Timing of emergence of U.S. summer temperatures with explainable AI

Labe, Z.M. Timing of emergence of U.S. summer temperatures with explainable AI, Building AI-Ready workforce for heat resilience, 6th NOAA AI Workshop, USA (Sep 2024) (Invited-Remote).

References...
Eischeid, J.K., M.P. Hoerling, X.-W. Quan, A. Kumar, J. Barsugli, Z.M. Labe, K.E. Kunkel, C.J. Schreck III, D.R. Easterling, T. Zhang, J. Uehling, and X. Zhang (2023). Why has the summertime central U.S. warming hole not disappeared? Journal of Climate, DOI:10.1175/JCLI-D-22-0716.1

Labe, Z.M., N.C. Johnson, and T.L. Delworth (2024), Changes in United States summer temperatures revealed by explainable neural networks. Earth’s Future, DOI:10.1029/2023EF003981

1. TIMING OF EMERGENCE OF U.S. SUMMER
TEMPERATURES WITH EXPLAINABLE AI
Zachary M. Labe
NOAA Geophysical Fluid Dynamics Laboratory (GFDL)
with…
Nathaniel C. Johnson, NOAA GFDL
Thomas L. Delworth, NOAA GFDL
19 September 2024 – 6th NOAA AI Workshop
“Building AI-Ready Workforce for Heat Resilience”
@zlabe.bsky.socialhttps://zacklabe.com/

2. NOAA/NESDIS/NCEI NOAAGlobalTempv6.0.0

3. A "warming hole”
NOAA/NESDIS/NCEI NOAAGlobalTempv6.0.0

4. Temperature anomalies [ °C ] relative to 1981-2010
Climate model data from GFDL SPEAR_MED
United States – Summer
1920 2020

5. Temperature anomalies [ °C ] relative to 1981-2010
United States – Summer
1920 2020
Dust Bowl – July 1936
Mt. Pinatubo
2022

6. Eischeid, J. K., Hoerling, M. P., Quan, X. W., Kumar, A., Barsugli, J., Labe,
Z. M., ... & Zhang, X. (2023). Why Has the Summertime Central US
Warming Hole Not Disappeared?Journal of Climate,36(20), 7319-7336.
https://doi.org/10.1175/JCLI-D-22-0716.1
Persistence is consistent with
unusually high summertime
rainfall over the region
Large ensembles demonstrate
that this rainfall trend can arise
from atmospheric internal
variability alone
Recent trend in tropical Pacific
SST can also reinforce this
pattern

7. TEMPERATURE

8. TEMPERATURE
We know some metadata…
+ What year is it?
+ Where did it come from?

9. TEMPERATURE
We know some metadata…
+ What year is it?
+ Where did it come from?

10. We know some metadata…
+ What year is it?
+ Where did it come from?
[Labe and Barnes, 2022; ESS]
TEMPERATURE

11. TEMPERATURE
Neural network learns nonlinear
combinations of forced climate
patterns to identify the year
We know some metadata…
+ What year is it?
+ Where did it come from?
[Labe and Barnes, 2022; ESS]

12. ----ANN----
2 Hid den Layers
10 Nodes each
Ri dge Regulariz ation
Early Stopping
[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 Te mpe rature Map Prec ipitatio n Map
+
TEMPERATURE
We know some metadata…
+ What year is it?
+ Where did it come from?
[Labe and Barnes, 2022; ESS]

13. Seasonal Maps of T2M, TMAX, TMIN
Input
NOAA GFDL – SPEAR_MED
Fully-Coupled (AM4/LM4/MOM6/SIS2)
Historical + SSP5-8.5
0.5° land/atmosphere, 1.0° ocean
https://www.gfdl.noaa.gov/spear/
Delworth et al. (2020, JAMES)

14. Seasonal Maps of T2M, TMAX, TMIN
Hidden Layers
Artificial Neural Network
Input

15. Seasonal Maps of T2M, TMAX, TMIN
Hidden Layers
Output
1921
2100
Artificial Neural Network
Input

16. Backpropagation
Seasonal Maps of T2M, TMAX, TMIN
Hidden Layers
Output
1921
2100
Artificial Neural Network
Input
Post hoc – XAI methods

17. Neural Network
Predictions
for SPEAR/Obs

18. 1921 2021 2100
ACTUAL YEARS
PREDICTED YEARS
NOAA Monthly U.S.
NClimGrid v1.0
1921
2021
2100

19. Max year predicted in the 1921-1950
baseline for observations
Timing of
Emergence
1921-1950
Skill
1:1 Perfect Prediction

20. June – August – Timing of Emergence (ToE) For Observations Over United States

21. How is the neural network able to detect the year prior to ~1990?
Temperature anomalies [ °C ] relative to 1981-2010
Machine learning predictions GFDL SPEAR_MED simulation

22. Machine Learning Explainability Methods – Ad Hoc Feature Attribution
Decrease
likelihood of year
Increase
likelihood of year

23. 1) Is it
aerosols?
(all forcings)
(all forcings, but no anthropogenic aerosols)

24. Coherence
Check!
(all forcings)
(a natural-only forcing simulation)

25. 50 km resolution 100 km resolution
2) Is it related to resolution?
SPEAR_MED SPEAR_LO
MAE (years)
Mean Absolute Error (MAE) for ensemble member predictions over 1921 to 1989 for different spatial resolutions

26. 3) Is it systematic in CMIP6?

27. 4) So, what is it? The land surface?

28. TRENDS IN EVAPORATION
SPEAR_MED, but NO anthropogenic aerosols SPEAR_MED, but NO anthropogenic forcings
IncreaseDecrease
Fully-Coupled [Historical]

29. MAX TEMPERATURE (JUN-AUG)

30. Eischeid, J. K., Hoerling, M. P., Quan, X. W., Kumar, A., Barsugli, J., Labe, Z. M., ...
& Zhang, X. (2023). Why Has the Summertime Central US Warming Hole Not
Disappeared?Journal of Climate,36(20), 7319-7336.
https://doi.org/10.1175/JCLI-D-22-0716.1
MOVING FORWARD
•Increasing probability of daily
summertime temperatures
exceeding the Dust Bowl with
future climate change
•However, large differences
remain across climate models
•XAI provides a tool for comparing
across models and observations
in near real-time, such as for
extreme heat

31. INPUT
[Large Ensembles]
PREDICTION
Machine
Learning
Explainable (or interpretable) AIIdentify patterns of
climate change
signals

32. KEY POINTS
1.Forced temperature changes have emerged in observations during
summer in the United States as detected by an artificial neural network
2.Increasing spatial resolution improves neural network skill for predicting the
year of a given summer temperature map
3.Western United States land surface climate properties contribute to earlier
timing of emergence predictions for the SPEAR climate model
[email protected]
Thursday, 19 September 2024
19 September 2024 – 6th NOAA AI Workshop
“Building AI-Ready Workforce for Heat Resilience”
Labe, Z.M., N.C. Johnson, and T.L Delworth (2024). Changes in United States summer temperatures
revealed by explainable neural networks, Earth’s Future, DOI: 10.1029/2023EF003981