Table of Contents 
Previous Chapter 5 Deriving Weather Elements
Whether a forecaster chooses to initialize forecast fields from the current forecast or based on a numerical model, it will be necessary to generate grids of weather elements. These grids will at least be used for reference.
A difficult problem that must be addressed when initializing weather elements is related to the representativeness of the basis data. Neither grids nor points can be expected to adequately represent small-scale topographic and geographic features, which may produce significant localized changes in weather. We will provide means for forecasters to include such local effects in the forecasts (see Section 6.2.6), but a much better solution would be to incorporate such techniques in the initialization.
As a rule, numerical models do not directly produce weather elements. The obvious solution is to generate the elements directly from a model's output variables. Such algorithms are being developed. Examples include work at FSL using the Local Analysis and Prediction System (LAPS), Mesoscale Analysis and Prediction System (MAPS, soon to become operational as NMC's RUC), and NMC's eta model. Much of the present work is directed at aviation-specific weather elements, in support of the FSL/FAA Aviation Gridded Forecast System. Similar algorithms could be developed for other NMC models and weather elements.
Currently, most weather elements must be derived using statistical or interpretive approaches. Regardless of the success of developing direct-from-model fields, it is likely that statistical methods will always produce better first-guess forecasts at stations, since they are tied to station climatological records.
NMC model output will be interpreted in terms of probability and best-category forecasts by statistical techniques. These statistical forecasts will then be used to derive forecasts for weather elements which constitute the hydrometeorological fields required for the official forecast. Some elements, such as temperature, dew point, and wind speed and direction, can be obtained directly from statistical guidance.
The ICWF's transformation of MOS point data into grids of "explicit weather" elements is a multistep process:
1. MOS station guidance, as updated by the most recent LAMP output, is mapped to a grid. For continuous MOS elements (those for which an average is meaningful), the forecasts for one or more stations contribute to the value assigned to a grid point. For categorical elements, each grid point is assigned data from a single station.
2. If any grid points remain undefined due to missing MOS data, they take on the value of the nearest data-bearing grid point.
3. Selected continuous elements are spatially smoothed.
4. Weather elements are derived from the guidance at each grid point.
5. Areas of precipitation and cloud layers are identified and mapped.
The early versions of AFPS will rely primarily on statistical guidance and thus will use a similar scheme. The increased number of stations available in the future will improve this scheme, but any statistical approach is likely to have difficulty in complex terrain.
Even though the inherent accuracy of Perfect Prog is less than MOS, the results achieved from calibrated Perfect Prog forecasts are expected to be sufficiently accurate for grid initialization.
Since present NWS statistical techniques do not directly produce forecasts of certain weather elements, algorithms will be used to translate probability and categorical forecasts for various weather elements into forecasts of other weather elements.
As an example, consider the determination of precipitation type and intensity. The ICWF employs a three-step process, using currently available MOS parameters. First, the phase of precipitation (liquid, freezing, or frozen) is determined by analyzing the probabilities of each phase, the best category forecast of precipitation type, and 3-hour temperature forecasts. The nature of the precipitation (general, showery, or drizzle) is then determined by analyzing probabilities of rain, rain showers, and drizzle. Finally, precipitation intensity is determined from an analysis of quantitative precipitation forecasts. In the 1- to 20-hour time frame, LAMP forecasts can be used in basically the same manner.
Table of Contents 
Next Chapter