This document discusses simulating mineral deposits to estimate grade dispersions where true grade values are unknown. It involves:
(i) Interpreting actual grade data as a realization of a random function, characterized by its first two moments and distribution.
(ii) Generating multiple simulations/realizations of this random function that match actual data points. These simulated deposits have the advantage of known grades at all points, not just data locations.
(iii) The simulations represent different variants of the mineralized phenomenon as characterized by the model, but have the same dispersions as the true deposit, allowing dispersions to be estimated where true values are unknown.
This document discusses simulating mineral deposits to estimate grade dispersions where true grade values are unknown. It involves:
(i) Interpreting actual grade data as a realization of a random function, characterized by its first two moments and distribution.
(ii) Generating multiple simulations/realizations of this random function that match actual data points. These simulated deposits have the advantage of known grades at all points, not just data locations.
(iii) The simulations represent different variants of the mineralized phenomenon as characterized by the model, but have the same dispersions as the true deposit, allowing dispersions to be estimated where true values are unknown.
This document discusses simulating mineral deposits to estimate grade dispersions where true grade values are unknown. It involves:
(i) Interpreting actual grade data as a realization of a random function, characterized by its first two moments and distribution.
(ii) Generating multiple simulations/realizations of this random function that match actual data points. These simulated deposits have the advantage of known grades at all points, not just data locations.
(iii) The simulations represent different variants of the mineralized phenomenon as characterized by the model, but have the same dispersions as the true deposit, allowing dispersions to be estimated where true values are unknown.
This document discusses simulating mineral deposits to estimate grade dispersions where true grade values are unknown. It involves:
(i) Interpreting actual grade data as a realization of a random function, characterized by its first two moments and distribution.
(ii) Generating multiple simulations/realizations of this random function that match actual data points. These simulated deposits have the advantage of known grades at all points, not just data locations.
(iii) The simulations represent different variants of the mineralized phenomenon as characterized by the model, but have the same dispersions as the true deposit, allowing dispersions to be estimated where true values are unknown.
r:lrliltl()t.tS ltt tlvcrhurtlcll iIn(l mincrlrlizctl llrickncss' .llre hletrtlrltll l)r()c(\\ lrrttl tltc llcxitlrlitl'oI t]rc plirnl will dcpcrtd ()ll (l)c .l,rl,c,rt,,,l ol: tllt' rir:rtle t tc(e ttctl llt ;tll stltlcr (tllrill" rnonthly' l'clrrlr')' ll tlrr. lll \ltli lCillll\ \\.. le [)crlct'll\ ktlrlrr,tt' tllc reiltltrctl tlisl^.rsitlrts':rrttl, ,f,ru..,f,.tttoslrtrtlitLrlcrvtrrkillglllel[l(xls'ctltrltlheclctcrmincdbyapplying r rrr i.,u: srrnurirlctl [)r(]ce ssc5 1,, tlris rcillity. UnIortunatcly. the pcrIcct krrrrirlcdt1c ol tttts lrt stltl rcltlill' l\ llot avirilirblc at thc planninS stages oI an ()pcrxtr();.-llrc rrrIernurtion rrvuillble lt this stagc is usually very frag- ,,.l.n,rry, and iinrirccl (o thc grlrdcs of a [crv samplcs' The estimrtions ..l..lucc..lfronltlrisirifclrtrrlrtit>n-cVUllrhrorrghkriging_are[artooinrprc. e rsr. [0r tlte crlrct clrlcullrtiotls of dispe rsiorls thiit are reqtrired' r\s rt is tnl;lrlssrble ttl cstinllttc thc irr sittr rcality correctly.in strfiicicnt tlctlril..,rc.ir:,rpri. itlca rs to sintullltc it on tlrc brrsis of zl rnodel' In a rvay' it rcl! cicposit lrtrr-l srrnullrtiorls of rhis cicposit lrre dillcrcnt varients of thc ,r,u.," nli,.,.,,rlizcti plrenon)cnon, as characterizcd by a givcn model' Consitlcr, for cxirrnplc, thc true grirdc variablc z0(x) at each point -x of a <1cposit. The gc<lsraristical approich cons.ists in interpreting the spalial tlistributicln of the 8r:rdc u6(x) as I particular realiz-ation of a random Itrrrction z(.r ) -l-]ris i{F i. ch,,(,clcrizc(i by its tirst tw() momenls and its drslrrtrr.rtion Iurrctton. u'hich are cstlmJte d Irilnt rhe cxPcrimentirl d:rta This rrr,r.lcl r:. lititt:. strtrltlrlc ftrr t]tc llrircticlrl prtlblcrlr tlI dctcrmining r:trttrus drspersitlns oI rhc [radus in rhc dcposit' sirrcc thc drspcrsion varianccs oI Z(x\c..lnbccxprcssctlas;tfttnclitlnt,fthcsectlnd.tlrdermomcnlonly_ c()virri,ncc ()r !'irrloQrllm - cI f.rmtrla (ll '16) A simulation thus consists of (lrir\\'rntltttothcrrc:rlizitlton'-.(r)trf thisRFZ('t) l-hclworcalizations'rcal .rrtrl rtnlulltlctl. tlrllcr Ironr c:rch olhe r ltl givcn loclllittns but comc Irtlm the 'llltlcl{ljZ(,r)'rhclirstlw,<rm()IllCnr\:rntlthctrtllvitriatcdistrihtrrion Iurrctrtln o[ whrch arc lixed r\s Iirr as thc clrspcrsrtln oI thc sirrlullrtcci variirblc is conccrned' therc ts rr<: ditlcrcncc hlt:lwccn rlrc simulared <ieposit {:,(x)l and the real dcposil {.-.,(r)}..Illcsinlularcrldcprlsithlrstheadvantagcofbcingknownatall p()rnls r and not only at lhc c'rpcrrmcntal <jata points x"' This simulated .1"p,-,rir is also called, "'nume'ic"l nrodel" of the real deposit' \ilNl\(; (ilr()s L'\ I lsl l( s 2,.(x") = zo(r"), Vr.. C<trulitionin! l l-hcrc are an inlinite numbcr of possiblc realizations {2,(r)' s : I to co} of a ii Ztrl From among rhis infinity of rcalizations, the simulations {2,"(r)} rlrur are chosen are tliosi that mcer thc experimental data valucs at the :rcluirl clata locations x., i'e', thosc simulations for rvhich 'l-tris is kn, tlrc srnrul:ttc, poor tlata i r()l)u\lr'itss t thi: r,:ll dat: A. Journcl t Ioclrl drrf t. stirtlonar)'r lcIt'irrrntl st Iurlhtr rntp frorn thc re w'iistc scltml .Sitrr ri /alrort Sirrtuitrtccl r trons, and i as (he real ( Irorn an esl (i) 1-tre ol z.(x)w'hic criteria [or minimal m 'fhcre is n< variabilitY minimizati d ispersionl polvgonal constl}nt a cstlnliltc tl rs, thus. a i: ,( r )) {ii) On srnrulittior (mtitn artr grirtlcs (;1 lrut tIra(l( ' I r ) l\ I :ltown llt; lrrtctl vtlr- lrr gcnr bic. it ca I I I I I i ,r(/l (lc[)(l)(l ()ll tl}U \lilll) \,ilrliltl()t.tS lll ()VcrbUr(lcll iIn(, mincrillizctl lhickncss' .llretrletrtlrlttl)r()cc\\lrrttltltcl'lexitrrlitl'oIt]rcplirnlwilldcpcrtd()rltl)c .l,rl,e,rt,,,l ,,1:,1,.' ,1,,',.lt\ rctul\('(l ltt ltll stltlc: (tllrill" rnonthly' 1'clrrlr')' ll tlrr'rrr \llli lciltll\ \\clu [)crlctll\ !'tl0Ntt' tlrc retltrtrctl tlisl^-rsiorts' lttttl' ,rilu'.,r,.ttttlsIrrrtlitLrlcrvtlrkttlgttlellttlr'ls.ctltrltlhcclctcrmincdbyapplying r lrr i.,u: srrnulirlctl [)r(]ce sse \ 1., tlrrs rclrltly. UnIor(utratCly' the PCf Icct krrrrirlcdtc ol tttts lrt srltl rcltlill' l\ llot avirilablc at thc planninS stages oI an ()pcrarr();. 'llrc rrrIor nurtiort lrulrililble lt this stage is usually very f rag- ,.r.l.n,rry, and iinrrrccl to thc grlrclcs of a fcrv samplcs' The estimrtions .l..lucc..lfronirlrisirifclrtrtlrtit>n*cVUllthrotrqlrkriging-are[artooinrpre. e rsu frtr tltc e rlrct clrlcullrtiiltls t>f tlispe rsiorls thiit arC required' ,,\sitisinl;lrlsstblet()cstinliltCrhcr.rrsittrrcalitycorrectly.instrfllcient tlctlrrl..,tc rir:,lpric itlcli rs trl sinrulittc it on tirc bnrsis of zl rnodcl' In a rval" it rcll rlcposit trr'rd srntuletiorls of this dcposit lrrc <Jillcrcrrt varients of thc .r,,.,.," nli,.,.,,rlizctl plre non)cnon, as characteriz-cd by a given model' Corrsidcr,.for cx:rrnplc, thc true grirtlc variablc z0(x) at each point x of a <1cposit. The gctlsratistical approich cons.ists in interpreting the spatial rlistributitlnofthcgr'rdc:o(r)asaparticullrrrealizationofarandom Itrrrcri.n Z(,r). -l-his i{F i. ch,,',crrrizcii by irs firsr tw() momenls and its drstrrtrution Iurrctton. *'hie h arc cstlmJtcd Irilnt rhe cxpcrimentrrl data T]ris nr,rrlcl r:.. lititt:. strltlllrlc ftrr t]tc llrirctical prtlblcrrl tlI dctcrrnining rltrttrus c!rspersitlnsrlIrhcIradcsinrhctJcpilsit'sirlccrhcdrspcrsionvarianccsoI Z(x\c..tnbccxprcsscdasltIttnctitlrrtlfthcseC()nd.()r(lermomcnlonly_ c()virri.ncc ()r 'irrloQrllm - c[' f.rmtrla (ll '16) A simulation thusconsistsof (lrir\\'rtrrtltttttlhcrrc;rlizitlton'-'(t)tlf rhisRFZ('t) l-hclworcalizations'rcal ;.rrtrl rtnlulltlctl. tlrllcr Ironr c:rch olhe r ltl givcn loclllirtns but comc frtlm the .lurc ll.lj Z(.r ). rhc lirst lw<r m()rllcnl\ :lnd thc ttnlvitrialc distribtrrion Iurrctron o[ ivhrch ilrt lixed r\s Ilrr as thc clrspcrsrt)n oI thc sirrlullrtccl vari:rblc is conccrned' therc ts rr<: ditlcrcncc bctwcctl rltc simulatecl <ieposit {:'(x)l and thc real dcposil {--,,(r)}.'llrc sinlulatcrl dcposit h:rs the atlvantage of bcing known at all P()|ntsrandntltonlyatlhccxpcrrmcntalclatapointsr.,.Thissimulated .1"p,-,rit is also callecl , "'nume'ic"l nrodel" of the rcal deposit' \ilNl\(i (ilr()s L'\ I lsl l( s 2,.(r") = zs(r"), Vxo. Cctrulitioning -fhcrc are an intinitc number of possible realizations {2,(r)' s : I to co} of a ii Ztrl From among rhis infinity of realizations, the simulations {2,"(r)} rlrur arc chosen are tliosi that meer thc experimental data values at the :rclual clata locations x., i'e', thosc simulations for rvhich 'l'tirs is kn, tlrc srntulutc, poor data i r()l)u\l 11L:ss t thi: rrll dat: A. Journcl r Iocirl drrf t. statlonar)'r leIt -irlrntl st Iurlhcr rrrtp frorn {hc re *'irstc sctml St tttu-latiort Sirrtuitrtccl r tions, and I as (he real ( Iront an esl (i) 1lrc ol z.(x)whic critcria [or minimal m 'fhcre is n< variability minimizati d ispersionl polygonal constlnt a csttrrtatc ll rs, thus. a i: ,( r )) {ii) On sr mu lit tior (nltitn artr grirtlcs (:, truc gratl< :.,(r ) ts t :ltow'tl llt; lrrtcrl vllu lrr gcnt blc. lt ca -r(/l