Coniglio Workshop

Impact of Doppler Radar Observations on the EnKF Analysis of a Developing MCS: 

Impact of Doppler Radar Observations on the EnKF Analysis of a Developing MCS Mike Coniglio David Dowell Lou Wicker Dave Stensrud National Severe Storms Laboratory

Motivation: 

Motivation Limited real-data EnKF experiments Does EnKF work with a messy convective event? Obstacles to getting robust storm-scale analyses using WSR-88D data? Refl-only vs. refl&vel assimilation? Build catalog of cases from diff convective modes to prove storm-scale EnKF concept Start with the 11-12 June 2003 MCS

11-12 June 2003 MCS: 

11-12 June 2003 MCS 2129z 2144z 2159z 2214z 2229z 2243z 2258z 2313z 2328z 2343z 2357z 0013z 0142z 0127z 0112z 0058z 0043z 0028z 0207z 0221z 0236z 0251z 0306z 0321z

11-12 June 2003 MCS: 

11-12 June 2003 MCS

Configuration of Experiments: 

Configuration of Experiments NCOMMAS ne=20-50 members ~ 400 km by 400 km domain Dx,y=2-3 km in model and obs analysis Data assimilated every 1-5 min for various 45-90 min periods Modified & perturbed OUN sounding LFO ice with various slope/intercept parameters Only reflectivity assimilated initially

Sense from Initial Experiments: 

Sense from Initial Experiments Solutions very sensitive to initial sounding Too little CIN – noise & spurious storms Need to set missing dBZs to a low value (0) Too much CIN – storms die after assimilation Storms/cold pool structures during assim. too small & too weak, not persistent Perhaps REFL-only assimilation not feasible? Try using reflectivity & radial velocity Too much convection in ground clutter Automated velocity unfolding and ground clutter removal implemented

Runs after 3-4 months testing: 

Runs after 3-4 months testing Runs are stable with good representation of two main clusters but getting too much convection in between main storms Obs Analysis EnKF Analysis

Runs after 3-4 months testing: 

Runs after 3-4 months testing Too much ground clutter getting through? No! Anomalous cold pool is an issue:

Dowell 8 May 2003 experiment: 

Dowell 8 May 2003 experiment What if TH isn’t updated with a velocity ob?

No update of TH with Vel: 

No update of TH with Vel Anomalous cold pool gone: BUT……..

PROBLEM: 

PROBLEM Noticed storm/cold pool growth slow Increase in initial sounding variability and bubbles didn’t help much dBZ and QR too low in low levels!

PROBLEM: 

PROBLEM dBZ and QR too low in low levels

Lack of Low-Level Reflectivity: 

Lack of Low-Level Reflectivity Could explain why updraft/cold pool structures too transient Microphysics or obs problem? What if reflectivity is withheld in precipitating areas? Don’t assimilate reflectivity < 15 dBZ LFO large hail setting not appropriate? -Change to graupel/hail mixture Change raindrop size distribution Are there more problems with using velocity? Don’t assimilate radial velocity at all

no assimilation of reflectivity > 15 dBZ, more graupel: 

no assimilation of reflectivity > 15 dBZ, more graupel Reflectivity important for maintaining storms!

Fewer but larger raindrops: 

Fewer but larger raindrops

Fewer but larger raindrops: 

Fewer but larger raindrops

Fewer but larger raindrops: 

Fewer but larger raindrops

No assimilation of velocity: 

No assimilation of velocity

No assimilation of velocity: 

No assimilation of velocity

Where to go from here: 

Where to go from here Nail down anomalous bubble problem Explore microphysics impacts Thorough comparison of refl only vs refl & vel experiments New ways to initialize ensemble/maintain spread Try new cases from “scratch”