Automated and Integrated Products
In the current NEXRAD system, automated products derived from radar base data provide important information directed toward specific needs and users. These products have provided valuable support to the forecast and warning process. Although the mechanisms discussed in Chapter 2 are providing continuing improvements in these products, operational experience has identified shortcomings. These shortcomings can be traced to the quality of the radar data and to the original NEXRAD requirements. Investigations into nowcasting and into the assimilation of radar data into numerical weather prediction (NWP) and hydrologic models have revealed additional demands on the radar data and the derived products. In this chapter, radar data issues will be presented from the viewpoint of the support needed for these automated analyses.
It is convenient to discuss the role of radar data in its support of different information horizons:
diagnosis—a quantified description of the current atmospheric conditions,
nowcast—short-term, precise forecasts of specific events (0–2 hours),
short-term forecast—the prognosis of the state-of-the-atmosphere and specific events in the next several (2–6) hours, including products for the forecaster and for data assimilation,
longer-range forecasts and support of climatological studies—the analysis of atmospheric information on longer time scales such as 6–48 hours and longer, and
off-line product development.
Although the exact boundaries between the first three time horizons may
blur, the distinctions are often meaningful from the viewpoint of current product development techniques. The ideal system would have seamless continuity, in which a fully implemented four-dimensional data assimilation (4DDA) system could render these distinctions obsolete. The integrated observing system should provide a comprehensive 4-D database from which various users can draw those portions that are pertinent to their applications.
RADAR COVERAGE AND DATA QUALITY
The following types of data limitations reduce the effectiveness of the current NEXRAD products:
degraded resolution at long range, and
data latency (update rate).
Data voids have many causes (as discussed in Chapter 2), but the end result is the denial of desired data to a product algorithm in some portion of the coverage volume.
Data corruption usually results from a combination of factors, and the impact varies between minimal and severe. Product degradation can take the form of an enlarged data void, when contaminated data are detected and masked, or of erroneous product results, when incorrect data are passed on to the product algorithms. Advanced engineering techniques can reduce, but probably not totally eliminate, data corruption. Experience has shown that the integration of data quality analysis (DQA) with a data assimilation system is an effective way of detecting and masking erroneous data in order to prevent the introduction of faulty information into the product algorithms.
Data latency results from the use of a scan rate that is slow compared with the requirements of an effective product algorithm. Different products and different atmospheric situations impose different data update requirements. To alleviate data latency issues for all algorithms, it is necessary either to have a fixed scan strategy that is suitably rapid for all circumstances or to have an adaptive scan strategy that can concentrate radar resources effectively. Operational mechanical scanning radars are usually restricted to 360-degree scans, with each rotation taking 15–20 seconds. Faster scanning results in reduced sensitivity. Single-beam radar collects data from one tilt per revolution. Simple counting provides the relationship between the rotation time, the number of tilts, and the volume scan time. Continuing the current design would provide little room for improvement over the NEXRAD. Fundamental changes in the basic system, such as use of phased array technologies, would provide data updates at rates necessary for all critical products.
The original NEXRAD design focused primarily on requirements to address diagnostic severe storm products. Shortfalls in these products can be attributed primarily to coverage and data-quality issues. In addition, new applications have been developed that the original design did not anticipate. Overshoot and beam blockage are responsible for serious limitations in the observation of low-level precipitation and boundary layer winds. Surface winds, wind shears, and convergence lines are essential for the diagnosis and nowcasting of the evolution of convection and of the transport of pollutants and hazardous materials and for the diagnosis of damaging winds and aviation hazards. Special attention should be given to providing adequate coverage in the boundary layer.
The automated nowcast applications of the current NEXRAD are mostly tracking algorithms. In addition, human forecasters make extensive use of the radar products in support of their nowcast functions. Some prototype expert nowcast systems have been developed in research programs, especially in the areas of the growth of convective storms and quantified precipitation estimation (QPE). It has not been determined whether nowcast products will remain the responsibility of forecasters or whether they can be fully automated in the near future. Regardless of the nowcast agent, forecaster, or algorithm, there is a requirement for high-quality and comprehensive radar data.
The nowcast data requirements are the same as the diagnostic product requirements:
Comprehensive and accurate boundary layer wind and reflectivity information is required for the successful nowcast of convective storm development and precipitation rate.
Accurate precipitation-type identification is required for improved QPE nowcast.
FORECASTS AND ASSIMILATION OF RADAR DATA INTO NWP MODELS
There has been major progress in NWP of synoptic scales of weather, i.e., for spatial scales on the order of a few hundred kilometers or longer and for time scales of one day or longer. With the improvement of models and of methods of data assimilation, and with operational ensemble forecasting, such forecasts are now becoming routinely skillful for five days or longer. Because initial value problems are an important part of NWP, the improvement in the initial conditions for the forecast models is an essential component in this evolution.
There have been quite a number of experimental and operational studies on the assimilation of radar observations from NEXRAD in the United States and from other Doppler radars in Europe and Japan. Alberoni et al., (2001) provide a review of these efforts with an extensive reference list. We now review briefly the experience in assimilation of wind, water and precipitation data and indicate future requirements. Several methods for the physical initialization of models using estimated precipitation rates from satellite observations have been applied in global models in the tropics (e.g., Krishnamurti et al., 1993; Treadon, 1996; Falkovich et al., 2000). These methods change the model initial conditions during a period before the start of the forecast, forcing the model to produce rain where observed and eliminating it where satellites indicate no precipitation. Generally they result in improved short-range forecasts of precipitation, but the improvements do not last long. Regional experiments assimilating radar/rain-gage estimates of precipitation have also been successful in Japan (Matsumura et al., 1997) and have been implemented in the National Centers for Environmental Prediction (NCEP) Regional Reanalysis System (DiMego et al., 2001).
Rain and cloud water content can also be assimilated with relatively crude methods, using radar-precipitation relationships to estimate rain (e.g., Xue et al., 1998, Haase et al., 1999, Zhang, 1999). Results generally indicate an improvement due to a phase correction of the location of the convective systems Grecu and Krajewski (2000) used variational assimilation of radar data into models to develop rainfall forecasts. As indicated before, radar is the only observing system with the potential of providing initial conditions for very high-resolution numerical weather prediction models. The NWS has implemented a new 12-km-resolution Eta model at continental scale, and in the next decade a continued increase in resolution can be expected. The future radar should provide accurate information on winds and precipitation fields for these models, which should lead to significant improvements in the 6- to 72-hour forecasts. Information about the error statistics (e.g., Ciach and Krajewski, 1999; Keeler and Ellis, 2000) will also be needed for effective assimilation of the radar data into these models.
By comparison, it is particularly challenging to predict short-lived phenomena such as thunderstorms. One of the most promising and active scientific frontiers in numerical weather prediction is the prediction of severe weather (tornadoes, squall-lines, intense summer convection, etc.) with mesoscale or even storm-scale models. In order to predict the evolution of these phenomena, the models must have very high resolution (grid sizes on the order of 1–10 kilometers), comprehensive boundary layer and precipitation physics, and the ability to provide initial conditions with sufficient accuracy and comparable spatial resolution, especially in the boundary layer. The use of models initiated with high-resolution radar data offers promise for improving the skill of forecasting these short-lived phenomena. The three most important variables required for initialization of storm resolving models are:
water substance (including phase), and
Radar provides high-resolution (space and time) information regarding the first two, and the third one can be indirectly derived from the winds.
A number of useful applications have already been demonstrated for use of powerful radars such as the WSR-88D in optically clear air. In such cases the echoes are due either to the windborne insects or to backscatter from turbulent fluctuations in refractive index (Wilson et al., 1994; Gossard, 1990; Serafin and Wilson, 2000). In either case one may use a single Doppler radar to obtain quite good measurements of the winds. Sun and Crook (1997a, b) have demonstrated a method for obtaining high-resolution wind and thermodynamic data from a single Doppler radar and a numerical model. Because a dense network of surface observing stations in Oklahoma have proved to be extremely valuable over a broad spectrum of human and industrial activities (Morris et al., 2001), such radar-derived wind and thermodynamic data have a wide variety of applications to the aviation, construction, agriculture, energy, and trucking communities and to hazardous chemical, biological, and nuclear accidents (Serafin and Wilson, 2000).
Among the techniques which have already shown promise for thunderstorm nowcasting are the Thunderstorm Auto-Nowcaster developed at the National Center for Atmospheric Research (NCAR) and the ITWS developed by MIT Lincoln Laboratory. These systems use radar, satellite, and conventional meteorological observations in expert systems for short-term forecasting. They have already proved to be valuable in air traffic control. When combined with storm-scale numerical models, they are also expected to be of great value in quantitative precipitation forecasts and flood warning (Serafin and Wilson, 2000). Subjective forecasting and nowcasting would also be greatly benefited by such datasets.
The value of radar data, as part of an integrated observing system, in diagnostic applications nowcasting systems, and hydrologic and numerical weather prediction models should be considered in the design of the next generation weather radar system. The characteristics of radar observations and associated error statistics must be quantified in ways that are compatible with user community needs.
LONGER-RANGE FORECASTS AND CLIMATOLOGY
Radar climatology involves classification of the historical patterns of atmospheric events as they were observed by the radar system. Issues include event frequencies, geographic distributions, distributions of event phenomenologies,
error characteristics, etc. To characterize event behavior over a protracted period of time, it is important that there be consistency in the radar information, which is archived.
The current NEXRAD system is providing important data for a more comprehensive understanding of the climatic variability over the United States. These data include inferences on rainfall, storm intensity, storm tracks, winds, boundary layer changes, urban effects, land-sea interactions, and human interactions. Radar variables in themselves, in contrast to meteorological variables derived from them, can also provide a set of consistent data from which climatic inferences can be made. Furthermore, radar data can be applicable at scales ranging from the microscale to the macroscale. When a replacement radar system is installed, it will be critical to retain continuity in observational content and, as much as possible, to retain formats to ensure preservation of the climatic record for analysis and assessment of changes. The possible relationships of weather variability associated with global climate changes such as temperature and sea state changes are of particular importance, and the availability of a consistent weather radar database within this context will be critical to making and/or validating associated assessments.
Mesoscale radar climatology has been helpful in advancing our understanding of mesoscale convective systems (e.g., Bluestein and Jain, 1985; Houze et al., 1990; Parker and Johnson, 2000), a key factor in the development of convective nowcast products. A practical application is the use of counts and characterizations of microbursts, based on radar observations at several airports, as an input for cost-benefit analyses to support Federal Aviation Administration (FAA) decisions regarding the deployment of wind-shear detection systems (McCarthy and Serafin, 1984).
To support the use of radar data in the climate observing system and other research areas, standards for calibration and continuity of observations should be established and implemented.
DATA ARCHIVING AND ANALYSIS
There are several reasons for the support of a complete and accessible radar data archive:
support for radar climatology,
support for research and testing for improved data quality analysis (DQA) and product algorithms, and
quantification of the data error characteristics, required for hydrology and NWP models.
Climatological and off-line product development involves working with data from an archive. The official NEXRAD archive is maintained at the National Climatic Data Center (NCDC). The archived data must be used in the state in which they were archived, including the selected scan strategy, resolution, and format. The future system should maintain continuity in observational content and, as much as possible, formats in order to ensure preservation of the climatic record for analysis and assessment of changes.
It is unreasonable to attempt to anticipate all uses for these data. However, it is important to keep the data in their most basic form and to let the researchers apply data quality analysis. A desire for consistency also suggests that the data archive be subjected to some core standards. On the one hand, data consistency may facilitate the accurate estimation of data errors. On the other hand, such standards may be at odds with adaptive scanning strategies. “Do no harm” should be an imperative. For example, care must be taken that a faulty control algorithm not be able to deny to the archive data that are essential for the improvement of that algorithm. These considerations are far-reaching.
Plans for next generation weather radar systems should include provisions for real-time dissemination of data to support forecast, nowcast, and warning operations and data assimilation for numerical weather prediction, and certain research applications. Routine reliable data archiving for all radars in the system for research, climatological studies, and retrospective system evaluation must be an integral part of the system. Convenient, affordable access to the data archives is essential.
SUPPORT TO USER DECISION PROCESSES
Long-term consideration should be given to the development of a multitude of products that describe the atmosphere in new and exciting ways and that have a high degree of impact to end users. End users include farmers and ranchers, those involved in transportation on the ground, on the water, or in the air, and, of course, the general public. Many such products cannot be conceived today, until an advanced system is developed.
Also, many tactical decision aids can be generated to change how the user community looks at weather science capability. For example, better weather information would provide the means for traveling from one place to another with much greater efficiency, or the means for maximizing crop yields.
Going beyond this concept is a fully integrated observing system (IOS), which will use radar as but one of many observational capabilities, so that many years hence, a four-dimensional data system (space and time) will be used to characterize all manner of weather situations. The IOS, which may in fact be a
number of such systems that address the needs of a variety of users (including many with limited meteorological knowledge), will provide the ability to extract weather features critical to users’ needs. User needs include information on severe weather, wind and moisture fields, or on weather as it may impact aviation. A current aviation-specific example illustrates such a capability. With the concept of collaborative decision making—whereby the three elements of the aviation system (the pilot, the air traffic controller, and the airline dispatcher) work with the same weather data system (likely the same data source but displayed in different user-specific ways)—a safer and more efficient utilization of the airspace is achieved.
The end state user may provide input to two types of decision aids: (1) highly directed aids for meteorological users that provide forecasters with improved assistance in operational warning and forecast needs and (2) a potpourri of nonmeteorological products and decision aids for nonmeteorological users such as those in the civil and military aviation sectors, the newly emerging intelligent surface transportation systems (e.g., smart cars and trucks), agriculture, and U.S. military. A user group should be formed in parallel to the evolution of the IOS to ensure effective development of an advanced product suite.
Tactical Decision Aids and means for collaborative decision-making capabilities should be developed for both meteorological and nonmeteorological users of the system, with attention to the demands on the integrated observing system.