Note:
The information on this page is being revised.
We hope to update this page by the end of 2005.
Meanwhile, for information on probability to categorical outlook conversion,
please see table in HTML,
MSWord or
PDF formats.
The SPC produces probabilistic Convective Outlooks
in conjunction with the traditional categorical Convective Outlooks. These
outlooks are done for all Day 1,
Day 2, and Day 3 periods.
Categorical Convective Outlooks
The traditional Convective Outlook is a categorical forecast that specifies
the perceived level of threat via the descriptive wording: Slight Risk,
Moderate Risk, and High Risk. This structure does not give the
forecaster's expectations of the threats of individual severe weather
hazards (large hail, damaging winds, and tornadoes). While the accompanying
discussion for the outlook may describe the forecaster's thoughts about the
individual hazards, the accompanying graphic does not explicitly depict these
expectations. Further, even though the SPC may have
strict definitions of
what a Slight, Moderate or High Risk means, this definition may easily be
confused, or even replaced, by the users of the product. A more thorough
description of the conventional Convective Outlooks may be
found at this location.
Example Day 1 categorical Convective Outlook
The graphic clearly defines the geographic threat areas. A large Moderate Risk
is forecast for the Central Plains. The Moderate Risk is surrounded by a
Slight Risk area from Texas through Wisconsin. The graphic does not provide
much information concerning forecaster expectations of the individual severe
weather hazards which the forecaster may have. This outlook will
be discussed later in conjunction with the probabilistic forecasts
for this event.
Probabilistic Convective Outlooks
Forecasting rare events such as tornadoes and the occurrence of large hail and
damaging wind gusts is a very difficult process and one that contains a large
amount of uncertainty. In the traditional Convective Outlooks this
uncertainty is conveyed via the Slight/Moderate/High Risk terminology. A more
direct method of expressing the forecaster's uncertainty is to use
probabilities. Probabilities directly express a level of confidence that an
event will or won't occur. While probabilities may seem somewhat difficult to understand
at first, once you have a grasp of how to interpret them you will
quickly gain an appreciation for how much more information they provide.
(A great introduction to why probabilistic forecasting is so useful may be
found in an
online essay by
Chuck Doswell
and
Harold Brooks.)
Definition of the probabilities
The probabilities used in the SPC Convective Outlooks are what are known as
subjective probabilities. The forecasters make their best estimate of
the probability of an event occurring. They may have guidance
concerning what probabilities are appropriate but the values
they choose are not created automatically by a computer or via statistics.
The probabilities that you see on the images represent the probability of
one or more events occurring within 25 miles of any point during the
outlook period.
This is done because the probability of severe weather at an given point is
quite small and because the Convective Outlook is not a smallscale, shortterm
forecast but one that covers the entire U.S. for periods up to 24 hours in
length. There is a large amount of uncertainty in forecasting severe weather
on these scales. How many times have you experienced a tornado
in your neighborhood? Not very many times, if ever! Now think of how
many times you can think of severe weather occurring within 25 miles of your
location. It's probably safe to say that you can think of some examples. How
large of an area is a circle with a 25 mile radius? Below you'll see an
example of the Oklahoma City metro area where the large blue circle
represents such an area.
You should be able to imagine that the probability of having severe weather
occur within such an area is much larger than the probability of having it
occur specifically within your neighborhood. Keep this in mind as we head
into the discussion of the probability values that you may see expressed
in the outlooks.
So, how do you interpret the forecast values?
For example, a 15%
contour on the hail forecast outlines an area within which the probability
of 1 or more reports of large (3/4" or greater) hail occurring within 25
miles of any point in that area during the forecast period is 15%.
Larger values imply greater risk. For example,
if you find that the probability for large hail in your area is 35% on
a given day and it was 25% the day before, you can tell that the SPC believes
there is a higher threat of large hail for your area on that day.
You may be thinking that 35% is not a very large probability of having 1 or
more of these severe weather hazards occur near you on a given day, but 35%
is roughly a 1 in 3 chance. To better understand this, let's put the
probability values into perspective in terms of climatology.
The climatology of severe weather is very different than the climatology of
precipitation. Compare the number of days that you have experienced rain at
your home to the number of days that you have had large hail at your location,
or even a tornado. The number of times a given location experiences severe
weather in a year is much less frequent than the amount of time it experiences
rainfall. Because rainfall occurs so frequently (on average) the daily
climatological values approach 20% on any given day in many locations (i.e., a
1 in 5 chance, or it rains 1 day out of 5 on average).
Suppose you hear a forecast calling for a 40% chance of rain.
Given your knowledge of climatology, you can immediately say that the
forecaster believes the chance of rain is twice as high as normal (40%/20%).
This does not mean that rain will definitely occur but does mean that the
forecaster believes that there is a higher than normal risk of
precipitation
occurring on that day. A knowledge of the climatology of the event being
forecast is useful, even necessary, in order to interpret the probabilities
being forecast. In the case of probability of precipitation forecasts, these
values typically run from 0% (certainty that it will not rain) to 100% (which
implies certainty that it will).
As stated previously, the probabilities of severe weather occurring at any
given location are much lower than those for precipitation. How much lower?
The following image shows the probability of 1 or more tornadoes occurring
within 25 miles of a point for the week of April 29May6.
The image shows that the probabilities for this week range from 0 to 1.5%.
These are very small values! Climatological values of rare
events such as severe weather are much, much smaller than the climatology of
cloudy days, or the probability of precipitation occurring.
As a part of the probabilistic forecasting program at the SPC, a
representative severe weather climatology has been developed by members of
the National Severe Storms Laboratory
(NSSL) and the SPC for use by the
SPC, the emergency management community, and the general public. This
project is
available on the NSSL web site.
You can find a tremendous amount of information there to assist yourself in
understanding the severe weather climatology for your area.
Because severe weather occurs relatively infrequently, there is a large
amount of uncertainty as to precisely where it will occur. Accurate Yes/no forecasts
of whether or not you will experience a tornado in your neighborhood in the
next 24 hours are simply not possible many hours ahead of time. Further, the
role of the Convective Outlook is not to do such pinpoint forecasts. The
product is a nationalscale forecast that highlights
areas where severe weather is possible over the lower 48 states. The product
is simply not meant to
be a localized, shortterm forecast of severe weather for your neighborhood.
Because of this and the fact that the climatological probabilities of severe
weather are so small, the probabilities that you will see used in the forecasts
will generally be much smaller than you might expect.
The following table shows the probabilities that you may see in the various
probabilistic outlooks:
Day 1 
Tornadoes  2%, 5%, 15%, 25%, 35%, 45% 
Large Hail  5%, 15%, 25%, 35%, 45% 
Damaging Wind  5%, 15%, 25%, 35%, 45% 
Day 2 
Any severe weather  5%, 15%, 25%, 35%, 45% 
Day 3 
Any severe weather  5%, 15%, 25% 
These values represent the only probabilities that you will see in the SPC
forecasts. How should you interpret them? Very simply put, the smallest values
represent areas where the most uncertainty exists and correspondingly where
the smallest expected coverage of storm reports exists. The higher the
probabilities, the greater the perceived threat and the greater the expected
coverage of that hazard being forecast. The highest probabilities are generally
reserved for the more significant severe weather events and may be used very
infrequently, if at all, during the year.
Another way of thinking of the values
is related to climatology. Consider our earlier discussion of tornado
probabilities for the first week of May where the peak values were
approximately 1.5%. Let's assume that the SPC forecaster drew a 35% area
which included northwest Texas and southwestern Oklahoma. The ratio of the
forecast to climatology (35%/1.5%) yields a value of approximately 23. The
SPC forecaster is stating that they believe the risk of tornadoes in that
region is more than 20 times larger than climatology. By
comparing the forecast probability to climatology, you can more easily
understand the magnitude of the risk on a given day.
Description of the probabilistic outlooks
Day 1
The most specific Convective Outlooks are those issued during the Day
1 period. Accordingly, the SPC forecasters have the most information available
to them to differentiate the threats of the individual severe weather hazards.
During this period, the SPC produces probabilistic outlooks for each primary
severe weather hazard (tornadoes,
damaging wind, and
large hail) separately. By producing
separate forecasts for tornadoes, damaging wind, and large hail, the user is
given substantially more information upon which to make decisions than in the
traditional outlook. In addition to the probabilities for separate types of
severe weather
occurring, areas are shown where there is a 10% or greater chance of significant severe
weather occurring. Significant severe weather is defined as
F2 or greater
tornadoes, damaging winds with speeds greater than 65 knots, or large hail 2"
or greater in diameter. If the forecaster believes that there is less
than a 10% chance of significant severe weather occurring in the outlook area,
then that threat will not appear on the graphics.
Day 2
Probabilistic Outlooks are issued for the Day 2 period as well. Because many of
the specific details of severe weather forecasting can only be known hours
ahead of time, rather than several days, the severe weather probabilities for
the Day 2 Outlooks represent the probability of any severe weather hazard
(large hail, damaging wind, or tornadoes) occurring (rather than producing
individual forecasts for each hazard). Areas where there is a 10% or greater
probability of significant severe
weather events (again, defined as 2" or larger hail, 65 knot or stronger
winds, and F2
or stronger tornadoes) are also indicated on the graphics.
Day 3
On November 7, 2001, the SPC began issuing Convective Outlooks
for the Day 3 period to the public. These outlooks will be similar to the
Day 2 outlooks in that the probabilities represent the probabilities of
any type of severe weather hazard (tornadoes, large hail, damaging wind)
within 25 miles of any point. Because of the large and increasing amount
of uncertainty forecasting severe weather 3 days ahead of time no attempt
is made to forecast areas of significant severe weather hazards like is
done for the Day 2 period.
Example Day 1 probabilistic Convective Outlook along with the corresponding
Categorical Outlook
click to view the full image
This image shows the categorical Convective Outlook issued at 2000 UTC on
May 16, 1999 as well as the corresponding probabilistic forecasts valid for the
same time period. The Bottom right panel shows the conventional outlook
which depicts a large Moderate Risk area (shown in yellow) for portions
of the southern and central Plains. A Slight Risk (green lines) surrounds
the Moderate Risk area and extends from Texas to Wisconsin.
The unlabeled brown lines represent regions where general thunderstorms
are forecast. An emergency manager, or storm
spotter, or member of the general public, may use this graphic to determine the
relative level of threat for their area. However, forecaster expectations of
locations of tornadoes, large hail, and damaging winds are not provided.
The remaining 3 panels of the figure show the forecast probabilities of
hail (upper right), damaging wind (lower left), and tornadoes (lower right).
Probabilities shown are 2% (green), 5% (brown), 15% (yellow), and 25% (red).
The use of
similar colors for 15% for probabilities and also Slight Risk is purely coincidental and is not meant to imply that the two are equivalent. Light blue hatched
areas represent areas where a 10% or greater chance of significant severe
weather (as defined above) is forecast. As you can see, the forecaster has
identified different areas for large hail, damaging wind, and tornadoes. While
one may have thought that the greatest risk of severe weather was forecast for
the center of the Moderate Risk area, the probabilistic forecasts show that
the greatest threats (highest probabilities) of large hail and tornadoes are
in western Iowa and northern Missouri while two separate areas of damaging winds as identified by 15% contours are found in Oklahoma, and across Iowa, northern
Illinois and southern Wisconsin.
Benefits of the probabilistic Convective Outlooks
We believe the new Probabilistic Convective Outlooks issued by the SPC do a better
job of expressing uncertainty, as well as detail, compared to the traditional
Convective Outlooks. These new outlooks directly express forecaster uncertainty
through the use of probabilities. Further, in the Day 1 period, forecaster
expectations of large hail, damaging winds, and tornadoes are explicitly
conveyed through individual forecasts. By producing forecasts of each hazard
individually, users who are sensitive to one particular threat (e.g., car
dealers and large hail) can make more informed decisions.
Even without a complete understanding of what
the probabilities mean, you can directly assess from the graphics:
 Geographic areas where the various severe weather hazards are expected. These areas may or may not overlap with one another.
 The perceived levels of threat for the severe weather hazards. The higher
the probabilities are, the increased threat of that hazard occurring. Refer
to the discussion above concerning the probabilities used in the outlooks
and especially the range of probabilities used.
 Areas where significant severe weather is expected.
Comments/suggestions?
Richard Thompson
