The NWS also plans to upgrade its method of producing forecasts. Instead of the current method of typing repetitious forecasts, a new, more efficient method of interactive forecast preparation (IFP) will be introduced. There are two planned IFP systems. Initially, The Interactive Computer Worded Forecast (ICWF) will be deployed (Ruth and Peroutka, 1993). It is a forecast zone-based matrix of numbers that represent the periods and parameters of the forecast. The forecaster enters the forecast into the matrix and text products are automatically generated.
The AWIPS Forecast Preparation System (AFPS) will be the second IFP system (NOAA, 1993). AFPS is grid-based and contains a series of two-dimensional grids on which the forecaster "draws" the weather. Each grid represents a parameter for a given period of time. Once the forecaster has finished drawing the gridded forecast, textual, graphical, and gridded products are automatically generated.
AFPS, under development since 1990, is a joint project between the Forecast Systems Laboratory (FSL) and the NWS Techniques Development Laboratory (TDL) in Silver Spring, MD. AFPS is expected to be field-tested for several years beginning in 1996.
AFPS provides tools to allow the forecaster to view and edit these grids intuitively. All of the displays and tools are highly interactive thus allowing the user to quickly draw the forecast.
Algorithms are provided to initialize all or portions of the database from various models, such as the Rapid Update Cycle (RUC), Model Output Statistics (MOS), or from previous forecasts, observations, and climatology (Wier and Wakefield, 1996).
Product generators access the forecast grids from the database and produce a variety of output products. These products exist in textual, graphic, and/or gridded formats.
Software interfaces allow individual sites to integrate the output of local models into the system. Any model that is grid-based can be integrated into AFPS by the WFO staff. The local staff may also write custom product generators to format site-specific products.
There are many benefits to the IFP method. Since all products are derived from one database, all products will be consistent with each other. Forecast monitoring and verification are easier when forecasts and observations are represented by gridded fields instead of free-form text. Forecast grids may be shared across WFOs, thus providing data for both reconciliation of forecast differences at the WFO boundaries and WFO backup operations.
Figure 1
Main Control provides an inventory of databases that are available to the forecaster. One database of significant importance is the "official forecast" database that contains the grids from which all forecast products are generated. The other databases represent the forecasts derived from numerical models. To examine the data within a database, the forecaster can simply click on the menu which creates a worksheet for that database.
A worksheet, which represents the view for the user of a database, shows the available forecast parameters along with the currently defined data grids. The worksheet is used for controlling the data seen in the editors, initializing other worksheets, and modifying the valid time of the data grids. From the worksheet, the forecaster can create editors.
The editors provide viewing and editing capability for the data grids. Forecast data may be viewed in different formats and edited using interactive tools.
There may be multiple worksheets and editors open on the display at any time.
Figure 2 
The user clicks on the desired button and AFPS brings up a worksheet containing the information.
Figure 3 
The worksheet presents other important information to the user including the editable state, editors attached to the worksheet, and the data loaded into the editors. The worksheet also serves three other purposes: time editing, editor control, and data selection.
The time editing function allows the user to modify the valid times of each grid. Each grid in the database is applicable for a specified time period. For example, a sky condition grid with a time period of January 1 at 1200UTC to January 2 at 0000UTC represents the sky condition expected over the 12-hour period. The editing function allows the user to temporally move, expand, contract, and split the grids.
The worksheet opens various types of editors from which the data grids may be viewed and modified. The selection of parameters and the data grids for each editor is controlled from the worksheet. The worksheet is used to step an editor forward and backward in time.
The data selection feature may be the most powerful capability of the worksheet. Time ranges and parameters may be selected using the mouse. Once a selection is made, the user may choose to copy data between worksheets. This is the primary means of initializing the forecast worksheet from model worksheets. Other selection-based programs may be initiated such as derivation of additional parameters, interpolation (Wier, 1995), and consistency checking.
Most of the worksheets in AFPS cannot be modified and are intended for viewing only. Their primary function is to provide the forecaster with guidance. The official forecast worksheet is the only worksheet that can be modified.
The worksheets and editors are coupled so that editing changes made on the worksheet will immediately be seen on the editors.
Each of the editors is capable of displaying the data in multiple formats. User controls permit changes to color enhancement tables, graphic colors, and display presentation attributes. Fields may be toggled on and off as well as overlaid to facilitate comparisons. Zooming and panning allow more detail to be seen.
Legends are provided to aid in the interpretation of the color displays and are also used to set the value of the editing tools (Figure 4).
Figure 4 
Figure 5 
There are many editing tools available to the forecaster (LeFebvre, 1995). The tools are categorized into gridpoint and freehand. The gridpoint tools allow the forecaster to define areas of selected grid points on the display and then apply edit operations to those points. The gridpoint tools include the "Set Value," "Push-Pull," "Move/Copy," "Define Grid," and "Fill In Hole" tools. Set Value is an absolute tool that sets all selected grid points to the same value. Push-Pull is an incremental tool that increases or decreases the value of all selected grid points by the same amount. The Move/Copy tool permits grid values to be moved or copied to other areas on the data grid. Define Grid provides the user with the power of identifying just a few grid points and their values, and the remainder of the grid will be derived. Fill In Hole is the opposite of Define Grid. It uses interpolation to determine the values of selected portions of the grid from areas outside the selection.
There are three ways to select the set of grid points to edit. The "Area" and "Point" tools allow the user to interactively draw and select grid points, predetermined named sets of grid points may be accessed via a pull-down menu, and the user can select grid points based on data values (e.g., all grid points with temperatures less than 35 degrees).
The free-hand tools, consisting of the "Paintbrush," "Spraycan," and "Bulldozer," do not require the forecaster to select grid points ahead of time. As the tools are passed over portions of the grid, the data values are immediately modified. Grid points "painted" by the Paintbrush are set to the value of the paintbrush. The Spraycan increments or decrements the "sprayed" grid points by a user-selected amount. The Bulldozer smooths gradients that it passes over.
Since all of the editors for a worksheet share the same database, changes made on the spatial editor will immediately be shown on all other spatial and temporal editors connected to the same worksheet.
Figure 6 
An edit tool allows the user to adjust the color rectangle or time line until the desired value is shown. The changes made on the temporal editor are propagated back to the spatial grid.
The user may select one of three influence methods in applying the changes to the spatial grid. The single-point method changes only the single grid point for which the temporal editor represents. The absolute selected point method applies the new value to all selected points on the grid. The relative selected point method applies the amount of change made on the temporal editor to all selected points on the grid. Figure 7 shows the locator map interface which depicts the spatial area of influence for the temporal editor.
Figure 7 
The user has the capability of modifying large amounts of data with little effort when using the selected points influence.
Forecasting using grids potentially eliminates many of the current drawbacks. Various product formats may quickly and automatically be produced at little expense to the forecaster. Products will always be consistent since they are generated from the same data grids. The forecaster should be able to concentrate more on the forecast and less on composing texts.