FORECASTING
This exercise was originally created by Dr. Marion Alcorn,
at Texas A&M's Department of Atmospheric Sciences







This is a hands-on laboratory exercise that offers you a chance to learn about weather forecasting methods. There are 10 problems to be completed. If you would like to complete this exercise and have it graded by a meteorologist with the National Weather Service, simply mail your answer sheet (along with the required attachments) to:

Herron/Massie
National Weather Service Forecast Office
500 Weather Station Road
Old Hickory, TN 37138


Now, have fun! and take a dive into the world of weather forecasting!







METHODS OF FORECASTING



Introduction

In this exercise, we will look at some methods used in making weather forecasts and will make forecasts of various weather elements for several locations. In weather forecasting, a meteorologist is attempting to predict how the weather will change during a specified period and what the weather conditions will be during the period of the forecast. Actually, making a forecast of weather conditions is quite easy. The difficulty arises when a forecaster wishes to be accurate. Of course, a forecast isn't much good if it does not accurately state the future weather conditions that will occur. Thus, meteorologists are continually attempting to improve the accuracy of forecasts and to accurately predict further into the furture.

To make an accurate forecast, a meteorologist must first understand what processes are occurring in the atmosphere to produce the current weather at the location for which the meteorologist is forecasting. This is done by measuring certain elements (making observations) of the atmosphere; i.e., temperature, pressure, wind direction and speed, humidity, cloud cover, precipitation, etc. The more complete measurement coverage across the earth's surface and vertically through the atmosphere of the those elements which affect the "weather" we experience, the better "picture" we have of the processes producing the weather we are currently experiencing. By observing the changes which take place to these elements over time and comparing the changing patterns with historical patterns, an understanding of expected future weather conditions can be made.

If meteorologists can understand how the atmosphere changes over time in response to various factors; i.e., differences in warming across the earth's surface from solar radiation, radiational cooling at night, warming of the atmosphere due to latent heat release during condensation, etc., and can write mathematical equations to express these changes, then a useful tool becomes available to the forecaster - computer models - which can be constructed to express how the atmosphere is changing and will appear at some future time. The output from these models can be used as an aid to forecasters in preparing the forecasts.

Note: these computer models are by no means perfect and should never be relied on exclusively by forecasters when preparing forecasts. They are a tool and should be used in conjunction with the forecaster's understanding of the changing weather patterns, as determined from a close examination of measured weather data to determine if the actual weather conditions are changing in the manner that computer models are predicting that the weather conditions will change.

Major research and development effort is ongoing in improving all areas of the process, from development of better observational techniques (both surface systems, upper air systems, and satellite systems), development of forecasting techniques to be used by forecasters, to development of better mathematical equations and computer models, to procedures to communicate weather information to users in a timely and reliable manner.

One might think of the forecast preparation process to provide users (the public, industry, etc.) with needed information as depicted in the following figure. (1) Observations give the forecaster information about what is actually occurring in the atmosphere. The computer models also use these observations to provide information concerning possible future conditions. (2) The forecaster uses the latest observations and computer model information, a forecaster-machine mix, to develop (3) a forecast which is then distributed to users. Some attempts have been made to (4) control and modify the environment but more research needs to be done in this area before it ever becomes widely used.


There are several methods used in forecast preparation, depending on the time element involved and the weather element for which the forecast is needed.



Methods of Forecasting

  • Persistance Forecasting. Persistance forcasting is a prediction that the weather in the future will be the same as it currently is; that there will be no change to the weather conditions. Persistance forecasts are generally good only for short periods of a few hours and become less accurate as the time period lengthens. If it is raining now and a forecast that it is going to rain for another three hours would be a persistance forecast. In the tropics, especially at island stations, where day after day the weather is basically the same because the station is affected by the same air mass with no passages of fronts, a persistance forecast that tomorrow is going to be the same as today is usually quite accurate.
  • Steady-state or Trend Forecasting. In this method of forecasting, the forecaster is looking at the changes that are occurring in the weather systems; the fronts, air masses, high and low pressure systems, which are affecting the station. The forecast is based on the assumption that these changes will continue at the same rate they have been occurring. Thus, if a cold front is approaching the station at 20 miles per hour, then it will continue to move at 20 miles per hour in the same direction; so, the forecaster can determine the weather conditions based on the location of the front determined by extrapolating its position assuming its rate of movement doesn't change. Similarly, if a cold air mass is moving toward the station and temperatures at stations within the air mass are dropping at one Fahrenheit degree per hour, then temperatures at the station for which you are forecasting will drop at one Fahrenheit degree per hour. However, rarely will a front move at a consistant rate of motion for 24 hours or more. Thus, steady-state forecasting gives a good guide to follow at least for short periods of time. Attempting to use such a method for a forecast greater than 24 hours will usually prove inaccurate. For forecasts of a few minutes to several hours, the method has proven successful. The method called nowcasting which refers to forecasts for the next several hours, are often based on such steady-state techniques.

Open the meteogram image for Tallahassee, Florida.

Problem
1.
This meteogram shows the measured air temperature, along with other meteorological variables, at the National Weather Service Office in Tallhassee, Florida.

Assume that this meteogram is showing measured weather information for today. (It is actually data from July 6 and 7, 2002.) Assume also that it is 10PM EDT at night, tonight, (0200Z on the image) and you must make a forecast for the minimum temperature for the upcoming morning. The minimum temperature typically occurs between 1000 and 1100Z (06AM and 07AM EDT in Tallahassee) assuming no major changes in the air mass across the station. Using the steady-state method (trend), and a 3-hr trend calculated from the temperature at 23z and 02z (units=degrees/hour), determine what the minimum temperature should be at 1100Z.

Note: "Z-time" is also known as Universal Coordinated Time (UTC) or Greenwich Mean Time (GMT). This time is given according to a 24-hour clock and is the "local time" at Greenwich, England. Thus, 18Z means 1800Z, which means 6:00 p.m. in Greenwich, England. "Z-time" is always ahead of what our clocks say here in the United States. For instance, when it is 18z in Greenwich, England, the time is noon Central Standard Time or 1 pm Central Daylight Time. To learn more about how to convert from "z-time" to "U.S. time" [go here].

Record your answer on your answer sheet.

Based on the temperature indicated on the meteogram for 10Z, do you think your "steady state" forecast will end up being a "good" forecast or a "bad" one?

Now, consider how to make a forecast for the maximum temperature for tomorrow from this observational information. Remember, you want to be accurate. This may be somewhat more difficult since you have no trend to follow (ie, any "trend" ending at 10z only indicates rate of nighttime cooling).

Make a forecast for the maximum temperature for tomorrow and explain what method you used, either persistance or steady-state (trend) and how you used this technique to arrive at the maximum temperature value you chose.

Record your answer on your answer sheet.

Close the meteorogram.
Open the image, "TEMP MAP".

Problem
2.
This map shows an analysis of surface data, including temperature (the red-colored numbers).

What is the average temperature over southeast Colorado?

Suppose that the air mass causing this temperature regime in southeast Colorado was moving southeast at a rate of 150 miles in 24 hours.

What will the temperature be at Wichita Falls, Texas 24 hours from now? (Hints: Wichita Falls is located in north-central Texas, near the Oklahoma border. The station identifier on your surface map is "SPS." The distance from southeast Colorado to Wichita Falls is about 150 miles). What will be the difference in temperature at Wichita Falls between now and tomorrow at this same time (warmer, cooler, roughly the same)?

Record your answers on your answer sheet.

Print the surface map and attach it to your answer sheet(s)

The temperature you obtained for Wichita Falls is probably not going to be the actual temperature Wichita Falls receives for several reasons.

  1. The air mass will probably not continue to move at the rate predicted for the full 24-hour period.


  2. As the air descends from the elevations of southeast Colorado (1447 meters) to Wichita Falls ( 291 meters) it will experience adiabatic warming at a rate of 1oC per 100 meters, or almost 12oC; and,


  3. If the land at the more southerly latitude of Wichita Falls is warmer than in southeast Colorado, the air mass will be warmed as it progresses southward.


These type of considerations must be made by a forecaster when preparing a forecast.

  • Analogue method. This technique utilizes the fact that existing weather patterns on weather charts which resemble previous weather patterns on previous weather charts should produce the same type of weather elements, or phenomena, that the previous patterns produced. These previous patterns can then be used as a guide for making forecasts of weather elements. For example, if from previous weather maps it is seen that an intense ridge at 500-hPa during the warm months over the west coast produces a surface, high-pressure center located over the Nevada-Utah regions and this pattern produced strong Santa Ana winds along the west coast of the United States, then when a forecaster sees a pattern developing which has a strong ridge at 500 hPa developing near the west coast, a forecast of strong, easterly winds at coastal stations will likely be accurate. Similarly, such a pattern in the cold months tends to direct those polar and arctic air masses located in central Canada toward the southeast and brings cold outbreaks of air across the central and eastern part of the United States.

    This analogue method, or pattern recognition - the ability of a forecaster to recognize weather patterns which will produce certain weather phenomena - is a vital skill a forecaster needs to be able to prepare accurate forecasts. Sometimes, these patterns can be used to predict the conditions for several days in advance, although the predictions of specific conditions; how strong the Santa Ana winds will be or how cold the temperature will be, are often not adequately predicted by such techniques.

  • Climatological Forecast. This method uses such guidelines as the average value of weather elements for a region, the maximum and minimum values of weather elements, the most or least time of occurrence of certain weather phenomena, etc. to make a prediction of the value of those weather elements for some future period. It is based on the assumption that the specific element value will not be significantly different than the values of that element from previous observations. As an example, if you were making a forecast as to whether College Station would have snow for Christmas, then, knowing that during the 30-year climatology record for College Station only a trace of snow was ever recorded, a forecast of no snow for Christmas next year would probably be quite accurate. Simalarly, a forecast of no snow for Christmas in the year 2010 would also probably be quite accurate. Climatology can be used as a guide for making both short-term, hours to days ahead, and long-term forecasts, such as 30-day and 90-day forecasts or longer.

Open this image.

This image is a Climatic Outlook chart produced by the National Climatic Center. These charts provide expected conditions for up to a year in advance of the time the chart was produced. Each chart normally covers a three month period. This chart shows the outlook for temperature.

Problem 3.
At Nashville, Tennessee during the months indicated on the image you just opened, are the average temperatures expected to be above (A), below (B), the same as - normal (N), or unable to determine (CL) as indicated by the National Weather Service outlook map. If the temperatures are expected to be above or below normal, indicate by what percent; i.e., between 0 and 5%, 5 to 10%, or greater than 10%.

Note: The months for which this forecast is valid are indicated by the first letter of each month (ie, JAS= June, August, September; ASO= August, September, October; etc.)

Answer the same question for Richmond, Virginia.

Record your answers on the answer sheet.

Print the surface map and attach it to your answer sheet(s)

Open this image.

This chart is the Climatic Outlook for precipitation for the same 3-month period as shown by the temperature outlook.

Problem 4.
At Nashville, Tennessee during the months indicated on the image you just opened, is the average precipitation expected to be above (A), below (B), the same as - normal (N), or unable to determine (CL) as indicated by the National Weather Service outlook map. If precipitation is expected to be above or below normal, indicate by what percent; i.e., between 0 and 5%, 5 to 10%, or greater than 10%.

Answer the same question for Richmond, Virginia.

Record your answers on the answer sheet.

Print the surface map and attach it to your answer sheet(s)

Below is a table giving the climatological conditions for Nashville, TN and Richmond, VA which has been determined from past observations at these two stations.

NASHVILLE STATIONMEANS
Temperature (C)JanFebMarAprMayJunJulAugSepOctNovDec
Mean Maximum4752 60717987908983736050
Mean Minimum2831 39485765706861493932
Mean Monthly3841 50606876807972615041
Precipitation
(inch)
JanFebMarAprMayJunJulAugSepOctNovDec
Mean 3.583.81 4.854.374.88 3.573.973.46 3.462.624.12 4.61
Record Monthly Maximum13.9210.3112.358.41 11.0411.957.75 8.3111.446.13 9.0413.63
Record Monthly Minimum0.190.641.180.520.690.450.710.690.28trace0.540.98
RICHMOND, VAMEANS
Temperature (C)JanFebMarAprMayJunJulAugSepOctNovDec
Mean Maximum475059 697885 888781716151
Mean Minimum283036455564686760483831
Mean Monthly384048576775787771605041
Precipitation
(inch)
JanFebMarAprMayJunJulAugSepOctNovDec
Mean 3.243.16 3.512.963.843.625.034.403.343.533.173.26
Record Monthly Maximum7.975.978.657.318.879.2418.8714.1016.609.397.647.07
Record Monthly Minimum0.640.480.940.640.870.380.510.520.260.010.170.40

Based on the above climatology table, answer the following question.

Problem 5.
What is the sum of the monthly mean (average) precipitation that Nashville normally receives during the same months used in problems 3 and 4?

Answer the same question for Richmond, Virginia.

Record your answers on the answer sheet.

By knowing the temperature and precipitation that is normally received at a station by keeping a record of past observations, such as was done to create the above table, and by having an Outlook chart, one can now have an idea as to whether the temperature and precipitation will be above or below normal and some idea as to how much; between normal and 5% above (or below) normal or greater than 5% above or below normal, etc..

  • Numerical Weather Prediction . Numerical Weather Prediction involves using mathematical equations which describe the processes occurring in the atmosphere that cause changes to weather elements; such as, temperature, pressure, wind speed, wind direction, moisture content, etc., those elements used to describe the state of the atmosphere. Typically, this "state of the atmosphere," or "picture," is defined by the value of weather at many different locations, called grid pints, not only at ground or sea level, but also vertically in the atmosphere (troposphere and lower stratosphere). The horizontal distance between these grid points is different for different models and the number of levels from sea level up to the lower stratosphere differs for different models. Once weather observations are made and the value of the measured weather elements are entered into the program, the computer can then solve the equations to determine new values of the weather elements for some period in the future; for example, ten minutes past the time the observation measurements were made. The computer then uses these new values to determine subsequent values at each grid point ten minutes later. This procedure continues until values have been determined for 12, 24, 36, 48 hours, and for some models even longer, into the future. The computer then prepares prognostic charts based on these calculated values, analyzes the data and determines locations of fronts, pressure centers, highs and lows on upper air charts, etc. The charts can then be printed or displayed on computer terminals for forecasters to use in preparing forecasts.

    Prognostic maps produced by numerical methods are only as good as the equations defining the processes, (and for some processes, no equations exist), the accuracy and coverage of the observational data used by the models, the techniques used to develop the model and the ability of the computer itself to accurately run the model.

    It is important for forecasters to not rely solely on model output, but rather to use them as another tool in preparing their forecasts.

    The maps prepared from these computer models, as well as analyses of observational information, are hanging on the 12th floor map display. Some of the models in current use are the RUC Model, the Nested Grid Model, the ETA Model, the Aviation Model, the Rapid Update Cycle (RUC) Model, the Medium-Range Forecast Model. The following links provide access to the prognostic charts produced from these models.

NGM 24-hr prognostic map ETA 24-hour prognostic map Aviation Model
to 72 hr
RUC Model
to 12 hr
Medium-Range Forecast Model
204 to 240 hours.

The Medium-Range Forecast Model prepares prognostic charts out to 240 hours (10 days).

Compare the NGM map and the ETA map. Look at the 24-hr surface pressure, location of fronts, temperature forecast and dew point forecast. Notice that for the same forecast (prognostic) time, there are differences in the two maps.

To make a forecast using these various prognostic maps, a meteorologist must first decide which of the maps is closest to being correct.

Problem 6.
Pick two differences between the NGM 24-hr prognostic map and the ETA 24-hour prognostic map. Explain what those differences are. You do not have to explain why there are those differences.

Record your answers on the answer sheet.

Print these maps and attach them to your answer sheet(s)

Problem 7.
Look at this Aviation Model map. This map shows a 24-hour forecast of sea-level pressure, (solid lines), 6-hour accumulation of precipitation, (shaded areas), and the 1000-500 mb thicknesss, (dashed lines). Remember that wind blows clockwise (anticyclonically) about high pressure centers and counter-clockwise (cyclonically) about low pressure centers.

Look at the 24-hour sea-level pressure forecast. From what direction should the winds be blowing across Texas A&M University (in east-central Texas) at the valid time of this map?

Record your answers on the answer sheet.

Print the surface map and attach it to your answer sheet(s)

Problem 8.
Look at this site. This site provides links to several maps generated by the Rapid Update Cycle (RUC) model. Under the Forecast Time(s) column, check the box labeled loop all times. Then choose the link titled: MSLP/Winds. Notice the movement of the High and Low pressure centers and the changing wind directions during the next 12 hours shown on the five maps.

Go BACK one page and again, make certain the loop all times box is checked. Now select the link titled: Clouds all levels. Notice at the bottom of the image that the colors indicate the composition of the clouds; whether they are of liquid water droplets, supercooled droplets, or ice crystals.

Is the model forecasting for: no clouds, for the clouds to increase, decrease, or stay the same at Texas A&M University (east central Texas) during the next 12 hours?

Record your answers on the answer sheet.

Print the first and last map in this loop and attach the copies to your answer sheet(s). To copy the first map, simply use the left-most toggle button to the upper left of the graphic to step to the initial time. Print the copy. Then, to copy the last map in the loop, simply use the right-most toggle button to step to the last frame. Print the copy.


  • Model Output Statistics. Model Output Statistics (MOS) is an objective weather forecasting technique which consists of determining a statistical relationship between the element being forecast and values calculated by a numerical model. For example, the probability of precipitation occurring at a location during a 12-hour period may be based on the season of the year, the sea level pressure value at that location, the surface, 700-mb and 500-mb temperature values calculated at the location, the relative humidity values calculated at levels from the surface to 200 hPa, or the difference between the actual mixing ratio and the saturation mixing ratio at various levels. These are typical element values which may be used in an equation to determine the probability of precipitation at a location. The equation used for determining cloud amount may use a completely different set of element values. Bulletins containing the MOS forecasts are computer generated for various stations in the United States and are a tool used by forecasters in preparation of their local forecasts. Since the MOS forecasts are generated from statistical considerations using values generated from computer models, and thus can only be as accurate as the models, the forecaster should use them only as a guide and should modify the MOS information as necessary for local conditions.

    Model Output Statistics can be presented in a common table format or in the less-common form of a meteogram, with which you should be familiar. In a MOS meteorogram, the computer model generated weather element values are projected into the future in a graphical format. For more information on meteograms, you can check out Technical Attachment No. 01-13 (October 2, 2001), written by Beth McNulty and published by the NWS Western Region Headquarters. The title of the article is, "Meteogram Analysis and Interpretation".

    As the numerical models evolve, new sets of MOS information become available and, conversely, others disappear. For instance, one set of MOS data which was based on the old LFM model is now defunct. The following sets of MOS guidance are currently available: FWC data [based on the NGM model], MAV and MEX data [based on the Aviation model], ETA data (based on the Eta model), FMR data (based on the extended MRF model). A directory which includes all of this MOS data, for cities in the United States, is available from the NWS Southern Region Office

Problem 9.
Using the NWS's directory of NGM Model Output Statistics (FWC MOS), prepare forecasts of the indicated elements for the following cities for the days and times indicated. Need help interpreting MOS codes? [Go Here] Since you do not have observational data or climatological data for these locations, you cannot be expected to determine whether actual conditions are changing in the same manner that the computer models are expecting the conditions to change. Thus, you cannot be expected to modify the MOS data to produce a more accurate forecast. Rather, simply use the information that MOS is presenting.

Most of the information requested below is for a specific time (i.e., 18z).

When calculating wind direction, remember to multiply the value for "WDIR" by 10. For instance, if WDIR=32, the actual wind direction is "320 degrees." This is a compass direction, with 360 degrees indicating a wind from the North, 090 degrees from the East, 180 degrees from the South, etc.

When determining the "Forecast Weather at 18z," all you have to do is give the FWC code for "OBVIS" (visibility obscuration), "POP06" (6-hour probability of precipitation) and "CLDS" (clouds). Also, if "PTYPE" is listed on the FWC, indicate the code value on your answer table. This is only a seasonal element carried during the cool season and will obviously not be carried during the summer months, when precipitation type is assumed to be rain.

Once you have produced a MOS collective, using the directory of NGM Model Output Statistics, use the copy function on your computer to highlight and copy all of the MOS data into an empty file (perhaps, with WORDPAD). Print a copy of the MOS data and circle the elements that you use to complete your answer table.

Attach a copy of the MOS data with your answer sheet(s). Record your answers in a table similar to the one below.

Tomorrow
Max.
Temp.
Min.
Temp.
Wind
Dir.
at 18z
Wind
Speed
at 18z
Visibility
at 18z
Range in Height
of lowest cloud layer
(at 18z)
Precip.
Amount
( total for 6 hrs, at 18z)
Forecast
Weather
at 18z
Boston MA
_______ _______ _____ _____ ______ ______ _______ _______
Houston TX
_______ _______ _____ _____ ______ ______ _______ _______
Jacksonville, FL
_______ _______ _____ _____ ______ ______ _______ _______
Richmond VA
_______ _______ _____ _____ ______ ______ _______ _______
San Antonio TX
_______ _______ _____ _____ ______ ______ _______ _______
Seattle WA
_______ _______ _____ _____ ______ ______ _______ _______
Sault Ste Marie MI
_______ _______ _____ _____ ______ ______ _______ _______

Problem 10.
Using the Forecast Weather Codes determined in Problem 9, provide a text-style forecast for each station. Assume "CLDS" codes of "SC" and "BK" mean "partly cloudy."

For example, if the weather codes were:

OBVIS=H

POP06=30

CLDS=BK

PTYPE=R

The text-style forecast might read something like, "It will be partly cloudy and hazy with a 30 percent chance of rain."

Record your answers on the answer sheet.





Copyright © 1996-2000 Texas A&M University, Department of Atmospheric Sciences and Marion Alcorn.