1. Neil Taylor1 and Bill Burrows2
Using GEM REG Output for Forecasts of Convective Initiation on the Canadian Prairies
Experimental Techniques Under Development
in the Hydrometeorology and Arctic Lab (HAL)
Neil Taylor1 and Bill Burrows2
1Hydrometeorology and Arctic Lab
2Cloud Physics and Severe Weather Research Section / Hydrometeorology and Arctic Lab
PASPC Spring Workshop
5/7 May 2009

2. Outline Introduction and review from 2008
Some parameter relationships to observed lightning from 2007 and 2008 convective seasons
Working towards objective forecasts of convective initiation – combined fields
Preliminary objective verification of combined fields
Summary / Future work

3. Introduction
Timing and location of convective initiation (CI) is of primary importance to forecasters during the summer months
On short timescales, and once convection is underway, observational data are the key to success
On longer timescales, and to fill in data voids, an (intelligent) integration of observational and NWP data can be used
GEM REG will not explicitly resolve storm-scale processes but may help identify CI and storm-supporting environments
Can we apply scientific understanding of CI to available model data to generate (better) objective first-guess convective initiation forecasts?
That will be consistent with forecaster conceptual models and observed data and will not replace a rigorous convective analysis / diagnosis / prognosis
In the future, can these forecasts then be integrated into the next generation of forecast production tool / nowcasting system?

4. Review
Climatological studies suggest the troposphere is conditionally unstable on most days during the summer months
ABL processes are important for CI
Observational and modeling studies on CI suggest that thermodynamic and kinematic processes related to CI are linked
CI has long been associated with convergence boundaries (70% or more of storms associated with detectable boundaries, e.g., Wilson and Schreiber 1986)
can influence local climatology, tornadogenesis & other severe weather
We have targeted the following with respect to CI:
ABL water vapour (including moisture depth)
ABL convergence (including convergence depth)
Low-level wind shear and buoyancy (affecting parcel trajectories in / near the ABL)
Convective inhibition

5. ABL Moisture
Higher Td (mixing ratio) can promote CI and lead to stronger updrafts
Lifted parcels entrain dry air below cloud base
ML lifting attempts to account for this
deeper mixed moist layers should contribute to less entrainment and stronger ABL updrafts
Depth of moisture is an important factor for CI
Consider a Surface-Based Parcel
LCL
LFC
Shallow ABL Moisture
Deep ABL Moisture
Lifting parcels using a mean layer mixing ratio is an attempt to account for dry-air entrainment below cloud base.
Unsaturated lifted parcels will entrain dry air
A deep, well-mixed ABL will reduce sub-cloud entrainment => stronger updrafts

6. ABL Convergence
Convergence boundaries can act to provide lift and deepen moisture
Strength of convergence => strength of ABL updraft
Boundaries associated with secondary circulations
Divergence below LFC can offset low-level convergence
Parcels have a better chance of reaching their LFC if they remain under the influence of low-level convergence (e.g., Ziegler and Rasmussen 1998)
Both magnitude and depth of convergence are important factors for CI

7. ABL Convergence
Divergence occurring below the LFC can weaken ABL updrafts and inhibit or delay CI
LFC
Convergence
Divergence
Convergence boundaries are often associated with secondary circulation
Convergence locally deepens moisture

8. ABL Convergence
Deep
Convergence
Divergence
LFC
Deep layer convergence acts to reduce divergence below the LFC and acts to maintain the influence of low-level convergence (i.e., deeper lift) on ascending parcels thus promoting CI

9. Low-Level Vertical Shear
Vertical shear below the LFC can act to tilt parcel trajectories downshear and increase entrainment (e.g., Markowski et al. 2006)
Updrafts should remain vertical and deeper if associated with convergence and if shear on either side of the boundary is equal in magnitude (Rotunno et al. 1988)
Applies for both stationary and moving boundaries
Shear near and just above the LFC can act to tilt updrafts downshear to delay CI and storm organization
Strong shear below the LFC and near and just above the LFC may act to delay or inhibit CI
If updrafts persist this shear may act to enhance storm organization over time

10. Low-Level Vertical Shear
Weak shear near the LFC promotes deeper penetration of ABL updraft and organized convection
Strong shear near the LFC (within 2 km ?) may delay or inhibit CI and organized convection
LFC
Vertically oriented updraft minimizes dry air entrainment below cloud base
As trajectories are tilted parcels may entrain more dry air to reduce buoyancy and weaken updrafts
Weak Shear Below LFC
Strong Shear Below LFC
A role of convergence boundaries may be to promote vertical updrafts (Markowski et al. 2006) to support CI

11. Progress To Date…
In 2007 a suite of NWP fields were introduced for preliminary consideration and evaluation in operations
Consisted of new fields targeting the concepts previously mentioned and a number of severe weather forecast fields that were previously unavailable (at least hourly)
Preliminary and informal evaluation conducted on RSD in 2007
Fields available in 2008 and used in support of UNSTABLE operations (15 km, 2.5 km and 1.0 km GEM) – but not yet formally evaluated
Parameter values interpolated to CG lightning flashes in space and time using data from May-Sep 2007
Results and subjective estimates used to define combined fields associated with observed lightning
Previous and new combined fields interpolated to observed CG flashes using 2007 and 2008 data (1,865,694 CG flashes)
Preliminary objective verification underway
Real-time subjective evaluation on RSD in 2009

12. New Combined Fields Moisture Supply and Depth
Depends on
50 mb mean Td
Mixed Moist Layer Depth
BLUE
50 mb Td ≥ 4 °C
Moist. Depth ≥ 200 m
GREEN
50 mb Td ≥ 8 °C
Moist. Depth ≥ 500 m
YELLOW
50 mb Td ≥ 10 °C
Moist. Depth ≥ 750 m
RED
50 mb Td ≥ 12 °C
Moist. Depth ≥ 1000 m

13. New Combined Fields Convergence and Depth
Depends on
surface divergence (SDIV)
convergence depth (COND)
ratio of convergence depth to MLLFC height (RCLF)
BLUE
SDIV ≤ -3x10-5 s-1
GREEN
SDIV ≤ -5x10-5 s-1
COND ≥ 500 m
YELLOW
SDIV ≤ -10x10-5 s-1
COND ≥ 1000 m
RCLF ≥ 0.5
RED
SDIV ≤ -15x10-5 s-1
COND ≥ 1500 m
RCLF ≥ 0.75

14. New Combined Fields Low-Level Shear and 0-3 km SBCAPE
Depends on
0-MLLFC Bulk Shear (0-MLLFC Shr)
LFC + 2 km Bulk Shear (LFC+2 Shr)
0-3km surface-based CAPE (SBCAPE)
Most unstable CAPE (MUCAPE) ≥ 100 J kg-1
BLUE
0-MLLFC Shr ≤ 40 kt
LFC+2 Shr ≤ 25 kt
GREEN
0-MLLFC Shr ≤ 30 kt
LFC+2 Shr ≤ 15 kt
0-3 SBCAPE ≥ 75 J kg-1
YELLOW
0-MLLFC Shr ≤ 20 kt
LFC+2 Shr ≤ 10 kt
0-3 SBCAPE ≥ 125 J kg-1
RED
0-MLLFC Shr ≤ 10 kt
LFC+2 Shr ≤ 5 kt
0-3 SBCAPE ≥ 200 J kg-1

15. New Combined Fields Convective Inhibition
Depends on
Most unstable CIN (MUCIN)
Difference in MULFC and MULCL height (MULFC-MULCL)
Most unstable CAPE (MUCAPE)
BLUE
MUCIN ≤ 100 J kg-1
MULFC-MULCL ≤ 1000 m
GREEN
MUCIN ≤ 50 J kg-1
MULFC-MULCL ≤ 500 m
MUCAPE ≥ 0 J kg-1
YELLOW
MUCIN ≤ 25 J kg-1
MULFC-MULCL ≤ 250 m
MUCAPE ≥ 100 J kg-1
RED
MUCIN ≤ -1 J kg-1
MULFC-MULCL ≤ 100 m
MUCAPE ≥ 250 J kg-1

16. New Combined Fields Objective CI Forecast
Uses selected components of previous fields
Moist. Depth
SDIV
COND
RCLF
MUCIN
MULFC-MULCL
MUCAPE
0-3 km SBCAPE
0-MLLFC Shear
MLLFC+2 km Shear

17. Some Preliminary Verification
Combined fields were also interpolated to observed CG lightning flashes (nearest grid point value) using 2007 and 2008 data from 1 May to 30 Sep
Contingency tables were constructed to evaluate POD, FAR, CSI and other parameters for comparison with existing objective forecasts of thunderstorms
Kain-Fritsch deep convection parameterization scheme
A pseudo-SCRIBE method (total 1 hr precipitation ≥ 0.2 mm and Showalter Index ≤ 0 – does not include use of 30% PoP as in SCRIBE)
Cloud Physics Thunder Parameter (Bright et al. 2005) to determine if the “environment” will support lightning
Parameters designed to target CI
Working on verification of CI-only flashes
Are certain time periods more useful than others (e.g., for SFC-based storms)?
Model inability to specify “right place, right time” influences objective verification scores (e.g., position / timing of convergence features, thermal or other features aloft, etc.) => subjective verification req’d

18. Probability of Detection 12 UTC Run: T+3 (T+1) to T+18 for 2007 (2008)
Good POD
Bad POD

19. False Alarm Ratio

20. Critical Success Index

21. POD By Hour: Low-Level Shear and 0-3 km SBCAPE
Marked Improvement in Typical CI Period

22. FAR By Hour: Low-Level Shear and 0-3 km SBCAPE
Some Improvement in Typical CI Period

23. CSI By Hour: Low-Level Shear and 0-3 km SBCAPE
Marked Improvement in Typical CI Period

24. Future Work
Real-time subjective verification of parameters will be main focus of RSD work in 2009
Verify parameters against CI-only flashes (CI flash identification under development)
Design fields targeting elevated convection specifically
Generate hourly fields incorporating surface observations (a poor man’s RUC system) to move towards a more useful nowcasting tool
Incorporate selected predictors into a new statistical lightning forecast
Formulate statistical forecasts of CI (similar approach to existing statistical lightning forecasts)
Application to ensemble forecast data (?)
Exploit this dataset of lightning-correlated NWP parameter values to characterize GEM REG behaviour (e.g., regional, SFC-based vs. elevated storm environments, etc.)

25. Summary
GEM REG (or other models) won’t consistently pinpoint CI (model-generated cold pools, boundaries in wrong place / wrong time) but may provide useful information about the mesoscale environment
Parameters have been developed to characterize processes in the ABL that have been shown (in reality) to be important for CI
New combined fields will be available to operations in 2009
Keep in mind these are still be verified / adjusted – experimental only
If you do use them please let me know what you think
Interpolation results suggest some preferred parameter values that may be useful for forecasting lightning / CI
Preliminary objective verification of combined fields indicates reasonable POD but poor FAR
CSI poor for all parameters
KF and pseudo SCRIBE methods => low POD but slightly better FAR
Some improved stats during typical CI period for daytime convection