This series of Forecasting Hints looks at 10 difficult to forecast events or situations. Difficult to forecast means there is not a definite reliability of the forecast and the reliability is not as high as would be desired. The following topics will be covered:

6. FOG


A specific event narrows it down to a select time and place. For example, meeting for lunch at noon on a certain day at a particular restaurant is a specific event. It has been narrowed down to a specific time and place. In meteorology, this specificity is more generalized. For example, a forecast that gives a 70% chance of thunderstorms during the afternoon within a forecast city. The time frame is not for a specific time but rather covers a range of hours. A forecast city also will cover an area thus a forecast has to be made for citizens scattered across an area. The percentage is introduced since an event may or may not happen at the city within the given amount of time. Uncertainty is present in weather forecasts. Also, forecast reliability decreases with time. For example, a forecast for 5 days out is typically less reliable than a forecast for the next day. This occurs since small changes and small size phenomena are more likely to influence observed weather events as time advances (Butterfly Effect). It is more difficult to analyze phenomena as the size gets smaller, thus it is difficult to know how the extreme multitude a tiny phenomena will impact observed weather as time moves forward.

Why is it difficult to give a specific forecast? For example, a forecast for thunderstorms between 1:30 pm and 2:30 pm that is forecasted a day in advance. The reason is due to the enormous complexity of factors involved. Weather is influenced from all scales of motion and the properties at all these scales of motion cannot be analyzed. Enough information is gathered that a good general idea can be obtained for the expected weather. The amount of weather information gathered is limited by cost and feasibility. Since this gathered weather information is what is used to make the forecast, limitations on data gathering will limit forecasting ability. The ability to pinpoint exactly where a tornado will occur 3 hours in advance for example is not feasible. Too many factors limit being able to make this type of prediction. The good news is that forecasts will improve through greater computer power and data collection techniques. This improvement though will suffer from limitations on how much it can improve.


Hurricanes bring significant destruction to the coastline that they impact. The impact is much more significant over and near where the eye makes landfall. Much of the coast though is spared the worst of the destruction. More of the coast will be at a threat before landfall due to the uncertainty of where exactly the hurricane will make landfall. This writing examines the reasons for the difficulty of predicting hurricane landfall.

Uncertainty will be higher when the hurricane is a couple of days or more from making landfall. Generally a broader region of the coast will be in the threat area from the hurricane when it is far from land. As the hurricane approaches the coast, the region of threat is typically better and better defined. This is especially true for faster moving hurricanes that have well defined upper level steering mechanisms.

What are some reasons why it is difficult to forecast landfall location? Besides a forecast being several days out, another reason is due to the shifting position of the eye. Eyewall replacement cycles can cause the center of the circulation to drift one direction or another. This can make it appear the hurricane is changing directions but often this shift is temporary. These shifts in direction though can cause the path to alter from what was expected. The convective storm dynamics in the eyewall can cause the center of circulation to be pulled one direction or another. On satellite, it can be noticed that the eye does seem to shift somewhat from the prevailing tendency that the eye has from time to time. Eyewall replacement cycles and storm dynamics in the eye wall play a role in this shift. Another reason for a change in hurricane track is due to changes in the steering currents that can also be difficult to forecast. The steering currents can be weak and subtle and thus any change in this steering will cause the hurricane track to change. Steering is influenced by upper level winds associated with troughs and ridges. If the steering is weak then often the hurricane will move slowly and it can be more difficult to anticipate how the steering will influence the storm in the future. A third reason can be due to a significant strengthening or weakening of the hurricane. When the hurricane changes in intensity then this is often accompanied by some sort of path shift. Interactions with land/islands, a very warm area of sea surface temperature, and upper level wind shear can cause the path to shift from the previous path. All these reasons cause it to be more difficult to forecast where a hurricane will make landfall.


Winter weather can be very difficult to forecast in the south within the United States. Examples of states include Texas, Arkansas, Mississippi, Alabama, Georgia and several others. Of course, this is a generalization since winter weather as a whole can be difficult to forecast anywhere. However, there are some reasons why it is particularly difficult to forecast winter weather in the south. This writing will go over these reasons.

One reason is due to the forecast having to cover a range of different precipitation types. In the north, the precipitation type is often snow in the winter while in the south the precipitation type is often rain. This is another generalization. However, in the south there are often difficult situations in which the precipitation type could either be a cold rain, snow, sleet, freezing rain, or a combination of 2 or more of these. It is difficult enough to forecast if precipitation will occur, predicting the precipitation type adds another level of complexity to the forecast.

Another reason is because winter precipitation is less common in the south. Forecasters in the north often have more experience with forecasting snow and other winter weather types. Going further south, these experiences are less. This is another generalization since a forecaster in the south can be great with winter weather forecasts. On the whole though, having less opportunity to forecast winter precipitation types for the local area can make it more difficult to forecast winter weather when it does happen.

A third reason is due to public attention. Winter weather forecasts for snow and ice often get more attention in the south than in the north. Big winter storms can be attention getters in the north also but they can be especially attention getting in the south, even when small accumulations are expected. Reasons for this include winter precipitation being less common, less road preparations for winter weather and less overall preparation people have in general. Thus, in the south, the forecast will be monitored very closely. With such a strong scrutiny, any forecast mistakes on precipitation occurrence, precipitation type and precipitation amount will be strongly noticed.

Finally, these forecasts are very sensitive to slight changes in the weather analysis. A small change in temperature can mean the difference between snow, rain and another precipitation type. A slight change in the low pressure track will change the areas that are expected to receive a certain precipitation types. A slight change in lifting can mean the difference between precipitation that produces icing and no precipitation. This sensitivity to many weather variables makes it difficult to pin point a winter weather precipitation forecast in the south.


Often forecasts are made for a specific location but the residence of the city and surrounding area are scattered over a wide ranging spatial scale. Thus, the conditions and future weather for the downtown or airport of a city can be very different compared to nearby cities. This is important since residence in nearby cities often base what they think will happen at their house based on the forecast for the downtown area or airport. Variations in weather over short distances occur since weather, especially wild weather, occurs over localized areas. Examples of these localized events include hail storms, tornadoes, thunderstorms, heavy snow bands, micro-bursts, and fog events. Reasons for local variation include elevation changes, vegetation changes, changes in the amount of urbanization, changes in latitude, changes in the nearness to lakes/ocean, and the small area that many weather events take place over.

Local variations can have a significant impact on the high and low temperatures and also the precipitation probability and expected accumulation. It is common for residence that live a little ways from the city to say for example “the temperature is often cooler here than the city” and “the elevation is 500 feet higher in my city thus it is often cooler here and more snow falls”. The larger a forecast area is, then the more likely a forecast is going to have problems if the same forecast is given for the entire area. Because of local variations, the forecast area will often be broken up into regions where the forecast numbers are adjusted for each region given the numerous local impacts that can influence the forecast. Two big influences are latitude and elevation. Temperatures tend to be cooler with increasing elevation in many forecast situations (although the opposite can be the case when a shallow layer of cold air pools into the lower elevations). Temperatures tend to be cooler in higher latitudes. Thus, for example, it is common for the northern part of the viewing area in a state like Oklahoma to be cooler than the southern part of the viewing area. Elevation can influence precipitation probabilities. For example, in some cases the precipitation chance will be higher closer to mountains where topographic lifting takes place. It is important to keep local variations in mind when making a forecast for an area, especially if the area is large in size. When making a forecast it can be important to emphasize the localized nature of certain weather events. This will help residence to be aware of the potential of certain weather events such as a damaging storm without expecting damage everywhere within the entire forecast area.


Forecasting snowfall amounts can be a challenge. There are several reasons for this. One reason is that it is on the condition that snow falls in the first place. This makes it doubly difficult since not only does the forecaster have to forecast if precipitation will occur but on top of that it needs to be forecasted how much will occur. A second reason is that a small difference in the precipitation amount will have a big difference in the accumulation in inches of snow. For example, a 1/10th of an inch of liquid equivalent can produce 1 inch of snow while 4/10ths of an inch of liquid equivalent can produce 4 inches of snow. For a rain forecast, this difference is not that apparent but with snowfall accumulation it is very apparent. Another reason is due to the variety of liquid equivalent amounts that can occur. 1/10th of an inch of liquid equivalent can produce anywhere from less than a half inch of snow up to several inches of snow depending on the snow to liquid equivalent that occurs. A good typical value is a 10:1 ratio which means 10 inches of snow will occur from each inch of liquid equivalent. Depending on the temperature profile, the snow can be a fluffy 20:1 ratio or a wet snow with a 5:1 ratio. Thus, it is important to forecast the temperature profile in order to determine how fluffy or dense the snow will be. Another difficulty is compaction which means the snow becomes denser each hour the snow is on the ground. This occurs when gravity and melting cause the snow to pack more densely. Thus for example, 6 inches of snow on the ground in the evening can compact to 4 inches by morning. Also, a warm ground temperature can melt snow and cause compaction. If the ground is too warm, then much of the potential accumulation can be lost to melting. The wind can add difficulty to snowfall measurements since the wind will drift and pile the snow to different depths in different places. This can make it difficult to determine a precise number of inches that has fallen. These various reasons make snowfall accumulation forecasts a challenge.


Fog is often an underrated weather hazard. At times, it can be just as dangerous as a severe storm. Fog is most dangerous to travelers. With lowered visibility, it can be much more difficult to judge where other vehicles are. This reduces the reaction time and leads to numerous accidents and accidents where many vehicles can be involved (chain reaction accidents). The ingredients for fog formation are fairly intuitive to understand (saturated air, cooling air, wet ground) but it can still be difficult to forecast if fog will develop and how dense it will be. Often fog advisories occur as the fog event is happening but it can be difficult to forecast in advance. This writing goes over reasons for the difficulty in fog forecasting.

One reason for the difficulty in fog forecasting is that it can be easy to forget about. With all the attention that is paid to high temperature, low temperature and precipitation, fog can be a forgotten weather forecast event. Often fog has to develop before it is remembered that it is an event that is important to forecast for. Fog is not sensational like a severe storm and it is not sought after like knowing the high or low temperature. Often fog is just not a priority and not at the forefront of the mind when developing a weather forecast.

Another reason in the difficulty of forecasting fog is that the density of fog is so variable across the local forecast area. Some locations can have dense fog while others have only a light fog. Fog tends to be densest in wet vegetated area, lower elevations near streams that are cooler and wetter and where emissions from factories add condensation nuclei to the air which leads to a higher density of cloud droplet formation. Fog is a cloud on the ground. Just like the sky can have thinner and thicker clouds, fog can be variable in density. Fog will tend to be thinner where the wind speed is higher, where there is more urban surface and less wet vegetation, over warmer surfaces and sometimes higher elevation surfaces when exposed to more wind and less cold air pooling. It is important for a forecaster to recognize the “fog prone” areas within the local forecast region. These regions are especially important to know about where major highways systems move through these areas. When driving, the sudden increasing density of fog can make highway travel very dangerous.

Fog can be difficult to forecast for since slight changes in weather variables can mean the difference between dense fog and very little fog. Several of these variables are explained below. These are forecast generalizations that have exceptions but often these tendencies will be noticed. These factors work together to help determine how dense fog will be:

Wind: A higher wind speed will mix the air more and this tends to reduce fog density while a light wind tends to increase fog density.

Lifting: Any amount of low level dynamic lifting will help increase fog density. This lifting will help condense moisture in the air. For example, a slight warm air advection pattern in saturated air can produce enough lifting to help develop a dense fog. Fog can also be produced from an upslope flow of saturated air (such as air flowing up a mountain slope). Since rising air cools, if it is already saturated, the upslope flow will ensure a relative humidity of 100% which aids in fog formation.

Wet soil/ground: One of the most important ingredients for dense fog is a wet ground. A previous rainfall that soaks the soil, vegetation and ground can mean the difference between dense fog at the surface and a light fog at the surface. The wet ground and soil contributes to a constant supply of moisture that can be evaporated into the air and when this combines with overnight cooling, it can cause the relative humidity to stay at 100% which increases the likelihood of dense fog.

Overnight cooling: To aid in fog development, it helps to have clear skies at night. The clear skies aids in the ground cooling rapidly at night. This cooling helps bring the temperature to the dewpoint. With a high dewpoint from a wet ground, overnight cooling can help saturate the air and this can lead to dense fog.

Precipitation: Precipitation adds moisture to the air and saturates the ground which helps raise the relative humidity toward 100%. Fog can many times occur with a sustained rainfall. With light wind and lifting, fog will many times develop.

Cool air pooling: Since cold air is denser than warmer air, the colder air will flow into the valleys. This region also tends to have more moisture due to streams and denser vegetation. Thus, cooler air pooling into valleys can lead to dense fog in these areas.

Lake/Ocean moisture input: Lakes and ocean water add a continuous supply of moisture to the air. When other factors are in place that can cause fog to develop, often fog will be produced densest over and near these moisture sources.


The general ingredients for tornadoes are intuitive to understand but forecasting the location they will occur is very difficult. Because of this, tornadoes are generally forecasted for a large area. Thus, a tornado occurring anywhere in that large area indicates confirmation of a forecast that had a tornado risk. The general ingredients for tornadoes are the interaction between speed/directional wind shear, adequate instability, low cloud base (low lifted condensation level), and lifting mechanism(s). With these ingredients in place, tornadoes can occur but where exactly they occur is very difficult to forecast in advance. The same can be said of where a severe thunderstorm will occur in general.

The forecasting of where a tornado will strike and the speed and direction of tornado movement is generally done at the formation stage of the tornado within its parent thunderstorm. This is accomplished with skilled, experienced and trained storm spotters along with using state of the art radar data that can display the precise 3-D wind field that is updated frequently within a storm. This can allow for several minutes of warning before a tornado strikes a location. Thus, when a tornado warning is issued, there is often only a short amount of time to get sheltered. The dissemination of a tornado warning issued by the National Weather Service is typically done through sirens, weather radio, television/radio, and smart phones.

The inability to pin point where a tornado will touchdown far in advance is one reason why tornadoes are so deadly. Severe storms are capable of producing tornadoes thus any severe storm should be taken seriously. Only a tiny fraction of land area experiences a tornado moving over it in any one year thus the rareness of the event causes some people to not pay as close attention to the threat. It is important to stay aware of the storm situation. Tornadoes are difficult to forecast and for this reason it is important to be able to take shelter at a moment’s notice.


Weather forecast accuracy decreases with time. The greatest accuracy occurs with a Nowcast which is giving the current weather conditions and expected weather for a short time after. The least accuracy occurs in the longer term forecasts such as 6 days out or more. Another consideration is the specifics that are asked within the forecast. Shorter term forecasts tend to focus more on specifics (i.e. this area will get storms, the high will be near 80 F). Medium term forecasts such as in the 3 to 5 day time period have more generalities (i.e. high temperature in low 80’s, low temperature in upper 60s, chance of storms). Longer term forecasts that are beyond 5 days (and especially beyond 10 days out) start trending toward a comparison to climatological values (high temperature above average, threat of precipitation less than climatological normal). Thus, shorter term forecasts allow for more specifics while longer term forecasts tend to be more general and climatologically leaning. This helps compensate for the decrease in forecast accuracy with time.

Why does forecast accuracy decrease with time? One reason is due to error magnification and analysis error magnification. Small errors in the weather analysis will lead to big errors with time. This is a reason why different forecast models can have varying outputs in the long term forecast. An analogy is time spent on reading. Think of the difference between reading for 15 minutes and day and reading 30 minutes per day. After one day, the difference is reading time is only 15 minutes. But after 10 days the difference in reading time is 150 minutes. This comes out to 2.5 hours in difference in reading time after 10 days when comparing the two. Thus, a small amount of change in the decision for what to do each day at the start results in large differences in what can be accomplished in the long term. The atmosphere works generally the same way in that a small change in the short term analysis will lead to big changes in the cumulative long term. Another reason for forecast accuracy decreasing is the inability to analysis the atmosphere at all scales and all locations within the atmosphere. It becomes impractical and too expensive to analysis the atmosphere with a very fine pattern. Only so many locations can be monitored for weather data and only a small fraction of the atmosphere can be examined with direct measuring devices. Thus, not knowing exactly what is happening between observation points introduces inaccuracies into the weather data that magnify with time. A third reason is due to scale amplification. Very small size events, such as a leaf falling from a tree and moving the air molecules around it, have basically no impact on the weather in the short term. However, that small change in air motion will influence the weather in longer time frames for the same reasons as explained in the first reason in the paragraph. For these reasons, it is safe to say that long term weather predictions that give specifics will often be incorrect. Chaos wins over certainty in long term weather forecasts.


With the advancements in technology there can be an expectation that more and greater things can be accomplished. This is generally true. This works in forecasting also. Each 10 years generally experiences an increase in forecast accuracy. The problem though is that these expectations can be taken to the extreme. This is called unrealistic expectations. An example of an unrealistic expectation is a question like, “for our outdoor event in 5 days from now, will it be raining?”. This is such an easy question to ask and seems reasonable. However, the advancements in meteorological technology have not allowed a certain answer to this question to be formulated. A reason why future predictions of weather can not be answered with a simple yes or no is because the science of prediction deals with probability and not certainty. The probability forecast is a permanent fixture of weather forecasting. It is true that more accurate forecasts can be obtained with more data and better analysis techniques but a goal of having certain predictions is a limit that can not be obtained.

As technology improves, expectations tend to increase right along with it. For example, 5 day forecasts are more accurate today than 20 years ago due to better analysis and data techniques. The increased accuracy may not be noticed by the public though since there will be an expectation for even more improved forecasts. Increasing expectation examples include forecasts with better temperature accuracy at longer time frames, a more precise time that rain will fall in the short term forecast, and not only better 10 day forecasts but the addition of a 15 day forecast. Forecasting improvements tend to be taken for granted as they happen and then an expectation for additional improvement is requested. This can cause great frustration in weather forecasters. The public thinks in terms of certainty while forecasters have to think in terms of probability. One way to explain this to the public is to use examples of sporting events, news events and political events. For example, it is not certain which team will win the Super Bowl, the value of the stock market in 5 days from now, and who the next president of the United States will be. The weather is no different than a probability that occurs with future news, the stock market, sporting events, and elections but somehow the weather is often viewed differently as if we will somehow be possible to predict weather with certainty. All predictions have to have probability introduced into the forecast to have a thorough comprehension of what may happen.


The weather can be significantly different just in the span of a couple of kilometers. This is certainly the case when it comes to storms. One location can experience heavy rain, high wind and hail while another location nearby misses the experience of these weather elements. Because of the high variability of experiences over short distances, forecasting who exactly will experience the storm conditions is not feasible. This is why weather events are often referenced in probability. For example, “there is a 40% chance of thunderstorms” and “there is a 15% chance of a tornado occurring within 25 miles of your location”. Severe weather is commonly forecasted in this probabilistic fashion. Forecasters tend to focus on a region that weather event(s) will occur.

The use of probability removes the problem of having to give a yes/no forecast, such as there will be severe weather at your house. Probability ranges from 0 to 100%. The higher the probability, then the greater the likelihood is of a particular weather event occurring. Some weather events are fairly rare at any one location but are more common over a region. This is why the probability forecasts (such as severe weather damage within 25 miles of your location was developed). For example, a tornado has only a tiny probability of striking any one location but that probability is higher if it is considered as the probability of a tornado striking anywhere within 25 miles of a location (which is about a 2,000 square mile area).