The Research Support Desk (RSD) Initiative at Environment Canada: Linking severe weather researchers and forecasters in a real-time operational setting

1
The Research Support Desk (RSD) Initiative at
Environment CanadaLinking severe weather
researchers and forecasters in a real-time
operational setting

• David M. L. Sills1 and Neil M. Taylor2
• 1Cloud Physics and Severe Weather Research
  Section, Environment Canada
• 2Hydrometeorology and Arctic Lab, Environment
  Canada

2
Outline

• Background
• The RSD Initiative
• Objectives
• The RSD at OSPC in Toronto
• Scientific focus
• Field Experiment Support
• Verification
• Future Plans
• The RSD at PAPSC in Edmonton
• Scientific focus
• Field Experiment Support
• Verification
• Future Plans
• Summary

3
Forecast-Researcher Collaboration
Sustained collaboration between forecasters and
researchers, particularly when they are
co-located, shown to be a very fruitful endeavor
NSSL Spring Program 2003
Sydney 2000 Forecast Demonstration Project
4
National Laboratories
5
The RSD Initiative
The Research Support Desk (RSD) initiative takes
this one step further and facilitates the
interaction of forecasters and researchers in
real-time - a unique approach
6
The RSD Initiative

• Researchers also gain first-hand knowledge of
  the science gaps facing forecasters

7
Staffing an RSD

• Who should work at a RSD?
• Ideally, someone from research community with met
  expertise and some experience in operations
  (could be EC or perhaps university)
• Or could be a subject matter expert from
  operations with some research experience (MTs
  working in the national labs, or nearing
  retirement)
• In either case, excellent communication skills
  essential

8
Currently Two RSD Locations
ASPC
PASPC
QSPC
PSPC
OSPC
9
The RSD at OSPC in Toronto

• Located in operations area adjacent to summer
  severe weather lead desk and student desk

• One or more ResMets work shifts to cover as many
  potential severe weather events as possible
  between May and September

10
The RSD at OSPC in Toronto

• RSD tools have evolved since 2004 Aurora
  prototype object-oriented nowcasting workstation

• Two 21 LCD touch screens for use with Aurora
• operational forecast workstation (IFW)
• HAILCAST, RAOB, Burrows lightning, GEM-LAM2.5,
  etc.

11
Convective Nowcasting

• Science gap identified mesoscale analysis and
  convective nowcasting
• Extrapolation of radar echoes can give a first
  order estimate of intensification / dissipation
  of deep moist convection (MAPLE)
• Mesoscale boundary information can improve these,
  and allow convective initiation nowcasts
• Once boundary locations are known, other factors
  can be examined such as CAPE, CIN, shear,
  cloudiness, etc. ? obs, 1-hour RUC fields and
  conceptual models are used for this purpose
• Storm intensity and convective mode also need to
  be nowcast

The basis of a good nowcast is a good
mesoanalysis!
12
Optimal Human-Machine Mix

• Make best use of human strengths and expertise?
• Judgment / decision-making
• Pattern recognition
• Conceptual and mental models
• Adaptive strategies
• Make best use of machine strengths?
• Dealing with large volumes of data
• Integrating multiple datasets
• Handling complex calculations and complicated
  parameter interactions
• Automating product generation
• Enhance human expertise?
• Regular hands-on analysis, diagnosis and
  prognosis
• Conceptual models and dynamic mental models
  required

13
Aurora Nowcasting Prototype

• Aurora (Greaves et al. 2001) is a prototype
  nowcasting platform being used to test
  nowcasting concepts and techniques that may be
  implemented in NinJo
• Prototype is focused on optimization of the
  human-machine mix

• Uses a modifiable database made of gridded
  fields and object fields, including user-entered
  objects such as fronts

14
Aurora Nowcasting Prototype

• Ingests radar, satellite, surface observations,
  lightning and NWP (GEM-REG, GEM-LAM, RUC)
• Can automate the generation of any number of
  products from the database
• Has the capacity for expert system algorithms
  using objects and fields in the database,
  including those modified by the nowcaster

15
(No Transcript)
16
The basis of a good nowcast is a good
mesoanalysis!
17
Automated daily convective forecast verification
for 27 Jun 08
18
OSPC RSD on the Web

• Current and archived RSD products are easily
  accessed via an internal web site
  ospc76.to.on.ec.gc.ca/rsd
• Aurora data are archived to facilitate case
  studies of important events

19
Field Experiment Support

• Supported the BAQS-Met 2007 lake breeze / air
  quality / severe weather field study in SW
  Ontario
• BAQS-Met prognosis and analysis products were a
  subset of regular RSD products
• Communicated with field team during briefings
• Forecasters had access to special data sets in
  real time plus preliminary results

20
OSPC Forecaster Survey Results

• Anonymous survey results after 2005 season
• ? 100 were comfortable with a researcher in the
  operational area (initially 63 2004)
• ? 83 interacted with the RSD occasionally to
  very often (78 2004)
• ? 88 thought that the RSD improved the quality
  of OSPC forecasts / watches / warnings (64 2004)
• ? 72 thought the RSD provided an enhanced
  learning environment (59 2004)
• ? 83 had a positive overall impression of the
  RSD (81 2004)
• ? 94 wanted to see the RSD continue into the
  future (81 2004)

21
An Interesting Survey Result

• When asked about their preferred learning method
• mentoring during real-time events rated first
• simulations and COMET-type training modules
  tied for second
• Reading scientific papers and attending
  seminars rated a distant third and fourth,
  respectively

Mentoring is labour intensive for the researcher,
but over a period of years may be more efficient
than traditional broadcast methods
(presentations and papers).
22
Quantitative Verification

• Also working on ways to quantitatively verify RSD
  experimental techniques, and the impact of the
  RSD on operations more difficult than
  qualitative
• RSD operated on 50 of days in 2008 in order to
  compare severe weather warning verification stats
  for RSD and non-RSD days

23
Future Plans

• Possible operation of Ontario RSD in winter
  mode focusing on mesoscale features such as snow
  squalls and rain-snow boundaries
• - Test in winter 2004-2005 showed promise,
  especially with detecting rain-snow boundaries
  using newly developed polarimetric radar fields
• For summer 2009, invite researchers for day or
  two of double-banking on the RSD volunteers?

24
Future Plans

• Also for summer 2009, a storm-scale nowcasting
  prototype currently under development will be
  tested
• Science gaps identified -gt need for warn on
  forecast approach at storm scale and vastly
  improved warning region selection technology to
  improve lead times
• Forecaster should focus on meteorology, not
  product generation
• - Nowcaster would work with storm objects within
  Aurora (NinJo) and utilize conceptual / mental
  models
• - Warning bulletin generation would be automated

25
URP Cells, Tracks and Ellipses
ingested by Aurora
26
Summer Convective Nowcasting with Aurora

• Prototype based on input from URP cell tracker
  and URP severe weather algorithms
• Forecaster is able to add and/or adjust forecast
  tracks using the line editing tools in Aurora
• Forecaster is also able to delete tracks, or
  filter tracks by changing the cell intensity
  threshold (i.e. only display cells / tracks with
  centroids higher than 45 dBz)
• Modified forecast track used to provide content
  for warnings (e.g. affected regions / locations
  along track, speed, direction, etc.)
• Would eliminate the need for a warning region
  selector in NinJo

27

• Ellipse colour indicates convective trend for
  next 30 min
• red intensification
• yellow initiation
• green dissipation
• grey no change

• Cell node indicates location of reflectivity
  core, ellipse size indicates approx. size of 40
  dBz area (threshold is adjustable)

Radar echo
URP track

• Past cell nodes are open circles, current cell
  node is filled square, future cell nodes are
  filled circles

Surface boundary position from most recent
mesoscale analysis
2000 UTC
28

• Forecasters mental model supercell will
  continue veer rightward and intensify upon
  encountering the surface boundary, after which it
  will begin to dissipate and return to non-deviate
  storm motion

• Forecaster chooses 1-hr track (options are 30,
  60, 90, 120 min) and draws the curves
  representing the future tracks

2000 UTC
29

• New node labels are automatically placed along
  the track at 10 min intervals with equal spacing
  based on the track option chosen (in this case, a
  1-hr track was chosen)

• Node labels are enhanced (lighter grey) at 30
  min intervals

2000 UTC
30

• Aurora automatically places ellipse objects
  around cell node locations along the track

2000 UTC
31

• Trend is INTENSITY (current cell) INTENSITY
  (previous cell)

• Based on the forecasters mental model, the
  ellipses are modified to indicate intensity trend
  and cell size
• Intensity trend for each ellipse is chosen via a
  pop-up menu

Display cell attrbutes including SCIT data trends
(meso, hail, WDRAFT, etc.) for cell Convective
Trend Options O Intensification O Initiation
O Dissipation O No change
2000 UTC
32

• Based on the forecasters mental model, a severe
  thunderstorm warning area is drawn that
  encompasses all locations where severe weather is
  likely

• Nowcast uncertainty built into shape of warning
  area increasing coverage with time

2000 UTC
33

• A pop-up window allows the forecaster to check
  off which severe weather elements to include in
  the warning and add uncertainty information via
  free-form text

Severe Thunderstorm Warning O Hail O Damaging
Winds O Heavy Rain O Possibility of tornado
2000 UTC
34

• At 2010 UTC, new URP track guidance arrives and
  is displayed
• Red ellipses indicate a positive intensity trend
  forecast by URP
• Previous forecaster-entered track is also
  displayed

2010 UTC
35

• By default, the new guidance is accepted
• This allows for automated cell tracking if the
  forecaster is busy with other activities
• The forecaster can adjust this track if needed
  using Aurora line editing tools

2010 UTC
36

• However, the forecaster can also choose to draw
  a completely new forecast track
• A 90 min track is shown here, and cell node
  labels would be automatically added at equally
  spaced 10 min intervals along the track
• Labels are highlighted at 30 min intervals

2010 UTC
37

• In this case, the forecaster chose to use the
  previous track instead of the guidance
• Thus, the track start is automatically shortened
  by 10 min and filled square placed at new first
  cell node from the old forecast track

• The forecaster can adjust this track if needed
  using Aurora line editing tools

• Track length can be increased by 10 min to
  maintain a 1-hr track, or left as is to forecast
  the termination of the cell in 50 min

2010 UTC
38

• Also need the ability to easily add a new track
  for a developing cell in the case of convective
  initiation (a shower forecast to become a
  thunderstorm)
• In this case a 30 min track was selected
• Yellow and grey default ellipses yellow
  indicates cell initiation

Defaults
2010 UTC
39

• Forecaster adjusts convective intensity trend
  using the pop-up menu for each ellipse

Display cell attrbutes including SCIT data trends
(meso, hail, WDRAFT, etc.) for cell Convective
Trend Options O Intensification O Initiation
O Dissipation O No change
2010 UTC
40
Other considerations

• Dave Patrick at PASPC Winnipeg has made extensive
  revisions to the URP cell tracking algorithm
• Bill Burrows at HAL is working on a lightning
  tracker, so in the future either a blended track
  could be used, or the forecaster decides which
  track is best
• Similar approaches are being pursued in Australia
  (TIFS) and France (SIGOONS)
• Obvious challenge will be to keep workload
  reasonable, user interface must be intuitive and
  fast
• Approaches for other seasonal severe weather
  nowcasting still need to be explored

41
RSD at PASPC in Edmonton

• Beginning in 2006, a desk in strategic location
  in operational area
• RSD focus in Edmonton is on forecasting and
  nowcasting convective initiation
• Experimenting with new convective initiation
  parameters and indices using full-resolution 1 h
  GEM data e.g., mixed moist layer depth,
  low-level convergence depth, 0-LFC bulk shear

Convergence Depth
42
RSD at PASPC in Edmonton

• Experimented with graphic showing meteorological
  features of interest for convective initiation,
  and areas where convective initiation is
  predicted
• Accompanied by daily convective initiation
  discussion addressing con-cepts in more detail

http//hal-bobk.edm.ab.ec.gc.ca/RSD.html
43
Field Experiment Support

• Supported the UNSTABLE 2008 convective
  initiation field study to lee of Alberta Rockies
• Communication with field team during briefings
• Forecasters had access to special data sets in
  real time plus preliminary results
• Final experimental results disseminated to
  operations rapidly via the RSD

44
PASPC Forecaster Survey Results
From survey after 2006 season ? those
comfortable with a researcher in the operational
area grew from 79 to 89 over the season ? 63
interacted with the RSD occasionally to very
often ? 64 thought the RSD resulted in better
watches / warnings ? 53 thought the RSD provided
an enhanced learning environment ? 84 had a
positive overall impression of the RSD ? 100
would like to see the RSD continue into the
future ? Positive results are similar to those
from OSPC
45
Quantitative Verification

• Real-time subjective evaluation of selected
  existing and new fields will be conducted on RSD
  against radar, satellite, lightning data
• At end of season will verify combined fields via
  2x2 contingency table to assess skill in
  forecasts of CI

46
Future Plans

• Continue to develop and evaluate new NWP-derived
  fields related to convective initiation and
  severe storm forecasting / nowcasting
• - Will use the results from this work to improve
  first-guess forecasts of CI in future forecast /
  nowcast tools
• Investigate combining hourly updates of surface
  fields with GEM NWP data above the surface
• - Will serve to provide rapidly updated surface
  information valuable for storm environment
  analysis and diagnosis

47
Summary

• The Research Support Desk initiative at
  Environment Canada seeks to increase
  collaboration between researchers and forecasters
  by having researchers work directly and
  interactively with forecasters in a real-time,
  operational environment, particularly during
  severe weather events
• A RSD at the OSPC has been active during the
  summer seasons of 2004-2008, while an RSD at the
  PAPSC has operated during the summer seasons of
  2006-2008
• Surveys of forecasters at both locations have
  shown strong support for the initiative
• It is hoped that the concept will be implemented
  by the remaining SPCs across Canada

48
Acknowledgments

• Norbert Driedger, Emma Hung, Bob Paterson, Brian
  Greaves, Bill Burrows and Ron Goodson for
  technical invaluable assistance with programming
  and data feeds
• Stewart Cober, Ed Becker, Marie McPhee, Gary
  Burke, Bob Kochtubajda for supporting the
  initiative
• RSD ResMets Mike Leduc (OSPC 2005, 2007), Yoseph
  Mengesha (OSPC 2006), Patrick King (2006), Isabel
  Ruddick (OSPC 2007), Bryan Tugwood (OSPC 2008),
  Dave Sills (OSPC 20042008), and Neil Taylor
  (PASPC 20062008).

49
Questions?
50
References
Greaves, B., R. Trafford, N. Driedger, R.
Paterson, D. Sills, D. Hudak, and N. Donaldson,
2001 The AURORA Nowcasting Platform Extending
the Concept of a Modifiable Database for Short
Range Forecasting. Preprints, 17th International
Conference on Interactive Information and
Processing Systems for Meteorology,
Oceanography, and Hydrology, Albuquerque, NM,
Amer. Meteorol. Soc., 236-239.