]\gt ; 2:33 On the and of By HAROLD JEFFREYS , M.A. , D.Sc .
( Communicated by Sir Napier Shaw , F.R.S. Received June 3 , 1919 .
) In studying the relation between the observed wind at various heights , and the pressure distribution at the surface of the one usually takes as the standard of comparison the " " geostrophic\ldquo ; wind , which is determined by the pressure distribution alone .
If its velocity be denoted by is defined by the equation 2 siu where is the earth 's ular velocity of rotation , is the latitude , is the density of the air , is the pressure , and is the rate of increase of pressure per unit distance normal to the isobars .
Self-consistent units are of course to be employed .
The direction of this ideal wind is , so that the low pressure is on its left .
If certain conditions were satisfied in the atmosphere , it is easily shown* that the actual wind would always be strictly equal to the geostrophic wind apart from conceivable free vibrations compatible with constant pressure ) .
These conditions are that ( 1 ) the pressure distribution should not be changing with the time ; ( 2 ) friction should be absent ; ( 3 ) the velocity should be so small that its square can be neglected .
None of these conditions holds in practice .
The failure of the first is the chief reason for official encouragement of meteorology .
Sir Napier Shaw has shown that the instantaneous centre of the motion in a cyclone moving from west to east would be expected to be north of the centre of the isobars , causing the win in front of the cyclone to blow outwards from the region of low pressure , while that on the following side blows inwards .
The second disturbing influence\mdash ; friction\mdash ; is ordinarily the most important near the surface , but dies out at a height of a few thousand feet .
Its amount depends on several factors , but chiefly on the temperature lapse rate and the wind velocity ; and G. I. Taylor has shown that it causes the surface wind to deviate towards the region of * See , for instance , ' Phil. Mag vol. 37 , pp. 1-3 ( 1919 ) .
Geophysical Memoirs of the Meteorological Office , ' No. 12 , p. 35 .
VOL. XCVI.\mdash ; A. Dr. H. Jeffreys .
On the Relation between low pressure , while its velocity is decidedly less than that of the ic wind .
The third condition is usually approximately satisfied , except near the centres of cyclonic storms .
If the pressure distribution were stationary , the terms neglected.would amount simply to the " " cyclostrophic\ldquo ; term of the ordinary theory ; but , owing to its variation with time , they become more complex .
The observed wind at the surface has usually between 30 per cent. and 80 per cent. of the geostrophic velocity , and it generally deviates by 1 to 5 points from the isobar towards the region of low pressure .
Now , the first and third of the above-mentioned disturbing influences would make the wind deviate clockwise and counter-clockwise equally often , since they produce opposite effects on two sides of a ession .
The deviation in direction is therefore to be attributed mostly to friction .
The same applies to reduction of velocity .
Accordingly , friction is to be regarded as dominant cause of the difference between the actual and theoretical winds The present investigation was undertaken to find out how much of the large range of variation in the difference , which occurred even with constant geostrophic velocity , was due to variations in the friction itself , and how much to the other neglected factors .
If denote the velocity of the surface wind , and its deviation from geostrophic wind , measured counter-clockwise , and if denote then Taylor showed that , provided the turbulence was the same at heights and the system was steady , would be equal to .
This result independent of the amount of turbulence , so long a@ it does not change with height , and of the geostrophic velocity itself .
This fact provided the for the first attack on the problem .
The correlation betweer and , and between and , were calculated from a series oi 142 observations taken on merchant ships in the North Sea , supplied the Marine Division of the Meteorological Office , the geostrophic wind found from the working charts ( .
interval ; scale 1 : 10000000 ) .
Only those observations for which good determinations of the gradient available were employed .
Now would on the assumption to tested always be unity , except for the eifects of the variation with time in the pressure distribution and accidental errors in the determination of and .
Hence should be apart from these effects .
, on other hand , would depend on the variations in the pressure distribution , the turbulence , and also on itself , as well as on the accidental errors of and G. Thus would be expected to be considerably greater than for the contrary to hold would imply either that there were great accidental errors in , or that the assumption that the turbulence is independent of the Wind and bution of Pressure .
height is seriously in error .
Actually , the results were found ; the winds have been classified according to the quadrant in which the trophic wind lies .
All velocities are in metres per second .
Thus , in three of the four quadrants , reater than , contrary to what migbt have been expected ; while the remaining one is based on too few observations to be relied upon .
In no case does approach unity .
This result might be attributed to accidental errors in .
For so low a correlation coefficient to be due to this would , however , require the standard error of to be greater than the ratio of the standard deviation of to the mean value of , and the range of wind velocity was such that this was of order 1 .
Thus would have to be liable to errors of several points .
Now , the error of is the algebraic difference between those in the directions of the true and geostrophic winds .
The former class are probably small ; sailors ' estimates of the direction are said to be very concordant and trustworthy , even in spite of the complication arising from the motion of the ship .
Estimates of the direction of the geostrophic wind are open to more suspicion .
The isobars on the charts are obtained by drawing srnooth curves that will fit a comparatively small number of observations , and the error is therefore the sum of the error of this interpolation and that introduced in measuring the charts .
As the days were selected so as to give good determinations of the gradient , the latter error was unlikely to exceed a point , but the former is , from the nature of the case , unknown .
It is quite possible that the true isobars are really wavy in form , the wavelength being of the order of 100 km .
, and that the practice of smoothing these waves out leads to considerable errors .
The smallness of may then be attributed to two main causes : first , great in turbulence with height ; and second , irregularities in the isobars .
The fact that is systematically less than might be held to support the former view , for contortion of the isobars would presumably increase and decrease equally often ; but , as a systematic under-estimation of the surface wind by the observer would act similarly , one cannot place much reliance on this .
The vertical distribution of turbulence is closely connected with the .
H. .
On the Relation between temperature lapse-rate , and this in its turn presumably with the difference in temperature between air and sea , which is given in the ships ' logs .
An attempt was made to test the turbulence hypothesis on this basis , but it proved a failure , for the temperature differences recorded were less than could be accounted for by the conditions of exposure in a ship , and accordingly no significance could be attached to the small correlation found .
A series of obSeryabionS , taken on ships of H.M. Navy , and kindly supplied for this purpose by the Hydrographer , was made use of in a further investigation of this question .
Enough data were available for it to be possible to diyide them into 16 classes , whose average number was about 40 , according to the strength of the geostrophic wind and the quadrant whence it blew .
Thus Tables showing the frequencies of values of and in the separate classes could be constructed ; and if quadrants differed systematically among themselves , or if strong winds systematically gave larger or smaller values of or than light winds , one would expect these facts to be discovered by this method .
Two separate classifications were carried out : one gave in each class the number of times the surface wind veered or backed by any number of points from the geostrophic wind and the other the number of times lay between any two multiples of 12 per cent. The results are given in the two following Tables ( p. 239 ) .
The number in each compartment is the number of times a geostrophic wind of the appropriate class , indicated in the first column , was found to be associated with a direction deviation or velocity ratio within the limits indicated at the top .
The column includes those few instances where or lay outside the limits of the Table .
These frequency distributions are shown in diagrammatic form in the figures on pp. 237 , 238 .
In these the length of any projection from a circle is proportional to the percentage of the winds of the appropriate class that have the value of or indicated by the number just inside the circle ; is expressed in points , numbers on the clockwise side of the zero denoting veer of the geostrophic wind from the surface wind , while numbers on the counter-clockwise side denote backs ; is expressed in percentages , and all instances where it lay between consecutive multiples of 12 per cent. are rouped together .
With regard to the differences in direction we notice ( 1 ) veers of over 4 points and backs of any amount are very rare .
( 2 ) there is a sudden rise in frequency between and 1 point ; the decline for greater values of is much more gradual , so that the frequency curves are violently asymmetrical .
In every class where it is possible to identify the commonest Wind Distribution of deviation with any certainty the number of deviations greater than this is greater than the number of smaller deviations .
Thus the asymmetry general property of the classes , and not a peculiarity of any particular group of wind-strengths or directions .
LIGHT BREEZES : under STRONG BREEZES : to This asymmetry provides a further argument against the attribution of unexplained variations in to observational error .
For the directions of the geostrophic and true winds are both directly observed , and there is no reason in either case why errors of opposite signs should be unequally probable .
Thus the deviation-frequency curves , if observational error played a large Dr. H. Jeffreys .
On the Retation between part in their determination , should be symmetrical ; or if the mber of observations per class were insufficient to give symmetrical curves , the mode STRONG WINDS : to GALES : over would not lie systematically oftener on one side of bhe mean than the other .
We must , therefore , look elsewhere for the cause of the skewness .
Wind Distribution of Pressure .
Table I.\mdash ; Deviation of the Surface Wind from the Gradient .
Deviat.long.ovorove .
Table II.\mdash ; Frequencies of different Values of Dr. H. Jeffreys .
On thoe Relation between The error introduced in smoothing the isobars may perhaps be a oause , but there are difficulties in the way of supposing it to be the chief .
Suppose for tho sake of definiteness that the smoothed isobars are straight and parallel , and that the corresponding geostrophic wind is south-west .
Then the above counts show that the surface wind is likely to be anything from south-west to south , and it is more likely to be one of south-west and south-west by south than to be one of south-south-west , south by west , or south .
If the true isobars were sine curves , and the deyiation caused by turbulence constant , the deviation-frequency curve would be symmetrical ; for the sine curves would give a frequency distribution symmetrical about , and the constant effect of turbulence would merely amount to shifting this curve bodily to one side .
Suppose , however , that the turbulence gives a constant deviation and that the true isobars systematically cross the smoothed ones from west to east at a smaller angle than from east to west .
Then throughout the long stretches where the mutual inclination is small the effect of turbulence is partly counteracted by the slope , while they combine in the short stretches where the slope is great and reversed .
Thus most of the deviations would be small ; there would be few great ones , and the frequency curve would be notably asymmetrical in just the observed way .
So far this estion , therefore , appears satisfactory ; but its fundamental hypothesis is difficult to grant .
Why should the true isobars systematically cross the smoothed ones at a smaller angle one way than the other ?
Why should this ence be practically independent of the direction of the isobars themselves ?
Until we have direct theoretical or observational reason to believe in this peculiar systematic contortion of the true isobars , we should prefer a hypothesis that does not require it .
There remains yet another difficulty .
If the normal distance between consecutive smoothed isobars be , and if be the deviation caused by turbulence and the inclination of the true isobar to the smoothed one , we have Also the normal distance between consecutive true isobars is reduced to , so that the true geostrophic wind is proportional to .
Hence , since , the geostrophic wind derived from the smoothed isobars , varies only slowly , the value of obtained should show an intimate relation with , i.e. , with where is a constant .
The theory leaves little room for inaccuracy in this ; but it will be seen later that the correlation between and is so slight that it is even possible to doubt its This fact amounts practically to disproof of the hypothesis thab the asymmetry is due to contortions in the true isobars ; such contortions probably give only Wind and Distribution of Pressure .
a fairly symmetrical variation in which is largely masked by the asymmetrical variation due to some other cause .
in the distribution of pressure cause the velocity and direction to differ somewhat from the values they would have if the distribution were steady .
If , however , such changes could give rise to an asymmetrical frequency distribution , the asymmetry in a cyclonic disturbance would presumably be different in front from behind .
Suppose , for instance , that this caused a frequency distribution in the south-west quadrant in which the commonest deviation was , and more winds deviated by more than A than by less than A. Then if the commonest deviation in the north-east quadrant were , we should expect that more winds there would deviate by less than than by more .
A similar reversal would between the north-west and south-east quadrants .
Actually we find that though the asymmetry not equally pronounced in all quadrants , it nevertheless acts in the same direction in all .
Thus the most important cause to consider is one that will always make more winds deviate by more than the mode than deviate by less .
Causes acting sometimes one way and sometimes the other may be fairly important ; but it is clear that they cannot be the most important , for then the frequency distribution would sometimes have its asymmetry reversed in an indubitable manner , which never occurs .
Thus the motion of the isobars is not the most important cause , whether it operates through the inertial terms in the equations of motion or through the difference in time between the wind and pressure obseryations .
The disturbing cause not yet considered and rejected is turbulence .
This term is here used in its widest possible sense .
Atmospheric eddies are of many kinds , being produced by any agency that can cause local instability .
The friction of the wind over a flat surface , for instance , causes small eddies to form near it , and these rise through the air , spreading out and becoming less violent as they go ( like ordinary smoke rings ) and taking with them their momentum and heat .
Wind blowing past an obstacle would in the ideal ses considered in hydrodynamical theory give a definite surface with differenlt velocities on its inside and outside ; actually this is only an approximation , for this state is unstable , eddies form along the surface , their motion causes a transfer of momentum which reduces the discontinuity .
Air heated uniformly.all over the bottom oo.radually acquires a tempel.ature lapse-rate greater than the adiabatic ; when this happens the system ) Ules unstable , and convection currents are formed .
Lord Rayleigh has that the diameters of the regions of rising air will be comparable with , ut obably greater than , the height of the homogeneous atmosphere : thus the are on a 'Phil .
Mag vol. 32 , pp. 529-546 ( 1916 ) .
Dr. H. Jeffreys .
On the Relation between Wind and Distribution of Dr. H. Jeffreys .
On the between very different scale from the ordinary frictional eddy .
Yet they leseinble it .
in providing a mode of transfer , by nlass-motion , of momentum and heat ; they thus produce a pseudo-viscosity .
Local kinds of instability , such as great temperature lapse-rates at particular heights , also give rise to eddy motion on an intermediate scale ; examples are afforded by sea-fog and by most clouds .
* The combination of a great lapse-rate with a wind naturally produces eddies more readily than either of these acting alone .
Enough has been said , therefore , to indicate that the distribution of turbulence in the atmosphere depends mostly on the wind velocity and on the temperature lapse-rate .
Both of these are very variable , and accordingly so is the turbulence .
The effect of turbulence is proportional to a certain coefficient , the eddy-viscosity , whose value at any point is proportional to an average of the product of the range of variation of the wind velocity there into the size of the average eddy that gets there .
Near the ground the size of the eddies is small , but the velocity varies a great deal .
The higher one goes , the more of the turbulence takes the form of convection ; the size of the eddies becomes comparable with the height of the troposphere , while the relative velocity of their parts diminishes .
As these two changes produce opposite effects on the eddy-viscosity , it is impossible to predict on these grounds whether it should increase or diminish with height .
Taylor dealt with the simplest case , where the coefficient is independent of the height , and showed that if this were we should always have The deviation , measured counter-clockwise from the isobar , should always .
be positive and less than points .
It is , of course , a function of and .
An intimate correlation should therefore be traceable between and .
It has already been pointed out that such a correlation would always exist , also , if the deviations were mostly due to contortion of the true isobars ; thus both these hypotheses can be tested ) looking for it .
This was done by means of simple 4-cell contingency tables .
In each class the wind observations were reclas according to the values of and It was found that in the 16 classes the signs of the contingency coefficients .
occurred as follows:\mdash ; Definitely positive , 1 ; doubtfully positive , 2 ; quite uncertain , 5 ; doubtfully negative , 5 ; definitely negative , 3 ( pp. 242-243 ) .
Thus no uniformity was traceable ; even where the sign of a contingency coefficient was clear the magnitude was not great ; so that altogether , while G. I. Taylor , " " The Formation of Fog and Mist ' Proc. Roy .
Met .
Soc vol. 43 , pp. 241-268 ( 1917 ) .
Capt. C. K. M. Douglas , " " The Lapse-Line and its Relation to Cloud Formation ' Journ. Scott .
Met .
Soc , pp. ( 1917 ) .
Wind and Distribution of Pressure .
there is a Saint connection , large values of tending to correspond to small values of , the effect is nificant in comparison with the actual range of variation of both quantities .
The principal cause of the variation is therefore one that does not lead to a strong correlation between and and both of the hypotheses that led us to expect such a correlation are therefore untenable .
We therefore conclude that the turbulence coefficient is ordinarily a function of the height above the ground .
Let us then see whether variations with height are capable of producing such rularities as are observed .
As the effects of changes in the distribution of pressure and wind with time , and second order terms in the equations of motion , have been shown to be of secondary importance in this connection , they will be neglected .
Let the co-ordinates be , where is measured vertically upwards , and let the horizontal components of velocity be and .
Put .
With the appropriate approximations the equations of horizontal motion reduce to the form Take the axis of along the surface isobars , so that If were there would be no friction , and the eostrophic condition would hold .
Hence we have and therefore the equations of motion are Put .
Then these equations combine into one So far the law of ariation of has not been speciiied .
Suppose , illustrative purposes , that it can be expressed in the form Dr. H. Jeffreys .
On the Relation betueen where and are constants .
Put Then The solutions of this are of the form where and are the rates of the equation in and A and are constants as yet undetermined .
We have , and therefore the sum of their real parts is 1 .
Hence either or must have a positive real part .
Let this be .
Then tends to infinity with , and hence the condition that the velocity at a great height shall not tend to infinity is that .
Then At the surface Hence The skin friction at the surface is equal in magnitude to per unit area , where is a constant .
It acts in the direction opposite to the velocity .
It must be equal to times the rate of shear in the air near the surface ; hence Hence has the same phase as .
If the argument of \mdash ; is the constant , we have by equating the ratios of real and imaginary parts on the two sides whence Put 8 .
Then is a constant between and and .
From this we can see that vanishes when , that is , when the turbulence is infinite , and reaches unity when .
The latter is Taylor 's case , corresponding to .
Thus when the turbulence is the same at all heights Wind and Distribution of Pressure .
but when is different from zero , whether positive or negative , is reater than is given by this equation ; it is never greater than It must be noted that when is positive , the eddy viscosity is finite on the ground and steadily increases with height ; when , is negative , it is fiuite on the ground , decreases to zero at a height , and then tends to infinity with the .
We may then infer that great turbulence aloft increases ; presumably if the turbulence decreases with height this ratio is reduced .
Taylor found* that the means of large numbers of observations made this ratio practically unity ; this indicates that if we write , the mean value of is about 1 , though the study of individual observations here carried out shows that it must vary over a wide range .
Deviations of more than four points can easily occur , for is only restricted to be less than , which may reach .
Negative values of seem impossible to account for by turbulence alone .
They rarely occur , and when they do are probably to be attributed to contortion of the isobars of sufficient amount to counteract the effect of turbulence .
The occasional surface winds equal to or even greater than the geostrophic wind are probably due to the same cause .
Such contortion is presumably frequent ; for in the straight isobar regions on the map local showers are common , indicating areas of low pressure and , therefore , curvature of the isobars , Though it is not the chief factor in determining the relation of the surface wind to the radiant , as has already been shown , it may , therefore , have a very important complicating influence .
We may , therefore , conclude that the chief , but not the only , cause of the observed variations in the relation of the surface wind to the radiant is variation in the vertical distribution of turbulence .
This vertical distribution is not , however , usually known , though information is gradually accumulated from various sources It is presumably closely correlated with the lapse-rate of temperature ; this in its turn is correlated with the time of day and plays an important part in determining the occurrence of cloud and fog .
Now as a rule the wind determined by pilot balloon ascents is found to agree fairly closely with the geostrophic wind at altitudes of 3,000 feet and more ; above this the change of velocity is usually small , and the value of the eddy viscosity is , therefore , immaterial , since it only enters the equations of motion through vertical variations in the velocity .
The lapserate above this height , therefore , has little effect on the relati of the surface wind to the gradient .
It may , however , be more or less correlated with the lapse-rate below this level , and in that case there should be a correlation 'Phil .
Trans vol. 116 , pp. 1-26 ( 1915 ) .
C. K. M. Douglas , .
cit. tion between and Distmbution of Pressure .
between cloud-form and the relation of the lower winds to the gradient .
year 's observations at South Farnborough , provided by Mr. B. A. W. Watt , were worked up for this purpose , bnt no satisfactory correlations were tl.aced .
Ths deviations were reater than at sea , presumably owing to reater surface friction ; but those at 1,000 feet and above were never sufficiently greater than their standard deviations ( within classes for which the cloud-form was the same ) for any prediction to be possible .
In the class , where the clouds were st. or st. more than half the sky , the mean value of was , and its standard deviation points ; when there were no clouds below the al. level the mean value of was 3 points and its standard deviation point .
Thus there was no striking correlation between cloud-form and deviation in direction .
A classification of some 600 wind observations over the North Sea , according to their velocities and directions , showed that the most striking feature of the values was their asymmetrical frequency distribution .
From the fact that this was noticeable in nearly every class , it was inferred that it could be produced only by variation in turbulence or systematic contortion of the isobars on a scale too small to be recorded on the weather map .
The latter cause , however , and also such variations in turbulence as keep the coefficient of eddy viscosity the same at all heights , would lead to strong correlations between and , which are not observed .
Hence it is concluded that the principal cause of variation in the relation of the surface wind to the gradient is variation in the vertical distribution of turbulence ; and it is shown that such variation could the effects actually observed .
To complete the test of the theory it is necessary to have direct quantitative information on this distribution , which is not yet available ; but the evidence already accumulated is sufficient to make it exlremely probable that some information on the distribution , and hence on the lapse-rate below 3000 feet , could be obtained by prediction from the observed relation between the surface wind and the geostrophic wind .
July physical account may be given of the influence of tical variations oi turbulence .
Suppose the turbulence to be confined to a very thin Iayer over the ground .
Then above this the wind will have practically the geostrophic value , while in the layer itself the friction on the lower surface will far exceed the geostrophic terms in the equations of motion , on account of the rapidity of the shearing .
Hence the motion in this will be much as if the rotation of the earth were nored , and hence On New of off Poisonous the deviation.in direction wiJl be insignificant .
The friction will theless make the surface velocity less than the geostrophic velocity .
In this case , therefore , is nearly 1 , and is less than 1 .
Thus , if the turbulence decreases with height , we should expect to find less than ; it was found previously that if it increases with height will be greater than ] On New Method of offses .
By .
HERTHA AYRTON .
[ PLATE 4 .
] ( Communicated by Prof. T. Mather , F.R.S. Received August 29 , 1917 .
) On May 6 , 1915 , shortly after the first German gas attack , I read a paper before this Society , dealing with the variations of pressure , and consequent currents , set up in oscillating water by an obstacle the ground under the water .
After the meeting , it occurred to me that , by oscillating an obstacle in an appropriate manner in the air close to our trenches , it ought to be possible to set up currents similar to those in the water , which , while sending the noxious gas back towards the enemy , would , at the same time , } ) our l1len well supplied with fresh air from behind .
The obstacle must be kind of fan , and I thought smoke might represent the gas for experimental purposes .
Apparatus .
Smoke of the right kind was not easy to produce , the difficulty being that it had to be heavier than air , as the noxious gases were said to be .
At last , however , after many trials , I found that smouldering brown paper gave very satisfactory results , if treated in the right way .
It was burnt in a biscuit tin , ( fig. 1 ) , from which the smoke rose through a tube , , to another box , issuing through a second tube , I ) .
When this whole arrangement , except the box , was kept wrapped in wet , the smoke , cooled by the evaporation of the water and by passing through the box , became heavier than air , and fell almost like water from the mouth of the tube 1 ) on to one end of the Table ( 6 feet by 3 feet ) , which had a barrier , , at that end , to protect the smoke from accidental draughts .
On reaching table , the snloks * Held by the Society for deferred publication after the War . .
XCVI.\mdash ; A.

