Note the candidate's conclusion: "It is also possible that a time series – whether it be AR(p), MA(q), ARMA(p,q) or otherwise – is simply not the best way to fit this data.  Changes in the global climate such as El Niño / La Niña, sea surface temperatures, and even dust from the Sahara Desert can drastically affect the formation of tropical cyclones in the Atlantic from year to year.  If these conditions change, the number of past storms may not be an indicator of future storms and a time series is no longer appropriate." Distributions with few occurrences that depend on various external forces are often not well modeled by an ARIMA process. Your student project should not force an ARIMA model onto the data if it is not appropriate.