Search for: All records

Abstract This paper proposes algorithms for estimating parameters in Earth System Models (ESMs), specifically focusing on simulations that have not yet achieved statistical equilibrium and display climate drift. The basic idea is to treat ESM time series as outputs of an autoregressive process, with parameters that depend on those of the ESM. The maximum likelihood estimate of the parameters and the associated uncertainties are derived. This method requires solving a nonlinear system of equations and often results in unsatisfactory parameter estimates, especially in short simulations. This paper explores a strategy for overcoming this limitation by dividing the estimation process into two linear phases. This algorithm is applied to estimate parameters in the convection scheme of the Community Earth System Model version 2 (CESM2). The modified algorithm can produce accurate estimates from perturbation runs as short as 2 years, including those exhibiting climate drift. Despite accounting for climate drift, the accuracy of these estimates is comparable to that of algorithms that do not. While these initial results are not optimal, the autoregressive approach presented here remains a promising strategy for model tuning since it explicitly accounts for climate drift in a rigorous statistical framework. The current performance issues are believed to be technical in nature and potentially solvable through further investigation. 

Abstract The impact of interactive ocean dynamics on internal variations of Atlantic sea surface temperature (SST) is investigated by comparing preindustrial control simulations of a fully coupled atmosphere–ocean–ice model to the same atmosphere–ice model with the ocean replaced by a motionless slab layer (henceforth slab ocean model). Differences in SST variability between the two models are diagnosed by an optimization technique that finds components whose variance differs as much as possible. This technique reveals that Atlantic SST variability differs significantly between the two models. The two components with the most extreme enhancement of SST variance in the slab ocean model resemble the tripole SST pattern associated with the North Atlantic Oscillation (NAO) and the Atlantic multidecadal variability (AMV) pattern. This result supports previous claims that ocean dynamics are not necessary for the AMV, although ocean dynamics lead to slight increases in the memory of both the AMV and the NAO tripole. The component with the most extreme enhancement of SST variance in the fully coupled model resembles the Atlantic Niño pattern, confirming the ability of our technique to isolate physical modes known to require ocean dynamics. The second component with more variance in the fully coupled model is a mode of subpolar SST variability. Both the reemergence of SST anomalies and changes in ocean heat transport lead to increased SST variance and memory in the subpolar Atlantic. Despite large differences in the mean and variability of SST, atmospheric variability is quite similar between the two models, confirming that most atmospheric variability is generated by internal atmospheric dynamics. 

Abstract This paper derives statistical models for predicting wintertime subseasonal temperature over the western US. The statistical models are trained on two separate datasets, namely observations and dynamical model simulations, and are based on least absolute shrinkage and selection operator (lasso). Surprisingly, statistical models trained on dynamical model simulations can predict observations better than observation-trained models. One reason for this is that simulations involve orders of magnitude more data than observational datasets. 

Abstract This paper proposes an approach to diagnosing the skill of a machine-learning prediction model based on finding combinations of variables that minimize the normalized mean square error of the predictions. This technique is attractive because it compresses the positive skill of a forecast model into the smallest number of components. The resulting components can then be analyzed much like principal components, including the construction of regression maps for investigating sources of skill. The technique is illustrated with a machine-learning model of week 3–4 predictions of western US wintertime surface temperatures. The technique reveals at least two patterns of large-scale temperature variations that are skillfully predicted. The predictability of these patterns is generally consistent between climate model simulations and observations. The predictability is determined largely by sea surface temperature variations in the Pacific, particularly the region associated with the El Nino-Southern Oscillation. This result is not surprising, but the fact that it emerges naturally from the technique demonstrates that the technique can be helpful in “explaining” the source of predictability in machine-learning models. 

This paper shows that skillful week 3–4 predictions of a large-scale pattern of 2 m temperature over the US can be made based on the Nino3.4 index alone, where skillful is defined to be better than climatology. To find more skillful regression models, this paper explores various machine learning strategies (e.g., ridge regression and lasso), including those trained on observations and on climate model output. It is found that regression models trained on climate model output yield more skillful predictions than regression models trained on observations, presumably because of the larger training sample. Nevertheless, the skill of the best machine learning models are only modestly better than ordinary least squares based on the Nino3.4 index. Importantly, this fact is difficult to infer from the parameters of the machine learning model because very different parameter sets can produce virtually identical predictions. For this reason, attempts to interpret the source of predictability from the machine learning model can be very misleading. The skill of machine learning models also are compared to those of a fully coupled dynamical model, CFSv2. The results depend on the skill measure: for mean square error, the dynamical model is slightly worse than the machine learning models; for correlation skill, the dynamical model is only modestly better than machine learning models or the Nino3.4 index. In summary, the best predictions of the large-scale pattern come from machine learning models trained on long climate simulations, but the skill is only modestly better than predictions based on the Nino3.4 index alone. 

This paper derives a criterion for deciding conditional independence that is consistent with small-sample corrections of Akaike's information criterion but is easier to apply to such problems as selecting variables in canonical correlation analysis and selecting graphical models. The criterion reduces to mutual information when the assumed distribution equals the true distribution; hence, it is called mutual information criterion (MIC). Although small-sample Kullback–Leibler criteria for these selection problems have been proposed previously, some of which are not widely known, MIC is strikingly more direct to derive and apply.