John G. W. Kelley, Meredith Westington, and Eugene Wei
January 29, 2001
1. INTRODUCTION
The primary purpose of National Ocean Service's (NOS') Operational Data Acquisition and Archiving System (ODAAS) is to acquire, archive, and subset observations and output from the numerical weather, ocean, and river forecast models operated by the National Weather Service (NWS). The ocean and weather models are run at the NWS' National Centers for Environmental Prediction (NCEP) in Camp Springs, Maryland while the river models are run at NWS' regional River Forecast Centers (RFC). ODAAS supports the development and operation of NOS real-time nowcast/forecast systems for estuaries and ports. An experimental version of ODAAS is run by NOS' Coast Survey Development Laboratory (CSDL) and monitored weekdays by CSDL personnel. A portion of ODAAS is maintained on a workstation at the NOS' Center for Oceanographic Products and Services (CO-OPS) to support NOS' Chesapeake Bay Operational Forecast System (CBOFS) (Gross, 1999).
ODAAS is composed of shell scripts and FORTRAN programs which acquire analyses and forecasts from operational and experimental atmospheric and oceanographic forecast models from the NCEP supercomputers. The ODAAS scripts and programs use File Transfer Protocol (FTP) over the NOAA Campus Telecommunication Network.
This web report provides an overview of ODAAS. The discussion is divided into sections representing atmospheric and oceanographic information. Prior to this discussion, a description is given of the information requirements of NOS estuarine nowcast/forecast systems and how these requirements are being addressed by ODAAS. For simplicity, NOS estuarine nowcast/forecast systems will be referred to as estuarine prediction systems in this technical report.
2. INFORMATION REQUIREMENTS FOR NOS REAL-TIME ESTUARINE PREDICTION SYSTEMS
The numerical and statistical estuarine prediction systems being developed and implemented by NOS require real-time or near-real-time observations and analyses, and short-term forecasts of atmospheric, oceanographic, and river variables (Figures 1 and 2). This information is used for specifying surface and lateral boundary conditions for NOS estuarine prediction systems. The different types of information are briefly discussed in the following sections.
2.1 Atmospheric Information
NOS estuarine prediction systems require analyses and forecasts of surface wind stress on a grid to provide surface boundary conditions. In the future, these systems may also require analyses and forecasts of surface heat and moisture fluxes.
2.1.1 Observations
The creation of analyses of surface heat flux on the estuarine model native grid requires the interpolation of hourly or more frequent surface weather observations and cloud cover. The generation of analyses of surface wind velocity or stress requires observations of wind direction and speed. Moisture flux estimation requires estimates of precipitation and evaporation. Precipitation can be estimated from surface weather observations of rainfall amounts via rain gages or NWS Doppler weather radars.
Surface weather observations are available over NOAAPORT Satellite Broadcast Network (SBN), NWS' and NOS' FTP servers, and also from the observation files on NCEP supercomputers. Presently, only observations from NOS Physical Oceanographic Real-Time System (PORTS) stations in Chesapeake Bay and the National Data Buoy Center (NDBC) Thomas Point Lighthouse Coastal-Marine Automated Network (C-MAN) station (TPLM2) are stored in the ODAAS directories.
2.1.2 Analyses
Gridded analyses of surface wind stress and heat flux can be obtained in two ways: 1) interpolation of gridded surface flux analyses from an atmospheric model's data assimilation system to the model grid of the estuarine prediction system or 2) calculation of surface fluxes from gridded meteorological fields of data analysis or assimilation systems.
Analyses of surface fluxes and meteorological fields are available from NCEP's Eta Data Assimilation System (EDAS) and Rapid Update Cycle (RUC). EDAS analyses are available at 3-h intervals at a horizontal resolution of 22 km for North America and coastal waters. RUC analyses are available at 1-h intervals at a horizontal resolution of 40-km over contiguous United States and immediate coastal waters.
With these analyses, a NOS developer can calculate surface heat flux and wind stress via the two methods described earlier; however, the EDAS analyses do not contain precipitation analyses. The best substitution for a precipitation analysis would be to acquire the EDAS 3-h or RUC 1-, 2-, or 3-h forecasts of precipitation or estimate of precipitation amounts based on NWS weather radar reflectivities.
Presently, the ODAAS acquires hourly analyses of surface wind velocity and mean sea-level pressure from RUC and also NOS experimental Chesapeake Bay Local Analysis and Prediction System.
2.1.3 Forecasts
Short-term forecasts of surface wind stress and heat flux can be obtained in same two ways: 1) interpolation of gridded surface flux forecasts from the atmospheric models to the model grid of the estuarine forecast system or 2) calculation of surface fluxes from gridded meteorological forecast fields.
Gridded analyses of surface fluxes and meteorological fields are available
from the NCEP global spectral atmospheric short-term prediction model (commonly
referred to as the ‘Aviation Model') and Eta Mesoscale Atmospheric Prediction
Model. The Aviation Model has an equivalent horizontal grid resolution
of 105 km. There are four cycles of the Aviation model per day, occurring
at 0000, 0600, 1200, and 1800 UTC. The 0000 and 1200 UTC cycles produce
forecasts out to 126 hours, while the 0600 and 1800 UTC cycles generate
forecasts to 84 hours.
Eta-22 model forecast fields are available at 3-h intervals at a horizontal
resolution of 22 km for North America and coastal waters. This version
of the Eta model is commonly referred to as the Eta-22 model. There
are four cycles of the Eta-22 model per day, at 0000, 0600, 1200, and 1800
UTC. The 0000 and 1200 UTC cycles produce forecasts out to 60 hours
while the 0600 and 1800 UTC cycles generate forecasts out to 48 hours.
Other NCEP atmospheric forecast models which could provide short-range
surface fluxes and/or meteorological fields are the Nested Grid Model (NGM)
and RUC. The NGM provides 48-h forecasts for North America and adjacent
waters. The RUC generates 3- to 12-hour forecasts for the contiguous United
States and immediate coastal waters at a 40-km horizontal resolution.
With these forecasts, a NOS developer can calculate surface heat flux
and wind stress via either of the two methods described earlier.
In addition, Eta-22 model, the NGM, and Aviation models generate
precipitation forecasts which could be used in the calculation of surface
moisture flux.
Presently, ODAAS acquires output from the 0000 and 1200 UTC cycle of
the Eta-22 and Aviation models.
.
2.2 OCEANOGRAPHIC INFORMATION
2.2.1 Observations
The generation of nowcasts require water level observations from NOS' water level gage stations and Physical Oceanographic Real-Time System (PORTS) stations near the open boundaries of the estuarine prediction systems. Presently, NOS observations are obtained from CO-OPS by each developer separately. The PORTS observations are formatted in the NOS PORTS Uniform Flat File Format (PUFFF) (Evans et al., 1998).
In the future, these estuarine prediction systems will require real-time observations of water temperature, currents, and salinity for forcing at their open boundaries, and for possible assimilation into the models. PORTS could provide this information in real-time.
2.2.2 Analyses
In the future, NOS estuarine prediction systems may extend beyond the mouth of an estuary and into the coastal ocean. This would require information along an extensive boundary which could not be provided by distant observation sites on the coast. The open boundary conditions could be provided by climatology or by real-time analyses or nowcasts from coastal ocean prediction systems, such as NOAA's experimental Coastal Ocean Forecast System (COFS). Although available, ODAAS does not presently acquire three-dimensional nowcasts from COFS.
2.2.3 Forecasts
In order to generate forecasts, the NOS estuarine prediction systems
presently require hourly forecasts of either subtidal or total water level
on or near the open boundary. ODAAS obtains subtidal water level point
forecasts from the NWS Meteorological Development Laboratory‘s (previously
called the Techniques Development Laboratory) (MDL's ) Extra-tropical Storm
Surge (ETSS) Model. The ETSS model generates 48-h forecasts of subtidal
water level at various locations along the East Coast and Gulf of Mexico
twice a day at 0000 and 1200 UTC. Hourly forecasts of total
water levels are also available from COFS. COFS has one forecast
cycle per day at 0000 UTC which generates 24-h forecasts of total water
level at locations along the East Coast.
In the future, NOS estuarine systems may forecast for temperature,
salinity, currents and its grid domain may extend beyond the mouth of an
estuary and into the coastal ocean. The model's open boundary
condition would require short-term forecasts along an extensive boundary
which could not be provided by distant forecast points on the coast.
The alternative would be to use forecasts from coastal ocean prediction
systems, such as COFS. Although gridded forecasts of temperature,
salinity, and currents are available, ODAAS does not presently acquire
three-dimensional forecast fields from COFS.
2.3 River Outflow
NOS estuarine prediction systems also require information on river and sewage outflow in the future. At NOS/CSDL, river information has been based on 1) monthly mean based on historical data, 2) near-real-time observations, and/or 3) NWS river forecast guidance. The different types of river information are discussed below. A detailed description about the river information and the requirements of NOS estuarine forecast systems can be found in Thomson (2000).
2.3.1 Climatological Data
Information on mean monthly or annual outflow is available from most major rivers in the United States. CBOFS uses monthly mean river flow for the Susquehanna, Chester, Choptank, Nanticoke, Patuxent, Potomac, Rappahanock, York, and James Rivers.
2.3.2 Observations
It will be important for NOS estuarine prediction systems to use real-time
observations of river flow rates or estimates based on river stage, especially
in situations of high flows following heavy rains during tropical or extra-tropical
cyclone, mesoscale convective system or in cases of low flows during a
drought. Near-real-time river observations are available from the
U.S. Geological Survey, from NWS' Regional River Forecast Centers
(RFCs), or from the NWS Office of System Operations (OSO).
RFCs receive data from the gages in a variety of ways.
Most gages have Geostationary Operational Environmental Satellite (GOES)
Data Collection Platform (DCP) telemetry and transmit their data
every hour typically (R. Shedd, 1999, personal communication). These
data are processed into a Standard Hydrometeorological Exchange Format
(SHEF) message by the Hydrometeorological Automated Data Systems (HADS)
processor at the Office of Hydrology in Silver Spring, Maryland and transmitted
via the NOAAPORT SBN under the WMO header SRUS20 KWOH. Alternatively,
a smaller portion of the gages have telephone interrogable units.
Some these data are also processed by Centralized Automatic Data Acquisition
System (CADAS) and HADS and distributed in the same product (Shedd, personal
communication, June, 1999).
Presently, ODAAS receives river discharge estimates based on river stage observations from the NWS Middle-Atlantic River Forecast Center in State College, PA.
2.3.3 Forecasts
NOS estuarine prediction systems may also benefit from NWS short-term forecasts of river outflow, especially in situations of rapidly changing conditions in the hours or days following heavy rains or during a drought. Short-term river predictions can be obtained from forecast guidance generated by river forecast models operated at the RFCs. The forecasts are usually made once a day before Noon local time and at 6-hour intervals out to 54 hours. The forecast guidance includes the observed river stage at 1200 UTC. However, in flood or near-flood conditions, the guidance may be updated up to four times per day.
Presently, ODAAS at CO-OPS receives once per day river discharge forecasts
from the NWS Middle-Atlantic River Forecast Center in State College, PA.
The forecasts are valid for for head water locations of the Chesapeake
and Delaware Bay and the Port of New York/New Jersey.
3.0 OVERVIEW OF ODAAS
Descriptions of the sources, software, hardware, and scheduling procedures
are given in this section. Directories and file names are shown in
bold typeset.
3.1 Information Sources
The primary source of data archived by ODAAS is the NCEP IBM System Parallel (commonly referred to as the IBM SP) supercomputer located at the Commerce Department's Census Bowie Computer Center in Bowie, Maryland. Output from NCEP operational models is located on the IBM SP in operational subdirectories of the directory: /com. The subdirectories are named according to the modeling system. For example, the subdirectories containing output from the Eta atmospheric forecast models are named eta. It is anticipated that ODAAS will acquire NCEP model output from NCEP data servers in mid-2001 due to changes in NCEP policies regarding access to NCEP computers by other NOAA agencies.
3.2 Software
ODAAS is developed from a variety of software sources which are used to acquire, archive, translate, and subset NCEP model output. Both C and Bourne UNIX shell scripts and FORTRAN 77 programs are used to acquire and subset the data. ODAAS relies on the NCEP Gridded Binary (GRIB) decoder, wgrib, for extracting specific GRIB records within a NCEP model output GRIB file and for degribbing the GRIB records into ASCII or traditional binary format (e.g. IEEE binary). Additional information on wgrib and GRIB can be found the http://www.ncep.noaa.gov and ftp://wesley.wwb.noaa.gov/pub/wgrib. Presently, ODAAS obtains model output from NCEP via FTP "pulls". However, it is anticipated that by mid- 2001, NCEP's Distributed Brokered Networking (DBNet) software will be incorporated into ODAAS to obtain model output from the NCEP data servers.
3.3 Documentation
All ODAAS shell scripts and FORTRAN programs include documentation blocks located at the beginning of the main sections of each program and subroutines. Originally, these documentation blocks were based on those used by NCEP (Peters, 1998) which included the following information: Program Name, Purpose, Contact Person(s), Attributes (language, computer system), Programs Called, Input Files, Output Files, and History; however, the blocks were changed to meet NOS/CO-OPS standards. The NOS documentation blocks include the following information: Program Name, Program Location, Technical Contact, Author, Date, Revisions (Date, Author, Reason), Abstract, Language, Target Computer, Equation Used, Program Execution, Input Parameters, Stored Procedures, Tables, Subroutines or Functions Called (Name, Location, Description), Input Files (Unit No., Name, Location, Description), Output Files (Unit No., Name, Location, Description), Libraries Used, Error Condtions, Remarks. In addition, an effort has been made to include comments within the code to identity major tasks performed by the program.
3.4 Description of Directories
The model directories include the following sub-directories: archives
(data), exec (executables called by the scripts), execlog (log files created
when script is executed/completed), idl (IDL programs/images associated
with the model), info (information about the model or grid) scripts (operational
and manual scripts), and sorc (Fortran programs called by the scripts).
There is also a modlog file in each model directory. This file is
updated by the ODAAS staff when any changes are made to the script, log
files, etc.
3.5 Hardware
Presently, ODAAS runs at CSDL and the smaller version of ODAAS is run on the CO-OPS SGI workstation used to support CBOFS . It is anticipated that ODAAS will also be ported to a backup CBOFS server at CO-OPS.
3.6 Schedule
The scripts responsible for acquiring and processing NCEP model output
from NCEP are scheduled using the UNIX utility ‘crontab'.
The scripts are designed to sleep for 15 minutes if the model output is
not present on the specified directories at the NCEP IBM SP computer. The
scripts are located inside a cron file called cronodaas to be executed
automatically. The automated status messages are also generated by
this cron file and are located in the directory /ODAAS. For a script
to be executed by the cron file, the entry must contain the time of execution,
full directory path, and be directed either to an output file or to /dev/null
(trash).
ACKNOWLEDGMENTS
ODAAS was originally created in CSDL by Charles Sun in 1996. The
responsibility of maintaining ODAAS and its development was transferred
to Eugene Wei in 1997. In January, 1998, the responsibility was given
to John Kelley. The monitoring, maintenance, and development of ODAAS was
carried out by John Kelley, Sarah Maxwell, Eugene Wei, and intern Allison
Thomson between 1998 and March 2000. Presently, CSDL's Meredith Westington,
John Kelley and Eugene Wei are responsible for ODAAS at CSDL and CO-OPS.
Thanks to CO-OPS' Kate Bosley and Tom Bethem, and CSDL's Frank Aikman and
Kurt Hess for reviewing this technical report.