Why Should I Care?

This time a year, those involved with forecasting tropical storms and hurricanes in the Atlantic Ocean tend to get busy.  On June 1st we kicked off the official start of the 2009 hurricane season for the Atlantic, and while we have not had any name storms yet there was already a close call (Tropical Depression 1).  Generally speaking however, the main activity for the season does not usually get under way until August, so for most folks what you tend to hear about this time of year is the latest forecasts for the projected seasonal activity.  With all the significant landfalling hurricanes along the US gulf coast the last few years, these forecasts have gotten a great deal of attention in recent years.  However, do these predictions real provide value?  And if so, for whom?

Each year a group of organizations issue their projections for how many hurricanes and tropical storms we can expect in the current season.  These groups include the US Climate Prediction Center (CPC), Colorado State University (CSU), Tropical Storm Risk (TSR) and the UK Meteorology Office (UKMT).  A couple of these organizations issue an initial forecast in April and then revise their forecast right around the start of the season or during its first week.  While the others just wait until the first month of the season to make their projections.  The image to your right shows the forecasts made by these entities to date and compares them with the seasonal average for the last 10 years of 15 tropical cyclones per season.  There are a couple of things that probably strike you right away – the variance that has occurred in just two months and the broad number range of the forecasts.  Which leads one to ask – why are they different? why such a change in just two months? and why can’t they all just forecast a number (like CSU)?  Now that first question is an item that is worthy of its very own post, but just imagine watching your local television stations and seeing that each is forecasting a different temperature range and chance of rain for tomorrow.   This happens because each forecaster puts more trust in a source or sources of information than does a different forecaster.  They may even have access to different types of information.   But in the end it will generally lead them to different forecast, except in those case where all the source information are agreement.

When dealing with these seasonal forecasts we are considering the large scale weather behaviors that set the table for favorable formation (forecasting actual individual storm formation is a whole different challenge).  But it is not like we can look to one magical source of climate control and expect it provide all the answers.  These forecasters look to a variety of predictors to reach help in driving their forecasts.  One example of these predictors is El Niño/La Niña.  The state of this oscillating region of sea surface temperatures in the Pacific is actually known to impact large scale weather behaviors in the tropical Atlantic region.  In addition to being just one potential indicator of behavior, because we are most interested in how it will be have during the core of the hurricane season (Aug to Oct); we actually need to predict its behavior as well.  Unfortunately, predicting El Niño/La Niña also has its own challenges (sometimes referred to as the spring barrier).  This is an example of why predictions change between April and June.

Now imagine multiplying the El Niño / La Niña challenges across all the various predictors.  This leads to the challenge of how do we properly bound the limits of this forecast.  Well if the predictors have lots of uncertainty, the reasonable thing to do is show this wide range of uncertainty, hence these wide range estimates.  These sorts of estimates even happen with your local daily forecast.  You may be presented with a number for a high temperature, but usually that is positioned logically within a range of probable high temperatures.   These bounds help place a level of confidence around a forecasted value.  In general, a highly confident forecast would have very small ranges.  However, as you can see with tropical cyclone seasonal forecasts that these ranges tend to be +/- 3 tropical cyclones or greater.  They also tend to be greater for the April forecasts because these have even more uncertainty.

So one would wonder, do these forecasts actually provide value?  Well, that very much depends on your perspective – Do you need a know the number of tropical cyclones within +/-  1 or so?  Do you need to know if a hurricane is going to make landfall near where you live or have a vacation home?  If so, you are out of luck with our current predictive ability.   However, for researchers and risk analysts there is value to be gained from these forecasts. 

Another reasonable question is are these forecasts any good?  We will explore that topic in a future post, but in the mean time you can read some excellent posts by Jeff Masters of Weather Underground who has written about this issue in multiple past blog entries (Post 1, Post 2, Post 3).