How to Make a Weather Forecast

Objective

The purpose of this web page is to describe, in a conceptual manner, how weather forecasts are prepared.

The Evolution of Weather Forecasting

The “art and science” of weather forecasting has evolved over the last 60 years from a forecaster-based analysis and synthesis process to a computer-dominated process. The purpose of this section is to provide an overview of the changes that have occurred during this period.

Forecasts 60 years ago were prepared by an analysis of current surface and upper air data, combined with a series of forecast rules and schematic atmospheric models based on meteorological science. These forecasts were strongly influenced by the experience of the forecaster. The first computer forecast, the 3-level 500 mb barotropic model went operational in 1955 with a 3-level baroclinic projections available by the early 1960s. These model projections increased the level of science in the forecast preparation process. They allowed forecasters to anticipate surface changes based on movement of mid-tropospheric features as depicted by the computer forecast models. They also extended the timeframe of the forecast beyond today and tomorrow.

As computer size and power increased, meteorological forecast models improved how they could describe the atmosphere. The late 1960s into the 1970s saw the introduction of multilevel atmospheric models. By the 1980s a sophisticated boundary layer was added and forecasts of surface system improved significantly. During the next two decades the increased computer size and capability incorporated a wide variety of physical processes that increased the quality and extent of the numerical forecast. In addition, limited-area regional models become operational.

By the early 2000s the numerical forecasts were sufficiently accurate in time and space that computers could produce a worded forecast without human intervention. By the 2010s the development of mesoscale forecast models improved the ability to forecast convection, including significant thunderstorms.

The ”art and science” of weather forecasting has evolved from a highly human-based process (the art) to a highly computer-based process (the science) over this sixty year period. The remainder of this web page will focus on human-based processes and describe these processes in a conceptual manner.

How Do You Prepare a Weather Forecast?

All forecasts start with weather data. After an initial review of data, an initial analysis produces an understanding of what is happening in the atmosphere and, more importantly, why it is happening. After a final synthesis of all available information, a final forecast is prepared. In the following sections, each pieces of the process will be discussed in more detail. Remember that forecast preparation is not a "cook book" process. Even though you may start your forecast process in a similar way each day, at some point, the current weather situation will dictate what you examine and how you use this information.

It All Starts with Data!

All weather forecasts start with real weather data. These data are collected from a variety of sources including surface observations, rawinsonde observations, wind profilers, aircrfaft, satellites, and radar. Each observation platform provides one piece of the puzzle that makes up a total picture of the atmosphere at some point in time. Past observations provide a time series of information to evaluate changes in the current state of the atmosphere.

Real data tell you what is actually happening in the atmosphere. In a mathematical sense, forecasting is an initial value problem. If you don't have a good picture of where the atmosphere is starting, it can be difficult to accurately determine its future state. This is particularly important for forecasts of short time durations (0 to 6 hours). Be sure the forecast follows from what you see outside the window!

With the extensive use of computers to display data, ensure that you are looking at actual data and not a combination of real data and computer enhancement. For example, radar displays (as seen on television) typically show precipitation type. Radar observes radar return intensity. The television radar display combines other information, for example, computer model information and/or surface observations, to infer precipitation type. Be aware of what you are examining.

Analysis

In order to make sense of the vast amount of data available, you must "analyze" it. What does analysis mean? The American Heritage Dictionary of the English Language dictionary defines analysis as:

The Glossary of Meteorology defines analysis as follows:

This process usually consists of drawing and interpreting the patterns of winds, pressure, pressure change, temperature, humidity, clouds, and hygrometers. Each element is examined separately in order to provide the analyst with a better understanding of the current state of the atmosphere and on-going physical processes. The manual drawing of isopletehs has been replaced with computer drawn charts. When charts were drawn by hand, the person drawing the isopleths was also looking at data details and, in a sense, starting the analysis process. Using computer drawn charts adds a burden to the forecaster to examine patterns more carefully.

Part of this analysis process is the use of derived parameter, that is, quantities calculated from observational data via objective computer analysis. Derived parameters include such things as moisture flux divergence, temperature advection, or vorticity. These parameters can be evaluated qualitatively, but objective analysis often gives a more detailed picture.

Initial Synthesis

Once you have taken things apart, you need to put them back together using synthesis. Synthesis is defined as:

The overall purpose of the analysis-synthesis process is to develop a complete understanding of why the atmosphere is doing what it's doing. Why is the current weather occurring?

Final Synthesis and the Forecast

This step brings together the real data and the computer forecasts for one final review and synthesis. You have to ensure that you have a good understanding of what has happened and what will happen. You need to ensure that the temporal flow of your forecast is consistent. The methods you applied during the analysis-synthesis process will also be useful here.

The forecast brings together in words (or code) what you want to tell your users. It provides the temporal and spacial details consistent with the specific forecast product being issued. The idea here is to convey useful information within the state of the art and provide as much accuracy as possible without hedging any more than necessary.

Numerical Weather Prediction

It is beyond the scope of this web page to examine numerical weather prediction (NWP), i.e., computer forecast models, in detail. Nevertheless, a few comments are in order. Listed below are several components of NWP.

Objective Analysis: Running parallel to the human analysis-synthesis process is the computer processing of weather data. In most cases this processing starts with objective analysis or OA. The purpose of OA is to transform weather observations that are irregularly spaced onto to a regularly spaced grid. These gridded data sets are used to initialize numerical forecast models and also allow the analyst to calculate "derived parameter" such as temperature advection or vorticity.

NWP Models: The objective analysis provides an initial set of data that are used by numerical models. Numerical models are mathematical representations of atmospheric parameters and processes. Most models use some form of the hydrodynamic equations to predict wind, temperature, moisture or other atmospheric parameters. Several numerical prediction models are run operationally and are available for forecaster use. Check the National Centers for Environmental Prediction website for the latest information on currently available forecast models.

Computer Forecasts: Numerical forecast models produce a wide variety of output. Some forecasts are basic weather charts such as 500 mb or 300 mb contour, temperature, and/or isotach fields. Derived parameters such as divergence are also available. Basic model data are processed via statistical methods to produce products such as Model Output Statistics (MOS). Further processing of these data are now used to produce computer worded forecast products.

One very valuable aspect of today's computer forecast models is the gridded data sets that are routinely available. These data sets allow anyone with the proper software to manipulate the model forecast data into any form that can be represented by equations or algorithms. In some ways, what you, as a forecaster, examine is limited only by your imagination.

Computer Forecast Evaluation NWP: The goal here is to bring together real data interpretation with the computer products. At this point you need to ask yourself questions like:

Another part of the NWP evaluation involves the relative influence on the forecast of observations versus model output. Beyond 18 hours from issue time, numerical forecast model output is the primary basis for the forecast. The main issue at this point in time is to determine which model will have the most influence on your forecast.

On the other hand, within the first 12 hours from issue time, observations and the conclusions drawn during the analysis-synthesis process should have an impact on the forecast. As time progresses into the forecast, the influence of "what's outside the window" slowly diminishes while the influence of the computer model increases. In any given situation you have to judge the relative importance of these two factors.

Some people would say that the information provided by the computer forecast is all you need to issue a forecast. It is true that there is sufficient model output that one can issue a forecast based only on computer output with little or no intepretation of any other information or data. However, a real meteorologist needs to be able to understand what the atmosphere is doing and why things are going on, not just be able to interpret computer output. If forecasts become no more than a repetition of what the computer says, why do we need people (meteorologists) in the middle?

In Summary

A nice way to summarize the weather analysis and forecast process is stated by the following six questions:

In discussing these six questions in his article in Weather and Forecasting (2003), Dr. Bosart reinterates what was stated above. With the availability of computer forecasts, it is easy to go straight to "e" and not concern youself with the "why" questions. However, real meteorologists need to answer all six questions as they prepare their forecasts. A large part of a meteorologist's experience is gained by watching the weather "over the years" and answering the "why" questions. If you skip these "why" questions and rely on computers to do all your thinking, your experience level with atmospheric processes will be weak and your claim to the title "meteorologist" can be questioned.

One last thought, perhaps that of person who has examined a lot of data and analyzed a lot of maps and believes that this information tells you a lot about the future state of the atmosphere:

Review Questions

Instructions: Place the cursor over the answer of your choice. If you are correct, it will be highlighted in green; if you are incorrect, it will be highlighted in red.

The weather forecast process starts with:

1.  analysis
   
   
2.  synthesis
   
   
3.  weather data
   
   
4.  computer forecasts

The weather data process encompasses:

1.  collecting and examining weather observations
   
   
2.  checking time series of past information
   
   
3.  looking out the window
   
   
4.  all of the above

Derived parameters refer to quantitues calculated from observational data via computer objective analysis.

1.  true
   
   
2.  false

The overall purpose of the analysis-synthesis process is to develop a complete understanding of why the atmosphere is doing what it is doing.

1.  true
   
   
2.  false

Numerical weather prediction is based on:

1.  mathematical representation of atmospheric processes
   
   
2.  observation trends
   
   
3.  statistical analysis of observations
   
   
4.  computer operating system biases

You should never compare short term forecasts based on observed data to short term computer forecasts.

1.  true
   
   
2.  false

As "time after forecast issuance" increases, the influence of observations on a forecast normally:

1.  is not considered
   
   
2.  decreases
   
   
3.  increases

Real meteorologists can answer the question: Why is today's weather occurring?

1.  true
   
   
2.  false