Efficient screening of climate model sensitivity to a large number of perturbed input parameters

Covey, Curt; Lucas, D. D.; Tannahill, J.; Garaizar, Xabier; Klein, Richard

doi:10.1002/jame.20040

Cited by 31 publications

(51 citation statements)

References 46 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In practice, however, most parameterizations do not contain this information. The uncertainty ranges of the parameters in this study are based on Covey et al (2013) and previous CAM5 tuning exercises Qian et al, 2015). They are listed in Table 1.…”

Section: Model Parametersmentioning

confidence: 99%

Automatic tuning of the Community Atmospheric Model (CAM5) by using short-term hindcasts with an improved downhill simplex optimization method

Zhang

Lin

et al. 2018

Geosci. Model Dev.

View full text Add to dashboard Cite

Traditional trial-and-error tuning of uncertain parameters in global atmospheric general circulation models (GCMs) is time consuming and subjective. This study explores the feasibility of automatic optimization of GCM parameters for fast physics by using short-term hindcasts. An automatic workflow is described and applied to the Community Atmospheric Model (CAM5) to optimize several parameters in its cloud and convective parameterizations. We show that the auto-optimization leads to 10 % reduction of the overall bias in CAM5, which is already a well-calibrated model, based on a predefined metric that includes precipitation, temperature, humidity, and longwave/shortwave cloud forcing. The computational cost of the entire optimization procedure is about equivalent to a single 12-year atmospheric model simulation. The tuning reduces the large underestimation in the CAM5 longwave cloud forcing by decreasing the threshold relative humidity and the sedimentation velocity of ice crystals in the cloud schemes; it reduces the overestimation of precipitation by increasing the adjustment time in the convection scheme. The physical processes behind the tuned model performance for each targeted field are discussed. Limitations of the automatic tuning are described, including the slight deterioration in some targeted fields that reflect the structural errors of the model. It is pointed out that automatic tuning can be a viable supplement to processoriented model evaluations and improvement.

show abstract

Section: Model Parametersmentioning

confidence: 99%

Automatic tuning of the Community Atmospheric Model (CAM5) by using short-term hindcasts with an improved downhill simplex optimization method

Zhang

Lin

et al. 2018

Geosci. Model Dev.

View full text Add to dashboard Cite

show abstract

“…For comparison, with a prescribed SST, Covey et al (2013) evaluated the sensitivity of the radiation balance at the top of the atmosphere in CAM to a large number of input parameters with both an OAT exploration, similar to the pEKS, and an alternative Morris one-at-a-time (MOAT) sampling. For example, they found with both methods that the highest sensitivity of the upward shortwave flux (solar energy reflected back to space by the atmosphere and surface) was with respect to cldfrc_rhminl, out of 21 evaluated CAM parameters.…”

Section: Resultsmentioning

confidence: 99%

Parameter space Kalman smoothers for multi-decadal climate analysis in high resolution coupled Global Circulation Models

García-Pintado

Paul

2018

Preprint

View full text Add to dashboard Cite

Abstract. In climate reanalyses for multi-decadal or longer scales with coupled atmosphere-ocean General Circulation models (CGCMs) it can be assumed that the growth of prediction errors arises chiefly from imprecisely known model parameters, which have a nonlinear relationship with the climate observations (paleoclimate proxies). Also, high-resolution CGCMs for climate analysis are extremely expensive to run, which constrains the applicability of assimilation schemes. In a model framework where we assume that model dynamic parameters account for (nearly) all forecast errors at observation times, we compare 5 two computationally efficient iterative schemes for approximate nonlinear model parameter estimation and joint flux estimation (taking the specific shape of freshwater from melting in the Greenland ice sheet), and its physically consistent state. First, a trivial adaptation of the strong constraint incremental 4D-Var formulation leads to what we refer to as the parameter space iterative extended Kalman smoother (pIKS); a Gauss-Newton scheme. Second, a so-called parameter space fractional Kalman smoother (pFKS) is an alternative controlled-step line search, which can potentially be a more stable approach. While these 10 iterative schemes have been used in data assimilation, we revisit them together within the context of parameter estimation in climate reanalysis, as compared to the more general 4D-Var formulation. Then, the two schemes are evaluated in numerical experiments with a simple 1D energy balance model (Ebm1D) and with a fully-coupled Community Earth System Model (CESM v1.2). Firstly, with Ebm1D the pFKS obtains a cost function similar to the adjoint method with highly reduced computational cost, while an ensemble transform Kalman filter with an m = 60 ensemble size (ETKF 60 ) behaves slightly worse. The pIKS 15 behaves worse than the ETKF 60 , but an ETKF 10 (m = 10) is even worst. Accordingly, with CESM we evaluate the pKFS and the ETKF 60 along with an ETKF with Gaussian Anamorphosis (ETKF-GA 60 ). From all the options, the pFKS has the lowest cost function and seems the favored overall option under heavy computational restrictions, but the ETKF obtains better estimates of the flux term.

show abstract

“…In this study, parameter sensitivity in the fully coupled Community Earth System Model (CESM1) is examined for global warming projection runs, as well as historical climate, specifically for key parameters in the deep convection scheme. Insufficiently constrained parameters affecting deep convection are known to affect simulation of current climate and global average climate sensitivity [e.g., Neale et al ., ; Sanderson , ; Covey et al ., ; Guo et al ., ; Qian et al ., ] and are hypothesized here to contribute strongly to sensitivity in projections of regional precipitation change.…”

Section: Introductionmentioning

confidence: 97%

“…A number of these studies indicate that climate sensitivity in their PPE is similar to or greater than climate sensitivity in the multimodel ensembles [ Murphy et al ., ; Stainforth et al ., ; Collins et al ., ; Rowlands et al ., ; Yamazaki et al ., ] although it is not clear that PPE typically spans the range of the CMIP archive [ Yokohata et al ., , ]. Some have sought to reduce error in the model climatology by tuning parameter values or identifying more probable values based on error with respect to observed climatology [ Jones et al ., ; Severijns and Hazeleger , ; Jackson et al ., ; Zhang et al ., ; Covey et al ., ; Tett et al ., ].…”

Section: Introductionmentioning

confidence: 99%

Identifying sensitive ranges in global warming precipitation change dependence on convective parameters

Bernstein

Neelin

2016

Geophysical Research Letters

View full text Add to dashboard Cite

A branch‐run perturbed‐physics ensemble in the Community Earth System Model estimates impacts of parameters in the deep convection scheme on current hydroclimate and on end‐of‐century precipitation change projections under global warming. Regional precipitation change patterns prove highly sensitive to these parameters, especially in the tropics with local changes exceeding 3 mm/d, comparable to the magnitude of the predicted change and to differences in global warming predictions among the Coupled Model Intercomparison Project phase 5 models. This sensitivity is distributed nonlinearly across the feasible parameter range, notably in the low‐entrainment range of the parameter for turbulent entrainment in the deep convection scheme. This suggests that a useful target for parameter sensitivity studies is to identify such disproportionately sensitive “dangerous ranges.” The low‐entrainment range is used to illustrate the reduction in global warming regional precipitation sensitivity that could occur if this dangerous range can be excluded based on evidence from current climate.

show abstract

Efficient screening of climate model sensitivity to a large number of perturbed input parameters

Cited by 31 publications

References 46 publications

Automatic tuning of the Community Atmospheric Model (CAM5) by using short-term hindcasts with an improved downhill simplex optimization method

Automatic tuning of the Community Atmospheric Model (CAM5) by using short-term hindcasts with an improved downhill simplex optimization method

Parameter space Kalman smoothers for multi-decadal climate analysis in high resolution coupled Global Circulation Models

Identifying sensitive ranges in global warming precipitation change dependence on convective parameters

Contact Info

Product

Resources

About