCUBE

Results



Temperature response to 2xCO2

Uncertainty (ensemble range)Ensemble mean response

Results



Low snow albedo vs. high snow albedo

Models with low present-day snow albedo show weaker warming

Results



Snow albedo feedback

Present-day snow albedo can be used to constrain ΔTemp (cf. Qu and Hall 2006, Levis et al. 2008)

Results



DJF variability response to 2xCO2 Variability reduction over strongest snow melting!

Mechanisms: (1) change in surface energy budget, (2) snow/no snow transition, (3) synoptic variability

Results



Role of reduced DJF variability

Reduction in variability amplifies change in cold extremes

Models with strong DJF warming show strong variability reduction

1xCO2

2xCO2 model 1

2xCO2 model 2

Results



Cold extremes are amplified

• More DJF warming-> amplified warming of cold extremesDJF 5th percentile

DJF 95th percentile

Results



Uncertainties in cold extremes (DJF 5th perc.)

Uncertainties regionally exceed the uncertainties of 6 CMIP 3 models providing daily output

Conclusions



Conclusions

• Land surface parameterizations induce major uncertainties in response of temperature extremes

• Variability response to 2xCO2 is particularly sensitive to land surface parameters

• Variability and mean response often relate to similar processes -> uncertainties add up

• Hot extremes show somewhat smaller uncertainties• Response for precipitation-related indices, e.g. consecutive

dry summer days differs in sign over Mediterranean, Australia, Central North AmericaReferences: • Fischer, E.M., D. Lawrence, B. Sanderson, 2010, submitted to Clim. Dynam.



Summer precipitation and droughts



Precipitation response to 2xCO2 (JJA)

Fischer et al. 2009c, in prep

Precipitation response varies between -17% and +12%

• Parameters explaining highest variance are identified

• Model developers will constrain sensitive parameters



Consecutive dry days (



What mechanisms are driving DJF variability?• Warm models with low

snow cover and warm temperatures-> low DJF variability

• More DJF warming-> lower variability



Present-day climate



Role of parameters

Region Expl. Var. Veg. alb. Snow alb. Wat. table Vcmax Rough. l.CN America 96% 62% 2% 13% 19% 1%

Australia (DJF) 65% 15% 0% 40% 11% 0%

N. Asia 96% 34% 38% 9% 16% 0%

• Parameters add nearly linearly in present-day (not for response to 2xCO2)

• Vmax and Wat. Table important over dry mid-latitudes



Experimental setupParameter perturbations• Selected land surface parameters are poorly constrained by

observations• Perturbations should be justifiable and if possible based on literature

and/or observational studies• Parameters include 2 optical properties, 2 properties affecting water

cycle and 1 turbulent property

Caveats• Experiment covers by far not full range of uncertainty!• Ocean circulation feedbacks are not accounted for• Initial condition uncertainties are not systematically addressed• Extremes are ‘not very extreme’



Momentum roughness length (doubled, to values used in the ECMWF Tessel)



How well is variability simulated in CCSM3.5?

Overestimation over middle East



How well is variability simulated in CCSM3.5?

Scale and shape of distribution are remarkably well captured

HadOBSCCSM3.5



JJA variability response to 2xCO2• Variability increase

associated with soil drying

• More JJA warming-> higher variability



What mechanisms are driving JJA variability?• Interannual var:

atmospheric circulation and soil moisture

• Day-to-day var: Relative humidity points to a combined effect of moisture and temperature



Latent cooling as a damping factor• Ratio of latent to radiative

cooling (Gregory et al. 1995)



Hot extremes (Summer 95th percentile)



Heat wave duration in summer half-year

• Large uncertainties due to 2 uncertain variables



Number of tropical nights

• Small sensitivity of night-time temperatures to land surface parameters



Vcmax (max. of carboxylation)

Vcmax: maximum carboxylation capacity of Rubisco at 25°C, which controls photosynthesis and affects transpiration (Thornton et al. 2007)

SLA: specific leaf area, ratio of leaf area to leaf mass

CNL: leaf carbon:nitrogenratio (gC gN-1)

Perturbations based on 1 standard deviation given in White et al. (2000)



Ensemble range for JJA precipitation



Vegetation albedo

Vegetation albedo perturbed by +/-20%, whichrepresents approx. the maximumregional bias



Low veg. albedo vs. CTL (JJA)

Annual global land energy budget:net radiat.: +2.0 W/m2latent heat: +0.7 W/m2sensible heat: +1.3 W/m2

Changes in net radiation mainly over crop and grass

(largest perturbation)



Snow albedo – snow aging



Snow albedo (near-infrared) – snow aging



Snow albedo – snow aging

Global land net radiation -0.64 W/m2

Effects strongest in spring and over Tibetan Plateau



Decay factor affecting water table depthWater table depth

Latent heat flux



Roughness length

PFT ztop [m] z0m CLM3.5 z0m ECMWFNET temperate 17 0.935 2NET boreal 17 0.935 2NDT boreal 14 0.77 2BET tropical 35 2.625 4BET temperate 35 2.625 4BDT tropical 18 0.99 2BDT temperate 20 1.1 2BDT boreal 20 1.1 2BES 0.5 0.06 0.1BDS temperate 0.5 0.06 0.1BDS boreal 0.5 0.06 0.1C3 Grass 0.5 0.06 0.02-0.05C3 non-arctic grass 0.5 0.06 0.02-0.05C4 Grass 0.5 0.06 0.1Corn 0.5 0.06 0.15Wheat 0.5 0.06 0.15



Roughness lengthSensible heat flux

Mean wind speed



Perturbed land parameter experiment



Regional change of 95th perc. (JJA)

• In contrast to current climate state, the response two 2xCO2 depends on parameters in a highly non-linear way!



Vcmax (perturbed vs. CTL)

Annual global energy budget:latent heat -1.95W/m2transpiration -3.76W/m2ground evap. +1.70W/m2sensible heat +2.11W/m2

Half of the reduction in transpiration is

compensated by enhanced ground evaporation

Precipitation is substantially reduced over some regions



Global annual land temperature

1xCO2 2xCO2



95th percentile of daily JJA temperatures

• Ensemble range is larger for temperature extremes

• LSM parameters affect not only mean but also temperature variability



Coupling CLM pert. exp. vs. CMIP3

• Range of correlations in CLM ensemble exceeds the range of CMIP3 models!

• Note that simulation length is much shorter (-> larger spread)

Notaro et al. 2008



Soil moisture-precipitation coupling

• Very simple approach to determine soil moisture-precipitation coupling indicates large range of coupling strengths

Soil moisture leads Precip leads

CLMCUBE - A perturbed land�surface parameter experiment��Erich Fischer1,2��1 Institute for Atmospheric and Climate Science, ETH Zurich, Switzerland� 2 Climate & Global Dynamics, NCAR, Boulder, CO��Thanks to: Dave Lawrence, Ben Sanderson, Keith Oleson and Jerry MeehlMotivationExperimental setupExperimental setupResponse to 2xCO2Global land temperature responseGlobal land temperature responseTemperature response to 2xCO2Low snow albedo vs. high snow albedoSnow albedo feedbackDJF variability response to 2xCO2Role of reduced DJF variabilityCold extremes are amplifiedUncertainties in cold extremes (DJF 5th perc.)ConclusionsSummer precipitation and droughtsPrecipitation response to 2xCO2 (JJA)Consecutive dry days (

CLMCUBE - A perturbed land surface parameter experiment · 2010. 7. 17. · CESM workshop, June...

Documents

Transcript of CLMCUBE - A perturbed land surface parameter experiment · 2010. 7. 17. · CESM workshop, June...

Perturbed State-Transition Matrix for Spacecraft Formation ...

Green's Functions in Perturbed Space-time

Assessing perturbed parameter ensembles as a tool for … · Assessing perturbed parameter ensembles as a tool for sampling model uncertainties and making climate projections James

The Global Atmospheric Circulation...The Global Atmospheric Circulation Exercise sheet 1 Due by 07.03.2019 Tutors: Jake Casselman (jake.casselman@env.ethz.ch) Bernat Jim enez (bernat.jimenez@env.ethz.ch)

LabVIEW Perturbed Particle Swarm Optimization Based ... · LabVIEW Perturbed Particle Swarm Optimization Based Approach for ... LabVIEW Perturbed Particle Swarm Optimization Based

Using perturbed parameter ensembles to quantify uncertainty in …€¦ · Quantifying Uncertainty in Model Projections (QUMP) •Investigate climate feedbacks to doubling CO2 •HadCM3

Intermittency in stochastically perturbed turbulent …people.fisica.uniroma2.it/~biferale/a32.pdfIntermittency in Stochastically Perturbed Turbulent Models L. Biferale, 1'2 M. Cencini,

Measure-perturbed one-dimensional Schrödinger operators · 2014-04-09 · Measure-perturbed one-dimensional Schrödinger operators Acontinuummodelforquasicrystals Dissertation ...

Hybrid Algorithm for Singularly Perturbed Delay Parabolic ...

Singularly perturbed boundary-equilibrium bifurcations

PART !: NEUTRON SPECTRUM PARAMETER iEASURE^ENTS · Relative gold-copper, lutetium-copper, or indium-manganese, lutetium-manganese activity ratios were determined in the perturbed

Multivariable PID Using Singularly Perturbed Systemeprints.utm.my/id/eprint/54205/1/...MultivariablePIDusingsingularly.pdf · Multivariable PID Using Singularly Perturbed System Mashitah

Research Paper Perturbed hematopoietic stem and progenitor ...

Computing Singularly Perturbed Di erential EquationsComputing Singularly Perturbed Di erential Equations Sabyasachi Chatterjeey Amit Acharyaz Zvi Artsteinx Abstract A computational

Evolutionary global optimization posed as a …Evolutionary global optimization posed as a randomly perturbed martingale problem and applied to parameter recovery of chaotic oscillators

Computing Singularly Perturbed Di erential Equations

More about extreme eigenvalues of perturbed random matricesbenaych/slides_conf_2010/Maida.pdf · More about extreme eigenvalues of perturbed random matrices ... 1

Physico -Chemical Properties of Perturbed water

A perturbed land surface parameter experiment with the ECMWF

Perturbed Microbiota/Immune Homeostasis in Multiple Sclerosis