EXTENDED RANGE FORECAST OF ATLANTIC SEASONAL HURRICANE ACTIVITY,
INDIVIDUAL MONTHLY ACTIVITY AND
We foresee an active Atlantic
basin tropical cyclone season in 2006; however, we have reduced our projection
for 2006 hurricane activity from our earlier forecasts. Landfall probabilities for the 2006 hurricane
season are projected to be above their long-period averages for the East Coast
and near their long-period averages for the
(as of 3 August 2006)
By Philip J. Klotzbach[1] and William M. Gray[2]
with special
assistance from William Thorson[3]
This forecast as well as past forecasts and verifications are available via the World Wide Web at http://hurricane.atmos.colostate.edu/Forecasts
Emily Wilmsen, Colorado State University Media Representative, (970-491-6432)
is available to answer various questions about this forecast
Department of Atmospheric Science
Email: amie@atmos.colostate.edu
|
Forecast Parameter and
1950-2000 Climatology (in
parentheses) |
Issue Date 6 Dec 2005 |
Issue Date 4 April 2006 |
Issue Date 31 May 2006 |
Observed Activity Through July
2006 |
Forecast Activity After 1
August |
Total Seasonal Forecast |
|
Named Storms (NS) (9.6) |
17 |
17 |
17 |
2 |
13 |
15 |
|
Named Storm Days (NSD)
(49.1) |
85 |
85 |
85 |
5.5 |
69.5 |
75 |
|
Hurricanes (H) (5.9) |
9 |
9 |
9 |
0 |
7 |
7 |
|
Hurricane Days (HD) (24.5) |
45 |
45 |
45 |
0 |
35 |
35 |
|
Intense Hurricanes (IH)
(2.3) |
5 |
5 |
5 |
0 |
3 |
3 |
|
Intense Hurricane Days
(IHD) (5.0) |
13 |
13 |
13 |
0 |
8 |
8 |
|
Net Tropical Cyclone
Activity (NTC) (100%) |
195 |
195 |
195 |
6 |
134 |
140 |
POST 1-AUGUST PROBABILITIES FOR AT LEAST ONE MAJOR (CATEGORY 3-4-5) HURRICANE LANDFALL ON EACH OF THE FOLLOWING COASTAL AREAS:
1) Entire
2) U.S.
East Coast Including Peninsula
3)
4) Above-average
major hurricane landfall risk in the
Notice of Author Changes
By William Gray
The order of the authorship of these forecasts has been reversed from Gray and Klotzbach to Klotzbach and Gray. After 22 years (since 1984) of making these forecasts, it is appropriate that I step back and have Phil Klotzbach assume the primary responsibility for our project’s seasonal, monthly and landfall probability forecasts. Phil has been a member of my research project for the last five years and has been second author on these forecasts for the last four years. I have greatly profited and enjoyed our close personal and working relationships.
Phil is now devoting more time to the improvement of these forecasts than I am. I am now giving more of my efforts to the global warming issue and in synthesizing my projects’ many years of hurricane and typhoon studies.
Phil Klotzbach is an outstanding young scientist with a superb academic record. I have been amazed at how far he has come in his knowledge of hurricane prediction since joining my project five years ago. I foresee an outstanding future for him in the hurricane field. I expect he will make many new forecast innovations and skill improvements in the coming years. I plan to continue to be closely involved in the issuing of these forecasts for the next few years.
ABSTRACT
Information obtained through July
2006 indicates that the 2006 Atlantic hurricane season will be more active than
the average 1950-2000 season; however, we have reduced our prediction from our
earlier forecasts. We estimate that 2006
will have about 7 hurricanes (average is 5.9), 15 named storms (average is
9.6), 75 named storm days (average is 49.1), 35 hurricane days (average is
24.5), 3 intense (Category 3-4-5) hurricanes (average is 2.3) and 8 intense
hurricane days (average is 5.0). The
probability of
We expect Atlantic basin Net
Tropical Cyclone (NTC) activity in 2006 to be about 140 percent of the
long-term average. This early August
forecast is based on a newly devised extended range statistical forecast procedure
which utilizes 57 years of past global reanalysis data. Analog predictors are
also utilized. This 3 August forecast reduces our forecast from our early
December 2005, early April 2006 and late May 2006 predictions due to small
changes in June-July atmospheric and oceanic fields that indicate conditions
are less favorable for tropical cyclone development in the tropical
Acknowledgment
We are grateful to the National Science Foundation (NSF) and Lexington Insurance Company (a member of the American International Group (AIG)) for providing partial support for the research necessary to make these forecasts. We also thank the GeoGraphics Laboratory at Bridgewater State College (MA) for their assistance in developing the Landfalling Hurricane Probability Webpage (available online at http://www.e-transit.org/hurricane).
The second author gratefully
acknowledges valuable input to his CSU research project over many years by
former graduate students and now colleagues Chris Landsea, John Knaff and Eric
Blake. We also thank Professors Paul
Mielke and Ken Berry of
1 Introduction
This is the 23rd year in which the CSU Tropical Meteorology Project has made forecasts of the upcoming season’s Atlantic basin hurricane activity. Our forecast team has shown that a sizable portion of the year-to-year variability of Atlantic tropical cyclone (TC) activity can be hindcast with skill exceeding climatology. These forecasts are based on a statistical methodology derived from 57 years of past data and a separate study of analog years which have similar precursor circulation features to the current season. Qualitative adjustments are added to accommodate additional processes which may not be explicitly represented by our statistical analyses. These evolving forecast techniques are based on a variety of climate-related global and regional predictors previously shown to be associated with the forthcoming seasonal Atlantic basin tropical cyclone activity and landfall probability. We believe that seasonal forecasts must be based on methods that show significant hindcast skill in application to long periods of prior data. It is only through hindcast skill that one can demonstrate that seasonal forecast skill is possible. This is a valid methodology provided that the atmosphere continues to behave in the future as it has in the past.
A variety of atmosphere-ocean conditions interact with each other to cause year-to-year and month-to-month hurricane variability. The interactive physical linkages between these many physical parameters and hurricane variability are complicated and cannot be well elucidated to the satisfaction of the typical forecaster making short range (1-5 days) predictions where changes in the momentum fields are the crucial factors. Seasonal and monthly forecasts, unfortunately, must deal with the much more complicated interaction of the energy-moisture fields with the momentum fields.
We find that there is a rather high (50-60 percent) degree of year-to-year hurricane forecast potential if one combines 4-5 semi-independent atmospheric-oceanic parameters together. The best predictors (out of a group of 4-5) do not necessarily have the best individual correlations with hurricane activity. The best forecast parameters are those that explain the portion of the variance of seasonal hurricane activity that is not associated with the other variables. It is possible for an important hurricane forecast parameter to show little direct relationship to a predictand by itself but to have an important influence when included with a set of 4-5 other predictors.
In a five-predictor empirical forecast model, the contribution of each predictor to the net forecast skill can only be determined by the separate elimination of each parameter from the full five predictor model while noting the hindcast skill degradation. When taken from the full set of predictors, one parameter may degrade the forecast skill by 25-30 percent, while another degrades the forecast skill by only 10-15 percent. An individual parameter that, through elimination from the forecast, degrades a forecast by as much as 25-30 percent may, in fact, by itself, show much less direct correlation with the predictand. A direct correlation of a forecast parameter may not be the best measure of the importance of this predictor to the skill of a 4-5 parameter forecast model. This is the nature of the seasonal or climate forecast problem where one is dealing with a very complicated atmospheric-oceanic system that is highly non-linear. There is a maze of changing physical linkages between the many variables. These linkages can undergo unknown changes from weekly to decadal time scales. It is impossible to understand how all these processes interact with each other. It follows that any seasonal or climate forecast scheme showing significant hindcast skill must be empirically derived. No one can completely understand the full complexity of the atmosphere-ocean system or develop a reliable scheme for forecasting the myriad non-linear interactions in the full-ocean atmosphere system.
2 Newly-Developed
1 August Forecast Scheme
We have recently developed a new 1 August statistical seasonal forecast scheme for prediction of Net Tropical Cyclone (NTC) activity. This scheme was developed on NOAA/NCEP reanalysis data from 1949-1989. It was then tested on independent data from 1990-2005 to insure that the forecast shows similar skill in this later forecast period. As a rule, predictors were only added to the scheme if they explained an additional three percent of the variance of NTC in both the dependent period (1949-1989) and the independent period (1990-2005)
The pool of four predictors for this new extended range forecast is given and defined in Table 1. The location of each of these new predictors is shown in Fig. 1. Strong statistical relationships can be extracted via combinations of these predictive parameters (which are available by the end of July), and quite skillful Atlantic basin forecasts of NTC activity for the season can be made if the atmosphere and ocean continue to behave in the future as they have during the hindcast period of 1949-2005. Sixty percent of the variance in NTC is explained over the 1949-2005 period, and on independent data (1900-1948), using the same equations and predictors, 49 percent of the variance is explained. This is comparable to what would be expected with independent data as a jackknife regression technique on the 1949-2005 period indicated 52 percent of the variance could be explained. This gives us increased confidence that the new statistical scheme should be of considerable value in the future.
Our statistical forecast for the other predictors (i.e., named storms, hurricanes) is then adjusted by the predicted statistical value of NTC. For example, if a typical season has 10 named storms and the predicted NTC value is 120%, the predicted number of named storms for the season would be 12 (10 * 120%).
Figure 1: Location of predictors for the 1 August forecast for the 2006 hurricane season.
Table 1: Listing of 1 August 2006
predictors for this year’s hurricane activity.
A plus (+) means that positive values of the parameter indicate
increased hurricane activity this year, and a minus (-) means that positive
values of the parameter indicate decreased hurricane activity this year. The combination of these four predictors
calls for about an average hurricane season.
|
Predictor |
Values for 2006 Forecast |
|
1) June-July SST (20-40°N, 15-35°W) (+) |
+0.8 SD |
|
2) June-July SLP (10-25°N, 10-60°W) (-) |
+0.3 SD |
|
3) June-July NINO3 Index (5°S-5°N, 90-150°W) (-) |
+0.3 SD |
|
4) Pre-1 August Named Storm Days – South of
23.5°N, East of 75°W |
0 |
Table 2 shows our statistical forecast for the 2006 hurricane season and the comparison of this forecast with climatology (average season between 1950-2000). Our statistical forecast is calling for about average activity this year, which adds additional support for the reduction of our forecast from our previous early-season predictions.
Table 2: 1 August statistical forecast for 2006.
|
Predictands and Climatology |
Statistical Forecast Numbers |
|
Named Storms (NS) – 9.6 |
10.0 |
|
Named Storm Days (NSD) – 49.1 |
51.1 |
|
Hurricanes (H) – 5.9 |
6.1 |
|
Hurricane Days (HD) – 24.5 |
25.5 |
|
Intense Hurricanes (IH) – 2.3 |
2.4 |
|
Intense Hurricane Days (IHD) – 5.0 |
5.2 |
|
Net Tropical Cyclone Activity (NTC) – 100 |
104 |
2.1 Physical Associations among Predictors Listed in Table 1
Brief descriptions of our 1 August predictors follow:
Predictor
1. June-July SST in the Northeastern
Subtropical
(20°-40°N, 15-35°W)
Warm sea surface temperatures in this area in June-July
correlate very strongly with anomalously warm sea surface temperatures in the
tropical
Predictor
2. June-July SLP in the Tropical
(10-25°N, 10-60°W)
Low sea level pressure in the tropical Atlantic in June-July
implies that early summer conditions in the tropical
Predictor
3. June-July Nino3 Index (-)
(5°S-5°N, 90-150°W)
Cool sea surface temperatures in the Nino3 region during
June-July imply that a La Niña event is currently present. In general, positive or negative anomalies in
the Nino3 region during the early summer persist throughout the remainder of
the summer and fall. El Niño conditions
shift the center of the Walker Circulation eastward which causes increased
convection over the central and eastern tropical Pacific. This increased convection in the central and
eastern Pacific manifests itself in anomalous upper-level westerlies across the
Caribbean and tropical
Predictor
4. Named Storm Days South of 23.5°N,
East of 75°W (+)
Most years do not have named storm formations in June and
July in the tropical
2.2 Hindcast Skill
Table 3 shows the degree of hindcast variance (r2) explained by our new 1 June forecast scheme based on our 41-year developmental dataset (1949-1989), our skill on the independent dataset (1990-2005), and our skill over the entire dataset (1949-2005). Note that the scheme generally shows improved skill in the independent dataset, which lends increased confidence in its use.
Table 3: Variance (r2) explained for our new 1 August forecast scheme for NTC in the developmental dataset (1949-1989), in the independent dataset (1990-2005), and over the entire dataset (1949-2005).
|
Variable |
Variance (r2)
Explained Developmental Dataset (1949-1989) |
Variance (r2)
Explained Independent Dataset (1990-2005) |
Variance (r2)
Explained Entire Dataset (1949-2005) |
|
NTC |
0.52 |
0.76 |
0.60 |
3 Predictions
of Individual Monthly Atlantic TC Activity for August, September and October
A new aspect of our climate research is the development of TC activity predictions for individual months. There are often monthly periods within active and inactive Atlantic basin hurricane seasons which do not conform to the overall season. For example, 1961 was an active hurricane season (NTC of 222), but there was no TC activity during August; 1995 had 19 named storms, but only one named storm developed during a 30-day period during the peak of the hurricane season between 29 August and 27 September. By contrast, the inactive season of 1941 had only six named storms (average 9.3), but four of them developed during September. During the inactive 1968 hurricane season, three of the eight named storms formed in June (June average is 0.5).
We have conducted new research to see how well various sub-season or individual monthly trends of TC activity can be forecast. This effort has recently been documented in papers by Blake and Gray (2004) for August and Klotzbach and Gray (2003) for September. These reports show that it is possible to develop skillful prediction schemes for August-only and September-only Atlantic basin tropical cyclone activity. We have also developed a separate October forecast scheme. On average, August, September, and October have about 26%, 48%, and 17% or 91% of the Atlantic basin's NTC activity. Initial August-only forecasts have now been made by Blake for the last six years (2000-2005), and the verification of these forecasts looks promising. The verification of the September-only and October-only forecasts also appears to show skill.
3.1 Independent August-Only Statistical Forecast
Figure 2 and Table 4 list the predictors used in the August-only hindcast (Blake and Gray 2004) for each of the seven different forecast parameters. The table also shows hindcast skill for the 51-year period 1950-2000, as well as the independent jackknife hindcast skill over this period. Table 5 gives the predictor values for August 2006. Table 6 gives our independent statistical prediction for August 2006. These predictors indicate well above-average activity for August 2006. The most skillful August predictors, in general, call for a very active month, so we are calling for considerable activity during the month.
Figure 2:
Global map showing locations of August-only TC predictors. Table 4 provides a listing and description of
these predictors. The numbers in the
boxes are keyed to the descriptions given in Table 4. The numbers in parentheses beneath each box
indicate how many individual parameters (NS, NSD, etc.) are obtained from each
predictor.
Table 4: Listing of predictors chosen for each forecast parameter and the total hindcast variance explained by these predictors for the August-only forecast. The name and atmospheric parameter utilized in each predictor is given below – where the number is keyed to Fig. 2.
|
Forecast Parameter |
Number of Predictors |
Predictors Chosen From
Table |
Variability Explained
by Hindcast (r2) (1949-1999) |
Estimated Independent
Forecast Skill (Jackknife) |
|
NS |
5 |
3, 6, 7, 9, 11 |
0.55 |
0.41 |
|
NSD |
5 |
1, 2, 3, 8, 10 |
0.71 |
0.61 |
|
H |
4 |
1, 2, 8, 10 |
0.57 |
0.47 |
|
HD |
5 |
3, 4, 8, 9, 10 |
0.69 |
0.59 |
|
IH |
5 |
1, 3, 5, 8, 12 |
0.68 |
0.59 |
|
IHD |
5 |
1, 4, 5, 6, 9 |
0.78 |
0.72 |
|
NTC |
5 |
1, 2, 8, 10, 12 |
0.74 |
0.66 |
Table 5: August 2006 predictors. The sign of the predictor associated with increased tropical cyclone activity is in parentheses.
|
Predictors |
2006 Observed Values |
Effect on 2006 Hurricane Season |
|
Galapagos July 200 mb V (-) |
-1.0 SD |
Enhance |
|
|
-1.0 SD |
Enhance |
|
|
+0.8 SD |
Suppress |
|
SE Pacific July 200 mb U (-) |
+0.2 SD |
Suppress |
|
|
+1.0 SD |
Suppress |
|
|
+0.7 SD |
Enhance |
|
Galapagos July 200 MB U (-) |
-0.5 SD |
Enhance |
|
|
+0.5 SD |
Enhance |
|
Northwest Pacific June SLP (+) |
+1.0 SD |
Enhance |
|
|
-0.6 SD |
Enhance |
|
|
+0.4 SD |
Suppress |
|
|
-1.0 SD |
Enhance |
Table 6: Independent August-only prediction of 2006 hurricane activity based on Blake and Gray (2004). August climatology is shown in parentheses.
|
Parameter |
Statistical Model |
Qualitative Adjustment |
|
NS |
3.3 (2.8) |
4 |
|
NSD |
21.1 (11.8) |
22 |
|
H |
2.9 (1.6) |
3 |
|
HD |
8.1 (5.7) |
11 |
|
IH |
0.7 (0.6) |
1 |
|
IHD |
2.0 (1.2) |
3 |
|
NTC |
53.6 (26.1) |
50 |
3.2 Independent September-Only Statistical Forecast
Figure 3 and Table 7 list the predictors used in the September-only hindcast (Klotzbach and Gray 2003) for each of the seven different forecast parameters. The table also shows hindcast skill for the 51-year period 1950-2000, as well as the independent jackknife hindcast skill over this period. Table 8 gives the predictor values for September 2006. Table 9 gives our independent statistical prediction for September 2006. Predictor values for September 2006 are mixed, so our final forecast is calling for slightly above-average activity for the month.
Figure 3:
Predictors selected for the end of July forecast of September tropical
cyclone activity. The numbers in each
area are keyed to the description given in Table 7.
Table 7: Listing of predictors chosen for each forecast parameter and the total hindcast variance explained by these predictors for the September-only forecast. The name and atmospheric parameter utilized in each predictor is given below – where the number is keyed to Fig. 3.
|
Forecast Parameter |
Number of Predictors |
Predictors Chosen From
Table |
Variability Explained
by Hindcast (r2) (1950-2000) |
Estimated Independent
Forecast Skill (Jackknife) |
|
NS |
3 |
2, 3, 5 |
0.29 |
0.19 |
|
NSD |
5 |
2, 3, 4, 5, 8 |
0.54 |
0.44 |
|
H |
3 |
2, 3, 8 |
0.38 |
0.28 |
|
HD |
5 |
2, 3, 4, 5, 8 |
0.60 |
0.51 |
|
IH |
5 |
1, 2, 3, 5, 9 |
0.63 |
0.53 |
|
IHD |
4 |
3, 4, 5, 9 |
0.63 |
0.54 |
|
NTC |
5 |
2, 3, 4, 5, 9 |
0.75 |
0.68 |
Table 8: September 2006 predictors. The sign of the predictor associated with increased tropical cyclone activity is in parentheses.
|
Predictors |
2006 Observed Values |
Effect on 2006 Hurricane Season |
|
South Atlantic April 1000 mb U (-) |
-0.6 SD |