Does nudging squelch the extremes in regional climate modeling?

1. Does nudging squelch the extremes in regional climate modeling? Tanya L. Otte, Martin J. Otte, Christopher G. Nolte, and Jared H. Bowden 10th CMAS Conference Chapel Hill, North Carolina 26 October 2011

2. Our Research Problem…Simplified Constraint of RCMtoward GCM Freedom of RCM todevelop smaller-scaleprocesses delicate balance Keeps RCM climateconsistent with GCM Decreases variability Allows RCM climateto deviate from GCM Increases variability More constraint toward GCM Less constraint toward GCM

3. unrelated solutions related solutions Connecting Larger-Scale Fields to Smaller-Scale Fields Larger-Scale Model Smaller-Scale Model output from larger-scale model regional output from larger-scale model output from smaller-scale model Need smaller-scale model to connect to and build upon larger-scale model. output from smaller-scale model taking cues from larger-scale model’s features 3

4. Nudging…Simplified • Used to constrain smaller-scale fields toward larger-scale fields • Need to determine: • How many to use? • Where to use? • Which color? 4

5. Constraining RCM to GCM with “nudging” • Use “observed” state to influence model (GCM  RCM) • Adds a non-physical term to the model equations • Term is based on difference between model and observation • Can be either toward: • Analyses(direct differences of fields in time/space) – Typical for CMAQ • Attractive when background fields are not significantly more coarse than target resolution • Horizontal wind components, temperature, moisture • Spectra (differences in wave decomposition in time/space) – New in WRF • Attractive when background fields are significantly more coarse than target resolution • Horizontal wind components, temperature, geopotential (~pressure) 5

6. WRFv3.2.1 forced by 2.5° × 2.5° NCEP Reanalysis 2 (R2) Three 20-Year continuous WRF runs with hourly output. Compare to 3-h, 32-km NARR, and 1-h, 0.31° CFSR products. 108-km: 81 x 51 36-km:187 x 85 • RRTMG Longwave and Shortwave Radiation • WSM6 Microphysics • Grell Ensemble Convective Scheme • Yonsei University PBL Scheme • NOAH Land-Surface Model Figure from Bowden et al., J. Climate, in press

7. Mean 2-m Temperature Difference (1988-2007) Nudging reduces warm bias in Plains and cold bias in Canada. Differences in complex terrain from topography in WRF vs. CFSR.

8. Mean Annual Precipitation (1988-2007) WRF overpredicts precipitation by ~250 mm annually. Patterns improved with nudging, particularly AN.

9. Monthly Area-Average Temperature Difference from NARR Both types of nudging consistently reduce error.

10. Monthly Area-Averaged Precipitation Total Higher highs Lower lows Greater variability without nudging: Is it real?

11. Monthly Area-Averaged Precipitation Total Compared to NARR, WRF is too wet. Nudging reduces erroneous peaks.

12. Annual Days with Temperature Relative to NCDC Threshold < 32°F > 90°F < 0°F NARR NN AN SN Nudging adds variability for extreme cold. Nudging lowers extreme high temperatures…will this verify?

13. Annual Days with Precipitation Exceeding NCDC Threshold > 0.1 in > 0.5 in > 1.0 in NARR NN AN SN WRF is generally too wet compared to NARR. Nudging, especially AN, makes extremes more realistic.

14. Nudging Does Not Appear to Squelch the Extremes in RCM • Both AN and SN improve means • 2-m temperature slightly warmer with AN than SN • Precipitation totals simulated better with AN than SN • Precipitation overpredicted by WRF, especially without nudging • SN has more variability than AN • Spectra (not shown) suggest AN coefficients are too strong for RCM • Will weaker AN coefficients improve variability, retain value? • Need hourly observations to validate variability of SN surface fields • Performance is consistent in most regions • Complex terrain qualitatively affects results