]\gt ; The tionship between Pressure at the Same Lerel in the Free Atmosphere .
By E. H. CHAPMAN , M.A. , D.Sc .
( Communicated by Sir Napier Shaw , F.R.S. Received September 21 , 1920 .
) An important set of correlation coeffic showing the relationship between pressure and temperature at the same level in the free atmosphere has recently been published by W. H. Dines .
* The coefficients are reproduced here in Table I. I.\mdash ; Correlation Coefficients between Pressure and Temperature at the same Level in the Free Atmosphere .
.\mdash ; All available observations from Pyrton Hill , Ditcham Park , and Benson .
These correlation coefficients , ranging as they do from to for the layers between and , present indisputably strong evidence of a very close relationship between pressure and temperature at the same level .
Now it is easily proved that a correlation coefficient is always lowered in the numerical sense by errors of observation .
Hence a point of considerable importance arises with regard to the correlation coefficients of Table I Granted that errors of observation exist in the observations from which the coefficients are calculated , what are the true values of these coefficients , " " The cteristics of the Free Atmosphere W. H. Dines , F.RS .
, 'Geophysical Memoirs , ' No. 13 , Meteorological Office , London , 1919 .
See pp. 34 , 35 , 'Computer 's Handbook , ' Section , Sub-section 2 , Meteorological Office , London , 1916 .
236 Dr. E. H. Chapman .
Pressure and Temperature at the effects of errors of observation being elimin ated ?
It is the of the present paper to answer this question from the data available for the purpose .
Correction of Coefficients of for Probable Errors of Observation .
There is first a piece of statistical theory to consider in its special plication to our problem .
Suppose that we have a series of departures from a mean value .
Let this series be , , , the terms in the series being true values .
Let the errors of observation be , , , respectively .
Then the observed will be , , observed standard deyiation will be where since and are mutually independent .
Now let us suppose that we have a second series of departures from a mean value .
Let this series be , , .
Let the errors of observation be , . .
, respectively , so that the observed values are .
We have as before where is the ubserved standard deviation and the true standard deviation .
We shall now correlate the two series .
The true correlation coefficient is iven by The observed correlation coefficient since and and and are mutually independent .
We havo thus that But and Hunce we get ' or .
( 1 ) Thus , if we know the values of we can calculate the true correlation coefficient from the observed correlation coefficient We shall now find values of for temperature and pressure Qervations in the free atmosphere .
Suppose that we have a series of the Same Level in the Free Atmosphere .
observations at a certain level .
Let the true values of these temperature readings be , . .
Let the errors of observation be , .
( so that the observed readings are , , .
Now , let us put a restriction on the errors of obscl'vation of our temperature readings in this way , let the possible error be imes the true reading , i.e. , } } , and so on .
Suppose that the grcatest term in the X series is .
Then the errors of observation will have a from to , the mean error being zero since positive and ative errors are equally likely .
Thus the of errors of observation is But if be the standard deviation in our " " \ldquo ; series this range is approximately , since errors of observation follow the norn ) law and practically covers the whole in a normal distlibution , being the mean and the standard deviation .
We have then or Now our X series is a series of temperature readings which are likely to be distributed according to the norlnal law .
If we write for the mean of this X series , a close approximation to the value of would be given by .
With this value of we 3 or .
( 2 ) suppose that we have a series oS pressure readings taken at the same levels as our temperature readings .
Let our pressure series be , , taken at the same time as as , and so on .
Let the errol .
S of observation in our series be , , , respectively .
Let us assume that } } , etc. We shall then have .
( 3 ) Then , if we know the values of and , equation ( 2 ) will give us the value of in terms of , and equation ( 3 ) will give us the value of in terms of .
We should then be able to substitute these values of in equation ( 1 ) , and thus obtain the value of the true correlation coefficient , between pressure and temperature at the same level , in terms of and and the observed correlation coefficient Estimation of the Probable of Mecxsurement of Temperature the : From Dines ' figures , * average values of ( temperature ) and ( pressure ) for the year are as given in Table II .
* See pp. 11 , 12 , ' Quart .
Jour .
Roy .
Met .
Soc vol. 40 , No. 169 , January , 1914 .
238 Dr. E. H. Chapman .
Pressure and Temperature at Table II.\mdash ; Average Values of ( temperature ) and ( pressure ) for Winter ( Nov.-April ) , 36 observations , and Summer ( May-Oct .
) , 50 observations .
Height in kilom .
Height in kilom .
Height in kilom .
Height in kilom .
Height in kilom .
Height in kilom .
Height in kilom .
Height in kilom .
Height in kilom .
Height in kilom .
Height in kilom .
Height in kilom .
Height in kilom .
Height in kilom .
Height in kilom .
Height in kilom .
Height in kilom .
Height in kilom .
Height in kilom .
Height in kilom .
Height in kilom .
Height in kilom .
Height in kilom .
Height in kilom .
Height in kilom .
Height in kilom .
Height in kilom .
Height in kilom .
Height in kilom .
Height in kilom .
Height in kilom .
Height in kilom .
Height in kilom .
Height in kilom .
Height in kilom .
Height in kilom .
Height in kilom .
Height in kilom .
Height in kilom .
Height in kilom .
Height in kilom .
Height in kilom .
Height in kilom .
Height in kilom .
Height in kilom .
Height in kilom .
Height in kilom .
Height in kilom .
Height in kilom .
Height in kilom .
Height in kilom .
Height in kilom .
Height in kilom .
Height in kilom .
Height in kilom .
Height in kilom .
Height in kilom .
Height in kilom .
Height in kilom .
Height in kilom .
Height in kilom .
Height in kilom .
Height in kilom .
Height in kilom .
Height in kilom .
Height in kilom .
Height in kilom .
Height in kilom .
Height in kilom .
Height in kilom .
Height in kilom .
Height in kilom .
Height in kilom .
Height in kilom .
Height in kilom .
Height in kilom .
Height in kilom .
Height in kilom .
Height in kilom .
Height in kilom .
Height in kilom .
Height in kilom .
Height in kilom .
Height in kilom .
Height in kilom .
Height in kilom .
Height in kilom .
Height in kilom .
Height in kilom .
Height in kilom .
Height in kilom .
Height in kilom .
Height in kilom .
Height in kilom .
Height in kilom .
Height in kilom .
Height in kilom .
Height in kilom .
Height in kilom .
Height in kilom .
Height in kilom .
Height in kilom .
Height in kilom .
Height in kilom .
Height in kilom .
Height in kilom .
Height in kilom .
Height in kilom .
Height in kilom .
Height in kilom .
Height in kilom .
Height in kilom .
Height in kilom .
Height in kilom .
Height in kilom .
Height in kilom .
Height in kilom .
Height in kilom .
Height in kilom .
Height in kilom .
Height in kilom .
Height in kilom .
Height in kilom .
Height in kilom .
Height in kilom .
Height in kilom .
Height in kilom .
Height in kilom .
Height in kilom .
Height in kilom .
Height in kilom .
Height in kilom .
Height in kilom .
Height in kilom .
Height in kilom .
Height in kilom .
Height in kilom .
Height in kilom .
Height in kilom .
Height in kilom .
Height in kilom .
Height in kilom .
Height in kilom .
Height in kilom .
Height in kilom .
Height in kilom .
Height in kilom .
Height in kilom .
Height in kilom .
Height in kilom .
Height in kilom .
Height in kilom .
Height in kilom .
Height in kilom .
Height in kilom .
Height in kilom .
Height in kilom .
Height in kilom .
Height in kilom .
Height in kilom .
Height in kilom .
Height in kilom .
Height in kilom .
Height in kilom .
Height in kilom .
Height in kilom .
Height in kilom .
Height in kilom .
Height in kilom .
Height in kilom .
Height in kilom .
Height in kilom .
Height in kilom .
Height in kilom .
Height in kilom .
Height in kilom .
Height in kilom .
Height in kilom .
Height in kilom .
A curious result may be noticed in Iable II .
If and were constant , the values of ( temperature ) in summer , and ( pressure ) in winter and summer would decrease with increasing height .
The values of in winter , however , would decrease from 2 to 6 kilom .
, and then increase from 6 to 10 kilom .
According to W. H. Dines , *there is good reason to suppose that the probable error of a recorded in the upper air does not exceed 1o C. Now , as we have already stated , practically covers the range observations in a normal distribution , being the mean and the standard deviation .
Since approximately , where is the error , we can write this range as C. , we get C. for our maximum error expressed as excess over mean .
On a true mean temperature of A. , this gives us a value of , or approximately .
Let us assume , then , as a value of , and , as seems reasonable enough , let us assume the same value for .
Putting in Table II , and substituting the values obtained for and in equation ( 1 ) , we get the equations of Table III .
The equations of Table III can be applied directly to the correlation coefficients of Table I. The winter equations can be used for the trimesters Janualy-March , and the summer equations for the trimesters April-June , July-September , without appreciable error .
Calculated in this the true correlation coefficients between temperature and pressure at the same level are as in Table Table presents some very striking features .
The highest corrected coefficients are for the trimester April-June at heights of 4 and 6 kilom .
These coefficients are and respectively .
If the sumptions we have made can be justified , then we have every reason to assert that the * See p. 30 , ' Computer 's Handbook , ' Section II , Meteorological Office , London , 1917 .
the Same Level in the Free Atmosphere .
Table III.\mdash ; Values of True Correlation Coefficients , , in terms of observed correlation coefficients , , for maximum possible errors of observation in both temperature and pressure readings of 2 er cent. of true readings .
Table \mdash ; Trtle Correlation Coefficients between temperature pressure at the same level , calculated from the observed correlation coefficients of Table I , on the assumption that both temperature and pressure readings are subject to maximum possible errors of 2 per cent. of values .
in kilom .
2 .
4 .
6 .
S. 10 .
Janu farch April-June . .
July-September October-December relation of change of temperature to change of pressure at the same level is practically that of direct proportionality during the trimester April-June at heights 4 to 6 kiloms .
The same assertion may also be made for the trimester January-March for height8 kilom .
, the corrected coefficient in this case being .
Table can be summarised in the statement that , throughout the year , from 4 to 8 kilom .
the relationship between change of temperature and change of pressure at the same level is an exceedingly close one , being sometimes , but not always that of direct proportionality .
Further Inquiry into tloe agnitude of the Probable Error .
The results of Table are sufficiently encouraging to warrant our proceeding to obtain more definite information as to the possible values of el.rors of observation of pressure and temperature readings in the upper air .
This we shall do in the next part of the paper .
Neglecting the variation of gravity with height , and the effect of humidity on the density of the air , the equation may be written lir .
E. H. pman .
Pressure S is the pressure at the surface , is the pressure at height ( ( is harmonic mean temperature of the column from the stlrface to height metres .
Suppose that is found from temperature readings taken ; equal intel.vals of height , and that the of these readings are , , .
Let these true be subject to errors of observation , error of observation being within times true reading .
Then i vere the maximum possible observed value of we should have Hence , if there were no errors of observation in the and , we hold $o the maximm possible observed height , where from ( 4 ) .
the same way we should get the minimum possible observed height where Thus , if there were no errors of obseryation in pressure , and the errors of obseryation in temperature were not greater than times the true reading in each case , we have a possible range for errors in height where is the true height .
Let us now assume a possible error of observation in pressure of times the true value in each case .
Then the maximum value of and being true values , would be , and the minimum value .
Supposing there to be no error in temperature , the maximum value of the observed height , would be given by In the way the minimum value of the observed height would be given by the Same Level in the Free Atmosphere .
If , now , we have errors of observation in temperature up to ti1nes the true values , and at the satle time errors of observation in pressure up to times the true values , , the maximum value of the observed would be given by value of the observed height would be iven by being the true in each case .
The range over which the observed could vary would be eoivon by .
( 5 ) Prof. W. R. Blair*has the of a series of registering balloon ascents made America , in which the balloons followed by two theodolites and the heigbts calculated by triangulation , as well as from the temperature-pressure records of the instruments .
Since the base-line used was 5088 metres ( over 3 miles ) , the heights found by triangulation must be fairly accurate , and we can reasonably take them 10 be true .
1 Table are given those of Prof. Blair 's results of which direct use is made in paper .
" " Free Air Data by Means of Sounding Balloons ' Monthly Weather Rev vul .
44 , , Washington Dr. E. H. Chapman .
Pressure Temperature Table V. \mdash ; - Height fulation .
temperature- pressure record .
Lowest .
Highest .
Metres .
If , then , we put , and in equation ( o ) , and assume a harmonic mean temperature of A. for the air column 0-3000 metres , we get It is convenient to reduce this equation to the form .
( 6 ) Since is posilive is positive and therefore we must have i.e. , Also , since is positiye , the maximum value of is i.e. , the maximum value of is .
This a maximum value of .
If we put in equation ( 6 ) , we get gives us gives us , for we have , and for we the value When , we have metres .
Hence , from equation ( 5 ) , assuming a harmonic mean temperature of A. for the air column 0-6000 metres , we get This equation reduces to .
( 7 ) Reasoning , as before , we get a maximum value for , and for Other values of and from equation ( 7 ) are At 9000 metres , we have from equation ( 5 ) the Same Level in the Free Atmosph assuming a harnlonic mean temperature of A. for the air column to 9000 metres .
This equation leduces to .
( 8 ) From this equation , the maximum value of is , and the maximum value of .
Other values are ; It will be convenient to tabulate the corresponding values of .
and Thi is done in Table Values of and .
If and be truue .
of temperature and pressure at the same level , and are the maximum possible errors of observation in and respectively .
From the numerical values given in Table , we can assume certain possible values of and for the layer of the atmosphere from ) to 9 kilom .
We shall not lay much stress on the values riVen for 9000 metres , since the number of observations for that leight is small .
If reasonably possible value of is .
Assuming and , from Table II and equation ( 1 ) we the results given in Table Table , i.e. , no errors in temperature readings , maximum possible error in pressure readings 2 per cent. of true values .
where inter .
' Dr. .
H. Chapman .
Pressure Temperature at In the results set out in Table , we assumed tb.at there verc no errors of observation in temperature .
Let us now assume that th( are no of observation in ) readings , i.e. , let us assume Table we get a value of , corresporiding to .
Putting , from Table Il and equation ( 1 ) we the results iven in Table VIII .
Table VIII .
, maximum possible error in temperature readings 5 per cent. of true values , no errors in pressure readings .
Valueso fa where rWinter .
Summer .
' It is instructive to compare Tables and VIII , and to see from such a comparison that there is a distinct possibility of errors of observation in temperature having a much more marked effect than errors of observation in pressure increasing the observed correlation coefficients of Table I. The numerical quantity which we have called is defined by the relationship , the true correlation coefficient between temperature and pressure at the same level when errors of observation in both elements have been eliminated , and being the observed correlation coefficient .
All the values of which have been calculated in connection with the of this paper are iven in Table IX .
Reference to Table shows us the true correlation coefficients between temperature and pressure at same level , obtained by using the values of corresponding to .
It is clear that , to get higher values of the true correlation coefficients than those of Table , we must use higher values of than those corresponding to .
Such higher values of , on the whole , are found in Table IX , corresponding to ; ( ii ) ; ( iii ) ; ( iv ) Since it is hardly likely that pressure readings will be subject to no error whatever , we can neglect ( i ) , ( iii ) and ( iv ) in which .
The remaining case , , gives us a maximum possible error the Level in the Atmosphere .
Dr. E. H. Chapman .
Pressure Temperature at of observation in a temperate , ur reading of 3 per cent. of the true value and a maximum possible error of observation in a pressure of 1 per cent. of the true value .
With ordinary temperatures and pressures in the upper air this would mean admitting a maximum possible error in temperature of about C. , and a maximurll possible error in pressure of 8 .
We must emphasise the fact that these are maximum possible errors which would seldom occur .
The corresponding probable errors would be just under C. and 2 .
If Prof. Blair 's observations are correct , and our calculations based on those observations are correct , such errors are certainly possible .
The Corrected Values of the Coefficients .
Assuming , , that we can have , we obtain certain possible values of from Table IX .
These values of are iven for 2 , 4 , 6 , 8 , 10 kilom .
By interpolation , we can get values of for 3 , 5 , 7 , 9 kilom .
Applying these values of to the correlation coefficients of Table I , we get the true correlation coefficients of Table X. Table X.\mdash ; True Correlation Coefficients between Pressure and Temperature at the same level , assuming a maximum possible error in temperature readings of 3 per cent. , and in pressure readings of 1 per cent. of the true value .
If the assumptions and calculations of this paper are correct ( and there appears to be no reason to doubt them ) , we have obtained some striking results in Table X. Three of the coefficients in that Table are actually unity .
Even if we cannot conclude that the relationship between pressure and temperature at the same level is always that of direct proportionality , we have proved that it is so at 4 and 6 kilom .
for the first three months of the year , and at 8 kilom .
for the second three months .
The correlation coefficients of Table X are shown raphically in the accompanying figure .
The present paper had its origin in a gestion made to me by Sir Napier Shaw , F.II .
S. I am glad to have this opportunity of expressin my thanks the Same Level in the Free Atmosphere .
to Sir Napier for his original suggestion , and also for subsequent estions , which led to the calculation of the correlation coefficients of Tables etres and X. I cannot do better than conclude the paper by quoting the following remarks made by Sir Napier when the paper was submitted to " " The least that can be said about the extrordinary results of Table X is that the relation of change of temperature to change of pressure has been proved to be that of direct proportionality at 4 kilom .
and 6 kilom .
in the second trimester and at 8 kilom .
in the first trimester , while from 3 to 8 kilom .
during the whole year the departure from proportionality is ever more than 20 per cent. " " It would , therefore , be very profitable to consider what it is that in certain trimesters spoils the relation which is obviously operative in others .
We know , for example , that there is a seasonal variation of pressure and temperature at all levels , so that there must be presumably a seasonal change VOL. XCVIIL\mdash ; A. Sir A. Schuster .
in the constant of the regression-equation within a trimester which will affect the coefficients .
We know also that at the surface there are speciinens of at the same pressure with very different temperatures ; and if these specimens passed to the pressure of the assigned levels by change of pressure alone , the same pressures would have differeIlt tenlperatures at those levels .
Hence in random sampling one get temperature changes without pressure changes , and versed .
Also the water-content of the a great influence upon the temperature change due to pressure change .
Further , Dines has indicated that if at any level isobars are also isotherms , there is no change of wind with .
This is sometimes true for a considerable range of height , sometimes not .
Why ?
The present paper is an vitation to somebody to take the next step in explaining the departures from these laws of the upper atmosphere .
On the Absorption Scattering of Light .
By SIR SCHUSTER , F.R.S. ( Received October 28 , 1920 .
) When a beam of light passes a material medinm it suffers gradual extinction , mainly through molecular scattering , but partly also through conversion into heat .
The two processes should be clearly distinguished , as there is some danger of confusing the issue if the term ' absorption\ldquo ; is applied to what is a mere diversion of the light energy from its original direction .
It is easy to form an idea of the mechanism involved in scattering , but the process of absorption is more difficult to represent analytically .
In order to obtain a clearer view of its fundamental featm.es , I pose to discuss the transmission of light through a medium the molecules of which have only one free period of vibration .
We may base the discussion on th differential equation the position of an oscillator .
( 1 ) Its well-known solution is , ( 2 ) where

