Climate indices in historical climate reconstructions: a global state of the art

Abstract

Abstract. Narrative evidence contained within historical documents and inscriptions provides an important record of climate variability for periods prior to the onset of systematic meteorological data collection. A common approach used by historical climatologists to convert such qualitative information into continuous quantitative proxy data is through the generation of ordinal-scale climate indices. There is, however, considerable variability in the types of phenomena reconstructed using an index approach and the practice of index development in different parts of the world. This review, written by members of the PAGES (Past Global Changes) CRIAS working group – a collective of climate historians and historical climatologists researching Climate Reconstructions and Impacts from the Archives of Societies – provides the first global synthesis of the use of the index approach in climate reconstruction. We begin by summarising the range of studies that have used indices for climate reconstruction across six continents (Europe, Asia, Africa, the Americas, and Australia) as well as the world's oceans. We then outline the different methods by which indices are developed in each of these regions, including a discussion of the processes adopted to verify and calibrate index series, and the measures used to express confidence and uncertainty. We conclude with a series of recommendations to guide the development of future index-based climate reconstructions to maximise their effectiveness for use by climate modellers and in multiproxy climate reconstructions.

Quantifying and reducing researcher subjectivity in the generation of climate indices from documentary sources

“…This study set out to explore and quantify the degree of error between individuals generating ordinal-scale climate indices from a historical documentary dataset. Ultimately, our findings reinforce the conclusions of Nash et al (2021) that climate indices should be derived by those with training in historical climatology methods, as well as a detailed knowledge of both the climate of the area in question and the nature of the documentary sources. However, we have demonstrated that generating such indices using groups of two or more researchers will substantially reduce the issue of researcher subjectivity.…”

“…Raters were also asked to provide a brief explanation of why they assigned each rain-year to a particular category, and a summary of rainfall conditions. This approach to index development represents a variation on the 'Pfister method' (see Pfister et al, 2018;Nash et al, 2021), using five categories rather than seven, and generating a series for an entire season rather than monthly. Training Partnership.…”

“…There is no consistent global approach to climate index generation (Nash et al, 2021), the closest being the method developed by Christian Pfister in the 1980s for the creation of Swiss temperature and precipitation series (Pfister, 1984); a method widely adopted across Europe and, in the case of rainfall reconstructions, across Africa (e.g. Vogel, 1989;Nash and Endfield, 2002), Australia (Fenby and Gergis, 2013) and India (Adamson and Nash, 2014).…”

“…Here, the accuracy of any index series depends upon the ability of the analyst to interpret and evaluate collections of raw weather descriptions and then categorise them into appropriate classes. As Nash et al (2021) note, transforming information from historical documents to numbers on a scale requires considerable expertise to minimise subjectivity. Consequently, this should ideally be undertaken by researchers with a good scientific knowledge of the regional climate, and an understanding of the language of the time period during which the sources were written.…”

A Bayesian Approach to Historical Climatology for the Burgundian Low Countries in the 15<sup>th</sup> Century

“…Historical texts, both descriptive sources, such as chronicles, and written records of phenology, such as grape harvest dates, have enabled high-resolution reconstructions of temperature and precipitation for periods prior to the modern instrumental record (White et al, 2018). The principal approach for these reconstructions has been the "index" method (Pfister et al, 2018;Nash et al, 2021). Historical climatologists working with these texts have converted their information into ordinal indices (typically -3 to +3), which approximate departures from average monthly or seasonal conditions, and they have calibrated these indices to early instrumental data to obtain reconstructed values for the pre-instrumental period.…”