100 MONTHLY WEATHER REVIEW. Apan, 1923
“In the West Atlantic, and particdarl between the
the charts with precision, thanks to the complementary information which we received. The mean number of observations received dail was more than 10. For some days, which coincidei with the passing of a very strong tempest in the North Atlantic! we received an average of 20 Simultaneous observations from ships. The record was reached on the 22d of Feburary at
1
%n the East Atlantic, where the radio traffic is vary intense, as usual, we received only one rather feeble mean of observations of ships, but thanks to the message received from the Eiffel Tower, and to our communica- tions with the steamers of the line from New York, the tracin of the charts was not less precise. We retrans-
tions of the steamship Fmnce; those 01 the Lafuyette, bound to the southwest of $ Rin toward the Azores,
York. As soon as this ship lost direct contact with Brest, we took charge of the retransmission of tho messages. Two appointments were made a day for this purpose, and the observations of the Paris wern ratrans- mitted to the Office National MBteorologique with ours immediately afterwards. “The Weather Bureau is disposed to maintain tlii.4 service of retransmission by the L‘Angot” message; it wishes only when the Jacques Cartier is on the sea, to receive our obseivations, too, These will be added by the Office National MBtQorologique to tho European mess es of “Lyon-Annapolis.I1 We are favorable to
true r61e of ‘ ‘meteoroloaical cruiser” and centrtilizntion station. There would &Lo be a chance to authorize us tn transmit clear1 the positions of the centers of depression. information w&ch would be exchanged between the two meteorological services via Annapolis-Lyon.
Bermudas and the Azores, it was possib 7 e to trace all
m., with 30 simultaneous observations.
mitte f to the Office National MBt6orologi ue the observa-
and finally those of the steams E ip Paris, bound for New
this i T ea, which would make the Jacques C’artirr play a
CONCLUSION.
“We believe we can a f i m that from the point of view of receipt, as well as of trltnsmissions and retransmissions, we have obtained in the course of the voyage of January and Februar 1923, results which are incompnrably
astonishment at the results obtained. He declared to us that he had never had an idea of the precision with
superior to a gl preceding ones. Mr. Bowie showed great
which one could follow the evolution of meteorological
States has verified this success officially, the Office National MBt6orologique will leave no stone unturned to keep up the present organization of the J
by citing the number of meteorologicd messages which were received by the two operators of the Jcsc~lus C‘artier, and the number of telegrams of information, general or particular, which were sent. The first number is 500 messages received; the second, 140 messages sent. Finally, we will remark that the results obtamed are largely due to the reaularity of the radio service and the prcc&ion with whicg the messages are received. The two radiotelegraphers are at present iving us incom- parable service, and the trainmg of ot % er operators for this work would seriously upset the enterpnses by dis- si mting our energy for an undetermined period of time. d e are therefore of the opinion that it is indispensable to rant to these officers a special indemnit in keep’
MQtOorologique and to Atlantic meteorology.” The single esitinple of the weather information broad- cast by the Jac ws Cartier here given will convey an ex-
of that vessel:
phenomena and forecast their effects, while ocean. We believe that from the moment
which on this voyage exceeded all hopes.
wit fl the services which they render to the 0 A ce Nation 3
cellent idea of % t e scope of the meteorological activities
BRO.iDCAST-3fbRCH 2OTH~REENWICH NOON.
Meteo. Jqi~ti C‘nrtier to all sKips: North Atlantic weather aituaiion AfIsrch 211th. Greenwich noon.-Pressure is h over mrthwestcrn Enrope :3).3k also from Bermuda to Azores an 9 northward ale0 over American Atlantic States. Trough of low pressure extending from Iwlanrl (29.60) southesstwsrd to Gibralter (29.50). Disturbance of great intensity (23.00) over Newfoundland movi D. eaet-northeaetward. Forccasls.-Track. Azores to Gibralter-unseaed weather eastern port ion-northerly fresh winds, squally weather and overcant weather we&m portion. Stop. Tracks. Azorea to English Channel--over- cast weather and moderate esaterly winds eastern portion-freah or
strong nortlierly winds and squally weather western portion. Stop Azores region-fresh northerly winds moderating b d a and shifting to southeast and south to-morrow-xercast weather L o m i ? fair. Track, Azores to Bermuda-Moderating northerly winds an fair weather hetween longitudes 30’ and 42’. Southerly increasing winds shiftin6 to eouthwest west of 43; weather becoming overcast and prob- a1 11y rainy. Advisory for westbound ships over transatlantic tracks west of Ire- land to south of Grand Banks: Disturbance of great intemity (29.00)
over Xewfoundland moving toward Iceland will be sttended by strong ehiftiii winds to-day and to-nioht lmtween lon ‘tudes 35’ & 55O fol- lowed !y fresh or strong northeAy winds. W& will shift to-morrow morning between 40° 91 50’.
CONCERNING THE ACCURACY OF FREE-AIR PRESSURE MAPS.’
By C. LE ROY MEISINGER. Meteorologist.
[Weather Bureau, Washington, D. C., April 1, 1823.1
INTRODUCTION.
The necessity for knowing the accuracy of the charts.- The computation of baronietric pressure at free-air levels is a purely mechanical operation after one has constructed the necessary tables; but the verification of the results of such computations is a much niore diffic.ult matter, since a ri idly accurate standard of comparisori does not exist. f t is true that free-air pressures and temperatures are measured b means of kites, and wind
one reason or another, frequently not comparable with the results of computation. Experience has demonstrated that maps depicting the barometric distributions at free-air levels are often con-
velocities by pilot balloons, t ut these measures are, for
1 Presented before the Amerlarn Meteorological Scciety, at Washington. D. C.. Apr. 16.192(.
sid erably a t variance with those representing sea-level conditions. If these differences are real, it is reasonable to su pose that they are physically significant, and one niny {ope to learn by study and experience what the significance is. On the other hand, if they are false, one may be seriously inisled in attempting to interpret them. Therefore, if we are to have confidence in these charts, it is imperative that we ascertain, in every possible way,
the degree of accuracy that may be expected of them. This is the purpose of the present inquiry. Tire maps .upon which the inquiry 2s based-The maps esaniined in the preparation of this pa er were con- structed dnil during the months of Decem % er, 1922, and January a n i February, 1923, from postcard reports mailed to the Central Office of the Weather Bureau at Washington, from 29 Weather Bureau stations in the central and eastern United States. Under the authority
APRIL, 1923. MONTHLY WEATHER REVIEW. 191
of Weather Bureau Oirdar, dated November 20, 1922,
these stations were supplied with .the necessary reduction tables and instructions for reducing barometric pressure to the levels of 3,281 feet (1 kilonieter) ttnd GJ5G3 feet
(2 kilometers) above sea-1evel.l means of wliicli the reduction tebleq were constructei has been completely set forth in MONTHLY WXATHBR REVIEW- SUPPLEMENT No. 21," and will not be discussed here. The postcard reporting was discontinued at the end of February in order that time might be afforded for a complete and careful study of the results. It was believed that if these months-the season of the weather's most annoying ca rices-should yield results of vdue,
seasons.
The method b
even brighter shou P d be the prospects during less active
1
nor simple; for, YI etween the vibrant en of the meteoro-
rally t a rough the character of tho recorded data and in
COMPARISONS OF ORSERVEZ> Ah'D COMl'UTED PRESSURES.
The problem of obtaiirhg comparable datu..-In seekin to arrive at an estimate of the accuracy of the results o
computation, one of the most obvious methods is t.0 compare the computed pressure (based upon surface
observations) with that reduced from the record of the kite meteorograph. At first sight, this seems a direct and simple com arison, but, in truth, it isneither direct
graph, tracin its record among t e clouds, and the columns of co Y d figures whence the residuals are derived, is a tedious series of computations, interpolations, and extrapolations. This is no fault of the methods of
aerolo ical reduct.ion, for these complexities arise natu-
the reduction of the kite observation to the epoch of the computation. The chief dilEculty lies in knowing the magnitude of the effect pf pressure and .temperature variations between the time of computation and the
time when the kite attained the level in question. However, the values of barometric ressure corre- sponding to S a. m., 75th meridian time, gave been pre- ared with scrupulous care by members of the Aerological hivision of tlie Weather Bureau and subjected not infrequently to review and check. Frepency distributions of residuuls.-During the three months, there was made a t the sis aerological stations of the Weather Bureau, a total of 351 kite flights yielding pressure values for the two selected levels suitable for comparison with compuf.ation. After rendering these vklues as.nearlV as poss?b.le representative of 8 a. m.
(75th median time) conditions, they were compared with the computed values, and the difference was regarded as
negative in sign when the computed pressure was less than the observed, nncl osit,ive when the com uted ressure was greater. CEssif Sing these residuae by Rundredths of an inch, frequency distributions have been obtained. Fjaure 1 contains curves of the frequencies of errors of t%e several magnitudes for the two levels separately and combined, usin data from all the sta-
R
tions. I!he principal features o P these curves are:
(1) For the 3381-foot level, there are no errors in excess of 0.1 inch. The mode of the curve is at the zero class, but the curve is slightly as etrica1,show-
(2) For the 6,562-foot level, the ran e of error is some-
expected from the nature of the hypsometric equation, in which the length of the reduction column is an impor- tant term. Other conditions being equal, a ven error in
the United States east of the 100th meridian where the longer air column is nearly twice that of the shorter, an error in pressure at the upper level nearly twice the magnitude of that at the lower level. The curve is nearly synimetrical, but has its mode in the - 0.03 class. Won1 this it would appear that, during the period of the observations, the temperature argu- ments should have been,. in general, somewhat hi her.
but would simply shift the curve to the right. A large number of these negative residuds was contributed by the comparisons of data from Due West, S. C. Ths station has been established since the method for com-
puting was devised and, as a consequence, none of its data has contributed to the ori nal investigation.
(3) When the residuds for the t4wo evels are combined, the curve has its mode practically in t'he zero class but is
somewhat skew, wit,h an excess of negative values. This curve, which is simply the sum of the two mentioned above, derives its characteristics very obviously from the features of its components. lnd ividzial station histogram.s.-Since it appears that more is to be learned through dividing the data into groups than by studying them en masse, it is desirable to subdivide the curves of fi re 1 into separate station
tions beconies too small to subject to statistical investi- gation. Indeed, it has been deemed advisable to present in the station histograms only the combined residuals from b0t.h levels. These histo airs are given in figure 2.
In spite of the aucity of E t a , there is a certain-ob- eographica influence o erative in the distribu- y.ious tion o f residuals in the severa diagrams. Stations in lower latitudes show extremes larger on the negative side than do the northern stations. Drexel, Nebr., with its large number of observations, shows an almost Ellendde, N. Dak., shows t e ltwgest positive residuals; and Groesbeck, Tex., the most sout,herly of the stations, the largest ne ative errors. Owing to the unique charactenstics o f Due West, S. C., explained above, we may, for the moment, disrecmrd its histogram. WE, should the southern stations yield a lar er num-
t.hab in the ori 'na'l stud , the number of available kite
reduction method was greatest from Drexel and filen- dale; but, if the negative tendency at Groesbeck, for instance, were the result of nothing other than poor estimates of the mean tern erature of the air column, one
wide, should follow the Gaussian law. The cause seems to lie rather in the geographical and seasonal character- istics of temperature at the particular stations.
ing a larger number of negative residua s"" s than positive.
what less than twice that of the lower P evel. This is to be
the mean temperature of the air column wi f produce, in
This would not decrease the number of large resi % uals
The corn arisons for Diie West are, t rl erefore, on a sorne- what cli K erent footing, as will be es lained later.
P
groups. When thus subdivi C Y l ed, the number of observa-
? P
metrical) distribution;
ber of negative residuals than the northern? f t is true
observations wfich coul c9 be utilized for fornulatin the
should expect that the 8 istribution of errors, however
192 MONTHLY WEATHER REVIEW. b, 19228
Since negative errors in the mean teniperature of the air column are productive of negative errors in reduced pressure, it appears t.hat the southern stat.ions must possess characteristics that will produce such negative errors occasionally and more frequently than northern stations. Such errors may result from: (1) An under- estimation of the difference between the surface temper- ature and the mean temperature of the air column; or,
(2) a surface temperature that. is abnormally low owing
54
59
50
48
46
44
42
40
38
36
34 W VI
5 32
LL
0 30
a w
(II 28 I 3 Z 26
24
12
20
I8
16
14
12
IO
0
E
4
2
the ot,her winds, except west, and one with a calm. The largest residual, -0.1s inch, occurred with a sout-h- west surface wind. An investi ation of the vertical
discrepancy, reveals an inversion of estraordinary magnitude. The barometric situation a t the surface showed high pressure to the south of the station. The weather was clear and the surface winds light. At the
ten1 erature gradient as observe d at Groesbeck by kites on t K e morning of January 10, 1923, the date of this large
F I ~. l.-Frequency dlstributions of differences between (Ibserved and computed freeair pressures for the 3,281-foot leyel and 6,5Bz-f00t level, and for both levels combined. Data obtained from the SIX aerologcal stat.ions of the Weather Bureau diirmg the wmter months of 1822-23.
t o radiation conditions; local condit,ions, such ns espo- sure of thermometers ; or thermal peculiarities associated with particular wind directions. For instance, consider Groesbeck (which yielded the largest number of large negative residuals) in the light of the surface wind direction curre.nt at the t,inie of the reduction. Of the 17 cases of residuals ab t.hc G,562-foob
(2 kilometer) level of -0.1 inch or greater, G occurred with southwest winds, 5 wit.h south winds, one each with
inversion level, the wind was west-southwest and of greater speed. Surface ten1 erat.ure measured simul-
daily to Gashin ton, was 46' F., as compared with 38' If at Groesbeck. h e esposure of thermometers at Groes- beck is 11 feet above the round, while at Palestine these
temperatures indicate, therefore, that the low pressure computed for the upper level was not so much the result
taneousl at Palestine, Tes., t R e nearest station reportin
instruments are on a roo f 64 feet above ground. These
APRIL, 1933. MONTHLY WEATHER REVIEW. 193
Ro. O.-Frequency histograms
8
6
4
7
-
.20.19.18.17 .I6 15.15.1 3.12.1 1 .10.0S.08.07.06.05.0~.03.02 .Ol.CO .Ol .02.03.04.05.06.07.08.09.10.11 .12.13.1~.15.16.l7.18.19.20 \ v \- -v + -
I lor the six aerological stations of the Weather Bureau, showlng the number of eases @I diUerences of dlfferent magnitude3 computed free-air pressures; both lev& combined.
I between obserpsd and
194 MONTHLY WEATHER REVIEW. APRIL, 1983
-
1922. Deremh15 _...__...._ 30 .._._.__... 22 ....__...._ 25 _....___... 29 ..___.___..
l923. January 4 ....._._... 11 ....__..... 7 ___._..____ 8 ....._._._. 10 ...._____.. 11 .....__.... 17 ....____... 24 ....___.... 25 ....___.... February 17 _....___.__ 19 .._.._.__.. 20 .....____..
of warm air introduced at and above the inversion level as of an unusual coolin at the surface, yobably through
cooling was registered by thermometers so near the surface. In Table 1 have been classified the data similar to those given above for each of the 17 casea of large dis- crepancies at Groesbeck.
TABLE 1.-Tenperatrcre and wind conditions at Groesbmk, Tm., upou the ocensioiur of the 17 krgest iwgaiive error8 ~Jconrputaliorr.
nocturnal radiation. 5 he full eflect o this radiational
-.-
Inches. OF.
-0.15 4i -0.15 44 -0.14 50 -0.16 56
-0.15 64
-0.10 50 -0.12 51
-0.12 60 -0.17 55 -0.18 55 -0.11 61 -0.10 51 -0.12 47 -0.11 51 -0.11 37 -0.10 44 -0.12 50
--
O F .
-10
-13 -12 -14 -13
- 8
-12 -10 -12 -17 -10 -9
-10
- 9
- 9
-9 -10
Com- puted.
OF.
--
37 31 38 42 41
42 39 50
43 38 51
42
37 45 28 35 40
. __
n. e.
SU'.
1.
8.
sw. ne. sw. calln.
sw.
SW.
s.
5.
se. nr.
8.
sw.
7
3 8
12
+7 +8 +9
+1
1 Positive sign (+) indicates hlgher temperature at Palestine thm at Groesheck.
I I I I I I I J
0 IO 20 30 40 50 60 70
TEMPERATURE F)
FIO 3 -S ecimens of vertical tan eratnrecumes determined byplesns of kites at wv- erh bf t i e aerological@atlons. e) qroesberk, Tex Dwember.15 19% ressure error at upper level, -0.15 inch: snrface wmd north (6; Groesberk "ex.. Eccernber 20,
1922' ressuM error at up er level -0.i5 Inch: surface wind,ssbnthrast le) Uroea- beck $ex December 25 Lz. prAure mor at'u er level, -n.m inch: iurface wind. eas&uth&st (d) Dreh Aebr January 20 %3; pressure error at upper level -0.15 inch; &ace whd, &st-no&hm-est. (e) %Ileadale, N. Dak., January 20, 1923; pressure mor at upper level, +0.13 inch; snrface wind, northm-est.
Figure 3 (a), (b), and (c), shows a ou of vertical temperature curves as determined at roes eck on the ascent of the kite, selected from amon the dates in
temnerature inversions which were resnonsible for the Table 1. These show clearly the marke d nature of the
large pressure residuals. Southerly stations , such as Groesbeck, do not, show, on the average, a marked tendency toward inversions of temperature. When such esceptional cases do occur, it is clear that abnormally lar e errors will be produced in the computed pressure d o !i t. At northern stations, such as Ellendale, in winter, the inversion is a relatively constant condition in the morning hours, and the method eniplo ed in reducing pressure allows for this more adequate T y as a consequence. At this station, for esample, there were only three cases of errors greater than 0.1 inch, two of which were positive and one negative. The two ositive cases indicate that a greater inversion was ai'owed for than actually
0 .Ol .02 .03 -04 .05 .06 .07 .08.09 .IO .I! .It .I3 J4 .I5 ./6 ./7 .I&' .I9 .20
Difference between o2served and co.mpot ed pressures. (Hundredths of an inch)
Fro. 4.-Curves for tho two levels separately and cmnbincd. slio;-ing the prrrentage or errors of computed pressure of giveu maguitudes or a!aaller.
occurred. One case of a marked in- version a t Dresel is shown in figure 3 (d). Drexel's large number of obscrvstions and syninietrical distribu- tion of residuals, as well as its geographical situation, serve to rocluce the most satisfactory of d the histo-
corn arisons to justify reliable conclusions. Tfe outstandmg feature of the Due West histogram is
its tendency for negative residuals throughout. (Fig.
2.) It yielded the largest, ne tive departure of m y sta-
The curve is decidedly skew, and is also markedly dis- ,placed from symmetry with the zero cJnss. The skew- ness may be attributed to the same causes as at Groesbeck and other stations, but the rnoclnl displacement to theleft is probabl the result of faulty geographical interpolation
the eastern part of the United States were derived in the original papers from the stations at Mount Weather, Va.
(near Washington, D. C.), and Leesburg, Ga. The record at the Virginia station was probably ade uate for the
!ur values were based wa.s quite short. Due West, S. 6, occupies a position approximately halfway between
(See fig. 3 (e).)
grams. 1; roken Arrow and Royal Center afford too few
tion, and the smallest nun1 f? er of posit.ive departures.
in one of t TI e steps in the original research. The data for
urpose, but the length of the period upon w 1 ich the Lees-
APRIL, 1923. MONTHLY WEATHER REVIEW. 195
the two mentioned above. If? on the average, the tem- perature values had been enou rh higher to shift the histo- gram several cla.sses to t,he right, this st.at.ion would have stood out as one of tlie best. One hestitates? however, to reconiiiiend that the values for this stiltion be increased upon the basis of these t,hrer. nionths’ esperience. It is wiser to withhold judgment u on the best solution of the problem until other evidence {as been presented. Thus far considerable attention has been devoted to the large residuals. But sight should not be lost of the large number of very satisfactory and arcurate calculations. Figure 4 shows for the two levels separately and, in a third curve, combined, the percentage of residuals of
any given value or smaller. The curves take no account of the sign of the residual. For esam le, reference to
cent of all the comput.at,ions were within 0.05 inch of the observed value; for the higher level the percentage is about 54, and for the two levels combined it is about 73.
When it is recalled that the isobars on the daily weather map are drawn for intervals of 0.1 inch and that, in tele-
graphin pressures, only the even hundredths a.re coded,
The first two curves are based on 251 values each, wlu maf e the last is upon the combined number, or 502.
the curve for the 3,3Sl-foot level shows t r iat about S9 per
errors o r? 0.05 inch will esercise but 1itt.le effect upon the
.
COMPARISONS OF OBSERVED AND ESTIMATED WINDS.
The djfictrrlties o obtmiiihg relidlo standards of com-
that some clue as to the re1id~ilit.y of the free-air maps may be adduced from the coni arison of estimated wind
winds :tctuillly observed by means of pilot balloons. Here, as in the case of kite dtita, the ccjnipnrison is not direct, mcl, niorwver, there are severd sources of dis- agreement that tire quite im iissible of elimination and yet which may apparently, gut not really, throw dis- credit upon the usefulness of the ninps for purposes of estimating the wincl direckion. In the first )lace, the theory of tho gradient wind states
by friction or visc.osity, is perpendicnlar to the gradient a t the level in question. A synoptic chart gives an instantaancous pic.ture of thc pressure distribution but gives no clue as to t.lie true horizontal trajectory of an air particle unless the whole barometric formation is stationary-a condition which seldom occurs in nature. Neverthekss, the agreement between the direction of the observed wind at a given level and the trend of the isobars at the same level is usually sufficiently close to permit of a satisfmtory det.ermination of the general drift of the air. Thus, in these compmisons, the estimate of wind direction was based entirely on the rtbove-mentioned assumptions, and it is to be supposed, therefore, that,. in the case of rapidly moving formations or those in which the intensity was c.hanging rapidly, many discrepancie.s between observed and estininted winds must occur. In addition to the complexities introduced into the corn arison by the gradient wind assumptions, there is
balloon observations themselves are representative of the air drift at a given level. It should be remembered that ilot-balloon flights give but momentary indications
Por each level, since the balloon is constantly ascending. Moreover, under variable or rapidly changing conditions, two ascents separated by n short time interval ma
pa.rison -from pibot- d alloou da.tn.--It hits been pointed out
directions, b:ised upon the grnc 7 ient mind relations, with
that the win d direction, escept so far as it is influenced
the f urther question as to the clegree to which the pilot-
winds varying in direction by several points at t z e reveal same
level. In general, this is true especially in the case of light winds, and on a warm summer day all manner of vagaries, real and apparent, may be introduced into the honzontal trajectory of the balloon by convective forces. Again, difficulties may be encountered in attem ting. to read from the altitude-direction curve the irection corresponding to certain arbitrary levels, such as those chosen for reduction purposes. The winds may have been turning rapidly just a t the level in question, so that
a small difference in elevation would show a quite dif- ferent wind direction. Figure 5 shows a series of altitude- speed and altitude-direction curves in which this is clearly brou ht out. I t a as been necessary to make estimates for station lying on the western frontier of the computations, an hese estimates were probably in greater error tha
7000
6QOO
h
+ W
E 5000
W U
c
a
4000
z v1
W >
4
W P 3
3000
t
a 5 2000
1000
0
FIG. .i.-Altitiide-Telrlocitv and altitude-direction curves for observations made bjzilot balloons at Broken -4riow. Om.. sho\vlng example8 of rapid turning 01 win irec- tinn at certain L P T ~S , and the difficulty of determinmg at such clrvations the true drift of thc wind.
stat.ioiis lying within the network, since the proper west- ward estemion of the isobars was uncertain. These factors all tend toward confusion in the deter- mination of an absolute standard of accuracy in wind direction. How th,e e.Qtim.ntes were m.ads.-The method used in arriving a t the errors of estimation of free-air wind direc- tion was as follows:
(1) Having completed maps for tlie two levels for a given date, a form, carrying as its first column L list of pilob-balloon st,ations of t-he Weather Bureau, the Army, and the Navy, was entered with the estimated direction of t,he wind ?or each station and level. This, it will be noted, was done in every case before actual observa- ticiiial niaterinl l i ~l yet been received, and was, there? fore, elitmirely free from bias or influence which might un- conciouslg arise if observed direction arrows had been plotted before the estimate was made. (2) ZTpon completing this table for the period of a
month, it was handed to the balloon section of the Aero- logical Division, where there were entered, for all stations at which observations were made at the required time of day, the direction recorded u on the observational form.
on the 16-point scale. The observed wind speed was also entered.
All directions, both estimate B and observed, were given
196 MONTHLY WEATHER REVIEW. APRIL, 1923
(3) The classification of the errors of estimation was then made in the following manner: Facing the direction of the observed wind, the number of points on the 16-
point scale, either to right or left, between this direction and that estimated was determined. For esample, suppose a west wind was observed and B northwest wind was estimated: Facing the west, the northwest wind would be two points to the right in error. The observed velocity was also recorded in a class correspond- ing to a 2-point error to the right. Thus was formed a
frequency table of 16 classes corres onding to the number
mated wind. The velocities also were tabulated in this wa so as to make class averages possible. $hat constitutes a correct estmate?-One is next faced with the problem of determinin what shall constitute satisfactory agreement between %e two. Owino to the difficulty of determining in many cases what t,he &served wind direction really was, it has been unanimously agreed
of points of difference between t R e observed and esti-
considering errors in estimatiii direction, one must also
curves at the top of fiwure G there is shown for the two levels the averawe veKcity of the wind, in miles per hour, corresponhg to the several classev of errors. Owing to the very small number of large errors, it is ap- parent at once that the mean velocities at the extremes of the dirtgram would hardly be comparable with those at the center. Hence, the veloc.itg curves were discon- tinued at class 4, although the points were plotted. It i s clearly s f m f h t grea.tent accwmy i s associated with highcst velocity of the: wind In these cases the wind velocitj- wa5 about 30 miles er hour for the correct
of discontinuance of the curves. The two curves passin through the center of the dia-
observations upon which the means of the velocity are based. Since this quantity gives a measure of the rela-
have regard to the correspon a ing velocity. In the two
class, falling off to about 15 mi P es per hour a t the pointa
gram give the values of t a e square root of the number of
ERROR IN DIRECTION 16 POINT SCALE
Fro. 6 -Up (4 seetion: Mean wind velocity corresponding to the several clsrres of error in estimating freeair wind dirertion. Middle swtion: The square rnot of thp number of odserva~ons u on which the upper curves are based. Lower section: Curves 01 pcrrentage frequency of occurrence of errors of diffrrent magnitudes in estimating free- sir wind directfons.
among those with whom this matter was discussed to dqim (1s correct an esi?i.m.a,te tha.t fell o m p0i.n.t eith.cr ai&
qf the obsenrcd urind on. the 16-point ~c d e . Successively, to either side of this class, the errors were desi nated as 1-point, 3-point, etc., on the same scale. I% this way, the two classes of 7-point error coincide and re- present the maximum possible error of estimation. Error frequencri.es i n u6nd direction.-In the lower part of fi ure 6 are shown two curves, one corresponding to the %iP and one to the lower level. These curves show t e percentage of errors falling within different classes, as defined above. The relatively high maxima of these curves as well as the sharp decrease of ordinates on either side of the correct class show at once the high de ee of accuracv of estimation.
!&TOIS in wind direction are not serious unless the velocity of the wind is high. Wind directions in nearly calm conditions have no significance. Therefore, in
tive value of the means, it is seen that the more accurate estimates are associated with more reliable mean veloci- t.ies. Indeed, the ordinnt.es of the curves at the extremes are less than 1/35 those in the center. It is reasonable to sup ose that if enou h large errors of estimation were
would fall very low near the ends of the diagram. It has been noted in this figure that the curves for the two levels are slightly displaced, the upper curve lying to t.he right of the lower. The reason for this will not, be investigat,ed, owing to the minor practical im- portance of the It, is suspec.ted that it is asso- ciated with the e e ,"b f the eastward motion of transla- tion of pressure systems introducing a component into t.he motion of air within the system.
of estimation ab t,he upper level is a k most, the
availn t! le to give reliab k e means of velocity, thosc. moans
It, is more significant, that the curve representin
identica accuTacT with t.hat for the lower. This shows that, in
APRIL, 1983. MONTHLY WEATHER REVIEW. 197
spite oE the lower degree of accuracy indicated in the pressure computations for the upper level, compared with those for the lower, the accuracy of est,iniated wind direct,ion at the upper level is of quite as high degree as that, at the lower. This is a highly import.ant feature of this diagram. Figure 7 shows in the heavily-drawn upper curve t.he ercentage of cases in which the error was n given num- ger of points or less, the nuniber of oints being clesig-
figure 4, no distinct,ion is made between errors to right or left. Nor are the levels separated, for it was found t,hat t.he percentage of accuracy for the two levels was so
nearly identical in each class h a t the curve representing the niean of the two is wholly adequate. I t is seen t,lint 63 per cent of the estimates were correct nnd that 91 per
cent were within two points of error. The t,wo CZI~VCA
iiated on the axis of abscissae. In t P iis diagram, as in
700 2
0 I-
600 % a
m 500 0
0
400 E
m I 3 300 z
w In
LL
200
IO0
ERROR IN DIRECTION (16-POINT SCALE)
FIG. ?.-Upper curve show the peneutage of errors ofestimstedfrec-air wind dircction tudes or smaller. Lower curves give the number of observations for of given the two I x u p o n which the upper curve is based.
in the lower part of thc figure show the number of obser-
vations in the several classes as indicat.ed on t.he scnlr. of ordinates at the right. For tlie lower level at tot.al of 984 comparisons was made; for t,he upper level the number was 637.
THE EFFECT OF PRESSURE ERRORS ON THE MAPS.
The txposure (of fh~r~noi)tt*ttrs a.nd its efeet wpon the
. free-air maps.-One of the important points made in
previous papers on this subject concerned the inconse- uential effect upon the horizontal pressure gradient in %e free air of temperature errors, providecl errors of similar magnitude were made at nd'acent st,at.ions. This argument is intrinsically sound. IN ecert,heless, i b
has been found that the inffuence most ac.t,ive in pre- venting similar errors at adjacent, stat.ions is not, the factor applied to tlie surface temperature to obtain the
mean temperature of thc air column, but the surface t,eniperat ure itself, owing t,o tlie non-uniform ex osure of thrrmonieteis. Thus, on mornings following stilf clear night.s, or at, thp_ times of a cold wave, the t>hernionieters at. ciby stnt.ions roundabout, aerological st,at,ions may
regist.er several tlegrces higher than the latter. Under such conditions, t,he resid tiiig reduhons are niore nearly
correct. at, cit.y st,at,ions than at kite stations, and irregu- Inrit,ies in tlie isobars occur in the ricinit.y of the aero- logical stations. An interesting demonstration of this difference in current teniperatures between near-by stations when the thermometers are differently esposed is to be found be- tween Drexel nncl Oninha, Nebr. ; and between Groesbeck and Palestine, Tes. 1933, only have been esamined, but they show cleary i? what a more eshaustive investigation probably would disclose to a nicety, namely, that there is an approsi- niately linear inverse relation between the velocity of tabe wind at the aerological station and the current taemperature t lifference between the two stations. In other words, high wind velocity is associated with small .
cliff erence of temperature (because of turbulence) and conversely. No variat.ion with wind direction was so apparent. Therefore, when the kite station values are in harmony with those of surrounding stations, it is found that there is higher wind velocity and less differ- ence between the current .surface temperatures. Thus, at times of litrgest errors, t,he city exposures are better for t,he purpose thnn are those a t aerolo4cd stations. At the. outse,t, it niny seem paradosicti? that the aero- logical st.:ttions from which the original data for the investigation were derived ahoulcl be founcl less suitable than cit,y st:ttions. But the reason is ap arent, namely,
occttsions are in bett.er agreement with average condi- t.ions at aerological stations than are the temperatures clt aerological stat,ions themselves. Moreover, if all the stations where these free-air reductions are made were nerolouical st,at.ions, the map would probably not be so seriouag affected. But since they are not, it is neces- sary for the minoiity to yield. How may the situation he twnedied ?-The following schemes sug est themselves in consideration of means t.o prevent t fl e occurrence of large residuals when com- parisons are made with observed pressures, and also to rocluce more accurate maps:
K 1 TO correct the aeroloaical station temperature tables by amounts which wouTd bring them into accord with surrounding stations; or,
(2) To eliminate the aerologicd stations from the net- work, using only regular Weather Bureau statioim There is hardly justification for the first plan in such t,ests as this, based upon a single season of comparisons. Nest. wint,er may be singularly free from such inversions
a,s were productive of lar e negative errors during the
the temperature cliff erences between city and country esposures were. small except when large pressure discrep- ancies occurred at the free-itir levels. Therefore, it seenis that t.lie second suggestion, i. e. , the elimination of aero- logical stations froin the network, would yield the more
SR tisf a c tory m nps . T ~G efeet qf occu8ion.d lu;rge errors on the map.-Grant- ing occasional large coniput,ational errors, what will be the effect on the mnp! This question is not raised to condone the errors, for t>lie effort and desire are for the elimination of error so far as possible, but to esaniine qualitatively the results of the comparisons.
Records for the month of Januar
that the city station teiiiperatures on t !I ese particular
present. sea.son. The signi fi cant point is that, in general,
198 MONTHLY WEATHER REVIEW. APRIL, 1923
A scrutiny of the various dates when comparisons of In other words, the data submitted on post cards from pressure were made indicates that December 33, 1932, city stations have been used in each of the maps; but the was the occasion. when several of the largest errors of aerological stations have been varied, using first the com- which we have knowledge occurred simultaneously at the puted pressure; second, no values a t all; and, third, the
aerolo icd stations. A map of this date must be a fair observed values. e x d p e of just how seriously the isobaric distribution A there is a questionable-looking southward
is affected. Figure 8 shows for sea-level, and for the loop of t e isobar about Royal Center, Ind. In map B 7 R
In ma
DECEMBER 22, 1922.
8a.m.. 75th Meridian Time.
I -_..-
b c _,_.C-.--
i
Fro. &-Sea-level and 6,5B%foot-levd maps for December 2?,1922,8 a.m., 75th meridian time. For differcncrs in drawing maps A, n, and C. see discilssion in Wxt. .
6,562-foot level (maps A, B, and C) the barometric charts for the date in question. Map A is drawn from the corn- ost card. Map B is
Eased upon the regdar Weather &u.reau sta.tions only,
from using the ressures meammd by kites, together with those compute1 at regular Weather Bureau stations.
the loop is eliminated, and practically no change results elsewhere. In map C the pressure a t Broken Arrow, Okla., and Groesbeck, 'rex., are increased so as to pro- duce a northward rojection of high pressure from the
In general, these maps do not disagree as to t e major In limited regions, such as central Texas, it is
lete list of stations as reported by
R
aerological stations being eliminated. Map C results Gulf region. The E oyal Center loop is not ap arent.
features.
APRIL, 1928. MONTHLY WEATHER REVIEW. 199
clear that under such conditions the wind estimate from either map A or B would be somewhat in error! assuming the gradient-wind conditions obtained. Nevertheless, when we consider that the difference between the sea- level map and those for the upper level is so striking, and that the differences between individuals A, B, and C me so small, there is little cloubt that the important and sig- nificant features of the true isoba.ric distribution at the u per level are de ictecl. The anomalous loops, such as
a single station, are easily reco nize a,s erroneous and
, and i t is contended that if the type of map exem- pl' sax ed hy map B were used-that is, if the aerological stations were not used but near-by replar stations sub- stituted, so far as possible-the resulting maps would be highly accurate and dependable so f a r as free-air baro- metric gradients are concerned. This procedure is sug- gested in further activities of this character.
B
t R at about Royal e enter, Ind., on ma A, when based on
allowance can be made. But t B lis should not be neces-
CONCLUSION.
An effort has been made in tahis paper to set forth clearly and impartially the resu1t.s of n st&st,ical e s m - ination of a t,hree months' series of free-air pressure ma.ps.
It was first shown that a very large percentage of coiii- parisons between observed and coniputed pressures lay within small error limits, better resu1t.s being obt.ainec1 at the 3,2Sl-foot t,han at the 6,563-foot level. Wit.h all the uncertainties of the gradient-wind assumpt,ions and the short-period vagaries of t.he observed mind, an extreme1 high percentage of amreenlent (practically the same at 55 ot,h levels) was fount? between the wind e&- mated from t,he isobaric chnrt4s of the free iiir, and that observed by ilot balloons. Tlie average relocit,y of the wind falls o rp appreciably with the larger errors of esti- mated wind direction, so that, usually, when a. serious error of estimat.ion is macle, the velocity of the wind is so small as t.o render the wind direct,ion inconsequentid. This is es ecially significant. in est,imat,ing winds for
aircraft. 8 n the poorest day nvnilable the difference .between maps drawn with and wit,hout, t,he aerological st,at,ions and also wit.h the observed pressures a t aerolog- ical stations was so small t.hat lit,t,le-c!iffering conclu-
sions might be drawn from each. It was st,ateil in t-he beginning that, if the maps could be proved t.o be accurate t,heir scientific value would
stand unchallen ed. If this discussion of the accuracy
is here provided a means of knowing, quite independently of current aerological observat;ion, t-he current wind conditions in the free air over large areas of the country, whether the weather be clear or cloudy. At present, the pilot-balloon stations telegraph the free-air winds as soon as possible for the use uf the forecaster. Often such observations are impossible because of cloudiness, rain, snow, or fog. When used in conjunction with the aero- lo 'cal reports, these charts should constitute a most
continuous pressure system, a speculative and uncertain matter a t present. The advant,age of such maps for aviation is obvious. I t would be quite impossible, even if it were wit-hin t,he scope of this paper to discuss the point, to indicate t>he precise manner in which these charts should be utilized in general forecasting. That must be deter- mined by actual trial; and by actual trial is meant day- to-day telegraphing of free-air pressures in order that the may be available to the forecaster for comparison wit K other data while the current weather situation is fresh in mind. It is common knowledge that the r0wt.h of aviation implies increasing deniands u on the $eather
be employed to ac.quire esperience ancl familiarity with actual physical processes in the free air, while the science of aeronautical meteorology is yet in its infancy? The dt~y will come when this knowledge ,udZ be required, and since it can only be BC uired by es erience it is the
This trial should continue for at least a year, and, during that time, all possible constructive crit.icism should be h o u ht to hear upon the maps.
S ~~~r o ~o l e ~~r n e ,I t s .-T l ~e large amount. of observat,ional data used for com arison purposes in t,his paper could not
of Mr. L. T. Samuels, Mr. W. C. Haines, nnd t,heir asso- ciates in t,he Aerological Division of t.he Weather Bu- renu. They manifested considerable personal int.erest in the results and contributed many helpful supes- t,ions: t.hey were particularly rompt,, also, in supp gin
errors. For this cooperation t.he aut,lior records his earnest thanks.
of t,he maps has Q een convincing, it will be seen that there
vauable T means of weaving the various reports into a
Bureau. Is it not essential, +erefore, t 3: at every means
esercise of foresight to ilia qh -e a practica H trid of the maps.
have been assemb s ed without. the enthusiast.ic assistance
'i data and in ferreting out the P act.s concerning suspeckec
COX ON THERMAL BELTS AND FRUIT GROWING IN NORTH CAROLINA.'
By ALFRED J. HENRY.
[Weather Bureau, Washington, D. C., May 7.19.23.1
NoTE.-According to custom an abstract and review of MONTHLY WEATHER REVIEW SUPPLEMENT No. 19 is presented below. Theae SUPPLEMENTS contain the results of the more comprehensiye st,udies made b Weat,her Rureau officials or others. They appear at irregular intervai and are generally too voluminous to appear in the REVIEW proper. As a result of economies in print.ing that have recently become effective the edition of the PUPPLEMENT~ is not large enough to supply the regular REVIEW readers. The SUPPLEMENT Kill he sent free, however, to those who may have a practical interest in the subject, investigated, but on1 upon application, and 80 lon aa the bureaus' supply laeta. b f k r &at, is exhausted applicants wifl be referred to the Superintendent of Documents, Washington, D. C. The price of
t.his supplement ie 50 cents.-EDITOR.
The phenomenon of the stratification of the air tem- perature over valleys and the inclosin slopes was brought to the attention of sr.ientm5c men o f the United States
~
1 Cox, H. J.: Thermal belts and fruit prowing in North Carolina: with an appendix: Thermal belts from the horticultural vicn'point, by W. W. Hutt, former State horti- culturist, YONTHLYWEATHEB REVIEW SUPPLEMENT No. 19.
by Silas McDowell, of Franklin, Macon County, N. C., more than 60 years ago. Mc.Dowell was a farmer with leanings toward botany and geology, who spent! his entire life in the mountain region of western North Carolina. His first published account of tliermal belts or verdant zones, as he called them, appeared in the report of the Commissioner of Patents for 1S61, and the sub- stance of that report was later presented to the Philo- sophical Society of Washington (D. C.), by Dr. J. J. Chickering. The late Prof. John LeConte, of Berkeley, Calif., writing in confirmed McDowell's obser- vations and added the statement that the ground in the thermal belts freezes in winter, a fact that might not be inferred from McDowell's description. The esplana.tion of the henoniena oil'erecl by McDowell
-- would hardly he accepte B at this time, but nevertheless --
2 1853, vol. 1, p. 278.