This page contains the explanation of the process used to make the seasonal forecasts for formation and landfall. As always, these are not "official" forecasts, and therefore are not to be used for planning purposes.

HurricaneAlley.net

2005 Atlantic Hurricane Forecast Methodology

Introduction
"Track Record"
Forecast Philosophy
Forecast Methodology
Variables Influences Details

Introduction

 The Hurricane Alley forecast for the tropical cyclone season in the North Atlantic, Caribbean Sea, and Gulf of Mexico is issued publicly for the sixth season in 2005.  The first 2 years the forecast was only for the number of storms that were expected to form in the North Atlantic/Caribbean/Gulf of Mexico region.  The last 4 years the forecast has included the forecasted formation and landfall zones.  The results in the first 3 years of this particular forecast have been very promising.  The forecasts were not 100 % accurate, but they were well within the range of better than just a “lucky guess”.   Those 3 forecasts and the results can be seen at the end of this document.  The greatest of thanks and appreciation is given to 2 gentlemen who have aided tremendously in this forecast.  Bryan Ray,  and Michael Bryson. 

Much has been said and written concerning the perceived inability to forecast tropical cyclone formation and movement very far in advance of 24 hours.  While there is certainly an abundance of proof to demonstrate that this is true for certain “odd” situations, it is the belief here at Hurricane Alley that the overall picture is one that is much more predictable.  This means that while there will be a storm that forms “out of season”, or loops and jumps and jogs in seemingly “crazy” ways during its lifespan, much more often than not the storms are “typical”.  The same can be said for “seasons”.  There are certainly years when there are storms that number 21 (1933) and years when they number 4 (1983), but for the most part the number of storms in any given year remains within a very narrow window.  It is for this basic reason that we believe the forecasting of how many will form, where they will form, and where they will eventually track is fairly predictable.  This is not to say that we are yet to the point that we can predict exactly which storm will make landfall, but we feel we are to the point that we can predict with reasonable accuracy that a storm or storms will have a direct impact on a particular area of coastline during the next season.  This predictive process becomes “clearer” for each month that we get closer to the season itself.  That is to say, the forecast issued in January will not be as accurate as the forecast issued in June.  But, by the same token, there is rarely a significant change during that time span.  The particular areas that are forecasted to be impacted may shift slightly east or west, north or south, but the general area will stay the same.  Again, there is rarely a significant change, although that is not to say that there is no chance that there won’t be. 

"Track Record"

The following tables demonstrate the “track record” of the Hurricane Alley forecasts since 2002.  That year was the first time that Hurricane Alley publicly presented a forecast that included forecast regions and landfall areas.

 STORM FORMATION 

^{1 (2002 forecast only for August, September, October)}

 STORM LANDFALLS
 

^{1 (2002 forecast only for August, September, October)}

Forecast Philosophy

It is our belief that the atmosphere, as complex as it is, still has general conditions that exist.  Combinations of certain aspects of the atmosphere are common, and the results of those combinations are predictable.  While certainly not 100% predictable, the results are predictable enough that the future outcomes are reasonably accurate.  The definition of reasonable varies by individual, but for the purposes of this forecast that definition is accepted as a 75% probability that the forecasted event will take place within the window of time and space being referenced.  The relationship is not a perfect 1 to 1 relationship.  Meaning that the higher or lower a certain variable is in value does not necessarily correlate to an equal or predictable increase or decrease in the resultant.  Granted, the events that we know as tropical cyclones are relatively rare.  They only occur on the average of 10 events a year in the North Atlantic region.  This makes them rare by our definition.  But, even in this rarity, we believe that there is a sequence of events preceding the formation and subsequent track of these rare events that is both determinable and thus valuable as a predictive tool.  It is upon this basis that we have studied the various measurable components of both sea and air and have determined a set of parameters that we feel offer the best opportunity to “read” and thus use as a forecasting tool of these rare events.

We also believe that the forecasting of numbers of storms means very little.  Obviously, the more storms, the higher probability that one or more may reach the coastline.  But this alone does not always follow.  There have been years when there were numerous tropical cyclones, yet none impacted the coastline.  There have been years, 1992 (Andrew) for example, when the numbers were low but the impact was tremendous.   It is therefore highly important that the attempts to forecast these events be directed toward the prediction of where they will form, and if they will be a landfall impact.

Forecast Methodology

             We have researched the various components of the atmosphere and ocean that are at this date discernible.  These components are listed in Table 1.  It is certainly not to be construed that we believe that these are the only components, just that these are the ones that we are currently aware of and have been measured over a length of time that would provide some amount of correlation with the tropical cyclone formation, track, and landfall event. The variables that are used are used without reference to their individual “forecasts”.  For example, the el Niño forecasts are not used, the actual readings at the time are used.  This process is chosen so as to eliminate as much of the inherent “error” of forecasting as possible.  Unless the forecast is "guaranteed" then the impact cannot truly be determined, as the impact varies with the degree of the variability of the component being measured.  The same thinking is applied to all of the other components, with the exception of the QBO (Quasi-Biennial Oscillation).  This is the only component where we have a fairly accurate time frame of easterly versus westerly wind-flow and can be reasonably confident of what the state of the component will be at a given time in the future.

            Each of the variables listed in the following table is measured against the standardized normal.  The anomaly, or difference from this normal is then used to determine the influence in relation to succeeding tropical cyclone activity for June through November.  This relationship is then measured to provide a “weighting” factor to the variable.  This factor is then applied to the inclusion of the variable as related to the current year anomaly versus the standardized values.  The resulting values are then compared to the total of the weighting values.  Those years in which the percentage of the total weighted value equals or exceeds 80 % are used as analog years.  It is from these years, and a measure of forecaster adjustment that the results are obtained. 

             The influence on the forecast that the different variables used in the forecast methodology has is listed below, with accompanying weight and factor.

Blue signifies enhancing, Red signifies restricting

The influence of each of the above factors changes relative to the month of the year.  Those changes will be explained in the next release of this document.