Process-based Evaluation of Temperature and Precipitation Projections and Downscaling Methods over the CONUS: Charting a Path for End-Users from the CMIP6 Ensemble to Multivariate Facility-Level Risks (RC19-1391)��Daniel Feldman (LBNL, PI)

Objectives

Robust 21^st Century temperature and precipitation projections are needed at the DoD facility scale.

This project is seeks to navigate the CMIP6 Earth System Model (ESM) archive and downscale it to produce T and Pr at 6 km resolution across the CONUS.

The project selects CMIP6 models and downscaling techniques to estimate future risks from changes in T and Pr distributions.

​

Takeaways:

• Analyzed historical observations to train downscaling techniques.
• Developed and running statistical downscaling (LOCA2) for CMIP6.
• Ran dynamical downscaling (WRF) for Western United States.
• Established pipeline to provide LOCA2 solutions to 5^th National Climate Assessment.

​

​

Earth System Model Downscaling

• Robust 21^st Century projections of temperature (T) and precipitation (Pr) at the facility level require supplementing Earth System Model simulations with fine-scale information
• Statistical methods (LOCA2) use historical observations of relationships to estimate facility-level effects based on ESM predictor variables.
• Dynamical methods (WRF) use an atmospheric circulation model to predict facility-level T and Pr that have no analogs in historical observations.
• Historical observations test ESMs and downscaling.
• We use the strengths of observations, WRF, and LOCA2 to develop facility-level projections of T and Pr from the CMIP6 ensemble of ESMs.

Technical Approach

Milestones: Critically evaluating observational data (FY20); Developing and implementing downscaling (FY21); Analyzing results (FY22) Producing detailed projections (FY23).

Products: LOCA2-downscaled CMIP6, WRF-downscaled CMIP6 sub-set

Anticipated completion date: December, 2023.

Feldman et al, 2021, JWRPM

• Historical observations form the foundation for ESM model evaluation.
• Widely-used gridded observational products mute precipitation extremes through gridding (Risser et al, 2021).
• They also are not homogenized to account for temperature measurement errors (Charn et al, 2021, in prep).
• Comparisons between observations and models can and must account for these observational product biases.

Results to Date (1 of 2)

Charn et al, 2021, In Prep

Risser et al, 2021, Climatic Change

• LOCA2 statistical downscaling is under-way, with 31 candidate CMIP6 models. Solutions delivered to NCA5 in spring, 2022.
• Western U.S. dynamical downscaling completed for 3 CMIP6 ESMs for SSP370 (Rahimi et al, In Prep).
• MPI-ESM1-2-LR requires more bias-correction than CESM2 or CNRM-CM6.
• Bias-corrected dynamical downscaling has similar large-scale trends as non-bias-corrected dynamical downscaling (Risser et al, in prep).

Results to Date (1 of 2)

Rahimi et al, 2021, In Prep

Risser et al, 2021, In Prep

• Observational data is central to robust ESM downscaling, but it must be used cautiously.
• Model bias-correction based on historical observations appears to be useful for future projections.
• Downscaling should focus more on multiple models rather than multiple ensemble members.
• CMIP6 downscaling is under-way!

Investigator Contact Information:

drfeldman@lbl.gov

https://downscaling.lbl.gov

https://tinyurl.com/m9z6hmxb

Publications:

https://tinyurl.com/yfn6x8zu

https://tinyurl.com/vwzyft8e

https://tinyurl.com/c7aubre5

https://tinyurl.com/m5yzmjnk

Additional Resources:�http://loca.ucsd.edu/

​

Longmate et al, 2021, In Prep

Lessons Learned and Next Steps