1 of 80

Climate Model Ensembles:�Why they are complicated, valuable, and fascinating�

May 22, 2024, 5:00 – 6:30 p.m. ET

Benjamin Brown-Steiner, Ph.D.

Research Associate, PRI

2 of 80

3 of 80

4 of 80

1

5 of 80

1

6 of 80

1

7 of 80

4

8 of 80

4

9 of 80

4

10 of 80

Surface Temperature, Two Layers

11 of 80

Surface Temperature, Many Layers

12 of 80

4

13 of 80

16

14 of 80

16

15 of 80

160

16 of 80

160

17 of 80

Simple Climate Models

Atmosphere

Ocean

18 of 80

Simple Climate Models

Atmosphere

Land

Ocean

19 of 80

Simple Climate Models

Atmosphere

Land

Ocean

20 of 80

Simple Climate Models

Atmosphere

Land

Ocean

21 of 80

Global Climate Models

22 of 80

23 of 80

24 of 80

25 of 80

Q&A Session #1

26 of 80

The Spark

Wendell Berry:
“Raindrops that pass in random fashion through an imaginary plane above the forest canopy…tend to leave the ecosystem as they entered it, in randomized fashion.”
“Does ‘random’ in this (or any) context describe a verifiable condition or a limit of perception (i.e. random as far as we can tell)?”
Berry’s answer is: “It describes a limit of perception…based on the belief that pattern is verifiable by limited information, whereas the information required to verify randomness is unlimited…”
“…should have said that rainwater moves from mystery through pattern back into mystery...”
“To call this mystery ‘randomness’ is to take charge of it on behalf of those who do not respect pattern. To call the unknown ‘random’ is to plant the flag by which to colonize and exploit the unknown. To call the unknown by its right name, ‘mystery’…”

t=0

t=t₁

27 of 80

Perhaps the most familiar attempts to get at uncertainty/mystery…

IPCC, Guidance Notes, July, 2005

IPCC, AR5, Summary for Policymakers

“There is high confidence that the ENSO will remain the dominant mode…and precipitation variability on regional scales will likely intensify. Natural variations…are large and thus confidence in any specific projected change in ENSO …remains low.”

28 of 80

What is science?

the study of or

knowledge about

the natural world

29 of 80

One more complexity…all that stuff science cannot do

30 of 80

Facts

Values

Knowledge

Wisdom

Decisions

Actions

31 of 80

Facts

Values

Knowledge

Wisdom

Decisions

Actions

Very Certain

Uncertain

32 of 80

Facts

Values

Knowledge

Wisdom

Decisions

Actions

“One of the purposes of objectivity, in practice, is to avoid coming to a moral conclusion.”

- Wendell Berry

33 of 80

REAL WORLD�(OMNISCIENT)

the past today the future

34 of 80

OUR VIEW AS OBSERVERS�(NOT OMNISCIENT)

AVAILABILITY (missing data)

ACCURACY (inexactness)

INDETERMINANCY (unobservable)

IGNORANCE (mystery)

UNPREDICTABILITY (the unexpected)

PATH DEPENDENT (what will we decide?)

the past today the future

“…we will always experience [data] as probabilistic, as shimmering, rather than fixed.”

“…all knowledge about [the earth system] depends fundamentally on modeling.”

- Paul Edwards, A Vast Machine, page 352

35 of 80

SCIENTIFIC METHOD�LETS US PUT THINGS TOGETHER�

We make Presuppositions

We take observations

We build a body of Evidence

We apply Logic

We develop a Model

We get Output

Now what?

“the world exists”

“what we see is what is there”

Parameterizations: ß, Ω, µ

Theoretical Knowledge

Empirical Knowledge

Computational Capabilities

Structure

modus ponens (if A 🡪 B, A .˙. B)

“we can represent the world with mathematics”

BC & IC

PEL Model,

Scientific Method in Practice,

Hugh Gauch, Jr. (2003)

36 of 80

This is important

But the interesting stuff is out here

Understanding = Knowledge + Ignorance

Bill Vitek and Wes Jackon, Virtures of Ignorance, 2005

We must characterize what we don’t know as much as what we do know.

37 of 80

BUILDING TRUST �THROUGH CHARACTERIZING UNCERTAINTIES

EPISTEMIC

UNCERTAINTIES

ALEATORY

UNCERTAINTIES

Observational Uncertainty (the past shimmers)

Chaos / Dependence on Initial Conditions

Internal Variability

Surprises (non-linearities, unobservables)

Known-Unknowns: uncertainties in the parameters or model components

Unknown-Knowns: the very rare happy surprise (more often luck)

Unknown-Unknowns: things we cannot or do not know how to describe

38 of 80

We want to know how much

rain will pass through

that forest canopy.

ALEATORY

UNCERTAINTIES

EPISTEMIC

UNCERTAINTIES

If we look at the aleatory

uncertainties as chaos

or randomness, we have to

consider: “Does ‘random’

in this (or any) context

describe a verifiable condition

or a limit of perception?”

When we use random do we

mean “random as far as we can tell?”

- Wes Jackson, Virtues of Ignorance, page 22

MYSTERY?

RANDOM?

39 of 80

PERTURB BOUNDARY CONDITIONS�(a.k.a. CREATE SCENARIOS)

What does this do for us?

SRES, RCPs, etc. allow us to proceed under uncertainty, very practical
Allow us to explore the boundaries of what is possible (in our models) through exploration of epistemic uncertainties (via model structure)
Let us use our expert knowledge (which is good, but fallible)
Lets us sidestep the responsibility of saying which path is the best/worst
Historically, we have been hesitant to apply probabilities to scenarios
CAVEAT: Only explores what we think is possible, so we may be missing a lot

40 of 80

41 of 80

PERTURB INITIAL CONDITIONS

t₀

t₅₀

t₁₀₀

Ensemble Forecasting (or Monte-Carlo Approach)
Keep everything in the model the same (parameters, structure, etc.)
Exploring aleatory uncertainties…
…but not all of them. Just the ones your model can simulate.

Adapted from Wilks 2006

Full

Range

of

Possibilities

42 of 80

Figure 2: Scenarios of future emissions for various greenhouse gases and other pollutants. Image from Riahi, K. et al. The Shared Socioeconomic Pathways and their energy, land use, and greenhouse gas emissions implications: An overview. Global Environmental Change 42, 153–168 (2017).

43 of 80

44 of 80

Q&A Session #2

45 of 80

46 of 80

47 of 80

D6

48 of 80

D6

D4

D8

D12

D20

D10

49 of 80

Group Name Dice

“Today” D6 + D6

“2.6” D6 + D4

“4.5” D6 + D8

“6.0” D6 + D12

“8.5” D6 + D20

50 of 80

Group Name Dice

“Today” D6 + D6

“2.6” D6 + D4

“4.5” D6 + D8

“6.0” D6 + D12

“8.5” D6 + D20

51 of 80

https://g.co/kgs/NTi1z2a

52 of 80

X