The research team at Colorado State University uses a few more advanced metrics than I do to come up with a seasonal tropical forecast, but many of the premises are the same. One area my “light” study does not focus on is the expected landfall zones this year. Rather, I will just attempt to put a number on the forecast and show you some insight on why I forecast the amount of storms that I do.
My 2013 Seasonal Forecast calls for 15 named storms, 4 Hurricanes and 2 Majors.
Background:
First my data set will be from a time frame of 1960 to 2012. This will keep most of my data confined to a time within the “satellite era” of weather forecasting. Secondly, there are a few atmospheric signals that I will use within those years to narrow down the data set even further. Those “signals” in my research will be
- Negative Analog Winter Temps (cold AO)
- Sustained (or back-to-back) Neutral ENSO Years
- Similar ENSO Anomaly Years
- Equivalent Sunspot Years
Data Sets:
AO years are the simplest data set to pull, and includes the cold winter seasons of 1963, 1985, 1996, 2000, and 2010
There are numerous back-to-back neutral ENSO events over the last 50 years, including lengthy blocks from the 70′s, 80′s and 90′s. Specifically, the data set features 1977, 1978, 1979, 1980, 1981, 1983, 1984, 1985, 1993, 1994, 1995, 2004, 2005 and 2006.
Current forecast anomaly for this year’s neutral ENSO is around -.3º. This helps narrow down the ENSO events even more for my study. Those years include 1961, 1966, 1978, 1983, 1984, 2001.
Finally, I introduced a new piece of theory into the science of hurricane forecasting, sunspots. We find ourselves currently in a sunspot minimum year, but we have seen a cyclical pattern develop over time. Sunspot research is relatively new, so my data set is not quite as deep as I would like, but does include similar sunspot minimum years of 1974, 1977, 1985, 1994, 1997, 2003 and 2004.
Linkage:
I used three different methods to narrow down all of the data in hand. First, I took every year separately and analyzed those hurricane seasons individually. Next, I re-classified the yearly data sets into their categories of either negative AO winters, ENSO neutral years and/or sunspot years. Some of the years fell into multiple categories, while others stood independently. Doing this enables me to fine tune which years I want to analyze closer for the 2013 Seasonal Forecast.
Field Narrowing:
For my study, I narrowed the data down to the years of 1977, 1981, 1985, 1994, 1996, 2004, and 2006. Each of these years had multiple listings within an ENSO, sunspot or negative AO category. From this point, the research is simple: average out the number of named storms and hurricanes from this 7 year sample.
Conclusions:
The 7-year sample yields an average of 10.4 named storms, 6.5 hurricanes and 2.85 major hurricanes per year. You may note that this sample is lower than my seasonal forecast of 15 named storms, 4 hurricanes and 2 major hurricanes. Why the contrast? Simple! We are still in a period of increased hurricane activity (multidecadal variability) as well a period of enhanced satellite detection. I have essentially attributed 2 extra named storms to the multidecadal variability and 2 more named storms for our advances in remote sensing and studying open-ocean cyclones.
To contrast, I am forecasting lower hurricane and major hurricane activity right now due to the near or below normal water temperatures in tropical breeding grounds of the western Atlantic Ocean. The eastern Atlantic remains very warm into early April, but is not considered a true tropical breeding zone until August and September. Because of this, I am “subjectively” lowering hurricane estimates by 2 and cutting the major hurricane forecast by 1.
Limitations:
My research is essentially a data set pull and is not proven to be a statistically significant way to forecast tropical weather. I also do not incorporate any new data from the Atlantic thermohaline circulation theory the research team at Colorado State uses. I also do not incorporate any African monsoon “predictors” into my research nor do I use any hindcasting validation.
Regardless, I hope I have provided some insight into just “some” of the data analysis that goes into a seasonal forecast. This is a unique challenge to actually come up with, and stay firm with specific numbers. Maybe years from now our forecast skill will become even better that we can pinpoint specific areas of coastline that may be at high risk.





















