The Tropical Meteorology Project (TMP) at Colorado State University
(CSU) has recently released its latest hurricane forecast. This blog
briefly discusses how this forecast was made and what it entails for the
remainder of the Atlantic hurricane season.
CSU began issuing seasonal forecasts in 1984, when Dr. Bill Gray noted a significant relationship between El Niño (warmer than normal sea surface temperatures (SSTs) in the eastern and central tropical Pacific) and Atlantic hurricanes. When El Niño conditions are present in the Atlantic, vertical wind shear levels increase in the Atlantic, tearing apart tropical cyclones that are trying to develop.
Figure 1. Typical vertical wind profile observed in El Niño and La Niña years in the Main Development Region.
Since that time, the forecast model has been expanded to utilize a variety of parameters that include Atlantic basin sea surface temperatures, sea level pressure anomalies and low-level wind anomalies. The primary tool that is used for making the various predictions is a statistical model based on historical data. In the case of the August seasonal forecast, three predictors are utilized, that when used in combination, explained significant amounts of variability over the period from 1979-2012. The target predictor is Net Tropical Cyclone (NTC) activity. This parameter convolves the frequency and intensity of tropical cyclones into one number, where the 1950-2000 average of this parameter is 100% by definition. It weighs major (Cat. 3-5) hurricanes more strongly than Cat. 1-2 hurricanes and tropical storms.
Figure 2. Early August Statistical Model Predictors
Figure 3. Observed versus hindcast values of post-31 July NTC for 1979-2014 using the statistical model.
Like all of the statistical models, the August model did not predict the very low activity experienced in 2013. It did an excellent job predicting the below-average seasonal levels that occurred in 2014. One challenge is deciding whether to use longer-term datasets for model development or more reliable shorter-term datasets. The problem is that the farther back in time one goes, the more uncertainty there is in both hurricane statistics as well as large-scale prediction information (e.g., SSTs, wind speeds, etc.). Consequently, the choice is to have a model with more realizations of how the atmosphere/ocean operates with less certainty in the data, or fewer realizations of the atmosphere/ocean system with more data certainty. For example, plotted below are July upper-level winds over the tropical Atlantic from the NCEP/NCAR Reanalysis. This product has an unphysical strengthening of upper-level westerly winds from the late 1940s to the late 1960s. This would preclude the use of upper-level winds in the tropical Atlantic as a predictor prior to the early 1970s.
Figure 4. July tropical Atlantic upper-level winds
The statistical model for the remainder of the 2015 Atlantic hurricane season came in quite low, primarily due to unfavorable upper-level winds over Africa, as well as stronger-than-normal trade winds over the Caribbean. Both of these conditions are typical of strong El Niño conditions, similar to what is currently being experienced. The output from the statistical model is displayed below:
Figure 5. Statistical model output
I also utilize an analog technique, whereby I look in the past for years that had similar conditions to this year. Our primary qualification for this year's analogs is that strong El Niño conditions must be present. When this qualification is made, our best analog years are: 1965, 1972, 1982, 1987 and 1997. Four of the five years listed had well below-average hurricane activity, with the notable exception of 1965. Not only did 1965 had two major hurricanes, one of these was Hurricane Betsy, which did massive amounts of damage to New Orleans. A summary of activity for these five years is listed below:
Figure 6. Analog years selected for the 2015 Atlantic hurricane season.
Our final forecast for the 2015 Atlantic hurricane season is a combination of our statistical model, our analog model as well as any adjustments for other factors not explicitly considered in our statistical scheme. This seasonal forecast is the lowest August seasonal forecast that our project has ever issued, except 1991, where our project predicted zero major hurricanes:
Figure 7. Final August seasonal forecast values.
The low seasonal forecast is due to a couple of factors. The primary factor, as noted earlier, is the presence of a strong El Niño. For a full discussion of this year's El Niño event, please check out my earlier blog:
http://www.wunderground.com/blog/PhilKlotzbach/co mment.html?entrynum=1
El Niño implies warmer than normal SSTs in the eastern and central tropical Pacific, which can be clearly seen in the figure below:
Figure 8. Current Pacific sea surface temperature anomalies
As mentioned in the introduction, El Niño tends to increase vertical wind shear in the tropical Atlantic and the Caribbean. Shear throughout the Caribbean has been very strong over the past couple of months, with record high levels experienced in July.
Figure 9. July Caribbean shear anomalies
In addition, the tropical Atlantic has been cooler than normal over the past few months. Cooler than normal waters mean less fuel for developing storms, as well as generating pressure gradient patterns that typically drive stronger low- and mid-level winds, promoting dry air outbreaks from Africa. The first figure below shows how much cooler the tropical Atlantic and Caribbean are than they have been over the past twenty years, which were typically associated with active hurricane seasons. The second figure below shows mid-level moisture anomalies across the Atlantic, which are much drier than normal:
Figure 10. July 2015 SST minus July 1995-2014 SST
Figure 11. Mid-level relative humidity anomalies
All in all, conditions are incredibly harsh across the Atlantic basin this year. While overall levels of hurricane activity are expected to be quite low this year, it only takes one system making landfalling near you to make it an active year. For example, 1983 is considered to be the quietest hurricane season since 1950. Only four tropical cyclones formed all year, but one (Hurricane Alicia) made landfall near Houston and did considerable damage. 1992 was also a well below-average season and had only one major hurricane, which happened to be Hurricane Andrew. It must be emphasized that now is the time to prepare for potential hurricane impacts, as it only takes one!
For the full seasonal forecast, please check out:
August Seasonal Forecast
For frequent updates on tropical weather throughout the year, feel free to follow me on Twitter at @philklotzbach.
CSU began issuing seasonal forecasts in 1984, when Dr. Bill Gray noted a significant relationship between El Niño (warmer than normal sea surface temperatures (SSTs) in the eastern and central tropical Pacific) and Atlantic hurricanes. When El Niño conditions are present in the Atlantic, vertical wind shear levels increase in the Atlantic, tearing apart tropical cyclones that are trying to develop.
Figure 1. Typical vertical wind profile observed in El Niño and La Niña years in the Main Development Region.
Since that time, the forecast model has been expanded to utilize a variety of parameters that include Atlantic basin sea surface temperatures, sea level pressure anomalies and low-level wind anomalies. The primary tool that is used for making the various predictions is a statistical model based on historical data. In the case of the August seasonal forecast, three predictors are utilized, that when used in combination, explained significant amounts of variability over the period from 1979-2012. The target predictor is Net Tropical Cyclone (NTC) activity. This parameter convolves the frequency and intensity of tropical cyclones into one number, where the 1950-2000 average of this parameter is 100% by definition. It weighs major (Cat. 3-5) hurricanes more strongly than Cat. 1-2 hurricanes and tropical storms.
Figure 2. Early August Statistical Model Predictors
Figure 3. Observed versus hindcast values of post-31 July NTC for 1979-2014 using the statistical model.
Like all of the statistical models, the August model did not predict the very low activity experienced in 2013. It did an excellent job predicting the below-average seasonal levels that occurred in 2014. One challenge is deciding whether to use longer-term datasets for model development or more reliable shorter-term datasets. The problem is that the farther back in time one goes, the more uncertainty there is in both hurricane statistics as well as large-scale prediction information (e.g., SSTs, wind speeds, etc.). Consequently, the choice is to have a model with more realizations of how the atmosphere/ocean operates with less certainty in the data, or fewer realizations of the atmosphere/ocean system with more data certainty. For example, plotted below are July upper-level winds over the tropical Atlantic from the NCEP/NCAR Reanalysis. This product has an unphysical strengthening of upper-level westerly winds from the late 1940s to the late 1960s. This would preclude the use of upper-level winds in the tropical Atlantic as a predictor prior to the early 1970s.
Figure 4. July tropical Atlantic upper-level winds
The statistical model for the remainder of the 2015 Atlantic hurricane season came in quite low, primarily due to unfavorable upper-level winds over Africa, as well as stronger-than-normal trade winds over the Caribbean. Both of these conditions are typical of strong El Niño conditions, similar to what is currently being experienced. The output from the statistical model is displayed below:
Figure 5. Statistical model output
I also utilize an analog technique, whereby I look in the past for years that had similar conditions to this year. Our primary qualification for this year's analogs is that strong El Niño conditions must be present. When this qualification is made, our best analog years are: 1965, 1972, 1982, 1987 and 1997. Four of the five years listed had well below-average hurricane activity, with the notable exception of 1965. Not only did 1965 had two major hurricanes, one of these was Hurricane Betsy, which did massive amounts of damage to New Orleans. A summary of activity for these five years is listed below:
Figure 6. Analog years selected for the 2015 Atlantic hurricane season.
Our final forecast for the 2015 Atlantic hurricane season is a combination of our statistical model, our analog model as well as any adjustments for other factors not explicitly considered in our statistical scheme. This seasonal forecast is the lowest August seasonal forecast that our project has ever issued, except 1991, where our project predicted zero major hurricanes:
Figure 7. Final August seasonal forecast values.
The low seasonal forecast is due to a couple of factors. The primary factor, as noted earlier, is the presence of a strong El Niño. For a full discussion of this year's El Niño event, please check out my earlier blog:
http://www.wunderground.com/blog/PhilKlotzbach/co mment.html?entrynum=1
El Niño implies warmer than normal SSTs in the eastern and central tropical Pacific, which can be clearly seen in the figure below:
Figure 8. Current Pacific sea surface temperature anomalies
As mentioned in the introduction, El Niño tends to increase vertical wind shear in the tropical Atlantic and the Caribbean. Shear throughout the Caribbean has been very strong over the past couple of months, with record high levels experienced in July.
Figure 9. July Caribbean shear anomalies
In addition, the tropical Atlantic has been cooler than normal over the past few months. Cooler than normal waters mean less fuel for developing storms, as well as generating pressure gradient patterns that typically drive stronger low- and mid-level winds, promoting dry air outbreaks from Africa. The first figure below shows how much cooler the tropical Atlantic and Caribbean are than they have been over the past twenty years, which were typically associated with active hurricane seasons. The second figure below shows mid-level moisture anomalies across the Atlantic, which are much drier than normal:
Figure 10. July 2015 SST minus July 1995-2014 SST
Figure 11. Mid-level relative humidity anomalies
All in all, conditions are incredibly harsh across the Atlantic basin this year. While overall levels of hurricane activity are expected to be quite low this year, it only takes one system making landfalling near you to make it an active year. For example, 1983 is considered to be the quietest hurricane season since 1950. Only four tropical cyclones formed all year, but one (Hurricane Alicia) made landfall near Houston and did considerable damage. 1992 was also a well below-average season and had only one major hurricane, which happened to be Hurricane Andrew. It must be emphasized that now is the time to prepare for potential hurricane impacts, as it only takes one!
For the full seasonal forecast, please check out:
August Seasonal Forecast
For frequent updates on tropical weather throughout the year, feel free to follow me on Twitter at @philklotzbach.
No comments:
Post a Comment