Table of Contents Previous Chapter 11 Weather Element Initialization
This chapter describes the initialization methods used for each of the weather elements. Weather elements are initialized from various sources including statistical guidance, algorithms, observational data, model output grids, and other weather elements.
Several techniques are used for MOS initialization.
MOS forecasts are generally made for stations, not grids. Thus, algorithms must be defined to populate a grid of weather elements using a collection of MOS station forecasts. The first step in this process is for the office to specify a set of station-to-grid maps. These maps specify which station(s) should be used at each grid point. Up to four stations may be specified for each point for those elements for which an average is meaningful. Figure33 on page116 shows two examples of using data from three stations to initialize a grid. For AFPS, it is envisioned that the station-to-grid mapping scheme may also incorporate other correction factors, such as multiplicative or additive for corrections for topography. Some of these correction factors may change depending upon the weather pattern or season.
Figure 33 - Station-to-Grid Mappings
Station-to-grid maps are shared among the MOS elements as follows:
- max/min temperature, temperature, dew point
- wind speed, u-component, v-component
- 6- and 12-hour probability of precipitation
- thunderstorm probability, severe weather probability
- best category precipitation type, conditional probability of freezing precipitation, conditional probability of frozen precipitation, conditional probabilities of liquid precipitation type (rain/shower/drizzle)
- 6- and 12-hour quantitative precipitation forecast, snow amount
- visibility, cloud
- obstructions to vision
For most continuous weather elements, the MOS forecasts are applied to the grid using the appropriate station-to-grid map. In most cases, they are then smoothed using one pass from a Shuman filter. The filtering step "softens the edges" and produces a more realistic initialization.
???
Transforming probabilistic MOS forecasts for stations into bounded areas of weather is a multi-step process:
1. PoP06, PoP12, TRW, SVR, conditional probabilities, QPF06, QPF12, visibility, and conditional visibility are interpolated to hourly resolution and applied to grids using the appropriate station-to-grid maps.
2. The probabilities are then converted into "explicit weather" forecasts for each point in space and time. The decision trees which guide this process are illustrated in Figure34. Coverage/probability elements are coordinated with the PoP12 weather element.
Figure 34 - Explicit Weather Decision Trees
3. A visibility is chosen for each grid point: conditional visibility if the weather coverage/probability is chance or slight chance, and prevailing otherwise.
4. For each hour, adjacent grid points with identical weather elements are coalesced into bounded areas with one-hour time spans. At this point the weather elements type, intensity, coverage/probability, and visibility are initialized.
5. Bounded areas which have identical spatial extent are combined.
6. Bounded areas are compared with those at adjacent hours. If an identical bounded area exists at an adjacent hour, the time range for one area is expanded and the other area is deleted.
Transforming LAMP and MOS forecasts for stations into bounded areas of cloud is a multi-step process:
1. LAMP forecasts of cloud height and amount (both prevailing and conditional) are applied to grids using the appropriate station-to-grid maps.
2. At each point in space and time, sky condition is initialized with the highest amount of the prevailing layers present.
3. For each hour, adjacent grid points with identical cloud amounts and coverages are coalesced into bounded areas with one-hour time spans. Conditional cloud layers at any point in space and time are scanned only if the weather coverage/probability at this point in space and time is chance or slight chance. At this point, the cloud elements base, amount, and probability are initialized.
4. For each hour, adjacent grid points with identical sky cover are coalesced into bounded areas with one-hour time spans.
5. Bounded areas which have identical spatial extent are combined (for both clouds and sky condition).
6. Bounded areas (for clouds and sky conditions) are compared with those at adjacent hours. If an identical bounded area exists at an adjacent hour, the time range for one area is expanded and the other area is deleted.
Speed - MOS surface wind is applied to the grid using the station-grid mapping technique and a one- pass Shuman filter with an office-specified filter factor. The values are loaded into 1-hour time spans beginning at the nominal MOS projection (every 3 hours).
Direction - MOS u and v wind components are applied to the grid as above, then the components are resolved into direction. Values are loaded into time periods as above.
These are initialized as described for wind speed, above.
MOS 12-hour PoPs are converted from UTC to local time and applied to the grid as described above. Values are loaded into 12-hour time spans ending at the nominal MOS projection.
MOS 6- and 12-hour PoPs, best category precipitation type (POPT), conditional probability of freezing (POZR), conditional probability of frozen (POF), temperature, conditional probabilities of liquid precipitation type, 6- and 12-hour QPFs, and probabilities of thunderstorms and severe thunderstorms are mapped to a grid using the station-grid mapping technique. The grid is spatially smoothed using a one-pass Shuman filter with an office-specified filter factor. Refer to Figure35 for details.
Figure 35 - Converting MOS To Explicit Weather
Type - For each cloud bounded area, if an overlapping or coincident weather area contains a thunderstorm (any mention), then the lowest cloud layer is marked with CB.
Base - LAMP cloud heights for all prevailing layers are transformed to MSL and then to the grid using the station-grid mapping technique. Groups of grid points with identical cloud groups are coalesced into a cloud bounded area. Bounded areas with identical geographic coverage are combined.
Probability - If a LAMP conditional cloud guidance was used for this layer, cloud probability is set to CHANCE. Otherwise, the probability is set to NULL.
Amount - LAMP cloud amounts for all prevailing layers are transformed to tenths of sky cover, then processed as described above for cloud base.
Probability "A", "B", "C" - MOS and LAMP both forecast QPF as an ensemble of probabilities for various QPF ranges. These are converted into the required probabilistic amounts and applied to the grid using the station-to-grid map. A Shuman filter is used with an office-supplied parameter. QPF at a grid point is set to zero if the POP12 weather element is too low.
As for QPF, MOS forecasts snow as an ensemble of probabilities for various ranges. These are processed as described above for QPF. Snow at a grid point is set to zero or missing if the POP12 weather element is too low.
The MOS Maximum and Minimum Temperatures are processed using the station-to-grid mapping technique and the Shuman filter.
Model grid sources are the eta and NGM. It is expected that, in the future, many more general weather elements will be derived directly from model grids
Starting from the lowest layer moving upward, the algorithm finds the first layer that is freezing or below in the temperature field
Using model temperature fields, inversion tops and bases are determined by the temperature gradient, grouped into similar regions (such as tops within 200 feet), and converted to bounded areas.
For the time specified and for every grid point, gridded model output is retrieved for site-selected layers and used to compute weighted averages of the wind speed and u- and v-components. Transport wind speed is the mean wind speed through the site-defined layer. Transport wind direction is derived by resolving the mean u- and v-components.
Speed and direction are read directly from the model grids at a site-selected height, for the time specified and for every grid point.
For the times specified and for every grid point, the Haines Index is computed using a site-selected model's grids. The Haines Index is defined as:
where T is the temperature at two pressure surfaces (p1, p2) and Tdp1 is the dew point temperature at the lower level. The country is divided into three elevation areas, low, mid, and high, which determines the pressure surfaces to use for the equation. a and b are functions which return integer values.For example, the high elevation Haines Index is derived from the 700 and 500 hPa levels. The two functions are defined in this case to be:
Table 60 - High-Elevation Haines Index Categories
--------------------------------------------------------------------------
a (700 hPa T - 500 hPa T) Value b (700 hPa T - 700 hPa Td) Value
--------------------------------------------------------------------------
less than 18 degrees Celsius 1 less than 15 degrees Celsius 1
18 to 21 degrees Celsius 2 15 to 20 degrees Celsius 2
22 degrees Celsius or greater 3 21 degrees Celsius or greater 3
--------------------------------------------------------------------------
The individual factor values are added together to determine the Haines Index which is converted into categories per Table61.
Table 61 - Haines Index Categories
-----------------------
Factor Values Category
-----------------------
2 or 3 very low
4 low
5 moderate
6 high
-----------------------
For the times specified and for every grid point, the height where a dry adiabat from the surface at the forecast temperature intersects the sounding is computed.
This is computed from the forecast surface temperature and the model sounding, for the times specified and for every grid point.
Swell height and direction are taken directly from special oceanographic swell models, such as the one produced by the Ocean Products Center in Monterey.
To be determined.
To be determined.
For each grid point, the "highest" Weather Coverage/Probability element is found, and transformed into a POP12 value using an office-specified table. An example of such a table is shown in Table62.
Table 62 - Sample Conversion from Weather to Pop
---------------------------------------------
Weather Coverage/ Assigned PoP
Probability
---------------------------------------------
Isolated 10%
Slight Chance, Widely Scattered 20%
Chance, Scattered 40%
Likely 70%
Occasional, Numerous 80%
Widespread 100%
---------------------------------------------
These are simply extracted from the forecast temperatures in the appropriate time range.
After calculating the relative humidity from temperature and dew point at each appropriate time step, the extrema are extracted.
At each grid point and within a specified time range, sky condition is initialized with the highest amount of the prevailing cloud layers. For each time range, adjacent identical sky condition values are gathered into bounded areas. If all bounded areas are identical for adjacent time ranges, the time ranges are combined.
Visibility - Visibility is determined by examining the Weather Type and Weather Intensity general weather elements and using a lookup table.
TAF-specific weather elements are initialized from the GWE grids. A single grid point is used for each TAF site.
Prevailing wind speed, direction, and gust are initialized directly from the Wind Speed, Wind Direction, and Wind Gust GWE grids, respectively.
Where Cloud Probability is DEFINITE, Prevailing cloud type, base, top, and amount are initialized from the Cloud Type, Cloud Base, Cloud Top, and Cloud Amount GWE grids, respectively. Conversions from MSL to AGL are applied for cloud base and top.
Chance and Occasional cloud elements are initialized in the same way as Prevailing, but where Cloud Probability is CHANCE or OCCASIONAL, respectively, at the TAF site's grid point.
Where Weather Probability is WIDESPREAD or LIKELY, Prevailing weather type and intensity are initialized using the Weather Type and Weather Intensity GWEs, respectively. A type conversion is performed from weather types listed as general weather elements to weather types acceptable to TAF forecasts.
A similar process is used for Chance weather, where Weather Probability is SCATTERED or CHANCE, and for Occasional weather, where Weather Probability is OCCASIONAL.
Prevailing, Chance, and Occasional Visibility are initialized directly from the Visibility GWE where the Weather Probability is WIDESPREAD or LIKELY, SCATTERED or CHANCE, or OCCASIONAL, respectively.
TWEB special weather elements are initialized from the general weather elements, along a corridor of grid points representing the TWEB route. Weather elements and coverages are initialized as specified in the matrix below.
Table 63 - TWEB SWE Initialization Matrix
-----------------------------------------------------------------------------------------------------
Cloud (Base, Weather (Type Surface Visibility Hazards
Top, Amount) and Intensity)
-----------------------------------------------------------------------------------------------------
Initialized from: corresponding corresponding Visibility where a comparison of
Cloud GWE Weather GWE where Weather Coverage/ Cloud Base and
where Cloud Weather Coverage/ Probability is: topography, where
Probability is: Probability is: Cloud Probability is:
Coverage
Isolated Slight Chance Isolated Isolated Slight Chance
Local Chance Widely Scattered or Widely Scattered or Chance
Slight Chance Slight Chance
Areas Occasional Scattered or Scattered or Occasional
Chance Chance
Widespread Definite Widespread or Widespread or Definite
Likely Likely
-----------------------------------------------------------------------------------------------------
For example, Local Surface Visibility is initialized from the Visibility GWE, where Weather Coverage/Probability is either WIDELY SCATTERED or SLIGHT CHANCE.
The 10-Hour Fuel Moisture is derived from multiple information sources, both observed and forecast. The temperature and dew point are used to derive the relative humidity. Further information may be obtained from TSP 91-10, Format Fire Weather Forecasts. The algorithm is outlined in Figure36.
Figure 36 - //////10-Hour Fuel Moisture Algorithm
For the time period specified and for every grid point, Precipitation Duration is produced by accumulating the number of hours with one or more weather bounded areas multiplied by an office-specified factor derived from the highest Weather Coverage/Probability element. An example of such a table is shown in Table64.
Table 64 - Sample Precipitation Duration Conversions
------------------------------------------------
Weather Coverage/ Precipitation
Probability Duration Factor
------------------------------------------------
Isolated 0.0
Slight Chance, Widely Scattered 0.20
Chance, Scattered 0.40
Likely 0.60
Occasional, Numerous 0.80
Widespread 1.00
------------------------------------------------
For each grid point and time period, the cloud and weather GWEs are examined, and the algorithm shown in Table65 on page126 is applied to calculate the lightning activity level.
Table 65 - Lightning Activity Level Algorithm
-----------------------------------------------------------------------
Weather Condition Cloud Condition LAL
-----------------------------------------------------------------------
Thunderstorm with Probability of not relevant 3
slight chance or greater
not relevant scattered or greater coverage, 2
cloud type CU or CB
other than above other than above 1
-----------------------------------------------------------------------
Chance of Wetting Rain is derived from the PoP fields directly. No special derivations are required.
The time range for Dew Intensity and Dew Dryoff Time is nominally midnight to noon. For each grid point, the Cimino algorithm is applied using temperature, dew point, wind, and precipitation history to get dew intensity and dew dryoff time values. Values are converted to the nearest hour, and are jointly gathered into bounded areas.
Drying Conditions is categorical and covers 6 am to 6 pm. It is computed from sky condition, dew point, temperature, wind, and PoP. First, the estimated pan evaporation is calculated from the dew point, temperature, sky condition, and wind for each grid point and time period. These values are summed for each grid point to obtain a totalized estimated pan evaporation. A numerical value for drying conditions is calculated based on the estimated total evaporation modified by the probability of precipitation. The numerical value is converted to a category and like areas are gathered into bounded areas. This algorithm is described in detail in TSP-88-27 and is sketched in Figure37.
Figure 37 - Drying Condition Algorithm
The calendar day Hours of Sunshine is calculated from sky condition and sunrise/sunset times. The sky condition for each time period is examined during daylight hours and a percentage factor used for each cloud category. Table66 shows a sample algorithm for determining the factor based on sky condition.
Table 66 - Cloud Sunshine Factor
--------------------------------------------------------
Sky Condition Factor Sky Condition Factor
--------------------------------------------------------
Clear (CLR) 100% Broken (BKN) 30%
Scattered (SCT) 60% Thin Overcast (-OVC) 20%
Thin Broken (-BKN) 40% Overcast (OVC) 0%
--------------------------------------------------------
Hourly dew points (GWE) are analyzed for each grid point and the minimum is determined for the nighttime period (6pm-noon next day).
Wave heights are very complicated since they effectively require a model. Some primitive calculations can be obtained using sea surface temperature, observed wave heights, basin information, wind, and air temperature. The basin information along with wind can be used to determine fetch, which is a very important factor in wave heights.
Table 67 - Wave Height Algorithm
---------------
---------------
get from Alaska
---------------
Superstructure icing is determined from the air temperature, sea surface temperature, wind, and wave heights. A table lookup, such as shown in Table68 on page128 is used to determine the category.
Table 68 - Superstructure Icing Algorithm
---------------
---------------
get from Alaska
---------------
Table69 lists the weather elements that must be manually initialized, i.e., they are not initialized from other sources.
Table 69 - Weather Elements Not Initialized
-----------------------------------------------------
Category Weather Element
-----------------------------------------------------
GWE Wind Gust
Cloud Top
Inversion Type
SWE - TAF Chance Wind Speed, Direction, and Gust
Occasional Wind Speed, Direction, and Gust
-----------------------------------------------------
Footnotes
- (1)
- Local Analysis and Prediction System under development at the Forecast Systems Laboratory.
Table of Contents Next Chapter