WWRP Liaison: K.A. Browning (Univ. Reading)
2. Environs and Climatology
Sydney, the Olympic venues, and associated observing
network, as shown in Fig.1 are located in a subtropical environment on coastal plains
100-150 km east of the Australian Great Dividing Range ( > 1000 m height). The
coastline is complex with many inlets and harbours. The S2000 Olympics are to be conducted
during spring when the weather is usually pleasant with mild to warm daytime temperatures
(maximum) and cool temperatures (10-120C) at night. Humidity is moderate.
However, the weather is diverse with potential for significant severe weather, lightning,
rain and wind events primarily associated with frontal passages and east coast lows. On
average there are 10-12 rain days per month at Sydney during the FDP period with the
monthly rainfall mean for September being 70 mm. The rainfall is variable ranging from a
minimum of 0.6 mm to a maximum 355 mm. Thunderstorms and severe weather days increase in
frequency during spring. Thunderstorm days range from an average of 2 per month in
September to 6 per month during November. Light to moderate SW to NW winds prevail during
the morning with sea breezes developing during the afternoon. Many of the early period sea
breezes will not reach the main Olympic venue but are important for the sailing courses.
September is quite windy with an average of 3.5 days per month recording winds more than
22 knots.
3. Goal and Emphasis
The goal of S2000 FDP is as follows:
To demonstrate the capability of modern
forecast systems and to quantify the associated benefits in the delivery of a real-time
nowcast service.
The emphasis is to be placed on the 0-6 h time
frame but incorporate a 0-24 h outlook. A combination of systems and techniques will be
employed. Observationally based techniques with special emphasis on radar systems will
provide the primary focus for the shorter duration nowcasts (0-2 h). It is anticipated
that these observationally based approaches will demonstrate the most impact over existing
capabilities. High resolution NWP techniques making use of mesoscale data sources and
state-of-the-art assimilation procedures will be employed for the longer duration
forecasts (2-24 h).
4. Proposed Project
Components
BOM Observational Network
For the September-November 2000 period, the BOM
observational network will be enhanced by the addition of a C-Band Doppler/polarimetric
radar at Badgery=s Ck. This will provide improved Aclear-air@
radar coverage, a dual-Doppler capability and polarimetric monitoring of the Sydney basin.
In addition, the BOM will deploy two additional boundary layer profilers; six portable AWS=s
(see mesonet in Fig.1) and four times per day rawinsonde soundings at Sydney Airport
during the 1999 trial and September-November 2000 FDP period.
Canadian Radar Processing and Decision System
(0-1 h time frame)
The Atmospheric Environment Service in Downsview,
Ontario, Canada has developed a system to help forecasters decide on the severity of
individual thunderstorms. This system is called the CAnadian Radar Decision System
(CARDS). It ingests 3-dimensional radar data and Doppler data, classifies individual radar
cells for hail, tornadoes, downbursts, heavy rain etc. and alerts forecasters to
potentially severe storms in prioritised order. Components include automatically generated
vertical cross sections through the bounded weak echo region of the cell, numerous cell
property displays and wind gust potential. Various components of the CARDS have been used
in operational forecast offices for up to the past ten years. These components are being
integrated into a comprehensive radar processing system and the CARDS will be ready at the
end of 1998.
Warning Decision Support System (0-2 h time
frame)
The National Severe Storms Laboratory (NSSL) in Norman
Oklahoma, USA has developed a system for helping forecasters make severe storm warnings
called the Warning Decision Support System (WDSS). The WDSS is a system that ingests
Doppler weather radar data, lightning detection data, surface data, and other weather data
sources that are available. It processes the data streams in real-time using image
processing, artificial intelligence and statistical techniques and provides important
warning information for meteorologists as well as innovative ways to view and access the
data in a timely manner. Components of the WDSS include a number of algorithms that
automatically detect and predict severe weather phenomena (e.g., tornadoes, hail, and high
winds) and an interactive display that was developed explicitly for meteorologists in a
busy warning environment. The WDSS has been tested in 16 different offices in the USA over
a four-year period.
Auto-Nowcaster (0- 2 h time frame)
The National Center for Atmospheric Research (NCAR) in
Boulder Colorado, USA has developed a nowcasting system for thunderstorms which is called
the Auto-Nowcaster. It is an expert system that automatically provides time and place
specific 0-120 min forecasts of thunderstorms. It ingests and integrates data from Doppler
radar, satellite, surface stations, upper-air soundings, lightning sensors and numerical
models. A recent unique feature is the ingest of output from a numerical model and its
adjoint which retrieves from single Doppler radar data very high resolution wind and
thermal fields in the boundary layer. The Auto-Nowcaster system automatically detects,
extrapolates and characterizes boundary layer convergence lines and convective storms.
Thunderstorm initiation, growth and dissipation are forecast by considering stability,
storm motion relative to convergence lines, low-level shear relative to convergence lines,
storm characteristics, and the presence of cumulus clouds in the vicinity of convergence
lines. The above data are converted into fields of forecast parameters which are then
operated on through a combination of conceptual model forecast rules and fuzzy logic
techniques. The Auto-Nowcaster has been tested in two operational forecast offices during
the past two years.
Generating Advanced Nowcasts for Deployment in
Operational Land-surface Flood Forecasts (0-3 h time frame)
The United Kingdom Meteorological Office (UKMO) and
University of Salford have developed a system called Generating Advanced Nowcasts for
Deployment in Operational Land-surface Flood forecasts (GANDOLF). This system for
forecasting convective cells utilizes an object-oriented procedure. Cells are classified
into five stages of development and cell development may be instigated using a life cycle
model with an objective method of recognising cell characteristics from three-dimensional
radar data linked to satellite and mesoscale model data. New cells are triggered at an
appropriate stage of their life cycle. Cells are moved using either model winds or motion
vectors derived from cross correlation of radar images at different times. Ten minute to
three hour forecasts are made on a 5 min cycle. GANDOLF has been evaluated operationally
over three summers by the U.K. Environmental Agency Thames Region covering the London
area.
NIMROD (0-6 h time frame)
The UKMO, Bracknell, Berkshire, U.K. has developed a
very short period forecasting system for precipitation, cloud and visibility called NIMROD
that has been operational in the UKMO since 1995. This fully automatic system ingests
radar, satellite, surface reports and numerical weather prediction model products. The
basic approach is to use linear extrapolation of present features and to incorporate non
linearities of evolution from the NWP model. Forecasts of 0-6 hr are made with a 30 min
update cycle although a 15 minute cycle is possible.
5. Mode of Operation and
Interactions with BOM
The BOM will provide an enhanced forecasting service to
support its commitment to S2000. Twenty-one extra forecasters will operate twenty-four
hours per day for forecast/weather duties, to provide warning services and interact with
media and Olympic venue operators. They will be located primarily at the BOM New South
Wales Regional Forecast Centre in central Sydney with a smaller group at Rushcutters Bay (
in support of the sailing community). Additional computer, engineering and technical
support will be available to ensure fault free operation. The BOM also has a nowcast
developmental program underway and this will provide support for the WWRP S2000 FDP
functions.
The WWRP FDP will involve a two-year program leading up
to the S2000 games as summarised in Appendix I. Each participating country will have one
person responsible for overall national coordination and to act as a focus for
interactions with Australia. These coordinators are J. Wilson (USA), B. Golding (UK), P.
Joe (Canada) and T. Keenan (Australia). Issues to be addressed include, tuning of
algorithms, system development, interfacing to different data sources, communications and
trial real-time testing, e.g. tuning of NSSL algorithms for the southern hemisphere,
gaining access to the BOM radar network and WWRP NWP systems will require adaptation to
local data feeds etc. It was recognised by the participants that the provision of high
resolution quality controlled radar (currently unavailable in Sydney) is necessary for all
the proposed systems. During the first year of the program, a small subcommittee involving
participants from NCAR, NSSL, Bureau of Meteorology Research Centre (BMRC), Environment
Canada (EC) and the UKMO will investigate issues related to radar data quality control,
e.g. mitigation of the effect of clutter, second trip, and anomalous propagation. This
subcommittee will interact with the European Cooperation in Science and Technology Program
on this issue.
A 1999 pilot program will be undertaken in
Sydney for installing, testing, interfacing and initial data collection. A final system
setup will be undertaken starting mid-August 2000 with all algorithm tuning to be
completed by 1 September 2000. After this date tuning of algorithms will be >frozen=.
The final FDP will be conducted through the period September 1 to 30 November 2000. The
extension past the formal Olympic period is dictated by the increasing frequency of severe
weather events throughout the spring. This extension will likely double the number of
thunderstorm cases available for study.
Each WWRP nowcasting/forecasting system will
have a representative at the BOM Sydney Office during the prime S2000 period. These
representatives will form a core of WWRP expertise within the project and provide a basis
for interaction with the BOM. The role of these WWRP participants will be to Achampion@
the cause of their particular system within the joint WWRP FDP, and to undertake Atraining@
of BOM Atrainers@. A rotating WWRP Manager will be selected from this
group to act as the local overall WWRP advocate. This manager will interface directly with
BOM forecasters to present the consensus WWRP FDP forecasts. This person will participate
in daily BOM functions including policy weather discussions etc. and ensure proper
day-to-day management of the WWRP FDP program.
The BOM will employ one person in Sydney to support the
WWRP S2000 FDP requirements in Sydney. This person will be the focus for local issues for
WWRP participants. This person will be responsible for monitoring the operational status
of WWRP systems, assuring proper functioning and ongoing evaluation of their utility
(especially during and after the 1999 trial), and undertaking training of BOM forecasters.
The training role will be essential for optimum use of WWRP FDP products during the trial.
Space will be provided by the BOM adjacent to the
forecaster area for use by the WWRP participants to house equipment and other facilities.
BOM will provide upgraded communication facilities to support primary WWRP activities,
e.g. timely communication of radar data and agreed upon access to data sources and
numerical data.
6.S2000 Forecast Process
and FDP Products
The BOM will provide for all venues 0-48 h forecasts of
weather elements (temperature, humidity, wind, rain and pressure) at 3 hourly intervals.
These forecasts will be updated twice per day. Briefings of Olympic Organisers will also
occur twice per day. An 8 am briefing will provide a basis for a review of existing plans
given current forecasts and latest information. A 4 pm briefing will be employed to
address weather related issues affecting the activities planned for the following day.
Special forecasts will be issued as required in the event of rain, thunderstorms or strong
wind conditions. These warnings will be updated at 15 minute intervals. For sailing venues
there will an additional briefing and requirements include hourly forecasts of conditions
for the period 0800-1600 relating to wind speed and gusts. The WRRP FDP will provide
information relevant to enhancing the information and quality of these forecasts.
Diagnostic and end-to-end user FDP forecasts
will be provided in real time meeting the same operational deadlines as the overall BOM
S2000 forecasts. Forecast users in this sense include Aend users@ and ABOM
forecasters@. The user for a diagnostic FDP product is the official BOM forecaster.
The intention is to provide additional products, currently unavailable in their
operational setting, which will improve the quality, extent and value of their forecasts.
This information will be primarily in the form of new fields, e.g. tracks of radar
diagnosed severe and ordinary thunderstorms, size and swath of hail, areas of convective
initiation, NWP fields based on latest assimilation procedures at 0.5-1 km resolution
(suitable to resolve complex topographic influences including those associated with Sydney
Harbour). End-to-end products will be those in a form suitable for use by all forecast
users including the BOM forecasters. Meteograms for each venue will be the main form of
such products. Non-NWP techniques (WDSS, Auto-Nowcaster, NIMROD, GANDOLF) will derive
forecasts on a common 15 minute cycle.
Some example FDP meteogram type products include.
NIMROD: 15 minute cycle, 0-6 h forecasts at 15 minute
intervals of accumulated and instantaneous rainfall.
GANDOLF: 5 minute cycle, 15 minute interval forecasts
of instantaneous rain rate and storm positions.
Auto Nowcaster: 15 minute cycle, 15, 30 ,60 and 120
minute interval forecasts of storm position, rain rate, windspeed and direction.
WDSS: 5 minute cycle, storm location, storm motion,
forecast storm location, probability of hail, probability of a tornado, probability of
severe winds (20 minute forecasts) and downburst prediction (10 minute forecast)
CARDS: 5 minute cycle, point probability forecasts of
rain (60 minute).
NWP Schemes: Temperature, humidity, pressure, wind
and rainfall forecasts to 18h.
The time interval for NWP meteograms will be 5 minutes.
A limited set of agreed upon WWRP FDP products will be
made available to S2000 BOM forecasters on their BOM Australian Integrated Forecast System
(AIFS) workstation. AIFS is the primary machine for diagnosis and forecast preparation in
use by the BOM. The primary fields to be available in this manner will include:
a) Hourly analyses and forecasts of surface
temperature, humidity, wind vectors, divergence, CAPE, equivalent potential temperature,
pressure and precipitation.
b) WWRP FDP meteograms at Olympic venues including
those of primary meteorological variables and forecast weather elements
c) Radar product information (WDSS, NIMROD, GANDOLF
display), Auto-Nowcaster product display.
d) WWRP IBM INFO 2000 public analysis product.
These products will be provided in agreed upon form and
this will be developed during the period leading to September 1999 trial. If forecasters
require additional information on WWRP products, it will be available through local WWRP
displays located adjacent to the BOM forecaster area.
A set of WWRP FDP and BOM forecasts will be
subjected to evaluation. To this end BOM official forecasts and BOM model forecasts, Model
Output Forecasts (MOF) and WWRP Ameteogram@ type products will be archived in
a common format on a common system for use by the WWRP verification procedure. All WWRP
diagnostic field type products will be archived separately.
It is also proposed to have an agreed upon WWRP/Nowcast
graphical product showing current weather available for general access from the BOM web
and the IBM INFO2000 system. The latter is the primary form of communication available to
all S2000 participants and includes venue managers, athletes, and commercially interested
parties. The form of this WWRP product will be derived after consultation with the BOM and
the WWRP (see next section).
7. Verification and Impact
The verification process will be aimed at assessing the
additional benefit provided by WWRP FDP products. The verification will be based on
objective measures and subjective forecaster type evaluations. WWRP assistance is to be
sought for this through the establishment of a verification advisory group. The charter
will be to derive details of evaluation/verification procedures in accordance with the
aims of the WWRP and this FDP. Indeed the S2000 FDP will serve to focus attention in this
area. The verification advisory group will be responsible for neutral evaluation of the
FDP products.
A common forecast data set will be created as indicated
above. No skill and control forecasts such as persistence, climatology and linear
extrapolation, NWP and Model Output Statistics (MOS) will be included in the forecast data
set. In the case of venue specific meteograms, verification will be undertaken against
automatic weather station data. Standard verification parameters will include mean, root
mean square error, bias, correlation, probability of detection, false alarm rate, threat
index, critical success index etc.
NWP fields (predictions and analyses) of surface
temperature, humidity, and wind will be evaluated against analysis fields and AWS data.
Forecast rain fields (NWP and radar-based) will be evaluated against radar fields or
whenever possible with ground truth information. Storm track and associated variable
forecasts (mesocyclone, hail) forecasts will be evaluated against radar derived
information. The use of phase shifted cross-correlation schemes for field verification in
situations with time and space errors are being investigated.
Forecaster evaluations although often less objective
are considered an essential component of the whole process. The verification process will
include forecaster assessment of the various WWRP tools. Each day as part of the
forecasters activities a summary will be produced providing an assessment of the WWRP
products. These subjective interpretations will provide ongoing feedback to the
forecasters and the WWRP participants on each system and their products.
A small advisory group based on existing WWRP
expertise will be established to consider such issues related to the impact of FDP
products on end-users. This group will advise on the design of user-oriented products and
appropriate questionnaires etc. An analysis of frequency of Ahits@ or usage
of the WWRP products on INFO2000 and the BOM system is possible. Comparison with other
products could be used to evaluate the impact of WWRP products.
8. FDP Qualifying
Attributes
The International Science Steering Committee of the
WWRP has stated the following attributes must be addressed in a FDP proposal. Each is
specifically addressed below.
Attribute a) The project addresses forecasts
of weather of international applicability, with emphasis on high impact weather.
The primary forecast products for S2000 will be severe
weather, lightning, wind and rain. As indicated above in the climatology discussion for
Sydney, convective severe weather is a definite possibility during the Olympics,
particularly so during the Paralympics and the following one month extension of the FDP.
Nowcasting severe weather is obviously a high impact weather phenomenon that affects many
locations all over the world. We make specific note of wind forecasts that will be made at
each venue and the Sydney airport. Forecasting wind shifts when winds are above 15 knots
is particularly important for the efficiency of aircraft operations at airports around the
world.
Attribute b) Existence of clear evaluation
protocol
As discussed in section 7 objective verification of the
WWRP FDP products will be undertaken.
Attribute c) Expectation of success and level
of support available
The likelihood of success is very high. BOM support for
the S2000 FDP is considerable given the resource provisions and the commitment of staff to
the program. Examples include staff support in Sydney for the year long WWRP system
development and forecaster training phase starting September 1999; provision of improved
observational networks and communication systems, ongoing scientific support for the
implementation of radar quality control etc.
The FSL LAPS and NSSL WDSS were previously used in the
Atlanta Summer Olympic games in 1996 and were both sited in forecaster surveys to be of
significant benefit and added value. In addition the accuracy of preliminary severe storm
warning statistics from the Sterling, Virginia National Weather Service Office for the
summer of 1998 has shown a major improvement in accuracy. It was during this time that
both the WDSS and Auto-Nowcaster systems were in place at that office.
The level of support from the contributing partners to
this proposed FDP is covered in Appendix II. Four countries have stated their intentions
(based on approved funding) to contribute significant hardware, US$1.8 million of in-kind
contributions and US$1.8 million of specific additional funds toward this FDP.
Attribute d) Prospect of a clear advance on
current local or global operational practice
Presently the Sydney forecast office does not have a
Doppler radar or concentrated nowcasting system support. Thus the advances will be clearly
apparent as has been noted world wide when Doppler radars are installed. The more germane
question is whether the forecast procedures offered in this FDP will show advances over
what would be expected if the Sydney office had a Doppler radar for some time. Insight to
this question can again be obtained by looking at the Atlanta Olympic experience where two
of the eight forecast techniques being proposed here were implemented. In that case the
improvements were clear and there is reason to believe advances will be even more apparent
for Sydney when even more advanced tools are implemented.
Attribute e) Forecasts will be provided in
real-time and forecast information will be communicated for user utilization and
subsequent impact evaluation
This attribute is specifically addressed and satisfied
as discussed in sections 5, 6, and 7.
Attribute f) Suitability of the techniques,
systems, or skills comprising the demonstration project for subsequent transfer to members
for operational use.
The suitability of transferring the techniques
discussed here depends heavily on the ability of some member nations to acquire Doppler
radars and high speed computers. Clearly those countries with Doppler radars could utilize
the WDSS, Auto-Nowcaster, CARDS, and GANDOLF to significantly improve their ability to
nowcast high impact convective weather events. Those countries with non-Doppler radars
could likely use NIMROD and GANDOLF to improve precipitation and convective storm
forecasts. The use of ARPS and LAPS presently requires expensive very high speed
computers. However the cost of such computers is constantly decreasing and before too many
years will be affordable to most countries.
The creation of the radar quality-control subcommittee
will be well received by member weather organisations. This is a fundamental issue in the
use of radar data and a comprehensive and international discussion including a comparison
of techniques will be of great side benefit.
Tuning of algorithms before the pilot and final
demonstration projects in Sydney will be of benefit to member countries in establishing
the evaluation procedures and protocols required in technology transfer.
9. Summary
A proposed WWRP FDP to provide enhanced nowcasting
support for the S2000 Olympics has been described. The primary emphasis for the proposal
is supporting 0-6 h nowcasts but including a 0-18 h outlook. It represents a unique
opportunity to demonstrate the utility of modern nowcast systems and procedures in a high
impact weather situation. Component systems have been demonstrated in operations in the
U.K. (NIMROD, GANDOLF), Canada (CARDS) and U.S.A (WDSS, Auto-Nowcaster,). The execution of
this program involves synergistic action from various agencies as summarised in Appendix
I. Funding agencies have been approached and will look seriously at supporting the
respective components subject to the usual caveats.
This FDP will provide enhanced observing capabilities
coupled with nowcast systems and procedures currently not available within the BOM. In
this sense it will provide a clear advance over current operational practices. The S2000
FDP will provide specific products to forecasters and end-users. Meteograms for the
various S2000 Olympic venues with diagnostic and prognostic fields relevant to storm
occurrence, rainfall, severe weather, wind-shifts etc. will be included in the BOM
forecast system for easy access by the forecasters. These products will be derived from
observationally based techniques and NWP systems. All of these products are important in
the context the S2000 forecasts. A specific end user WWRP product will be available
through INFO2000 and from the BOM. WWRP products will be subjected to forecaster and
objective evaluation/verification and their impact on users assessed. The involvement of
WWRP expertise is expected to support the design of products, evaluation procedures and
study of product impact upon users.
Appendix I. Time Schedule
Appendix II. Resource Allocations
The following is an estimate of resource
allocations for the FDP. Staff allocations include those working on system support
(software, engineering) and scientists. In-kind represents resources, e.g. salary provided
from existing revenue base but applicable to total FDP costs. Additional funds represent
new funds being sought to support the actual FDP.
Note all costs are quoted in US$.
*Support being sought from BOM over two years.
Fig 1. Olympic Venues