Semi-automatic tuning of coupled climate models with multiple intrinsic timescales: lessons learned from the Lorenz96 model
This paper discusses a method for improving climate models by adjusting their parameters based on historical data. It uses a simplified model to demonstrate how this method works and emphasizes the importance of expert input in the process.
Content & Liability Disclaimer
This article and its accompanying video are automated summaries derived from the original research paper by Unknown authors. The original research was conducted solely by the paper's authors; PDFdigest did not conduct any of the research and makes no claims of ownership over the underlying scientific work.
The video narration is generated by artificial intelligence and references the paper's authors for attribution. The video is not narrated by any of the paper's authors. This content may contain inaccuracies, omissions, or misinterpretations of the original research. First-person language (e.g., "we found", "our results") reflects the original authors' voice, not PDFdigest's. Always read the original paper for accurate, verified information before making any decisions based on this content.
This content is provided "as is" without any warranties, express or implied. Simulated systems OÜ, its officers, directors, employees, and agents shall not be liable for any direct, indirect, incidental, special, consequential, or punitive damages arising from your use of, reliance on, or access to this content, including but not limited to errors, omissions, or misinterpretations of the original research. This disclaimer applies to the fullest extent permitted by applicable law.
- 1 As there are many measures of validation, objective methods can rapidly become expensive if they are all to be taken into account.
- 2 A second aspect of objective calibration that we consider, is the independence of metrics.
- 3 It serves as a simple test bed to investigate the performance of algorithms related to dynamical system forecasting, parameterization, data assimilation and more .
- 4 In the machine learning community, the process of using cheap emulators with the aim of derivative-free calibrating of costly black-box models, belongs to the field of surrogate modeling or meta-modeling .
Introduction
Climate models, or Earth system models (ESMs), have become a primary means of exploration of our changing climate. Numerical models of the Earth system were among the earliest applications of digital computing .
This soon gave rise both to numerical weather prediction, and studies of the climate, what John von Neumann called the “infinite forecast” , the statistics of weather fluctuations over long time periods.
The inclusion of the ocean circulation into the climate system, starting with Manabe & Bryan , also led to our first attempts to understand the radiative and thermal balance of the planet under changes in CO 2 concentration .
Research Question
It serves as a simple test bed to investigate the performance of algorithms related to dynamical system forecasting, parameterization, data assimilation and more . In the machine learning community, the process of using cheap emulators with the aim of derivative-free calibrating of costly black-box models, belongs to the field of surrogate modeling or meta-modeling .
Specifically, for the current paper, we aim to “history match” the metrics in Eq.
A second aspect of objective calibration that we consider, is the independence of metrics.
Methodology
A frequent method to validate the atmospheric component is to run experiments conforming to the Atmospheric Model Intercomparison Protocol . One recent approach applied for climate modeling is the Calibrate-Emulate-Sample method of the CliMA group where a limited set of runs of an atmospheric model is used for a broad characterization of an attractor in the parameter space of the model, by comparison with a reference, typically a model higher.
Study Design
In an alternate approach pioneered by the HighTune Group , a version of the History Matching (hereinafter noted HM) method developed by D.
In this method, the emulator is used in successive waves not to find optimal parameter values, but to eliminate implausible regions of parameter space, according to a chosen set of metrics (distance between model outputs and observations).
How PDFdigest Helps You Understand Research
Instant Paper Analysis
Get structured summaries and key findings from dense PDFs in seconds.
Visual Explanations
Turn complex methods, figures, and results into clearer visual breakdowns.
AI-Powered Q&A
Ask focused questions and get answers grounded in the paper.
Results & Findings
The anatomy of an ESM consists of several model components, such as the atmosphere, ocean, and land surface. The fundamental basis of an atmosphere or ocean model consists of solving the fluid dynamics of the medium, air or seawater, with the additional complexity of the myriad processes that govern the Earth system, including water in all its phases, as well as marine and terrestrial ecosystems.
- The anatomy of an ESM consists of several model components, such as the atmosphere, ocean, and land surface.
- The fundamental basis of an atmosphere or ocean model consists of solving the fluid dynamics of the medium, air or seawater, with the additional complexity of.
- Starting with the basic atmosphereocean general circulation model (AOGCM) of Manabe & Bryan , ESMs have steadily grown in complexity, adding more and more detail in.
- 1 in NASEM , synthesizing model evolution over several IPCC assessment reports).
- This scale range, in conjunction with complex physical, chemical and biological processes that govern the planet, make the construction of ESMs a challenging problem indeed.
When tuning the IPSL climate model for CMIP6, it was actually intentional to compensate for biases in the atmosphere, that initially limited dense water formation, by tuning the sea-ice leads fraction and finally enhance ventilation of the deep ocean .
Also when tuning a climate model, metrics refer to observations, that may encompass intrinsic uncertainty which should be taken into account, as well, this limitation cannot be envisioned in a perfect model setting.
Practical Applications
We finally discuss, in Section 4, the lessons learned from applying HM to L96 and the open research avenues that could lead to an efficient application of HM unto coupled climate models. Alternatively, a finite number of metrics may be allowed to have implausibility higher than 3.
In the aforementioned example, each bin has a restricted space for b and c but contains all the possible values of a.
The iterative aspect of the HM provides a certain flexibility that other approaches may not.
In addition, a difficulty arises with multi-wave design right after the first wave : LHS cannot sample the NROY space as it is in general not a hyper-rectangle anymore and may contain several disconnected regions.
The L96 model
The L96 model, introduced by Edward Lorenz, serves as a simplified test bed for investigating algorithms related to forecasting and parameterization. It consists of slow and fast variables with specific equations governing their dynamics.
Space filling design
Space filling design techniques are essential for efficiently sampling parameter space in experiments. The paper discusses the use of maximin Latin Hypercubes Sampling (LHS) to achieve a uniform representation of the parameter space.
Frequently Asked Questions
A second aspect of objective calibration that we consider, is the independence of metrics. As there are many measures of validation, objective methods can rapidly become expensive if they are all to be taken into account.
A frequent method to validate the atmospheric component is to run experiments conforming to the Atmospheric Model Intercomparison Protocol . HM is a well published and established method used in several science and engineering applications such as galaxy formation models and infectious.
In the “traditional” approach to model calibration, described above, the model is run forward for a sufficient time for a given choice of parameters to validate it against observations. The emulator is then used to perform the explorations of parameter space for.
Alternatively, a finite number of metrics may be allowed to have implausibility higher than 3. Overall, the stopping criteria are problem-dependent, and the limitation of computational resources most often come to play.
Overall, the stopping criteria are problem-dependent, and the limitation of computational resources most often come to play. When tuning the IPSL climate model for CMIP6, it was actually intentional to compensate for biases in the atmosphere, that initially limited dense water formation.
This paper discusses a method for improving climate models by adjusting their parameters based on historical data. It uses a simplified model to demonstrate how this method works and emphasizes the importance of expert input in the process.