Finding a Needle in a Climate Haystack

Abstract:

We are studying the regional impact of volcanic eruptions that might have caused global cooling using high-quality annual-resolution proxy records of natural phenomena, such as tree-ring widths, and cultural events, such as the dates of the beginning of grape and rye harvests. To do this we need to determine if the year following an eruption was significantly colder and wetter than preceding or subsequent years as measured by any given proxy and if that year is consistently cold and wet across different proxies. The problem is complicated by the fact that normal inter-annual variations in any given proxy can be quite large and can obscure any volcanological impact and by the fact that inter-annual variations for different proxies will have different means and standard deviations. We address the first problem by assuming that on a regional scale, the inter-annual variations of different proxies are at best only weakly correlated and that, in the absence of a volcanological signal, these variations will average out on a regional scale. We address the second problem by renormalizing each record so that it has the same mean and standard deviation over a given time interval. We then sum the re-normalized records on a year-by-year basis and look for years with significantly higher total scores. The method can also be used to assess the statistical significance of an anomalous value. Our initial analysis of records primarily from the Northern Hemisphere shows that the years 1601 and 1816 were significantly colder and wetter than any others in the past 500 years. These years followed the eruptions of Huayanaputina in Chile and Tambora in Indonesia, respectively, by one year. The years 1698 and 1837 also show up as being climatologically severe although they have not (yet) been associated with specific volcanic eruptions.