Seasonal Forecast for the 2005 Hurricane Season

Issued May 10th, 2005

* UPDATE:  June 26th, 2005
** UPDATE:  July 31st, 2005
*** UPDATE:  August 21st, 2005
**** UPDATE:  September 11th, 2005

**** This is the final seasonal update for 2005.  After Katrina this may seem almost academic; it's obvious who's taken the brunt of this season.  But, from climatological statistics, the season is only half over as of the 10th of September.  One would sure hope that activity-wise we are, in actuality, far past the half-way point.  But, certainly, there've got to be at least a few storms left in the cue.  In fact, due to the scattered locations of the storms and the very warm Atlantic sea surface temperatures at the start of the season, there really has not been a tremendous sapping of the Atlantic heat content.  At any rate, things are pretty stable with respect to the season forecast and there aren't many changes in the update.

*** Changes in the driving atmospheric parameters this month were far more subtle.  Not that there were any major changes last month, but there were enough to shift around the risk assessment a bit.  Any risk assessment shifts this month will be largely based on subjectivity, which I try to keep to a minimum, but is inevitable to a degree.  The various predictors have not just remained completely unchanged for the past month, but the changes this month were fairly subtle.  Not only were they subtle, but they offset each other somewhat.  For example, the QBO continues to progress towards its peak for tropical cyclone inhibition (not that it's had much success this season inhibiting activity so far) while the Nino3.4 region's sea urface temperature anomaly (SSTA) has cooled during the past several weeks.  For stability, we use a month-long SSTA average. because of that and the fact that we're using data only through the end of July, we don't see the full impact of this Nino3.4 SSTA drop in this analysis.  However, there is still some slight drop (by about 0.1C).  Another reason this has difficulty impacting the risk assessment is that this largely reverts us back to earlier seasonal values, as the Nino3.4 SSTA had previously been rising slowly.  At any rate, very few modifications this month.  Amendments to this assessment for this update are inserted where appropriate and preceeded by a triple asterisk (***).

** A bit late with the July update again.  My intentions are to get these updates out around mid-month, not late month.  This is especially important for the later-in-the-season updates.  So, I'll try to get the August update issued earlier.  The reason for the lateness is probably obvious... the Atlantic has been roaring through the month of July, in just incredible fashion.  At any rate, we have seen some changes over the past month or so in the various parameter predictions.  So, you will see some modifications to the seasonal forecast below.  However, most of the changes witnessed were expected.  Moreover, the changes did not represent an actual shift in the direction of the parameter.  Just as one example, we discussed the QBO in last month's update (see below).  It has continued to shift in the anticipated direction.  So, no great shocks there.  Of come significance is how strongly negative (easterly) the QBO has become.  We're ending up with a battle of atmospheric forces.  The QBO is becoming a rather strong inhibitor.  The ENSO phase is getting just strong enough (warm east Pacific SSTs) to become a slight inhibitor.  Meanwhile, the Atlantic SST anomaly continues to climb into record territory, acting as a strong counterpoint to the QBO and ENSO.  Anyway, in general, though, the changes are minor.  And, as always, the original updates are left intact.  Updates for this July post will be instered where appropriate, preceeded by a double asterisk (**).

* Getting a bit of a late jump on this month's updated seasonal forecast.  That is not, however, due to any need for great adjustments to the forecast.   The parameters, analog seasons, and end results remain largely the same.  One parameter has shifted significantly in the past month or two; that is the Quasi-Biennial Oscillation.  Nonetheless, even this has only a minimal impact, as the change is in only one parameter and it is in the direction anticipated.  Still, it is the primary reason that there are any changes at all in the forecast.  As the analogs (not the risk assessment) are obtained in a wholly objective manner, the fact that we "expected" the QBO to change in the manner it did is somewhat irrelevant.  If that change did not previously appear in our analogs (and, incidentally, it did, but not to the degree that the change is occurring) then our expectations were not factored into our May update.  At any rate, the updates, as always, are interspersed below, preceeded, for this June update, with a single asterisk (*).

Last year was probably the easiest year to hit our risk assesment successfully, simply from a statistical standpoint, as our maximum risk areas were Florida and the Carolinas.  Statistically speaking, that's like predicting that in an eight man race you'll finish in the top six... no guarantee, but odds are, the prediction will verify.  2004 was just like that.  Given climatology, the two areas in the U.S. most prone to hurricane landfalls are Florida and North Carolina.  So, putting them in the maximum risk area is hardly a rip-roaring success.  In fact, digging into the specifics of the 2004 risk assessment, we see that there were some subtle gradations in my 2004 forecast (this is especially evident for those with access to the risk assessment map).  Within those gradations, every hurricane except the relatively modest Alex ran ashore in a slightly lower risk assessment region.  So, certainly, the 2004 forecast wasn't great.  Nonetheless, it was a reasonable success as we predicted that the prevailing pattern would yield some sharply turning tracks, keeping landfalls away from the western Gulf of Mexico and the Northeastern U.S.  In fact, of five storms reaching the Gulf, none struck Texas and only two tropical storms hit Louisiana (including the ultra-bizarre longshot... a revived Ivan, after dropping DOWN the East Coast - after its earlier landfall and traversing the Eastern U.S. inland - moving across Florida, and redeveloping); and, with 9 named storms making it past 70W longitude, only one impacted the Northeast as a tropical storm (Hermine).  So, no hurricanes struck those lower-than-normal risk regions, and only a few minimal tropical storms had an impact.  Therefore, overall we can probably consider the 2004 risk assessment a reasonable success.

Verification:

** With the significant activity to date, we can already do a very preliminary verification for 2005.

For the above verification we've used the following color coding:
Red = Poor forecast
Yellow = Marginally bad forecast
Green = Generally successful, but not perfect forecast
Blue = Near-perfect forecast

In 10 years of risk assessment predictions only one season, 2000, was an unmitigated disaster.  And only one other season, 2002, was generally "bad".  In contrast, six seasons turned out reasonably accurate while another two pretty much nailed the risk assessment.  While the high risk focus of the forecasts is almost always a climatologically high risk region, which may seem to give these forecasts less gravity, keep in mind, there are wide climatologically risky regions routinely left out of these forecasts.  For example, three of these ten seasons focused merely on the "Central Gulf Coast", excluding Texas, the entire Southeastern U.S. and all of Florida except the westernmost portion of the Panhandle.  Eight of ten seasons may not yet be statistically significant to prove skill, but barring some successive failures in the next couple of seasons, we're likely well on our way.

* Obviously, it's a bit early to say much in terms of a verification for 2005.  However, it's only late June and we've already had one storm, Tropical Storm Arlene, and nearly a second (a much-watched wave developed a low pressure just before moving inland into eastern North Carolina on June 26th).  Though that second system doesn't officially count for verification purposes, both Arlene and that system fell into two of the higher risk regions.  Arlene may have been a shade east of the LA/MS/AL zone (Moderate-High Risk), actually coming ashore in the slightly lower Moderate Risk NW Florida region, but... Obviously, these are not discreet breakdowns - there is not some wall between regions - so, if we were to look at "Moderate Risk" as the mid-range, Arlene still came ashore in an above normal risk region.  The second system, though not truly counting, struck our lone "High Risk" region of the Carolinas.  Also, since there is only, on average, one tropical cyclone every two years in June, we're off to a fast start.  Though there is no historical statistical link between early season and overall activity, there is good meteorological reason (warm ocean temperatures) why there could be this year.  So, all in all, both the risk and activity forecasts seem to be getting off on the right foot.

** Arlene, Cindy, Dennis and Emily all clipped the Gulf of Mexico high risk region, with only Arlene officially right dead on in it.  Bret and Gert did not strike any high risk areas.  So, thus far I'd have to say this has been only a modest success.  This is especially true since the Carolinas have the highest risk of any region in my forecast.  Of course, given trough penetration, the Carolinas are far more likely to be struck in late August, September or even early October than they are in July or even June.  So, let's wait and see how that pans out.

*** Since Gert (the most recent at the time of the last update) things have slowed down a bit (though still on record pace) with only Harvey and Irene.  Both remained offshore and impacted no land.  They did fit nicely, however, with the high risk region along and off the U.S. East Coast.  The way this is laid out on the risk map, for subscribers with access to that map, while North Carolina is in the high risk region, they are on the western edge of it, as the core of it is offshore between Bermuda and the U.S. East Coast.  Harvey and Irene moved, essentially, right through this high risk region.

**** The last month has been mediocre with regard to the validation, but there is a change in the air as of this writing.  It's "mediocre" because Katrina behaved well swinging through two of the higher risk areas, but the lone region identified as absolutely high risk, the Carolinas, has remained untouched.  However, the "change in the air" is that as of the date of this writing Hurricane Ophelia is predicted to strike North Carolina within the week.  It remains to be seen whether or not that prediction will come to pass, but there is certainly the threat there.

Pattern:   Frankly, we can probably do away with this section, in the efforts for brevity; in many ways, it is redundant, as the risk assessment gives a general indication of the overarching pattern.  Moreover, it is this very risk assessment that is the heart of what we are forecasting, not the pattern itself, regardless of how related they are.  So, let's keep this portion of the discussion extremely brief...  The global atmospheric and oceanic parameters are not pointing towards any real strong bias in one direction or the other.  However, these neutral conditions can often be harbingers of fairly amplified and progressive patterns.  That means some significant weather swings across much of the U.S. this summer.  Obviously, as we get into the dog days troughs will have more difficulty digging.  So, we're not saying that massive cold fronts will sweep through all the way into the Southeast in the middle of July and August.  That is highly unlikely in any summer.  But we could see a number of decent fronts make it into the Northeast and their remnants, in the form of slightly cooler drier air, pushing deep into the Southeast.  And, while that's not occurring, some genuinely hot weather.  It will likely average out near normal in the Central and Eastern U.S., perhaps a shade above normal in the Southeast and a shade below normal in the Upper Midwest, but with changeable conditions.

* Not much to add here.  I suppose this is averaging about right.  It's a bit difficult to tell with the somewhat up and down weather in parts of the east.  Even my local area (Washington, D.C.) began the month quite pleasantly cool.  We also had our share of rain.  Since then, we've gotten rather hot, then moderate, then hot again, but dry since the early month rain.  I suppose with the talk of the fronts sweeping through the area, these changeable temperatures is exactly what would be expected.  However, I would have thought we wouldn't have had quite the dry stretch we've had (although, it has been rather localized and certainly not extreme and, as I write this, there is rain in the forecast).  So, I'd be hesitant to say that the pattern forecast is right on target, but it's roughly on track.

** There continues to be a bit less troughing in the East than I would have expected.  Still, there has been some, and I would hope/expect that once we move towards the end of August and into September, those troughs will show more strength and penetration.  And with that occurring near the peak of the hurricane season, it will have the effect of drawing storms more northward near or just off the U.S. East Coast, which goes in line with the risk assessment.  So, there's not enough in the pattern to convince me that we're off the mark, but I'm am a bit uneasy.  I realize this is probably right about the climatological peak for the minimum in trough/frontal southward penetration, but I still would've expected just a bit more from these systems.  So, I'm a bit nervous about the risk assessment, but not enough to make any changes to this patter forecast.

*** On thing I've been concerned with in both the assessment of the analog seasons and the pattern thus far this season is the wide, flat nature of the Atlantic subtropical ridge.  I have New England in a rather higher than normal risk.  This requires a rather "tall" Atlantic ridge, forcing storms to ride up the East Coast into New England.  Several storms did so in the analog seasons, which is why the risk assessment came out the way it did.  However, there were also a large number of storms in the analog seasons that hooked hard out to sea.  Irene acted in a fairly similar way.  Also, as you see above from my previous update, I'm concerned about the flat troughs, which is a symptom of the same issue, just a very flat pattern overall.  At this point, since we're climatologically in just about the flattest time of year (July and August), I'm not going to "pull the plug" on the forecast.  I'm encouraged also by a couple of deeper troughs recently.  So, nothing is "officially" changing in my forecast, but I do at least want to re-iterate this concern.  If, by the beginning of September, there are no higher amplitude troughs and ridges in the pipeline, New England and, to a slightly lesser degree, the Canadian Maritimes can figure on a significantly reduced threat.

**** We've seen some good sharpening of the pattern, which is not tremendously surprising as we move into September and the high latitudes begin to cool.  This is having a very real impact even at the very date of this update, as much of the numerical model guidance shows Ophelia running up into New England.  Because of the specific details, IF this verifies it does appear that she's likely to be a mere tropical storm at that time, but that's not really the point.  The point is the track, as controled by the pattern.  We have, in fact, seen an amplification, returning us back to our original expectations.  And given that we're only moving deeper and deeper towards autumn, I can't foresee this pattern flattening out substantially for any extended period of time.

Factors:   Here's where we've begun to change our methodology a bit.  In past years I've focused on the same parameters the Dr. William Gray has used in his activity assessment.  However, I migrated away from that somewhat.  For one thing, the African rainfall factor hasn't even worked that well for the general activity forecast in the past few years.  I would still bet my life that there's a "connection" to activity, given that it's highly logical.  However, the connection to landfall activity, though apparent in some publications, is less.  And simply based on my own interpretation, I would view the LOCATION of such landfall as having an even lesser link to African rainfall (though perhaps not zero).  So, I have eliminated this fact.  I have also ceased using the Caribbean sea level pressure anomaly (SLPA).  This is a valuable tool, BUT... it is highly dependent on the Atlantic sea surface temperature anomalies (SSTA).  These Atlantic SSTAs are already utilized as a heavy predictor.  As such, including on top of that the Caribbean SLPA corrupts the data.  Because of Dr. Gray's emperical statistical (not PURELY statistical) methodology, this non-independence of the data does not harm him; it may do so in our case.  In fact, there's even been shown to be some modest link between the Quasi-Biennial Oscillation (QBO) and El Nino, two of our parameters.  So, we still can get the point of totally independent factors.  However, we need to strive to get as close as we can.  So, Caribbean SLPA has been removed.

There are two other significant changes.  First, with respect to the Atlantic SSTA, NOAA's Climate Diagnostics Center now provides a far superior dataset for this parameter.  Previously, the only available dataset had very limited regional coverage over the Atlantic.  This new dataset covers a much greater region over the tropical north Atlantic.  Second, we are now including the Atlantic Multi-Decadal Oscillation (MDO).  This now gives us the ability to quantify the cyclical nature of the Atlantic Basin's circulation.  Actually, I was prepared to do this using the Thermohaline Cycle Index (TCI).  However, it is a simplistic calculation corrupted for our usage by its strong dependence on the Atlantic SSTA.  The MDO is correlated strongly to TCI, so it is surely not entirely independent either.  But it does provide us with a slightly more independent source.

In all, these are significant changes.  Still, we do continue to use a core of factors... QBO, El Nino, Atlantic SSTA and the Atlantic multidecadal circulation... much the same as in previous forecast.  So, while there are significant adjustments, the basic risk prediction scheme remains intact.

* As noted up front, the QBO is the big change this month.  It is the large factor in eliminating a couple of analog seasons and adding in a few.  Although, some subtle shifts in the other parameters are also aiding in moving thins around.

** I've pretty much discussed this already... The QBO continues shifting in the negative (relative easterly direction), rather strongly.  Meanwhile, eastern Pacific SST anomalies continue to warm.  Although, those are warming slowly and have still only reached the level to qualify as a weak El Nino phase right now, and increasing slowly (so, likely to remain weak throughout the hurricane season).  The Atlantic SST anomaly, however, has reached extreme ranges.  It's up to nearly one degree Celcius above normal (in a three month average) in the "TNA" (Tropical North Atlantic) sector.  While that may not sound like much (one mere degree), the highest three month average on record (since 1948) is 0.70C above normal in February-March of 1998.  The actual value over the past three months right now is 0.95C above normal, with the latest month (June... the July data is not available yet) at 1.01C above normal.

*** Mainly small changes this month, and we have a problem with one of the critical parameters not being available for some unknown reason.  We've seen the QBO continue to progress in the negative direction.  However, its rate of change has slowed dramatically as it approaches its peak.  The Nino3.4 SST anomaly has actually changed rather significantly, dropping into neutrality, very close to zero in the past couple of weeks.  This most recent data, though, does not make it into this assessment.  We need to be consistent in our use of these parameters.  As such, in using the July Nino3.4 monthly average SST anomaly we see a much more modest shift in the SSTA... a drop from about +0.54 to +0.45C.  Note that the latest Nino3.4 weekly SSTA value from early August is exactly 0.0C... no anomaly whatsoever.  We have a serious problem with the Atlantic SST anomaly, as the database for this parameter still has not been updated since June.  This creates a major problem in doing the analog assessments below because this is one of the more heavily weighted parameters.  We can take a wild guess that it is probably a bit cooler than the June number, but still very high... we can make that guess because 1) June was a record, difficult to maintain and, 2) with record tropical cyclone activity it may be safe to presume that some energy/heat was sapped from the Atlantic.  But it's nearly impossible to put a number on it.  As such, we can really just leave this parameter alone.  In doing so, it is difficult to get much shift at all in the analogs or the risk assessment.

**** Here's why there are almost no changes with this update in terms of risks.  Our driving factors have really held pretty steady.  The QBO has continued to increase in a direction unfavorable for development, but the increase this month has been the smallest all season; it is virtually unchanged from last month.  The El Nino phase continues to fade towards neutral.  This parameter has changed measurably, but still only a very small change this month... from a +0.39C anomaly to a +0.17C anomaly... certainly a measurable change, but both are nearly neutral.  Meanwhile, for the Atlantic SST anomaly, since I use a three month running mean, there is zero change in this parameter; the value dropped from the mean was +0.85C, the value added was +0.86C.  There was a similar bounce-back in the Atlantic Multi-decadal Oscillation (AMO) parameter.  On the whole, the AMO has decreased some but, again, by a rather trivial amount.

Analogs:  We've actually added this section this season, but primarily to split the "risk discussion" section into two.  That section had discussed the analogs and the overriding, resulting general risk assessment (not the specific region by region risks).  The table below shows the top 9 analog seasons, in the order of their analog strength.  The numbers under each regions column represent the total categorical landfalls for that region in that year.  What we mean by "categorical landfalls" is the following...  In the given region, if a Category 2 hurricane strikes the region 2.0 points are assigned; if a Category 4 hits, it's 4.0 points.  Get the idea?  Very simple.  It a tropical storm strikes, it's worth 0.5 points.  This season we are also attempting to break down the regions in a more even distribution (similar lengths of coastline per region), plus we're including non-U.S. regions.

* Because I am not doing weighted averaging with these analogs, they needn't be in analog-strength order.  I simply did that with the original post to provide you with a feel for how each season ranked.  I mention that, because I'd like to leave the original post intact as much as possible.  So, for this update here's how we're going to modify the table below...  Seasons 1964 and 2002 have been eliminated.  We will simply place an "X" in front of them and change the background shade.  Meanwhile, seasons 1969, 1958 and 1952 have been added.  We will simply place them at the end with an asterisk to denote them as additions for this update.  The means and medians will simply be updated, without notation.  I'd prefer to leave the original values so that you can see the changes, but that would get ugly.  Plus, all of the data is contained herein (I'm not removing 2002 and 1964 from the table) so you can pretty easily see the impact of the changes.

** The changes this month are nearly indecipherable in the numbers in my raw database.  However, right at the cusp of what I include, we've seen just enough change to "toggle" a few seasons off and on.  In fact, quite interestingly, what we mainly see occurring is a shift back towards what we had before the June update.  That is, the changes in the atmospheric parameters this month result in some slight slippage for 1969 and 1958, and a slight boost for 2002 and 1964.  The changes are extremely subtle, but it's just enough to turn the former two off, and return the latter two back to the list.  I'll use the same notation as with the last update, but I'll put the previous asterisks and "X"s in parenthesis, and use the double asterisk notation for this update.  Note that the 1952 addition from the last update maintains its status.

*** Last month's changes were small; this month's are even smaller.  As noted in the "factors" section above, one of our more critical factors failed to get updated this past month.  So, we're rather "stuck" here.  We can say, however, that the modest changes in QBO and Nino3.4 SSTA do secure 1952's tenuous existence on the list.  Also, 1963 is on the cusp of making the list, but it falls just short.  The small changes do re-drop 1964 from the list.  Depending on the true state of the Atlantic SSTA, 1970 may get dropped from the list; unfortunately, this forces some subjectivity on my part and I'm going to elect to drop 1970 (note the "X" in front of the year).  The 1980 season also tries to crack the list with improvements from the minor QBO and Nino3.4 changes, but it's not quite enough.  The 2002 and 2004 seasons get threatened with removal from the list due to their increasingly poor QBO analogs, but they remain in place, saved by their increasingly strong Nino3.4 analogs.  So, that's it... we removed 1964 and 1970 and added no replacements, leaving us with eight analog seasons.

**** The progressively smaller changes continues.  It's difficult this month to even come up with any meaningful changes to the assessment.  Part of the problem we face is that the El Nino changes are the most significant and they only reinforce what we already modified previously.  For example, 2002 and 1952, added in earlier months are boosted.  Meanwhile, 1964, 1969 and 1958, which were already previously eliminated, fall further.  So, there are no changes to the analysis below.  I will say, however, that 1960 and 2002 get a boost from this change, while 2003 takes a hit.  It is not sufficient enough to remove 2003 from the list, but in the risk analysis that follows this section, I will attempt to take this into account.

Risk Discussion:   The data in the table above shows several aspects of the risk assessment.  The median line points us to the midpoint for each region, so that we are not misled by singular major seasons.  So, this helps us assess risk by pointing to those regions more likely than not to be impacted by a tropical cyclone.  The median will point to the fifth of the nine analog seasons.  Values of 0.5 indicate that that region received one tropical storm in that median year.  Greater values indicate multiple and/or stronger tropical cyclone impacts.  Therefore, locations with any non-zero median can be presumed more likely than not to be impacted by "something".  Meanwhile, regions over 0.5 are under a rather high risk.  These regions in the Carolinas, Bermuda and, oddly, the Canadian Maritimes.

Risks:  The regional risk assessments follow...

*For the update, where a change occurs you will see the old risk filled in a dull shade and the new risk filled in normally, with an asterisk in the field.

The above color-coded table takes into account not only the Mean and Median for that region, but also the weighting of the analog strengths.  For example, New England, with only a 0.61 mean and a zero median could easily be dropped into the low or at least the low to moderate category.  However, in two of the three strongest analogs that region has a 1.0 "score", with the third of those years being the out-of-whack 1962 season.  And of the next two seasons in the list, 2003 and 1960, New England has a 2.5 in one of them, and that is in 1960, which MAY be elevated after April's data is analyzed, if the QBO changes rule the day.  On the flip side, I tried not to shift things around too heavily in a subjective manner.  For example, I easily could have dropped Texas down one category given that all of its action is in the lower analog seasons.  However, I decided that all nine of these are good analogs, so getting too deep into weighting is counterproductive, and Texas has some impressive numbers.  Similarly, I could have dropped Nicaragua/ Honduras and the Yucatan, but chose not to.  Actually, those are more compelling than even Texas.  There was much debate over them, but I opted to stick with the objective numbers for now.  But don't be surprised if either or both of them are dropped down to low in the next update.  I also could have argued to move South Florida and the Bahamas up to a high risk, as they're sporting an impressive 2.22 mean, the highest that exists on the table.  But, with only a 0.50 median, that region is clearly aided by some hefty seasons (last year in particular).  So, I opted to keep them out of the high risk for now.  The result is that only the Carolinas get pegged as high risk (though I'm concerned about their relatively modest mean value), with higher than normal risk values throughout the East Coast.

* Most of the changes have an obvious reasoning behind them as the medians and/or means changed notably for these areas.  But, you may wonder why Texas and LA-MS-AL have not been dropped back a category to "Moderate Risk".  For Texas, I have trouble doing that for any region with a non-zero median.  In order to have a non-zero median that means that half, or more, of the analog seasons contained a tropical cyclone landfall in that region.  That is, a strictly objective assessment states a MINIMUM of a 50-50 shot of a landfall for Texas.  Meanwhile, the re-ordering of the analogs pushes the fairly active - in Texas - 1970 season to the head of the class.  So, I've left Texas alone.  As for LA-MS-AL, that's a tougher one, as their mean is under 1.0 and their median is zero.  There is only one reason I've kept this area as Moderate to High, and it is a subjective reason that, I admit, may well be incorrect.  That reason is 1969 getting launched up to number three in this update.  That was the year of Camille.  Of course, that was the only storm to hit that region in that season or either of the other two added seasons, while the two eliminated seasons were both active in that region.  So, I'm largely basing this logic on one storm... probably unwise, but I've also considered that 2002 and 1964 are still "good" analogs... just not good enough to rate on the top ten (2002 is now the 13th best analog, and 1964 is the 15th best); so, it is fair to keep those two seasons in consideration.

** No changes whatsoever to the table.  To be clear, there are some slight changes to the risks, just nothing for which I can justify any changes to this table.  For example, the NW Carib has a pretty extreme median now, at 2.25, as does S FL and the Bahamas.  However, I cannot move them fully into "High Risk" with a median still below 1.00.  The mean is the average, whereas the medium is more representative of the "typical" season.  If the typical analog season doesn't contain even a single category one hurricane strike for a region, I cannot consider it "High Risk".  I'd be pretty firm in that.  Also, I was tempted to bump the NE Carib and Bermuda up a category, while dropping the Maritimes and New England down a category.  However, all of them are on the threshold of those categories, and I would remind you that the shift in analogs is among the weakest analogs.  That is, we've not dropped our strongest analogs or pushed 2002 and/or 1964 up to the top of the list.  So, it seems prudent to leave well enough alone.  I will tell you, though, from a subjective standpoint, I am most inclined to reduce the risk in New England.  With a mean now at 0.55 and a median dropped to a flat-line zero, it is difficult to justify a "moderate" risk; low-moderate may be more appropriate.  But with the mean still over 0.5, and the aforementioned logic (the fact that this change is due largely to the weaker analogs), I've decided to leave them alone.

*** Just a couple of minor changes with this update.  The risk assessment for Texas and for NW Florida have been decreased a category.  I was hesitant to decrease Texas since they've already been threatened this season.  Actual activity has to speak loudly in such an assessment.  However, it's also safe to say that outliers can easily occur from time to time and, which is why there is never any guarantee... note that we have no "zero" risk category.  So, we don't want to lean TOO heavily on activity that has already occurred, it won't necessarily repear.  As a result, I just followed the numbers and decreased the risk for those two regions.  I also faced a quandary for New England.  As mentioned in the discussion earlier, there is reason to think that the risk for New England should be lowered.  However, from a pure numbers standpoint, the removal of the two analog seasons actually caused New England's raw risk numbers to increase.  Thus, in the end, it seemed best to just leave them alone.

**** The only change is the reduction for Texas.  They were already borderline for the last update.  Now, with the reduction of the influence of 2003 the risk category for Texas needs to be dropped down a notch.  I did not increase any risk categorization for East Coast locations, because they may have already been a bit "generous".  So, the re-prioritization of a few of the analog seasons only helps to solidify the current assessment.  Also, I did not raise MS/AL/LA to a full fledged "High Risk" because Katrina is passed.  Obviously, as a full 2005 seasonal assessment, that region would, in hindsight, certainly qualify as high risk.  But with half the season left and that area already getting clocked once, they may well see nothing else... still, I have to at least follow the numbers for the most part.  So, I need to keep their risk relatively high... getting one storm does not preclude getting another (see: Florida, 2004)... but I'm not going to bump it up based on a hair-splitting analysis for this update.

Conclusion:  Overall, this seasonal risk assessment is rather "messy".  There is not really any good, clean High Risk region that we can focus in on.  The Carolinas seem like such an obvious High Risk, with landfalls in all but one analog season.  But most of those strikes were so mundane that their highest seasonal total score was a 2.5; not that a Category 2 hurricane should be taken lightly - it shouldn't - but the point is, they received no major hurricanes in any of these analog seasons.  I was compelled to push them high merely because of the consistency with which they were struck.  This "lukewarm" picture yields a fairly modest confidence level in the forecast.  It is further complicated by issues like the aforementioned complexities in Texas... several impacts, but all in lower analog seasons.  Plus, there seems to be a rather wide broadbrush, with many regions in Moderate or Moderate-High Risk.  But, we can't call it low confidence or just a whitewashed prediction, as there are also some notable signals.  Bermuda and the Canadian Maritimes are, for example, at a strikingly high risk.  Stretching down from the Maritime provinces and up from the High Risk Carolinas, all of the Mid-Atlantic and Northeastern U.S. are in higher than normal risk.  On the flip side, Mexico and Central America are noticeably absent from any great risk.  So, there are, at minimum, some fair signals as to where the greatest threats and risks are for 2005.

* Not a heck of a lot of change in the overall risk assessment.  The updated analogs caused a bit of a focus in the corridor along or just off the U.S. East Coast.  So, regions "tucked in" a bit, like northeastern Florida, Georgia, Virginia, Maryland, Delaware and New Jersey see a slight dip in their risks.  The same is true for areas east of this corridor, e.g., Bermuda.  And though the Canadian Maritimes and New England don't see enough increase in their raw numbers to bump them up a risk category, their raw numbers did, in fact, increase.  I should also point out that one should not be misled by the increase in the risk category in the Yucatan.  With all other Gulf risk categories remaining as is, one might think this indicates a steady or slightly increased Gulf risk, in general.  On the contrary, all other Gulf regions saw a decrease in raw risk numbers... just not enough to pull them down a category.

** After all that, and with some significant changes in the atmospheric parameters, we're essentially left with very little change in the risk assessment... so little that I cannot change any of the categorizations.  However, high tier subscribers with access to the risk map will see some minor differences.  There will be an expansion of some of the higher risk regions over the northern Caribbean.  Meanwhile, off to the north, there will be some slight reduction through New England and the Canadian Maritimes, offset by a slight increase out towards Bermuda.  The Carolinas remain our sole high risk region, but I do have some concerns with the pattern being able to sustain this risk.  I'm not yet convinced that the pattern can't support a Carolina storm or two, as troughs dig deeper in the latter part of the season.  But, as the Carolinas are far and away my high risk region, I am concerned.  Basically, for this season's risk assessment to be correct, NC/SC really MUST be struck by at least one minimal hurricane... at least.  For the sake of those in the Carolinas, hopefully, I'll just be wrong.  At this point, though, my forecast remains unchanged and focused on that area, the Carolinas, as the high risk.

*** All in all, this August update seems to have resulted in a general, albeit slight, risk reduction for most of the Gulf.  Specifically impacted in the analysis were Texas and northwestern Florida.  Early in the season the Gulf was fairly active and this assessment is still supposed to apply to the entire season, including that which is behind us.  So, I'm not overly confident following the numbers in that regard.  But, that's what those numbers show and, besides which, most readers of this product will be forward looking.  That is, they're not going to read this to see where the storms have been!  We'll do that when it's time for the final validation; even then, the Gulf is still showing some pretty reasonable risk... so, if they quiet down a bit from this point on and the higher risk areas (like NC) heat up, even the current forecast will validate just fine.  Of course, that's an "if", so, we'll see what happens.  At any rate, the primary change for this update does have a slightly reduced risk in the Gulf and nearly everyone else remains the same.

**** Very tiny changes for this final update.  The atmospheric and oceanic parameters are just very stable, not changing much at all over the past month.  This leaves us with very little modeification to the risk assessment.  The only actual change in a risk category is a one notch lowering for Texas.  And the one area where the forecast is really "on the hook" is the Carolinas.  For the sake of folks living there, I obviously hope my forecast is wrong.  But as they are assessed as "High Risk" no landfalls there at all would be a true "bust" to the forecast.  As it stands, as of September 11th, the seasonal risk assessment is poised to verify with Hurricane Ophelia.  But, that hasn't occurred yet, so we can only wait and see what happens and what else is in store for the remainder of this season.

-Gary