ML20080M195
| ML20080M195 | |
| Person / Time | |
|---|---|
| Issue date: | 03/23/2020 |
| From: | Ahijevych D, Powers J, Prein A, Schwartz C, Sobash R Office of Nuclear Regulatory Research, National Center for Atmospheric Research |
| To: | Office of Nuclear Regulatory Research |
| T. Aird | |
| Shared Package | |
| ML20080M135 | List: |
| References | |
| Download: ML20080M195 (20) | |
Text
How well can Kilometer -Scale Models Capture Recent Intense Precipitation Events?
Andreas F. Prein , D Ahijevych, J Powers, R Sobash, C Schwartz National Center for Atmos pheric Res earch Photo by @KenGeiger 5th Annual Probabilis tic Flood Hazard As s es s ment Works hop, Feb. 19, 2020 This material is based upon work supported by the National Center for Atmospheric Research, which is a major facility sponsor ed b y the National Science Found ation und er Coop erative Agreement No. 1852977.
Convective outbreak Model Observation Correct representation of:
- Spatial structures
- Intensities
- Time evolution
Step Improvement in Simulating Intense Rainfall Storms x = 12 km (K-F scheme) x = 4 km x = 1 km
Deep convection in atmospheric models 16 times 625 times more more grid grid cells cells GCM grid spacing (~100 x 100 km)
- Deep convection is sub-gridscale process 100 km
- Needs cumulus parameterization When do we start to resolve deep convection?
- ~4 km horizontal grid spacing (Weisman et al. 1997) 100 km
Resolution of State-Of-The-Art Climate Models Resolution of State-Of-The-Art Climate Models Resolution of State-Of-The-Art Climate Models Resolution of State-Of-The-Art Climate Models Resolution of State-Of-The-Art Climate Models NRC project NR. 31310019S0015 Convection-Permitting Modeling for Intense Precipitation Processes Probable Maximum Precipitation (PMP) Convection-Permitting Models Does not allow quantification of Can they facilitate a more physically-based uncertainties in hazard estimates in probabilistic flood risk assessments?
either a physical or a risk sense.
Intense Precipitation Events in Eastern CONUS Daily, 1-in-5-yr precipitation amount for Evaluation in Four Regions 3646 stations for the period of 1950-2010 Kunkel et al. 2012
Convection-Permitting Model Simulations Dataset x Elements Period Region References NCAR Real-time 3 km 10-member 5/1/2015- CONUS Schwartz et al. (2014, Ensemble ensemble 12/31/2017 2015a, 2015b),
forecasts Romine et al. (2014)
NCAR MPEX 3 km & 10-member 5/15/2013- Central / Schwartz et al. (2017)
Ensemble 1 km ensemble 6/15/2013 eastern forecasts U.S.
NCAR Severe 3 km & Deterministic 2010-2017 Central / Sobash et al. (2019),
Weather Study 1 km forecasts; 500 eastern Schwartz et al. (2019) cases U.S.
- 10,570 36-hour WRF simulations/forecasts at 3-km horizontal grid spacing (1.8 mi)
- 810 36-hour simulations at x=1 km (0.6 mi)
Are Intense Precipitation Events Harder to Simulate?
Equitable Threat Score (ETS) Southern U.S.
Equitable Threat Score (ETS) [km]
Equitable Threat Score (ETS)
Thresholds
>= 5 mm/d Observed Storm >= 20 mm/d Simulated Storm >= 50 mm/d 0 1 Overlap []
Model skill increases with intensity of event Observed Precipitation Rate [mm/d]
Case Selection l Top 20 Events in Each Region Top 20 Events in Appalachia Region
Lagrangian Evaluation Framework West Virginia Flooding of 2016 Simulation has to capture:
- Track
- Movement speed
- Size evolution
- Precipitation volume
- Peak accumulation
West Virginia Flooding of 2016 Observed Accumulation Storm Speed Peak Peak Storm Intense Size Displacement Precipitation Volume Accumulation Intensity vs. Rainfall Elevation Strom Tracks
West Virginia Flooding of 2016 Observed Precipitation Best Peak Accumulation Best Peak Location Best Volume l 1 km Worst Overall Simulation
- Large spread due to initial condition perturbations
- 3 km and 1 km results are comparable
- 3 km seem to have too much rainfall on lee-side
Tropical Storm Bill l June 2015 Observed Precipitation Best Simulation Peak Accumulation Precipitation Volume Peak Displacement
Next Steps
- Assessment of model performance based on ensemble of intense events
- Quantification of systematic model biases
- Analyses of uncertainty sources to model performance
- Conceptual framework to use CPM simulations in Monte Carlo rainfall-runoff simulations Uncertainty Source Setting Horizontal grid spacing (x) 3 km, 1 km (1.8 mi, 0.6 mi)
Precipitation observations Stage-IV (Crosson et al. 1996, Fulton et al. 1998)
Mosaic WSR-88D (Zhang and Gourley 2018)
PRISM (Daly et al. 1994, 2002, 2008)
Newman (Newman et al 2015)
Initial Conditions Ensemble datasets to be used reflect initial condition perturbations
Summary and Conclusions
- Convection-permitting models can capture recently observed intense rainfall events east of the Continental Divide
- Predictability increases with rarity of event
- Sensitivity to initial condition perturbations is large
- 3 km and 1 km simulations show comparable results This work is sponsored by NRC under the Interagency Agreement Number 31310019S0015 prein@ucar.edu