Modeling – Center for Earth System Modeling, Analysis, and Data

Introduction

We are providing help on ideas on how to use climate models to explore specific science questions that require more intensive analysis within the climate systems. These processes could be in the atmosphere, oceans, land or ice systems and also include biogeochemistry. Our goal is to work with interdisciplinary groups to provide “appropriate” models that will address your research needs. These models can range from traditional climate system models (e.g., atmosphere-only or coupled atmosphere-ocean-land models) to the most complicated Earth system models (e.g., fully coupled atmosphere, ocean, land, ice, vegetation, and biogeochemical systems models). We also use models with different levels of model complexity (e.g., a single column atmosphere-landSurface model or a zonal mean Earth System model) or with multiple configurations using different model components. To help you further, …

Current Models

Community Earth System Models (NCAR)
Single Column CESM
MIT MESM
and others

Descripton of Earth System Model research

Modeling the Earth System has multiple research areas that are now linked in the current era.

Numerical Earth System Models

In the first research area, we have the Earth System Models (ESMs). ESMs are numerical models that solve the set of coupled climate system equations that govern the dynamics of the atmosphere, ocean, land, ice, vegetation, and biogeochemical systems. At their core, these models are solving a discretized set of coupled partial differential equations and require the largest high performance computer (HPC) platforms to simulate the Earth system behavior over periods from seasons to decades to millenia. These Earth System Models provide outputs that represent possible states of the Earth which can be used to for many purposes. A few of these are: estimating the state of Earth’s climate responding to the industrial era anthropogenic forcings, estimating paleoclimates during prehistoric eras (e.g., past glacial periods, paleocene-eocene thermal optimum, snow-ball Earth conditions, and others), and investigating climate system processes to understand interactions across the different components. These numerical models are currently saved on remote data servers with include typically 100+ Petabytes of datafiles.

Statistical Earth System Models

In the second research area, we have statistical models. In climate research, at least two classes of statistical models should be considered. One model class is used to perform state estimation; a second class is used for estimating predictions. The state estimation models are typically used to understand past data (e.g., model outputs or observational data) and to generate a statistics that represent possible states of the world for a given situation. As an example, we often define a climatological mean for a region and examine the distribution of deviations from that mean state. Together, the mean state and the deviations describe the observed data for a specific region. We can simplify it further by estimating the standard deviation statistic. If we consider future climate changes, we can also use this approach and predict the change in the mean or a change in the distribution of the deviations. Similarly, we could characterize the distributions of of extreme values to estimate how they have changed or will change in the future.

The second class of statistical models is statistical prediction models. For these models, we develop a statistical model that uses past data to predict a specific future event. Based on past data, the simplest statistical model would only use past observations to make predictions. In more sophisticated approaches, a statistical model will be trained using multiple data sets including historical observations as well as past numerical model outputs. As an example of this, we could use simulations of historical climate change from multiple models, compare them with past observations, and then base a prediction on those statistics. Another approach would be to compare model predictions against past climate and use data assimilation methods to calibrate predictive numerical model.

Contact Info