Machine learning for evaluating
climate model projections
@ZLabe
Zachary M. Labe
Postdoc at Princeton University and NOAA GFDL
7 December 2022 – Tech Talks 2.0
IEEE Student Branch ITTI & IEEE GRSS IITI
https://zacklabe.com/
Machine Learning
is not new!
But…
Machine Learning
is not new!
Artificial Intelligence
Machine Learning
Deep Learning
Computer Science
Computer Science
Artificial Intelligence
Machine Learning
Deep Learning
Supervised
Learning
Unsupervised
Learning
Labeled data
Classification
Regression
Unlabeled data
Clustering
Dimension reduction
• Do it better
• e.g., parameterizations in climate models are not
perfect, use ML to make them more accurate
• Do it faster
• e.g., code in climate models is very slow (but we
know the right answer) - use ML methods to speed
things up
• Do something new
• e.g., go looking for non-linear relationships you
didn’t know were there
Very relevant for
research: may be
slower and worse,
but can still learn
something
WHY SHOULD WE CONSIDER
MACHINE LEARNING?
GROWING DATA
Adapted from: Kotamarthi, R., Hayhoe, K., Mearns, L., Wuebbles, D., Jacobs, J., & Jurado, J.
(2021). Global Climate Models. In Downscaling Techniques for High-Resolution Climate
Projections: From Global Change to Local Impacts (pp. 19-39). Cambridge: Cambridge University
Press. doi:10.1017/9781108601269.003
CLIMATE MODELS
Horizontal Grid
Vertical Levels
Past/Present/Future
Fully-Coupled System
20-40 Petabytes of data
ADAPTED FROM EYRING ET AL. 2016
CMIP6
GROWING TOOLS
Python tools for machine learning
Today’s weather or climate
scientist is far more likely to be
debugging code written in
Python… than to be poring over
satellite images or releasing
radiosondes.
“
D. Irving| Bulletin of the American Meteorological Society| 2016
Machine learning for weather
IDENTIFYING SEVERE THUNDERSTORMS
Molina et al. 2021
Toms et al. 2021
CLASSIFYING PHASE OF MADDEN-JULLIAN OSCILLATION
SATELLITE DETECTION
Lee et al. 2021
DETECTING TORNADOES
McGovern et al. 2019
Machine learning for climate
FINDING FORECASTS OF OPPORTUNITY
Mayer and Barnes, 2021
PREDICTING CLIMATE MODES OF VARIABILITY
Gordon et al. 2021
TIMING OF CLIMATE CHANGE
Barnes et al. 2019
INPUT
[DATA]
PREDICTION
Machine
Learning
INPUT
[DATA]
PREDICTION
~Statistical
Algorithm~
INPUT
[DATA]
PREDICTION
Machine
Learning
NSF AI Institute for Research
on Trustworthy AI in Weather,
Climate, and Coastal
Oceanography (AI2ES)
https://www.ai2es.org/
E.g.,
Research to
Operations (R2O)
Tornado Warning
Special Marine Warning
Severe Thunderstorm Warning
Flash Flood Warning
E.g.,
Establish robust,
responsible AI for
severe weather
detection
Tornado Warning
Special Marine Warning
Severe Thunderstorm Warning
Flash Flood Warning
Artificial Intelligence
Machine Learning
Deep Learning
X1
X2
INPUTS
Artificial Neural Networks [ANN]
Linear regression!
Artificial Neural Networks [ANN]
X1
X2
W1
W2
∑ = X1W1+ X2W2 + b
INPUTS
NODE
Artificial Neural Networks [ANN]
X1
X2
W1
W2
∑
INPUTS
NODE
Linear regression with non-linear
mapping by an “activation function”
Training of the network is merely
determining the weights “w” and
bias/offset “b"
= factivation(X1W1+ X2W2 + b)
Artificial Neural Networks [ANN]
X1
X2
W1
W2
∑
INPUTS
NODE
= factivation(X1W1+ X2W2 + b)
ReLU Sigmoid Linear
X1
X2
∑
inputs
HIDDEN LAYERS
X3
∑
∑
∑
OUTPUT
= predictions
Artificial Neural Networks [ANN]
: : ::
INPUTS
Complexity and nonlinearities of the ANN allow it to learn many
different pathways of predictable behavior
Once trained, you have an array of weights and biases which can be
used for prediction on new data
INPUT
[DATA]
PREDICTION
Artificial Neural Networks [ANN]
TEMPERATURE
TEMPERATURE
We know some metadata…
+ What year is it?
+ Where did it come from?
We know some metadata…
+ What year is it?
+ Where did it come from?
TEMPERATURE
We know some metadata…
+ What year is it?
+ Where did it come from?
TEMPERATURE
Neural network learns nonlinear
combinations of forced climate
patterns to identify the year
----ANN----
2 Hidden Layers
10 Nodes each
Ridge Regularization
Early Stopping
We know some metadata…
+ What year is it?
+ Where did it come from?
[e.g., Barnes et al. 2019, 2020]
[e.g., Labe and Barnes, 2021]
TIMING OF EMERGENCE
(COMBINED VARIABLES)
RESPONSES TO
EXTERNAL CLIMATE
FORCINGS
PATTERNS OF
CLIMATE INDICATORS
[e.g., Rader et al. 2022]
Surface Temperature Map Precipitation Map
+
TEMPERATURE
----ANN----
2 Hidden Layers
10 Nodes each
Ridge Regularization
Early Stopping
We know some metadata…
+ What year is it?
+ Where did it come from?
[e.g., Barnes et al. 2019, 2020]
[e.g., Labe and Barnes, 2021]
TIMING OF EMERGENCE
(COMBINED VARIABLES)
RESPONSES TO
EXTERNAL CLIMATE
FORCINGS
PATTERNS OF
CLIMATE INDICATORS
Surface Temperature Map Precipitation Map
+
TEMPERATURE
[e.g., Rader et al. 2022]
THE REAL WORLD
(Observations)
What is the annual mean temperature of Earth?
What is the annual mean temperature of Earth?
THE REAL WORLD
(Observations)
Anomaly is relative to 1951-1980
What is the annual mean temperature of Earth?
THE REAL WORLD
(Observations)
Let’s run a
climate model
What is the annual mean temperature of Earth?
THE REAL WORLD
(Observations)
Let’s run a
climate model
again
What is the annual mean temperature of Earth?
THE REAL WORLD
(Observations)
Let’s run a
climate model
again & again
What is the annual mean temperature of Earth?
THE REAL WORLD
(Observations)
CLIMATE MODEL
ENSEMBLES
What is the annual mean temperature of Earth?
THE REAL WORLD
(Observations)
CLIMATE MODEL
ENSEMBLES
Range of ensembles
= internal variability (noise)
Mean of ensembles
= forced response (climate change)
What is the annual mean temperature of Earth?
Range of ensembles
= internal variability (noise)
Mean of ensembles
= forced response (climate change)
But let’s remove
climate change…
What is the annual mean temperature of Earth?
Range of ensembles
= internal variability (noise)
Mean of ensembles
= forced response (climate change)
After removing the
forced response…
anomalies/noise!
What is the annual mean temperature of Earth?
• Increasing greenhouse gases (CO2, CH4, N2O)
• Changes in industrial aerosols (SO4, BC, OC)
• Changes in biomass burning (aerosols)
• Changes in land-use & land-cover (albedo)
What is the annual mean temperature of Earth?
• Increasing greenhouse gases (CO2, CH4, N2O)
• Changes in industrial aerosols (SO4, BC, OC)
• Changes in biomass burning (aerosols)
• Changes in land-use & land-cover (albedo)
Plus everything else…
(Natural/internal variability)
What is the annual mean temperature of Earth?
Greenhouse gases fixed to 1920 levels
All forcings (CESM-LE)
Industrial aerosols fixed to 1920 levels
[Deser et al. 2020, JCLI]
Fully-coupled CESM1.1
20 Ensemble Members
Run from 1920-2080
Observations
So what?
Greenhouse gases = warming
Aerosols = ?? (though mostly cooling)
What are the relative responses
between greenhouse gas
and aerosol forcing?
Surface Temperature Map
ARTIFICIAL NEURAL NETWORK (ANN)
INPUT LAYER
Surface Temperature Map
ARTIFICIAL NEURAL NETWORK (ANN)
INPUT LAYER
HIDDEN LAYERS
OUTPUT LAYER
Surface Temperature Map
“2000-2009”
DECADE CLASS
“2070-2079”
“1920-1929”
ARTIFICIAL NEURAL NETWORK (ANN)
INPUT LAYER
HIDDEN LAYERS
OUTPUT LAYER
Surface Temperature Map
“2000-2009”
DECADE CLASS
“2070-2079”
“1920-1929”
BACK-PROPAGATE THROUGH NETWORK = EXPLAINABLE AI
ARTIFICIAL NEURAL NETWORK (ANN)
INPUT LAYER
HIDDEN LAYERS
OUTPUT LAYER
Layer-wise Relevance Propagation
Surface Temperature Map
“2000-2009”
DECADE CLASS
“2070-2079”
“1920-1929”
BACK-PROPAGATE THROUGH NETWORK = EXPLAINABLE AI
ARTIFICIAL NEURAL NETWORK (ANN)
[Barnes et al. 2020, JAMES]
[Labe and Barnes 2021, JAMES]
LAYER-WISE RELEVANCE PROPAGATION (LRP)
Volcano
Great White
Shark
Timber
Wolf
Image Classification LRP
https://heatmapping.org/
LRP heatmaps show regions
of “relevance” that
contribute to the neural
network’s decision-making
process for a sample
belonging to a particular
output category
Neural Network
WHY
WHY
WHY
Backpropagation – LRP
LAYER-WISE RELEVANCE PROPAGATION (LRP)
Volcano
Great White
Shark
Timber
Wolf
Image Classification LRP
https://heatmapping.org/
LRP heatmaps show regions
of “relevance” that
contribute to the neural
network’s decision-making
process for a sample
belonging to a particular
output category
Neural Network
WHY
WHY
WHY
Backpropagation – LRP
LAYER-WISE RELEVANCE PROPAGATION (LRP)
Volcano
Great White
Shark
Timber
Wolf
Image Classification LRP
https://heatmapping.org/
LRP heatmaps show regions
of “relevance” that
contribute to the neural
network’s decision-making
process for a sample
belonging to a particular
output category
Neural Network
Backpropagation – LRP
WHY
WHY
WHY
LAYER-WISE RELEVANCE PROPAGATION (LRP)
Image Classification LRP
https://heatmapping.org/
NOT PERFECT
Crock
Pot
Neural Network
Backpropagation – LRP
WHY
[Adapted from Adebayo et al., 2020]
EXPLAINABLE AI IS
NOT PERFECT
THERE ARE MANY
METHODS
[Adapted from Adebayo et al., 2020]
THERE ARE MANY
METHODS
EXPLAINABLE AI IS
NOT PERFECT
OUTPUT LAYER
Layer-wise Relevance Propagation
“2000-2009”
DECADE CLASS
“2070-2079”
“1920-1929”
BACK-PROPAGATE THROUGH NETWORK = EXPLAINABLE AI
WHY?
= LRP HEAT MAPS
[Labe and Barnes 2021, JAMES]
Layer-wise Relevance Propagation
BACK-PROPAGATE THROUGH NETWORK = EXPLAINABLE AI
WHY?
= LRP HEAT MAPS
Machine Learning
Black Box
[Labe and Barnes 2021, JAMES]
Layer-wise Relevance Propagation
BACK-PROPAGATE THROUGH NETWORK = EXPLAINABLE AI
WHY?
= LRP HEAT MAPS
Find regions of “relevance”
that contribute to the
neural network’s
decision-making process
[Labe and Barnes 2021, JAMES]
1960-1999: ANNUAL MEAN TEMPERATURE TRENDS
Greenhouse gases fixed
to 1920 levels
[AEROSOLS PREVAIL]
Industrial aerosols fixed
to 1920 levels
[GREENHOUSE GASES PREVAIL]
All forcings
[STANDARD CESM-LE]
DATA
1960-1999: ANNUAL MEAN TEMPERATURE TRENDS
Greenhouse gases fixed
to 1920 levels
[AEROSOLS PREVAIL]
Industrial aerosols fixed
to 1920 levels
[GREENHOUSE GASES PREVAIL]
All forcings
[STANDARD CESM-LE]
DATA
1960-1999: ANNUAL MEAN TEMPERATURE TRENDS
Greenhouse gases fixed
to 1920 levels
[AEROSOLS PREVAIL]
Industrial aerosols fixed
to 1920 levels
[GREENHOUSE GASES PREVAIL]
All forcings
[STANDARD CESM-LE]
DATA
1960-1999: ANNUAL MEAN TEMPERATURE TRENDS
Greenhouse gases fixed
to 1920 levels
[AEROSOLS PREVAIL]
Industrial aerosols fixed
to 1920 levels
[GREENHOUSE GASES PREVAIL]
All forcings
[STANDARD CESM-LE]
DATA
CLIMATE MODEL DATA PREDICT THE YEAR FROM MAPS OF TEMPERATURE
AEROSOLS
PREVAIL
GREENHOUSE GASES
PREVAIL
STANDARD
CLIMATE MODEL
[Labe and Barnes 2021, JAMES]
OBSERVATIONS PREDICT THE YEAR FROM MAPS OF TEMPERATURE
AEROSOLS
PREVAIL
GREENHOUSE GASES
PREVAIL
STANDARD
CLIMATE MODEL
[Labe and Barnes 2021, JAMES]
OBSERVATIONS
SLOPES
PREDICT THE YEAR FROM MAPS OF TEMPERATURE
AEROSOLS
PREVAIL
GREENHOUSE GASES
PREVAIL
STANDARD
CLIMATE MODEL
[Labe and Barnes 2021, JAMES]
HOW DID THE ANN
MAKE ITS
PREDICTIONS?
HOW DID THE ANN
MAKE ITS
PREDICTIONS?
WHY IS THERE
GREATER SKILL
FOR GHG+?
RESULTS FROM LRP
[Labe and Barnes 2021, JAMES]
Low High
RESULTS FROM LRP
[Labe and Barnes 2021, JAMES]
Low High
RESULTS FROM LRP
[Labe and Barnes 2021, JAMES]
Low High
RESULTS FROM LRP
[Labe and Barnes 2021, JAMES]
Low High
Earth is warming!
https://research.noaa.gov/article/ArtMID/587/ArticleID/2756/Simulated-
geoengineering-evaluation-cooler-planet-but-with-side-effects
Could we detect whether we were under the
influence of stratospheric aerosol injection (SAI)
using regional climate patterns?
TEMPERATURE
YEAR 2045
SAI? SAI?
PRECIPITATION
YEAR 2045
SAI? SAI?
LET’S TRY ANOMALIES
YEAR 2045
PROJECTIONS OF
TEMPERATURE
PROJECTIONS OF
PRECIPITATION
CAN WE DETECT A SAI WORLD?
LOGISTIC REGRESSION
CLIMATOLOGICAL MAPS OF ARISE-SAI-1.5 IN 2050-2069
MEAN STATE
DECADAL TRENDS
TEMPERATURE
DECADAL TRENDS
TEMPERATURE
DECADAL TRENDS
PRECIPITATION
DECADAL TRENDS
PRECIPITATION
N
Y
HIDDEN LAYERS
INPUT LAYER
INPUT LAYER
SAI WORLD?
or
or
map of near-surface temperature
map of near-surface temperature
map of total precipitation
map of total precipitation
Years Since
SAI Injection
OUTPUT
LOGISTIC
REGRESSION
ARTIFICAL
NEURAL
NETWORK
softmax
N
Y
HIDDEN LAYERS
INPUT LAYER
INPUT LAYER
SAI WORLD?
or
or
map of near-surface temperature
map of near-surface temperature
map of total precipitation
map of total precipitation
Years Since
SAI Injection
OUTPUT
LOGISTIC
REGRESSION
ARTIFICAL
NEURAL
NETWORK
softmax
CAN WE DETECT A SAI WORLD?
Central
Africa
HOW DID THE ML MODEL KNOW?
…Using regional climate patterns!
CAN WE DETECT A SAI WORLD?
PRECIPITATION
PRECIPITATION
PRECIPITATION
PRECIPITATION
PRECIPITATION
N
Y
HIDDEN LAYERS
INPUT LAYER
INPUT LAYER
SAI WORLD?
or
or
map of near-surface temperature
map of near-surface temperature
map of total precipitation
map of total precipitation
Years Since
SAI Injection
OUTPUT
LOGISTIC
REGRESSION
ARTIFICAL
NEURAL
NETWORK
softmax
TEMPERATURE
PRECIPITATION
INPUT
[DATA]
PREDICTION
Machine
Learning
Explainable AI
Learn new
science!
MACHINE LEARNING IS JUST
ANOTHER TOOL TO ADD TO OUR
WORKFLOW.
1)
MACHINE LEARNING IS
NO LONGER A BLACK BOX.
2)
WE CAN LEARN NEW SCIENCE
FROM EXPLAINABLE AI.
3)
KEY POINTS
1. Machine learning is just another tool to add to our scientific workflow
2. We can use explainable AI (XAI) methods to peer into the black box of machine learning
3. We can learn new science by using XAI methods in conjunction with existing statistical tools
Zachary Labe
zachary.labe@noaa.gov
@ZLabe
Labe, Z.M. and E.A. Barnes (2021), Detecting climate signals using explainable AI with single-forcing
large ensembles. Journal of Advances in Modeling Earth Systems, DOI:10.1029/2021MS002464
Labe, Z.M., E.A. Barnes, and J.W. Hurrell (2023), Identifying the regional emergence of climate patterns
in a simulation of stratospheric aerosol injection, to be submitted in December 2022

Machine learning for evaluating climate model projections

  • 1.
    Machine learning forevaluating climate model projections @ZLabe Zachary M. Labe Postdoc at Princeton University and NOAA GFDL 7 December 2022 – Tech Talks 2.0 IEEE Student Branch ITTI & IEEE GRSS IITI https://zacklabe.com/
  • 2.
  • 3.
  • 4.
  • 5.
    Computer Science Artificial Intelligence MachineLearning Deep Learning Supervised Learning Unsupervised Learning Labeled data Classification Regression Unlabeled data Clustering Dimension reduction
  • 6.
    • Do itbetter • e.g., parameterizations in climate models are not perfect, use ML to make them more accurate • Do it faster • e.g., code in climate models is very slow (but we know the right answer) - use ML methods to speed things up • Do something new • e.g., go looking for non-linear relationships you didn’t know were there Very relevant for research: may be slower and worse, but can still learn something WHY SHOULD WE CONSIDER MACHINE LEARNING?
  • 7.
  • 9.
    Adapted from: Kotamarthi,R., Hayhoe, K., Mearns, L., Wuebbles, D., Jacobs, J., & Jurado, J. (2021). Global Climate Models. In Downscaling Techniques for High-Resolution Climate Projections: From Global Change to Local Impacts (pp. 19-39). Cambridge: Cambridge University Press. doi:10.1017/9781108601269.003 CLIMATE MODELS Horizontal Grid Vertical Levels Past/Present/Future Fully-Coupled System 20-40 Petabytes of data
  • 10.
    ADAPTED FROM EYRINGET AL. 2016 CMIP6
  • 11.
  • 14.
    Python tools formachine learning
  • 15.
    Today’s weather orclimate scientist is far more likely to be debugging code written in Python… than to be poring over satellite images or releasing radiosondes. “ D. Irving| Bulletin of the American Meteorological Society| 2016
  • 16.
    Machine learning forweather IDENTIFYING SEVERE THUNDERSTORMS Molina et al. 2021 Toms et al. 2021 CLASSIFYING PHASE OF MADDEN-JULLIAN OSCILLATION SATELLITE DETECTION Lee et al. 2021 DETECTING TORNADOES McGovern et al. 2019
  • 17.
    Machine learning forclimate FINDING FORECASTS OF OPPORTUNITY Mayer and Barnes, 2021 PREDICTING CLIMATE MODES OF VARIABILITY Gordon et al. 2021 TIMING OF CLIMATE CHANGE Barnes et al. 2019
  • 18.
  • 19.
  • 20.
  • 21.
    NSF AI Institutefor Research on Trustworthy AI in Weather, Climate, and Coastal Oceanography (AI2ES) https://www.ai2es.org/
  • 22.
    E.g., Research to Operations (R2O) TornadoWarning Special Marine Warning Severe Thunderstorm Warning Flash Flood Warning
  • 23.
    E.g., Establish robust, responsible AIfor severe weather detection Tornado Warning Special Marine Warning Severe Thunderstorm Warning Flash Flood Warning
  • 24.
  • 25.
  • 26.
    Linear regression! Artificial NeuralNetworks [ANN] X1 X2 W1 W2 ∑ = X1W1+ X2W2 + b INPUTS NODE
  • 27.
    Artificial Neural Networks[ANN] X1 X2 W1 W2 ∑ INPUTS NODE Linear regression with non-linear mapping by an “activation function” Training of the network is merely determining the weights “w” and bias/offset “b" = factivation(X1W1+ X2W2 + b)
  • 28.
    Artificial Neural Networks[ANN] X1 X2 W1 W2 ∑ INPUTS NODE = factivation(X1W1+ X2W2 + b) ReLU Sigmoid Linear
  • 29.
  • 30.
    Complexity and nonlinearitiesof the ANN allow it to learn many different pathways of predictable behavior Once trained, you have an array of weights and biases which can be used for prediction on new data INPUT [DATA] PREDICTION Artificial Neural Networks [ANN]
  • 31.
  • 32.
    TEMPERATURE We know somemetadata… + What year is it? + Where did it come from?
  • 33.
    We know somemetadata… + What year is it? + Where did it come from? TEMPERATURE
  • 34.
    We know somemetadata… + What year is it? + Where did it come from? TEMPERATURE Neural network learns nonlinear combinations of forced climate patterns to identify the year
  • 35.
    ----ANN---- 2 Hidden Layers 10Nodes each Ridge Regularization Early Stopping We know some metadata… + What year is it? + Where did it come from? [e.g., Barnes et al. 2019, 2020] [e.g., Labe and Barnes, 2021] TIMING OF EMERGENCE (COMBINED VARIABLES) RESPONSES TO EXTERNAL CLIMATE FORCINGS PATTERNS OF CLIMATE INDICATORS [e.g., Rader et al. 2022] Surface Temperature Map Precipitation Map + TEMPERATURE
  • 36.
    ----ANN---- 2 Hidden Layers 10Nodes each Ridge Regularization Early Stopping We know some metadata… + What year is it? + Where did it come from? [e.g., Barnes et al. 2019, 2020] [e.g., Labe and Barnes, 2021] TIMING OF EMERGENCE (COMBINED VARIABLES) RESPONSES TO EXTERNAL CLIMATE FORCINGS PATTERNS OF CLIMATE INDICATORS Surface Temperature Map Precipitation Map + TEMPERATURE [e.g., Rader et al. 2022]
  • 37.
    THE REAL WORLD (Observations) Whatis the annual mean temperature of Earth?
  • 38.
    What is theannual mean temperature of Earth? THE REAL WORLD (Observations) Anomaly is relative to 1951-1980
  • 39.
    What is theannual mean temperature of Earth? THE REAL WORLD (Observations) Let’s run a climate model
  • 40.
    What is theannual mean temperature of Earth? THE REAL WORLD (Observations) Let’s run a climate model again
  • 41.
    What is theannual mean temperature of Earth? THE REAL WORLD (Observations) Let’s run a climate model again & again
  • 42.
    What is theannual mean temperature of Earth? THE REAL WORLD (Observations) CLIMATE MODEL ENSEMBLES
  • 43.
    What is theannual mean temperature of Earth? THE REAL WORLD (Observations) CLIMATE MODEL ENSEMBLES Range of ensembles = internal variability (noise) Mean of ensembles = forced response (climate change)
  • 44.
    What is theannual mean temperature of Earth? Range of ensembles = internal variability (noise) Mean of ensembles = forced response (climate change) But let’s remove climate change…
  • 45.
    What is theannual mean temperature of Earth? Range of ensembles = internal variability (noise) Mean of ensembles = forced response (climate change) After removing the forced response… anomalies/noise!
  • 46.
    What is theannual mean temperature of Earth? • Increasing greenhouse gases (CO2, CH4, N2O) • Changes in industrial aerosols (SO4, BC, OC) • Changes in biomass burning (aerosols) • Changes in land-use & land-cover (albedo)
  • 47.
    What is theannual mean temperature of Earth? • Increasing greenhouse gases (CO2, CH4, N2O) • Changes in industrial aerosols (SO4, BC, OC) • Changes in biomass burning (aerosols) • Changes in land-use & land-cover (albedo) Plus everything else… (Natural/internal variability)
  • 48.
    What is theannual mean temperature of Earth?
  • 49.
    Greenhouse gases fixedto 1920 levels All forcings (CESM-LE) Industrial aerosols fixed to 1920 levels [Deser et al. 2020, JCLI] Fully-coupled CESM1.1 20 Ensemble Members Run from 1920-2080 Observations
  • 50.
    So what? Greenhouse gases= warming Aerosols = ?? (though mostly cooling) What are the relative responses between greenhouse gas and aerosol forcing?
  • 51.
  • 52.
    INPUT LAYER Surface TemperatureMap ARTIFICIAL NEURAL NETWORK (ANN)
  • 53.
    INPUT LAYER HIDDEN LAYERS OUTPUTLAYER Surface Temperature Map “2000-2009” DECADE CLASS “2070-2079” “1920-1929” ARTIFICIAL NEURAL NETWORK (ANN)
  • 54.
    INPUT LAYER HIDDEN LAYERS OUTPUTLAYER Surface Temperature Map “2000-2009” DECADE CLASS “2070-2079” “1920-1929” BACK-PROPAGATE THROUGH NETWORK = EXPLAINABLE AI ARTIFICIAL NEURAL NETWORK (ANN)
  • 55.
    INPUT LAYER HIDDEN LAYERS OUTPUTLAYER Layer-wise Relevance Propagation Surface Temperature Map “2000-2009” DECADE CLASS “2070-2079” “1920-1929” BACK-PROPAGATE THROUGH NETWORK = EXPLAINABLE AI ARTIFICIAL NEURAL NETWORK (ANN) [Barnes et al. 2020, JAMES] [Labe and Barnes 2021, JAMES]
  • 56.
    LAYER-WISE RELEVANCE PROPAGATION(LRP) Volcano Great White Shark Timber Wolf Image Classification LRP https://heatmapping.org/ LRP heatmaps show regions of “relevance” that contribute to the neural network’s decision-making process for a sample belonging to a particular output category Neural Network WHY WHY WHY Backpropagation – LRP
  • 57.
    LAYER-WISE RELEVANCE PROPAGATION(LRP) Volcano Great White Shark Timber Wolf Image Classification LRP https://heatmapping.org/ LRP heatmaps show regions of “relevance” that contribute to the neural network’s decision-making process for a sample belonging to a particular output category Neural Network WHY WHY WHY Backpropagation – LRP
  • 58.
    LAYER-WISE RELEVANCE PROPAGATION(LRP) Volcano Great White Shark Timber Wolf Image Classification LRP https://heatmapping.org/ LRP heatmaps show regions of “relevance” that contribute to the neural network’s decision-making process for a sample belonging to a particular output category Neural Network Backpropagation – LRP WHY WHY WHY
  • 59.
    LAYER-WISE RELEVANCE PROPAGATION(LRP) Image Classification LRP https://heatmapping.org/ NOT PERFECT Crock Pot Neural Network Backpropagation – LRP WHY
  • 60.
    [Adapted from Adebayoet al., 2020] EXPLAINABLE AI IS NOT PERFECT THERE ARE MANY METHODS
  • 61.
    [Adapted from Adebayoet al., 2020] THERE ARE MANY METHODS EXPLAINABLE AI IS NOT PERFECT
  • 62.
    OUTPUT LAYER Layer-wise RelevancePropagation “2000-2009” DECADE CLASS “2070-2079” “1920-1929” BACK-PROPAGATE THROUGH NETWORK = EXPLAINABLE AI WHY? = LRP HEAT MAPS [Labe and Barnes 2021, JAMES]
  • 63.
    Layer-wise Relevance Propagation BACK-PROPAGATETHROUGH NETWORK = EXPLAINABLE AI WHY? = LRP HEAT MAPS Machine Learning Black Box [Labe and Barnes 2021, JAMES]
  • 64.
    Layer-wise Relevance Propagation BACK-PROPAGATETHROUGH NETWORK = EXPLAINABLE AI WHY? = LRP HEAT MAPS Find regions of “relevance” that contribute to the neural network’s decision-making process [Labe and Barnes 2021, JAMES]
  • 65.
    1960-1999: ANNUAL MEANTEMPERATURE TRENDS Greenhouse gases fixed to 1920 levels [AEROSOLS PREVAIL] Industrial aerosols fixed to 1920 levels [GREENHOUSE GASES PREVAIL] All forcings [STANDARD CESM-LE] DATA
  • 66.
    1960-1999: ANNUAL MEANTEMPERATURE TRENDS Greenhouse gases fixed to 1920 levels [AEROSOLS PREVAIL] Industrial aerosols fixed to 1920 levels [GREENHOUSE GASES PREVAIL] All forcings [STANDARD CESM-LE] DATA
  • 67.
    1960-1999: ANNUAL MEANTEMPERATURE TRENDS Greenhouse gases fixed to 1920 levels [AEROSOLS PREVAIL] Industrial aerosols fixed to 1920 levels [GREENHOUSE GASES PREVAIL] All forcings [STANDARD CESM-LE] DATA
  • 68.
    1960-1999: ANNUAL MEANTEMPERATURE TRENDS Greenhouse gases fixed to 1920 levels [AEROSOLS PREVAIL] Industrial aerosols fixed to 1920 levels [GREENHOUSE GASES PREVAIL] All forcings [STANDARD CESM-LE] DATA
  • 69.
    CLIMATE MODEL DATAPREDICT THE YEAR FROM MAPS OF TEMPERATURE AEROSOLS PREVAIL GREENHOUSE GASES PREVAIL STANDARD CLIMATE MODEL [Labe and Barnes 2021, JAMES]
  • 70.
    OBSERVATIONS PREDICT THEYEAR FROM MAPS OF TEMPERATURE AEROSOLS PREVAIL GREENHOUSE GASES PREVAIL STANDARD CLIMATE MODEL [Labe and Barnes 2021, JAMES]
  • 71.
    OBSERVATIONS SLOPES PREDICT THE YEARFROM MAPS OF TEMPERATURE AEROSOLS PREVAIL GREENHOUSE GASES PREVAIL STANDARD CLIMATE MODEL [Labe and Barnes 2021, JAMES]
  • 72.
    HOW DID THEANN MAKE ITS PREDICTIONS?
  • 73.
    HOW DID THEANN MAKE ITS PREDICTIONS? WHY IS THERE GREATER SKILL FOR GHG+?
  • 74.
    RESULTS FROM LRP [Labeand Barnes 2021, JAMES] Low High
  • 75.
    RESULTS FROM LRP [Labeand Barnes 2021, JAMES] Low High
  • 76.
    RESULTS FROM LRP [Labeand Barnes 2021, JAMES] Low High
  • 77.
    RESULTS FROM LRP [Labeand Barnes 2021, JAMES] Low High
  • 78.
  • 81.
  • 82.
    Could we detectwhether we were under the influence of stratospheric aerosol injection (SAI) using regional climate patterns?
  • 83.
  • 84.
  • 85.
  • 88.
  • 89.
  • 90.
    CAN WE DETECTA SAI WORLD? LOGISTIC REGRESSION
  • 91.
    CLIMATOLOGICAL MAPS OFARISE-SAI-1.5 IN 2050-2069 MEAN STATE
  • 92.
  • 93.
  • 94.
  • 95.
  • 96.
    N Y HIDDEN LAYERS INPUT LAYER INPUTLAYER SAI WORLD? or or map of near-surface temperature map of near-surface temperature map of total precipitation map of total precipitation Years Since SAI Injection OUTPUT LOGISTIC REGRESSION ARTIFICAL NEURAL NETWORK softmax
  • 97.
    N Y HIDDEN LAYERS INPUT LAYER INPUTLAYER SAI WORLD? or or map of near-surface temperature map of near-surface temperature map of total precipitation map of total precipitation Years Since SAI Injection OUTPUT LOGISTIC REGRESSION ARTIFICAL NEURAL NETWORK softmax
  • 98.
    CAN WE DETECTA SAI WORLD?
  • 101.
  • 102.
    HOW DID THEML MODEL KNOW?
  • 103.
  • 104.
    CAN WE DETECTA SAI WORLD?
  • 105.
  • 106.
  • 107.
  • 108.
  • 109.
  • 110.
    N Y HIDDEN LAYERS INPUT LAYER INPUTLAYER SAI WORLD? or or map of near-surface temperature map of near-surface temperature map of total precipitation map of total precipitation Years Since SAI Injection OUTPUT LOGISTIC REGRESSION ARTIFICAL NEURAL NETWORK softmax
  • 111.
  • 112.
  • 113.
  • 114.
    MACHINE LEARNING ISJUST ANOTHER TOOL TO ADD TO OUR WORKFLOW. 1)
  • 115.
    MACHINE LEARNING IS NOLONGER A BLACK BOX. 2)
  • 116.
    WE CAN LEARNNEW SCIENCE FROM EXPLAINABLE AI. 3)
  • 117.
    KEY POINTS 1. Machinelearning is just another tool to add to our scientific workflow 2. We can use explainable AI (XAI) methods to peer into the black box of machine learning 3. We can learn new science by using XAI methods in conjunction with existing statistical tools Zachary Labe zachary.labe@noaa.gov @ZLabe Labe, Z.M. and E.A. Barnes (2021), Detecting climate signals using explainable AI with single-forcing large ensembles. Journal of Advances in Modeling Earth Systems, DOI:10.1029/2021MS002464 Labe, Z.M., E.A. Barnes, and J.W. Hurrell (2023), Identifying the regional emergence of climate patterns in a simulation of stratospheric aerosol injection, to be submitted in December 2022