Selecting Domain and Grid Resolution
When applying a meteorological model, modelers must decide what domain will cover the area of interest. There are many issues to consider when deciding on a modeling domain. Some of the most important issues are the following:
In this section, we will apply some of information you have learned in this course and also give you examples of how meteorological modelers address these three major issues.
Meteorological models need meteorological data in order to model the actual weather conditions. The more accurate and detailed the data you provide to the meteorological model, the greater the probability of correctly simulating the actual weather conditions. Most models require initial conditions in order to start the simulation. These initial conditions include wind, temperature, pressure, moisture, and cloud cover data.
Most models include a preprocessor system in order to create these initial conditions (e.g. MM5). These preprocessors can use National Weather Service (NWS) data or data from a meteorological model applied on a larger domain that contains the region of interest. If necessary, this information is then interpolated to the grid and subsequently used as the initial conditions for the simulation.
Models must also have information on the terrain, such as surface characteristics and the elevation above sea level. We covered previously how terrain characteristics affect the wind field and the transport of heat to and from the surface of the earth (see Session 2, Part 3: Factors Inluencing Energy Balance and Session 6, Part 1: A Review of Lifting Mechanisms in the Atmosphere). Typically, terrain information is only available at certain horizontal resolutions (e.g. 54 kilometers). Model results can vary significantly when using land characteristics at different horizontal resolutions. In general, a higher resolution of data will provide a more accurate simulation of the atmosphere. This is one of the factors that will affect the modeler's decision regarding what horizontal resolution to use in the modeling application.
Temporal and Spatial Distribution of Data
One way in which modelers evaluate the performance of a meteorological model is by comparing the predicted values to the observed values. The best way to make this comparison is to have an observation data set that covers the spatial and temporal characteristics of the application. Longer simulations tend to stray further and further from observed values. This is often the result of either inaccurate input information or the many approximations that have to be made inside models (refer to Session 9).
Modelers sometimes use a tool called Four Dimensional (length, width, height, and time) Data Assimilation (FDDA) which helps prevent the stray of predicted values from observed values (which we introduced in Session 9). The goal of FDDA is to incorporate actual observations into the model in an effort to create a better description of the atmosphere. This is normally accomplished by first comparing the predicted and observed values and then calculating a nudging factor. The user typically has some control over the "strength" of the nudging by adjusting a nudging coefficient. This factor nudges the predicted value towards the observed value. The nudging factor is calculated and used at a certain time step which the user may be able to modify.
The actual observations typically used for FDDA are National Weather Service (NWS) surface and upper air observations. In addition, modelers can also use ship, buoy, and satellite data. Occasionally, additional meteorological observations are available due to an intensified meteorological or air quality study. The modelers must whether they have enough data throughout the domain and throughout the simulation period to apply FDDA. To apply FDDA with inadequate data would deteriorate the results of the model. The method of applying FDDA varies from model to model. During this tutorial, we will discuss the manner in which FDDA is applied in the MM5. Most models that have FDDA allow the user to enable and disable the FDDA tool.
Computer Resources for Meteorological Modeling
Three dimensional meteorological models can require a tremendous amount of computing time depending on the size of the domain, horizontal and vertical grid resolution, and the length of the meteorological simulation. Thus, modelers must consider what computing resources are available for these simulations. There also may be a time constraint to finish all meteorological modeling for a certain project. Modelers can usually estimate computing time based on information within the User's Guides or by visiting their web page. Another method of estimating the computing time is to conduct a simulation on a very small domain for a very short simulation time for different horizontal resolutions. Modelers can then use the computing times from these simulations and calculate computing times per grid cell. When a larger domain is then chosen, the modeler can estimate the computing time by multiplying the computing time per cell by the number of cells. These are just some methods for helping to determine the domain and spatial resolutions while weighing the computer resources available.
The Shodor
Education Foundation, Inc.