Analysis and Prediction of Convective Initiation on 24 May, 2002 June 14, 2004 Toulouse 2 nd IHOP_2002 Science Meeting Ming Xue, William Martin and Geoffrey Stano School of Meteorology and Center for Analysis and Prediction of Storms University of Oklahoma

Our work on May 24, 2002 CI Case MS Thesis work of Geoffrey Stano: Multi- scale study based-on ADAS analyses including special IHOP data sets, and by examining observations directly. High-resolution simulation study (focus of this talk) Very large (~2000) ensemble and adjoint I.C. sensitivity study. dP/dqv sensitivity fields.

Objectives Understand convective initiation in this case Predict and understand the convective systems involved Assess the sensitivity of precipitation forecast to initial conditions

Methodology Make use of special data sets collected during IHOP Create high-resolution gridded data to perform diagnostic analysis and for model initialization Verify model simulations against available data Analyze realistic high-resolution model simulations to understand CI process

Synopsis of the Event Convection started between 20:00 and 20:30 UTC in Texas panhandle area along a dryline. An intensive observation of IHOP_2002. Rapidly developed into a squall line and advanced across Oklahoma and northern Texas SPC reported almost 100 incidents of large hail, 15 wind reports, and two tornadoes in central Texas

Time and Location of Initiation (17UTC – 22 UTC)

Isotach Analysis: 250mb TX Panhandle located at left-rear of upper level jet A short wave trough moved over western TX 19 UTC 20 UTC

CIN and CAPE Analyses CIN 19UTC CAPE 19UTC CAPE 20UTC CIN 20UTC

Results of Stano’s diagnostic study Favorable condition pointing to initial initiation near Childress, TX –Placement of 250mb jet max and minimum –Surface heating and high surface dew points –Low Convective Inhibition values Causes for initiation –Weakening of cap over the boundary layer by turbulently mixing –Break down of cap led to higher CAPE values –Convergence along the dryline or possibly from the cold front approaching dryline More specifics limited by data resolutions

Model Simulation Study Model can provide much more complete data in both space and time Easier to examine cause and effect Model fields are dynamically consistent Caution - model solution may deviate from truth therefore verification against truth is necessary

Forecast Grids 4

Model Configurations 1 km grid nested inside 3 km one ADAS analyses for ICs and 3 km BCs NCEP ETA 18UTC and 00UTC analyses and 21UTC forecast used as analysis background ARPS model with full physics, including ice microphysics + soil model + PBL and SGS turbulence LBCs every 3h for 3km grid and every 15min for 1 km grid 6 hour simulation/forecast, starting at 18 UTC

OBS Used by ADAS ARM COAG IHOP Composite Upper Air - rawinsondes KS Ground Water District 5 OK Mesonet SAO SW Kansas Mesonet Western TX Mesonet Profiler data absent

Surface Data Sets

Upper-Air Observing Sites

Dropsondes on 20 UTC Isodrosotherms

20:02UTC

20:32UTC

21:02UTC

21:32UTC

22:02UTC

22:58UTC

23:58UTC

Animation 20UTC-00UTC, KLBB radar

Animation 20UTC-00UTC, KFRD radar

3km model simulation/forecast

1 km simulation

T=2.5h 20:30UTC

T=3.0h 21:00UTC

T=3.5h 21:30UTC

T=4.0h 22:00UTC

T=4.5h 22:30UTC

T=5h 23:00Z T=5h 23:00Z

T=5.5h 23:30Z T=5h 23:00Z

T=5h 23Z

Animations Surface reflectivity 2km level w, winds See movies (18:30 UTC – 00:00 UTC)

See Movies

19:16UTC

19:45UTC

20:30UTC

23:00UTC

Vertical cross-section animations w and  w and qv qv and  e See Movies

+

+

+

X-section at y=170 km See movie

Conclusions Dryline convection occurred in favorable synoptic-scale environment (upper jet, sfc trough, convergence, moisture supply, CAPE, etc.) Realistic CI predicted by the model, especially at 1 km resolution Model CI associated with dryline was ~ 30min-1h late (related to spinup?) Active development of BL eddies observed before C.I. in convergence zone (50-100km wide) and in the moist air (underneath CAP)

Conclusions – continued A number of successfully penetration by B.L. parcels through the CAP before sustained convection is achieved Sustained convection occurred on the moist side near but not in the well mixed convergence zone or at the location of strong Td gradient Sloping terrain and surface heating played key roles Cold frontal convection started earlier, due to lifting Cold front or the triple point did not directly trigger dryline convection – dryline did it on its own and initiated convection further south Cold air surge to the N.W. of dryline did not directly trigger convection (was in dry air, no CAPE)

Conclusions - continued Role of surface wind convergence induced by mixing unclear to C.I. – perhaps indirectly by destabilizing the zone B.L. depth increases with time, so does its temperature, due to surface heating and eddy mixing) Any role of played by gravity waves or their interaction with B.L. eddies – not clear. Eddies organize into rolls/cloud streets near the edges of the convergence zone, where background wind is present Fine westward-propagating echoes observed by KLBB radar at the low levels are associated with the rearward propagating outflow boundary

Comments very welcome! Like to compare with fine-scale observations

Posters Dawson and Xue: Data sensitivity study on June 16 MCS case Liu and Xue: 3DVAR assimilation GPS slant path water vapor data with an isotropic spatial filter. Tested on simulated data from June 18, 2002 dryline case. Tanamachi: June 12, 2002 boundary layer water vapor oscillation case - undular bore? Observation and simulation study Sensitivity Study – Title Slide

TANGENT-LINEAR AND ADJOINT MODELS VERSUS PERTURBED FORWARD MODEL RUNS William J. Martin Ming Xue Center for Analysis and Prediction of Storms and the University of Oklahoma, Norman, Oklahoma

6 hr. forecast total accumulated precipitation

The sensitivity of a defined response function, J, such as the total rainfall in some area, is sought as a function of the initial fields. This is done by making a large number of forward model runs, each with a perturbation in a different location. The location of the perturbation is varied so as to tile a 2-D slice of the model domain. dx=dy=9km, no Cb parameterization

1 g/kg qv pert Example of an initial perturbation

Response Function Defined As: Sensitivity Defined As:

Sensitivity field for the dependence of total PPT on initial boundary layer moisture perturbations +1 g/kg perturbations