Questions about air quality in Britain in late 18th and early 19th centuries

[Roger J Smith]

While working on my project to discover natural variability cycles in daily CET data, I decided to have a look at temperatures by day of the week. The only plausible cause of any variations on that exact time scale (7.000' days) would be human influences.

 

As you may know, this daily data extends from 1772 to the present, and as that came to 240 years at the end of 2011, I have divided my study into eight periods of 30 years.

 

Somewhat to my surprise, there is a clear trend from 1772 to about the mid-19th century for temperatures to be colder mid-week (Wednesday and Thursday in particular) than over weekends. This trend is basically established in the winter half-year and degrades to almost random in the summer half year. So looking at only the winter half year, the amplitude of this weekly cycle is on the order of 0.3 C degrees.

 

Towards the Victorian era and what I would understand to be the rise to industrial production peaks, this cold signal shifts slowly towards the earlier part of the week resting on Tuesday from about the late 19th to the mid 20th century. The trend seems to flatten out about the time that coal transitioned to cleaner heating fuel sources and in the past thirty years the main signal has been a warm peak late in the week rather than a conspicuous cold signal earlier.

 

I would like to display and discuss this data for general interest, but before getting to that stage, I am wondering if any NW readers can offer any insight into atmospheric quality in the first 60-90 years of the data. My understanding is that the Industrial Revolution was gradually adding soot to the air during these years but I'm wondering about the social issues about working days and non-working days, was there any slowdown in production over weekends in those years, or were the factories producing on a similar pace? If so, what else might account for any human modulation of a weekly cycle at that time? I can see where a weekly cycle becomes more and more plausible as legislation improved working conditions to give working people days off that in most cases would be Saturday and Sunday. Also Monday data would include more and more holidays as time goes on.

 

I have also looked at the various years in the early half of the data by dates of Christmas and New Year (which would be the same day of the week) to find out if the weekly cycle changes its appearance depending on what day the holidays occurred.

 

Anyway, the main question I am trying to solve here before trying to present the data in a coherent way would be this: how much soot and coal smoke would be in the air around 1790, 1820, 1850 compared to the heavier loading that could be assumed from 1870 to 1960. And also, my memory from early visits "home" was that coal smoke was still quite prevalent in the early 1970s but not in the late 1970s and 1980s. Is there any data on this transitional period?

 

If anyone has any other ideas about what could cause a cold signal to appear in CET data mid-week let me know. I have had a preliminary look at weather types and signals, and figure that the years most affected were "moderately cold" years. Very cold and also milder than average years show smaller variations. So this seems to be inversion-related, a cold signal that peaks in conditions that are below average but not record cold (those winters generally had flatter profiles).

[Gray-Wolf]

Kinda on topic but I do remember a study showing that 'modern day' polution builds up through the week leaving weekends weather blighted....peeing off workers who work through a lovely week only to see cloud/rain for the weekend? 

[SomeLikeItHot]

Perhaps this could help

 

http://www.pnas.org/content/105/34/12140.long

 

Meant to be some record of London's air pollution around accoring to this Lomborg article. CAn't find the series he refers to though.

http://www.guardian.co.uk/environment/2001/aug/15/physicalsciences.globalwarming

 

We often assume that air pollution is a modern phenomenon, and that it has got worse in recent times. However, air pollution has been a major nuisance for most of civilisation, and the air of the western world has not been as clean as it is now for a long time. In ancient Rome, the statesman Seneca complained about "the stink, soot and heavy air" in the city. In 1257, when Henry III's wife visited Nottingham, she found the stench of smoke from coal burning so intolerable that she left for fear of her life, and in 1285 London's air was so polluted that Edward I established the world's first air pollution commission. Shelley wrote: "Hell must be much like London, a smoky and populous city."

For London, the consequences were dire. In the 18th century it had 20 foggy days a year, but this had increased to almost 60 by the end of the 19th century: this meant that London got 40% less sunshine than the surrounding towns, and the number of thunderstorms doubled in London from the early-18th to the late-19th century.

We have data for air pollution in London since 1585, estimated from coal imports till 1935 and adjusted to measured pollution from the 1920s till today. This shows how levels of smoke and sulphur pollution increased dramatically over the 300 years from 1585, reaching a maximum in the late 19th century, only to have dropped even faster ever since, such that the levels of the 1980s and1990s were below the levels of the late 16th century. And despite increasing traffic, particulate emissions in the UK are expected to decrease over the next 10 years by 30%. Smoke and particles are probably by far the most dangerous pollutant, and London's air has not been so free of them since the middle ages.

[SomeLikeItHot]

AS to the working week question victorian times had a 6 day working week tpyically with only sunday off.  

 

In the earlier part I would have thought much of the polution was private heating rather than industry which would have really kicked off more 1850s.

 

A interesting comparison would be if you could get temp records for France during the period of the revolutionary calendar when they had a ten day week with one day off. -  Oct 1793 to April 1802 I think.

 

Also if its a local efect you could perhap trace the strength of the effect to temperature records in various global regions with different industrialisation dates (eg Nth america, Germany etc)

Edited June 12, 2013 by SomeLikeItHot

[Gray-Wolf]

You might also want to look at the development of smoke stacks which moved the pollution higher and higher through the period?

[Roger J Smith]

Very interesting replies, thanks for all that. I am going to show some of the findings in preliminary form here, the graphs could use a little work but you'll see what I have found in this methodology.

 

First, I should mention, the data have been arranged into days of the week by a careful iteration process that took notice of the following facts:

 

** the first Monday in the data set was 6 Jan 1772.

 

** there were no leap year days in 1800 or 1900

 

** with those exceptions, first Mondays are one day earlier each year with the exception of years that follow leap years, and those are two days earlier

 

** with that in mind, the data set was then constructed so that each year began with the first Monday which of course falls in the range of 1-7 January.

 

** the winter portion of the year, approximately 37%, includes all weeks that start in Jan, Feb, Mar until the 20th, last few days of Oct (day 300 of year began this portion), Nov and Dec. The rest of the year is essentially the portion where ambient temperatures and solar effects make the need for residential or business space heating less continuous or absent. This does not imply no output of coal soot in early periods or greenhouse gases in later periods, but from this source, reduced amounts.

 

** daily CET anomalies are based on my research study (see historical weather section) where I averaged CET values for each day of the year. Any variations in that daily record not related to the annual curve related to insolation would of course be interesting too, for example, when assessing any differences around Christmas and New Years which fall randomly throughout the week. These anomalies are then fed into the tables for days of the week and anomalies are then averaged for each day (of the week).

 

I will show the overall averages later, they are not all that interesting anyway since we're looking at different historical periods that have different dynamics in play, so it's the 30-year (and from those the 15-year) periods that are more interesting.

 

Looking at the first 30 years (1772-1801) this graph shows that the winter portion of the year showed a tendency for progressive cooling from Monday to Thursday, then a progressive rise to Sunday with a flat-topped peak Sunday-Monday. The range of anomalies for the winter period is .25 C deg with the lowest value -0.83 (compared to 240-yr average) on Thursdays. Meanwhile the non-winter portion of the year varies somewhat less and for this interval was generally between 0.2 and 0.4 positive (this may indirectly feed into discussions that early CET daytime readings were perhaps somewhat skewed by poor shading, the mean daily anomalies start to get closer in winter and non-winter as we move forward). The year is a weighted average of these two periods since winter (as defined for this study) is only three-eighths of the total data. Inspection of weekly averages through the heart of the summer season suggests that the weekly variation goes virtually extinct in the warmest months then begins to return in later September and through October.

 

 

The second graph here shows how similar the next 30 years were, especially for winter data. The data for 1802-31 are shown superimposed on the data for 1772-1801. The winters were slightly milder in general but a very similar trend curve was obtained for winter. The annual curves are almost overlapping and the non-winter half year is also similar although now the data are closer to long-term averages (perhaps once again feeding into that discussion about improving observation techniques, better shading of thermometers).

 

 

The last item I'll post today is this more complex graph which shows the 30-year averages for eight intervals (labelled on the graph). In general, each 30 year period is a bit milder than the one before it when compared to long-term averages so the layers go up the graph, but there's one outlier, the seventh interval (1952-81) is somewhat cooler than the two before it, and so the graphed curve for that interval is placed 0.3 deg above its actual position, as illustrated by the telephone pole like symbols -- the values are at the base of these "poles" and not within the colour code of the last thirty year interval which shows the sharp jump that has been labelled the AGW period and "m'oder'n winter" by many.

 

 

Now, looking at this composite in some detail, the Thursday cold feature in the first two intervals begins to migrate earlier in the week towards the Victorian era. The third interval has Wednesday as its coldest day, then the fourth and fifth intervals Tuesday (this spans 1862-1921). The sixth interval, 1922-51 has Monday as its coldest day, then it's back to Tuesday in the presumably coal-choked period 1952-81. The most recent thirty years show up with a Sunday minimum.

 

I will comment further tomorrow, the 15 and 10 year sub-intervals are also interesting especially for the transitional period 1982-91 which is essentially a flat signature. It appears from this analysis that the change to a warmer work week began in the 1992-2001 data and has peaked more strongly since then. Although not part of the data set, 2012 and what we have had so far of 2013 continue the trends of 2002-2011 and if included would not change the appearance of the graphs very much.

 

 

 

 

 

 

[Roger J Smith]

Meanwhile, that third graph can also be drawn up in a less confusing way with successive colour layers, see the previous post for the discussion, but this shows the eight intervals more clearly. Just realize that they all get a bit warmer except for the 7th out of 8 and that one is slightly cooler than either 5 and 6, so that the actual temperature increase from interval to interval is sometimes less than implied by these equal separations. The graph does show the eight intervals in chronological order within equal strips -- the inter-segment variation is sometimes a little more or less than the average which is about 0.25 C deg (range from warmest to coldest day). Below the graph, I have attached a table of numerical values for the eight periods below this alternate graph. See the previous post for a discussion of how those values are derived.

 

 

Basically, the white segments of the eight interval graph represent the colder parts of the week. Notice how that begins at Thursday at the bottom (first interval 1772-1801) then shifts gradually towards Monday then back to Tuesday as time moves forward. The major change over time is that the second half of the "work week" gets relatively warmer towards the end of the series.

 

DAYS OF THE WEEK, ANOMALIES in 30-YEAR INTERVALS (C deg)

 

Winter season defined as late October to late March

______________________________________________________

 

Period __________ MON __ TUE __ WED __ THU __ FRI __ SAT __ SUN

 

1772-1801 ______ --0.59 _  --0.58 _ --0.72 _ --0.86 _ --0.79 _ --0.64 _ --0.59

1802-1831 ______ --0.33 _ --0.31  _ --0.41 _ --0.39 _ --0.40 _ --0.33 _ --0.35

1832-1861 ______ --0.16 _ --0.33 _ --0.35 _ --0.32 _ --0.23 _ --0.20 _ --0.10

1862-1891 ______ --0.07 _ --0.18 _ --0.23 _ --0.13 _ --0.10 _ --0.17 _ --0.07

1892-1921 ______ +0.11 _  +0.04 _ +0.14 _ +0.28 _ +-0.35 _ +0.24 _ +0.19

1922-1951 ______ +0.15 _ +0.17 _  +0.26 _ +0.24 _ +-0.25 _ +0.28 _ +0.25

1952-1981 ______ -+0.23 _ +0.09 _ +0.16 _ +0.19 _ +-0.26 _ +0.21 _ +0.19

1982-2011 ______  +0.81 _ +0.89 _ +0.95 _ +0.91 _ +-0.96 _ +0.84 _ +0.86

 

Today I will also add the details of 10-year averages over the last two 30-year segments of the study. The graph below shows these six decades (starting with 1952-61, ending with 2002-11) with the six segments separated by roughly equal graph portions. Letters on the graph identify the decades, A is the first of them and F is the most recent. 

 

 

This is what the graph tells us about recent trends.

 

A (1952-61) looks quite similar in profile to the early years of the study (1772-1831) and somewhat continues a trend that was typical of the previous 30 years, which had a cold start to the week and a slightly milder end including the weekend. The range is quite large, from +0.10 on Thursdays to +.42 on Sundays. Note this was the decade that includes the London smog episode of Nov-Dec 1952.

 

B (1962-71) was a colder decade but returns to the profile typical of the early 20th century, a cold start to the week (Tuesday -0.31 C) and a relatively mild end of the week (Friday -0.04 C). The very cold year 1963 had a different profile from this average so it doesn't drive the signal here.

 

C (1972-81) was a mild winter decade and the mildest part of the week was Thursdays and Fridays which averaged about +0.65, the coldest part was from Saturday to about Tuesday with days around +0.40 C.

 

D (1982-91) had quite variable temperatures in the winter seasons and on average this decade had a very weak variation of temperature. Looking at individual years, the colder winters of the mid-decade seemed to be coldest around mid-week.

 

E (1992-2001) was a much milder winter regime and there was a tendency for the mildest weather to come around Friday and the weekends especially towards the warmer end of the period.; The data for the week continue the low-variance regime of the previous decade to some extent but it does appear that the next decade;'s signature was being gradually established after the major El Nino warming year of 1998.

 

F (2002-11) has not totally warmed beyond the previous decade, weekends and Mondays are colder in general than the 1992-2001 decade, but a higher peak of warmth occurs on Thursdays when the anomaly reaches +1.18 C.

 

The anomalies for 2012 (not part of the overall study) and the first half of 2013 were

 

MON +0.87, --0.73

TUE +1.01, --0.55

WED +1.26, --0.96

THU +1.00, --1.42

FRI +0.74, --1.05

SAT +0.85, --1.10

SUN +0.26, --1.60

 

Note that Sundays turn quite a bit colder in both recent years. In the cold year of 2010 (note that the study looks at calendar years and not winters so this includes Jan and Dec 2010), the averages were

 

MON --3.01

TUE --1.96

WED --1.55

THU --0.98

FRI --0.79

SAT --1.49

SUN --2.57

 

In 2011, a much milder year the averages showed a similar late work-week peak:

 

MON +1.30

TUE +1.45

WED +1.80

THU +2.37

FRI +2.00

SAT +1.70

SUN +2.08

 

That's it for now, a discussion of this would be most welcome.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

[SomeLikeItHot]

Its interesting stuff, Would it be worth looking at diurnal range as well? I would think that pollution might have a stronger impact on that than the absolute level. Remember seeing that the freeze on air traffic after 9/11 had the impact of temporarily increasing the diurnal range from lack of contrails. Would this be another factor affecting recent years. I believe weekends are busiest for airflights typically.

[Roger J Smith]

I believe that diurnal range has been narrowing since about 1980 at many stations, as the recent AGW period (whatever its cause) has featured milder nights more than warmer daytimes. This is reflected in the larger deficit of record lows than surplus of record highs. This is quite anecdotal, but at Toronto which has a record going back to 1841, only January has broken a monthly record high (2005) since December did so in 1982. Otherwise, going back, the monthly records are in these months: Jan 2005, Feb 1954, Mar 1946, Apr 1842, May 1962, June 1964, July 1936, Aug 1918, Sep 1953, Oct 1963, Nov 1950 and Dec 1982. Record lows only happen nowadays in most North American stations when there is a highly anomalous winter cold air mass, the sorts of radiation cooling of stale polar air masses that used to edge out old record lows just seem to cloud up more and more with unspectacular minima. That's where a lot of this recent "warming" actually shows up in the hourly breakdown, in overnight lows on dry nights that are colder than average. Otherwise it's hard to find the signal of warming when you compare air masses. Of course a separate issue is whether air mass distribution has changed. That gets you no closer to knowing whether the cause is man-made or natural variability.

 

I am not as intimately familiar with UK records despite having done those tables on daily records, but my impression is that the same principles apply, you can blow away a cold record with a cold enough anomalous air mass but otherwise setting a new record low overnight is increasingly difficult. For one thing, with growing urbanization, the spreading urban heat islands must be leaking a bit of warmth out into the rural areas that are still used for the major historical comparisons. I feel certain that mid and high level cloudiness is increasing. Going back to the Toronto record, most of the sunshine record highs come from the 1960s when there seemed to be more frequent anticyclonic spells of settled weather lasting many days. You might suppose that winter sunshine would be correlated with cold weather but actually at Toronto you can find a bimodal distribution of sunny months, some of them quite mild, most of them colder than normal. From about April to October, above normal sunshine will correlate mainly with above normal temperature. March and November tend to be totally randomly correlated between sunshine and temperature. I wouldn't be surprised if the same were true in the UK climate except that almost all sunny winter months would be cold ones.

 

 

[Roger J Smith]

This table shows which day of the week was warmest in each year of the period 1772-2013. The numbers are to be read as 1 = Monday, 2 = Tuesday etc to 7 = Sunday. 45 for 1856 indicates 4,5 tied that year. 15 in 1914 indicates 1,5 tied for that year.

 

DEC __ 0 _ 1 _ 2 _ 3 _ 4 _ 5 _ 6 _ 7 _ 8 _ 9____ most frequent___________ cum frequency

 

1770s__ --_ -- _ 6 _ 2 _ 1 _ 6 _ 3 _ 2 _ 6 _ 6____ 6 SAT________________ 01,02,01,00,00,04,00

1780s__ 7 _ 7 _ 2 _ 1 _ 4 _ 6 _ 7 _ 2 _ 3 _ 7____ 7 SUN________________02,04,02,01,00,05,04

1790s__ 5 _ 7 _ 3 _ 3 _ 2 _ 2 _ 5 _ 2 _ 6 _ 7____ 2 TUE________________02,07,04,01,02,06,06

 

1800s__ 7 _ 3 _ 4 _ 1 _ 5 _ 7 _ 2 _ 6 _ 4 _ 5____ 4,5,7 THU FRI SUN_____ 03,08,05,03,04,07,08

1810s__ 4 _ 4 _ 6 _ 1 _ 6 _ 6 _ 2 _ 1 _ 1 _ 7____ 1,6 MON SAT__________06,09,05,05,04,10,09

1820s__ 3 _ 6 _ 7 _ 2 _ 7 _ 7 _ 3 _ 2 _ 7 _ 5____ 7 SUN________________06,11,07,05,05,11,13

1830s__ 4 _ 3 _ 6 _ 7 _ 1 _ 1 _ 3 _ 4 _ 3 _ 5____ 3 WED_______________ 08,11,10,07,06,12,14

1840s__ 2 _ 7 _ 4 _ 2 _ 6 _ 2 _ 3 _ 7 _ 1 _ 7____ 2,7 TUE SUN___________09,14,11,08,06,13,17

1850s__ 5 _ 2 _ 7 _ 3 _ 2 _ 2 _ 45_4 _ 5 _ 2____ 2 TUE________________ 09,18,12,09*08*13,18

 

1860s__ 6 _ 5 _ 2 _ 4 _ 1 _ 4 _ 7 _ 7 _ 1 _ 5____ 1,4,5,7 MONTUFRSUN __11,19,12,11*10*14,20

1870s__ 1 _ 7 _ 6 _ 2 _ 7 _ 5 _ 1 _ 4 _ 1 _ 3____ 1 MON _______________14,20,13,12*11*15,22

1880s__ 6 _ 4 _ 7 _ 1 _ 3 _ 6 _ 1 _ 5 _ 6 _ 7____  6 SAT  _______________16,20,14,13*12*18,24

1890s__ 5 _ 3 _ 7 _ 3 _ 5 _ 4 _ 7 _ 5 _ 4 _ 5____ 5 FRI ________________ 16,20,16,15*16*18,26

1900s__ 3 _ 5 _ 7 _ 7 _ 6 _ 4 _ 4 _ 3 _ 6 _ 5____ 3-7 WED-SUN _________ 16,20,18,17*18*20,28

 

1910s__ 1 _ 6 _ 5 _ 2 _ 15_ 5 _ 5 _ 5 _ 1 _ 1___ 5 FRI _________________ 19*21,18,17*23,21,28

1920s__ 6 _ 1 _ 5 _ 1 _ 6 _ 3 _ 1 _ 3 _ 4 _ 6____ 1,6 MON,SAT __________22*21,20,18*24,24,28

1930s__ 6 _ 4 _ 3 _ 1 _ 7_ 4 _ 5 _ 4 _ 6 _ 1____  4 THU ________________24*21,21,21*25,26,29

1940s__ 1 _ 6 _ 2 _ 3 _ 5 _ 3 _ 2 _ 6 _ 1 _ 3____ 3 WED _______________ 26*23,24,21*26,28,29

1950s__ 4 _ 6 _ 2 _ 6 _ 5 _ 3 _ 1 _ 2 _ 7 _ 7____ 2,6,7 TUE SAT SUN ____ 27*25,25,22*27,30,31

 

1960s__ 7 _ 1 _ 5 _ 2 _ 3 _ 5 _ 7 _ 5 _ 3 _ 3____ 3,5 WED FRI __________28*26,28,22*30,30,33

1970s__ 2 _ 6 _ 3 _ 4 _ 6 _ 1 _ 4 _ 4 _ 3 _ 1____ 4 THU _______________ 30*27,30,25*30,32,33

1980s__ 2 _ 1 _ 2 _ 6 _ 3 _ 5 _ 2 _ 7 _ 3 _ 1____ 2 TUE _______________ 32*30,32,25*31,33,34

1990s__ 2 _ 5 _ 1 _ 4 _ 7 _ 2 _ 7 _ 7 _ 5 _ 5____ 5,7 FRI,SUN __________33*32,32,26*34,33,37

2000s__ 2 _ 3 _ 2 _ 3 _ 4 _ 5 _ 3 _ 3 _ 5 _ 6____3 WED _______________33*34,36,27*36,34,37

2010s__ 5 _ 4 _ 3 _ 2 _ - _ - _ - _ - _ - _ - ___________________________33*35,37,28*37,35,37

__________________________________

 

* add one-half to frequency

 

(no trend for last four years except for the steady 363d cycle from 2009 to 2013).

 

The cumulative frequency was quite lopsided towards weekend peaks especially Sundays until about the 1920s since when the weekdays have been catching up.

 

Interesting to note the consistent run of Fridays in WW-I (1914-18) which may indicate all-out industrial production then, probably by WW-II the even more extreme need for production wiped out the "work week" entirely although the trend for 1940-45 is near mid-week on balance, then while the 1990s were largely weekend peaks the 2000s have seen mainly mid-week peaks.

 

Buried in all this may be some natural cycle of about 363-364 days (see my other research thread for ideas on what that could possibly be), as there's a tendency for years to peak a day or two days earlier than previous years as a sub-theme of this data set.