1 THE FORECASTING PROCESS

with glacial valleys, and the continental plateau itself, may also give rise to local wind systems (e.g. katabatic winds; mesoscale lows).

· The location of any significant gaps in mountain ranges that might require special attention in the prediction of wind or cloud.

· The distribution of rock/ice/sea ice/open water. These factors may be important in determining local heat sources and sinks and mesoscale/micro–scale variations in low–level moisture.

· The presence of rocky outcrops close to or on a flight path might also be crucial to ameliorating the effects of white–out in overcast conditions.

6.1.2 Getting to know an area's climatology

Basically, climatology summarises various aspects of weather (e.g. average cloud amount; days of gales; blowing snow; fog, etc.) in a convenient form (e.g. frequency tables) and consequently provides some information on the normal expectation of weather at a particular locality. Appendix 2 contains climatological data for many of the places covered in this handbook. The value of these data for short–term forecasting varies greatly with locality and season. Frequently climate statistics only provide background information, but, where the weather is strongly linked to seasonal and diurnal patterns, climatology exercises a strong influence on the forecast. Nevertheless, climatology must not be insisted on at the expense of synoptic and dynamic reasoning.

Other types of climatological information have more definite relevance in everyday short–term forecasting. These include:

· typical diurnal behaviour of the katabatic wind at a particular station;

· mean hourly rates of change of temperature and dew–point temperature and the change under different synoptic conditions;

· typical rates of precipitation for different synoptic patterns.

An additional factor often overlooked in applying geographical and climatological data is the effect of actual antecedent conditions: for example, is there more open water nearby than usual that might indicate increased cloudiness or an increased risk of fog? In marginal situations the incorporation of knowledge of these effects could make the difference between a confident (and successful) forecast and an uncertain one.

6.1.3 Major steps in short–term forecasting

In practice, the best approach to short–term forecasting will depend on the particular time‑period of the forecast and the scale and life cycle of the weather phenomena that are expected in the forecast period. For example, in a blizzard situation persistence might be quite satisfactory for the next few hours to a day or so if a major low–pressure system has forced the blizzard and is slow moving. On the other hand, the onset or cessation of blizzard conditions may be more difficult to predict. Nevertheless, assuming familiarity with the forecast area and a knowledge of, or ready access to, local climatology, there are general principles that should be taken in a forecasting shift in order to develop a solid forecasting methodology. These principles are enunciated in a sequence of major steps that are recommended by Australian Bureau of Meteorology (1984), and outlined below and elaborated on in the subsequent sections.

· Step 1: Get a thorough briefing from the forecaster going off–duty (if there is one) or by examining observations/existing analyses/existing prognoses;

· Step 2: Plan task times and deadlines;

· Step 3: Do a comprehensive analysis of surface and upper–air conditions;

· Step 4: Formulate a model to explain current observations;

· Step 5: Make predictions;

· Step 6: Maintain a weather watch;

· Step 7: Repeat steps 3–7 if the forecast strategy is not working;

· Step 8: Brief the next forecaster thoroughly.

The mode of application of these steps will vary from place to place depending on such things as the specific work output requirements of the shift, observational and data processing facilities available, etc. Time will be an ever–present constraint and so the efficient forecaster will learn to take a number of steps simultaneously and will be able to judge just how much time to spend on each step in various situations.

6.1.3.1 A thorough briefing

A thorough handover/takeover briefing is needed at the start and end of each shift in order to:

· maintain continuity of short–term forecasting services;

· provide the new shift forecaster with sufficient background to act immediately to revise a forecast or to issue a new forecast if required;

· convey the degree of confidence held in the current forecast strategy;

· alert the new shift forecaster to any current problems with observing systems, communications systems, etc.; and

· enable the forecaster to properly plan the task times and deadlines for the forthcoming shift.

The maximum time available for briefing will vary from forecast office to forecast office but is normally a minimum of about 15 minutes. It is important that the briefing be done systematically to ensure that the key pieces of information are passed on and understood. In situations where offices do not provide 24–hour coverage or where the forecasts are undertaken by a single person the forecaster may have to brief him/herself at the start of a shift. In either situation, the briefing principles are the same, namely:

· Immediately identify the critical operational factors that will affect the provision of short–term forecasting services during the shift. Critical factors include:

(i) any warnings that are current or imminent;

(ii) any adverse weather, current or imminent;

(iii) any unusual deficiencies or variations in the observing system (e.g. apparent station dew–point errors, problems with autographic instruments, special observations being received, malfunction of satellite receiving equipment, special satellite picture enhancements in operation);

(iv) any problems with telecommunications (e.g. inward traffic delays, telephone system malfunctions).

· Review the meteorological situation of the past six hours or so with emphasis on the trend in local observations. Identify critical or unusual aspects of local autographic records, hourly aerodrome weather reports, radar, satellite data etc. Try to relate cloud and other observations to the latest radiosonde/upper wind data. Relate significant changes in the evolutionary weather pattern quite clearly to features analysed on various charts. For example, cloud may be evident over a coast where the winds have become onshore, or a coastal polynya may have formed as the wind has changed to off–shore.

· Briefly outline the broad–scale scene starting with analyses at T–24 hours and extending through to prognoses out to T+12 hours. If available, sequential image display sequences (e.g. movie loops or video) of satellite data are an ideal briefing tool. Offices should also have a convenient fixed sequence wall display of imagery and analyses specifically for briefing purposes. The upper–air pattern exerts a strong control and the significant features must be identified.

· Clearly state the short–term forecast strategy and the degree of confidence in it. In particular try to identify:

(i) the observations (location, time, type) and the specific analysis tools that are expected to be critical in the next six hours (these will vary according to the situation);

(ii) the expected evolutionary pattern (movement, development, etc.) of key components of the analysis pattern;

(iii) the expected influence of diurnal changes and local orographic effects;

(iv) the 'safety factor' built into the forecast that accounts for likely scope for error;

(v) any known special user requirement that might cause you to give particular attention to some aspect of the forecast;

(vi) there should be an alternative strategy that could be applied if the forecast starts to go wrong, particularly if confidence in the initial forecast is not high.

6.1.3.2 Task times and deadlines

All forecasting offices have a work schedule that sets out the deadline times for issue of forecasts. Sometimes these schedules also suggest optimum times for analysis, briefing, etc., although these are hard to fix because of the variable nature of the forecast request work–load from day–to–day. However, the office work schedules do not necessarily allow adequate time for forecast preparation and certainly do not cater for variations in work–load caused by such things as prolonged bad weather, observations/communication problems, or abnormal volumes of ad hoc requests. Therefore, to maximise the time available for scientific input to forecast preparation the forecaster must establish his/her own set of task times and deadlines at the start of each and every shift.

A critical factor is the actual and expected weather, details of which the forecaster will have obtained from the handover/takeover briefing. If the weather is good, and confidently expected to remain so, then forecast messages tend to be short and preparation time is minimal. In these situations forecasts can confidently be prepared well ahead of schedule (to spread the work–load more evenly over the shift) and the forecaster can devote more time to undertaking careful re–analysis of earlier data, to sharpen analysis skills and as an added check on the forecast.

The approach will be quite different in bad weather situations. Continuous monitoring of some weather elements will be essential and the forecaster will have to allocate extra analysis time according to the observational programme and facilities available. Forecast messages will be longer and more complicated, the detail of the forecasts will be more critical, and forecast accuracy confidence will tend to be lower than in good weather situations. In these situations it is very important to do a thorough analysis of all the data at strategic times rather than to become swamped by the mechanical aspects of physically writing or typing the forecasts and merely scanning incoming data.

In planning a 'bad weather' shift, try to set aside, in advance, at least one time–period for thorough analysis. During this period you will be able to formulate a good model of current atmospheric behaviour and develop a new forecast strategy (or extend the existing strategy). The actual preparation of a large number of final forecasts can then be done quite quickly.

Try to determine in advance precisely what data and what data sources are going to be the most critical in this weather situation and note the data receipt times so that you can schedule your other tasks around them. Set your own data input deadlines and remember that you can't afford to delay issue of a forecast because a critical observation has not arrived, otherwise you will get further and further behind in your work schedule to the detriment of your customers and perhaps also to your forecast accuracy. You have to do the very best with the data you already have, and to do justice to the data you must allow yourself time to analyse it. There is always scope for urgent amendment if some late data indicates that the forecast is seriously in error.

It is also essential for the forecaster to develop an understanding of how long certain forecasts take to prepare, what amendment/update criteria are specified, the time it takes for forecasts/amendments to reach the customer, etc. Incorporation of this information is important in developing a practical and realistic work schedule.

It is not possible to specify rigid time–scales for analysis and forecast preparation. Some forecasters work more rapidly than others; some situations are more complex and involve more data than others. However, each individual forecaster should be aware of his/her capabilities and allow for them in planning task times and deadlines.

In summarising the results of a three–year experimental programme in short–range forecasting in the USA, Scofield and Weiss (1972) found that, with all the available data at hand, operational meteorologists needed 10 to 15 minutes to integrate the information and to prepare and disseminate the short–term forecast. The more complex weather situations (e.g. an approaching squall line) usually involved more preparation time but, in almost all cases in the trial period, the forecasts were prepared and issued in less than 30 minutes.

6.1.3.3 Comprehensive analysis

Because of time constraints on data access and assimilation, the short–term forecaster has to be quite ruthless in selecting both the type of data and the extent of the geographical area to be analysed. All analysis must be directly relevant to the short–term forecasting problem otherwise valuable time that could be used in other facets of short–term forecasting will be lost. Nevertheless, the analysis that is done must be as thorough and as comprehensive as possible because extrapolation of existing weather trends is such an important input for the first hour or so of a short–term forecast.

Specific analysis techniques for use in the forecasting context are described in detail in Chapter 5 and will not be discussed here. However, at this stage it is valuable to summarise some of the principles involved.

The short–term forecaster's main concern should be 'primary analysis' of raw local data, but he/she must also incorporate what may be termed 'secondary analysis' of numerical models or of other externally prepared analyses of broader fields of data in terms of the relevance to the local area. The basic purpose of all analysis is to interpret the state of the atmosphere as a function of space and time so that the causes of current weather can be explained and thus used as a basis for prediction.

A comprehensive analysis involves more than just constructing isobars and streamlines, or looking at satellite displays and surface autographic instruments. It also involves either calculations of quantitative values, (or assessment of numerical model data (see Section 4.2) that can be used in forecasting and it involves an assessment of the main physical processes that prevail at the time. The main emphasis for short term forecasting should be on determining small–scale perturbations in the large–scale flow, and in identifying transitory features and minor changes in the atmosphere. Great attention must be paid to detail, and care should be taken not to ignore, change, or smooth data that at first may seem to be in error, until such time as the forecaster is absolutely sure that an error exists. Mesoscale numerical models are powerful analytical tools, but the forecaster must be able to monitor the quality and quantity of data input to them as well as understand their data assimilation processes. Moreover, in the Antarctic context such models are only now emerging with the main source of NWP data being the global models.

6.1.3.4 Formulating a model

A natural consequence of a comprehensive analysis is the formulation of a three–dimensional 'model' of the atmosphere that explains, as far as possible, all the current weather observations and ongoing changes. A numerical model will do this automatically although the forecaster must be aware of the model's structure and constraints (see Section 4.2). In a manual analysis environment the model will tend to be conceptual rather than mathematically exact (see, for example, the discussion on frontal models in Section 5.2.1). The important thing is that the forecaster must develop some sort of dynamic four–dimensional picture of the atmosphere, at least in his/her mind, so that it can be used as a basis for prediction. There are inherent dangers in blithely extrapolating parameters because of the complex nature of atmospheric processes.

There are of course many generalised physical models available of atmospheric structure that can be used as a basis for formulating a conceptual model to suit the situation at hand in mid latitudes. In the Antarctic context such models are less common. However, during their training course, or in subsequent reading of the literature, good forecasters will have become familiar with the various frontal models; models for “air–stream weather” with associated orographic interactive effects; models of katabatic winds and their interaction with synoptic systems; mesoscale lows, etc. Knowledge of these is important, but for short–term forecasting applications these preconceived models provide a general guide only, because it is most unlikely in the situation at hand that the atmospheric structure will match them exactly. Therefore, the forecaster has to be flexible and adapt preconceived models or develop new ones in situ.

A general approach to formulating a real–time model for short–term forecasting applications is:

· Develop an 'overview' of the atmosphere within about 500 km of your station or for the area for which you have to produce a forecast;

· Develop a mental picture of the mechanics of the airflow and in particular the main areas and causes of vertical motion (subsidence as well as ascent);

· Ensure you have a good idea of the vertical structure of any discontinuities and pay special attention to the three–dimensional structure of wind maxima (make them 'tubes' not 'flat arrows') and thermal and/or moisture concentrations;

· Focus on the mesoscale aspects likely to affect your station in the next six hours. In some situations a self–contained conceptual model of a mesoscale feature (e.g. a local katabatic wind) may dominate for a few hours, but don't lose sight of how this feature fits in, and interacts with, the surrounding air masses.

6.1.3.5 Making predictions

There are two stages in making predictions. The first involves the scientific prognosis of the weather, that is, preparing the forecast. The second stage may be described as 'forecast message formulation' and involves putting the results of the first stage in a form that the customer can understand.

Preparing the forecast

Techniques for the scientific prognosis of individual weather elements are described in detail in Section 6.6 but a general approach to the problem is summarised below.

· (i) Extrapolate existing trends in the observed weather, making sure to account for evidence of ongoing changes as distinct from past changes;

· (ii) Superimpose diurnal variations of the main meteorological elements and account for their interaction;

· (iii) Allow for broad–scale interaction;

· (iv) Take account of orographic and local features;

· (v) Apply any available objective aids or numerical model output for prediction of specific weather elements and adjust the forecast accordingly;

· (vi) Check final forecast for continuity and realism.

Forecast message formulation

This is a very important aspect and one that requires considerable skill. It is the vital cog in the total process of weather forecasting, the final link between the scientist and the customer.

Ideally the forecaster should convey his/her thoughts on the expected short–term weather directly, via personal or mass briefing. In this way the complete picture can be given, including suitability of the database and the range and success probabilities of alternative predictions. In many instances however, the forecaster must convey his/her predictions to the user via a single, fairly brief message.

· For some users, the forecast must be coded and the meaning of the expected weather is defined uniquely by the code. For other users the forecast has the form of a narrative statement, which could give rise to different interpretations by different users. Some users receive their forecast in hard–copy form that can be examined in detail and repeatedly referred to as the need arises. Other users receive their forecast verbally (e.g. via radio) and may only have one opportunity to absorb all the detail and significance of the forecast message.

· Whatever the case there are a number of general guidelines that the short–term forecaster should follow in preparing the forecast message:

-The forecast must be tailored to give the user the best possible meteorological basis for planning and decision–making. Even when standard forecast formats are prescribed there is usually an optimum user–oriented way of expressing the forecast.

-The forecast message must be as clear and concise as possible. Prime emphasis should be placed on hazardous weather or weather that has special significance for the customer, e.g. aircraft icing. State these things first (unless a rigid prescribed format prevents you) because that is what the customer will tune in to.

-Don't try to exceed reasonable limits of predictability, particularly in attempting to predict brief temporary improvements during bad weather situations, or the precise time of onset or cessation of hazardous weather. In such situations precise detailed forecasts have to be based on solid evidence, reliable techniques and very low probabilities of alternative results. (A simple forecast of "intermittent snow showers 0300–0600 UTC" may be far more effective than a four–line narrative that attempts to specify the precise characteristics and timing of every temporary fluctuation.)

-Don't be too vague or make too many qualifying statements. This not only makes the message hard to understand but can also reduce user confidence in the product. Many customers are interested only in the forecaster's opinion of the most likely weather, although some customers specifically ask for probabilities or other expressions of the confidence held in the forecast.

-If probability estimates are required make sure that they are realistic. Some guidance on preparation of probability forecasts is provided by USA Air Weather Service (1978).

-When the forecast message has been formulated, review it quickly before issue. This is a very important step that minimises the chances of non–technical errors that can occur when deadlines are tight. Always check the message to ensure that you have actually said what you intended to say. -Check that you have used the correct date, time, etc. If the message indicates a major change in forecast conditions from the previous forecast, check that you have acknowledged this in some form or other. This checking process should not take very long and it is certainly time well invested. Careless errors in final message formulation can ruin the best scientific input to the forecast.

-Don't stray into areas outside weather forecasting: for example, if preparing forecasts for aviation do not preempt the decisions or air operators reasons whether conditions are suitable for flying.

6.1.3.6 Maintaining a weather watch

Although this is shown as the last step in a systematic approach to shift–work forecasting, the maintenance of a weather watch is a continuous part of short–term forecasting. The meteorological situation must be monitored continuously so that the forecaster can take rapid action in response to signs of unexpected ongoing change.

The way in which a weather watch is maintained depends on the prevailing situation. If the weather is good and expected to remain so, the aim of the weather watch is to monitor a few key elements to confirm that no significant change is imminent. If a change from good to bad weather is expected, then the weather watch should be aimed primarily at detecting and confirming the approach or development of the bad weather. If the weather is bad, then the weather watch should be aimed mainly at monitoring the size, intensity and movement of the bad weather area. In all situations however, the basic objective is to avoid being surprised by new short–term developments.

An effective weather watch involves some or all of the following:

· Monitoring trends in visual observations. The main emphasis here should be on monitoring local station conditions but, if possible, a close check should be kept on reports from other stations in the area and from field parties. A lot of things about ongoing changes in the atmosphere can be inferred from visual observations, and often clues to changes are visually evident before they show up in instrumented observations. Changes in cloud types, cloud movement, or growth rates of individual clouds give an immediate indication of the net result of atmospheric processes that are occurring. (These changes may not be evident from instrumented observations until the next radiosonde or upper–wind flight, which may not be for six hours or so.) The significance of local conditions (e.g. poor visibility in a particular quadrant, persistent visible inversion layer, formation of rime etc.) should be taken into account. These things are not necessarily mentioned in conventional weather reports so the forecaster should work closely with the observer to ensure that a comprehensive observational watch is kept. Tell your observer what to watch out for in critical situations. However, care should be taken in interpretation of the information if the person making the report is not a trained observer.

· Monitoring trends in instrumented observations. All instrumented observations, their trends and inter–comparisons, should be noted. A close watch should be kept for any kind of discontinuities and general attention should be paid to factors such as:

-Changes in surface dew–point depression. Do these reflect local or broad–scale advection of dry/moist air, or are they indicative of vertical mixing processes?

-Surface temperature and dew–point temperature. Are the current values the same as have been forecast? If current trends are extrapolated, will they favour or inhibit the development of hazards such as fog? What inferences can be made about the onset of convection, cloud base etc.?

-The anemograph trace. Is the wind direction changing? It is consistent with the expected direction from the synoptic pattern? Does it contain a strong local component? Is the range of gustiness usual? Are the peak gust speeds similar to the 300 m or 600 m (~1,000–2,000 ft) winds?

-The barograph trace. Is the pressure change indicative of the normal diurnal variation or does it reflect some mesoscale or synoptic scale effects (e.g. an approaching or departing trough or mesocyclone)?

-Cloud ceiling indication. Is the ceiling changing significantly? How does this relate to existing precipitation and the vertical wind, temperature and humidity profiles?

-Vertical wind and temperature profiles. Do the observed profiles properly reflect the existing local weather? (If not they should be subjectively updated.) How does the observed movement of the clouds compare with the upper winds? How do the existing or updated profiles compare with the prognostic profiles? Has vertical mixing occurred in the low levels? Is it imminent? Is excessive low–level wind shear likely? Have stability indices changed? (Note: A thorough air–mass analysis should be performed in preference to the mere extraction of stability indices.)

-Monitoring diurnal variations. The normal diurnal variation of all the weather elements at stations in your area should be known and should be readily available in either tabular or graphical form. In monitoring each element therefore, the forecaster should note any departures from the normal diurnal trend and try to determine their cause. A plot of a time–section for various observations allows trends to be more easily monitored. It is useful to have the normal diurnal variation curves permanently plotted and to plot the current observations over them on a transparent overlay for easy comparison. This method also allows for convenient extrapolation of current trends.

-Monitoring over a larger area. AIREPS and pilot debriefings provide valuable information on the current state of the atmosphere outside the visual range of ground stations, but the primary monitoring tools are satellites. These data are very useful for monitoring broad–scale interactions and smaller scale features that may be moving into the forecast area. In analysing satellite pictures of the local area, concentrate your attention on small–scale changes in cloud structure, cloud–free areas etc. Specific satellite data techniques are discussed in Section 4.3. Facilities such as interactive graphics (for example the Man computer Interactive Data Access System (McIDAS) make the monitoring task much easier and more effective and permit superimposition of satellite, wind data, etc.

-Monitoring the synoptic situation. The key things to remember in monitoring the synoptic situation are:

(i) the short–term impact of each new observation should be assessed as it comes to hand. Do not wait till all the observations are plotted before looking at the chart;

(ii) pressure tendencies are very useful in determining short–term changes in the orientation of isobars;

(iii) keep informed on changes to the broad–scale situation.

6.1.3.7 Concluding remarks

The approach to short–term forecasting outlined in this section has been deliberately aimed at encouraging the forecaster to develop a systematic method of working; a method that has a solid scientific basis but one which can also be applied in the real–time situation where there are heavy work loads, severe time constraints, and where the latest most sophisticated technology is not necessarily available. Although new technology is being introduced into many operational areas, and in particular into the Antarctic forecast centres, with better forecasting methods being developed, and mesoscale numerical modelling will soon be an operational reality for some Antarctic operations, the general principles discussed in this chapter should still hold firm – but they should become easier to apply in practice.

An important practical point emphasised in this section is that effective short–term forecasting requires more than just the application of scientific and technical skill. Certainly scientific competence is the foundation for success, but to produce effective results in the short–term forecasting situation the forecaster also has to be a skilled manager. Rapid but accurate response is required in a situation where large volumes and different types of data may have to be assimilated and where uncertainty factors may be high. To cope with this the forecaster has to apply sound management principles to plan and organise his/her shift, direct and control the data flow and forecasting aids required to suit the situation, and to constantly review the quality of the final product.

Finally, every short–term forecaster should remember that his/her objective is to produce effective forecasts that give users the information they need, in a form they can understand, and in sufficient time for them to act upon it. As Golden et al. (1978) point out "... the best observations and forecasts are worthless unless someone benefits from them."' We are not engaged in an academic exercise in which perfection is the sole objective and success is measured only in terms of statistical parameters. We are providing a real–time service in which the measure of our success is the effectiveness of our forecasts from the user viewpoint.

6 THE FORECASTING PROCESS

6.1 The systematic approach to forecasting

6.1.1 Getting to know the physical environment of the area for which forecasts are being prepared