By Mark Mathewson, Thomas LeFebvre, and Tracy Hansen
Modernization of the National Weather Service (NWS) has placed the Advanced Weather Interactive Processing System (AWIPS) at the center of every forecaster's workspace to make their job more efficient. Now that AWIPS is operational at more than 120 forecast offices around the country, new technologies continue to be introduced that will improve and refine the weather forecast (Figure 1). Here we focus on the Interactive Forecast Preparation System (IFPS), which will revolutionize how the forecast is prepared and disseminated. Traditional time-consuming forecast preparation methods are being replaced with newly developed time-saving techniques that will help forecasters fully concentrate on the meteorological situation at hand through the use of interactive interpretation and editing techniques.
The nucleus of IFPS is a common digital database of weather elements from which routine forecast products can be automatically composed and formatted. The digital database, possible through advanced computer technology and high-speed communications, will eventually mean higher quality, more timely and more consistent forecasts among all forecast products.
History of IFPS
The concept of IFPS began with the digital forecast, simply a forecast that can be represented by numbers, which has been in use by the National Weather Service for decades. Perhaps the earliest type of digital forecast was generated using computers and statistics to represent forecasts at a point. In the 1960s advances in numerical weather prediction provided interpretation of model output statistics (MOS), such as daily temperature highs and lows, winds, clouds, and precipitation at selected sites across the United States. For a while after MOS was developed at NWS, it was viewed as additional guidance only, but over the years MOS forecasts have become accepted worldwide as the baseline by which human forecasts are judged. In the 1970s, when the NWS Techniques Development Laboratory [since renamed Meteorological Development Laboratory] began generating worded forecasts based on the use of MOS, the concept of automatic generation of products from "numbers" became a reality.
In the 1980s the NWS began a 15-year modernization program that touched practically every aspect of forecast operations. The integrated workstation – AWIPS – and other new systems were developed to help forecasters improve their understanding of the atmosphere. Larger and more capable computers installed at the National Centers for Environmental Prediction (NCEP) enabled researchers to run higher and higher resolution models. The integrated workstation brought to the forecasters' fingertips every possible data source. However, even though new higher resolution datasets were providing more detailed information for the forecast, the meteorologist had no better means of disseminating all of this detail.
The field forecaster's work routine underwent drastic changes, but until recently, one area remained unchanged: the method of producing the forecast. The IFPS now joins the long list of NWS improvements, modifying the forecast paradigm through the use of a digital forecast to generate new and existing forecast products.
In 1990 FSL began collaborating with the Meteorological Development Laboratory (MDL) in the development of a grid-based forecast preparation system called the AWIPS Forecast Preparation System (AFPS). This system, to be part of AWIPS, included the digital forecast database, grid viewers and editors, and modernized products, which were developed by FSL. Text formatting capabilities using components of a subsystem called the Interactive Computer Worded Forecast (ICWF) were developed by MDL. At that time, a true grid-based forecast preparation system had never been developed, although the Canadians had developed a spline-based system called the Forecast Production Assistant, from which many of the initial ideas for the AFPS were derived.
Within two years FSL had produced the first graphical forecast editor along with primitive associated programs designed to solicit input from forecasters to refine requirements. In a beta test, forecasters found this prototype inefficient to use because the edit tools were crude and the hardware was slow. A working group, including forecasters throughout the NWS, was organized to work closely with the developers in rapidly refining the system. This intense use and review of the system resulted in many suggestions that improved the user interface and software usability.
In 1995 FSL conducted a real-time experiment of a pre-AWIPS workstation using AFPS, and again received more suggestions for software improvements. More thought was needed as to how this grid-based system would be incorporated into NWS operations. Originally, the ICWF was to be deployed in AWIPS and then replaced with AFPS a couple years later. However, by 1997 NWS management realized that the introduction of two different interactive forecast preparation systems within as many years would be too disruptive and could cause problems in the field. Thus, a revised plan was to combine the ICWF and AFPS projects into a single system, the IFPS. The development roles of FSL and TDL remained the same, with one exception: FSL continued to improve the editors and derivations of sensible weather elements from models, but also branched out into development of modernized products (Figure 2). From this point, the IFP software developed at FSL became known as the Graphical Forecast Editor Suite (GFESuite). Before discussing FSL's work to ready the GFESuite for delivery to AWIPS, we will describe how the traditional forecast preparation process differs from the proposed IFPS digital process.
Traditional Forecast Preparation Process
The techniques used by NWS forecasters of yesteryear to generate products began with the meteorologists sitting down at their individual workstations to examine all available guidance (e.g., model data, observations, satellite, and radar data) in order to formulate a service-specific mental picture of the forecast. Duties within the forecast office were usually split across services, with one forecaster specializing in the public needs, another in marine matters, and yet another in aviation forecasts. After each meteorologist determined the forecast, a group meeting was held to ensure that their individual forecasts were consistent. Then they would return to their workstations (Figure 3) and type out all of the various forecast products for their respective user communities; for example, climatological summaries for the public, agricultural for the farmer, coastal marine for the boating/shipping industry, etc. In most cases, all forecast products were limited to text format only, without the benefit of graphical products relating to time and location.
IFPS/Digital Forecast Preparation Process
The IFPS method of generating forecasts represents a major paradigm shift for the forecaster, but offers great benefits. The process of producing the digital forecast begins the same, with the forecaster examining all available guidance, but the similarity ends here. The mental picture is no longer service-specific, but, rather, is a general view of the atmosphere in terms of sensible weather elements, which represent those forecast attributes that can be "sensed" or felt, such as temperature, wind, humidity, and precipitation type. As the mental composite of the forecast is jelling, the meteorologist enters this "picture" into the computer in digital form. For example, temperature values can be "painted," with each temperature range represented by a different color on a geographical map. Machines can now coordinate the forecasts among the various service-specific forecasters. Since the forecast is in digital form, the generation of forecast products is automated (see diagram in Figure 4), and can be of virtually any format and resolution.
The main characteristics that distinguish the IFPS preparation process from the conventional one are that 1) the official forecasts are represented in digital rather than textual form, 2) all forecast products are generated from the digital forecast, 3) graphical editors are used to define the digital forecast, and 4) numerical techniques are used to derive first-guess forecast fields directly from the models as well as for user-defined edit tools.
The most important benefit from the use of the IFPS digital process is that the forecaster can finally disseminate via the Internet, for example, the detailed forecast not available before increased frequency and resolution of the mesoscale models. More useful forecast details in addition to the text information include local effects from mountains, valleys, cities, and waterways. Other attributes, some already mentioned, include the fact that all forecast products from a site are consistent since all forecast products are generated from the same digital data source, product generation is no longer time-consuming, and new services (products) can be produced with little effort. The premise of the IFPS is high quality products provided quickly to the consumer. Thus, the GFESuite is designed to maximize the benefits of the digital forecast and better serve consumers. (A later article provides more information on the GFESuite.)
Development of the GFESuite
In the last two years, FSL developers have been concentrating on tasks that will result in an AWIPS delivery of flexible, configurable, and useable GFESuite software. After all, the software is destined for over 120 operational forecast offices, each dealing with varied terrain and forecast problems. To fill the current needs of the forecaster on a timetable suitable for ongoing AWIPS builds, we found the software components that best matched the plans for developing and disseminating the IFPS through the implementation of the Rapid Prototype Project (RPP).
The GFESuite is a complete end-to-end forecast preparation system: forecasters begin with model input, use their expertise and knowledge to edit grids, and generate the products. As we began building this system, it was apparent early on that the graphical editors in the GFESuite would not meet efficiency constraints because of their simple edit tools, such as assign value, increment/decrement values, and smooth. A higher quality class of editor was needed to accommodate a typical forecast shift in which the forecaster has to manage a dozen weather elements, hourly out to 7 days. This amounts to over 1,700 grids, each containing 5,000 grid points–at 10-km resolution! It was literally impossible to edit each and every grid; even if this magnitude of grids could be edited, the forecaster had no way of ensuring their consistency.
The only solution was to create a more "intelligent" system, so the simple "paint-like" edit tools were replaced with the meteorologically savvy "Smart Tools." As a flexible, user-configurable editing capability of the GFE, Smart Tools software allows forecasters to describe meteorological concepts, invoke the system to edit grids based on these concepts, and provide consistency checks as necessary. To illustrate, it was impossible to use the simple edit tools to modify a Quantitative Precipitation Forecast (QPF) grid and enhance it based on vertical motion. The forecaster would have to estimate the vertical motion field from wind and terrain information and then "paint" it into the grid. With Smart Tools, the forecaster writes an equation that adjusts the amount of QPF based on the surface wind and terrain information, and then a mouse click applies the concept to the grid. The QPF is enhanced in areas of upslope conditions and suppressed in areas of downslope conditions.
Consistency among weather elements is also a feature of GFESuite. In previous software versions, edits could be applied to any of the GFESuite weather elements (temperature, maximum/minimum temperature, and dewpoint), but this made them vulnerable to inconsistencies. However, with Smart Tools, a simple mouse click calculates the maximum and/or minimum temperature grids from the hourly temperature grids, and another click ensures that dewpoint can never exceed the temperature.
In mid-1999 the NWS created the RPP to address IFPS issues such as how to 1) improve the usability of GFESuite software, 2) determine and validate the forecast methodology using a grid-based IFP system, and 3) experiment and validate new modernized forecast products. Frequent interaction between field forecasters and FSL developers has been a highly successful means of achieving rapid turnaround of multiple software versions, incorporation in AWIPS build 5.0 last June, and field tests. (A later article provides more information on the Rapid Prototype Project.)
Acknowledgments – The authors wish to express special thanks to our other dedicated team members, Mike Romberg, Deborah Miller, and Dave Howard at FSL, and Matt Peroutka and Ronla Meiggs at the NWS Meteorological Development Laboratory for their continued cooperation and expertise over the years in the development of the IFPS.
(Mark Mathewson, Chief of the Enhanced Forecaster Tools Branch in the Modernization Division, leads the IFPS activities at FSL. Tom LeFebvre, meteorologist, and Tracy Hansen, software engineer, work in the same division. Mark Mathewson can be reached via e-mail at firstname.lastname@example.org, or phone 303-497-6713.)