Climate Models Overview

CEVE 543 - Fall 2025

2025-10-15

Reflecting on Module 1

Think-pair-share (2 minutes)

What are potential limitations of GEV and nonstationary GEV methods for climate risk assessment?

Key limitations:

  • Require long observational records (often unavailable)
  • Parametric representation of nonstationarity
  • Events with multiple drivers create multimodal distributions
  • Model structure uncertainty

In Module 2 we will explore how climate models can mitigate these limitations.

Logistics

  • Logistics

  • Climate Risks and Variability

  • Climate Models Overview

  • Model Limitations and the Need for Statistical Methods

Problem Set 2

Due Wednesday, November 19th.

Topics:

  • Climate model simulations
  • Statistical downscaling
  • Bias correction

Focus on practice problems for exam preparation.

Proposed Syllabus Update

Proposal: Remove final exam, redistribute points.

Current

  • Problem sets: 10%
  • Labs: 10%
  • Module tests: 40%
  • Final project: 20%
  • Final exam: 20%

Proposed

  • Problem sets: 15%
  • Labs: 15%
  • Module tests: 50%
  • Final project: 20%

Module 2 Structure

Three-part weekly rhythm:

  1. Monday - lecture on statistical/computational methods (whiteboard)
  2. Wednesday - Paper Discussion (slides and class discussion)
  3. Friday - Virtual lab implementing subset of methods

Module 2 Schedule

Week Method Lab
Oct 15-27 Climate data overview Climate data in Julia
Oct 20-24 Bias Correction QQ-mapping
Oct 27-31 Weather Typing PCA + \(K\)-means
Nov 3-7 Hidden Markov models HMMs
Nov 10-14 Stochastic Weather Generators TBD
Nov 17-21 Sensitivity Analysis Exam

Climate Risks and Variability

  • Logistics

  • Climate Risks and Variability

  • Climate Models Overview

  • Model Limitations and the Need for Statistical Methods

Climate Risks Cluster

Spatial clustering → portfolio risk. Temporal clustering → insurance/reinsurance viability.

Time series of the yearly number of 30-day extreme rainfall events exceeding the 10-year return level for the Rio Tinto portfolio computed using the 20CR dataset, using 3 windows (11, 21, and 114 years) to illustrate different aspects of long-term variability (Bonnafous, Lall, and Siegel 2017)

Climate Risks Vary on Many Timescales

Multi-timescale variability is essential for design and planning.

Global wavelet analysis of two (L) living Blended Drought Analysis and (B) Naturalized American River at Folsom (Doss-Gollin et al. 2019)

Nonlinear Dynamics

  • Recurrence and persistence of weather patterns (“regimes”) drive many extremes
  • Study of low-dimensional chaotic systems provides insight

Lorenz (1963) attractor

(a) Segment of the jet stream latitude time series corresponding to the first ten winters: DJF 1957/1967. (b) Trajectory of the winter jet latitude time series in delay coordinates using a 10-day lag. (c) Histogram and PDFs of the jet stream latitude anomaly using a kernel density estimate (solid) and a three-component Gaussian mixture model (dashed). In (b) and (c) the whole time series (DJF 1957–2002) was used. (Hannachi, Woollings, and Fraedrich 2011)

Conceptual sketch of nonlinear dynamics (Palmer 1999)

Climate Models Overview

  • Logistics

  • Climate Risks and Variability

  • Climate Models Overview

  • Model Limitations and the Need for Statistical Methods

What is a Climate Model?

Global grid example (Credit: CliMA)

Numerical solution of PDEs on global grids.

Think-pair-share (1 minute)

What equations do climate models solve?

Weather vs Climate

Weather

  • Initial value problem
  • Specific atmospheric states
  • Days to weeks
  • Sensitive to initial conditions

Climate

  • Boundary value problem
  • Long-term statistics
  • Decades to centuries
  • Forced by boundary conditions

History

Key milestones (per CarbonBrief):

  • 1956: Norman Phillips creates first general circulation model
  • 1967: Manabe & Wetherald publish groundbreaking GCM study with doubled CO\(_2\)
  • 1979: Charney Report provides critical assessment
  • 1988: James Hansen presents climate scenarios to Congress
  • 1995: CMIP launched for model intercomparison

Evolution: energy balance models \(\rightarrow\) GCMs \(\rightarrow\) ESMs

Using Climate Model Simulations

Climate models are computationally expensive. We rarely run them ourselves—instead, we use existing simulations.

Coupled Model Intercomparison Project (CMIP): Coordinated effort where modeling centers run standardized experiments.

Key features:

  • Common scenarios: All models run same experiments (historical, future emissions pathways)
  • Multiple models: Compare structural differences across models
  • Ensembles: Multiple realizations per model to quantify internal variability

Discussion

Why coordinate all models to run the same scenarios rather than letting each center choose their own?

CMIP6

Sixth phase (2015-present) addresses three questions:

  1. How does the Earth system respond to forcing?
  2. What are the origins and consequences of systematic model biases?
  3. How can we assess future climate changes given variability and uncertainty?

Scale: 100+ models from ~50 institutions running 23 endorsed experiments.

Experiments include: Historical simulations (1850-2014), future scenarios (SSP1-1.9 through SSP5-8.5), detection & attribution, and specialized experiments.

AI/ML Models

  • Weather models: GraphCast, FourCastNet, Pangu-Weather
  • Foundation models: Aurora, ClimaX
  • Climate emulators: LUCIE, etc.

Discussion

What are the advantages and limitations of AI vs. physics-based models?

Model Limitations and the Need for Statistical Methods

  • Logistics

  • Climate Risks and Variability

  • Climate Models Overview

  • Model Limitations and the Need for Statistical Methods

Coarse Resolution

  • CMIP6 models don’t resolve tropical cyclones, mesoscale convection, etc.
  • Sub-grid processes always require parameterization.

IPCC

Systematic Biases

Discussion: Model Limitations vs. Value

Think-pair-share (3 minutes)

Suppose you are analyzing extreme rainfall in a region where rainfall is strongly modulated by ENSO, but climate models struggle to accurately simulate ENSO dynamics.

Should you use the model simulations for your analysis? What factors would influence your decision?

Key considerations:

  • What specific information do you need from the models?
  • Can you bias-correct or statistically adjust for ENSO biases?
  • Are there alternative approaches (e.g., weather typing, conditional sampling)?
  • What are the consequences of getting it wrong?

References

Bonnafous, Luc, Upmanu Lall, and Jason Siegel. 2017. “A Water Risk Index for Portfolio Exposure to Climatic Extremes: Conceptualization and an Application to the Mining Industry.” Hydrology and Earth System Sciences 21 (4): 2075–2106. https://doi.org/10/f96k67.
Doss-Gollin, James, David J. Farnham, Scott Steinschneider, and Upmanu Lall. 2019. “Robust Adaptation to Multiscale Climate Variability.” Earth’s Future 7 (7): 734–47. https://doi.org/10.1029/2019ef001154.
Hannachi, Abdel, Tim Woollings, and K Fraedrich. 2011. “The North Atlantic Jet Stream: A Look at Preferred Positions, Paths and Transitions.” Quarterly Journal of the Royal Meteorological Society 138 (665): 862–77. https://doi.org/10.1002/qj.959.
Lorenz, Edward N. 1963. “Deterministic Nonperiodic Flow.” Journal of the Atmospheric Sciences 20 (2): 130–41. https://doi.org/10.1175/1520-0469(1963)020<0130:dnf>2.0.co;2.
Palmer, T N. 1999. “A Nonlinear Dynamical Perspective on Climate Prediction.” Journal of Climate 12 (2): 575–91. https://doi.org/10.1175/1520-0442(1999)012<0575:andpoc>2.0.co;2.