Satellite Remote Sensing vs Interpolation

Embed Size (px)

Citation preview

  • 7/25/2019 Satellite Remote Sensing vs Interpolation

    1/8

    Production of Site/Time Specific Hourly Irradiances

    ---

    Satellite Remote Sensing vs. Network Interpolation

    Richard Perez, Robert Seals and Antoine Zelenka *

    The University at Albany, Albany, New York.* Swiss eteorolo!ical "nstit#te, Z#rich, Switzerland

    A$STRA%T

    "n this &a&er, we co'&are the acc#racy o( satellite)derived ti'esite s&eci(ic ho#rly irradiances, with that o(irradiances obtained via e+tra&olation andor inter&olation o( nearby !ro#nd)'eas#rin! stations. Aco'&rehensive st#dy #ndertaken by the "nternational ner!y A!ency -Zelenka et al., //01 had addressed this2#estion, b#t had li'ited its sco&e to daily total irradiances. The &resent st#dy (oc#ses on ho#rly data.

    . "NTR34U%T"3N

    Siteti'e s&eci(ic irradiance data are needed to si'#late the o#tt o( solar ener!y syste's 5e.!., P67 in relationto other ti'esite s&eci(ic 2#antities that are indirectly related to the solar reso#rce 5e.!., the load re2#ire'ents o(a #tility, a s#bstation or a b#ildin!7. Standard cli'atolo!ical data s#ch as ty&ical years, or cli'atolo!icallyre&resentative stochastic data 'ay be a&&ro&riate (or syste' desi!n or (or ener!y &rod#ction calc#lations, b#tthey are not s#((icient to st#dy a syste'8s interaction with its intended end)#se i( there is any relationshi&between the two.

    9i!. : 4istrib#tion o( !ro#nd)tr#th stations

    $eca#se o( the scarcity o( irradiance 'eas#rin! stations, the #se o( re&resentative sites where irradiance data are'eas#red or 'odeled has been a co''on &ractice (or en!ineerin! calc#lations. "n the USA, the National SolarRadiation 4ata $ase -NSR4$, //;1 incl#des data (or 0

  • 7/25/2019 Satellite Remote Sensing vs Interpolation

    2/8

    Ten o( these stations ) so#thern New York State and assach#setts )) were o&erated by Ascension Technolo!y5AT"7 on behal( o( NYSR4A -//)/B1 and 'eas#red !lobal, direct and di((#se irradiance with AT")ty&erotatin! shadowband radio'eters 5RSR7. 3ne station located near the Albany Air&ort, o&erated by ACSScienti(ic on behal( o( Nia!ara ohawk Power %or&oration -//)//B1 #sed a ichalsky)ty&e RSR. The laststation, located at the University at Albany -//)//B1, incl#des C3 (irst class 'eas#re'ents o( !lobal,

    direct and di((#se irradiance.

    All data were care(#lly scr#tinized thro#!h both a#to'atic and vis#al 2#ality control. "n addition, instr#'entcalibration thro#!ho#t the network was veri(ied, and i( needed, corrected a &osteriori on a 'onthly basis, toins#re consistency thro#!ho#t the network. %alibration veri(ication was based on clearest)ho#rs interco'&arisono( direct nor'al irradiance at all sites. The 2#ality control and calibration control &roced#res have been

    described by Perez and Seals, -//B1.

    0.0 Satellite 4ata

    The satellite data consist o( inter'ediate resol#tion i'a!es (ro' the visible channel o( N3AA8s !eostationaryweather satellite @3S D. @ro#nd resol#tion o( these i'a!es in the considered re!ion is o( the order o( k'latit#de by < k' lon!it#de. Navi!ated i'a!es coverin! the north A'erican continent and the Atlantic 3cean are

    distrib#ted on an ho#rly basis tro#!h the "nternet 4ata 4istrib#tion Syste', -//B1. A &ortion o( these i'a!es,coverin! the 'id)Atlantic re!ion o( the US has been archived at o#r %enter since ay //B (or researchr&oses. An e+a'&le o( the archived (ra'es is shown in 9i!. 0.

    9i!. 0: +a'&le o( the northeastern US inter'ediate resol#tion visible (ra'es (ro' @3S)D archived at o#r

    Research %enter

  • 7/25/2019 Satellite Remote Sensing vs Interpolation

    3/8

    The inter&olation sche'e eval#ated in this &a&er is a wei!hted avera!e !ravity inter&olation. The inter&olatedesti'ate is a wei!hted avera!e o( all s#rro#ndin! sites within a !iven acce&tance radi#s. The wei!hts are a(#nction o( the s2#are o( the inverse distance (ro' each station to the inter&olated site 5hence the ter' !ravity7.This inter&olation 'ethod is deter'inistic, that is, the wei!hts are ass#'ed, a &riori, to vary as a !iven (#nctiono( the distance. ore so&histicated inter&olation 'ethods s#ch as kri!in! wo#ld reset the wei!htin! sche'e (or

    each case, based on the c#rrent s&atial str#ct#re o( the radiation (ield 5esti'ated (ro' the network8s stations7.

    @ravity and other inter&olation techni2#es are disc#ssed in detail by Zelenka et al. -//01.

    "nter&olation acc#racy is 2#anti(ied in ter's o( RS by co'&arin! each site8s 'eas#re'ents to inter&olateddata (ro' two or 'ore o( the other sites.

    $eca#se o( the li'ited n#'ber o( sites and the non)ideal distrib#tion o( stations in the network, and, in order to

    di((erentiate inter&olation (ro' e+tra&olation, we selected only a (raction o( the &ossible inter&olationco'binations where the considered site wo#ld be reasonably well s#rro#nded by the inter&olatin! sites. Aselection criterion based on the ratio between the avera!e wei!hted inter&olatin! distance 4, and the avera!edistance (ro' considered site to the wei!hted barycenter o( the inter&olatin! sites > was #sed. 3nlyco'binations where 4> e+ceeded .FB were selected. The 'athe'atical e+&ression o( 4 and > are !ivenbelow.

    4 G idiwi

    where diis the distance (ro' each inter&olatin! site to the considered site and w iis the assi!ned wei!ht o( eachinter&olatin! site.

    > G H -i5+i)+7 wi10I-i5yi) y7 wi1

    0J.B

    where +, y, +iand yiare the %artesian coordinates o( the inter&olated and each inter&olatin! site res&ectively.

  • 7/25/2019 Satellite Remote Sensing vs Interpolation

    4/8

    9i!.

  • 7/25/2019 Satellite Remote Sensing vs Interpolation

    5/8

    ;.0 "nter&olation

    RSs obtained (or inter&olation are &lotted in 9i!. ; as a (#nction o( the distance (ro' the closest inter&olatin!site. Also &lotted, (or co'&arative r&oses are e+tra&olation RSs &resented above.

    9i!. ;: "nter&olation RS as a (#nction o( closest site distance

    A s'all b#t noticeable i'&rove'ent over e+tra&olation is a&&arent as distance increases. RS at shortdistances re'ains basically #nchan!ed (or short station distances.

    This observation is &robably a res#lt o( the (act that, in relation to e+tra&olation, inter&olation is 'ore e((icientdetector o( lar!e at'os&heric &atterns 5e.!., &assa!e o( (ronts7 than o( s'all scale clo#d str#ct#res.

    Satellite

    Satellite !lobal irradiance esti'ates obtained (or the closest &i+el to Albany (ro' ay //B to Oan#ary //Ewere co'&ared a!ainst ho#rly val#es centered on the satellite i'a!e ti'e. The relative RS corres&ondin! tothe ay /B)Oan#ary /E is DF Cs2. ', that is 0;.BK in relative ter's, and well in line with earlier satellite

    eval#ations (or Albany -Perez et al., //;1.

    =ence, in ter's o( RS, the satellite breaks even with e+tra&olation at 0F k' 5E 'iles7 as shown in 9i!. B.This is considerably shorter than the &revio#sly re&orted break) even distance (or daily irradiance o( B k', and2#ite consistent with o#r e+&ectations -Perez et al., //;1. "ndeed, the 'ain so#rce o( error (or the satellite))deter'ination o( clo#d thickness and t#rbidity )) is considerably less sensitive to the e((ect o( ti'e scale than

    the so#rce o( error (or !ro#nd 5inter7e+tra&olation )) s&atial clo#d str#ct#res.

    This res#lt is re'arkable !iven the &reli'inary nat#re o( the considered satellite al!orith' that,

    did not acco#nt (or evolvin! !ro#nd albedo, did not o&erate with a thoro#!hly validated clear sky 'odel,

    did not acco#nt (or &i+el)s#n !eo'etry 5i.e., the relevant &i+el (or a considered station 'ay not always

    be the closest, b#t another &i+el or a co'bination o( the closest and other &i+els, de&endin! on the&osition o( the s#n in the sky7,

    did not incl#de a &osteriori !ro#nd tr#th navi!ation 5&i+el &osition7 veri(ication.

    "t is interestin! to co'&are the relationshi& between satellite)esti'ated irradiance and !ro#nd 'eas#redirradiance 59i!. E7 and the relationshi& between two stations e+hibitin! an RS o( the order o( the satellite8s59i!. F7. The noise &atterns are si'ilar overall, with a co#&le o( interestin! di((erences. The center core o( thee+tra&olation &attern is ti!ht and the distrib#tion o( o#tliers is !rad#al and 2#ite sy''etrical. The center core o(

  • 7/25/2019 Satellite Remote Sensing vs Interpolation

    6/8

    &oints (or the satellite is not 2#ite as ti!ht ))likely a by&rod#ct o( the satellite8s al!orith' li'itation in ter's o(t#rbidity esti'ation o#tliers are both (ewer and 'ore scattered.

    9i!. B: Satellite &rediction acc#racy relative to !ro#nd)based e+tra&olation

    9i!. E: Satellite derived irradiance vs. ho#rly !ro#nd 'eas#re'ents in Albany

    Addressin! the so#rce o( error o( these o#tliers 'ay be a worthwhile aven#e to e+&lore to i'&rove satellite

    'odels. %oncernin! this latter &oint, it is interestin! to note that, when co'&arin! satellite esti'ates a!ainst 'in#te !ro#nd tr#th val#es, the RS increases sli!htly to 0FK. =owever when co'&arin! the satelliteesti'ates a!ainst any one o( )'in#te 'eas#re'ents within hal( an ho#r o( the satellite i'a!e ti'e, the lowestachievable RS !oes down to FK )) that is o( the order o( the n#!!et e((ect 'entioned earlier. 4#rin!variable conditions, the satellite esti'ate is re&resentative o( the !ro#nd 'eas#re'ent at least at so'e &oint inti'e within the ho#r s#rro#ndin! it.

    B. 3N@3"N@ RSAR%=

    Several aven#es o( research are bein! e+&lored as a lo!ical (ollow)#& to this work, incl#din!:

    . The investi!ation o( other radiation co'&onents s#ch as direct irradiance, tilted irradiance, ill#'inance,

    etc. As a &reli'inary (indin! o( this e((ort, we &resent, in 9i!. D, the observed e+tra&olation RS o(direct irradiance co'&ared to !lobal, showin! a considerable error increase (or direct

  • 7/25/2019 Satellite Remote Sensing vs Interpolation

    7/8

    0. The eval#ation o( !ro#nd)aided satellite 'ethods, whereby satellite esti'ates co#ld be corrected #sin!strate!ically located !ro#nd network stations

    9i!. F: %o'&arative ho#rly !lobal irradiance at two sites distant o( 0F k'

    . The syste'atic i'&rove'ent o( the &i+el)to)irradiance al!orith', incl#din! so'e o( the obvio#s ste&s'entioned above

    0. The co'&arative investi!ation o( other satellite al!orith's.

    E. %3N%?US"3N

    Ce co'&ared ho#rly siteti'e !lobal irradiances obtained via e+tra&olation and inter&olation o( !ro#nd stations,to that obtained by &rocessin! widely distrib#ted satellite i'a!es with a &reli'inary test 'odel re&resentative o(e+istin! statistical 'odels.

    Res#lts show that, (or ho#rly data, the satellite beco'es 'ore acc#rate than a local !ro#nd station i( the distance(ro' the station e+ceeds 0F k' 5E 'iles7, down (ro' a &revio#sly re&orted B k' ran!e (or daily irradiances.

    9or a re!#larly s&aced network, the break)even distance is esti'ated to be ;B k', which, (or a co#ntry like theUS, wo#ld re&resent close to one tho#sand stations.

    9i!. D: Relative e+tra&olation error o( direct irradiance co'&ared to !lobal irradiance

  • 7/25/2019 Satellite Remote Sensing vs Interpolation

    8/8

    A%MN3C?4@NT

    This work is s#&&orted by NYSR4A 5contract No. ;0E)RT)TRR)/B, O. =arvey, 3((icer7 and NR?5s#bcontract No. >A=)BB)00)0, 4. RennQ, 3((icer7. any thanks to Ascension Technolo!y, "nc., (or theirti'ely distrib#tion o( New York8s radiation network data and to Nia!ara ohawk Power %or&oration and ACS

    Scienti(ic (or their irradiance data.

    R9RN%S

    Albany "4P)AR station5//)//B7. site descri&tion in "A)S=%P Task F)8s (inal re&ort. "A, Paris9rance.

    @a#tier, %., @. 4iak and S. asse, 5/D7: Si'&le Physical odel to sti'ate "ncident Solar Radiation (ro'@3S 4ata. O. %li'ate A&&l. eteorol., /, &&. B)0.

    "nternet 4ata 4istrib#tion Syste'. U%AR, $o#lder, %3

    Masten, 9., 5/D7: A Si'&le Para'eterization o( the Pyrhelio'etric 9or'#la (or 4eter'inin! the ?inke

    T#rbidity 9actor. eteorol. R#ndsch.,