Climate Modeling and Analysis

''This page is about the intersection of climate science and machine learning in the context of climate change adaptation. For an overview of climate science as a whole, please see the Wikipedia page on this topic.'' As described in "Tackling Climate Change with Machine Learning," The first global warming prediction was made in 1896, when Arrhenius estimated that burning fossil fuels could eventually release enough CO2 to warm the Earth by 5°C. The fundamental physics underlying those calculations has not changed, but our predictions have become far more detailed and precise. The predominant predictive tools are climate models, known as General Circulation Models (GCMs) or Earth System Models (ESMs). These models inform local and national government decisions  ), help people calculate their climate risks (see Policy, Markets, and Decision Science and Climate Change Adaptation) and allow us to estimate the potential impacts of solar geoengineering.

Recent trends have created opportunities for ML to advance the state-of-the-art in climate prediction. First, new and cheaper satellites are creating petabytes of climate observation data. Second, massive climate modeling projects are generating petabytes of simulated climate data. Third, climate forecasts are computationally expensive (some simulations have taken three weeks to run on NCAR supercomputers ), while ML methods are becoming increasingly fast to train and run, especially on next-generation computing hardware. As a result, climate scientists have recently begun to explore ML techniques, and are starting to team up with computer scientists to build new and exciting applications.

Uniting data, ML, and climate science

 * Data for climate models: Assimilation of diverse sources can improve climate models, and machine learning can transform raw sensor output into more relevant derived data. Well-curated benchmark datasets have the potential to advance several geoscience problems.
 * Accelerating climate models: Physical constraints are key ingredients for cloud, aerosol, ice sheet, and sea level models. Traditional solutions to these physics-based models are computationally expensive, but machine learning components can help alleviate the most problematic bottlenecks.
 * Working with climate models: It is possible to streamline existing climate models, pruning them down to key relationships simplifying computation with ensembles.

Forecasting extreme events

 * Storm tracking: While climate models can forecast long-term changes in the climate system, separate systems are required to detect specific extreme weather phenomena, like cyclones, atmospheric rivers, and tornadoes.
 * Local forecasts: Machine learning can be used to refine what are otherwise coarse-grained forecasts. These high-resolution forecasts can guide improvements in system robustness and resilience.

Textbooks

 * Introduction to climate dynamics and climate modeling (2010) : A technical treatment of the climate system, energy balance, climate modeling, and climate perturbations. Available here.
 * Principles of Planetary Climate (2010) : An introduction to the physics of climate, with examples in python.

Other

 * An Introduction to Climate Modeling (2014) : A video lesson from Climate Literacy's Youtube channel. Available here.
 * Oxford Research Encyclopedia of Climate Science : A collection of articles on the climate systems, impacts of climate change, and the methods used in climate science.

Major conferences

 * AGU Fall Meeting: A yearly conference organized by the American Geophysical Union. Website here.

Major journals
Climate science is a journal field. Noteworthy research appears in journals such as


 * Bulletin of the American Meteorological Society: A journal published by the AMS. Available here.
 * Geophysical Research Letters: The journal of the American Geophysical Union. Available here.
 * Proceedings of the National Academy of Sciences: A wide-reaching journal often featuring climate science. Available here.

Major societies and organizations

 * American Geophysical Union: An organization supporting work across the geophysical sciences. Website here.
 * Climate Informatics: An organization dedicated to computing in climate science. Website here.

Libraries and Tools
Pangeo: An open source python package for geoscience applications, available here.


 * Pangeo also maintains a list of packages useful for atmospheric, ocean, and climate science.

Data
The largest climate prediction datasets are ensembles of many climate simulations. These include simulations with varied physics, architectures, or initial conditions, and they are used to explore the range of possible climate futures. The largest ensembles include:


 * The Coupled Model Intercomparison Project (CMIP): A gateway to climate models in use and development, available here. CMIP is associated with the Earth System Grid Federation, which also provides data analysis tools and tutorials: https://esgf.llnl.gov/
 * The CESM Large Ensemble: Read about it in The Community Earth System Model (CESM) Large Ensemble Project. Available here.
 * Google Cloud Weather and Climate Datasets: Petabyte-scale weather and climate datasets from sources like NOAA’s NEXRAD and NASA/USGS’s Landsat, made available for free as part of Google Cloud’s Public Datasets Program. Available here.
 * EARTHDATA: NASA's gateway to earth science data. Data are available at multiple levels of processing. Available here.

N.B. Climate model data is typically presented in netcdf4 format. These may be smoothly converted to csv files or pandas dataframes, but be aware that the data lies on irregular 3D spherical grids.

The Earth and climate science community is also working to create benchmark datasets: https://is-geo.org/benchmarks/.