Sahelian Rainfall and its Correlation to Tropical Cyclone Activity
Sahelian rainfall patterns have long been suspected to have an affect on tropical cyclones in the Atlantic. Very recently researchers began to look deeper into this topic, but much remains unquantified and unknown. I have decided to look as deeply as possible into this topic to ascertain quantified trends, as well as gain a better understanding of changes in the Atlantic over time.
The data sets I used to do this study are tropical cyclone archives and the Sahelian rainfall index (courtesy of NineInchNails). I divided the tropical cyclone data set into three categories: number of named storms in the Atlantic per year, number of Cape Verde storms in the Atlantic per year (storms that formed east of 45 W and south of 25 N), and the number of storms generated by easterly waves per year. I feel each of these will reveal interesting results. I divided the Sahelian rainfall index data set into July though October averages (the main rainy months in the Sahel, as well as the main months for tropical wave genesis) and into the average of the past year.
To start, the 3 subsets of tropical cyclone data seemed to be proportional to each other. This is expected, here is a graph where that is easily visible.
Rather, the important fact seen here is the increase of tropical cyclones over time. The below scatter plot illustrates this. The increase has been very significant, from an average of about 8 storms to now approaching 15.
Now to look at the Sahelian rainfall index data sets. The data sets showed a surprising correlation, which suggests that the rainfall for the African Sahel during the main rainy months can be forecasted by looking at the past years rainfall. This method wouldnt be perfect however.
As can be seen on both charts, there was a massive plunge in index values in the late 60s that continues to this day. Perhaps there has been some slight improvement, but I stress slight.
Now to compare the tropical cyclone data to the Sahelian rainfall index data, to see if trends can be found between them. To start, I ran averages of the tropical cyclone types when there were positive anomalies and negative anomalies.
In both data sets, the correlation between Sahel rainfall and Cape Verde systems, as well as systems spawned by Cape Verde waves, is minimal if the link exists at all. There appears to be a significant link to overall named systems though in both data sets, with 17% more tropical cyclones in below 0 years in the J-O data set and 25% more tropical cyclones in below 0 years in the previous year data set. I'm going to try and isolate this trend.
I will use the previous year Sahel Index data set to isolate the trend, as it showed the greater correlation. I first took the years 1958 to 1967, the only years when the Sahelian rainfall index trended upward, and compared it to named storms. I exaggerated the Sahel index data so it can be easily seen. The graph is below.
As can be seen, as the index trended upwards, the # of storms trended downwards.
The next graph is the years 1968 to 2006, which were the below 0 years. Once again I exaggerated the Sahel Index so changes in it could be seen.
In this graph, it can be seen as the index trended downwards the number of storms< trended upwards.
Thus, a very significant trend has been found within the data. As the Sahel Index rises the # of storms decreases, and as the Sahel Index decreases the # of storms increases.
Correlation does not mean causation, and alot more research will have to be undertaken to determine the exact cause of this extremely interesting trend. I will attempt to discover the cause over the next few months. I hope you guys have found this as interesting as I have found it, and please do not copy this off to somewhere else without my permission.
Written By Zachary Gruskin For Hurricane Warning 4/22/07
