JUNE, 1915. MONTHLY WEATHER REVIEW. 245. is a portion (AZB in fig. 5) of the sky image upon the with the new camera would furnish detailed information negative which duplicates the first part of the picture. as to practically eve kind and phase of cloud form In order to obtain a c0rrec.t picture of the sky tlie dupli- within the area of a wexdeveloped cyclone or anticyclone. cating se-ment of the negitt,ivefihii is cut away and the The clesigu of the new instrument seeins to be entirely edges are brought together. (In fig. 5 the segment BZC' satisfactory, and the cloud photo a h here reproduced was cut away ancl ZB matched against ZC'.) The fi~id iii figures 5 and 6 wtts made by hfgueller by means of photograph, as shown in figire 6, is then niade by ineniis the temporary wooden cmiiera, constructed entirely by of transiuitt,ed liglit tlirough tlie conicid negative, wliicli liiinself and s iown in figures 1-4. It is very desirnble; represents the cloiiie of the sk-y. The st:iisit.ized circ.ul:w however, to provide for the construction of ti inore per- film may be replaced bg it conics1 fh,wit.11 8 slight iiiodi- iiiiiiient instrument, in order to facilitate further experi- fic.ut.ioiiin t,lia construction of tlie c'aiiier~. By tlic use of iiientation in the technique of cloud photography, and the coiiicd filii1 the size of the camera iuay be reduced, to cleiiionstrate the value of the camera as an instrument while the operiLtion of cutting :~wuyt.lit: w-erla ping seg- for systeinatic research, ancl for daily use at meteoro- ment of tiic? circular piiotograpii is eliniiuatocfl logical observatories.
-irlwniqis 91*tlww:~.~I curtmu,.
What advantages are offered by t.lie use of the new A TEST FOR PERSONAL ERROR IN METEOROLOGICAL forin of caiiiera, over the forms iiow in uso, especially iii OBSERVATIONS. the field of cloud photographyY In thc first place the new device caii be utilized for all purposes iiow served By ERICIt. MILLER,Local Forecaster. b existing forins, ancl in addition yields results not to be [Dated: \Venther Bureau. U'sdison, Wis., Mar. 23, ISIS.] otained by nieaiis of the latter. Our ignorance conceiiiing cloud forms and their sigiiii- The object of thiu is to call attention to the value came is still rather inore coiispicuous than our knowledge. of the daily rainfa. distribution as internal It is reasonable to assuiiie that n mote accurate survcy evidence as to the climatic data. Internal of cloud forins over it11 estendecl area and the ritpitl evidence is often of the cliniatologist changes w!iich they undergo will lend 11s to n better in weighing raw are necessarily under&mhg of the nature aiid niovenients of storms isolntecl, and hence are not, sub ect to personal inspection and weather climgc?~. A cloud is a visihle process rnthor and supervision. The degree oi neatiiess of the observer's than L finished lxodwt,, and is in iiimy iiistiiiices the only re orts affords only indirect evideiice of the fidelity with evidence we have of the forces nt work in the upper at- w fich the observations have been niade, especially as the ' mosphere. A photogra h which embraces the entire majoritmyof volunteer .observers-the rincipd source of field of view 0 )c1ied Ui) Eeftjre the observer, and clepicts cliniatic ilattk-we likely to be above tP ie average in gen- all of the clout'I forms --their relative positions ant1 their eral education, although they inay have liad no training extent-is iiiauifestly of urenter vnluc than a photograph whatever in scientific accurncy. Even this indirect evi- - which shows only :i siiis71 detached doncl, or portion of dence is not nvailnble to nimy who have only the neatly a cloud, and wliicli gives no uvidcnce of relationship with printed tables of data presented to t,lieir inspectZion. the eneral condition of the sky. Rilinfnll-frequency distributioii is deteriiiined by classi- ~i?epro~eius of cloutliiiess iii:i,y ])e profitu1)ly stuclied fying the recorcled dnily rainfalls according Go niagnitude, one cliiss for each uiiit of the scale used in niensurin raiii- ' by mealis of ti cletaileci study of individual cloud forins, 'ust as it is a coiiiiiioii and profitdde practice to study fall. In tlie United States rninfalls are recorcedf to lorestry by the .;tudy: of inclidud trees; there are, how- hundredths of inches, hence such. a table shows the num- ever, problems of clouds in association with other clouds, ber of daily rainfalls of "Trace," 0.01 inch, 0.02 inch, 0.03 whicli can not be successfully solved in this iiianner, just inch, niid so on, in ascending series. aa there itre )robleiiis of trees in nssociation with other The frequency distribution of daily rainfalls is of the trees, or tlie targer prohleins of forestry, which the study ' ' es tremely asyninietricnl or J -shaped f oriii " of tbp of individual trees will never su-grst. The new foriii of stat,isticiaii, the sinallest amount being most frequent.. camera will make suc~igeneral c€ou(~studies possitJle txnt~ In order to show that the J-shaped fre ueiicy distribution profitable. of ddvrainfadls ,revails throughout t7.l e widely differing, One of tho siiiipleut and nost obvious uses of the iicw rainfall regimes oI the United States, I present ui Table 1, camera is to obtain n permanent and true record of tlia and figures 1 to 5, inclusive, and figure .15, frequency dis- amount and chmcter of the cloud cover at stated int,er- tribution tables aiid curves for 15 United States Weather' v& during the da -. The on1 iiiethocl now in use to Bureau stations in the nricl Southwest, the north Pacific obtain this oncl is )y means oB eye obscrv:itions, which coast, the upper Mississippi and Ohio Valleys, the Gulf are liable to. errors of judgment, and at best arc only Stnt,es, the Lake Region and New England. These data rough estimates of the proportions of tho sliy covered by tire for the sis months, April to September, inclusive, the rincipd cloud forins. in tlie year 1914. Lon er series would give smoother as . dotograplis of t,lie sky at stated intervals, txlmi curves, is indicated %y taking tlie average shown &nultaneously :it R ncniber of stations within cycloiiic gmphicdly in figure 15. The six wanner months were adanticyclonic areas, ~vouldnff ord v:iluiible iiiaterinl for chosen in order to secure, as far as possible, im ersonal &e advilllceriieiit of our kno~vled~eof st,ornis. Qssuniiiim data, since the automatic raingnges at regular keather &at one of these ,Iiot,ograplis w31 sliow the amount an3 Bureau stations are then in operation, aiid while the re- Aaracter of the c1 oud corer over the area of from 200 to corded nieasurements are personal stick nieasurements, 1-30square miles, a comparatively few stations equipped they are likely to adhere uite closely to the automatic .._- .- records, especially for sinal1 amounts. 'Wr.MuellerIs ~OWwmtructing such a modified modelwith a conical amin accord -% thauggestions-0. L. Fassig. 'Yule, G. Udw. Theory of statjstics. London,1912. p. 98.
Unauthenticated | Downloaded 09/30/21 10:58 AM UTC 276 MONTHLY WEATHER REVIEW. JUNE, 1915 ’ TABLE1.-Rainfallfreqwnq dtstrikution-Nurnbe of times various daily rainfalls (April to 8eptrmber, 1914, idwive)hnw been reumikd by cnrtomatic raangages .
Amounts. Shh. - - - I. 0: .a .O - - - Wlnnamucca, Nev...... 15 3 3 2 29 16 4 2 _. 12 ...... :...... 2s 2 3 1 11 ...... 13 11 1 ._ 58 ...... 21 6 7 4 65 23 5 6 1 64 MdLSon Wfs ...... 23 10 6 3 60 mwadm, wis ...... 34 6 6 2 67 Gmm Bay, Wb...... 15 10 5 3 68 Vkksburg, Mlss ...... 22 8 4 2 63 New Orleans, La ...... 24 7 3 4 I2 .Mobile A& ...... 16 5 4 4 84 colnm.bus Ohio ...... 22 0 3 2 61 Bostm dsss ...... n 1 6 3 51 Ndd,Vt ...... 23 5 6 377 78 Total ...... I 322 89 D 797 Average ...... 21.5 ’.9 ,1 .5 ...... I - - -
TABLEZ.-Abnormalfieque~tcy dktrihition prodwed by observational error-Number of times various daily rainfalla (April to Sephnber, 1914, inch save) have been recorded by wopmakve obscrvcrs.
Amounts. Rainp Station. days. *
Arthur,Nev ...... 10 ...... 11 1 25+1 GilaBend,Ari ...... 1 ...... e ...... 19 ...... 44 27 5 52 ...... 2 ...... 1 ...... 1 ...... 2 ...... 1 ...... 1 1 1 12.... 13 11.... 11.... 1 ...... 7.3 8 51 ...... 11...... 25 8 68 ...... 2 .... 4 .... 1 ...... 19960 8 1 ...... 1 ...... 1 13 11 31 ...... 8 ...... 1 ...... 17 4 33 ...... 1 ...... 22360 TOW...... Average ......
* D3y3 having n fall of 0.01 inch m more. The advantages of the frequency-distribution table, Three principal types of personal error are apparent or.curve, of ramfall as a test for personal error are as from these esamples, vis: f QUOWS : (1) Those due to the failure to record light rainfalls, (a) The smallest amounts being most frequent make see figures 6, 7, 8, 9, and 10. it very sensitive to observational error of an systematic (2) Those showing that rainfalls have been noted and kind, since the smaller amounts are less lik to be con- not measured, but estimated as “Trace,” see figures 11 sidered important by the careless or neglec$ ul observer. and 12. The observer at Wapesboro has recorded a (6) The data are easily available, and easily reduced eater number of “Traces” than the regular station at for the purpose, the operation consisting merely of count- Few Orleans. ing. (3) Those showin that the rainfa have not been (e) It sffords a means of comparing the work of dif- accuratdy measure$ but recorded to the nearest tenth ferent observers in different parts of the country, since or twentieth of an inch. Figures 13 and 14 (Arthur and the recording of rainfall is a universal duty of meteoro- Hoepenick) show this practice to be consistently fol- logical observers, and because the J-shaped curve pre- lowed in the observations represented, while figures 10 v& in widely different districts. The number of small and 12 (Cornwall and Wapesboro), show it to be fol- rainfalls differs much less from one district to another, lowed in part only. This peculiarity is well known in than does the total rainfall. connection with the estimatwna of the tenths of degrees of markedly abnormal fre uency distribu- in observing thennometers graduated to full or half rainfalls are given in Tabi! e 2, and fi res degrees, but the measuring stick used in rainfall meas- The ma orit of these are from%- urements is full divided, so that this form of ersond consin bocause the had &ea$ come to my attention error can not be regarded as subconscious. TI% tend- in studies of local Jl ‘mate and not because they are more ency to estimate. to even arnounh appears very stro common in that State than elsewhere. These data are in measuremenb of snowfall (wherein the factor 7;; for the same period as the data presented in Table 1 a Walter, A. On errors of estimationin thermometric olmrmions. Q,ly. jour., Roy. andiigurea 1 to 5, and me comparable therewith. meteorol. ma, 100s. SLS: 149-261.
Unauthenticated | Downloaded 09/30/21 10:58 AM UTC JUNE, 1915. MONTHLY WEATHER REVIEW. 277 is used to reduce to water equivalent) as is seen in the A comparison of the total or average frequen of following record of rainfall frequency for Madison, Wis., different amount9 at the 15 stations in ~tt.61 for the month of January for the 20 yeam 1871-1890. will show a slight tendency to avor the fifth and tanth (The tendency has since disappeared at this station.) divisions, at the expense of the adjacent divisions. Among untrained observers the tendency becomes in TABLE3.--Frcqueneies of variour rain all3 al Madcm, Wk,during 1871-1890, ine usivc. some cases a ruling passion, as figures 13 and 14 and i the average for the 15 cooperative stations in figure 15 F~ clearly show. The followmg stations afford examples quency. of markedlv abnormal meference for the fifth and tenth divisions, correspondhi to rainfalls of 0.05(2n+ 1) and DUW. 11 0.1On inches, where n is some whole number, throughout 2 the whole range of recorded daily rainfalls during the 4 4 eriod considered, April to September, inclusive, 1914. .04 f 3 .1J ”1 .30 ti the .05 9 .15 0.3lto0.89 6 8ompared with the average frequency of at .06 9 .l6 ‘4 .40 ’I 15 regular Weather Bureau stations, 29 per cent, the .07 5 .1i 2 a41 to0.49 2 -08 5 .18 :o .50 6 ercentage of “Trace” is evidently in excess at some, and .OB 1 .l9 ;3 Seficient at others of these cooperative stations. - RA/NfALL /N H‘NDIQ€DlHS nF AN INCH- T -05 -10 -16 -20 T .Ob -10 -15 .20 T -05 -10 -15 -20 T -06 -10 -15 .20 ?’ .OB -10 a16 30 86 80
16 io
$6 3 -0 10 k Oa
$0 a 95 %a 16
10
6
0
FIG.1-14. Rainfall-frequencycurves for 14 btatioas, dprll to September, 1014. Abaebsa! show amount of the daily fall: ordinates Indicate number of times each ddy fall w recorded durlng the interval. [The drafhrnm transpsed ~omdof theqe flguray imaglne 13 under 6,11 and 14 under 8, tlaaUy 13 under 9.1 TABLE4.-Emnipk* of abnornurl freqmncie* of the “5’~’’aid “1O’a.” --______Consideration of the probability of occurrence show i how unnatural these records are. Hrlativo Irequenry u1 the wale d1vhiun.i: 1 - - The influence upon the total rainfall for a month or a S131inns. I-- I year, or u on the berage rainfall, of them personal errors 1 i ,, 1I1 25 3 ;, I,, TmL.. is probabf y not of any greater importance than the other 1 i ; &id9 errors affecting the measurement of precipitation. Their -a- -a- influence may, however, be felt stron ly in other data Per m”tb enick, \!‘Is...... n 100 relative to precipitation, perha s most oK all in the aver e m%i*au wls...... 91 0 number of “Rmy days.” effect is shown in Tab e MhW NbF ...... 69 0 TL 7 can& Vt ...... 3u 32 21 5, in which the regular Weather Bureau stations are set Qrfknts& Wis...... 30 30 4 N~IU0{Via ...... 44 35 in black-face type. W.yaesbab0 Miss ...... 28 21 30 Ymlklart, oh0 ...... 15 24 20
Unauthenticated | Downloaded 09/30/21 10:58 AM UTC MONTHLY WEATHER REVLEW. JCJWE,1915 server's habib of work; its use can not well be omitted by working climatologists or the directors of climatolo 'cal services without dan er of loss of valuable time and e%- ort in dealing with wort%l ess or vitiated data.
REMARKS. Prof. John F. Ha ord, director of the College of Engi- neering, Evanston, SIll., presenta by request the following comments on the above paper:
EVANSTON,ILL., Xay 17, 1915. The use of the daily rainfall fre uency distribution as a test for errors in the manner indicated in Mr. hdspaper, Neems to me to furnish interesting and suggestive inclicatims of the habits of the observer and the character of the errors in the record turnell in by him. The value of these indications depends primarily upon the use which is made of them. Two general classes of use may be considered: (1) They may be used as a means of inspection of the observer and his work; and (2) they may be used as a guide in testing the wcuracy of cooperative rainfall observations and as indicating how to reach safe conclusions from them. FIO. 15. Average frequency curves for daily If your observers would take suggestions liinclly and seriously, if &aUs April to Geptember 1914 as mborded at l5 regular Weat& Buku having found a cooperative observer's observations consillerably in statlous and at 15 cooperative stations. error it were possible to replace him by some one else, and if a consider- (See table below.) able amount of inspection of a cooperative observer's work were pos- sible, the indications obtainml from the frequency-distribution testa Summary of daily rabfalla recorded infi9utc 15. advocated by 51r. Miller woulJ be of conRillerable value as a part of inspection. * * * -4s a guille in testing the accuracy of cooperative rainfall observa- tions and as inllicating how to reach safe conclusions from them, it At 1Srqularstatims ...... 322 487 WO I 70 seem8 to me that the frequency-distribution testa are likely tc, prove At U cooperative stations...... 102 246 w valuable if carefully condered anel applied. In my opinion, the 250 I neceaeady careful consickration involves traveling farther along :&-} ...... 32% 51% 117% 129% Yr. Miller's line 01 thought than he has yet gone, 'udgrng by his paper. I It may lead in some cases to added confidence in the records rather than the reverse The following paragraph is, possibly. a fair esample of a reason for inereme11 confidence based on frequency-distribution tests. Consider the "Summary of c!dly rainfalls" under fi ure 15. Thie, summary intlicatep that the cooperative observers rnissef220 "Tmes 'I (322-102) which may bc estimotd as 0.005 inch each, or a total of 1.1 inch: an1.1541 rainfalls between 0.01 and 0.21 inch (487-Wj). which maybe estimated at 0.12 inch each, or a total of 2S.9 inches. The total lo= thus established in theee two group is, therefore, 1.1$28.9=30 inches. This is less than 10 per cent of the total shown in Table 2. This evidence indicates, therefore, that the total precipitation lost I Year.% I Dnp. from the recortl in the form of smaller rainfalls which are unrecognized ...... I s0 or erroneously recorilell by the cooperative observers as a group, is Barnm ...... 16 ...... :: I 78 possibly lese than 10 per cent and almost certainly less than 20 per cent Butternut...... 18 89 Downing ...... 19 100 of the total amount. Are not other errors. in p;ut unavoidable, prob- Dulntb...... 39 ...... 14 102 ably hrger t.han t.his? For example, ir, not the error in the toil rein- EauClatre...... 19 I 86 Amherst ...... 16 81 fall for a region prwlucel by eculiarities in thc geographical distri- Qmntsburg...... I Ap leton ...... 10 108 Hayward...... :; ! E Be.?oit ...... 17 77 bution of the stations ami in delocation with reference to the to og- Oscaola...... 17 I 79 Ihdhead...... 11 w raphy frequently much greater than 10 per cent? JVill not any reiuc- Red Wing...... 9 ...... 15 1 101 tion in the number of cooperative stations tend strongly to increase the p ...... 13 ...... 70 atdls...... 11 ...... :: I 81 errors referre.I to in the precailing sentence'? w erhawer ...... I 7 ...... 17 ! 78 The errors in the recorded total precipitation produced by the mere N%0hau...... I5 ...... I 17 j I habit of concentrating the readings at 0.05,0.10,.0.15, etc., are probably Dodgeville...... 10 ...... 23 ' 184 Dub ua ...... 36 ...... 17 ! 111 less than 5 per cent., ant1 possibly lesa than 1per ceuttoo small to be of Q&?Rapids...... 11 I 31 i 111 importance. I reacher1 thk conclusion from R hasty study of Tables I mnawk...... 18 Ysnitowoc...... 17 I 98 and 2 and figwe 15...... 13 Illlwaukee...... 39 I 188 Hlllsboro..Hattleld...... I 19 New London...... 14 80 It. seems to me that Nr. Miller's reasoning iR conclusive in showing kc::::::::; ;;!;J 18 96 that the number of rainy days per year haa one significance for regular 37 :t I 71 stations and qi1it.e a iliffereut significance for cooperative stationtx- Lancsater ...... ! 18 ...... 15 91 80 .lohit F. Hrixford. Msoston...... 14 ...... , I. Mesdow vsuey...... 19 89 Afsddold...... 19 ...... M ... 1 Naillsville...... 21 66 Shebopgan...... 10iy/18 94 '0- ...... 14 81 Watertown ...... 100 Tm*EOTTEST REGIOk IN TmUNITED STATES. hien M 88 waupaes...... I 14 8s PM duC ...... I I 1 By G. H. WILLEON,District Forecaster. *Iyears cooperative, 5 years regular station. (Dated: Weather San Francisco, Cal., The eat differences between the records of stations Bureau, July 5,1915.) manne rby professional observers and those made by When the old seekers and ioneers came to California amateurs may be reduced to some estent by considerin in the early sfties many of t.R em crossed the deserts in only the same eriod, but they are mainly due to dx the southeastern portion of the St.ate, and the intense ferences in the Fi delity to duty, as way be easily shown. hetit experienced in that region duriw the summer This suggests the possibility of classifying observers with nioiiths sooil bccame well kno\.vll and mu& feared. "he ms ect to fidelity to dut by arranging them in the sufferings of both inan and henst while traveling over or aer of the total number o9 rainy days reported by them. those dreary wastes have been t,he subject of many iu- The table, or curve, of fre uency distribution of daily terestin papers, some of which were based upon facts rainfall, however, gives a vaP uable insight into the ob- while ot% ers were pure fiction. Undoubtedly many lives Unauthenticated | Downloaded 09/30/21 10:58 AM UTC