Table of ContentsTable of Contents 
Previous Chapter Similar to today, the forecaster will start the shift with a briefing by the departing forecaster using a variety of techniques to display time sections, cross sections, and plan views. These techniques will include computation and display of weather elements as static charts, tables, and animated graphics.
The departing forecaster will have spent much of the previous shift maintaining a forecast and warning database. This will include reviewing incoming observations and central guidance, graphically preparing forecasts and warnings, reviewing machine-generated graphics, text, and broadcast-ready text, and then releasing the products to the end user. This database will support all NWS forecast programs including public, aviation, marine, hydrology, agriculture, and fire weather. Because a common database will support all forecast and long-term warning programs, inconsistencies from one forecast product to another will be minimized.
Forecast responsibilities will no longer be divided along product lines. The forecasters will work as a team to maintain the forecast database from the current time through the extended outlook. They will concentrate their efforts on the initial 24 to 48 hours, with special emphasis on 0- to 6-hour mesoscale events. As in the recently restructured public forecasts, text forecasts will include both short- and long-range predictions. It may be difficult for forecasters to improve on either numerical or NMC guidance at longer projections. Typically, longer-range (beyond 48 hours) guidance will be reviewed; occasionally, as time permits, the forecaster will modify this guidance.
During severe weather, it will be difficult for the forecaster to maintain (update) the routine forecast database to the same level as in fair periods. This may necessitate greater reliance on guidance and centrally-produced forecast grids for routine products so that the focus can remain on watches and warnings. NMC intends to prepare a full suite of gridded forecasts, which will be sufficient to generate all forecast products for projections of 6 hours and beyond. The forecast system is designed to run on "autopilot" if necessary; i.e., it will be possible for text forecasts to be generated from existing or updated information in the database that has not been edited by the forecaster. Such forecasts still will be reviewed and released by a forecaster. Certainly, the quality of these forecasts may not be as high as those issued under normal circumstances, especially in the first periods.
Most forecasts will be issued at scheduled times with updates as required by the meteorological situation. Guidance and observations will arrive continuously, and local applications will produce additional short-range statistical and dynamic forecasts. Automatic monitoring programs will compare incoming information with the current forecast database and advise the forecaster of significant differences.
The forecaster will have the option of using new guidance to update the current forecast in the database or staying with the current forecast. He/she will select forecast projections and weather elements from the various sources of incoming guidance. Whenever the database is updated, the forecaster, with the assistance of automated tools, will ensure that the forecast is internally consistent.
Other issues of this nature may arise in the future, when graphical and gridded forecasts are in use, and particularly were the NWS to adopt a "dial-in" facility for users to retrieve forecasts directly from a WFO database.
Figure 3 is a processing flow diagram illustrating how forecasts will be prepared.(2) Observations will arrive continuously at the WFO, as will central guidance from direct model output, from statistical models, and from the human/machine mix at NMC; all will be ingested by AWIPS. Some guidance parameters will be converted to weather elements before the forecaster views them.
The forecaster will be able to retrieve and review this guidance and graphically adapt it to the forecast of the day. For example, the guidance temperatures for 2200 UTC may be depicted as analyzed contours over the local forecast area. Precipitation areas may be depicted as enclosed areas with labels. The forecaster will manipulate contours and enclosed areas to construct forecasts. The greatest attention will be given to the forecast for the first 24 hours, with progressively less out to six days. If multiple sources of guidance are used, the forecaster may have to smooth the transitions from one model to another at their time boundaries.
When the forecast database does not provide adequate spatial resolution to represent significant features (local effects), the forecaster will have access to a set of reference areas. Local phenomena may be specified (e.g., canyon winds or lake-effect snows); these will be inserted in forecasts by the text generators, as appropriate.
Once the general forecast depictions are prepared, service-specific weather elements will be computed. For example, drying conditions will be computed for the agricultural forecast, and chance of wetting rain and lightning activity level will be computed for fire weather forecasts. These fields will be available for forecaster review and editing. Changes made to these elements do not reflect back to the general weather elements, so usually the forecaster will finish working on the general weather elements before starting work on service-specific elements.
Some NWS products are issued as combinations of forecast areas (zones, counties). Once the forecast has been created over the area of responsibility, appropriate combinations will automatically be chosen. Also, numerous forecast-specific product parameters (e.g., wind chill, heavy surf) will be computed.
Aviation terminal forecasts are issued for specific points. Aviation-specific weather elements will be derived from the general elements and will be modified independently of the general forecast.
When all forecast elements and computed parameters are complete, individual products will be generated.
This entire forecast process will be ongoing because new guidance will arrive periodically, forecast product release times will occur throughout the day, and forecasts will need to be updated because of changing weather.
Forecasters are responsible for approving each product for release. They may decide to enhance the wording of the text products to better describe the meteorological situation or to elaborate on how the forecast will vary in the local area of responsibility. This may also be the case for broadcast-ready products for NWR. However, if the forecaster manually edits an NWR product, the edited broadcast-ready text will also require checking to ensure that the words will be recognized by the NWR text-to-voice software.
The option to edit text products is advantageous due to the flexibility it gives the forecaster. The forecaster may edit the product to add details that the product generators are not programmed to include, to improve clarity, to incorporate local idioms, etc., but the text generators will be designed to produce an acceptable text product without human intervention. By choosing to edit the products, the forecaster spends extra time and becomes less efficient. He/she also sacrifices the consistency built into an automated method of producing text products: not only are the text products potentially inconsistent with one another, but also any product whose technical content has been changed will be inconsistent with the database. If so, the ability to use the database for monitoring and verification purposes is compromised.
Editing of graphic or gridded final products should not be necessary because the forecaster will be working with these products when the forecasts are created.
Later, output arrives from a model that often is better initialized because it has a more accurate and complete database of observations on which to draw. After reviewing these data, the forecaster decides that this model better handles the upper-level wave that is forecast for tomorrow morning. The forecaster generates a new set of general weather elements (stored in a new reference worksheet), then selects several hours of the wind forecast and includes them as part of the forecast worksheet.
With editing complete, the forecaster invokes the space/time interpolation and consistency checker. Informed by the latter that the dew point forecast is slightly inconsistent with the temperature,(4) the forecaster adjusts the dew point forecast, then stores the worksheet in the database, whereby it becomes the official forecast.
The text generators are then invoked using this forecast, and one or more text products are created and placed in a text editor. The forecaster reviews the text and, if required, makes a few changes to improve the wording, then stores and disseminates the product.
Later in the morning, extended forecast grids arrive from NMC. The forecast monitor program alerts the forecaster that the new guidance differs significantly from the current forecast. The forecaster reviews the guidance using standard AWIPS techniques and determines that the only significant difference is in the fourth and fifth days. The forecaster initializes another reference worksheet with this new guidance and then copies the fourth and fifth days, thereby replacing the old forecast.
By early afternoon, it is apparent that the atmosphere is less stable than forecast. Thunderstorms are becoming widespread instead of isolated. They are also appearing several hours ahead of schedule. Deciding to update the forecasts, the forecaster redraws some of the weather depictions to denote the widespread coverage of thunderstorms and copies (moves) later depictions to earlier in the day to adjust for the earlier thunderstorm appearance.
The forecaster then stores the revised official forecast and invokes the space/time interpolation and consistency checker. The system indicates that everything is consistent and the desired products can be generated. Selected products are then created, stored, reviewed, and disseminated.
W/W/As will be issued in text, voice, and graphical forms, and will be reviewed and, if necessary, edited by the forecaster before release. Products destined for NWR will be quality-controlled to ensure proper text-to-voice conversion.
The issuance of W/W/As will affect routine forecasts in at least two ways. Many of today's forecasts (e.g., zone, coastal marine) carry headlines or banners alerting users to severe weather. Further, if a severe thunderstorm watch or warning is issued, all appropriate products should reflect the severe weather. For these reasons, when the forecaster selects severe weather phenomena, AWIPS techniques must update the database to allow the product generators to include the new W/W/A banners in the products.
The second class of quality control routines will check the final text products before they are transmitted. These routines will be devoted to finding errors or problems in portions of the products that the forecaster manually edits. These checks can include spelling checks or, for NWR products, checks to ensure the forecasts can be converted from text to voice. These latter routines are not part of this project.
Guidance-versus-forecasts monitoring compares new guidance to the current forecast. This includes guidance from central sources as well as locally produced guidance. These techniques will compare appropriate product parameters and projections and alert the forecaster if the new guidance indicates a substantial change from the current forecast.
As the forecaster maintains the forecast database and makes changes, a monitor will compare the changed values with the current forecast and alert the forecaster if it is necessary to issue a revised forecast product. These techniques will make it easier to maintain current forecasts in each program area.
NMC may create the position of Chief Forecaster, part of whose job would be to perform some coordination activities. Clearly, this will be a significant and difficult task, for machine or human.
Although some aspects of verification may be tested during the development of AFPS, specification of verification techniques and procedures is not part of this project.
Table of Contents 
Next Chapter