SEARCH

SEARCH BY CITATION

Keywords:

  • historical climatology;
  • weather diary;
  • volcanic forcing of climate;
  • Tambora;
  • Galunggung

Abstract

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Materials and methods
  5. 3. Results
  6. 4. Discussion
  7. 5. Conclusions
  8. Acknowledgements
  9. Appendix
  10. References

An early 19th century diary, belonging to a farmer from NW England, contains detailed daily weather entries over a 15-year period from 1815 to 1829. The diary period encompasses the Tambora volcanic eruption and pre-dates the establishment of routine meteorological observations. The diary entries were imported into a database and the descriptive entries categorised and transformed into ordinal data for statistical analysis, measuring frequencies of weather events or related phenomena. In ranked comparisons, the diary is strongly correlated with established temperature and precipitation datasets, providing confidence in the reliability of the diary as a data source. From the categorised diary data, significant indicators of poor climate are identified, including higher-than-mean occurrences of low pressure, summer cold and rainfall. The years 1817, 1816 and 1823, respectively have the highest number of indicators of poor weather with high or very high significance, totalling 21 over these 3 years, whereas only 11 are recorded in total for the remaining 10 years of data. Thus significant climate anomalies are found in the 2 years following the Tambora 1815 eruption and to a lesser extent 1 year following another major eruption, Galunggung in 1822. As well as colder and wetter summers, other indicators such as optical phenomena, which are not routinely included in climate records, are recorded. There is no evidence that the diarist was aware of the Tambora and Galunggung eruptions at the time of writing. Although the diary covers a relatively short time span and is largely qualitative in nature, it is from an early period, pre-dating the establishment of the UK Meteorological Office, and is unusually methodical. The diary contains a wealth of data which could be further exploited. The current study demonstrates the potential of database technology for categorising and quantifying descriptive data sources in historical climate studies. Copyright © 2009 Royal Meteorological Society


1. Introduction

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Materials and methods
  5. 3. Results
  6. 4. Discussion
  7. 5. Conclusions
  8. Acknowledgements
  9. Appendix
  10. References

The global effects on climate of the Tambora 1815 eruption, because of the injection of large volumes of aerosol into the atmosphere, have been widely discussed (e.g. Rampino and Self, 1982; Stothers, 1984; Harington, 1992; Oppenheimer, 2003). The stratospheric loading of sulphate aerosol from Tambora is suggested to be between 93 and 118 Tg, and the area of the Tambora ash cloud estimated to be three times greater than that created by the Pinatubo 1991 eruption (Self et al., 2004). One of the most significant impacts in the years immediately following the Tambora eruption is the cooling of Northern Hemisphere summer surface air temperatures, with 1816 becoming known as ‘The Year Without a Summer’ (Harington, 1992). The averaged Northern Hemisphere summer temperature anomaly has been estimated at around − 0.5 °C (Briffa et al., 1998). Temperatures in Central England and locally in Lancashire (Manley, 1946, 1974) exhibit a notable cooling trend, with July 1816 the coldest July on record to date. Although less documented, the eruption of Galunggung 1822 has been implicated as demonstrating a cooling summer temperature signal (Fischer et al., 2007), and summer months during the year 1823 again rank as some of the coldest on record in England (Manley 1946, 1974). An investigation of the 200-year record of gale frequency in Edinburgh showed that there were clear peaks in storminess following three major volcanic eruptions, including Tambora 1815 (Dawson et al., 1997).

The eruptions of Tambora and Galunggung are the major events occurring in the period of a farmer's diary that is the focus of this study (1815–1829), and the volcanoes are located on a similar latitude in Indonesia, southeast Asia (Table I). The diary used in this study was written by John Andrew, who was born in 1763 and lived and worked as a farmer at Lumn. At GB grid reference [SJ 973 929] (lat 53.4329, lon − 2.0421), Lumn encompasses a small group of farm buildings, situated on the western edge of the Peak District between Manchester and Glossop, on land approximately 150–200 m above sea level (Ordnance Survey, 2002). Middleton (1900) described Andrew as a ‘remarkable character, a philosopher, poet and faithful chronicler of local events’. There is evidence of Andrew's interest in various aspects of earth sciences and mathematics, for example, his attendance at a geological lecture in 1828 and, amongst the Andrew papers, a handwritten book with meticulously drawn geometric diagrams and accompanying notes.

Table I. Major (VEI ≥5) volcanic eruptions in the period 1815–1829, along with Krakatau 1883 for comparison (NGDC, 2007)
VolcanoDateLatitudeLongitudeElevation (m)Volcanic indices
     DVIVEI
  • Volcanic indices are dust veil index (DVI) (Lamb, 1970) and volcanic explosivity index (VEI) (Newhall and Self, 1982).

  • a

    A DVI of 300 was originally calculated for Galunggung because of an ambiguous temperature signal, however, this was adjusted by Lamb to 500 using information from historical reports.

TamboraApril 1815− 8.25118.00285030007
GalunggungOctober 1822− 7.25108.052168500a5
KrakatauAugust 1883− 6.10105.4281310006

The diary entries are split over several volumes, or ‘Weather Books’ (Figure 1), but for the purposes of this study, the set will be referred to as ‘the diary’. It comprises 4652 largely contiguous daily entries over the period April 1815–June 1829 and a typical entry includes four descriptive weather observations, two wind direction readings and two readings labelled ‘G’, which are taken to relate to barometric pressure, but are not in any of the usual units for pressure. The diary also includes comments on optical phenomena and crop conditions which may provide useful climate indicators.

thumbnail image

Figure 1. The exterior (left) and sample page extract (right) from the John Andrew diary. The diary extract shows daily weather records from Friday 9 April 1824 to Tuesday 13 April 1824. Each weather record takes the same form (Table II). Interpolated between the weather records are more general diary entries, often relating to farming matters, and some newsworthy items, apparently added some time later. In this extract, for example, the death of Lord Byron is recorded

Download figure to PowerPoint

The weather records in the diary have been transcribed into a spreadsheet (Lada, 2005). The current study developed the electronic dataset into a database, allowing us to study the weather recorded by looking at the frequency of a defined set of keywords. This reduces the chances of erroneous interpretation, and eliminates the need for a comparison of the diary entries with modern-day numerical values (Manley, 1962). Details of the interpretation and manipulation of diary data are given below.

On an initial examination of the diary, it was noticeable that particularly severe weather events were being recorded by the author in the summer months of 1816, leading to the hypothesis that the entries described the climate influence of the Tambora eruption of 1815. Historical diaries have been used to describe the effects of large eruptions such as Tambora 1815, however, many of these resources are eyewitness accounts of the immediate effects of the eruptions, as described in Stothers (1984), Harington (1992) and Oppenheimer (2003). Less use has been made of historical diaries to assess the climatic effects in the years following large volcanic eruptions; furthermore, there is little demonstration in previous studies of the use of database technology to systematically interrogate and quantify descriptive data, which will be undertaken in this study.

The Andrew diary is an important primary source preceding any official meteorological service in Britain, and rare in the level of detail recorded, including not only numerous daily weather events, but also other phenomena and observations relevant to the investigation of volcanic forcing. This study sets out to extract key data from the diary in a form that can be used primarily to identify anomalous weather patterns.

This work will inevitably focus on NW England, where the diary was written, but this does have the advantage of allowing comparison with the Lancashire temperature series (Manley, 1946) and the Central England temperature (CET) series (Manley, 1953, 1974; Parker et al., 1992), the latter of which is one of the most important historical climate series. In addition to temperature, this study will discuss precipitation and pressure pattern changes.

We discuss below correlations between the diary data and the published temperature and precipitation datasets. Bearing in mind the age and mainly descriptive nature of the diary, its accuracy will be assessed and conclusions will be drawn on the usefulness of this type of record in climate studies.

2. Materials and methods

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Materials and methods
  5. 3. Results
  6. 4. Discussion
  7. 5. Conclusions
  8. Acknowledgements
  9. Appendix
  10. References

The spreadsheet of diary weather entries from a previous study (Lada, 2005) was imported into a new custom-designed Lotus Notes (IBM Corporation) database. Crop and other data, relevant to this study but not available from the spreadsheet, were added to the database directly from the diary. A random sample of electronic entries were checked against the original diary for inputting errors, and obvious errors in numerical values were also identified and corrected via specific views created for this purpose.

The database comprises a total of 4652 daily weather entries, each with 11 data fields (Table II). The fields are self-explanatory, except for ‘G’, a numerical entry ranging from ‘0’ to ‘22’. After some investigation, ‘G’ is taken to be a pressure reading evidenced by comments made in the diary, correspondence between low ‘G’ values and stormy periods and also between the high ‘G’ values and settled periods (high pressure). However, the numerical value does not conform to any of the usual pressure units, and the actual instrument used has not been identified to date.

Table II. Typical daily entry from the diary
Field nameData
Year1817
Month7
Day15
G morning3
Wind dir morningE
Weather morningDark-close-rainy
Weather forenoonGreat wind-rain—dark
Weather afternoonClose driving rain—FLOOD
Weather eveningFair
Wind dir eveningNE
G evening8

A hierarchy of categories and sub-categories was devised within the database (Table III) to facilitate frequency analysis of the data. Ordinal data is collated by examining the frequency of a weather event occurring over a given time period, for example, number of ‘rain days’ in a season. No attempt has been made to derive truly quantitative values such as temperature in °C/ °F or precipitation in mm, to avoid the potential large errors of attempting to convert the diary data to instrumental values. Safeguards were built into the system to ensure that multiple similar events in a single day did not lead to over-counting and that negatives (e.g. ‘no frost’) were not counted as positives.

Table III. Categories assigned to diary keywords and phrases
Main categorySub-categoryKeywords includedKeywords excluded
 12  
  • a

    Hail has been classified under ‘cold’ but could be extreme rain if part of a (thunder)storm, in which case it will fall under the ‘storm’ category as this encompasses ‘hailstorm’.

  • b

    Did not include ‘wet’ on its own because of ambiguity (could be for example damp fog or heavy dew); whereas ‘v.wet’ is assumed to relate to heavy or persistent rain.

  • c

    The low- and high-pressure thresholds for ‘G’ (4 and 15, respectively) were chosen as they encompass approximately the bottom and top 10% of data, and a sufficient quantity for statistical analysis.

  • d

    Red or rud has been interpreted as descriptions of skies, where higher incidences of red skies have been linked to volcanic aerosols in the atmosphere. The Shorter Oxford English Dictionary gives ‘rud: red, redness, ruddiness. Old English’.

  • e

    ‘Moorgrime’ is a north of England term that may relate to black sooty accumulations found in the wool of sheep grazing on upland pastures (Brimblecombe, 1992), although the association with soot was disputed by elderly residents of the area when we discussed the term in a local press article (Westmorland Gazette, December 14, 2007, and subsequent readers' letters).

All coldExtreme coldSnow‘Snow’‘Snow gone’ ‘snow waste’
  Hail/Sleeta‘Hail’ ‘sleet’ 
  Bitter‘Bitter’ ‘severe cold’ ‘severely cold’ ‘dreadful cold’ ‘intense cold’ ‘excessive cold’ ‘cold as christmas’ ‘cold as winter’ 
  Frost/ice‘Frost’ ‘ice’‘No frost’ ‘frost gone’
 Cold ‘Cold’ 
All rainExtreme rainStorm/TS‘Storm’ ‘thunder’ ‘lightning’‘Snowstorm’ ‘snow storm’
  Flood‘Flood’ 
  Heavy rain‘Heavy rain’ ‘excessive rain’: ‘excessive heavy rain’ ‘excessive day of rain’ ‘extreme rain’ ‘severe rain’ ‘severe day of rain’ ‘v.rainy’ ‘v.wet’ ‘excessive wet’ ‘dreadful rain’ ‘continual rain’ ‘continual day of rain’ ‘prodigious rain’ ‘rains prodigiously’ ‘driving rain’ ‘tremendous rain’ 
 Rainb ‘Rain’ 
 Showers ‘Shower’ 
PressurecLow If G (morning or evening) ≤4 
 High If G (morning or evening) ≥15 
OtherGloomy ‘Dark’ ‘black’ ‘gloomy’ 
 Red/Rudd ‘Blood’ ‘red rud’ ‘red-’ ‘rud-’ ‘red’ ‘rud’ ‘-red’ ‘-rud’ 
 Moorgrimee ‘Moorgrime’ ‘moor grime’ 

The analysis focussed on ten categories of weather and related phenomena: all cold; extreme cold; all rain; extreme rain; storm and thunderstorm; gloomy; red/rud; low pressure and high pressure (Table III); and also on crop conditions. Where relevant entries were available, a crop rating was given on a scale of − 2 to + 2, with − 2 relating to very poor crop conditions, e.g. ‘corn yet green—beaten down—none turned—dreadful prospects’ (9 September 1816); 0 relating to indifferent conditions; and + 2 relating to excellent crop conditions, e.g. ‘Hay harvest nearly finished, immense crop this year and well got (13 July 1827)’. For all categories except low pressure, high pressure and crop rating, the analysis included only the summer months of June, July and August (JJA).

Two types of statistical tests were used for this study, the Student's t-test and Spearman's rank correlation. To test the significance of weather anomalies after the eruptions of Tambora in April 1815 and Galunggung in October 1822, the ‘post-eruption’ years were defined as the 2 years following each eruption (1816, 1817, 1823 and 1824). Two-tailed t-tests were used with a null hypothesis that for each post-eruption year, the number of days of a particular event does not significantly differ from the mean number of days of the remaining (control) years. For example, the number of ‘cold days’ in summer months (JJA) of 1816 does not significantly differ from the mean number of ‘cold days’ for the summer months of the remaining diary years. T-test significance levels were set as V. High (0.1% chance that the null hypothesis has been discarded erroneously), High (1%) and Significant (5%). The Spearman's rank correlation non-parametric test was used to compare the diary temperature and precipitation data with established datasets (Table IV). These are CET (Manley, 1953, 1974; Parker et al., 1992); Lancashire temperature series (Manley, 1946); England and Wales precipitation series (EWP) (Alexander and Jones, 2001) and Edinburgh temperature and precipitation series (Mossman, 1896).

Table IV. Extracts of datasets ranked for summer season (JJA) years 1815–1829, with temperature ranked coldest to warmest and precipitation ranked wettest to driest
RankCentral England temperatureLancashire seriesEngland and Wales precipitationEdinburgh temperatureEdinburgh precipitation
 YearTemperature ( °C)YearTemperature ( °F)YearPrecipitation (mm)YearTemperature ( °F)YearPrecipitation (ins)
  1. Seasonal means were calculated where these were not available directly from the dataset. To enable comparisons, the summers 1822 and 1829 which were missing from the diary were also removed from the established datasets and the remaining summers re-ranked.

1181613.4182355.21829396.3181654.818174.63
2182313.6181655.41828355.9182355.118294.44
3181714.3181756.51817345.7181755.318163.13
4182114.5182057.41816311.6182955.618233.04
5182014.7182157.41823261.3182156.418282.94
6181514.8181557.61822236.2182756.618272.93
7182414.8182958.11824217.8182056.818222.75
8182914.8182458.11820203.7182857.218202.47
9182715.2182758.61815203.5181557.318182.03
10182815.6182259.31821199.8182457.918151.95
11181915.7182859.31819190.0182258.118241.70
12182515.9181959.51827172.5181858.418191.68
13182216.0182559.61825160.3181959.018261.48
14181816.6181860.31826121.8182559.418251.36
15182617.6182662.71818102.6182661.618211.20

For comparisons with the ranked established datasets, the diary summer seasons were ranked in descending order of cold-day count; and in descending order of rain-day count. A two-sided Spearman's Rank Correlation was used with a null hypothesis that there was no association between the diary data and the established datasets. The correlation significance levels were set as high (1%), with a correlation coefficient of >0.745, and correlated (5%), with a correlation coefficient of >0.566.

There were time periods within the diary where no data was available because of the author's illness or holidays. Rules were devised for handling the missing data in the statistical analysis (Table V). All the categories analysed were subject to these rules, except the crop rating where there were so few entries that it was deemed not practicable to interpolate the data.

Table V. Rules for handling of missing data for statistical analysis
Time period missingApplies toAction taken
Greater than 1 month1822 JJARemoved from statistics
in 3 month period1829 JJA 
 1815 DJF 
One month or less in1825 JJAInterpolated using mean of other months: Example: In 1827 there are three June
3 month period1826 JJAdays of extreme rain, but 20 days missing. Therefore total days extreme rain in
 1827 JJAJune 1827 will be: original 3 days + [mean of extreme rain in June of other years
 1823 DJF(7) × missing proportion (20/30 = 0.666)]
 1827 DJF3 + (7 × 0.666) = 8 days
 1829 DJF 

The methods were devised to maximise the potential of the data, but it is important to consider potential limitations which may affect the results. These include inconsistencies in terminology used by original diary author, typing errors in transfer from paper diary to electronic data, subjective decisions on categorisation, interpolation of missing data which may not represent true results and the limited number of years available for the study. These issues were mitigated as far as possible, for example, by using validation reports to check for typing errors and ensure the majority of keywords were correctly assigned, also using statistical tests designed for small samples.

3. Results

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Materials and methods
  5. 3. Results
  6. 4. Discussion
  7. 5. Conclusions
  8. Acknowledgements
  9. Appendix
  10. References

3.1. Rank correlations

Correlations between the established datasets and the diary were measured for temperature and precipitation (Table VI).

Table VI. Summary of Spearman's rank correlation coefficients measured for temperature and precipitation
 CETDiary
Temperature
CET 0.868a
Lancashire0.992a0.875a
Edinburgh0.922a0.805a
 EWPDiary (all rain)Diary (extreme rain)
  • Coefficients in italics are classed as correlated.

  • a

    Highly correlated.

Precipitation
EWP 0.805a0.710
Edinburgh0.6760.775a0.644

For temperature, the Lancashire and Edinburgh series were found to be highly correlated with the CET, with coefficients greater than 0.9. The comparison of the diary against each series revealed that although the correlation coefficients were slightly lower than between established series, correlation was still very good. The diary correlated most closely with the Lancashire series (Figure 2), then the CET, and finally the Edinburgh series.

thumbnail image

Figure 2. Spearman's Rank Correlation: Lancashire temperature series (Manley, 1946) versus number of ‘all cold’ days from diary; summer months only. Correlation coefficient is 0.875 (high correlation)

Download figure to PowerPoint

For precipitation, the correlation coefficients were generally lower than those measured for temperature, but correlation was still significantly high particularly for the diary ‘all rain’ days against the established datasets. One significant outlier with a large difference in ranking was 1824 (Figure 3), where the rainfall ranking was much higher (i.e. more rainfall) for the EWP than for the diary.

thumbnail image

Figure 3. Spearman's Rank Correlation: England and Wales precipitation series (Alexander and Jones, 2001) versus number of ‘all rain’ days from diary; summer months only. Correlation coefficient is 0.805 (high correlation). Outlier enclosed in dashed circle is the year 1824

Download figure to PowerPoint

3.2. Category analysis

For each year, each of the ten categories analysed was given a significance rating of very high, high, significant or none according to the t-test results. Histograms were compiled (Figures 4 and 5) based on the summary of significances for the ten categories (Appendix, Table AI), separated into indicators of poor or fair climate depending on their relation to the mean. The years 1816, 1817 and 1823 have the highest number of significant indicators of poor climate, whereas 1818 stands out as a year with a much fairer climate.

thumbnail image

Figure 4. Significant indicators of poor climate (higher than mean cold, rain, storms, gloomy, red/rud, low pressure; lower than mean high pressure and crop rating). Significance levels are: very high (0.1%) and high (1%). n/a = insufficient data for year

Download figure to PowerPoint

thumbnail image

Figure 5. Significant indicators of fair climate (lower than mean cold, rain, storms, gloomy, red/rud, low pressure; higher than mean high pressure and crop rating). Significance levels are: very high (0.1%) and high (1%). n/a = insufficient data for year

Download figure to PowerPoint

Breakdowns of the ‘all cold’, ‘all rain’ and ‘low pressure’ categories are presented (Figures 6–8). The ‘missing’ years are those where the quantity of data is insufficient to provide an analysis. In other categories, red/rud days with very high significance are 1817 and 1818, and gloomy years with very high significance were 1816, 1817 and 1820. Poorer years for crops and harvests were found to be 1816, 1817 and 1821. A full summary of significances for each of the ten categories for each year is available in Table AI in the Appendix.

thumbnail image

Figure 6. Number of ‘all cold’ days in the summer months (JJA) for each diary year. Cold years with very high significance are 1816, 1817 and 1823 with a p-value of 0.0000, and 1821 with a p-value of 0.0002. In related results, extreme cold years with very high significance are 1816, 1817 and 1821

Download figure to PowerPoint

thumbnail image

Figure 7. Number of ‘all rain’ days in the summer months (JJA) for each diary year. Rainy years with very high significance are 1816, 1817 and 1823 with a p-value of 0.0000. In related results, extreme rain years with very high significance were 1817, 1823 and 1828. A year with very high significance of storms/thunderstorms was 1817, but only one out of the 15 days recorded was a ‘low pressure’ day. The proportion of showers to other types of rain was fairly constant over the 15-year period

Download figure to PowerPoint

thumbnail image

Figure 8. Number of low pressure days (G ≤4) winter and summer seasons (DJF and JJA) for each diary year. Low pressure years with very high significance are 1816, 1817 and 1823 with a p-value of 0.0000. Years where high pressure (G ≥15) was lacking with high significance were 1816, 1823 and 1828

Download figure to PowerPoint

4. Discussion

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Materials and methods
  5. 3. Results
  6. 4. Discussion
  7. 5. Conclusions
  8. Acknowledgements
  9. Appendix
  10. References

The high significance of rank correlation of temperature for all the datasets compared (Table VI) suggests that temperature does not differ considerably within the spatial scale studied (the British Isles). The order of highest significance of diary correlation with established datasets is as expected: the Lancashire series (Figure 2); the CET, which uses records from Lancashire to London; then Edinburgh. These results provide independent support for Manley's 1946 Lancashire series work, and also suggest that event frequencies can provide a good approximation of temperature trends where actual instrumental temperature readings are not available.

The precipitation statistics are slightly less well correlated for all datasets. The diary ‘all rain’ days category is more highly correlated with the EWP dataset than it is with the Edinburgh series. This suggests that Lancashire represents the England and Wales mean fairly closely, but does not represent East coast rain, such as that recorded at Edinburgh so well. Variations in precipitation over small spatial scales are larger than variations in temperature; this conclusion was also reached by another study into volcanic signals and is expected from our understanding of these processes in the climate system (Robock and Liu, 1994). A significant outlier in the precipitation record was noted (Figure 3); there is much higher rainfall in the EWP than in the diary for 1824. To try to understand why this type of anomaly might occur, the descriptive agricultural records were consulted (Stratton, 1969) and it seems that thunderstorms were frequent over London and the southern counties in the summer of 1824, which would raise the EWP totals but not affect Lancashire. This type of local event is likely to account for ‘extreme rain’ recorded in the diary not correlating as well with the EWP as ‘all rain’. A major limiting factor of comparing ‘rain days’ with actual rainfall totals over a period of time is the inconsistency of measurement, for example, 1 day of continuous heavy rain may total more than 3 days of intermittent rain but would still only be classed as 1 ‘rain day’.

The overall high or very high correlation statistics of the diary with established datasets serves to add weight to the validity of the results from the ten categories analysed for significance within the diary. The significant indicators of poor climate are very prominent in the years 1816 and 1817 and also prominent in 1823 (Figure 4), and a lack of fair climate indicators in these years (Figure 5) is also apparent. One might question use of the terms ‘poor’ and ‘fair’ climate indicators, and in reality the assignment is more complicated; for example, lower rainfall than the mean has been classed as ‘fair climate’, but if rainfall was too low this could lead to a drought and so strictly should be ‘poor climate’. Despite this, the assignment of the climate indicators appears to be adequate for the purposes of this exercise. There is strong evidence to suggest that significantly poor summers occurred in NW England following the volcanic eruptions of Tambora and Galunggung.

Analysing some of the individual categories, it was possible to gain an insight into the detail of weather patterns over the 15-year diary period. For example, there was a very high significance in the number of cold (Figure 6) and extreme cold days in the summers of 1816 and 1817; whereas for 1823 the very high significance of cold days did not extend to extreme cold. It was noted that patterns of low pressure and rainfall did not always correspond as well as would be expected. This could be because of a number of factors including the subjective low pressure threshold of four, or the differences between frontal rain associated with low pressure and summer storms often associated with high pressure systems. An example of this is 1817, where 15 days of storms/thunderstorms were recorded in the summer, but only one of these was associated with low pressure.

The very high significance of low pressure in the diary years 1816, 1817 and 1823 supports Lamb's conjecture (1992) that the coldness of the 1816 summer in Britain was attributable to an anomalous equatorward extension of the Icelandic Low pressure system. This climate fluctuation is now commonly known as the North Atlantic Oscillation (NAO), where the weather in Northern Europe is heavily influenced by the positioning and resulting pressure difference between the Icelandic Low and Azores High. The shift in pressure pattern may be related to the positive-phase NAO observed after large tropical eruptions (Robock, 2000). However, this phase is generally considered to have most influence in the winter months with warmer and wetter conditions. Amplification of the NAO circulation pattern is explained by volcanic aerosols heating the tropical lower stratosphere, which increases the Equator-Pole temperature gradient and in turn causes a stronger polar vortex. Shindell et al. (2004) corroborates a winter warming response, and further states that cooling during the summer occurs because there are weaker planetary waves at this time of year, allowing the influence of the volcanic aerosols to become more dominant.

Wilson (1999) compared the timing of El Niño events and volcanic eruptions with Manley's Central England Temperature records. The onset of the eruptions of Tambora and Galunggung occurred outside El Niño periods, and a cooling of residual temperatures began almost immediately post-eruption, with the stronger eruption of Tambora having a more lasting influence. The occurrence of El Niño when temperatures were recovering appears to have caused a new temperature dip, a similar signal to the volcanic cooling. The evidence presented by Wilson (1999) suggests that there is no connection between the Tambora and Galunggung eruptions and El Niño events, but that each are separately capable of producing decreases in Central England temperatures. This concurs with the conclusion from a study of the 16 strongest El Niño events of the last 150 years (Self et al., 1997), that there is no general correlation between El Niño and volcanic aerosol perturbations. The debate continues, as subsequent studies have suggested that while El Niño events cannot be said to be triggered directly by large volcanic eruptions, evidence is presented for an increase in their likelihood in response to explosive volcanic forcing (Adams et al., 2003; Emile-Geay et al., 2008).

In a model simulation of the climate impact of the Pinatubo 1991 eruption, Thomas et al. (2009) showed a different dynamical response of the extratropical atmosphere during easterly and westerly quasi-biennial oscillation (QBO) phases, and also suggested that a disturbance of the polar vortex in the first winter after the eruption may have been influenced by El Niño; although these responses are still somewhat poorly understood. The interaction between volcanic forcing, atmospheric circulation patterns and precipitation is clearly complicated and likely to vary considerably on regional scales; consequently it is an area that warrants further research.

On assessing the red/rud category in the diary, assumed to relate to optical phenomena such as red skies, the number of days in 1817 and 1818 displayed very high significance, but the red/rud category of weather report was completely absent in 1816. If the red skies were as a result of Tambora, it would be expected that these would certainly have been visible by the summer of 1816. One possible explanation is that Andrew did not record red/rud as a weather event in his diary until these events became so common that he decided to do so. An indicator that red skies were observed in 1816 but not recorded as weather events is a general comment in the diary on 26 November 1816 which reads ‘sky as red as blood this E’. The optical effects caused by Tambora for several years following 1815 have been widely reported, including prolonged and brilliantly coloured sunsets and twilights observed near London in late 1815 (Stothers, 1984). It is apparent that the aerosol veil produced by Tambora reached England in about 3 months, and that some haze still remained after two and a half years. Bright sunset skies and red clouds were also observed in Britain and London after the Galunggung eruption, during 1822 and 1823 (Lamb, 1970).

There is now a good understanding of how major volcanic eruptions can affect the climate by disturbing the atmospheric radiative equilibrium (Figure 9). The life cycle of stratospheric aerosol is governed by the Brewer–Dobson circulation, where volcanic aerosol injected into rising stratospheric air close to the equator will spread meridionally, before removal processes take effect at high latitudes (Thomason and Peter, 2006). This means that as a general rule, tropical eruptions will result in poleward transport of aerosol in both hemispheres, whereas aerosol particles from eruptions in higher latitudes will be largely confined to their own hemisphere, being close to the exit of the Brewer–Dobson circulation. Ice core studies have revealed that global transport of aerosols is, in reality, complicated, with differences in hemispheric distribution influenced by the time of the year, location of the inter-tropical convergence zone and also the phase of the QBO (Thamban et al., 2006).

thumbnail image

Figure 9. Schematic diagram of the radiative effects of volcanic inputs to the atmosphere. Many gases are carried into the stratosphere by a large eruption and the sulphur gases, H2S and SO2, oxidise over weeks into H2SO4 aerosol. These small particles scatter short-wave solar radiation effectively in all directions, and this leads to decreased direct solar flux reaching the troposphere and earth's surface, causing net cooling. Heating of the stratosphere is because of volcanic aerosol particles absorbing some solar radiation. The minor greenhouse effect occurs where larger coagulated particles absorb outgoing radiation emitted from the earth's surface, however, this is a short-term effect as these heavier particles quickly settle out. (Amended from Figure 16.2 of Francis and Oppenheimer (2004) and Plate 1 of Robock (2000))

Download figure to PowerPoint

Satellite measurements of stratospheric aerosol before, during, and after the June 1991 Pinatubo eruption, were taken by the SAGE II satellite instrument (NASA, 2007). The volcanic aerosol was initially concentrated around the tropical stratosphere, then spread to the middle and high latitudes, and was still slowly dispersing two and a half years after the eruption. Typical responses of the disturbance of the atmosphere because of large volcanic eruptions include Northern Hemisphere summer cooling, winter warming, reduction of the diurnal cycle and optical effects (Robock, 2000). However, it is important to recognise, when studying any period of climate fluctuation related to volcanism, that other climate-forcing influences may also be acting, either additively or non-linearly.

Solar forcing is one such factor known to have a significant influence on global temperature, for example, the long-term Milankovich cycles. There are also shorter term decadal episodes which occur as a result of changes in solar activity and the ‘Little Ice Age’ encompassed two periods where there were fewer sunspots than normal, meaning the sun's magnetic activity was very low (Soon and Yaskell, 2003). This weak solar activity occurred during the ‘Maunder Minimum’ circa 1650–1715, and the ‘Dalton Minimum’, circa 1795–1823, and led to abnormally cold weather in the Northern Hemisphere (Soon and Yaskell, 2003). The Dalton Minimum is of particular relevance as the diary that is the subject of this work incorporates the latter part of this period. The massive eruption of Tambora in 1815 which occurred during this already cool episode is likely to have significantly amplified the effect on climate (Reid, 1997; Soon and Yaskell, 2003).

In ice core studies, a clear acidity peak (spike) has been attributed to the Tambora eruption; identified as a large spike which follows another spike attributed to an unknown eruption occurring around 1808 (Legrand and Delmas, 1987; Moore, 1991). Greenland profiles do not exhibit the Galunggung signal, and one debatable explanation is an unbalanced distribution of aerosol from this eruption between the hemispheres (Legrand and Delmas, 1987). The noticeable Galunggung signal in the dairy does not support such an explanation. In the East Antarctic record, there are several peaks that are difficult to attribute to specific eruptions around the time of Galunggung, as there are several possible volcanic sources in the 1820s and 1830s (Moore, 1991). Consequently there is more ambiguity surrounding the Galunggung ice core record.

Briffa et al. (1998) used a network of tree-ring-density chronologies to investigate the influence of volcanic eruptions on Northern Hemisphere summer temperatures. The past 600 years were ranked in order of tree-ring density lowest to highest, and very low summer temperatures were identified in 1816, 1817 and 1818 (ranked 2nd, 5th and 22nd, respectively) (Briffa et al., 1998). This may be indicative of a lengthy cooling influence from Tambora, with temperatures slowly recovering over a period of 3 years from 1816. Only the top 30 ranking coldest years were listed in this tree-ring study, and none were years immediately succeeding the Galunggung 1822 eruption.

Agricultural records provide an insight into the poor climatic conditions in the years following eruptions. In Britain, the cold and wet summer of 1816 is described, leading to a disastrous harvest and food riots; another late spring occurs in 1817, and 1823 suffers a poor harvest, with a cool summer merging into an autumn with strong winds and gales (Stratton, 1969).

The large peaks in poor climate indicators in the diary for the years 1816, 1817 and 1823 (Figure 4) are unlikely to have occurred by chance. It could be argued that the entire data series should be used to generate the mean to test against in statistical tests, rather than the mean of the ‘non-volcanic’ years. Using the overall mean produces little change in the findings because, although the new t-test values generally equated to lower significance for poor climate indicators in the post-eruption years, the significance for a fairer climate in the remaining diary years rises to balance the overall result.

Previous studies by Harington (1992) and others have put forward strong evidence of aerosol perturbation of the atmosphere in the years following the Tambora 1815 eruption. This study supports the evidence and suggests that the climate of the British Isles was affected significantly, at least during the summers of 1816 and 1817 with cold temperatures, high rainfall and low pressure. By 1818, the weather had much improved with no ‘cold days’ recorded in the summer. This differs from the Northern Hemisphere tree-ring data of Briffa et al. (1998), which records 1818 as a cold year. The indication is that there are regionally spatial variations in the hemispheric temperature trends. The residual aerosols in 1818 may still have caused optical effects with high incidences of red/rud in the diary; and this seems feasible with measurements of visual extinction at northern latitudes following Tambora (Stothers, 1984), and the known persistence of aerosols after Pinatubo. The implication is that classing just 2 years after a major event as ‘post-eruption’ is not sufficient for detailed studies, and if a larger time span were available then perhaps this period should be extended to three or more years.

As discussed previously, the smaller Galunggung 1822 eruption is less well-known and less studied than Tambora. Ice core records and tree-ring evidence for climatic effects following Galunggung are either non-existent or ambiguous, but the poor climate signals in the established datasets and the diary are quite strong. One study even acknowledged the cold summer of 1823 but stated that no volcanic forcing mechanism was involved (Sadler and Grattan, 1999), so Galunggung was evidently not considered. The climate signal is not as robust or long-lasting as the Tambora signal but this is not surprising because of the different scale of eruption. Weighing up all the factors including the poor climate in both the established UK records and diary for 1823, and the bright sunset skies and red clouds observed in Britain and London during 1822 and 1823 (Lamb, 1970), the case that the Galunggung 1822 eruption was responsible for Northern Hemisphere climate anomalies merits further investigation.

5. Conclusions

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Materials and methods
  5. 3. Results
  6. 4. Discussion
  7. 5. Conclusions
  8. Acknowledgements
  9. Appendix
  10. References

The early 19th century diary of John Andrew contains detailed daily weather entries over a 15-year period from 1815 to 1829. The diary period encompasses the Tambora volcanic eruption and pre-dates the establishment of routine meteorological observations. By importing the diary entries into a database, and categorising the descriptive entries, ordinal data for statistical analysis were produced. This ordinal data measured the frequencies of weather events or related phenomena, permitting time series analysis of the diary record. In addition to internal analysis of the diary data, they were compared against established temperature and precipitation datasets. The ranking of the diary summer seasons against established temperature and precipitation datasets resulted in surprisingly strong correlations. In particular this served to support Manley's Lancashire temperature series (Manley, 1946). It also instilled confidence that analysis of other types of data in the diary was likely to be reliable and thus an important primary source from a period which precedes the establishment of the Met Office in Britain. There was substantial evidence in the diary for poor summer weather in the British Isles having been strongly influenced by two major tropical eruptions. This signal was particularly robust in the 2 years following the Tambora 1815 eruption, but also significant in the year following the Galunggung 1822 eruption.

Acknowledgements

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Materials and methods
  5. 3. Results
  6. 4. Discussion
  7. 5. Conclusions
  8. Acknowledgements
  9. Appendix
  10. References

We would like to thank Christine Valentine who kindly loaned the diary and related materials of her ancestor, John Andrew, for this study and for being helpful and encouraging throughout. We gratefully acknowledge Konstantina Lada for her initial transcription of the diary records.

Appendix

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Materials and methods
  5. 3. Results
  6. 4. Discussion
  7. 5. Conclusions
  8. Acknowledgements
  9. Appendix
  10. References
Table AI. Summary of significances for each of the ten data categories analysed for each year
Post-eruption years1816181718231824
All cold JJAV. HighV. HighV. HighNone
Extreme cold JJAV. HighV. HighSigNone
All rain JJAV. HighV. HighV. HighNone
Extreme rain JJANoneV. HighV. HighHigha
Storm and Thunderstorm JJANoneV. HighNoneNone
Gloomy JJAV. HighV. HighSigNone
Red/Rud JJAHighaV. HighSigNone
Low pressure DJF and JJAV. HighV. HighV. HighHigh
High pressure DJF and JJAHighNoneHighSiga
Crop ratingV. HighV. HighNoneSig
Control years18151818181918201821182218251826182718281829
  • For example, in JJA 1816, the number of ‘All cold’ days (43) exceeded the mean of the control years ‘All cold’ days (11) with a very high significance.

  • a

    The result was lower than the mean, or in the case of high pressure and crop rating, higher than the mean and is therefore an indication of a fairer climate.

  • n/a denotes insufficient data available for the analysis.

  • V. High = significant at 0.1% level.

  • High = significant at 1% level.

  • Sig = significant at 5% level.

All cold JJANoneHighaSigaNoneV. Highn/aNoneNoneNoneNonen/a
Extreme cold JJANoneNoneNoneNoneV. Highn/aNoneNoneNoneNonen/a
All rain JJASigNoneNoneSigNonen/aNoneV. HighaNoneHighn/a
Extreme rain JJANoneHighaNoneHighNonen/aNoneSigaNoneV. Highn/a
Storm and Thunderstorm JJASigaHighaSigNoneNonen/aNoneNoneHighaHighn/a
Gloomy JJANoneNoneNoneV. HighSign/aNoneHighaHighaNonen/a
Red/Rud JJAHighaV. HighSigNoneNonen/aSigaNoneHighaNonen/a
Low pressure DJF and JJAn/aNoneNoneV. HighaNonen/aNoneSigNoneSign/a
High pressure DJF and JJAn/aNoneNoneNoneHighan/aSigaNoneNoneHighn/a
Crop ratingNoneHighaNoneSigaV. HighHighNoneNoneNoneNoneNone

References

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Materials and methods
  5. 3. Results
  6. 4. Discussion
  7. 5. Conclusions
  8. Acknowledgements
  9. Appendix
  10. References