The Jenkinson data are presented as annual totals expressed as a percentage of the mean annual average rainfall (AAR) for the period 1826–1975. Monthly values are provided as a proportion of the rainfall for each year. Annual anomalies and proportionate monthly values, presented in the Appendix of Jenkinson et al. (1979), were transcribed using double keying (by two of the authors) to minimise the risk of transcription errors. Table 1 provides an overview of the various data sources comprising the Jenkinson series. Figure 1 maps the location of contributing sources from Ireland and the UK, while Fig. 2 plots the number of constituent sources per year underpinning the series, along with annual anomalies (relative to 1826–1975) contained in the note for the years 1711–1977.

The earliest instrumental observations in the Jenkinson record originate from a rudimentary gauge operated by Thomas Neve in Derry during the period 1711–1725. Given the scarcity of observations in the early part of the Jenkinson record, at times annual totals were derived from regional series from the UK. For instance, during the period 1757–1839 data from NW England were used as a proxy Irish station. Similarly, Scottish data were used for the years 1757–1977. Prior to 1811, the number of sources is typically less than four, while for the years 1728–1756 and 1766–1791 the derived series is based on single sources – the weather diary of John Rutty (1770), who compiled monthly and seasonal weather summaries for Dublin (1716–1765), and a regional series comprising data from NW England and SW Scotland, respectively. The latter were compiled by Jenkinson et al. (1979) using available data from the UKMO. From the 1790s onwards, instrumental observations from across IoI became available, including stations at Dublin, Kilkenny, Derry, Belfast, Armagh, Cork and Sligo.

The methods employed by Jenkinson et al. (1979) to derive annual totals and proportionate monthly values are incompletely documented, but appear similar to those of Glasspoole (1925) and Nicholas and Glasspoole (1931) in deriving regional rainfall series for the UK and Ireland (see Appendix 1 in Wigley et al., 1984 for description). For available stations, Jenkinson et al. (1979) expressed monthly values as a fraction (percentage) of the annual value for each station, and these fractions were then averaged over available stations. Regional estimates of annual totals were constructed by taking an overlapping period of 20–50 years with the Glasspoole series for Ireland (1881–1924) and continuing this overlapping back to the earliest records. According to Jenkinson et al. (1979), overlapping periods were varied according to the available contributing stations with longer period records taken as optimum. For each overlapping period, estimates of AAR were derived iteratively to provide: (i) station AAR, (ii) station annual percentages of AAR and (iii) regional annual percentages of AAR (Jenkinson et al., 1979). Only the latter are provided in the Jenkinson note. For any station, estimates of AAR were taken as the arithmetic mean of the available annual totals. Individual years were recorded as percentages of this AAR. The mean of the percentage values for a given year across available stations was then taken as the estimate of the regional percentage of AAR for that year. Finally, Jenkinson et al. (1979) used the mean for all data years to derive the final estimate of the regional AAR (see description of methods in the original document in the Supplement).

The diaries of Thomas Neve (Dixon, 1959), Nicholas Blundell (1968) and John Rutty (1770) are critical sources for the early Jenkinson series. The Neve diary contains estimates of rainfall totals from Derry taken from a rudimentary gauge (described by Dixon, 1959). Nicholas Blundell's diary provides monthly and annual weather summaries for Crosby (Liverpool) in NW England for the years 1711–1715 and 1718–1727. The Rutty diary contains a chronology of weather in Dublin from 1716–1765, including monthly and seasonal weather summaries. To convert the qualitative descriptions of Blundell and Rutty to a quantitative measure of rainfall, Jenkinson et al. (1979) applied a graded scaling system, similar to Brázdil et al. (2010a) and Gimmi et al. (2007), to both diaries. Scores ranging from one (exceptionally dry) to nine (exceptionally wet) were used to rank individual months and aligned with seasonal summaries. Frequencies of rankings for each month over the period 1716–1765 were derived and internal percentages for each month estimated by comparison with frequencies of monthly rankings for the period 1840–1889 (Jenkinson et al., 1979). These values were used as a percentage of AAR for each month in the period 1716–1765. Monthly and annual percentages were processed to give internal percentages for each year, with the annual percentages of AAR (1825–1977) used to represent a single station (see the original Jenkinson et al., 1979 note in the Supplement). This combination of methods for quantitative and qualitative sources was used by Jenkinson et al. (1979) to produce the final IoI series of annual totals, expressed as a percentage of AAR for the period 1826–1975. The final series was calibrated against Glasspoole's EWR series, although details of this step are not provided by Jenkinson et al. (1979). Moreover, Jenkinson et al. (1979) report annual anomalies relative to 1826–1975, but this AAR value does not appear in the publication.

The IoI_1850 series is used to estimate this missing AAR and thus to reconstruct the full series. With the exception of the 24 years prior to 1850, there is a large overlap between the normal period applied to the Jenkinson series and the IoI_1850 series. Matching annual anomalies derived from the IoI_1850 series using the normal period of 1850–1975 with the anomalies derived from the Jenkinson series for concurrent years, reveals a strong positive correlation (Spearman's r=0.95). The time series of both sets of anomalies is plotted in Fig. 2c. Resampling was used to derive estimates of long-term AAR values from 1000 samples (with replacement) of long-term (100-year) AAR drawn from the IoI_1850 series over the period 1850–2015. Each AAR estimate was used to reconstruct annual totals and subsequently monthly rainfall totals for the years 1711–1849. Post-1850 data remain the homogenised IoI_1850 series of Noone et al. (2016). Resampling the missing AAR from IoI_1850 has two advantages. First, it ensures that the Jenkinson series is mean-adjusted to the homogenised IoI_1850 series. Second, confidence bounds can be generated for the reconstructed series to convey the uncertainty in estimating the value of AAR. The combined dataset is named the IoI_1711 series and provides a continuous monthly rainfall record from January 1711 to February 2016. Subsequent analysis is performed on the median of the IoI_1711 series.

England Wales Rainfall (EWR): Nicholas and Glasspoole (1931), building on earlier work by Symons (1866), presented a regional monthly precipitation series for the UK (1727–1931). Jenkinson et al. (1979) used annual EWR data to calibrate their series for Ireland, so there are obvious circularities. However, we include the years 1727–1765 prior to the modern England and Wales Precipitation (EWP) series that begins in 1766 (see below). Early EWR data were analysed by Jones and Briffa (2006) and provided by Phil Jones. We treat the EWR series by simply appending it to the EWP series (where EWR and EWP are appended in such manner we refer to the resultant series as EWR/EWP). Wigley et al. (1984) note inhomogeneities in the annual totals in the early EWR record, whereby there is a tendency to underestimate precipitation relative to EWP prior to 1870. Moreover, Tabony (1980) suggests that prior to 1820 EWR is unreliable. Knowledge of these discrepancies is useful for assessing the IoI_1711 series. The stations comprising the early EWR series are unclear and there is a need to reconcile the early EWR series of Glasspoole (1925) with the earlier Symons (1866) work. Glasspoole (1925) mentions that from 1727–1756 at least two stations were available, and from 1757–1774 at least three. Craddock (1976) provides a comprehensive list of stations that would have been available for this period of time – certainly more than used by Glasspoole (1925). Table 3 provides a list of stations that were operational (with at least one year of data) in England (note none listed for Wales) between 1725 and 1766. This covers the early period in the IoI_1711 series up to the start of the modern EWP series when traceability is clearer. This list of stations is used to identify potential circularity with the IoI_1711 series.

England and Wales Precipitation (EWP) Series: the England and Wales Precipitation series from 1766 onwards (Wigley et al., 1984; Alexander and Jones, 2001) and data for key EWP stations prior to 1766 are also employed. While EWP runs from January 1766 to present, some constituent records extend into the 1700s (e.g. Carlisle, Kew, Spalding at Pode Hole). The early part of these records often contains data from the stations listed in Table 3. However, many have also undergone adjustments to correct uncovered quality issues in the early series. An overview of each series is given in the following paragraphs. The EWP series was accessed via the UKMO website (https://www.metoffice.gov.uk/hadobs/hadukp/).

Kew Gardens Precipitation (Kew): routine rainfall measurements have been taken at Kew since 1871. Prior to this, annual and monthly totals were estimated by Wales-Smith (1971) back to 1697. During the years 1717–1724 and 1725–1744, the Richmond diaries and the diary of George Smith respectively, were used to derive estimates of monthly totals calibrated to observational series at Upminster, Fleet Street and Tonbridge (all in Table 3). For the remainder of the 1700s, data from Tonbridge, Lambeth and Somerset House feature strongly and the former two are noted in the series available to EWR in Table 3. However, monthly and annual totals for 1697–1870 were later revised by Wales-Smith (1980). Data used here were provided by Todd et al. (2013).

Spalding (Pode Hole) precipitation (Spald): the early record for Spalding (1726–present) is a composite of records from the East Midlands developed by Craddock and Wales-Smith (1977) (including Southwick (Oundle) 1726–1736 and Lyndon 1737–1798 from Table 3), while much of the record for the 19th Century derives from South Kyme and Pode Hole thereafter. Tabony (1980) corrected the series for being overly wet during 1770–1870. The quality of the pre-1770 record is largely unknown. Data used here were provided by Todd et al. (2013).

Carlisle Precipitation (Carl): the Carlisle series (1757–present) for NW England was constructed as a composite of stations prior to 1872. An annual series for Carlisle was first developed by Craddock (1976) and a monthly series subsequently developed for the years 1845–1976 by Tabony (1980). The series was further extended to 1757 by Jones (1983). Todd et al. (2015) updated the Carlisle series to present, and assessed the homogeneity of the record. Data for the years 1757–1783 come from observations taken by J. Carlyle, as listed in Table 3. Rainfall totals prior to 1790 were found to be under-representative and were increased by 24 %, while annual totals during 1827–1850 were also increased by 10 % (Todd et al., 2015). Data used here were provided by Todd et al. (2015).

Oxford Precipitation (Ox): the Radcliffe Observatory, founded in Oxford in 1772, is the longest and best documented continuous series of temperature and rainfall at any single site in the UK. Daily observations are available from April 1814, and continuously since 1827 (Burt and Howden, 2011). A reconstructed monthly precipitation record, first developed by Craddock and Craddock (1977), exists from 1767. Prior to 1815, monthly totals are estimated from the manuscripts of Thomas Hornsby (often incomplete and fragmentary) and data from Shirburn Castle (12 miles from Oxford), together with estimates from other stations around Oxford (Stroud, Sunbury and Lambeth, the latter is noted in Table 3). The Oxford record is unusual in that much work has been done to ensure that the entire record is homogenous, including the early measurements from the roof of the Observatory (Burt and Howden, 2011). Data used here were provided by co-author Tim Burt.

North Atlantic Oscillation (NAO) reconstruction (L-NAO): Luterbacher et al. (2001) developed a monthly resolution multi-proxy reconstruction of the NAO Index back to 1659 and seasonal reconstructions back to 1500. Full details of the reconstruction methods and constituent data sources are given by Luterbacher et al. (2001). Issues of circularity arise when using Luterbacher et al. (2001) to interrogate the IoI_1711 series because of the inclusion of Kew, Spalding (Pode Hole), EWP and other UK records in the reconstruction. During the 1700s the number of predictors falls rapidly from approximately 40 to less than 20, so these UK precipitation series are likely to be more heavily weighted in the NAO reconstruction for this period. Data were downloaded from the Climatic Research Unit (CRU) of the University of East Anglia (https://crudata.uea.ac.uk/cru/data/paleo/naojurg/).

European seasonal rainfall reconstructions (Pauling): Pauling et al. (2006) provide a seasonal gridded (0.5∘ resolution) precipitation reconstruction for European land areas for the period 1500–1900. From 1901 to 2000 the dataset is the gridded reanalysis of Mitchell and Jones (2005). Reconstructions are developed using a variety of long instrumental precipitation series (including for Ireland and the UK; see Fig. 1 of Pauling et al., 2006) and other documentary and proxy sources. Data were downloaded from the Climatic Research Unit (CRU) (https://crudata.uea.ac.uk/cru/projects/soap/data/recon/#paul05). The earliest observations that would have been part of EWR and used by Jenkinson et al. (1979) to calibrate their series, are also used as predictors in the Pauling et al. reconstructions (e.g. Kew, Spalding). Data for grids closest to each of the 25 stations comprising the IIP network (Noone et al., 2016) were extracted and seasonal regression adjustments developed over the period 1850–2000 using each IIP station. Average seasonal scaling adjustments applied to Pauling et al. (2006) data in bridging to IIP stations were 1.03 (DJF), 0.99 (MAM), 0.91 (JJA) and 1.01 (SON). The mean of the 25 regression adjusted series were then averaged to produce an island of Ireland series from the Pauling et al. (2006) data.

Rutty (1770) weather diary (Rutty): the Rutty diary provides qualitative insight to the weather conditions experienced in Ireland during the years 1716–1765. As outlined above, the Rutty diary was used extensively in the Jenkinson series. Nonetheless, drawing on the actual descriptions in the diary is useful to help contextualise conditions in the early record. A scanned copy of the Rutty diary is available online (https://archive.org/details/achronologicalh00ruttgoog).

Nine independent datasets consisting of observational and proxy records are also employed:

Central England Lake District (CELD) precipitation series: Barker et al. (2004) developed a 200-year homogenous monthly composite series for the Central England Lake District (CELD) (1788–2016). Available stations were bridged to Grasmere High Close using regression to develop seasonal adjustment factors. Much of the pre-1850 record is drawn from sites at Kendall and Keswick. Neither are key stations in the development of EWR/EWP (although Kendall was used as a station for regression development by Wigley et al., 1984). Wilby and Barker (2016) subsequently updated the CELD series with a further revision of adjustment factors undertaken in 2017. The data used here were obtained from Wilby and Barker (2016).

Hoofddorp (1735–1973) (Hoof): monthly precipitation data recorded for the Netherlands over the period 1735–1973 (Tabony, 1981; Slonosky, 2002), downloaded through the KNMI Climate Explorer website (http://climexp.knmi.nl/).

Central England Temperature (CET) record (Manley, 1974; Parker et al., 1992): temperature data coincident with IoI_1711 is used to assess consistency of rainfall, especially in winter given that warm winters tend to be associated with wet conditions. CET data were downloaded from the UKMO website (https://www.metoffice.gov.uk/hadobs/hadcet/).

Cork annual rainfall totals (Cork): in his list of rainfall records operating before 1780, Craddock (1976) cites a number of stations for Ireland including Derry (1711–1724), taken by Thomas Neve; Castle Dobbs, Antrim (1727), taken by A. Dobbs and annual totals for Cork (1738–1748) taken by Timothy Tuckey. While the first two sources are used in the Jenkinson series, the latter for Cork is not. Annual totals for Cork (south Ireland) were published by Wakefield (1812, p. 207) and transcribed here. Neither the exact location nor the design of Tuckey's gauge is known.

Westerly Index (WI): Barriopedro et al. (2014) developed a monthly index of atmospheric circulation variability over the North Atlantic from 1685–2008. Their index is based on direct observations of wind direction from Royal Navy logbooks from 1685–1850 provided by ship movements in the English Channel. After 1850, the CLIWOC v1.5 (Garcia-Herrera et al., 2005) and the ICOADS v2.1 (Worley et al., 2005) datasets for the same area are used. The so-called Westerly Index (WI) provides a measure of the persistence of westerly winds beneath the exit zone of the North Atlantic extratropical jet-stream (Wheeler et al., 2010). Before 1850, the WI contains only one record of wind direction (measured with a 32-point compass) per day, with data available for 95 % of days (Barriopedro et al., 2014). The monthly WI used here is defined as the percentage of days per month with prevailing wind from the west (i.e. blowing from between 225 and 315∘ from true north). WI data were obtained from Dennis Wheeler.

Paris–London (PL) Index: Cornes et al. (2013) developed the Paris–London index (1692–2004) as an indicator of the state of the NAO index. This index, developed from recovered and corrected mean sea level pressure (MSLP) data from the respective cities, provides a consistent measure of westerly air flow over northwest Europe. Full details of the construction of the PL index, together with a comparison with available NAO reconstructions, are provided by Cornes et al. (2013). Only seasons with no missing data in any constituent month were employed. Data were obtained from the CRU website (https://crudata.uea.ac.uk/cru/data/parislondon/).

London SLP (L-SLP): Cornes et al. (2012) present a 300-year (1692–2007) daily series of MSLP for the city of London. Digitised data were transcribed from multiple sources, quality controlled, corrected and homogenised to represent daily means of MSLP at standard modern-day conditions. Monthly values of MSLP are not reported when missing values exceed 20 %. Here, seasonal MSLP is only derived when all three months are reported. Data were obtained from the CRU website (https://crudata.uea.ac.uk/cru/data/parislondon/).

East Atlantic/Western Russia pattern (Eurasia 2 pattern EU2) (EU2 Index): the EU2 index (Barnston and Livezey, 1987) measures the zonal pressure difference across central Europe and is important in describing the variability of Eurasian climate, especially during boreal winter (Luterbacher et al., 1999). The EU2 pattern is characterised by two main large scale pressure anomalies located over the Caspian Sea and Western Europe and is found to be closely related to Rossby wave propagation (Lim, 2015). Luterbacher et al. (1999) reconstructed monthly EU indices back to 1675, with the derived data for the EU2 index provided for this analysis by Luterbacher.

Southern England tree ring precipitation reconstruction (Rinne): Rinne et al. (2013) present a 400-year long (1613–2003), annually resolved May to August precipitation reconstruction for southern England developed from oxygen isotope measurements of tree ring cellulose in pedunculate oak (Quercus robur). Using these reconstructions (1613–1893) and instrumental data (1894–2003), their derived precipitation series has been shown to be robust back to at least 1697 – the first year of the oldest existing instrumental precipitation series in England (at Kew, see above). Data were obtained from Katja Rinne-Garmston.

Figure 6a compares decadal moving averages for each long-term precipitation series. Strong agreement between all series is evident. In the pre-1766 record, EWR reveals the lowest decadal mean annual totals, consistent with concerns that EWR underestimates early totals (Jones and Briffa, 2006). In the IoI_1711 annual series the most recent decade (2006–2015) is the wettest. Indeed, the last 50 years has seen the most persistent period of decadal mean annual totals above the 1900–1950 mean. This is also the case for the Pauling data. The driest decade in the IoI_1711 annual series was 1740–1749. The 1740s was also notably dry in the other long series covering this period. For instance, the 1740s was the driest decade on record for Spalding and second driest for Kew (after 1893–1902). Although the 1740s were notably dry in the Hoofddorp series, decades commencing from the 1790s through to the early 1800s were even drier. Figure 6b plots the early annual totals from Cork taken by T. Tuckey relative to the raw IoI_1711 annual totals. These data confirm that the 1740s were notably dry. In both series, 1740 was the driest year of the decade, followed by 1747. Indeed, in the entire IoI_1711 series, 1740 is the second driest year after 1788.

Winters (DJF) are named by the year in which January falls, so that winter 1712, for example, comprises December 1711 together with January and February of 1712. Figure 7 plots decadal mean totals for all winter series standardised to the period 1900–1950. Consistency of all precipitation series (Fig. 7a) from about 1790 onwards is noted. Prior to this there is divergence between the available precipitation series. Evident also is the long-term increasing trend in winter decadal mean precipitation totals across all series. Despite this increasing trend, the decade 1730–1739 was the wettest in the IoI_1711 winter record. Other long-term UK precipitation series that extend back this far, together with the Pauling series, do not show a remarkably wet decade. However, the 1730s are well known for their exceptional warmth within the CET record (Jones and Briffa, 2006), which we also note shows strong coherence with IoI_1711 throughout the record (Fig. 7b). Additionally, the highest decadal value in the WI winter series is found for the 1730s and the winter L-NAO is in a positive phase at this time too.

Table 6 provides the seasonal summaries of Rutty (1770) during the 1730s. Evident for winter is the description of each year as open (frost free) and warm. In 1733 reference is made to primroses and violets blooming at Christmas. The only frosty winter is noted in 1739. While winter 1730 is noted as dry, each subsequent winter in the decade (with the exception of 1737 and 1739) is described as wet, wet and windy, or wet and stormy, which is consistent with strong westerly flow. While a number of independent series support the view that winters in the 1730s were remarkably wet in the IoI_1711 series, confidence is low in the winter decadal means from 1740 to 1785. Within the context of the entire series, decadal means for 1740 to 1785 are exceptionally low with little evidence for such dry winter conditions in other long precipitation series. We note that decadal means in the CET record are cold during this time, which is consistent with a negative winter NAO and WI. While dry conditions would be expected under such winter dynamics, the deviation between CET and IoI_1711 decadal means is largest during this period. At this time, the record of Jenkinson et al. (1979) is dependent almost entirely on qualitative information from weather diaries. Despite being biased too dry, it is likely that the decade 1777–1786 was at least among the driest decades in the winter series. This is consistent with the driest winter decade in EWP (1776–1785) and Carlisle (1776–1785). The years 1776–1785 rank amongst the top 15 driest decades in the Pauling series, while closely similar years (1781–1790) are also the driest winter decade in the Oxford record. The EU2 index is also markedly positive at this time. From 1790 onwards confidence in the IoI_1711 winter record is strengthened by the high degree of consistency with all other series.

The pre-1850 IoI_1711 record shows substantial decadal variability (Fig. 8). In IoI_1711 the wettest spring decade occurs at the beginning of the series (1715–1724), while the driest is 1831–1840. The latter is consistent with EWP (driest spring decade 1833–1842) and with Oxford (third driest decade 1831–1840). Figure 8a shows strong agreement in decadal totals throughout the record from 1740 across all precipitation series. However, the Pauling series does not show the same variability as in observed series throughout the 1700s. Prior to the 1740s there are few sources for comparison. Nonetheless, there is good coherence between Spalding, EWR and IoI_1711 during this period. Strong agreement is also evident between the WI, the EU2 index and IoI_1711. A positive WI index coincides with wet spring conditions during the earliest decades of the 1700s, at the turn of the 20th Century (1895–1910), during the 1950s and 1960s, and the last quarter of the 20th Century (from the mid-1970s). Conversely, predominance of negative NAO conditions during 1715–1724 is inconsistent with reconstructed exceptionally wet spring conditions (although we note that negative NAO conditions can result in diverse climate patterns). Neither the PL index nor L-SLP series contain sufficient data for the period to place much confidence in the comparator series. Throughout the latter half of the 1700s spring decadal means are almost always below the 1900–1950 mean, consistent with persistent negative NAO conditions and negative WI for the period.

The main feature of the IoI_1711 pre-1850 record is the persistence of wet summers throughout much of the latter half of the 18th Century (Fig. 9). The 1750s (1750–1759) is the wettest IoI_1711 decade. Decadal mean summer totals for this period are also notably wet in other long-term precipitation series. For instance, 1767–1776 is wettest in the Kew series (although it should be noted that the period is not as wet and is less protracted at Kew relative to other long-term series). The years 1755–1764 mark the wettest summer decade in the Hoofddorp series. The mid-1750s were also exceptionally wet at Spalding although the 1870s were marginally wetter. We note that the Pauling series does not show wet conditions throughout this period.

Figure 9b and c plots IoI_1711 summer decadal means for L-SLP alongside indicators of westerly flow (L-NAO, PL index and WI). The wet 1750s coincides with record low MSLP at London. Much of the latter half of the 18th Century is associated with strong westerly airflow, with a persistently positive summer NAO and markedly positive values for the PL index. The EU2 index is also negative throughout the latter half of the 18th century. Unfortunately, the summer WI has a lot of missing data for this period. Table 7 shows the seasonal summaries provided by Rutty (1770) for the 1750s. With the exception of 1759 all are described as wet. Figure 9d compares IoI_1711 with Kew, Spalding and the Rinne et al. (2013) tree ring reconstructions for southern England for the months May through August. The tree ring reconstructions are consistent with Kew in terms of the wetness of the period, although later decades are wetter (e.g. decades commencing in 1836 and 1877), suggesting that southern England, while wet throughout the latter half of the 18th Century, was not as wet as other parts of the BI. The driest summer decade of the IoI_1711 series is recorded for 1968–1977.

Autumn decadal means show strong coherence across all long-term precipitation series (Fig. 10). Of note is the large decadal variability in autumn series prior to 1900, with high decadal mean totals prominent during 1750–1870. The wettest decade in the IoI_1711 autumn series is 1770–1779, which partly overlaps with the wettest at Carlisle (1766–1775), Oxford (1767–1776) and Spalding (1768–1777). The 1770s are also notably wet in the Pauling series. The driest autumn decade in the IoI_1711 series is 1745–1754, which matches closely with Kew (1745–1754), Spalding (1748–1757) and the early EWR series (1748–1757). In autumn, there is a strong negative correlation with L-SLP (r=-0.65), with high mean L-SLP in the years 1745–1754 coinciding with dry rainfall conditions across many stations, and low L-SLP values in the 1770s consistent with wet autumn conditions at the time. Furthermore, the driest autumn decade (1745–1754) coincides with anomalously high values for EU2 index at this time, while the wettest autumn decade (1770s) coincides with a markedly low EU2 index, adding confidence to this finding.