3
National Weather Service Uses of Satellite Data
Satellite data and satellite products have been important elements of NWS operations for more than 20 years. They will become even more critical in the modernized weather service. When one examines the uses of satellite data specific to the internal forecast and warning operations of NWS, two main applications are evident. The first is the use of global data in numerical weather prediction models; the second is the use of satellite imagery for mesoscale and short-range weather warning and prediction. The former application has relied primarily on data from the POES sun-synchronous, polar-orbiting satellites, while the latter depends primarily on data from GOES satellites in geostationary orbit. It must also be noted that satellite data are widely used outside of the NWS. Users include private-sector weather service providers, other government agencies, the media, and educational institutions. Important uses of satellite data, beyond weather forecasting applications, include climate monitoring, climate research, and oceanographic applications (see Box 3-1).
Satellite Data in Numerical Weather Prediction
The primary NWS user of POES data is National Centers for Environmental Prediction (NCEP), formerly known as the National Meteorological Center. The NCEP uses the data to initialize numerical weather prediction models (see Box 3-2). Currently NCEP uses POES data, essentially on a global basis.
The data that have traditionally been used are the profiles of temperature and humidity inferred from multispectral radiances measured by the TIROS operational vertical sounder (TOVS).1 For comparable medium-range forecasts for the
|
BOX 3-1 Role of Satellite Data in Monitoring Climate and Other Applications The NRC has previously noted the importance of the NWS modernization in climate monitoring and research (NRC, 1991; NRC, 1992). Satellite data play a unique role because of their systematic and global nature. Estimates of critical variables, such as sea-surface temperatures, snow and ice cover, cloud cover, water vapor distributions, precipitation, and soil moisture are all provided globally from both geostationary and polar-orbiting satellites. Indeed, well-calibrated, space-borne radiometers may be among the most accurate means for measuring global temperature change (Spencer and Christy, 1992a and 1992b). The essential need for satellite data in climate applications is discussed in considerable detail in other NRC reports (NRC, 1988; NRC, 1990). NOAA's operational satellite data contribute significantly to the climate mission. Another important operational use of satellite observations (from both geostationary and polar-orbiting platforms) is estimating snowpack conditions for hydrologic forecasting of snowmelt and runoff. At the NWS National Operational Hydrologic Remote Sensing Center in Chanhassen, Minnesota, visible and infrared imagery from geostationary satellites is used to map the areal extent of snowpack on the ground. Highly reflective surfaces in the visible spectral range are classified either as clouds or ground snowpack, depending on the physical temperature derived from infrared channels. Under large-scale cloudy conditions, microwave measurements by polar-orbiting satellites replace geostationary satellite observations in estimating snowpack cover. Estimates of snowpack coverage and liquid water equivalent are made by merging satellite observations, ground measurements, and gammaray sensor aircraft observations. A gridded product has been developed and used at NWS River Forecast Centers to forecast snowmelt and run-off. Many spring floods and flash floods are associated with melting snow-pack caused by rain on snow or rapid changes in air dry-bulb or dew-point temperatures. |
southern hemisphere, these data have provided about a one-day advantage over forecasts made without satellite data. For example, a forecast that includes POES data will be as accurate for the fifth day as a forecast made without POES data would be for the third or fourth day.
New ways are being discovered for more effective, quantitative use of satellite data. In October 1995, NCEP implemented a new method to incorporate POES data. According to their tests, this method demonstrated the single largest
improvement in the analysis and forecast system in the past decade. The new method uses spectral radiances from POES directly in the data assimilation cycle, rather than first converting the radiances into vertical profiles of temperature and humidity. By incorporating the radiances directly, the model's temperature, water vapor, and wind fields adjust in a mutually consistent way to all available observations so that the radiances calculated from the model variables match the observed radiances more closely.
POES sounding data (see Box 3-3) are used in the NCEP's regional and global models; direct use of radiances will be operational in the regional systems by the end of 1996. NCEP also uses total precipitable (vertically integrated) water vapor, which is derived from the special sensor microwave/imager (SSM/I) on the DMSP spacecraft. Surface wind speed over the oceans, inferred from SSM/I data, and sea-surface wind speed and direction over the oceans, inferred from scatterometer data obtained by the European remote-sensing satellite ERS-1, are also used.
NWS is concerned that the quality and number of radiosonde observations worldwide are declining, particularly in developing countries. As a result, NWS will have to rely even more heavily in the future on both satellite information and aircraft data and, perhaps, on adaptive observations, for numerical weather prediction cycles. Another important application of POES data is for imagery of weather phenomena in northern latitudes beyond the range of GOES coverage, particularly for aviation forecasting in Alaska and the heavily traveled northern-Pacific commercial air routes.
Data from the PM2 satellite have historically been of primary importance for numerical models. Data from the AM satellite are used partly as a backup; however, they also contribute to the numerical weather prediction cycles. The AM satellite data will become even more important as the resolution of models increases and they require data with higher temporal and spatial resolution. Both the NCEP and the European Centre for Medium Range Weather Forecasts have already had a positive impact using data from a second POES. In any case, long-term operations with a single polar orbiter would be unacceptable because, if it were to fail, the delay in launching a replacement would lead to an unacceptable degradation of weather forecast skill and reliability in the intervening period.
Finding 1a. At least one operational POES is needed in orbit at all times to provide data vital to global numerical-prediction models. A backup POES in orbit also is required to ensure that unacceptable degradation of service does not occur when the operational POES fails. The backup satellite may also be operated simultaneously with the first satellite, thus providing global coverage four times a day. A replacement must be available for launch in case either of the orbiting spacecraft fails.
|
BOX 3-2 NCEP Models1 The NCEP routinely runs global and regional numerical forecast models in support of operational requirements of the NWS. Models that are run on a daily basis include the mesoscale eta model (named for the Greek letter used as a symbol for the vertical coordinate), the nested grid model (NGM), the rapid update cycle (RUC), the aviation (AVN) model, and the medium range forecast (MRF) model. The hurricane model is run whenever a tropical storm or hurricane develops in the Atlantic, east Pacific, or Caribbean basins (including the Gulf of Mexico). The suite of production jobs, which ingests meteorological information, analyzes it, and predicts the future state of the atmosphere for world-wide distribution, comprises a number of subsuites, or “runs.” The runs, although they differ in purpose and content, have much in common. Each run includes the establishment of initial conditions and describes the current state of the atmosphere, forecasts the future state, and disseminates the analyses and forecasts to many users. There are six such runs; all but one are repeated twice daily. The six runs are named either for their relative positions within each cycle (early or final) or for their general purpose (regional, aviation, hurricane, or medium-range forecast). The early (ERL) eta run provides a regional forecast over the United States as soon as possible after 0000 and 1200 Greenwich Mean Time (GMT) synoptic times for early guidance out to 48 hours to the NWS and the meteorological community at large. The ERL is run on a 48-km grid covering all of North America and has 38 layers in the vertical. At the initial and six-hour forecast times, it produces isobaric heights, temperature, and wind fields on standard output grids. Other information provided includes freezing levels, stability parameters, and relative humidity. The mesoscale (MSO) eta run provides forecasts over the United States at a very high resolution (29 km), from 0300 and 1500 GMT out to 33 hours, for internal distribution to the NWS and to the greater meteorological community via Internet. The MSO model produces weather forecast elements at the initial time and every three hours. |
|
The regional run produces forecasts to two days for the United States, using the regional analysis and forecast system. Sequences of analyses and short-range forecasts from the nested grid model (NGM) are produced every three hours during the 12-hour pre-forecast period. The analysis grid has a resolution of approximately 90 km over North America and 200 km elsewhere. The NGM forecasts are for 16 layers and a two-grid nested system out to 48 hours. The forecast grids have approximate resolution of 170 and 85 km at 45 degrees north latitude and a resolution of 320 km for the hemispheric grid. Model output statistics from the NGM are produced twice a day for more than 700 locations in the United States for a wide range of weather elements, such as maximum/minimum temperature, surface wind speed and direction, probability of precipitation, cloud ceiling height, visibility, and similar data. The RUC run provides high-frequency, short-term forecasts on a 60-km resolution domain covering the contiguous 48 United States and adjacent areas of Canada, Mexico, and oceans. Every three hours the RUC produces analyses and hourly forecasts out to 12 hours. The prediction system uses a hybrid vertical coordinate of 25 levels. The output from this model includes variables such as temperatures, heights of coordinate surfaces, relative humidity, and wind components. The AVN run is the first forecast in each cycle that is global in extent. Its primary purpose is to prepare guidance in support of NCEP's international aviation responsibilities. The forecast model and analysis system are identical to the MRF described below. The forecast is run to 72 hours with the production of pressure-level information at six-hour intervals. An array of weather elements similar to those produced in the NGM output are available for 225 locations in the contiguous United States, Alaska, and southern Canada from both the 0000 and 1200 GMT runs. The MRF run is generated only in the 0000 GMT cycle. Its purpose is to generate a global forecast for the medium-range scale, generally understood to mean the three-to ten-day range. Vertical resolution is 28 levels throughout. Data cutoff time is six hours after the synoptic observation times of 0000, 0600, 1200, or 1800 GMT. Objective guidance is available from this model for a wide range of weather elements for over 225 stations in the contiguous United States, Alaska, and southern Canada. Other runs, for example, the hurricane run and the ensemble run, are for special operational support or test purposes, respectively, and are not included in this summary. |
NCEP global models also use winds estimated from tracking the horizontal motion of clouds and water vapor viewed by GOES. The only GOES-7 sounder data used operationally by NCEP prior to 1996 were gradient winds. However, in 1997 NCEP plans to incorporate precipitable water data from GOES-8 into its regional models. Reasons for not previously using GOES temperature and moisture sounder data include the following:
- The GOES sounder was considerably less accurate than existing TOVS. Because there was no microwave instrument to remove the effects of clouds, sampling of GOES data was confined to clear regions of the atmosphere, which are relatively well predicted by forecast models.
- Experiments with first-generation sounder data from GOES-4 through GOES-7 in numerical weather-prediction models did not demonstrate more accurate forecasts.
Since the introduction of GOES-8 and GOES-9, NCEP has increased its emphasis on the operational use of sounder data. NCEP is evaluating the quality of GOES-8 and GOES-9 sounder products, including temperature and moisture retrievals and radiances within operational numerical models. They are collaborating with NOAA's Cooperative Institute for Meteorological Satellite Studies at the University of Wisconsin-Madison and with NESDIS to optimize the use of satellite information by using radiance data directly in NCEP's next generation analysis systems (NOAA, 1995).
Use of Geostationary Satellite Data by Weather Forecast Offices
The GOES data are a fundamental information source for NWS field offices. The first system that provided GOES images to some forecast offices was installed in the late 1970s. This analog facsimile transmission system required about 35 minutes for processing and delivery of a GOES image to the forecaster. The information was qualitative, and the forecaster could not control either the area covered or the resolution of the scenes. Digital data became available on a limited basis in 1982. The staff at the Storm Prediction Center state, without equivocation, that the availability of digital geostationary satellite data in 1982 led to an immediate “quantum improvement” in the accuracy of tornado watches. In 1985, improved analog satellite imagery was made available to 52 NWS field offices. Recently, a digital system that uses personal computers and data from the regional and mesoscale satellite data information system (RAMSDIS) was made available to 25 NWS field offices.3
|
BOX 3-3 Sounding Data To be useful for analyses and forecasts, sounding data must be gathered over large portions of the globe within fixed time constraints. Since the polar-orbiting satellites require time to traverse each orbit, and must rely on the rotation of the Earth beneath them to provide global viewing, the data are not gathered simultaneously or “synoptically,” as is desired by current operational numerical analysis models. The numerical models will, however, accept data within ±3 hours of the primary synoptic times of 0000 and 1200 GMT, and of the secondary synoptic times of 0600 and 1800 GMT. The present two-satellite POES system provides data valid (within ±3 hours) for use in numeric models initialized for both primary synoptic times. By international convention, radiosondes are released near 0000 and 1200 GMT each day. The combination of satellite data and radiosonde data is the principal source of data for the primary synoptic analysis periods. Satellite data coverage of data-sparse (no radiosonde data) regions of the world (oceans, southern hemisphere, and polar regions) is critical to the proper analysis of the atmosphere, which in turn forms the basis of the 24-hour to 5-day numerical forecasts of global and regional weather. The POES satellites are the principal source of sounding data that are valid for the “off-time” analysis cycles of 0600 and 1800 GMT. The present orbit of the afternoon satellite provides timely coverage over the eastern Pacific to gather data vital for the analysis and forecast models that are used for 12- to 48-hour forecasts of the weather over the continental United States. The proper analysis of weather systems that originate between 180 degrees west longitude and the west coast of the United States is critical to the accuracy of these models. These are the weather systems that will move eastward and affect U.S. weather in a 12- to 48-hour time period. Source: DOC (1985). |
Images of weather phenomena provided by geostationary satellites are essential for field office operations for mesoscale and short-range forecasting because they are continuously available in real time. Forecasters rely heavily on visible and infrared imagery for monitoring the tracks and evolution of severe storms, hurricanes, extra-tropical cyclones, and a host of other weather phenomena. Time-lapse imagery is particularly valuable. Data obtained over the Pacific and Atlantic oceans are critical for aviation and marine forecasts.
In the Weather Service Forecast Office in Honolulu, Hawaii, which has forecast and warning responsibility for areas as far west as Guam, Fiji, and Samoa,
satellite data are often the most reliable indicators of changing and significant weather. Indeed, forecasters in this office, as well as at the Aviation Weather Center, also use the Japanese geostationary meteorological satellite; forecasters on the eastern seaboard use the European METEOSAT. In addition to monitoring existing weather systems, the capability of geostationary satellites for detecting water-vapor fields and thin cloud lines and for deriving winds from the motion of clouds and water vapor has improved the accuracy of local weather forecasts.
Reliable and continuous service from the operational satellites remains a dominant national requirement. Since completely redundant systems and sensors are not provided aboard each satellite, sufficient GOES and POES must be launched to provide the necessary redundancy in orbit at all times.
Finding 1b. At least two operating GOES satellites are needed in orbit at all times to provide the necessary coverage from the central Pacific Ocean eastward to the coast of West Africa.4
Recommendation 1. To meet NWS high priority requirements for satellite coverage in support of weather forecasts and warnings, NOAA must ensure that the requisite data are available at all times from at least one POES and two GOES in orbit. To ensure this continuity, a backup POES and a backup GOES need to be available in orbit.
Finding 2. The GOES-8 and GOES-9 offer an opportunity to establish the operational utility of deriving soundings and upper air winds from GOES data and implement new operational techniques. Field office staff visited by the committee would like to have access to sounding and wind data from GOES.
Work at the Forecast Systems Laboratory, Aviation Weather Center, Storm Prediction Center, National Center for Atmospheric Research, and universities will also contribute to the recommended research and development effort. This effort is fundamental to improving forecasts and warnings and to providing sufficient information for NOAA's future decisions regarding the North American Atmospheric Observing System and the design of the next generation of satellites for the 21st century.
Recommendation 2. NWS/NOAA should fully support efforts to develop and demonstrate techniques for using GOES soundings and winds to improve warnings and forecasts. It is essential that the NESDIS and the NWS (particularly the NCEP) devote adequate personnel to processing, evaluating, and applying newly available satellite data as soon as possible. Science operations officers and other field staff could also contribute to this effort.
