air force geophysics lab hanscom afb ma f/g an … · an automated cloud observation system (acos)...
TRANSCRIPT
AD-AIOA 266 AIR FORCE GEOPHYSICS LAB HANSCOM AFB MA F/G 4/2AN AUTOMATED CLOUD OBSERVATION SYSTEM (ACOS).(U)DEC AD E B GEISLER, D A CHISHOLM
UNCLASSIFIED AFGL-TR-81-0002 NL
AFGL-TR-81-090211110111111111110AL MUAW PAMilaE N0, M :~
An Automated CloudObservation System (ACOS)
EDWARD B. GEISLER, Capt, USAF 1w
DONALD A. CHISHOLM I)TFC
I!ELECTE 3
17 December 1980
Approved for publie relea; dieribution onlmited.
METEOROLOGY DIVISION PROJECT 6670AIR FORCE GEOPHYSICS LABORATORYHANSCOM APB, MASSACHUSETTS 01731
M AIR FORCE SYSTEMS COMMAND, USAF
81 6 15 141
r--F
This report has been reviewed by the ESD Information Office (01) and isreleasable to the National Technical Information Service (NTIS).
This technical report has been reviewed andis approved for publication.
FOR THE COMMANDER
C Cf Scientist
Qualified requestors may obtain additional copies from theDefense Technical Information Center. All others should apply to theNational Technical Information Service.
Thi tehia reor haVenrvee n
UnclassifiedSECURITY CLASSIFICATION OF THIS PAGE (Whe.n 0.. Entered)
REPORT DOCUMENTATION PAGE BEOECMLEIGFR1. REPORT NUMBER 12 GOVT ACCESSION NO, I. ECIPIENT'S CATALOG NUMBER
A PG 1, -TR -8 1-0002 IQ4. TITLE (and SubliIII.J 5- TYPE OF REPORT & PERIOD COVERED
AN.JUTOXMATrD A LOUD OBSERVATION Scientific. Interim.SYSTEM (ALLIS) 6 PERFORMING ORG. REPORT NIMBER
ERP No. 7227AU TIORf.) 8. CONTRACT OR GRANT NUMBERt.)
\Ed'aHd B3.?Geisler,-C-apt, USAF >'J' / -
D lonald A.)Chisholni,.- ..
9. PERFORMING ORGANIZATION NAME ANO ADDRESS -- ."I'MORAM ELEM&NT.PRO2 ECT, T ASKARE A & WORK UNIT NUMBERS
Air Force Geophyvsics Laboratory (LYU)
Hanscom AFB ,- 62101F
Mlassachusetts 017:31 j A ~104 .-
II. CONTROLLING OFFICE N AME AND ADDRESS -~Ia. IfkRE-RIATK
Air Force Geophysics Laboratory (LYU) 1 17 Decemnb.,-1980
H-anscom A F13 AlNUMMER OF PAGES
Massachusetts 01731 2III MONITORING AGENCY NAME & ADDI1ESS(II diferent from Controling Ofibce) IS. SECURITY CLASS. t(1 h514 -Bect
UnclassifiedI.DECLASSIFICATION DOWNGRADING
SCHEDULE
IS. DISTRIBUTION STATEMENT (,[ this Report)
Approved for public release, distribution unlimited.
17. DISTRIBUTION ST. 4ENT (of I abstract entered in Block 20. it different from, Report)
IS SUPPLEMENTARY E S
IN. KEY WORDS si,,,on~.ode it ......... -d8 Identify by block nrob.r)
Rotating heam cei lometer Cloud base heightHierarchical clustering l,ow cloud amountHuman observation Automated observation ceilingWeather test facility Multiple cloud layers
20 ABSTRACT (C nofi~oo -0 r ...... o. IIt ....... is*rd Id-liIc by block OoO.b.,I
Analysis of cloud hase height data collected during a sevenl-month periodfrom a three Rotating Beam) Ceilometer (RBC) network on Otis AlPH. Xlassa-
chusetts. demonstrated the accuracy of an automated cloud observation system.IThe high degree of correspondence between thet automated and human observa-tions of cloud height, low cioud amount, multiple cloud layers, and ceiling con.firms the accuracy of the hierarchical clustering technique when applied to anetwork of IfC's confined to the immediate environs of anl airfield. Testsdemonstrated on)lyV slight improvements, ins automated cloud observation are
DD I JAN 73 1473 UnclassifiedSECURITY CLASSIFICATION OF THIS PAGE 'Whe.n P1.1. Fonr,.Il
SECURITY CLASSIFICATION OF TIlS PAGE(Wh, Dots Elfl.,.d)
20. WCont)
realized by incorporating additional information from a second and third R13Con or near an airfield.
14
7- 7-~- -
Accession For
NTIS P&
DTlC TAB
Preface
1Iris \v~is i hom-filed 1'1011 he0 hel'1 Of ManyV noeWithOUt Whom11 ,,'l,!'(SSfUl
* mplet ion Ao uld wit have, heen realzeij. \\f e lre espei'ialk, grniteftfl to Janre
llr~ocl! v 01)t llmAny\ Io'l: (li-,ttiorls onl A\ -AWOS; (afn. tames \\ evinan and
Hichl:11-d I*v11eh hor lhoei so.it on (to autormite the r'otating beat, ceilonlt!or0I Leo
Ia) n- illo~o. am lvdj'I.ane In)!v( form-hinitaintop thre jiek terit itinut rurrierta -
Rk.l -wle U i eng' l .\rwi ,ioal t.,wxdI : 0ato wM, P ) i r aka tWo'klaaon1 .
W' MOej)at i1W 1i' h lllIr! 21 01 ; anid Lx-it I 'l 1; 0!o! I irif'
* 1-
Contents
1. INTRODUCTION 7
2. HUMAN CLOUD OBSERVATIONS 8
3. CLOUD INSTRUMENTATION 9
3. 1 Rotating Beam Ceilometer 93.2 Network Spacing 10
4. AUTOMATED CLOUD PROCESSING 10
4. 1 Preprocessing 104. 2 Autcnated Cloud Observation Determination 11
5. DATA SETS 15
6. tOMPARATIvE ANALYSIS 16
7. SUMMARY AND CONCLUSIONS 21
I0FI EIRI'NCES 2:3
AI\''IINI)IX .\: (( \lgoritho 25
Illustrations
1. 0 ,ti- :\1,'!i RI(" Co tiguiration 10
2. Fxamph., of lhi'r:arrhi'al (11usterint 14
3. '1in Scries W Nt'()S and Hum an ()hiserwations 15
-. lint, Svriee t :AOS and Human ()hstcrvations (ceiling) 19
-- - -. - - - - S
Tables
1. Raw 1113C Data 12
2. Modtliods of' (1roupinr ('eilornetet D~ata 13
:1. AC OS S an ieJ Output 15
1.Autoplated Oir."ervatioll I )ata Reolative i'eqluencV l)istrihutioe s'for' Method No. 12 16
5. Sam pit Si zes or loipa tative Analvsi a I
6. Joint Occurrence or Ciling Reports (All ( ases 1 16
i onti rigene T abtes Comiparing Perctee Rel[at ive 1 requeneWVof 1.11Jv Cloud AirioUMS. 20
8. Contingenev Tables Comparing4 1' rrent Relative FvrequenecvOr 1 0%1. (' loud Ives22
C)I
An Automated Cloud Observation System (ACOS)
'1t11 ( SAl Air \eathor Se rvic ( AWS) has recognized the need to ndet ni ze its
has i woathert : olpport capabili ties. The Automtated Weather D~istribution System
(A\\ ItS) %hrlt i - ormitiand lleqni red Operational Capabilitv (110C 80 1-77) calls, in
irt. tor t1 ttilk l\'.towitk-td airfield weather (ibsorving and short range forecasting
iah1111% i t ))''H)lt 1x~1-baio permintt airt'ells and at hare-base tactical or
ii~riiv ~ twns, i tespnsw tot these, USAFI requiremients, the Air F~orce Ce,(o-
pdtv-i I .ahnnr'otnnv deve 1 loped a low-cost, hilly automated mircrocomtputer-
ha'ni-vf '.,\ M WS (?inilrl~lt Attl Weather SStitti1). The two-year test of
NI \\'.~ it 1n I'T tmistitat el and provided sufficient evidence that the require -
111"-'1 Ir -1 t itt1tonizted mrirticld obsi vat inn svstem is a satisfactorv solu-
tnt' Iir id) ' Aol11)1( hase Iweo!lnt d),s-ivatirlt front the rotating-beant ceiloni -
M I" . I 11,1,1illts . i1 ,hlnvn 11i1k satisticltoty attiirateil processing oif the
HIt -I I lI \ n-w 't the slint 0il Ill tinNIAWS lentonstration. Subsequent
(Wi int \W- ~ Ii'incstiiiint \ 'tArnard IVnchll have developedl and
I . h1isho'lill. I ) .A. I vni , HI. It. , %% evtinati, 1 .I.'. it t;n'islor, 1F. 13. (i1980) AI t-7n'titiitinti I ist it, thre MolaIr Autitwniti'1 \1.iather Svtt(i Mi1AWS),
2. \%(-Vrs:1r1, 1.( -ille I lyncwh. H. 11. (1 1) A Dligital P rocessing and ilisplayS','tn'r i 'n l~ntaiii4 Ca- o 'ilnteIr AN (PrIQ-fl), in pre))ratin
t ested hardware and software refinements w hich have SLLCrecSS t'ul lv od'.ed this
o rob Iew
T he purpose of titis study is to dev e op oibj ective pro( edu res to ( pmI i 1.1doi
base hiight(s) I. ow% cloud amiount categories, and ctoud .ciillrg, byv expanding Upon
he basi c- processing procedures previouslY d2v , 'ped. '1hipproach taken relies
a)n thet mret hodology initially developed by Duda et AI and aprpli ed by tire National
Weather Service Ii thiir test of the Aviation A utomated \\ rather Ohseivation
S\'stenr (A\ -A\\ OS).1 I'lhe A% -AWS test showed good agreene ent h 'tween alto -
rnatcd and human lrr(u,. observat ions. Thie autotrate-l % ine-; weore obtained front a
rtva;igrrar arvray of llHC' 7 mlites apart whrle' the hctrran ohA)el \ dion., were obtained
treatr Ih Irtianglc's center point . in the tests conducted here, hei( average length )I'
lhe t riarugle legs is aboir 1.5 a riles. 'ibis smaliler H tic network, dentloved ait (ho
AFG IC eat her lest Vacil Iitv I WI F) at Otis AFB [, Massarittise tta , allowecd a futhier
test of tilh lrareiial (.1inste rlng, algorithm conistrainred to) a tviirl aiie~ild environ -
Ilierit. '[his r'epoirt dtoumlents a 7 -tmotli test iof AFICl's Autoniated Cloud Obser~a-
lionmi Sva err IACOS I has ed onl compjari suns between ACOS an III-imoian roibserOvatilons.
2. IIL'MAN C;LOUD) OBSERVATIONS
F~ede rat \ leteorologi rat Handbook Ni). I (FtAiHI I' delineates the procedures to
be useit in fo rrnrr1ti a weather observation, Including de~tailed'u in atrct ions r'e -
lating thle statr' or appearance of' the, skv-lo-skv rover by cloulds adol. ii'ibscuritIg,
phenomena. It s ates 'a rootplete evauatiron of sky corndition irelirdes thle type of
c lourds or obscuiring phtictinwili presenrt, their strati fication, atmouittit, olpar.i lv
direction of trovetinent, height of' bases, arid tire effect onl vertical visibility of,
surface -baseil obscuri rig phenoirntiMa.
The tlirman obiserver wiho must describe lie state of tire sky las a most di ffi'rLt
task. Ani observer must identity cloud 'aver's, estimate the ewight Woa tcl layrer',
determine the ercenrtai' of sky cover, ant itentifl- tire tfl'p 'hil(t hresert.. It)
atioltnn, the observer musrt leterriior tilie stniunt if sky nt visiblte tue to strfaice-
traseit ohsr5'tiofls andt the vertical v islbiltiv Ii the obscuring, pliicnonienra. Tiierer
a rn' se'veral limitationls ar observer has to wi t'k with. I'r'errotty, the observer's
%lew" of, tire honrizorn i,, lirittd try pfirsirat obstructlonis surci as ain airport tertmoinail
31. D udta, Hi. 0. , M\lanruso, HI. L. , and lPaske rt, P-. V". (197 1) Analysis of' Tectini -(lilres for' I t'sci' 'burgE thle State of tithe Sky Through Auttomationi, Report No.P A A -U 13 --iI5 2.
4. Bradley, .1. ,.e ikow itz, M. , and Lewis, HI. ( 11179) Aviation Automoaterd Weathe'rOtns'rrvation Svsteri (A% -AWOS), Report No. FAA -HI)-79-63.
a. NOAA -Natlonrat We'athier Servicre (191701 Ferder'al M1eteorologi cal IHandbook No. 1.Sirrfanco Observat ions, I'. S. Governmrent t''rinti lri Offic',
',,d oither buildings. The cloud height measuring device (R BC) is often a mile or
more from the ob- -rver's location. Hence, the observer is r'equired to determine,
through visual observation. the representativeness of the cloud height measure -
mieits in the overall observing area. The subjectivitY of the process results in a
natural variability in cloud observatioins among observers. White visual acuity,
depth perception, and fatigue level play im~portant roles, the so -called packing
effect is a substantial contributor to this variability. Hlere the observer mistakenly
incorporates vertica development of clouds into the cloudI base extent resulting in
an inflated estimate of' cloud amount(s). lThis inherent variability in liunian cloudobserlvationl nedis to he icini zed in the cyvaluation of objetv adutmated
techiniques. Total replicationl of' human ohiervaitions should nut be anticipated nor
soughlt.
'lhe hiuman obscrvations usedill tis- study we~re I aken by FAA operational per -
sonnel located at tic( opposite sideo of the airfield onl Otis AFB'l, approximately I mile
'from the AFlG I. WI'FI. Supplemental observations, specifically for ihce purposes of'
this study. wker, not ohtainle(l due to inanipower coist constraints. Rather, we relied
oli the observations, cnitiliclv obtailied as req (uired by FIlI - I, transferred theml to
tliulter comlpatible tormi and uised them in Ihe comparative a-nalysis. Since thet
R It is not itntenicd to mewasure- the extenit anid idepth of obscuring pheinona nor, to
(dlnti tv clud type, thle auitinated tei'tutique developed and teated in this study was
Imitted to clouid hei iht Is) I, lo cloud amlount i'ategories, and cloitld ceiling.
:1. CIA)I I INSTRlI 'NIENT.VIION
.1. 1 Hol ing lk-mil Ciloneler
Ilii st id i rd W\ S climtlight si(t 4; \N"m( - I o l s use d toil balsic data act-
In I iis-illosc-pc ili'tiior. ihle lii'iicctiu' his ai ill tunlgstenl filziinent projec'itioni
i'1\s' l,t'( i l iaiic~lltiI 111iiililtel Lit t12(0 lIz, Mid oiiiitiiiioltvl 1ot~ited( L't 5 ~i i
.li'ir is t\pi pul> sit l1t0 ft froml t1n' lcii''bcM 'Aith its fielId (it \it'\% \ icliCAl :1n
'tlnit- Nitil the. rnotiig 11witul em.li \iili(it' 0 itiisi'i"'tiiii l(hi~eS
llird 's i'il plitiio~cti heii So'IOs toi tlhe %itc~l 1 lieu the \tIilhPuI inter ISeC tS
'nil, hi j'tel' oIi the 1wtil1ct1n be \ inbp.i liii's inl the( u'liin is ileticteil bYtit(' 'ind icdlihsplaiil onl the, inldicaitor In i' fourm of a hieighit \s illtellsit\ de(-
Iiela Ili' pr1i)hctoI' angle, %iwich the na~xmmun hbicksc~iter chll (t--- i s
Nwi.' s tli' il' nld Iwil:lit. kwei':usi' o)f know\i sil'il( .lid tiijololiut ic linnlit'itionsanid
hut' .)I iSe-ti'It'i ic llft uat.1 iibt~inml it (n)is 14 cI 'lu heights inl our'
in nr epO iflents boh ln~p £ ofile~ih Iit(wereuse ~euitflgin en
r'c P e toud he-i iht "CIfsP ' 1 1 n(t w ic n(f theiI ;'f1ti i eeet? 1ei
-, u ie S ~ d d o fi~ ithc~ ~ i t i l " ( I d -I *%
w h n p re C t , r ,~l4 ~ i e k r;,n h i e x t r; e t e i
1t ed V W i5P& I t io o t, '~ \V~ O f o t ' l O ' h t o not 1 i e i 'b I
)Ilkon pcgif(ift ilO 1 ~ t( i l II,- d t c ijfljtion of 1 i b '( I 'udII.
i' wil 1drCS~'thi- ill SL'ctiOll 0
0 Ic 1 1iP0~' he flet\OIB of 11S ujIed ill th s t1d> hy ir u; ie
N,,,tc1 A oppt a l Is h oti. rfl o h A, tOi 1ll's 'fr~ li j iw u 1
t i o d o s 11.' tinfi te if.,
obtin tl
t-( I d i t i L t i l , f I~ thi ' I. et i g a V r
also~iil jkocltigu)f tjitlFAA
0'0
W( Ilf) iirc obt~irid iihiv()Uflieftl\% loti tillvc IM1lS A ( )ti. tIld rnfld(d(II11g
flttIL. t~ipc .t A (G L. lizwch lxi(. l,111 rip hie twon ('131 \ tlets p(.r slen 1%.(. serisl
per nlifirrite). .,li (-;I h scifi Of k-t~ Jhunl). tht' iiigutilt( Ill thu w~k , ri-R-. rxri
Oriluvid i id tilt, tkwo 1.4 rgest pc(iks vctauincd is (fll'. , liusted rus AlrxI rid Alzix2 inl
Vitblt1. Als\1o, i 1-rin irirer (AM lii lur is rjrt~l dvt-rimriiiu kwn nuLzrimp
b ,sed i nll till tidY, Al.\ v I litir' fibti ft-ld e rung'l thuC , 11IrIlt Iiiiilit iid, LI! nYtii
~-u'~tnii-, dmljl i-u~~\nill rttiny.rrgtn ne linlY li.t'.in u
ItIOUI r ii - I \. - i t t.i n i! P11(1 in jl- : snenuirri 1 .'.1 =Ii U
'Ino 'In
- 1 ti, I I I I I I '- n I:!n 1 1 .
I. t(I H. l- .1
If( I! h.~ -i: *-OO it.i.-I !
\111.111h-d (Jo~d ()I-1atio l I e~e 116 1.i1lot
l averes, the accum ulatedl numtber of eI (vents tip to anid intc ludinrg tire aver a re used toi
c lassi rv the cLoud amount catego ry lfor the h irgher hi ve Is. Hase! )it data (collected
at Otis, we modi fied the AV -A\WOS population proportions to0 i. 04, 0. 5 1, and 0). 815
as break points front clear to scattered (SCI I. scatte'redi to broken MBKN I arid
broken to (oerc'as't (0%\C). Figure 3 is an example of outptit P i' 'h lustcrilc :(. I)
nique with k-lotid observations (heihts ando amounts) dlotted it ;I :i -tj 0 itttvil
overI a lb-hr priod. It, "cal demlonstrates; the, alijli t tb Ol v u rm,iii
to liamil.' t11tutiple kiver anld e'volvin Ii-mo con0f ditionls With 'IotCt......-s riterhI 1l,
cortsistoI''t .1id 1 illtCeolotgicallv sotiOi. l!ie corresjronti. iiii - ibs',rv Itilils.
plotted il AlIii tht tilnit axis, iolioistrite tie aultorriatel ti !" i~ iaai ltv
handle this coiiplox situation, IIIt fat, ieI( ittoitilti'ti'i i o- i
chanlges (imill roveroierits) inl ceilirg coirtionus w~ell hiibire t- 1,:1 111'Ir:
becollet tiamIv recoirdedi, iar'ticularl' after ilarkrtess fil. Ialir 31 Is :ilari
,f output tot-, it-.Ak0S observation.
NUMBER OF CLUSTERS 6 5 4 3 21
1.12
'Iy illN:. A O '11p eO t u
: I ser n . I ,, i630 7 S ( 'T S aIl I1 )in 30hssrAt
8(5 Spt 18" 163:1 11 St.' 1 I HN 28 M\ C
8,0 Sip t 18/ 1633 ( 7 SrIT 19 [NN 29 (A5.
to is it (Litii Xki'ItIlIOtri fVloi NhI~ls..I t'l WfIstsi-e 1980. In1 II.
timt\ -oili i3)issil5 fol iiitI[iiIIis experimlent t\ it-, sp~pronxilntek 600 comn-
jl' rismi r11it, be ht\%i (ti t ,tlisstr( ziId Isinim ibsi m' iOis. No 'siino t';5sir55
lniii* hct s. I he jiumimT iin(I i tonits (I b-. ri ;trii.-i ii in0 obsc!.. vitl]onlSi
1 t 1 :Ili III obscrvler. I\iS t 0 t sclectid is hsi, Is 11t thn' hf)oll ifsg
I.rite rjs: I )11 Ib jpis5551 Itigtl! N: -t leist lr Ii dural~tjion, (2) zit leist (Ine sft
I, ' t ir I .. ( IA, i- s~pI.oitinv . sod (,;) sen~t rid or. Inns's cloudis beloiw 81W0 ftR i
vrpsirtisI b\ tin' I AA\ obsir'si's. TI';vs 4 .5i1110!siiis ths rs'Iastivc( fssqns'ni' dist 'ibui-
tsirm i HI ID) .i lt' it sinsiitcdi cloudl ihsrv;stjOns- for lilt, toirt.\-olss (s pis sAI Sot.
tlnI ' I e tit undc r 6(iOhO In~n 11 hs I\ t InIs it th i , ill :it; '0 t~ r.
1,b'ie 4 . \lutoinitcd ( )1).,t- I\ it In I ) It I ,\(
I rk I ti n - I2(1 u (it r \ t1
0 \ 1 11' (1N 1 .1 t S) -
1,7 1. 1 *~
t'2 I _I
7r 7I
others one of them was not Livailable at ainy time. Of the fortY-one episodes used,
only ten episodes (spanning 266 hr's of com pa risons) included dab fromt aill three
RBK' s.
labhle 5. Sample Si/es for (rinpatraitivi(
Antilvsis
1-5 !--
6-10 .1
I11-)5 26,,
Questions I and 2 aibove were add reS,(.d using tia )Ise 1:cti for whuch
ceiling condition (miore than 50". cloud cover-i beloia 6kU ft) 'S -terimeiiid to
exist by eithe r the FAA obse rve r of the ZilutOlitte i) p oci -dIi IV ( ( )S). . hhe 6
lists the percentage of agreement realizied by -:~ tc thc fiftt en metthods (I' p touping
ceilomieter t' ata (see Ta',ble 2 for dlescription of g loupig c titci t t. Rega rd iIg tht-
comipa rison of inidividua scan datai to I -mn meatn 1ilas"vte tit, diffi-ernce be- It% fnA\Ithud 2 %s S 1. 6 v-ij rd 12 v,; 1 1. i Fes vilts stwiges~t nothi eg has beenf
lo-.t h\% lCpt'OCe.Sing c101iii sclm datai into 1-lain mean 4.lu )1. the clustering
Igmrithut. lecAll the Algi itillm deVrloptnettIt Ia1- breTt s t ructa red 1o Use ., weighted
30 min if dIA ta P provide ill possibil it% of' 80 -vents wocctinig when I -nun ita ns
tive eniploved. Tis numiber () events to-.,:' s timplirng s tabili t' Consistent v ith
tVIe reconirnetdition of DIh et Al. these tests confirm thait stabilit, and
dinsti-tate only% imargimil a dditive contrtibution using the g teat lv, inc reased numbe r
ot vents in indlividua scani data. lResitlt> of' MtIhodts 2 a nd :3, -1and 8, and 12 and
Im3 dentutiSt rate' SeconIl(vr peak info roimition does not ,ont ribute to imp roved s pecifi -
Ition. The Slight decrc:ise -,I i1g teiuint resiltiig from the tfwit- lp' algorithmls
is; IMiri d rit. to itilogiis 'iiie-lamip'' ilgirithtiis (1! Jc ~nil..tliod 5 vs
Mdethod A) is probably\ tcite d to litip misaignmenit a ithiti lie or more H1IA'5s.
T his ca n resuilt in Ch a~d heig hts frtomi ie twko Ia nips heinog s uffi cientl\ differtent to
alter the CILouti am)ount i(tCLminaitin mt sitwtititlts httdcrimg on thet threshold )f a
ceilinig coniditiot i tht is, 5 ' I 0thi cloud covetv ± 0. 05).
men eattn assess the Itddt jt e effect to ceiling spc ficttion otf mtultiple I 13( 's by,
Coltip tin9 reCsults IcVS tom tv (lit roka ii I ahli 6. 1-mt ixatileA( Methods I. 6 aInd
1 1 te Al I t, t ed ,th lo I( i(t 1-11nit utuati cloud bise height data aIS prOt'ided byv
the lpteIrtoti(5iti protcedutre, tv hile Methods 5, 11) iid 15 ar ti se oiltt ) the primiar.v
tirt secottdarv peaik valIes friim e;wich silt1. Iii tctl Itnstance, I s;light inc tease to
17k
aigreuent is r'ealized by u~dding a seconud IMCI inil then , tird 10WY. Tius suggeists
that at airfields with two 11I:' s presentl ' instaled, both C~OUld be itili i.ed with ai
miodest inc teaise in confidence in a utomrated cloud obsvv tAtiOns. Lcjuall% importajnt,
if nlot more1- so, anton11ited cloud obse r% it ions ftcm cc singl(. ItH(' have been shownl
to hliv e substaintial agreement with Ob 1'% a hcIlt 1011.1 1111(-! most Xe'Acr e situat ions.
This St ucl has demonst rated (-oilfi roting ftc', c'; iicr ~'\ \ ' tiidinuc 4 the addc
investment into a second or)- thircd C1loud h~s sc igh4 ~scc (:- toI th- ii cclic- iciflt
iwithini 2 to 3 miles) of -,i aiirfie-ld is not f r (rtd c itcice. c c-i
tion purposes e~xcept. perhaps, III UfUr cit_1X cc, i1c1lc-x I If III cI' I .1rcii
TFable 6. A-iirit ()kccurrrn'c- ol! t itliLI lw:cac t cr1 fi
I - lHIM ,' 2-1M I-l I 11
131
1, gue is :I c!'cc n.o s )1, Al I ti c m I'. 1,11( Vi IXK c- I H'a
oi !3S-pc if! I o. Iffwhw h.
2; t f ;! .-. j I
height lpte -t jo t o iI h" millc i Icc- cce i,-- ihler !1 Ip-if i r. c ): c',I,, Wi;
h-h" t. c-i -1l Ac I 'cci Ic Q <ic. I i ?631 '110. 1c'cc-' 7
LiII -! 1 i rl 1c iid r w c -I fl. 1 ~ O cc iifici -,sll c 111,,c
10 10o above in the l-RBC case). This impacts about 10% of the ceiling determina-
tions (fot- example, iri the l-RI3C case, human observations denote ceiling condi-ions 89% compared to 79! for the automated method). The difference can be due
to deficiences in the method by which the automated procedure deduces cloud amount
from t frequency oi occurrence of cloud height data (or population proportions as
discussed in Section 4. 2). Further refinement of the selected breakpoints between
cloud amount catelories could be accomplisied beyond the modification to the
AV -:\\WOS recommended values done in this study. For the purpose of this study,
however, additional refinement is not felt to be necessary. Conversely, the differ-
e'Cs 'ould be due to biases in the human observations caused by confounding cloud
1ittorns, contrast and illumination uncertainties (especially after darkness), or the
press of' lhr respons ib ities impacting on the time spent preparing the cloud
)bservatins. Note also that, for the cloud episodes used in this study, the human
o)bservation data set contained 2% or less clear conditions below 6000 ft while 8% of
the automated observations indicated clear conditions. This can be attributed to
distant, lower level clouds beyond the range of the three RBC's but observable by
the tower personnel.
HU IMPTtI . I, NO
FOP C N - 1 ,2 iC- ME TH " I E M TH -
7 -
A
5'PR T :r 18 SEt '?R bh 20b6C 0 L4 F 1j I , 139, 6:Of,
- I
* 1 -
4'f
, 0C' :. 3J
F'ig4ure 4. Time Series of ACOS and Human)bservations (Ceiling)
l19
[able Bi shows;- the percent relative frequency distribution of the number of
cloud layers below 6000 ft reported by the humian observers and the automated pro-
ed,(ures based on 1 -mni ni :'in values (Methods 1. 6, and 11). Here again, we find
atendcency for the automnatc.1 technique to in' ers pecit'v cloud c'ondi tions. Note in
he I -I i case, 250% of the ohs ervation,; fell below the diagonal indi cati ng ir ore
loud Lavers iii the hum an obse rvat ions t han in concomitant auto mated observations
%%Iiite ontv I 5%1 fell above the diagonal. The relative istribution of the ntumber of
Ict(i~ lavers, ts- re flectedt in thie 'Olumn and row tot al-, s without the bias re flerted
in ilre cloud amiount i'ategorv evaluation. Refer to Figure 2 for an example of the
illeteoroloi catI cons istency and i'''r seotlativerress of' the automated cloud observa -
tions in a mtitti t-layered cloud tgne
The results in Tales 7 and 8i support the g-eneral findling presented earlier that
loud spcilfication is not iminproved substantiallY byv adding basic cloud height infor -
orlation fromt a Second or thirid R13C in the i mmedliate vi ciriirv of the airfield. WhilIe,
in most situations, -a s econdl H13('s data w ill, not cont ribIut e to specification ito prox' C-
mrent, for certain multi -layered and variable c loud conditions it will have a positive,
stabilizing effect on the cloud charactercistics rlete rmi nation.
7.St NINI.Ul A~ND CONCLUSIONS
A na[vs is of cloud bas e hei ght data colltect ed ituring a s ('veil-month p( rid m Coin
at three H tC network on Otis AFB'l, Alass-actitsetia, deimcost rated Ihe accuralcy of
ain autotmrated c lou ohseorv'ing sys temi. 'Ilre hi_,,i degree oftencorres pondence between
he autotiratelI and thuman observatins o'toltud height, low cloud amnount, multiple
lorud layers, anrd ceilIintg confirmos the accuracy, of the hierarchical clustering tech-
iiqrle klren applied to a network of' IMCsr confined to the immediate environs of an
Al I'll t [ t. Vl'h 1 eotietoMIs of' 'omrblining basic cloud-height data were fortir iated
ani i eat m on data gathered for lurtv-orle epi soides of extensive tow cloudine(s5
tires,' tests demonstratedt only alight i inirovetrrerts in automatedt clond Observatiott
are r'eali zed lbv i ncorporati ni. addi ti onal informilation from a a e'rtri andi titrd H It00 ''C near an airfield. In aditiotn, there is a tron-signi fici'it itifferetic ini testtesults when peak r-eturtls or mrultiplte retturtns from inrdividiual cltoud heighrt scans
are utsed itnstead of 1-mn n mean cloudi height data sets. Baseil oil ('lo)inisonts with
ot ri'tiotiallv-obtained humian obtservations, the automratedt techniques itttret'v
loudi aioitt c'ateLgt'ies aind th(u numbet' of tow 'nuit [avers try about 1ff-1. fi'c'tt
2 1
It is concluded, based on these tests and earlier tests of' the A% -AWOS system,
that automnated procedures based on hierarchical clustering yield stable, reliable
and representative standard cloud observations. The ability to successfully inter-
race the H BC and] continuously process its output witlb a iroocsr-h d
system has been lemitIrated in tlicstc tecsts and] oachie ones. 2Wleother use-d as
a basic stand-alonc iutiiiitate' cloud )servation m-odule or as a re al-time support
to human dosorxves, i morlzatjoli W, h1 clou obsecrving Ifunkt in can Iw unider-
taken, even with -coptimunallv miat ure sensors like the 1313C.
The' heriCal 'UStermnl technique is being integrated into) the MAWS at lte
A PCL WI I as a part ()r a continuininvstv-igaticon )t' sh!ort rwige prediction svstenls.
Specifically, anl autonmate.l short-range Cto a sting (3hO to I O mmi)xp ~juVn t is
uiiderwaY utilizing the anltolicatti lhservations, as thie'vepen sample. lo re -
casts Of Cloud -Ccilog 'A ilili in made il lirobabi list i c and cate-gorical Cormnat.
References
1. Chisholm, 1). A. , lyvnch, R. 11. , \evnan, J. C. , ant CGeislei, F. 1. ( 19i8O)A ;) ittnstratin lest ort thei Modiulari Aitomtodv %keatter SNystem (MAV~ 5),A FGL -T 11- 8 U0087 Al1) A 087 07 0.
2. keman . .1.C. , and lynch. H. If. (1981) A lOigilal Processing and tDisplavSYstem11 tot' the Riotating Heaml Ceilonmetei (AN.U\AIQ -13), ill pieparationt.
3l )tida, H. 0. , IMancuso. HI. I.. . and laskeit. 11. F. ( 19i7 1) .\nalvsis of, TechniClues roii 111mo ce Siate )I' Owe Sk ' thitigh Automation, RepiitN o. F iT-71 -) 2.
4. Hzilev. .. , 1.ekiwiti' M.! . auld leowis, H. (199)71 A% iatioit Aiitiittiitei%k katlwh ()liseiv ati_1 "-e wr (A -ttt N.\% \oSl, icpoit Ni(. tAA -Im)77i-63.
5. .NotAA -v,;iotat \\ catier S4fr\ ieo ( 170)~ Federal MeHoilIainthook No. 1,Surtace t)hse rvat iOhs.; t S. Gwveinitleit l'ilnttig (ItHwo.
23
1) NO) N(K) I 111(1)
where
I) -I a-~1 *;(uzI-& listiin e,
ii (Ir 41t'r ht'2r2)fl
N\ \1N 1h l ol 'Amid0 hit. ill clu t or.
wl~~~ ~ ~ ~ It :)-( !1m 1 111,111h' ''
fl(') V,.1( )
(11) Round cluster heights to:
Surface to 5000 ft: nearest 100 ft.
5000 to 6000 ft: nearest 500 ft.
(12) Sky cover shall be calculated by using the following criteria:
(a) Cloud cover factor (R L ) is calculated starting with the
lowest cluster (layer)
n
(Total Number of Layer Hits)L= I
RL (Total Possible Hits)
where n is the cluster order starting from the lowest layer.
For subsequent layers (L> 1), the summation principal
from FMH-l is applied,
(b) If less than five hits from all ceilometers, "CLR BLO 60" is
stored,
(c) If R L -5 0. 04, height and "scattered" is stored,
(d) If R L ! 0. 85, height and "broken" is stored,
(e) If R L > 0.85, height and "overcast" is stored.
(13) Cloud data is displayed as follows:
(a) Clear is printed if no clouds,
(b) If one layer, report it,
(c) If two or three layers, report them with only one
overcast layer,
(d) If there are more than three layers, a total of
thret! layers is reporteoi in the foil, % ing order:
(1) lowest scatterd,
(2) lowest bro* *.m.
u.i) lowest ov er, -.1
(14) Generate a nrew hs-'"v ition '-verv TOinute.
I;
I I Il : " I -I III i - I : " - -' _L :'I