Oct. 20, 1870] 
NATURE 
501 
that the small sum of 1,000/. annually entrusted by Government 
to the Royal Society for miscellaneous experiments, is adminis- 
tered in a most praiseworthy manner ; and if Government would 
be ready to grant, and the Royal Society willing to undertake, an 
extension of this trust, it would, I think, bea great point gained 
for this class of physical experiments. (I speak only of experi- 
ments, but the encouragement of experimenters is a point of equal 
importance.) But when we come to experiments and observations 
requiring great time, the case is very different. Certain experi- 
ments, whether from the great time they require or the great 
expense they demand, cannot be well performed in a college ; 
while routine and long-continued observations such as those 
connected with the various branches of cosmical physics are 
of such a nature as to require a central establishment to super- 
intend their organisation and reduction. There is thus, I think, 
the necessity for a central establishment of some kind devoted 
to that class of experiments and observations requiring great 
time, great space, and great expense for their completion. 
Referring more particularly to Cosmical Physics, I feel con- 
vinced that meteorology should be pursued in connection with 
terrestrial magnetism and solar observations ; and were a well- 
considered scheme for solving this great problem fairly intro- 
duced, I am sure that scientific institutions and individuals 
throughout the country would do all that they possibly could to 
promote this most important branch of physical research. 
THE BRITISH ASSOCIATION 
SECTIONAL PROCEEDINGS 
tSecTrion A.—MATHEMATICAL AND PHYSICAL SCIENCE 
Barometric Predictions of Weather.—Mr. F. Galton, F.R.S. 
It has long been an established custom to consult the baro- 
meter to learn what the weather is likely to be. Now, I propose to 
investigate the value of this form of barometric authority by 
showing that it is possible to make strict use of the rules of pre- 
diction, notwithstanding the vagueness with which they are enun- 
ciated, and then, by comparing a series of carefully-made 
predictions with facts, to measure the degree to which they corre- 
spond. There is another form of barometric authority about 
which I do not propose to say anything here, namely, where the 
barometer is consulted in connection with the daily Weather 
Report. Owing to the new data thereby introduced, an inquiry 
into the value of those predictions would have to be conducted 
along an altogether different line to that which I am about to 
follow. 
My comparisons between predictions and facts will be based 
upon the tracings of the continuously self-recording instruments 
at Falmouth, established by the Meteorological Committee ap- 
pointed by the Royal Society, which have been published for 
the first quarter of the year 1869. Itis, however, right to add, 
that some years ago I made an elaborate inquiry into the Dublin 
observations during a much longer time, which led, so far as it 
went, to the same conclusions as now. 
I did not publish those inquiries, because I had a misgiving 
which was never wholly removed until I had the opportunity 
now afforded by the above-mentioned publication of studying the 
continuous records of instruments in large numbers. It is said 
that instrumental changes commonly occur in sweeps so large 
-and steady that future changes in them may be to some extent 
predicted by a knowledge of what has occurred. An analysis of 
the Dublin observations, made at intervals of three hours, con- 
tradicted this assertion, but I felt they might be held insufficient 
to dispose of it. It might fairly be said that three hours was too 
long an interval between the observations, and that if the instru- 
ments had been read off more frequently, I should have been led 
to different conclusions. It was necessary to settle this doubt, 
because, as there is certainly some correspondence between the 
barometer and the weather, it followed that if it be possible to 
predict the movements of the former, we shall also, as a matter of 
course, be able in some degree to predict the latter. I therefore 
examined the tracings which represent the continuous records 
of the barometer and other instruments with great interest and 
care, and soon convinced myself that the irregularities of the 
barogram and thermogram were far too great to enable us to 
predict their course from moment to moment. We have only to 
place a paper upon them, so as to hide what follows any given 
instant, and to expose what precedes it, and to move the paper 
forward, step by step, guessing beforehand what we are to see, 
checked by geographical conditions. 
to be convinced of the vanity of our expectations. This basis of 
weather prediction is so manifestly unsound, that I shall not take 
any further notice of it. 
We all know that the weather with which the barometer sym- 
pathises, is considered to consist of three independent variables 
—the velocity of the wind, its temperature, and its damp- 
ness. It is a question how far the direction of the wind 
need be reckoned as a fourth distinct influence. We also know 
that the velocity of the wind is the most important ; it is said 
that when the two other variables are unchanged, and the velocity 
of the wind alone varies, the barometer may range through two 
inches, but that it can only range througha quarter as much 
when either the temperature or the damp are the sole variables. 
I therefore feel at liberty to begin by simply comparing the 
changes of the wind’s velocity with those of the barometer, in 
order to obtain a provisional idea of the manner in which they 
go together. 
Two things are very clear at first sight—the one is that the 
wind’s velocity passes through numberless tumultuous variations 
of which the barometer takes no cognizance, and the other is 
that a connection decidedly exists between periods of storm and 
of fine weather, with barometric falls and rises. What, then, do 
we mean by feriods of storm? How long is the period during 
which the velocity of the wind should be summed up and averaged, 
in order to be made to accord most closely with the barometer ? 
I made several trials, and protracted the results on the same 
time-scale as the corresponding barograms. The ordinates of 
the different points whose position I calculated represented the 
average velocity of the wind during a definite period at the 
moment indicated by the point ; then I connected the points by 
a line drawn with a free hand. In this way I constructed a curve, 
every point of which represented the average velocity of the wind 
during the space of one hour, being half an hour before and half 
an hour after the instant corresponding to that point. In another 
curve, a three-hour period was adopted, and so on. Below all 
these I copied the barogram. 
There could be no doubt, on inspecting those lines, that a one- 
hour period is far too short, that a three-hour is better, a six-hour 
better still, and that a twelve-hour is as good as can be obtained. 
Any period between twelve and sixteen hours seemed equally 
suitable for adoption, some parts of the curve improving in 
correspondence as the period was lengthened, and others falling 
off; but, after a sixteen-hour period, the curve of wind velocity 
became less varied than the barogram, and the maximum ot 
correspondence was passed. Finding the twelve-hour system 
the most convenient to employ, I have adopted it here, leaving 
it to be understood that a different period might be taken within 
the limits named, without sensibly affecting the results. 
The correspondence between the wind curves thus obtained 
and the barograms is respectably close, there being hardly a dip 
or rise in the one which has not a counterpart in the other; but 
they are far from being exactly alike. Neither do the changes, 
of course, in the two curves, bear an invariable relation in point 
of time to one another ; but, as neither of them lags habitually 
behind, they must be considered ox the average simultaneous. 
I do not find the correspondence sensibly affected by making 
broad allowances for the neglected variables. Thus, on marking 
the epochs of cold and dry polar winds in one way and those of 
warm and moist equatorial winds in another way, little or no 
new light was thrown on the reason of the want of coincidence 
of the two curves. It seemed to me, from this trial, that the 
influence of temperature, damp, and wind’s direction, is consi- 
derably less than was commonly believed. 
The parallelism of tue curves was as close in extreme positions 
as in mean ones, which confirms the common statement that we 
must look to differences of barometric height and not to the 
absolute height for signs of changing weather. 
All this is easily compressed into a formula: 4, 4, are two 
successive barometric heights a few hours apart ; 7, (12) 7, (12) 
are the corresponding twelve-hour averages of wind velocity ; m 
isa simple factor to be determined by trial, then 
hy —hy =m {ex(12) — 2, (12)} 
-+ afunction of temperature and another of damp, neither of 
which is of much importance. 
m is strictly constant only for the same season, because the 
range of the barometer is wider in winter than in summer, for 
the same latitude because its range is smallest at the equator, 
and for the same locality because the wind’s velocity may he 
Bearing this in mind, the 
value of m for the first quarter of the year, at Falmouth, as 
