Global identification of tropical cyclone candidate events in NOAA CIRES 20th Century Reanalysis Data

Accurate forecasting of tropical cyclones in the future begins with a solid understanding of where they have developed, intensified, and moved in the past. Unfortunately, even as predicting tropical cyclone genesis remains an area of special concern and interest, the historical database of genesis events that inform such forecasts remains fragmentary. Prior to the advent of the satellite era, limited in situ synoptic observation networks led to an indefinite number of tropical cyclones remaining undetected in tropical basins globally. This era of low confidence in climatological data extends through roughly the mid-1960s in the Atlantic Basin and the late 1970s in other global basins. While prior studies (Vecchi and Knutson 2008) found that this undersampling of historical tropical cyclones exists, there remains disagreement regarding its magnitude, which has led to difficulties in interpreting long-term trends in activity. These uncertainties hamper not only operational forecasters, but more broadly the ability of public and private interests to accurately quantify and manage the risks tropical cyclones pose to coastal communities.

One means by which tropical cyclone climatology can be efficiently expanded and revised is thorough the use of reanalysis datasets. Truchelut and Hart 2011 utilized the NOAA/CIRES 20th Century Reanalysis, which assimilates only surface-based observations and extends back to 1871, to develop a scheme that identified previously unknown Atlantic Basin potential cyclones in the pre-satellite era. This was accomplished through compositing reanalysis synoptic fields of historical tropical cyclones and then manually identifying similar signatures in the reanalysis that did not correspond to known Best Track cyclones. Observational verification using historical ship observations showed the technique effectively identified around 1.5 potential missing tropical cyclones per year for the 1951-1958 Atlantic Basin hurricane seasons, which were subsequently suggested for addition to official records of tropical climatology.

This research expands on both the spatial and temporal scope of the earlier work by using a filtering algorithm that dramatically improves the efficiency and speed with which candidate events are identified in the reanalysis dataset. This scheme is applied to the Atlantic, East Pacific, West Pacific, South Pacific, North Indian, and South Indian Ocean basins for the years 1871-2008, allowing the first quantitative and objective global tropical cyclone candidate counts to be made for the decades prior to the advent of formal climatological records. While observational verification is not complete for each of the individual candidates, the analysis performed on a subset of the events confirms that the algorithm is able to identify potential missing tropical cyclones with a success rate approaching that of the earlier technique while demonstrating significant gains in efficiency. Extrapolating these success rates to the entirety of the candidate event dataset suggests that this method has likely identified hundreds of previously unknown tropical cyclones worldwide for future study and cataloging. As such, these results are a promising basis upon which to advance our understanding of long-term trends in tropical cyclone activity in order to improve the efficacy of operational and seasonal forecasting.