Introduction
The main purpose of Interactive Forecast Preparation (IFP) is to produce a gridded database containing a high-resolution forecast from which all official products are produced. These grids have to come from somewhere, and the numerical models are the source. Editing grids takes time, and the closer the grids represent what the meteorologist thinks the forecast should be, the quicker and more efficient the grid editing process becomes.
The derivation of sensible weather elements from models takes advantage of the quality, detail, and resolution of numerical models by providing initial grids of all surface weather elements used in GFESuite. Resolutions of the numerical models, typically 30-80 km, are not sufficient for the forecast resolution, which is typically 5-20 km. Additional algorithms are applied to simulate an increased model resolution.
The derivation of sensible weather elements from models provide the following benefits:
Weather Elements Needed
Numerical models do not directly provide the weather elements required. Many of the weather elements can be derived from data provided by the model through the use of sophisticated algorithms.
Table 1 lists the sensible weather elements that are produced based on models for the public, fire weather, and marine service areas.
Table 1. List of the Derived Sensible Weather Elements for Users.
How Weather Elements are Derived
Sensible weather elements are derived from the available model data through different algorithms. Typically the models provide the temperature, relative humidity, wind, and geopotential height fields at 50-millibar intervals. Some models provide additional information such as lifted index and other convective parameters.
Interpolation of model data is performed to derive the high-resolution model data at each GFESuite grid point. The algorithm uses a linear interpolation based on the nearest four model grid points for each GFESuite grid point. This technique increases the effective resolution of the model. Examples of some of the algorithms are explained in the next section.
Surface Temperature - Some models provide a surface temperature field. Unfortunately, this field is not directly usable for IFP due to its low spatial resolution. A temperature and geopotential height sounding is calculated from the model data. The model terrain and actual terrain are determined from the model data and high-resolution topography data.
The geopotential height of the actual surface is calculated and then used to determine the appropriate model layers to use for interpolating the model temperature. Not surprising, terrain influences the result as shown in Figures 1 and 2. Figure 1 is the Eta model surface temperature field at its available resolution of 80 km on AWIPS.
There is little detail shown in the field and it is difficult to discern any major major terrain differences. Figure 2 is the result of applying topography/terrain corrections to the model data based on high-resolution terrain information. It becomes quite apparent where the valleys and mountains are located.
Figure 1. Unmodified 80-km Eta model surface temperature field, showing lack of terrain detail.
Figure 2. IFP surface derivation from Eta model temperature soundings with terrain corrections, clearly showing the terrain influences.
Daily Maximum and Minimum Temperature - The derived surface temperatures are used to determine the daily maximum and minimum temperatures. Models usually do not generate the maximum and minimum temperature because they have no added benefit as model output. The maximum and minimum temperature, however, is one of the primary forecast weather elements produced by the forecasters.
Since many models only produce snapshot temperatures every 6 hours, it is not sufficient to take the maximum and minimum values of these sparse grids. A spline is fit through each grid point in the sequence of temperature grids produced by the models and then hourly temperatures are derived. The maximum and minimum values of these temperatures are then stored in the database.
Weather Type and Character - The public wants to know the weather prognosis, perhaps whether it is going to rain or snow. Numerical model output needs to be interpreted in order to determine the weather type and character. Models provide the amount of precipitation and temperature profiles, but they do not directly indicate whether it will rain or snow.
The type and character algorithm is quite simple, yet effective. It assumes that the model Quantitative Precipitation Forecast (QPF) is accurate, finds all areas that have a nonzero QPF, and then examines the surface temperature to determine the state, whether it is liquid or frozen. The "showery" character of the precipitation is determined by examining the ratio of convective precipitation to total precipitation. Thunderstorms are included if the calculated lifted index is lower than a specific threshold. The result of this algorithm is no weather, rain, rain showers, snow, snow showers, or thundershowers. Some of the newer models contain additional fields, such as the Convective Available Potential Energy (CAPE) and Convective Inhibition (CIn), which may be used to improve this algorithm. Figure 3 shows the calculated precipitation type and character based on QPF, the convective precipitation to non-convective precitation ratio, surface temperature, and atmospheric stability. In the northern portion of the display, the algorithms predict widespread rain, since the temperature is above freezing and there is little convective precipitation from the model. In the center of the display, the atmosphere is more unstable resulting in the depiction of thunderstorms. In the southern portion of the display, the atmosphere is more stable but the ratio of convective precipitation to total precipitation is higher, resulting in the depiction of rain showers. Figure 4 shows the QPF for the same time as Figure 3. Note that the precipitation areas in Figure 3 match the QPF in Figure 4.
Figure 3. Weather display showing the type and distribution of precipitation over the Charleston, West Virginia, forecast area.
Figure 4. Quantitative Precipitation Forecast field derivation for the same time period and area shown in Figure 3, showing the tight coupling between the weather derivation algorithms and the QPF field.
Snowfall Amount - Snowfall amount is determined from the surface temperature, the wet bulb temperature, and the QPF. The ratio of snowfall to QPF is determined based on the two temperatures and ranges from 10:1 in very wet snow conditions to 25:1 in extremely dry snowfall conditions. Of course, there is a check in the algorithm to prevent any snow at all above a certain temperature. The amount of snow is simply calculated as the QPF multiplied by the snow ratio.
Other Weather Elements - Some elements, including surface wind and QPF, are used directly from the model without modification.
Figure 5. Algorithm written in Python to calculate the Freezing Level.
Numerical python is both fast and flexible. It uses a slightly different syntax than that of smart tools. Below is an example of the python code which calculates Freezing Level.
def calcFzLevel(self, gh_c, t_c, topo): fzl = self._minus for i in xrange(gh_c.shape[0]): try: val = gh_c[i-1] + (gh_c[i] - gh_c[i-1]) / (t_c[i] - t_c[i-1])\ * (273.15 - t_c[i-1]) except: val = gh_c[i] fzl = where(logical_and(equal(fzl, -1), less_equal(t_c[i], 273.15)), val, fzl) fzl = fzl * 0.3048 return fzl
Summary
The derivations of sensible weather elements provide a key initial first guess for forecasters. Without the first guess, the editing process would quickly become too burdensome. As model physics and resolution improves, so will the quality of the derivations. With recent changes in the framework of the GFESuite, users may now adjust the algorithms to make their own improvements based on scientific knowledge.
(Mark Mathewson can be reached at mathewson@fsl.noaa.gov.)