Dynamically combining climate models to “supermodel” the tropical Pacific

Shen, Meiyu; Keenlyside, Noel; Selten, Frank; Wiegerinck, Wim; Duane, Gregory S.

doi:10.1002/2015gl066562

Cited by 40 publications

(29 citation statements)

References 28 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The CPT learning approach is based on an idea proposed by Smith (2001). CPT "crosses" trajectories of different models in order to create a larger solution space.…”

Section: Cross Pollination In Timementioning

confidence: 99%

Improving weather and climate predictions by training of supermodels

et al. 2019

Self Cite

View full text Add to dashboard Cite

Recent studies demonstrate that weather and climate predictions potentially improve by dynamically combining different models into a so-called "supermodel". Here, we focus on the weighted supermodel -the supermodel's time derivative is a weighted superposition of the time derivatives of the imperfect models, referred to as weighted supermodeling. A crucial step is to train the weights of the supermodel on the basis of historical observations. Here, we apply two different training methods to a supermodel of up to four different versions of the global atmosphere-ocean-land model SPEEDO. The standard version is regarded as truth. The first training method is based on an idea called cross pollination in time (CPT), where models exchange states during the training. The second method is a synchronization-based learning rule, originally developed for parameter estimation. We demonstrate that both training methods yield climate simulations and weather predictions of superior quality as compared to the individual model versions. Supermodel predictions also outperform predictions based on the commonly used multi-model ensemble (MME) mean. Furthermore, we find evidence that negative weights can improve predictions in cases where model errors do not cancel (for instance, all models are warm with respect to the truth). In principle, the proposed training schemes are applicable to state-of-the-art models and historical observations. A prime advantage of the proposed training schemes is that in the present context relatively short training periods suffice to find good solutions. Additional work needs to be done to assess the limitations due to incomplete and noisy data, to combine models that are structurally different (different resolution and state representation, for instance) and to evaluate cases for which the truth falls outside of the model class.tation of the physical processes the model is intended to describe. Nevertheless, with the best possible models in hand, more accurate predictions can be obtained by making good use of all of them, thus exploiting multi-model information.In order to reduce the impact of model errors on predictions, it is common practice to combine the predictions of a collection of different models in a statistical fashion. This is referred to as the multi-model ensemble (MME) approach: the MME mean prediction is often more skillful as model errors tend to average out (Weigel et al., 2008), whereas the spread between the model predictions is naturally interpreted as a Published by Copernicus Publications on behalf of the European Geosciences Union.

show abstract

“…The CPT learning approach is based on an idea proposed by Smith (2001). CPT "crosses" trajectories of different models in order to create a larger solution space.…”

Section: Cross Pollination In Timementioning

confidence: 99%

Improving weather and climate predictions by training of supermodels

et al. 2019

Self Cite

View full text Add to dashboard Cite

show abstract

“…There is evidence in Rodwell and Palmer (2007) that climatological errors develop quickly during the first few days of weather forecasts, implying that short-term training can reduce climatological errors. If this result carries over to the state-of-the-art models, then computationally expensive long climate simulations as in Shen et al (2016) can be avoided during training. Using relatively short integrations only can still improve the climatology of a supermodel.…”

Section: Discussionmentioning

confidence: 99%

An efficient training scheme for supermodels

Schevenhoven

Selten

2017

Earth Syst. Dynam.

View full text Add to dashboard Cite

Abstract.Weather and climate models have improved steadily over time as witnessed by objective skill scores, although significant model errors remain. Given these imperfect models, predictions might be improved by combining them dynamically into a so-called "supermodel". In this paper a new training scheme to construct such a supermodel is explored using a technique called cross pollination in time (CPT). In the CPT approach the models exchange states during the prediction. The number of possible predictions grows quickly with time, and a strategy to retain only a small number of predictions, called pruning, needs to be developed. The method is explored using low-order dynamical systems and applied to a global atmospheric model. The results indicate that the CPT training is efficient and leads to a supermodel with improved forecast quality as compared to the individual models. Due to its computational efficiency, the technique is suited for application to state-of-the art high-dimensional weather and climate models.

show abstract

“…Next to other model development ideas like the “Supermodel” in Shen et al . [], the main ideas of the EDF could live up in other techniques like combining neural networks with numerical models. The synchronization of members in terms of information exchange could be a valuable add‐on.…”

Section: Discussionmentioning

confidence: 99%

Decadal climate predictions improved by ocean ensemble dispersion filtering

Kadow

Illing

Kröner

et al. 2017

J Adv Model Earth Syst

View full text Add to dashboard Cite

Decadal predictions by Earth system models aim to capture the state and phase of the climate several years in advance. Atmosphere‐ocean interaction plays an important role for such climate forecasts. While short‐term weather forecasts represent an initial value problem and long‐term climate projections represent a boundary condition problem, the decadal climate prediction falls in‐between these two time scales. In recent years, more precise initialization techniques of coupled Earth system models and increased ensemble sizes have improved decadal predictions. However, climate models in general start losing the initialized signal and its predictive skill from one forecast year to the next. Here we show that the climate prediction skill of an Earth system model can be improved by a shift of the ocean state toward the ensemble mean of its individual members at seasonal intervals. We found that this procedure, called ensemble dispersion filter, results in more accurate results than the standard decadal prediction. Global mean and regional temperature, precipitation, and winter cyclone predictions show an increased skill up to 5 years ahead. Furthermore, the novel technique outperforms predictions with larger ensembles and higher resolution. Our results demonstrate how decadal climate predictions benefit from ocean ensemble dispersion filtering toward the ensemble mean.

show abstract

Dynamically combining climate models to “supermodel” the tropical Pacific

Cited by 40 publications

References 28 publications

Improving weather and climate predictions by training of supermodels

Improving weather and climate predictions by training of supermodels

An efficient training scheme for supermodels

Decadal climate predictions improved by ocean ensemble dispersion filtering

Contact Info

Product

Resources

About