AGU 2023 Oral Presentation

"Downscaling & bias-correction contribution considerable uncertainty to local climate projections in CMIP6" presented at AGU 2023 in San Francisco.

Transcript

1. Downscaling & bias-correction contribution considerable uncertainty to local climate projections in CMIP6 David Lafferty & Ryan Sriver University of Illinois 0
2. Managing local climate risks requires high-resolution, accurate climate information. Downscaled and bias-corrected model outputs are often used for this purpose. Downscaling & bias-correction can contribute considerable uncertainty to local climate projections. David Lafferty University of Illinois david0811.github.io 1
3. How much uncertainty in local climate projections arises from downscaling & bias-correction? Downscaling and bias-correction are important sources of uncertainty: • in the near-term (early-to-mid 21st century) • in projections of precipitation • in projections of extremes • in regions where observations disagree David Lafferty University of Illinois david0811.github.io 2
4. Ensemble # GCMs Spatial resolution Algorithm Training dataset Reference NEX-GDDP 22 1/4° BCSD GMFD (1960-2014) Thrasher, B., Wang, W., Michaelis, A. et al. NASA Global Daily Downscaled Projections, CMIP6. Sci Data 9, 262 (2022). CIL-GDPCIR 17 1/4° QDM + QPLAD ERA5 (1995-2014) Gergel, D. R., et al.: Global downscaled projections for climate impacts research (GDPCIR): preserving extremes for modeling future climate impacts, EGUsphere [preprint] ISIMIP3b 10 1/2° ISIMIP3BASD W5E5 v2.0 (1979-2019) Lange, S.: Trend-preserving bias adjustment and statistical downscaling with ISIMIP3BASD (v1.0), Geoscientific Model Development, 12, 3055–3070,. carbonplan GARD-SV 4 1/4° Generalized Analog Regression ERA5 (1981-2010) O Chegwidden, R Hagen, K Martin, M Jones, A Banihirwe, C Chiao, S Frank, J Freeman, J Hamman (2022) “Open data and tools for multiple methods of global climate downscaling" CarbonPlan. https://carbonplan.org/research/cmip6-downscaling- explainer carbonplan DeepSD-BC 2 1/4° SRCNN *at the time of publication David Lafferty University of Illinois david0811.github.io 3 We include all* global, publicly available, downscaled and bias-corrected CMIP6 outputs
5. Parent CMIP6 models Downscaling & bias-correction methods SSP scenarios Model uncertainty: variance across models, averaged over SSPs and downscaling methods David Lafferty University of Illinois david0811.github.io 4 We employ a simple variance decomposition approach to partition uncertainty
6. Downscaling & bias-correction methods SSP scenarios Model uncertainty: variance across models, averaged over SSPs and downscaling methods David Lafferty University of Illinois david0811.github.io 5 We employ a simple variance decomposition approach to partition uncertainty
7. Model uncertainty: variance across models, averaged over SSPs and downscaling methods David Lafferty University of Illinois david0811.github.io 6 We employ a simple variance decomposition approach to partition uncertainty
8. Parent CMIP6 models Downscaling & bias-correction methods SSP scenarios Model uncertainty: variance across models, averaged over SSPs and downscaling methods Downscaling uncertainty: variance across downscaling methods, averaged over SSPs and models Scenario uncertainty: variance across SSPs of the multi-model, multi-downscaling method mean David Lafferty University of Illinois david0811.github.io 7 We employ a simple variance decomposition approach to partition uncertainty
9. • The forced response is extracted via a 4th order polynomial fit • Inter-annual variability is characterized as the 10-year rolling variance of the residuals o Hawkins, E. & Sutton, R. The Potential to Narrow Uncertainty in Regional Climate Predictions. B Am Meteorol Soc 90, 1095–1107 (2009). o Hawkins, E. & Sutton, R. The potential to narrow uncertainty in projections of regional precipitation change. Clim Dyn 37, 407–418 (2011). David Lafferty University of Illinois david0811.github.io 8 We separate inter-annual variability from the forced response
10. David Lafferty University of Illinois david0811.github.io 9 03 4 674  9 -. 1). 3 %  66  ,25 ., 4. 4 (6.21 2 626 8 4. 1(  ( 1 4.2 2) 2915( .1, 4. . .6        9 -. 1). 3 66 ,25 ., 4.  
11. Downscaling and bias-correction are important sources of uncertainty: • in the near-term (early-to-mid 21st century) • in projections of precipitation • in projections of extremes • in regions where observations disagree David Lafferty University of Illinois david0811.github.io 10
12. David Lafferty University of Illinois david0811.github.io 11 Downscaling and bias-correction are important in the near term
13. David Lafferty University of Illinois david0811.github.io 12 Downscaling and bias-correction are important in projections of precipitation
14. David Lafferty University of Illinois david0811.github.io 13 Downscaling and bias-correction are important in projections of extremes
15. David Lafferty University of Illinois david0811.github.io 14 Thank you! Open Access Paper Interactive dashboard lafferty-sriver-2023-downscaling-uncertainty.msdlive.org
16. Supplementary slides David Lafferty University of Illinois david0811.github.io 15
17. David Lafferty University of Illinois david0811.github.io 16 CMIP6 model NEX-GDDP CIL-GDPCIR GARD-SV DeepSD-BC ISIMIP3BASD ACCESS-ESM1-5 SSP5-8.5 BCC-CSM2-MR SSP1-2.6; pr CanESM5 SSP1-2.6 SSP1-2.6 SSP2-4.5 CMCC-ESM2 CNRM-CM6-1 SSP2-4.5 CNRM-ESM2-1 SSP2-4.5 EC-Earth3 SSP2-4.5 EC-Earth3-Veg-LR GFDL-ESM4 HadGEM3-GC31-LL SSP3-7.0 SSP3-7.0 INM-CM4-8 INM-CM5-0 IPSL-CM6A-LR MIROC-ES2L MIROC6 SSP1-2.6; pr SSP2-4.5 MPI-ESM1-2-HR SSP1-2.6 MPI-ESM1-2-LR MRI-ESM2-0 SSP1-2.6; pr for SSP5-8.5 NESM3 SSP3-7.0 SSP3-7.0 NorESM2-LM NorESM2-MM UKESM1-0-LL
18. David Lafferty University of Illinois david0811.github.io 17
19. David Lafferty University of Illinois david0811.github.io 18
20. David Lafferty University of Illinois david0811.github.io 19
21. David Lafferty University of Illinois david0811.github.io 20
22. David Lafferty University of Illinois david0811.github.io 21