Today's second installment in the Learning Series deals with seasonal hurricane forecasting with an emphasis on the North Atlantic basin. Seasonal forecasting took off in the 1980s with the advent of the forecasts from Dr. William Gray (Colorado St. University) and has expanded in recent years to encompass many groups in academia (Colorado St., Florida St., TropicalStormRisk and others) and the private sector (AccuWeather, energy firms and others). Each group has different factors that they look at when making their forecasts – some factors of which are proprietary and not public knowledge – but, generally speaking, they come to similar forecasts as the underlying principles are the same. The article starts off with the basic factors that cause tropical cyclones to form and then moves off into a discussion of some of the parameters that two groups in particular use in making their forecasts. I close with a discourse on verifying these forecasts and a look at new frontiers being explored when it comes to seasonal hurricane forecasting. My primary resources for the discussion of parameters used come from the project webpages of Bill Gray at CSU and TropicalStormRisk.

It is well known that for tropical storms to form, you need something at low levels (a tropical wave or a convergence zone, for instance) that can draw moisture and heat from the oceanic surface. The oceanic surface must be warm enough – above about 26 deg C – to have sufficient heat to drive the cyclone, while a low vertical shear environment is needed to allow the storm to grow vertically and concentrate this energy in one location. The system can't be too close to the equator, must have some weak instabilities present, and can't be located in an overly dry (or absolutely moist) environment. These basic fundamentals, particularly SSTs and shear, form the basis for many seasonal and day-to-day forecasts of tropical cyclone formation. It is straightforward to apply them on a day to day basis: we know with some certainty where the favorable parameters are going to line up and best allow for tropical cyclone formation, with the Tropical Cyclone Formation Probability Product (http://www.ssd.noaa.gov/PS/TROP/genesis.html) a good example of this. But, it is very difficult to apply these parameters on long time scales with the exception of SST and, to a lesser extent, shear as our computer models are not advanced or accurate enough to forecast all of these parameters months in advance.

So, we start our discussion with SST and shear. SSTs are slowly varying features with both annual and multi-decadal cycles. Independent of whether they are being affected by global warming, SSTs have been on the increase lately as a result of the multi-decadal active period in the Atlantic (sometimes referred to as the AMO). With several months lead time and a winter's worth of storm activity in the past, SSTs can be reasonably well forecast into the summer months. Wind shear is a highly localized, highly varying (in both time and space) feature but also one that can be generally accounted for over the basin using an average. Seasonal forecasts don't necessarily take wind shear into account directly; instead, they use El Nino cycles as a proxy for whether the shear will be higher (El Nino) or lower (La Nina) than normal during the tropical season. In fact, El Nino cycles form a large basis – along with SSTs – of seasonal forecasts as they have a reasonably good predictability (save for events like 2006) and account for a large amount of the variance in year to year tropical cyclone activity.

Beyond these factors, numerous others are used to formulate seasonal forecasts anywhere from 1 to 7 months prior to the start of the hurricane season. For instance, Dr. Gray's team at CSU has recently used a combination of fall sea level pressure and mid-level height patterns over the north Atlantic and northeast Pacific along with stratospheric winds in the tropical western Pacific to forecast the following tropical season's activity. More recently, they have shifted to one that considers only eastern Atlantic SSTs and sea level pressures along with central Pacific sea level pressures. Each method has been shown to account for about half of the variability in year to year seasonal activity, which is either very good or not too good at all depending upon your perspective. It is something that has skill, but also something with a lot of room for improvement. Meanwhile, the group at TropicalStormRisk uses SSTs and ENSO cycles along with measures of trade winds across the North Atlantic. Their methods are newer but don't necessarily have less skill.

Two questions you might be thinking of here are 1) how did they arrive at these variables? and 2) why are they important? Let's tackle each of these now. First, the variables used in these forecasts are often arrived at through both meteorological and statistical analysis, generally in that order. A research team identifies all possible factors that realistically might control or influence tropical cyclone activity (the meteorological aspect), then runs statistical analyses to determine which factors are most significant and how to best combine them into a prediction scheme (the statistical aspect). With regards to the second question, the various parameters can be important for several different reasons. Sea level pressure and height fields can both help draw conclusions as to the current and future states of El Nino cycles as well as to how SST patterns might vary in those regions months down the line. SST fields themselves can help determine how sea level pressure and height fields might vary in those regions months down the line. Wind fields can give an idea for features such as El Nino and the Madden-Julian Oscillation (MJO). In particular, the trade wind factor that TropicalStormRisk uses in their seasonal forecasts can be used for determining the strength of the trade wind inversion over the eastern Atlantic during the tropical season; weaker trade winds suggest a weaker inversion. This is critical as a weaker inversion allows for Saharan dust – a common occurrence during the season – to mix out and be less of a hindering factor for Cape Verde storms. Other predictors that have been used in the past, such as rainfall over Africa, can help give a measure for Cape Verde activity during the peak of the season. Other groups likely have other parameters based upon atmospheric variables or climate indices.

In short, there is no one right way – or one common way – to do seasonal tropical cyclone forecasting. SSTs and El Nino patterns can give you about half of the story, but we don't know enough about (or have the ability to forecast well enough) the factors that control the remaining half of the story. As a result, seasonal tropical cyclone forecasts occasionally have significant errors, particularly if the underlying conditions change during the middle of the season such as in 2006, with an El Nino event developing during the tropical season. In many ways, seasonal forecasting is still a science in its infancy but one with a lot of growth potential in the years to come. Numerous industries and governments want to know how many events they can expect during a tropical season to better allocate their money, people, and resources. With that comes a desire to know where exactly these storms might threaten during the season, one of the expanding and changing areas of seasonal forecasting as we head into the rest of the decade. Some groups such as AcuuWeather feel that they can forecast regional threats before the season even starts, but there is (currently) little skill associated with such forecasts as it is mesoscale – smaller scale – features that tend to control regionalized impacts. That is not to say, however, that there aren't features and patterns that can be used to better highlight increased or decreased regional threats before the season starts, notable amongst which is a La Nina signal that tends to lead to an increased threat to the mid-Atlantic coastline of the US, just that the science is still in its infancy.

Finally, I close with a brief digression of the SOI, or Southern Oscillation Index. The SOI is one of several ways that El Nino events can be characterized and is merely the difference in sea level pressures between Tahiti and Darwin in the tropical Pacific/Indian Oceans. Negative values of the SOI, where the pressure at Darwin is greater than that at Tahiti, are often associated with El Nino events as a result of the oceanic heating in the central Pacific (where Tahiti is) leading to lower pressures there. Conversely, positive values of the SOI, where the pressure at Darwin is less than that at Tahiti, are often associated with La Nina events. The SOI can be calculated on a daily or monthly basis and can be normalized based on some given normalizing value. For more information, see http://www.bom.gov.au/climate/glossary/soi.shtml.

If you have any questions, please feel free to send me a PM or discuss this in the Blogger Discussion Forum. I'll tackle some of the other topics proposed earlier this month in future topics – probably as we move into April and May.