Distributed Modeling
Implementation Project Plan
1. Project
Team and Focal Points
Office of Hydrologic
Development (OHD) - Lee Cajina
Apex Digital Systems, Inc. (Apex) - Stephanie Liu
Arkansas/Red River Basin River Forecast Center (ABRFC) - Billy Olsen
West Gulf River Forecast Center (WGRFC) - Jerry Nunn
2. Background
The OHD- Hydrologic
Science and Modeling Branch (HSMB) proved there are potentially
large benefits in river forecasts from distributed modeling. The
results from using the Hydrology Laboratory Research Modeling System
(HL-RMS), a distributed modeling system prototype developed at OHD,
were very positive during the Distributed Modeling Intercomparison
Project (DMIP). To build on this work, OHD management has identified
implementing an operational distributed river forecasting model
as one of its priorities.
Although the
hydrologic aspects of distributed modeling have been researched
for several years, the software engineering challenges of implementing
a nationally supported distributed modeling river forecasting system
have only recently been considered. Implementing this system presents
a big challenge because the software and system architecture of
the National Weather Service River Forecasting System (NWSRFS) was
not originally designed for distributed modeling, and although there
have been significant enhancements to NWSRFS over the last 20 years,
the basic underlying architecture is still the same. Implementing
an efficient and robust distributed modeling system will require
significant changes to the current software and system architecture.
To assist OHD
in engineering a distributed model river forecasting system we have
teamed with Apex Digital Systems, Inc. Apex has been designing custom,
scientific software systems for over four years, and has expertise
in all phases of software engineering. To ensure we deliver a system
that is scientifically sound and useful for operational river forecasting,
we will also solicit feedback and participation from RFCs. We have
asked the ABRFC and WGRFC to be a part of our team. The collective
expertise of each team member will help engineer a useful system.
3. Goal
and Current Scope
The end goal
of this project is to develop a software system that will support
the distributed modeling requirements of a wide range of users.
Ultimately, this will include tools to support RFC forecasters and
developers, WFO forecasters, and research scientists. By working
through a formal software design process and identifying the needs
for modernizing the software and system architectures currently
used for river forecasting, the system that results from this project
should be capable of supporting existing scientific algorithms for
distributed modeling and, perhaps more importantly, be flexible
enough to accommodate evolving scientific capabilities and user
needs.
Since initial
successes with the HL-RMS prototype have been in the research setting,
an important part of the design process will be to identify the
improvements required for effectively using the model in an operational
setting. Thus, gathering feedback from forecasters is a critical
component of the design process. After some initial requirements
gathering work and discussions with ABRFC forecasters who have begun
testing the HL-RMS prototype, the project team has concluded that
without short term enhancements to the existing HL-RMS prototype,
it is not possible to understand the benefits and limitation of
the current system in an operational environment and to properly
define longer term requirements. Given this, the Distributed Modeling
System (DMS) development project has been divided into two phases.
DMS 1.0, which does not include any system architecture changes,
will be a short term project designed simply to enhance the current
HL-RMS prototype to facilitate testing and more meaningful requirements
gathering for DMS 2.0. Efforts for DMS 1.0 will focus on improving
the HL-RMS prototype so both RFCs testing the software can help
us determine some of the science and system related requirements
needed to deploy an operational distributed modeling system. At
the end of DMS 1.0, the DMS team and OHD/HL management will decide
whether it is best to address some of the RFCs science and/or system
related concerns through another version of the prototype (DMS 1.X),
or whether moving into DMS 2.0 is a better alternative. Efforts
for DMS 2.0 and beyond will focus more on software architecture
changes required to deliver a more robust and flexible distributed
modeling framework.
4. Design
Approach
Apex has
designed a requirements development approach that reflects the unique
nature of this project. The general tasks in the requirements definition
process are:
- Technical
review
- Define
user classes
- Define
high-level use cases
- Develop
high-level current state functional diagrams
- Develop
high-level future state functional diagrams
- Field
Review
- Define
detailed use cases
- Document
current data model and data flows
- Design
future state data model and data flows
- Design
interface mockups, where applicable
- Design
test plans
5. Deliverables
The deliverable
at the end of DMS 2.0 will be software integrated into AWIPS, including
documentation, which will allow all the identified users use distributed
modeling for river forecasting. For DMS 1.0, we will deliver an
enhanced prototype to two RFCs.
6. Success
Criteria
The final
indicator of success is the capability to generate river forecasts
using distributed modeling within AWIPS. As intermediate indicators
we have identified milestones as part of a detailed project schedule.
7. Project
Schedule Details
DMS
1.0
The current
project schedule for DMS 1.0 consists of tasks in the areas of project
planning, requirements development, implementation, testing, and
project summary.
DMS
1.0: Project Planning
This phase
of the project, which ends in late May 2003, is where we clearly
define the objectives and scope of the project using guidance from
OHD management. As part of project planning, we have already gathered
comments about the existing HL-RMS prototype and potential enhancements
from ABRFC, currently the only RFC with experience using HL-RMS.
The information was collected during a two-day discovery session
at ABRFC on April 15 and 16, where we discussed software engineering
issues related to implementing a distributed modeling system. Another
reason for our visit to ABRFC was to develop a better understanding
of the workflows at an RFC. We feel in order to produce an operational
system, it's necessary to have a basic understanding of what daily
operations consists of for an RFC forecaster. During project planning
we will also be selecting a second RFC to test the enhanced prototype
(DMS 1.0) and gather feedback from. A second RFC will give us a
better representation of future users.
As part
of project planning, we have also been planning on how to execute
the project by creating a detailed project schedule and this document
(the project plan). The project schedule and plan will help the
team coordinate our activities and make sure everyone has consistent
expectations.
DMS
1.0: Requirements Development
For the
initial phase of distributed modeling development, requirements
will be defined and documented on three levels, each of which is
described in more detail below:
Prototype
Data Management and Performance
Prototype
User Interfaces
Tools
Prototype
Data Management and Performance
Initial
indications are that an important area for prototype enhancement
is its data management processes. Due to the very large volume of
data, these processes require extremely long timelines, which makes
calibrating the system a difficult task. Initially, Apex will review
in detail the data structures, the current data management processes,
and other features that affect performance in the prototype. Subsequently,
Apex will define specific architectural modifications that will
improve the performance and reduce the times required for prototype
calibration and operation.
Prototype
User Interfaces
Currently,
the prototype does not provide any user interfaces, requiring users
to edit batch input files according to documented specifications.
Requirements definition for prototype user interfaces will focus
on basic capabilities required to enable effective testing and initial
production use. Some of the user interface requirements will correlate
with requirements defined under the "Prototype Data Management"
section described above.
Tools
To evaluate
and use the HL-RMS prototype effectively, users will require additional
visualization tools. Currently, it appears that the only primary
required tool is an adapted version of XNAV. Requirements definition
will focus on specific features of this tool.
DMS
1.0: Implementation
Implementing
the mutually agreed upon requirements, will consist of software
developers designing, reviewing the design with HSEB leadership
and building/coding the software. The HSEB leadership reviews will
keep them informed and give them an opportunity to tell us about
potential problems. In the area of data management, Apex will be
a key contributor to designing enhancements to the existing HL-RMS
code and HSEB resources will focus on coding and testing those enhancements.
Apex will also work with ABRFC to design and build the HL-RMS input
interfaces, while ABRFC will lead designing and building of the
enhanced prototype tool. The tasks in system integration and testing
and writing documentation will be shared among the three groups
with each group assigning a focal point for their respective areas
of development. After presenting OHD/HL leadership a summary of
the implementation phase, DMS 1.0 will be deployed to ABRFC in mid
September 2003 for more testing and use.
DMS
1.0: Testing
The DMS
1.0 testing phase is about a three month period starting in September
2003 where ABRFC and the second RFC about one month later (in early
October 2003) will evaluate DMS 1.0. As part of testing, both RFCs
will go through scenarios to determine the value of existing DMS
1.0 features and define others needed. We will combine the RFCs'
feedback into monthly status reviews for OHD/HL management.
DMS
1.0: Summary
The last
phase of DMS 1.0 is summarizing the project. OHD and Apex will gather
the comments and feedback from the RFCs, and then ask them to collaborate
with us to write a report to distribute to the other RFCs. The report
should be complete in early January 2004. After summarizing DMS
1.0 we will be ready to focus on DMS 2.0.
DMS
2.0
The project
schedule for DMS 2.0 has not been finalized. The details of DMS
2.0 will be clearer as DMS 1.0 concludes.
8. Issues
and Unknowns
Operational
distributed river forecasting is a new concept for OHD, therefore,
a detailed requirements analysis is necessary. At this time, we
do not know what new features we will submit to AWIPS.
The ideal
solution for implementing a distributed river forecasting model
may require some updates to the AWIPS baseline architecture. Therefore,
the full-featured implementation of an operational distributed river
forecasting may not be in place until AWIPS OB-5 and beyond.
Project
Schedule (PDF format)
(A
full version of the project plan is availble in Microsoft Word Format)
|