You are on page 1of 50

The NCEP/NCAR 40-Year

Reanalysis Project
E. Kalnay,* M. Kanamitsu,* R. Kistler,* W. Collins,* D. Deaven,* L.
Gandin,* M. Iredell,* S. Saha,* G. White,* J. Woollen,* Y. hu,* M. Chelliah,! W.
E"isu#a$i,!
W. %i&&ins,!J. 'ano(ia$.) K. C. Mo,! C. Ro*ele(s$i,! J. Wan&,!
+. Leetmaa,* R. Reynolds,* Roy 'enne,, and Dennis 'ose*h,
ABSTRACT
The NCEP and NCAR are cooperating in a project (denoted "reanalysis") to produce a !"year record o# glo$al
analyses o# at%ospheric #ields in support o# the needs o# the research and cli%ate %onitoring co%%unities. This e##ort
in&ol&es the reco&ery o# land sur#ace' ship' ra(insonde' pi$al' aircra#t' satellite' and other data) *uality controlling
and assi%ilating these data (ith a data assi%ilation syste% that is +ept unchanged o&er the reanalysis period ,-./"-0.
This eli%inates percei&ed cli%ate ju%ps associated (ith changes in the data assi%ilation syste%.
The NCEP1NCAR !"yr reanalysis uses a #ro2en state"o#"the"art glo$al data assi%ilation syste% and a data$ase as
co%plete as possi$le. The data assi%ilation and the %odel used are identical to the glo$al syste% i%ple%ented opera
tionally at the NCEP on ,, January ,--.' e3cept that the hori2ontal resolution is T04 (a$out 4,! +%). The data$ase
has $een enhanced (ith %any sources o# o$ser&ations not a&aila$le in real ti%e #or operations' pro&ided $y di##erent
coun tries and organi2ations. The syste% has $een designed (ith ad&anced *uality control and %onitoring
co%ponents' and can produce , %on o# reanalysis per day on a Cray YMP/8 superco%puter. 5i##erent types o# output
archi&es are $eing created to satis#y di##erent user needs' including a "*uic+ loo+" C5"R67 (one per year) (ith si3
tropospheric and stratospheric #ields a&aila$le t(ice daily' as (ell as sur#ace' top"o#"the"at%osphere' and isentropic
#ields. Reanalysis in#or%ation and selected output is also a&aila$le on"line &ia the 8nternet
(httpll9nic.#$.noaa.go&9:!!!). A special C5 R67' containing ,; years o# selected o$ser&ed' daily' %onthly' and
cli%atological data #ro% the NCEP1NCAR Re analysis' is included (ith this issue. 6utput &aria$les are classi#ied
into #our classes' depending on the degree to (hich they are in#luenced $y the o$ser&ations and1or the %odel. <or
e3a%ple' "C" &aria$les (such as precipitation and sur#ace #lu3es) are co%pletely deter%ined $y the %odel during the
data assi%ilation and should $e used (ith caution. Ne&er theless' a co%parison o# these &aria$les (ith o$ser&ations
and (ith se&eral cli%atologies sho(s that they generally contain considera$le use#ul in#or%ation. Eight"day
#orecasts' produced e&ery . days' should $e use#ul #or predicta$il ity studies and #or %onitoring the *uality o# the
o$ser&ing syste%s.
The ! years o# reanalysis (,-./"-0) should $e co%pleted in early ,--/. A continuation into the #uture through an
identical Cli%ate 5ata Assi%ilation =yste% (ill allo( researchers to relia$ly co%pare recent ano%alies (ith those in
earlier decades. =ince changes in the o$ser&ing syste%s (ill ine&ita$ly produce percei&ed changes in the cli%ate' par
allel reanalyses (at least , year long) (ill $e generated #or the periods i%%ediately a#ter the introduction o# ne( o$
ser&ing syste%s' such as ne( types o# satellite data.
NCEP plans currently call #or an updated reanalysis using a state"o#"the"art syste% e&ery #i&e years or so. The suc
cessi&e reanalyses (ill $e greatly #acilitated $y the generation o# the co%prehensi&e data$ase in the present
reanalysis.
>En&iron%ental 7odeling Center' National Centers #or
En&iron
%ental Prediction' ?ashington' 5.C.
"Cli%ate Prediction Center' National Centers #or En&iron%ental
Prediction' ?ashington' 5.C.
"National Center #or At%ospheric Research' @oulder' Colorado
Correspondingauthor address: 5r. Eugenia Kalnay' director'
En &iron%ental 7odeling Center' W/NMCZ, Roo% 4!
??@' ?ashington' 5C 4!4;;.
E"%ail9 ? 5 4;E K A = B N ,. ? ? @ .N6AA .C6D
8n #inal #or% . =epte%$er ,--..
E,--0 A%erican 7eteorological =ociety
1. Intro!ction
The National Centers #or En&iron%ental Prediction
FNCEP' #or%erly +no(n as the National 7eteorologi
cal Center (N7C)G8National Center #or At%ospheric
Research (NCAR)H Reanalysis Project $egan in ,--,
as an outgro(th o# the N7C Cli%ate 5ata Assi%ila"
,A list o# acrony%s used in this paper is included in appendi3 5.
Bulletin of the American Meteorological Society 437
tion =yste% (C5A=) project. The %oti&ation #or the
C5A= project (as the apparent "cli%ate changes" that 302
resulted #ro% %any changes introduced in the N7C
operational Clo$al 5ata Assi%ilation =yste% (C5A=)
301
o&er the last decade in order to i%pro&e the #orecasts.
These ju%ps in the percei&ed cli%ate para%eters o$
300
scure' to so%e e3tent' the signal o# true short"ter%
Tv 1000m b Pacific30N~30S 120E~90W
cli%ate changes or interannual cli%ate &aria$ility. An
o$&ious e3a%ple is presented in <ig. ,' (hich sho(s
large ju%ps in the analy2ed &irtual te%perature at
,!!! hPa in the Paci#ic 6cean (hen the %odel (as
changed. The i%pact o# syste% changes on other pa
ra%eters' such as esti%ated precipitation and its dis
tri$ution' is %ore su$tle and there#ore harder to
299
298
297
1980 1982 1984 1986 1988 1990 1992 1994
Iear
separate #ro% the true cli%ate ano%aly signals.
The $asic idea o# the reanalysis project is to use a
#ro2en state"o#"the"art analysis1#orecast syste% and
per#or% data assi%ilation using past data' #ro% ,-./
to the present (reanalysis). 7oreo&er' the sa%e #ro
2en analysis1#orecast syste% (ill $e used to continue
to per#or% data assi%ilation into the #uture (C5A=)
so that cli%ate researchers can assess (hether current
cli%ate ano%alies are signi#icant (hen co%pared to
a long reanalysis (ithout changes in the data assi%i
lation syste%. 8n addition' there (ill $e a one"(ay
coupled ocean reanalysis' in (hich the sur#ace #lu3es
#ro% the at%ospheric %odel (ill $e used #or the ocean
data assi%ilation. The NCEP8NCAR !"year reanaly
sis should $e a research *uality dataset suita$le #or
%any uses' including (eather and short"ter% cli%ate
research.
The project de&elop%ent has $een supported $y the
National 6ceanic and At%ospheric Ad%inistrationJs
(N6AA) 6##ice o# Clo$al Progra%s. An ad&isory
panel chaired $y J. Nogues"Paegle guided it through
out the de&elop%ental period (,-:-"-;). A#ter the
e3ecution phase started in ,--' the ad&isory panel
(as replaced $y a usersJ ad&isory co%%ittee' chaired
$y +. 6ort. The reanalysis syste% (as designed at
NCEP' (ith the participation o# o&er 4. scientists
#ro% NCEPJs En&iron%ental 7odeling Center' Cli
%ate Prediction Center (CPC)' the Coupled 7odel
Project' and Central 6perations. =cientists at NCAR
per#or%ed %ost o# the data collection and o$tained
%any special datasets #ro% international sources that
(ere not a&aila$le operationally through the Clo$al
Teleco%%unications =yste% (CT=). E. Kung (Bni
&ersity o# 7issouri) ac*uired early data #ro% China.
?e also had the colla$oration o#N6AAlNational En
&iron%ental =atellite' 5ata and 8n#or%ation =yste%
(NE=58=)' (ho pro&ided the T8R6= (Tele&ision 8n"
438
<8C. ,. Trace o# the l666"hPa&irtual te%perature a&eraged
#or the tropical Paci#ic ocean in the NCEP operational Clo$al
5ata Assi%ilation =yste% (solid line)' sho(ing the i%pact o#
changes in the %odel and in the reanalysis (dotted line).
#rared 6$ser&ation =atellite) 6perational Dertical
=ounder (T6D=) data) the Ceophysical <luid
5yna% ics Ka$oratory (C<5K) #or the ocean
reanalysis) the Bnited Kingdo% 7eteorological
6##ice (BK76)' (ho (ill supply their glo$al ice
and ==T reanalyses (C8==T) #or the earlier periods)
the Japanese 7eteo rological Agency (J7A)' (ho
pro&ided cloud"trac+ (inds and special ra(insonde
data not a&aila$le on CT=) and the European
Centre #or 7ediu%"Range ?eather <orecasts
(EC7?<)' (ho #illed so%e data gaps and pro&ided
a sea ice data$ase. The National Aeronautics and
=pace Ad%inistration1Coddard Ka$oratory #or
At%ospheres (NA=A1CKA) has pro &ided retrie&als
%issing #ro% the NCEP archi&es and o##ered to
per#or% T6D= retrie&als #or se&eral %onths o#
%issing data in ,-:0. The N6AAlEn&iron%ental
Research Ka$oratory (ERK)1Cli%ate 5iagnostic
Cen ter has pro&ided #unding #or archi&al and tape
han dling de&elop%ent and support.
The early design o# the project (as discussed in
an NMCINCAR reanalysis (or+shop that too+ place
at N7C in April ,--, (Kalnay and Jenne ,--,).
The (or+shop had the participation o#
representati&es o# all the groups planning to per#or%
reanalyses at that ti%e FCenter #or 6cean"Kand"
At%osphere 8nterac tions (C6KA)' EC7?<' and
NA=A1CKAG' as (ell as o# the %ajor types o# users
(e.g.' #or short" and long ter% dyna%ics and
diagnostics' transport o# trace gases' cli%ate
change' predicta$ility' angular %o%en tu% and
length o# day' coupled %odels' etc.). The near"
#inal design (as re&ie(ed in 6cto$er ,--; $y the
ad&isory co%%ittee' (ho suggested se&eral addi"
Vol. //' o.!" March #$$%
tional tests and %odi#ications $e#ore the start o# the
operational phase (started in 7ay ,--). Representa
ti&es o# the %ajor agencies interested in the project
FN6AA' National =cience <oundation (N=<)' NA=A'
5epart%ent o# Energy (56E)G and o# the other groups
per#or%ing reanalyses also participated in the re&ie(
o# the NCEP/NCAR project. The other plans pre
sented in the 6cto$er ,--; re&ie( (ere those o#
EC7?<' (hich is per#or%ing a ,."yr reanalysis
#or
,-/-"-) NA=A8CKA (=chu$ert et al. ,--;)'
(hich
per#or%ed a reanalysis #or ,-:."-!' the B.=. Na&y
(,-:. to the present)) and C6KA' (hich per#or%ed
an 8:"%on reanalysis #or the #$8&/8! E8 Nino. =uch
%ultiple reanalysis projects o##er a great opportunity
#or cooperation and interco%parison' (hich should
enhance each o# the projects. 8n particular' the
NCEP has $ene#itted #ro% the C6KA project
through the trans#er o# the Cridded Analysis and
5isplay =yste% (CrA5=)' (hich has greatly
enhanced the NCEPJs de&elop%ental graphical
display syste%' and #ro% the close interaction (ith
NA=A and EC7?< scientists per#or%ing a si%ilar
reanalysis.
The purpose o# this paper is to update the
docu%en tation o# the NCEP8NCAR syste% design'
output' and plans #or distri$ution. The $asic
characteristics o# the syste% are su%%ari2ed in
section 4' and the data to $e used in section ;. The
three %odules o# the reanaly sis syste% (data *uality
control preprocessor' analy sis %odule (ith
auto%atic %onitoring syste%' and output %odule)
are descri$ed in sections '.' and 0' respecti&ely.
The C5A=' (hich uses the sa%e #ro2en syste% $ut
continues the analysis into the #uture' is discussed
in section /. =ection : su%%ari2es the cou pling
(ith the ocean reanalysis. =ection - contains an
assess%ent o# the relia$ility o# the reanalysis output
and the i%pact o# changes in the o$ser&ing syste%.
=ection ,! su%%ari2es the project. 7ore detailed
docu%entation is a&aila$le #ro% the NCEP (Kalnay
et al. ,--;).
". #$er$ie% o& the
NCEP/NCAR
reanalysis syste' an e(ec!tion )lan
The NCEP/NCAR Reanalysis Project has t(o
uni*ue characteristics9 the length o# the period co&
ered and the asse%$ly o# a &ery co%prehensi&e o$
ser&ational data$ase. The reanalysis (ill co&er the
!"year period ,-./"-0 and (ill continue into the
#u ture (ith the C5A=. The o$ser&ations (ill $e
sa&ed in the ?orld 7eteorological 6rgani2ationJs
(?76)
Bulletin of the American Meteorological Society
$inary uni&ersal #or%at representation (@B<R)'
(ith additional in#or%ation' such as the #irst"guess
and *uality control decisions' incorporated into the
report. ?e are also considering the #easi$ility o#
e3tending the reanalysis$ac+ to ,-0' (hen the
Northern Le%i sphere (NL) upper"air net(or+ (as
esta$lished' as suggested $y se&eral researchers.
The length o# the reanalysis' and the desire to carry
it out as *uic+ly as possi$le to increase its
use#ulness' led us to design a syste% a$le to
per#or% one %onth o# reanalysis per day. =uch a
#ast pace o# e3ecution re*uired the de&el op%ent o#
a reanalysis syste% %uch %ore ro$ust and
auto%ated than the analysis"#orecast syste%s used
#or operational nu%erical (eather prediction. As a
result' the NCEP8NCAR reanalysis syste% has
%any no&el #eatures not yet present in operational
or research nu%erical (eather #orecasting syste%s.
As sho(n in <ig. 4' the NCEP8NCAR reanalysis
syste% has three %ajor %odules9 data decoder and
*uality control (MC) preprocessor' data assi%ilation
%odule (ith an auto%atic %onitoring syste%' and
archi&e %odule. The central %odule is the data
assi%i lation' (hich has the #ollo(ing
characteristics9
N T04 %odel (e*ui&alent to a hori2ontal resolution
o# a$out 4,! +%) (ith 4: &ertical le&els. The
%odel is identical to the NCEP glo$al operational
%odel i%ple%ented on ,! January ,--.'
e3cept #or the hori2ontal resolution' (hich is
T,40 (,!. +%) #or the operational %odel
(Kana%itsu ,-:-) Kana%itsu et al. ,--,).
N =pectral statistical interpolation (==8' or ;5
&aria tional) analysis' (ith no need #or nonlinear
nor%al %ode initiali2ation (Parrish and 5er$er
,--4) 5er$er et al. ,--,)) i%pro&ed error
statistics) and the $alance constraint on the ti%e
deri&ati&e o# the di&ergence e*uation
i%ple%ented at NCEP in January ,--. are also
included (5er$er et al. ,--).
N Co%ple3 MC o# ra(insonde data' including ti%e
interpolation chec+s' (ith con#idence corrections
o# heights and te%peratures (Collins and Candin
,--!' ,--4)) 6pti%al interpolation (6l)"$ased
co%ple3 MC o# all other data (?oollen ,--,)
?oollen et al. ,--).
N 6pti%al a&eraging o# se&eral para%eters o&er a
nu%$er o# areas' pro&iding %ore accurate
a&erages and esti%ates o# the error o# the
a&erage (Candin
,--;).
N 6pti%al interpolation ==T reanalysis (Reynolds
and =%ith ,--) starting #ro% ,-:4 and BK76
C8==T #or earlier periods.
439
.
Ke&el 88 data
OOll,Pi8OOHOOOO9O"GOH,,
Ko(er @oundary <ields
7onitoring
6ccasional
*uality control preprocessor and an
analysis output MC %onitoring %od
ule (<ig. 4)9 The data input is prepro
cessed' and all the analysis output
#ields are %onitored (ith a "co%ple3
MC" %onitoring syste%' in (hich the
statistics o# the data' ti%e tendencies'
etc. are co%pared to cli%atological
statistics in order to detect errors.
These statistics include tendency
chec+s. (These %onitoring syste%s
8nter&ention

(ill also $e i%ple%ented operation
ally' and their de&elop%ent consti
tutes a %ajor spin"o## #ro% this project
#or NCEP.)
It (as decided early in the project
7onitoring
6ccasional
8nter&ention
that one type o# output could not sat
is#y the needs o# the %any di##erent
types o# users. <or this reason the out
put %odule allo(s #or se&eral di##er
ent archi&es9
?ash. 5C
Ash&ille
@oulder
5ata
5istri$ution
N Ke&el"4 o$ser&ational data in
. @B<R' including MC' cli%atologi
cal' analysis' and 0"h #orecast in
#or%ation.
N Co%prehensi&e analysis' #irst
guess' and diagnostic #ields pre
sented in "synoptic" #or% (all
#ields e&ery 0 h) in the %odel sig%a
coordinates' as (ell as in pressure
<8C. 4. =che%atic o# the %ain co%ponents o# the NCEP1NCAR reanalysis
syste%
and their state o# readiness on January ,--.. "Bnder(ay" %eans that the
co%ponent is (or+ing $ut it still $eing i%pro&ed. ("N7C" has changed to
"NCEP.")
and isentropic coordinates' in
gridded $inary (CR8@) #or%at. A
restart #ile is included once a %onth
to allo( #or rerunning shorter pe
riods (ith enhanced diagnostics.
N 6ne"(ay coupled ocean %odel5 assi%ilation #or
,-:4 on(ard (Ji et al. ,--).
N The sa%e C5A= (ill $e used into the #uture' #ro%
January ,--. on(ard.
To support a rate o# reanalysis o# a$out one %onth
per day' it is necessary to ensure that the data input is
generally #ree o# %ajor data pro$le%s such as (rong
dates' (rong locations' gar$led in#or%ation' etc.' #or
$oth con&entional and re%otely sensed data. This is
particularly i%portant #or old data' (hich ha&e not
. $een pre&iously used at NCEP. =i%ilarly' the rate o#
one %onth o# reanalysis per day does not allo( #or
the detailed hu%an scrutiny that operational output
nor%ally recei&es. <or this reason' (e created a data
440
N A ti%e series archi&e in (hich each #ield is a&ail
a$le #or all ti%es' including standard pressure le&el
#ields' precipitation' sur#ace #lu3es' and other
(idely used diagnostic #ields. This #or%at (ill $e
the %ost use#ul #or %any users.
N A "*uic+ loo+" archi&e on C5"R67' one per year'
including the %ost (idely used #ields9 daily &alues
o# &aria$les at selected tropospheric and strato
spheric pressure le&els' sur#ace and top"o#"the"at
%osphere #lu3es' precipitation' %onthly and 2onal
a&erages o# %ost #ields' co&ariances' isentropic
le&el &aria$les' etc. A special C5"R67' contain
ing ,; years o# selected o$ser&ed' daily' %onthly'
and cli%atological data #ro% the NCEP1NCAR Re
analysis' is included (ith this issue (appendi3 E).
Vol. //' o.!" March #$$%
N Eight"day #orecasts per#or%ed e&ery . days' (hich
should allo( predicta$ility studies and esti%ates o#
the i%pact o# inho%ogeneities in the o$ser&ing
syste%s co&erage' (ith ano%aly correlation scores.
N A su$set o# the output is posted on the NCEP pu$
lic ser&er and is a&aila$le through anony%ous <TP.
N NCAR' National Cli%ate 5ata Center (NC5C)'
and CPC (ill distri$ute the $ul+ o# the reanalysis
data.
N Reanalysis in#or%ation and selected output is
also a&aila$le on"line &ia the 8nternet (http11
9nic.#$.noaa.go&9:!!!).
An i%portant *uestion that has repeatedly arisen is
ho( to handle the ine&ita$le changes in the o$ser&
ing syste%' especially the a&aila$ility o# ne( satel
lite data' (hich (ill undou$tedly ha&e an i%pact on
the percei&ed cli%ate o# the reanalysis. @asically' the
choices are (a) to select a su$set o# the o$ser&ations
that re%ains sta$le throughout the !"yr period o# the
reanalysis or ($) to use all the a&aila$le data at a gi&en
ti%e. Choice (a) (ould lead to a reanalysis that has
the %ost sta$le cli%ate' and choice ($) to an analysis
that is as accurate as possi$le throughout the ! years.
?ith the guidance o# the ad&isory panel' (e ha&e
chosen ($)' that is' to %a+e use o# the %ost data a&ail
a$le at any ti%e. Lo(e&er' in order to assess the i%
pact o# the introduction o# ne( o$ser&ing syste%s on
the percei&ed cli%ate o# the reanalysis' (e ha&e de
cided to produce a parallel reanalysis' at least , year
long' (ithout using a large ne( o$ser&ing syste%.
This (ill allo( the users to assess the e3tent to (hich
the ne( o$ser&ing syste% in#luences the percei&ed
cli%ate and the annual cycle.
The e3ecution phase started in 7ay ,-- on the
Cray I7P : superco%puter pro&ided $y the NCEP
#or this project. A$out 4 h o# the CRA I I7P (4"/
processors) are needed in order to per#or% one %onth
o# reanalysis and #orecasts per day. @y =epte%$er
,--.' ,; years (,-:4"-) should $e co%pleted (in
addition to se&eral years o# reruns per#or%ed to as
sess the i%pact o# changes in o$ser&ing syste%s' to
correct pro$le%s disco&ered in the data$ase' etc.).
Ne3t' the period ,-/-":4 (ill $e reanaly2ed and co%
pleted around the end o# ,--.' #ollo(ed $y the
,-./"/: decades. ?e e3pect to co%plete the ! years
o# reanalysis (,-./"-0) $y early ,--/. The e3tension
into ,-:"./' i# #easi$le' (ould $e done during ,--/.
This #irst phase o# reanalysis (ill $e #ollo(ed $y a
second phase in (hich a ,--: state"o#"the"art syste%
(ill $e used #or a second reanalysis. NCEP plans cur
rently call #or an updated reanalysis e&ery . years or
Bulletin o- the American Meteorological
Society
so. The successi&e reanalyses (ill $e %ade easier
$y the a&aila$ility o# the co%prehensi&e data$ase
in @B<R generated $y the present reanalysis.
*. The )re)aration o& ata &or reanalysis
The data collection is a %ajor tas+ that has $een
per#or%ed %ostly at NCAR. =ur#ace and upper"air
o$ser&ationsare $eing prepared #or the
reanalysis.The plan is to use the data a&aila$le #or
the original op erational NCEP analyses (a&aila$le
#ro% 7arch ,-04 on) and to add other datasets to
capture the older data #ro% a$out ,-: on.
Additional data inputs #or ,-04 on (ill pro&ide
%uch %ore data than (as #irst a&ail a$le
operationally and (ill $e %erged and #or%atted in
@B<R at NCEP. The co%ponent datasets are listed
$elo(. <or #urther details consult Kalnay et al.
(,--;) and the e3tended te3ts pro&ided $y NCAR
(see list o# te3ts $elo().
a. 'lo(al ra)insonde
data
NCAR has tapes o# the NCEP CT= data (ith up
per"air o$ser&ations #ro% 7arch ,-04 on' (hich
(ill $e the %ain data source #or reanalysis. ?e
plan to pro&ide $oth the CT= data ((hich also has
pi$als and aircra#t) and also rao$s #ro% national
archi&es in &ari ous countries. NCAR has rao$s
recei&ed directly #ro% so%e countries such as
=outh A#rica' Australia' Canada' Argentina'
@ra2il' the Bnited Kingdo%' <rance' and #ro% the
Bnited =tates (NC5C). The B.=. Air <orce (B=A<)
prepared a glo$al collection o# data (T5.) that is
%ostly #or the period ,-:"/!' (hich (ill $e
included. C<5K is helping (ith processing and
chec+ing this set' (hich (ill all $e ready #or the #irst
reanalysis. The Bni&ersity o# 7issouri (E. Kung) is
colla$orating (ith so%e o# the chec+ing $et(een
di# #erent sources o# the sa%e data and has the
o$tained daily upper"air data o# ;! stations o&er
China #ro% the Chinese =tate 7eteorological
Ad%inistration #or the period o# ,-."04. Bnder the
Bnited =tates"Rus sia $ilateral e3change e##ort led
$y R. Jenne (NCAR) and =. =hu%$era (NC5C)' the
Bnited =tates has re cei&ed 4! %agnetic tapes (ith
upper"air data #ro% ./
B==R stations #or ,-0,"/:. The Japanese 7eteoro
logical Agency has pro&ided NCEP (ith additional
data not a&aila$le o&er CT=.
5uring the reanalysis' it (as #ound that the count
o# signi#icant le&el (inds (as lo( #ro% August
,-:- to =epte%$er ,--,in the NCEP tapes $ut not
o&er the Bnited =tates and China. EC7?< supplied
their data
441
to #ill the gap.
8nterestingly' the
EC7?< had a si%i lar'
$ut co%ple%entary'
lo( count o&er the
Bnited =tates and China.
NCAR' NC5C'
Russia' Europe' and
other organi 2ations'
including ?76' ha&e
interests in i%pro&ing the
glo$al archi&e o#
ra(insonde data. ?e
anticipate &arious
colla$orations to
i%pro&e the $asic input
sets and to acco%plish
%erges. Lo(e&er' their
results (ill
$e a&aila$le #or later
reanalyses' not #or the
#irst one.
N
=8R=a8R
sounders
N
=8R=
on early
N7C
tapes
(not
radian
ces)
N
DTPRa8
R
sounders
N T6D=
sounders
(L8R
='
7=B'
ssur
N L8R=
data test
syste%
Apr
,-0-"
Apr
,-/,
No&
,-0-"
=ept
,--4
No&
,-/4"
<e$
,-/-
No&
,-/:"
prese
nt
Aug
,-/."
7ar
,-/0
h.
C*
A+
S
sur
fac
e
ma
rin
e
dat
a
The Co%prehensi&e
6cean"At%osphere 5ata
=et (C6A5=)'4 #irst
released in ,-:; and
recently up dated'
includes ships' #i3ed
$uoys' dri#ting $uoys'
pac+"ice $uoys' near"
sur#ace data #ro% ocean
station reports (Q@Ts'
etc.)' and other data.
An update #or
,-:!"-; has $een
co%pleted' and (or+ is
progress
ing on all the
sur#ace %arine
data #or ,-/"/-.
c
.

A
i
r
c
r
a
f
t

d
a
t
a
Aircra#t data is
a&aila$le #ro% the
NCEP CT= source
starting in 7arch ,-04.
Additional data ha&e
$een gathered #ro%
se&eral sources'
including data #ro% Ne(
Realand #or <e$ruary
,-:"June ,-::' so%e
o# (hich did not get into
CT=. Aircra#t data #ro%
e3peri%ents such as the
CARP (Clo$al At%o
spheric Research
Progra%) Atlantic
Tropical E3peri %ent
(CATE' su%%er ,-/)
and the <irst CARP
Clo$al E3peri%ent
(<CCE) (,-/-) (ill $e
used. =e lected B=A<
reconnaissance data is
a&aila$le' start ing #ro%
,-/. 5ata #ro%
Tropical ?ind Energy
Con&ersion Re#erence
Ke&el E*uip%ent
(T?ERKE) constant"
pressure $alloons #or
the =outhern Le%i
sphere (=L' July ,-/."
August ,-/0) (ill $e in
the dataset. These
$alloons pro&ide data
si%ilar to a single"le&el
ra(insonde near ,.!
%$.
d
.
S
u
r
f
a
c
e
l
a
n
d
s
y
n
o
p
t
i
c
d
a
t
a
Clo$al CT= data
(usually e&ery ; h) are
a&aila$le starting #ro%
,-0/ #ro% air #orce or
NCEP sources. Earlier
years are a&aila$le #ro%
the air #orce tape dec+
,; and #ro% B.=. hourly
data (#ro% NC5C). The
data co&erage is #airly
good #ro% ,-- on.
e
.

S
a
t
e
l
l
i
t
e

s
o
u
n
d
e
r

d
a
t
a
The $asic radiances
are a&aila$le #or the
#ollo(ing periods9
4C6A5= is a joint project o#
N6AAJERK' NCAR' and
NC5C. 7any other
organi2ations and countries
ha&e also contri$uted to its
creation.
J
=
e
e

a
p
p
e
n
d
i
3

5
.
4
4
2
8n the #irst phase o#
the reanalysis (e plan
to use the original
operational T6D=
retrie&als o# NE=58=
(4..S space resolution).
A syste% $ased on the
;5 &ariational
assi%ilation o#
&ariances (ill $e used
in the second phase o#
the reanalysis (to start
in ,--:). It should $e
noted that the pilot
e3peri%ents co%par ing
reanalysis (ith and
(ithout the use o#
satellite data' to $e
discussed in section -'
ha&e pro&ided use #ul
in#or%ation regarding
the uncertainties o# the
analysis (ithout satellite
data. This is &ery
i%portant #or the period
$e#ore ,-/- (hen no
T6D= satellite
soundings (ere
a&aila$le. ?e hope to
assi%ilate DTPR and
L8R= data a&aila$le
$e#ore ,-/- #or the
=outhern Le%isphere'
although (e ha&e no
recent e3perience (ith
that data' and there %ay
$e un#ore seen
pro$le%s. .
f
S
S
M
,
,
s
u
r
f
a
c
e
)
i
n
d

s
p
e
e
d
s
=pecial =ensing
7icro(a&ell%ager
(==788) data
$eca%e a&aila$le in July
,-:/' and at NCEP the
as
si%ilation o# sur#ace
(inds $eca%e
operational on ,!
July ,--;. ?e adopted
the neural net(or+
algorith% o#
Krasnopols+y et al.
(,--.)' (hich results in
(ind speeds
signi#icantly closer to
$uoys (ind speeds and
(ith $etter co&erage
under cloudy conditions
than the present
operational algorith%
used at NCEP. ?e ini
tially used a su$set o#
the high"resolution
SSM/- ra diance data
archi&ed $y NE=58= #or
cli%ate purposes.
Lo(e&er' a#ter o&er
years (ere reanaly2ed'
se&eral pro$le%s (ere
#ound that indicated that
it (ould $e necessary to
use the original dataset.
The high &olu%e o#
these data (%uch larger
than all other data
together) also resulted in
a signi#icant slo(do(n
o# the reanaly sis. <or
this reason it (as
decided that the #irst
phase o# the reanalysis
(ill not include (ind
speeds #ro% ==788
(e3cept #or li%ited data
i%pact studies). ?e plan
to use the ==788 (ind
speed' as (ell as total
pre cipita$le (ater and
other para%eters' in the
second phase o#
reanalysis.
g
.

S
a
t
e
l
l
i
t
e

c
l
o
u
d

d
r
i
f
t

)
i
n
d
s
=atellite cloud dri#t
(inds are used #ro%
the origi"
V
o
l.
/
/'
N
o.
;'
M
a
r
c
h
#
$
$
%
nal N7C tapes and #ro% the Ceostationary 7eteoro
logical =atellite (C7=) cloud dri#t (inds recei&ed
#ro% the J7A #or the period ,-/:"-,.
A te3t entitled "5ata #or Reanalysis9 8n&entoriesJT
has &arious %aps and displays that illustrate the typi
cal co&erage o# sur#ace and upper"air data that are
already a&aila$le. 7ost o# this in#or%ation co&ers the
period #ro% a$out ,-: on. The co&erage o# data is
rather encouraging' e&en #or the earlier years. ?e
note' ho(e&er' that ra(insonde o$ser&ing net(or+s
#or Antarctica and the (est coast o# =outh A%erica
did not start until July ,-./. 7any other reports ha&e
$een prepared that gi&e %ore in#or%ation a$out the
attri$utes o# di##erent datasets and the status o#
projects to prepare the data. Papers ha&e $een pre
pared that #ocus on di##erent issues) a selection o# these
papers is gi&en $elo(. Additional papers are a&aila$le
upon re*uest.
=election o# te3ts a$out data #or reanalysis (con
tact NCAR #or #urther in#or%ation)9
Text Date
5ata #or reanalysis9 8n&entories No& ,--4
=ea sur#ace te%perature data , <e$ ,--;
=ea ice data 4 Apr ,--;
Ra(insonde data #or reanalysis 4 6ct ,--
5ataset o# tropical stor% locations 40 Jan ,--;
N7C upper"air data9 ,-04"/4 4- 7ar ,--
Clo$al satellite sounder data ,4 Aug ,--
=ur#ace land synoptic data 7ay ,--
8ce cap $uoy update . Apr ,--
8n&entories o# data #or reanalysis 7ar ,--.
Analyses #or the =L #ro% ,-., on ,: 7ar
,--;
=tatus o# reanalysis data , Apr ,--;
4. +ata )re)rocessor
The purpose o# the preprocessing reanalysis %od
ule (<ig. 4) is to re#or%at the data co%ing #ro%
%any di##erent sources (<ig. ;) into a uni#or%
@B<R #or"
"Darious docu%ents descri$ing this dataset are a&aila$le #ro% R.
Jenne' NCAR' P.6. @o3 ;!!!' @oulder' Co :!;!/";!!!.
Bulletin o- the American Meteorological
Society
.EM/
+IRCR+0. 1C+Ri1MC 'M+lCM+ ,,+IlI1L
L+1D S0C 1C+RI1...,. 1C+Rl1MC2 3C)))))C4
M+RI1E NCAlJt1(l6A5U
<8C.;. =che%atic o# the data input archi&e structure #or the
reanalysis. The data are classi#ied into eight $asic types
(ra(insondes' aircra#t' land sur#ace' %arine' sur#ace $ogus'
satellite te%perature soundings' cloud"trac+ed (inds' and
==718). The PREP@B<R archi&e adds "e&ents" on each datu%
as it #lo(s through the reanalysis.
%at and to preprocess , or %ore years at a ti%e' $e
#ore the actual reanalysis %odule is e3ecuted at the
rate o# one %onth per day. This allo(s detection o#
%ajor data pro$le%s (ith su##icient lead ti%e (a #e(
days $e#ore the e3ecution o# the reanalysis)' so that
hu%an %onitors can try to ta+e correcti&e action.
The preprocessor thus %ini%i2es the need #or
reanalysis reruns due to the %any data pro$le%s that
#re*uently appear' such as data (ith (rong dates'
satellite data (ith (rong longitudes' etc. The
preprocessor also in cludes preparation o# the
sur#ace $oundary conditions (==T' sea ice' etc.).
a. Satellite data
A special satellite T6D= soundings data %onitor
ing syste% has $een de&eloped. It is intended as a
*uality control o# the data in the NE=58= archi&e
tapes that can su##er #ro% errors in dates and or$its
not li+ely to occur in the daily operational products.
=atellite data in grid $o3es o# ,!S 3 ,!S' as (ell as
single sat ellite o$ser&ations' are *uality controlled.
The a&er age in the $o3' the &ariance in the $o3'
and the a$solute &alue o# tendency o# the $o3
a&erage are co%pared (ith a cli%atologyto #lag
suspiciousgroups o# satellite data.
h. C.C )ith temporal
chec/
The co%ple3 *uality control (CMC) syste% (de
scri$ed in the ne3t section) is included in the
prepro"
443
cessor $ut (ithout the use o# the #irst guess o# the
%odel. The $aseline chec+ in the preprocessor (see
ne3t section) allo(s #or the detection o# changes in
the station locations' an i%portant pro$le% that inter
#eres (ith the accurate detection o# cli%ate change.
c. Climatological MC test of
data
The auto%atic %onitoring syste% de&eloped#or
the reanalysis output (section :) is $ased on
cli%atologi cal tests (ith three"di%ensional (grid
point) statistics co%puted #or each %onth. The
space"ti%e character o# the statistics pro&ed to $e
&ery success#ul in #ind ing pro$le%s in the pilot
e3peri%ents #or the reanaly sis (section :)' (hich
(ere then related to unusual data errors' leading also
to corrections o# se&eral errors present in the
operational syste%. This led us to chec+ the data
directly (ithin the preprocessor' $y e3press ing the
o$ser&ation ano%aly in units o# standard de &iations
(ith respect to cli%atology' a nu%$er (hich can $e
generated #ro% the @B<R "e&ents" archi&e (see
section .V). =uch a chec+ allo(s hu%an %onitors to
chec+ #or unusual data present in unusual a%ounts'
$e#ore the e3ecution o# the %onthly reanalysis' and
pro&ides the opti%al interpolation *uality control
(68MC) (ith additional in#or%ation that can $e used
$y its decision %a+ing algorith% (57A) as input to
the reanalysis.
d. Boundary fields
The #ollo(ing analyses and cli%atologies are
used #or the $oundary #ields9
,) ==T9 Reynolds reanalysis #or ,-:4 on' (hen
ADLRR data $eca%e a&aila$le' and the BK76
C8==T #or earlier periods (5. Par+er ,--4' per
sonal co%%unication).
4) =no( co&er9 NE=58= (ee+ly analyses and
cli%a
tology' updated (ee+ly (5. Carrett ,--.'personal
co%%unication))
;) =ea ice9 The ice #ield deri&ed #ro% =77Rl==78
data' and *uality controlled $y @. No%ura #or the
EC7?< reanalysis' has also $een adopted at
NCEP #or the period ,-/-"-;. @eyond ,--;' a
si%ilar algorith% de&eloped $y R. Cru%$ine is
$eing used. <or earlier periods (e plan to use the
analysis #ro% Joint 8ce Center analyses (hen
a&ail a$le' J. ?alsh and C8==T analyses
other(ise. These ha&e $een incorporated into the
==T analy sis so that all &alues $elo( "8.:SC are
considered sea ice. R. Cru%$ine has inserted a
%ore realistic glacial co&erage #or the Ross 8ce
=hel# and other
444
regions o# the Antarctic. A si%ple chec+
(co%pari son (ith %onthly cli%atologies and
standard de &iations) should help to ensure that
no %ajor er rors are present in the data or %ade
inad&ertently during their use.
) Al$edo9 7atthe(s (,-:.).
.) =oil (etness9 Bpdated during the analysis cycle.
The %odel uses the Pan and 7ahrt (,-:/) and
7ahrt and Pan (,-:) soil %odel. There is no
nudging o# the soil %oisture using concurrent
data' and a &ery s%all coe##icient (!.!.) is used
to nudge the soil %oisture to(ard cli%atology.
=oil %ois ture #ields sho( interannual
&aria$ility $ut no long"ter% dri#t (<ig. /).
0) Roughness length9 <ro% =i@.
/) Degetation resistance9 <ro% =i@ (5or%an and
=ellers ,-:-).
Lo(e&er' preli%inary reanalyses sho(ed that the
original resistance o&er regions dee%ed to $e
co&ered $y (inter (heat had e3cessi&ely high plant
resistance in the su%%er and #all' resulting in
te%peratures that (ere too high and lo(
precipitation in the eastern North A%erica su%%er
(=. =aha and L."8. Pan ,--' personal
co%%unication). <or this reason' (e are us ing the
%ini%u% %onthly resistance &alue #or each grid
point. 7onthly cli%atologies are the $ac+up o# the
analy2ed #ields (hen these updated #ields are not
a&aila$le.
,. +ata assi'ilation 'o!le
a. System
configuration
The C5A=1reanalysis is e3ecuted at the N6AA
Central Co%puter <acility in =uitland' 7aryland.
Bnli+e the operational NCEP syste%' (hich
currently is $ased on $oth 8@7"7D="type and
Cray"BN8Q co%puters' in the C5A=1reanalysis
syste% all pro cessing is done in the Cray"BN8Q
en&iron%ent. 6$ ser&ations (ill $e encoded in
@B<R' and gridded data in CR8@ the standard
?76 #or%ats. This syste% (ill soon $e also
adopted $y NCEP #or its nor%al opera tions.
The reanalysis (ill $e per#or%ed using the
present
Cray I7P : processors' ,4: 7? superco%puter'
and the s%aller Cray EK4. 6ther hard(are
includes a Ro$otic =ilo' upgraded in August ,--
(ith -!
=TK dri&es' (ith storage capacity o# !.0 C@ per
tape.
6&er 4!!! tape slots ha&e $een reser&ed #or this
project. =ince the Cray I7P (as saturated' the start
Vol. //' o.!" March /--0
*uired $y N7C to $e ins ta8led (early ,--) and #or
the operational syste%s to %igrate out o# the Cray
TA@KE
I.7odelle&els'%idle&elsig%a&a
geopotential thic+ness 5m6, and appro3i
%andatorypressurele&els(hPa).
lue'sig%athic+ness
'
%atelocation o# the
I7P (April ,--). =o#t(are used includes the
BN8C6= / operating syste%' N<= %ount o# Cray
co%ple3 #iles' @ourne shell BN8Q scripts' <ortran' 7odel 7idle&el 5elta
7andatory
pressure
so%e C' so%e Q ?indo(s' the 5ata 7igration <a"
le&el
sig%a sig%a Thic+ness le&el
o# the reanalysis had to (ait #or the ne( Cray C-! ac"
cility' the Cray Reel Ki$rarian' and the graphics sys"
tern CrA5= (C6KA). Recent changes include the
4: 4./; 0../
;.!
installation o# BN8C6= : and the replace%ent o# the
4/ ,!.!0 /.4- ..-- ,!.!
@ourne shell (ith the Korn (P6=8Q) shell. 8n addi"
40 ,:.; -.4;
;:4: 4!.!
tion' (e e3pect that the Cray I7P and Cray EK4 (ill
$e replaced in ,--. $y t(o Cray J-,0s. 4. 4:./. ,,.0! ;!.; ;!.!
4 ,./- ,.., 404,
(. Analysis scheme
4; 51W5 ,:.'!; 4;4 .!.!
The spectral statistical interpolation' a three"di%en"
sional &ariational analysis sche%e (Parrish and 5er$er
44 /:.,. 44.44 4,4 /!.!
,--4) 5er$er et al. ,--,)' is used as the analysis %od"
21 ,!4./: 4/.!- ,-: ,!!.!
ule. 8ts i%ple%entation in ,--, replacing an 78analy"
sis led to %ajor analysis and #orecast i%pro&e%ents'
4! ,;4.0, ;4.04 ,:.,
especially in the Tropics' and a %ajor reduction in the ,- ,0:.4; ;:.0/ ,/4- ,.!.!
precipitation spinup. An i%portant ad&antage o# the
,: 4,!.!0 ..!; ,0,4 4!!.!
==8 is that the $alance i%posed on the analysis is &alid
throughout the glo$e' thus %a+ing unnecessary the use
,/ 4'.:.4; .,.;. ,-. 4.!.!
o# nonlinear nor%al"%ode initiali2ation. Recent en"
,0 ;,4.: ./.,0 ,;/0 ;!!.!
hance%ents' such as i%pro&ed error statistics' and the
use o# the #ull tendency o# the di&ergence e*uation in
,. ;/4.!. 0,.-/ ,40! !!.!
the cost #unction (replacing the original linear $alance
, ;..0: 0..40 ,,;-
o# the incre%ents constraint)' ha&e also $een included
,; .!,.0: 00.0- lot/ .!!.!
(5er$er et al. ,--' Parrish et al. ,--.). The ==8 used
in the reanalysis is the sa%e as the syste% i%ple"
,4 .0:.!- 00.!0 :-.
%ented in the operational syste% in January ,--.' ,, 0;4'-! 0;./ 99:
(hich (as tested in parallel #or o&er ,! %onths and
,! 0-.40 .-.,- 00 /!!.!
resulted in signi#icantly i%pro&ed #orecasts.
- /.,,/0 .;./4 .0!
c. Model
: :!,.4 /.. 00
The T0414:"le&el NCEP glo$al spectral %odel is
used in the assi%ilation syste%' as i%ple%ented in the
/ :../- ,.,. ;: :.!.!
NCEP operational syste% in 5ece%$er ,--. The
0 ::;.: ;.-; ;,;
&ertical structure o# the %odel is sho(n in Ta$le ,.
. -,..-4 4-.,- 4.; -4..!
The %odel has #i&e le&els in the $oundary layer and
a$out se&en le&els a$o&e ,!! hPa. The lo(est %odel
-4... 4.!. 4!;
le&el is a$out . hPa #ro% the sur#ace' and the top le&el
; -0.;/ ,-..- ,04
is at a$out ; hPa. This &ertical structure (as chosen
so that the $oundary layer is reasona$ly (ell resol&ed
4 -:4.!: ,..:4 ,4-
and so that the stratospheric analysis at ,! hPa is not
--..!! ,!.!! :! ,!!!.!
greatly a##ected $y the top $oundary conditions. The
details o# the %odel dyna%ics and physics are de"
scri$ed in NOAA/NMC 5e&elop%ent 5i&ision %ajor physical processes' that is' con&ection' large"
(,-::)' Kana%itsu (,-:-)' and Kana%itsu et al. scale precipitation' shallo( con&ection' gra&ity (a&e
(,--,). The %odel includes para%eteri2ations o# all drag' radiation (ith diurnal cycle and interaction (ith
Bulletin o- the American Meteorological Society
44,
clouds' $oundary layer physics' an interacti&e sur#ace
hydrology' and &ertical and hori2ontal di##usion pro
cesses. A %ajor di##erence in the %odel as descri$ed
$y Kana%itsu et al. (,--,) is the use o# a si%pli#ied
Ara+a(a"=chu$ert con&ecti&e para%eteri2ation
sche%e de&eloped $y Pan and ?u (,--) $ased on
Crell (,--;). Prei%ple%entation e3peri%ents sho(ed
that the si%pli#ied Ara+a(a"=chu$ert sche%e results
in %uch $etter prediction o# precipitation than the
pre&ious Kuo sche%e o&er the continental Bnited
=tates' as %easured $y e*uita$le threat scores o&er
North A%erica. 8n addition' the precipitation patterns
o&er the Tropics are %ore realistic' (ith a s%oother
distri$ution and less concentration o&er tropical oro
graphic #eatures. T(o other recent i%pro&e%ents (ere
also i%ple%ented into the reanalysis %odel. The #irst
is a $etter diagnostic cloud sche%e (Ca%pana et al.
,--)' (hich has resulted in %odel"generated outgo
ing long(a&e radiation (6KR) in %uch $etter agree
%ent (ith o$ser&ations. The second is a ne( soil
%odel' $ased on Pan and 7ahrt (,-:/)' (hich has also
resulted in %uch %ore realistic sur#ace te%perature
and %ore s+ill#ul predictions o# precipitation o&er
North A%erica in the su%%er. These changes to the
%odel (ere syste%atically tested $y running 4 %onths
o# assi%ilations in su%%er and in (inter' and 4. #ore
casts #ro% each assi%ilation. =o%e tuning o# the
cloudiness and cloud optical properties (ere per
#or%ed to correct syste%atic te%perature and cloudi
ness errors. The #inal &ersion o# the %odel also
produced good ."day #orecast scores.
d. Comple0 1uality control of ra)insonde data
The %ethod o#CMC (Candin ,-::) is used to
*ual ity control the ra(insonde heights and
te%peratures. CMC #irst co%putes residuals #ro%
se&eral indepen dent chec+s (i.e.' it co%putes the
di##erence $et(een an o$ser&ation and the e3pected
&alue #or that o$ser &ation #ro% each chec+). It then
uses these residuals together (ith an ad&anced 57A
to accept' reject' or correct data (Collins and Candin
,--4). The chec+s included in the CMC code #or
ra(insonde heights and te%peratures used #or the
reanalysis include9 hydro static chec+' incre%ent
chec+ (ith respect to the 0"h #orecast' hori2ontal
interpolation chec+' and &ertical interpolation chec+.
8n addition' there is a $aseline chec+ $ased upon the
di##erence $et(een the station ele&ation and the
ele&ation that is consistent (ith the reported sur#ace
pressure and the lo(est t(o reported heights' using a
standard lapse rate and the hydrostatic e*uation.
Bsing the sa%e in#or%ation and assu%p"
446
tions' a %ean sea le&el pressure %ay $e o$tained
and co%pared (ith a #orecast %ean sea le&el
pressure. 8n this (ay' $oth an incre%ent and
hori2ontal residual o# %ean sea le&el pressure are
co%puted. The $ase line chec+ %ay allo( the
deter%ination o# errors in the location o# stations
as (ell as changes in their locations.
8n addition to these chec+s' used operationally at
NCEP' the reanalysis a##ords the possi$ility o# also
per#or%ing a te%poral interpolation chec+' (hich
cannot $e done in the N?P syste%. The &alue o# the
heights and te%peratures at o$ser&ation ti%e %ay $e
co%pared (ith those #or ,4 or 4 h earlier and later.
The te%poral residual is the di##erence $et(een the
reported height or te%perature and the &alue interpo
lated #ro% one &alue $e#ore and one a#ter' (hen they
are a&aila$le. =tatistics sho( this chec+ to $e o#
co% para$le &alue to the incre%ental chec+. It is
used along (ith other a&aila$le chec+s and is
particularly use#ul in the data preprocessor' (here
the #irst guess and' hence' the incre%ental chec+ are
not a&aila$le.
The CMC #or ra(insonde heights and
te%peratures per#or%s *uite (ell. The code has $een
running op erationally at NCEP #or se&eral years
and has under gone steady i%pro&e%ent. At present'
a$out /W o# the ra(insonde o$ser&ations are #ound
to ha&e at least one error. 6# the hydrostatically
detecta$le errors in %an datory le&el heights and
te%peratures' /.W are con #idently corrected' and
0!W o# the errors detected $y use o# the $aseline
chec+ are also corrected. The a$ solute nu%$er o#
corrections #or the early years o# reanalysis %ay $e
anticipated to $e s%aller' depend ing upon data
density' $ut there %ay $e a higher per centage o#
data that need to $e corrected.
The CMC %ethodology is also used to *uality
con trol the %andatory and signi#icant le&el
ra(insonde (inds. ?e e3pect to de&elop a li%ited
capa$ility to correct so%e (ind errors' #or e3a%ple'
(inds %anu ally entered o## $y %ultiples o# ,!!S in
direction or %ultiples o# ,!! +t in speed.
e. *ptimal interpolation 1uality control of all data
The 68MC (?oollen ,--,) ?oollen et al. ,--)
(as de&eloped as the #inal screening #or
o$ser&ations to $e used in the data assi%ilation. The
goal o# 68MC is to detect and (ithhold #ro% the
assi%ilation data containing gross errors generated
$y instru%ental' hu%an' or co%%unication"related
%ista+es that %ay occur during the process o#
%a+ing or trans%itting o$ser&ations. 8t also
(ithholds o$ser&ations (ith large errors o#
representati&eness that are accurate $ut
Vol. //' o.!" March /--0
(hose %easure%ents represent spatial and te%poral
scales i%possi$le to resol&e properly in the analysis
#orecast syste%. The 68MC syste% uses the sa%e sta
tistical representati&eness error %odel as the o$jecti&e
analysis syste% it precedes and' there#ore' (ill detect
o$ser&ations unrepresentati&e #or that syste%.
Three principles guide the 68MC algorith%9 86 use
o# %ulti&ariate three"di%ensional statistical interpo
lation #or o$taining co%parison &alues #or each o$
ser&ation #ro% near$y neigh$ors' 4) a co%ple3 o#
independent *uality control co%ponents consisting
o# interpolation and other types o# chec+s that
(hen e&aluated collecti&ely suggest (hether errors
e3ist in an o$ser&ation' and ;) "nonhierarchical"
decision %a+ing algorith% in (hich no #inal
accept1reject de cision is %ade #or any data until all
chec+s that %ay a##ect that decision are co%pleted.
The 68MC co%ponents are interpolationchec+s)
an
78 o# the appropriate &aria$le %ade to each datu%'
#ro% a group o# o$ser&ations near$y' #or%s a co%
parison &alue. Bni&ariate and geostrophic hori2ontal
chec+s are per#or%ed #or each datu% chec+ed' as
(ell as a uni&ariate &ertical (pro#ile) chec+ #or
sounding data. Te%perature data is con&erted to units
o# e*ui&a lent thic+ness di##erence #ro% a
$ac+ground (usually a 0"h #orecast) #or the chec+s'
(hile (ind data is chec+ed in ter%s o# &ector (ind
de&iations. A co% $ination o# indi&idual chec+
outco%es deter%ines (hether a datu% is accepted
$y the syste% (see Kalnay et a,. ,--;)?oollen
,--,) and ?oollen et a,.
,-- #or #urther details).
<or the reanalysis (e plan to add to the co%ple3
o# chec+s per#or%ed $y the 68MC t(o %ore *uanti
tati&e chec+s9 one is produced $y the ti%e interpola
tion chec+ o# the CMC (see pre&ious section)' and
the second is the de&iation (ith respect to
cli%atology' %easured in units o# the local analysis
standard de &iation cli%atology (see the discussion
in section c). @oth o# these should $e po(er#ul
additions to the MC and are %ade possi$le $y the
use o# the @B<R (ith a MC syste% #or storing data
(see section =#).
f B234 o(ser5ation 6e5ents6 files
The #inal step in o$ser&ation preprocessing' de
scri$ed in section ;' consolidates inco%ing data
#ro% all sources into @B<R #iles' an
internationally ac cepted standard #or%at #or le&el"4
data. Pro&isions ha&e $een %ade to archi&e in the
reanalysis @B<R #iles' along (ith each original
o$ser&ation (in a "push do(n" #ashion)' a spectru%
o# processing in#or%ation' collecti&ely +no(n as
"e&ents." At present' these in"
Bulletin o- the American Meteorological
Society
elude an indication o# the o$ser&ationJs source' all
*uality control decisions and their sources' a history
o# all %odi#ications %ade to the o$ser&ation prior to
the analysis (MC corrections' radiation corrections'
&irtual te%perature con&ersions' etc.)' and &arious
$ac+ground *uantities rele&ant to the analysis
process (i.e.' interpolated #irst"guess &alues'
interpolated ana ly2ed &alues' interpolated
cli%atological &alues' in terpolated cli%atological
&ariances' and o$ser&ation error esti%ates). As a
result' the o$ser&ational data $ase produced $y the
reanalysis syste% contains a #airly co%plete
processing history o# each o$ser&a tion' (hich can
$e use#ul in the e&aluation o# the per #or%ance o# the
analysis procedures the%sel&es' as (ell as to other
reanalysis projects carried out at NCEP or at other
research centers. The @B<R data $ase archi&e is a
%ajor enhance%ent o# prior NCEP o$ser&ational
#or%ats and as such has $een i%ple %ented as an
operational product in the NCEP Clo $al 5ata
Assi%ilation =yste%.
g. *ptimal a5eraging
The NCEP8NCAR reanalysis syste% (ill include
not only the co%putation o# gridpoint (and spherical
har%onics) &alues $ut also te%poral and spatial a&
erages o&er so%e prescri$ed areas. A ne( %ethod'
+no(n as opti%al a&eraging (Candin ,--;)' (ill $e
used in the course o# reanalysis. This %ethod assures
%ini%u% (in statistical sense) root"%ean"s*uare a&
eraging errors and' particularly i%portant #or the re
analysis purposes' it pro&ides this esti%atedr%s error
as a $y"product. The incorporation o# 6A (ill result
in an increased a$ility #or detection o# cli%ate
change' $ecause a&eraged &alues are less a##ected $y
the s%all scale e&eryday &aria$ility that acts as noise
and co% plicates the cli%ate change signal detection.
8n the reanalysis syste%' opti%al a&erages o&er
prespeci#ied areas are co%puted #or te%perature'
spe ci#ic hu%idity' u and & co%ponents o# the (ind'
and (ind speed at se&en (,!!!' :.!' /!!' .!!' ;!!'
4!!' and ,!!hPa) pressure le&els. The hori2ontal
areas cur rently include nine 4!S latitude $ands #ro%
the =outh Pole to the North Pole. The (eights are
co%puted $y nor%ali2ing the opti%al (eights so that
the su% o# the (eights o&er each area is e*ual to
one. ?e also in clude the geographical regions
chosen $y the 8nter go&ern%ental Panel #or Cli%ate
Change (8PCC) #or cli%ate change %onitoring
C8PCC,--!' p. ,./)' plus t(o regions co&ering
=outh A%erica9 tropical =outh A%erica' l6oN"4!!='
!!":!!?' and e3tratropical =outh A%erica' 4!!"
.!!=' .!!"/!!?.
447
T(o additional co%putations are included in the
6A co%ponent. 6pti%al a&erages o# the data incre
%ents (o$ser&ation &alue %inus #irst"guess &alue) are
calculated $y the sa%e %odule that currently a&erages
the actual o$ser&ations. 8n addition' a&erages o# the
#irst guess o&er the sa%e areas are co%puted directly
#ro% the spectral coe##icients o# the assi%ilation
%odel.
h. Periodic forecasts from the reanalysis
J. 7. ?allace suggested per#or%ing glo$al #ore
casts during the reanalysis. 8n order to +eep the co%
putational re*uire%ent at a #easi$le le&el' (e per#or%
one :"day #orecast e&ery . days. =uch #orecasts (ill
$e use#ul #or predicta$ility studies' indirect esti%ates
o# the accuracy o# the analysis' and esti%ates o# the
i%pact o# changes in the o$ser&ing syste%s. They (ill
also support the de&elop%ent o# adapti&e %odel out
put statistics #ro% the long reanalysis (P. 5alla&alle
,--;' personal co%%unication).
-. Reanalysis o!t)!t
The design o# the reanalysis output has $een a
%ajor co%ponent o# the project de&elop%ent Fsee also
=chu$ert et al. (,--;) #or a discussion o# the NA=A
reanalysis outputG. 5uring the April ,--, reanalysis
(or+shop' it $eca%e clear that there are %any di##er
ent types o# possi$le applications #or the reanalysis
output and that so%e o# the% (e.g.' transport o# green
house gases' (hich needs in principle all tur$ulent
transports $et(een any t(o layers) ha&e storage re
*uire%ents that #ar e3ceed (hat can $e handled $y the
project. <or this reason' it (as dec)ided that each unit
o# reanalysis output (, %on) (ill include restart #iles'
so that special purpose shorter reanalyses (ith e3
tended output can $e per#or%ed a posteriori.
The reanalysis gridded #ields ha&e $een classi#ied
into #our classes' depending on the relati&e in#luence
o# the o$ser&ational data and the %odel on the gridded
&aria$le. An A indicates that the analysis &aria$le is
strongly in#luenced $y o$ser&ed data' and hence it is
in the %ost relia$le class (e.g.' upper"air te%perature
and (ind). The designation @ indicates that' although
there are o$ser&ational data that directly a##ect the
&alue o# the &aria$le' the %odel also has a &ery strong
in#luence on the analysis &alue (e.g.' hu%idity and
sur#ace te%perature). The letter C indicates that there
are no o$ser&ations directly a##ecting the &aria$le' so
that it is deri&ed solely #ro% the %odel #ields #orced
448
$y the data assi%ilation to re%ain close to the at%o
sphere (e.g.' clouds' precipitation' and sur#ace #lu3es).
<inally' the letter 5 represents a #ield that is o$tained
#ro% cli%atological &alues and does not depend on the
%odel (e.g.' plant resistance' land"sea %as+). Appen
di3 A contains the co%plete classi#ication o# &aria$les.
Although this classi#ication is necessarily so%e(hat
su$jecti&e' the user should e3ercise caution in inter
preting the results o# the reanalysis' especially #or
&aria$les classi#ied in categories @and C. 8n addition
to this si%ple guidance' the user should +eep in %ind
that *uadratic &aria$les (e.g.' +inetic energy' transport
o# (ater &apor) are in general less relia$le than the
co%ponents #ro% (hich they (ere co%puted. Appen
di3 @ contains the list o# %andatory pressure' sig%a'
and isentropic le&els o# the output.
The reanalysis archi&e has $een designed to satis#y
t(o %ajor re*uire%ents9 8) the output should $e co%
prehensi&e' allo(ing' #or e3a%ple' the per#or%ance
o# detailed $udget studies' and 4) it should $e easily
accessi$le to users interested in long ti%e series o#
data. It $eca%e clear that it (as not possi$le to sat
is#y $oth re*uire%ents (ith a single archi&al #or%at.
<or this reason' the output %odule includes se&eral
di##erent archi&es. Reanalysis in#or%ation and se
lected output is also a&aila$le on"line &ia the 8nternet
(httpll9nic.#$.noaa.go&9:!!!). 8n this section (e de
scri$e #our types o# archi&es and the auto%atic %oni
toring syste% that (as designed to *uality control the
output.
a. B234 o(ser5ational archi5e
Reanalysis o$ser&ational data undergo %ultiple
processing stages' any o# (hich %ay in#luence the
*uality o# su$se*uent analysis and #orecast products
(see sections and .). <or purposes o# %onitoring and
re&ie(' and #or research $ased on the reanalysis' it is
use#ul to $e a$le to trace the progression o# MC and
related processing to (hich any particular o$ser&ation'
or group o# o$ser&ations' has $een su$jected prior to
its use (or nonuse) in the actual data assi%ilation. The
@B<R o$ser&ation e&ent archi&e #or%at (descri$ed in
section .#) has $een designed to pro&ide researchers
(ith this capa$ility.
Although the details o# the @B<R #or%at' and the
@B<R structures de&ised to support the o$ser&ation
e&ents archi&e' are rather co%plicated' a <6RTRAN
progra%%er inter#ace pac+age has $een de&eloped to
si%pli#y a userJs interaction (ith these #iles and to
ena$le #airly straight#or(ard access to all o# the ar
chi&e in#or%ation (ithout the need #or a great deal
Vol. //' o.!" March #$$%
o# technical e3pertise in @B<R. These
user"#riendly <6RTRAN inter#ace rou
tines' along (ith appropriate docu%en
tation and instructions #or their use'
(ill $e a&aila$le to reanalysis
in&estigators.
(. Main synoptic archi5e
This is the %ost co%prehensi&e ar
chi&e o# the reanalysis and (ill contain
a large nu%$er o# analysis and #irst
guess "pressure" #ields at !!!!' !0!!'
,4!!' and ,:!! BTC on a 2.se
latitude"
longitude grid) "#lu3'" "diagnostic'" and
"sig%a" #iles on the %odel Caussian
grid) ,-4 3 - points' in order to %ain
tain %a3i%u% accuracy) and restart
#iles at #ull resolution in order to
ensure re produci$ility. The co%plete
list o# out
TA@KE4. Co%prehensi&e "synoptic #or%at" archi&es. =ee appendi3 ; o# the
Kalnay et al. (,--;) #or a detailed list o# #ields' units' etc.' contained in each o#
the #iles listed $elo(.
Anal. Cuess Total 7@ 7@ C@ C@
<ile (day") 5day,6 day"
#ile"I day9J %onJ
(! yr6<)
Restart #iles (non"CR8@)
=ig%a : ,.- ,.. !./ 44..0
=ur#ace = : ,.4 -.0 !.4- ,;-.4
CR8@ ==T1sno(1sea ice
==T ! NA !.!- !.!! ,.4
=no( 8 (eeI>?8 7 8
(ee$?i NA" !.!! !')!! !.!!
=ea ice 8 ! 8 NA !.!! !.!! !.!,
@B<R o$ser&ations
put #ields (ith their classi#ication is
gi&en in appendi3 A. Ta$le 4 su%%a
ri2es the type and &olu%e o# the @B<R
data archi&e and the gridded synoptic
ar
Prep$u#r !
<nl$u#r
@ /. 4-./0 !.:- 4:..
() 789
:.0! ;.! 8.72 -..;0
CR8@ #iles (grid point)
chi&es.
c. 4educed 6time series6 archi5e
This archi&e contains $asic upper"air
para%eters on standard pressure le&els
(Ta$le ;)' selected sur#ace #lu3 #ields
(Ta$le )' and dia$atic heating and ra
Pressure : ,./; ,;.: !.4 ,--.;!
=ig%a : ;.;, 40.: !./- ;:,.;,
Cr$4d ! ,.;. ..! !.,0 //./0
Ar" 2d
()
0.;, 4..4 !./0 ;0;.0
8sen ! !.-- ;.-0 !.,4 ./.!4
Non"CR8@ #iles
diation ter%s #or each analysis cycle
#or the entire reanalysis period. 7ost o#
the
Ronal
6pta&rg J o
: !.!. !.4
!.!; !X,4
!.!, 0.,,
!.!! ,./4
data (ill $e sa&ed in CR8@ #or%at. The
pressure le&el data (ill $e sa&ed on a
4..! latitude"longitude grid' (hile the
sur#ace #lu3 #ields and
radiationldia$atic heating data (ill $e
sa&ed on a T04
Total (C@) .- 4;/0.0
=TK cartridge (!.0 C@ each) - ;-0!
Caussian grid (,-4 3 -). 8n addition' %onthly
%eans o# &orticity' di&ergence' &irtual te%perature'
speci#ic hu%idity' and sur#ace pressure are sa&ed in
spherical har%onic #or% on sig%a le&els. 7onthly
%eans o# the #lu3 ter%s are stored on the Caussian
grid.
The radiationldia$atic heating data (ill $e co%
posed o# t(o radiati&e ter%s (short" and long(a&e)
and #our dia$atic heating ter%s (large"scale conden
sation' deep con&ection' shallo( con&ection' and
&er tical di##usion). 7onthly %eans o# these data
are stored at each sig%a le&el o# the !!!!' !0!!'
,4!!' and ,:!! BTC cycles separately' so that the
%onthly %ean diurnal cycle in these #ields is
preser&ed.
The data storage order (ill $e %ar+edly di##erent
#ro% the %anner in (hich %odel data ha&e tradition"
Bulletin of the American Meteorological Society
ally $een stored at NCEP. =ince the cli%ate
research co%%unity generally uses indi&idual
para%eters at a single at%ospheric le&el $ut #or
%any ti%e periods (rather that all para%eters #or a
single ti%e)' %uch o# the data are stored in
chronological' not synoptic' or der. That is'
indi&idual #ields #or a single at%ospheric le&el are
a&aila$le #or "all ti%e" #ro% a single data
structure. The $asic pressure le&el data and sur#ace
#lu3 data (ill $e stored in this order' (hich is
re#erred to as "ti%e series" in the rest o# the paper.
d. .uic/:loo/ C+:4*M
output
<ollo(ing the suggestion o#the ad&isory co%%it
. tee' (e are creating a "*uic+ loo+" data$ase that
can
#it into a relati&ely s%all (, yr)6 nu%$er o#
C5"
449
TA@KE;. 5ata #ields on standard pressure le&els to $e sa&ed
on a 4..! lat 3 4..! long grid (, 3 /;). The ,/ le&els used
are
,!!!'-4.':.!'/!!'0!!'.!!'!!';!!'4.!'4!!',.!',!!'/!'
.!' ;!' 4!' and ,! hPa.
TA@KE. =ur#ace #lu3 data to $e sa&ed on the T04
Caussian grid (,-4 3 -).
Class <ield type
Class <ield type
A Ronal (ind
A 7eridional (ind
A Ceopotential height
A Dirtual te%perature
A A$solute &orticity
; Dertical &elocity (,!!!",!! hPa only)
; =peci#ic hu%idity (,!!!";!! hPa only)
R67s. The co%plete content o# the C5"R67s is
descri$ed in appendi3 C. 8t includes t(ice"daily &al
ues o# u" &' R' and T at three tropospheric and three
stratospheric pressure le&els (,!!!' .!!' 4!!'
,!!'.!'
4! hPa). 8n addition' the C5"R67 contains daily
&al ues o# total precipita$le (ater' sur#ace stress'
latent and sensi$le heat #lu3' net long" and
short(a&e #lu3 at the sur#ace' precipitation' and
sur#ace pressure' ==T' air sur#ace te%perature' soil
te%perature and %oisture (t(o le&els). It also
includes isentropic po tential &orticity) u" &' andp at
three isentropic sur#aces) %onthly a&erages) and
;
C
;
=ur#ace te%perature
=+in te%perature
4"% te%perature
; =ur#ace pressure
5 Al$edo
C =ur#ace sensiti&ity and latent #lu3es
C Top"o#"the"at%osphere #lu3es
; Ronal (ind at ,!%
; 7eridional (ind at ,! %
C =ur#ace (ind stress
A 7ean sea le&el pressure
; Precipita$le (ater
C =no( depth
; =no( co&er
C Precipitation (total and con&ecti&e)
; 7ean relati&e hu%idity (%ultiple layers)
C =oil (etness and te%perature
C =ur#ace runo##
C Cloud #raction (high' %iddle' lo()
C Cloud #orcing' clear"s+y #lu3es
C Cra&ity (a&e drag
; 7a3 and %in te%perature
#ields and their co&ariances. =ee appendi3 C #or the
contents o# the C5"R67s.
A special C5"R67' containing ,; years o# se
lected o$ser&ed' daily' %onthly' and cli%atological
data #ro% the NCEP1NCAR Reanalysis' is included
(ith this issue. =ee appendi3 E #or a list o# contents.
e. Automatic monitoring system for the
reanalysis output
As pre&iously noted' the NCEP reanalysis syste%
(as designed to per#or% , %onth o# analyses (ana
ly2ed and archi&ed e&ery 0 h) e&ery day'= days a
(ee+. =ince the &olu%e o# reanalysis output is &ery
large' it is not possi$le #or a hu%an %onitor to
re&ie( and chec+ all the reanalysis products' detect
%ajor er rors' dri#ts' etc. To #ul#ill this re*uire%ent'
(e de&el oped an auto%atic %onitoring syste%
(=aha and Chelliah ,--;) Kistler et al. ,--). At the
end o# each %onth o# reanalysis (e chec+ the ti%es
series o# geo potential height' 2onal (ind'
%eridional (ind' te% perature' and hu%idity at all
standard pressure le&els generated #or e&ery 0"h
period (!!!!' !0!!' ,4!!' and
,:!! BTC).
450
To %onitor the pressure ti%e series' (e use a pre
li%inary cli%atology $ased on NCEPJs C5A= o&er
a /"yr period #ro% , July ,-:0 to ;! June ,--;.
<ro% these daily &alues (e co%puted %onthly
a&erages' standard de&iations #ro% the %onthly
%eans' standard de&iations o# the tendency
(di##erence $et(een suc cessi&e analyses at !!!!
BTC) and standard de&ia tions o# the interpolation
chec+ (di##erence $et(een the analysis and the
interpolation #ro% analyses %ade
4 h $e#ore and a#ter). These statistics (ere
co%puted #or geopotential height' 2onal (ind'
%eridional (ind' and te%perature at each grid point
at ,4 %andatory pressure sur#aces and #or hu%idity
at 0 pressure sur #aces. <or %onitoring the sur#ace
#lu3 *uantities' a ," yr preli%inary cli%atology (,
<e$ruary ,--4";,
January ,--;) o# daily sur#ace #lu3 #iles #ro% a T04
%odel"$ased operational C5A= syste% at NCEP has
also $een created' since no long"ter% archi&e (as
a&aila$le. These short"ter% preli%inary
cli%atologies (ill $e later replaced $y longer ones
deri&ed #ro% the reanalysis itsel#.
Vol. //' o.!" March #$$%
?e use the %onthly statistics o# cli%atological
%eans' daily standard de&iations' and daily standard
de&iations o# the ti%e interpolation chec+' at each ;5
grid point and #or each %onth o# the year' to per#or%
se&eral statistical chec+s9 (e chec+ #or "#ield outli
ers" $y co%puting #or each &aria$le the percentage
o# points (hose distance to the cli%atological
%onthly %ean is larger than t(o standard de&iations.
8# this percentage is larger than the largest &alue o$
ser&ed in the /"yr cli%atology' (e identi#y the #ield
as an outlier and proceed (ith #urther chec+s and di
agnostics. These percentages are also graphically pre
sented to the hu%an %onitor so that trends or ju%ps
are i%%ediately apparent. 8n addition' (e also iden
ti#y "single gridpoint outliers" that di##er #ro% the cli
%ate %ean at each point $y %ore than a speci#ied
nu%$er o# standard de&iations. This chec+ has pro&en
to $e &ery i%portant in identi#ying $ad input
data' such as radiosonde data or satellite (ind data.
A "ti%e interpolation" chec+' is per#or%ed $y co%
puting si%ilar statistics #or the percentage o# the
points (hose di##erence (ith a ti%e interpolated #ield
(#ro% plus and %inus 4 h &alues) is larger than
e3pected.
@oth o# the a$o&e chec+s ha&e pro&en to $e &ery
e##ecti&e and ha&e $een conse*uently adapted #or use
in the data preprocessor (section ). A nu%$er o#
su$tle errors due to #or%at changes' unusual data
types' or *uality"control decisions (ere disco&ered not
only in the reanalysis syste% $ut also in the opera
tional syste%. The use#ulness o# the chec+ is due to
the #act that the cli%atological standard de&iations
(ere co%puted #or e&ery %onth and #or each grid
point in ;5 and (ere there#ore %uch %ore sensiti&e
than any other "gross chec+" pre&iously used at
NCEP. <urther re#ine%ents to the auto%atic %onitor
ing syste% (ill $e de&eloped using the #irst ."yr re
analysis "cli%atology" no( a&aila$le.
?ith respect to the le&el"4 data' the hu%an %oni
tor (ill also ha&e a&aila$le the #ollo(ing in#or%ation9
output o# the cli%atological chec+ o# the o$ser&ations
(a&aila$le #ro% the preprocessor $e#ore the %onthly
reanalysis)' the nor%al operational output o# the 68MC
and the CMC' and se&eral additional plots. These in
clude a plot o# the %ean and r%s data #its to the #irst
guess and to the analysis' classi#ied $y region) a "cur
tain" ti%e plot o# the daily nor%ali2ed r%s #its o# the
data at all le&els) and a plot o# the data tossed out $y
the 68MC. These plots should allo( a hu%an %oni
tor to chec+ large a%ounts o# data in order to detect
serious pro$le%s.
Bulletin o- the American Meteorological Society
.. The Cli'ate +ata Assi'ilation Syste'
The reanalysis project originated (ith the idea o#
per#or%ing a "post analysis" (ith a C5A=' (hich
(ould re%ain #ro2en into the #uture. 8n ,--!' 7.
Cane and J. 1ogues"Paegle o# the ad&isory
co%%ittee sug gested that a &ery long reanalysis
(ould $e %ore use #ul than the C5A= alone. The
de&elop%ent o# the reanalysis syste% (as then
started and $eca%e the largest co%ponent o# the
project. 8t is clear that the co%$ination o# reanalysis
#or the past and the C5A= into the #uture' $oth
using the sa%e #ro2en syste%' (ill $e %uch %ore
help#ul to researchers than either co% ponent alone.
The C5A= analysis (ill $e per#or%ed (ithin ;
days o# the end o# the %onth' (ith the sa%e
so#t(are as the reanalysis. This (ill allo( #or ti%e
to capture the $ul+ o# any delayed data and ser&e as
the $asis #or the generation o# the %onthly Climate
+iagnostics Bulletin o# ClJC.
As noted $e#ore' our plans include a second phase
o# the C5A=1reanalysis to start so%eti%e in ,--:'
a#ter the #irst phase is co%pleted. 8n the second
phase' the reanalysis"4 (ill $e per#or%ed (ith a
,--- state o#"the"artsyste%'coupled(ith a
correspondingC5A="
4' into the
#uture.=uchreanalysis(ouldthen$erepeated
e&ery . years or so using the %ost ad&anced syste%s
and the additional reco&ered data #ro% the past. The
C5A="l' ho(e&er' (ill $e continued into the
#oresee a$le #uture in order to %aintain the longest
ho%oge neous data assi%ilation product possi$le.
Ci&en that the C5A="l (ill $eco%e less e3pensi&e
(ith ti%e' it %ay $e #easi$le to consider running a
#i3ed o$ser&a tion syste% Fchoice (a) at the end o#
section 4G #or co%parison (ith the current
reanalysis' (hich has considera$le &ariations in the
o$ser&ing syste%s.
/. Co!)lin0 %ith the ocean
8n the #irst phase o# reanalysis (e (ill couple the
at%ospheric analysis (ith the opti%al interpolation
reanalysiso# ==T #or ,-:4on(ard.<or the earlier
peri ods (e (ill use the C8==T data that the BK76
has o# #ered to %a+e a&aila$le(Par+eret al.
,--;).TheBK76 C8==T analysis has recently $een
upgraded using E6<s' in colla$oration (ith NCEP.
8n addition' a one (ay coupled ocean reanalysis(ill
$e also per#or%ed.
a. ;he C<P SS;
analysis
The NCEP routinely produces a ,! gridded ==T
4,1
analysis using !,. The analysis is produced $oth daily
and (ee+ly' using / days o# in situ data (ship and
$uoy) and $ias"corrected satellite ==T data. The #irst
guess #or the ==T analysis is the preceding analysis.
@ecause ti%escales o# ==T ano%alies are o# the or
der o# %onths' the analysis #ro% the pre&ious (ee+ is
a %uch $etter esti%ate o# the current ==T than cli%a
tology. There is a large"scale $ias correction #or sat
ellite data' #ound necessary #ro% past e3perience and
$ecause the 78 %ethod assu%es that the data are un
$iased (Reynolds ,-::' ,--;) Reynolds and
7arsico
,--;).The present &ersion o# the 78(ith the $ias
cor
rection is a signi#icant i%pro&e%ent o&er the earlier
NCEP analysis and o&er any other analysis that uses
uncorrected satellite data. 8n the Tropics' the e*uato
rial eastern Paci#ic and Atlantic cold tongues are
%ore realistically sho(n in the !,. At higher
latitudes' the
78 sho(s tighter gradients in the Cul# =trea%' the
Kuroshio' and the <al+landsl7al&inascurrent
regions. The statistics esti%ated in the process o#
de&eloping the ==T 78 analysis sho( that ship ==T
o$ser&ations ha&e larger errors (,.;SC) co%pared to
the errors o# $uoy and satellite ==Ts (!.;S"6..SC).
8n addition' the e-#olding correlation scales ha&e
$een #ound to range $et(een .!! and ,4!! +%
(Reynolds and =%ith
,--).
The (ee+ly &ersion o# the 78 ==T reanalysis has
$een co%puted #or the period #ro% No&e%$er ,-:,
to the present. 8t is not practical to e3tend the period
prior to No&e%$er ,-:, $ecause the present opera
tional satellite instru%ent FAd&anced Dery Ligh
Resolution Radio%eter (ADLRR)G #irst $eca%e op
erational at that ti%e. To de&elop a %ethod to pro
duce relia$le ==T analyses $e#ore No&e%$er ,-:,'
e%pirical orthogonal #unctions (ere co%puted #ro%
the %onthly 78 analyses #or the ,4"yr period #ro%
January ,-:4 through 5ece%$er ,--;. A reduced
set o# spatial E6<s (ere then used as $asis #unctions
that (ere #itted to the in situ analyses to deter%ine
the correct te%poral (eighting o# each #unction.
7onthly ==T ano%alies (ere reconstructed #ro%
the spatial E6<s' #ro% the te%poral (eights #or
the period
,-.!":,' and #ro% a 4! grid #ro% .S= to 0-SN.
These #ields capture %ost o# the &ariance sho(n $y
in situ analyses (hile eli%inating %uch o# the noise
due to sparse in situ data sa%pling. 8n a colla$ora
ti&e e##ort' the BK76 is testing a %odi#ication o#the
original C8==T including the E6< e3pansion in or
der to enhance the signal"to"noise ratio #or the
periods $e#ore ,-:4 (N. Rayner ,--.' personal
co%%unication).
452
(. ;he ocean reanalysis system
The coupled %odel project at NCEP has $een
per #or%ing 5 ocean analyses #or the last / years in
or der to docu%ent %ore thoroughly current and
past cli%ate &aria$ility. These analyses also ser&e
as the initial conditions and &eri#ication #ields
#or the coupled ocean"at%osphere %odel used #or
%ulti season #orecasting (Ji et al. ,--). =ince %ost
o# the potential e3tended predicta$ility is thought to
$e the result o# coupled interactions in the Tropics'
the #o cus at NCEP has $een the de&elop%ent o# the
ocean analysis in this region' $ut (e plan to e3tend
the ocean analysis do%ain to the entire glo$e.
The ocean %odel used (as de&eloped at C<5K.
The Paci#ic %odel has a do%ain that e3tends #ro%
.S= to ..SN and #ro% ,4!!E to /!!?. The Atlantic
do%ain e3tends #ro% ,!!7? to 4!DE and .!7= to
0.SN. The $otto% topography is &aria$le' and there
are 4: le&els in the &ertical. The 2onal grid spacing is
,..! in the Paci#ic and l6in the Atlantic. The
%eridional grid spacing is one"thirdo# a
degree(ithin
,!S o# the e*uator and gradually increases outside
this 2one to ,S pole(ard o# 4!S. ?ithin ,!! o# the
north ern and southern $oundaries the %odel #ields
are re la3ed to cli%atological esti%ates. A
Richardson nu%$er #or%ulation #or the &ertical
%i3ing is used in the upper ocean. Kateral %i3ing is
#or%ulated as pro portional to the s*uare o# the
e*ui&alent hori2ontal (a&enu%$er.
The data assi%ilation syste% (5er$er and Rosatti
,-:-) is a ;5 &ariational techni*ue applied continu
ously in ti%e. Presently' only ther%al data is used in
the analyses. All a&aila$le te%perature data #ro%
ships' satellite esti%ates' dri#ting and %oored $uoys'
and e3penda$le $athyther%ographs (Q@Ts) are used.
E3tensi&e *uality control procedures ha&e $een de
&eloped to screen the data $e#ore they are used. Cor
rections are %ade to the %odel ther%al #ields in the
upper ocean do(n to /4! %. This depth range con
tains the %a3i%u% depth o# the $ul+ o# the a&aila$le
su$sur#ace ther%al data #ro% the T" and T/"type
Q@T pro$es. =ur#ace o$ser&ations are +ept in the
analyses #or 4 (ee+s' and su$sur#ace o$ser&ations
are +ept in #or (ee+s (ith (eights &arying
linearly in ti%e. 7a3i%u% (eight is gi&en at the
o$ser&ation ti%e' (ith %ini%u% (eights at the
$eginning and end o# the ti%e inter&al (hen the data
is $eing used.
Routine (ee+ly ocean"%odel"$ased analyses are
per#or%ed #or the Paci#ic and Atlantic $asins' (ith a
4"(ee+ delay in order to allo( to $e a$le to use the
Q@T data in the assi%ilation #or (ee+s. The %od"
Vol. //' o.!" March #$$%
els are #orced (ith a (ee+ly a&eraged stress #ield de
ri&ed #ro% the #our"ti%es"per"day near"sur#ace (inds
produced $y NCEPJs glo$al at%ospheric analyses.
These (inds are con&erted to stress using a constant
drag coe##icient o# ,.; 3 ,!";N The net heat #lu3 used
to #orce the %odel is set to 2ero in order to #acilitate
the e&aluation o# the heat #lu3es in the analysis sys
te%. The net #resh(ater #lu3 is also set to 2ero) once
a year the salinity #ield is restored to the %ean
cli%a tological #ield o# Ke&itus (,-:4).
Bsing this syste%' ocean reanalyses ha&e $een
per
#or%ed #or the Paci#ic and Atlantic $asins. <or this
purpose' all a&aila$le historical su$sur#ace data (ere
o$tained #ro% the archi&es and edited #or the ti%e
period #ro% June ,-:4 to the end o# ,--4. These
(ere %erged (ith the data that (ere a&aila$le in real
ti%e. The Paci#ic reanalysis #or June ,-:4"
5ece%$er ,--4 has $een co%pleted' and a
reanalysis #or the Atlantic #or the sa%e period is
under(ay.
?or+ on i%ple%enting a glo$al"%odel"$ased
ocean analysis syste% (ill start in ,--.. The routine
(ee+ly analysis capa$ility (ill $e i%ple%ented #irst)
the glo$al reanalysis #or the period ,-:4 to the
present (ill $e started in ,--.'using the at%ospheric
reanaly sis' and should ta+e a #e( %onths #or
co%pletion. ?ith respect to earlier periods' the
scarcity o# su$sur #ace ocean data i%plies that a
%eaning#ul reanalysis can only $e done #or the
Northern Le%isphere' and only #or the period
starting in the late ,-0!s. Ne&er theless' once a
consistent set o# #orcing #ields is a&ail a$le #ro% the
at%ospheric reanalysis' (e plan to per#or% the
ocean reanalysis #or the (hole !"year period
(,-./"-0).
1. Preli'inary res!lts an relia2ility o&
the at'os)heric reanalysis
The NCEP8NCARreanalysis (ill produce !
years o# daily at%ospheric and sur#ace #ields'
(hich' #or so%e &aria$les' are close to a $est
esti%ate o# the e&ol&ing state o# the at%osphere.
The analysis cycle' (ith the use o# the 0"h #orecast
as a #irst guess' is a$le to transport in#or%ation #ro%
data"rich to data"poor regions' so that e&en in
relati&ely data"&oid areas the reanalysis can
esti%ate the e&olution o# the at%o sphere o&er
$oth synoptic and cli%atological ti%escales.
A researcher using the reanalysis should $e
a(are'
ho(e&er' that the di##erent outputs are not uni#or%ly
relia$le. As indicated in section 0' #ields deri&ed
#ro%
Bulletin o- the American Meteorological Society
a #our"di%ensionalanalysis are not e*ually
in#luenced $y o$ser&ations. =o%e' such as upper"air
%ass and te%perature #ields (classi#ied as A in
appendi3 A) are generally (ell de#ined $y the
o$ser&ations and' gi&en the statistical interpolation
o# o$ser&ations and #irst guess' pro&ide an esti%ate
o# the state o# the at%o sphere $etter than (ould $e
o$tained using o$ser&a tions alone. 6thers
(classi#ied as @) are partially de#ined $y the
o$ser&ations $ut are also strongly in #luenced $y the
%odel characteristics. <or e3a%ple' the a%ount o#
%oisture that the tropical %odel at%o sphere can
hold depends on its para%eteri2ation o# cu%ulus
con&ection' since so%e con&ection sche%es tend to
dry out the at%osphere %ore than others.
There#ore' e&en i# the analysis incorporates ra(in
sonde and satellite %oisture data' the o&erall hu%id
ity (ill $e in#luenced $y the cli%atology o# the
%odel. This is e&en %ore true #or *uantities that are
not di rectly o$ser&ed or (hose o$ser&ations are not
cur rently assi%ilated into the present analysis
syste%s. E3a%ples o# these *uantities (classi#ied as
C) are pre cipitation and sur#ace #lu3es. To the e3tent
that the %odel and its physical para%eteri2ations are
realis tic' these #ields can $e relia$le and pro&ide
esti%ates as accurate as any other a&aila$le' e&en on
a daily ti%escale. Lo(e&er' they (ill ha&e regional
$iases i# the %odel tends to $e $iased. <or e3a%ple'
o'9er the southeastern Bnited =tates' the %odel
tends to $e colder and drier than the at%osphere
during the su% %er %onths. As a result' during the
0"h #orecast in the analysis cycle' the %odel tends to
precipitate the in cre%ent o# %oisture added $y the
ra(insonde o$ser &ations during the analysis. This
process o# per%anent spindo(n (ithin the analysis
cycle leads to e3cessi&e reanalysis precipitation in
this area.
8n this section (e present a #e( results #ro% the #irst
. years o# reanalysis (,-:.":-)' co%parisons (ith
the then"operational C5A=' and se&eral diagnostic
stud ies. The i%pact that una&oida$le changes in
the o$ ser&ing syste%s (especially the introduction
o# ne( satellite data) (ill ha&e on the reanalysis is
also as sessed. These results should pro&ide an
indication o# the relia$ility o# di##erent reanalysis
#ields. <urther results are presented in the
proceedings o# the Cli%ate 5iagnostics ?or+shop
(Chelliah ,--) =aha et al.
,--) Jano(ia+ ,--) =%ith ,--) ?hite
,--).
a. 'lo(al energy and )ater (alance
6utput #ro% the NCEP8NCAR reanalysis
includes %any diagnostics o# the physical #orcing o#
the at%o spheric #lo(' including co%plete sur#ace
energy and
4,*
III300
cY
>?????.
""
B
33??
. Y"Y"",,
C
150
c C77
" ?. Morel
9Y
! 25
""
_ 350
C, 325
=C@
=77
.Z
275
8D8 29@ I????
","
~ 225
)99
277
B
2< 200
B
9
,
s
!
C9@
r""
""
Ramanathan
III250
Reanalysis E
2@7
"" Reanalysis
EF
2C@
" ""
Ramanathan
C@7
"
D
- 175
!
....... CC@ """"
r?
I??
CE
125
Morel
,! 89@
100
999E
8C@
.. U""""
" ?
S% L% DSW ESW 1SW DLW ELW 1LW 1%0 1etRad
G6 C8)
8@7
?
75
iii
.!
iii
877
50
..c 9@
EllHt " " "
??.IF<r.
! @7 ? ?
a
0
-25
DSW ESW
C@
"
"
7
OlR 1et
<8C. . (a) Clo$al"%ean radiati&e $alance at the top o# the at%osphere' and ($) sur#ace energy $udget #or ,-:."-, #ro% reanalysis.
Lere' =L is sensi$le heat' KL is latent heat' =? is short(a&e radiation' K ? is long(a&e radiation' 6KR is outgoing long(a&e
radiation' 5 is do(n(ard' B is up(ard' N is net' L< is heat #lu3' R< is radiati&e #lu3' and Rad is radiation. Dalues #ro% reanalysis are
co%pared
to t(o cli%atological esti%ates9 Ra%anathan et ai. (,-:-) and 7orel (,--).
hydrological $udgets' the top"o#"the"at%osphere ra
diation $udget' angular %o%entu% $udgets' and
%onthly %ean dia$atic heating (?hite ,--).
<igure
a co%pares the glo$al"%ean radiation $udget at the
top o# the at%osphere and <ig. $ co%pares the sur
#ace energy $udget #ro% the reanalysis #or ,-:."-,
(ith cli%atological esti%ates #ro% Ra%anathan et
al. (,-:-) and 7orel (,--). <or %ost o# these
#ields' classi#ied as C' the reanalysis agrees
(ith the cli%atologies as (ell as the di##erent
cli%atologies
"agree (ith each other. At the top o# the at%osphere'
up(ard short(a&e radiation #ro% reanalysis appears
to $e ,, ? %"4 stronger than the cli%atological esti
%ates ((hich are #orced to $e in $alance)' and the
at %osphere loses ,, ? %T to space. There is
so%e e&idence that the ocean sur#ace al$edo in the
NCEP %odel is too high' and this %ay increase the
up(ard solar radiation. At the sur#ace the net
radiation is .":
? %"4 less than the cli%atological esti%ates' and the
3!-----~~--~--~!-------------~-!
a
a
G6
at%osphere loses ... ? rrrJ to the sur#ace.
Consis tent (ith the loss o# energy to space and
the sur#ace' the NCEP %odel cools slightly
during the 0"h #irst guess #orecast. The 2onal"
%ean and regional distri$u tions o# sur#ace #lu3es
in reanalysis also appear to $e consistent (ith
cli%atological esti%ates.
<igure . displays %onthly %eans o# the
glo$al %ean hydrological $udget and ,4"%on
running %eans o# net at%ospheric #lu3es #ro%
the reanalysis #or
,-:."-,. 6&er the entire period' e&aporation
e3 ceeded precipitation $y !.! %% day9J. An
annual cycle can $e seen' (ith %a3i%u% &alues in
July. Clo $al"%ean precipitation is (ithin the range
o# cli%a tological esti%ates. There is little
e&idence o# any long"ter% dri#t in glo$al a&erages
in reanalysis.
h. ;he effect of SSM-/ )ind speeds
==788 data $eca%e a&aila$le#ro% the 5e#ense 7e
teorological=atellite=yste% (57=P) in July
,-:/'and
U ,!
E
~ 2!41----------------------------------~
E
.
"
U
ia
4.:["
4.0 .8Y"""J.....I.K..
?IL3<
3I???????)????,)L.L??I??I?????)3
C,
III
r"".
"""
4.4?I??????????????????????????????????????H
U
E 4
\"""""""""""""""""""""""""""""""""""",
I
99
i
9
i
9
o
,.:
8.t=??????I?
*
? E ? /?E
t????????H
"
I
8.:MI?????????r??33??????????
3.
????????I
3
I3S0C
?.A+ ?.A.+L
,.4[ """""""""J.. """J8""""""""""""j
11----------------------------------~
iii !.:"8""""""""""""""""""""""""""""""""""O
.! 7.:MI??????????????????????????????????3
o !. ?I?????????????????????????????????????H
!.4"["""""""""""""""""""""""""""""""",
o
3.C "ffi~~~~~~~~~~~~#ffi~ffiffi~~
c ".
>u
9Y ",!
iii
.
o
!
",.
-20
,""
'4@M S '4: M S '49 M S '44 M S '4N M S 'N7 M S 'N8 M S '4@ M S '4: M S '49 M S '44 M S '4N M S 'N7 M S 'N8
<8C. .. (a) 7onthly %ean glo$ally a&eraged precipitation (P) and e&aporation (E)' and their di##erence) ($) ) 4"%on running
%eans o# the net sur#ace heat #lu3 (=<C)' net radiati&e #lu3 at the top o# the at%osphere (T6A)' and the net #lu3 out o# the
at%osphere (T6TAK) #or ,-:."-, #ro% reanalysis.
454 Vol. //' o.!" March #$$%
NCEP started to use operationally
ocean sur#ace (ind speeds deri&ed
(ith the algorith% o# Cood$erlet et
al. (,-:-) in July o# ,--;' a#ter par
allel tests sho(ed a positi&e i%pact
o# this data. ?e originally repro
cessed a data$ase o# SSM/- radi
ances archi&ed #or cli%ate purposes
pro&ided $y NE=58= (N. Crody and
R. <erraro ,--' personal co%%uni
cation) to deri&e esti%ates o# (ind
speeds. ?e used a neural net(or+ al
Estimated /re>i*itation 5mmDdoy6 -rom Reanalysis
'ul!+u& 8N44
@71
SA1
40N
$ON
!20N
ION
EA
IDS
4!=
30S
!=
50S
gorith% de&eloped $y Krasnopols+y
et al. (,--.). The neural net(or+ al
gorith%' (hich is nonlinear'
60S
o :7E 8C7E ,:!
120W 60W o
results in signi#icantly closer
collocations (ith $uoys than the pre
&ious operational algorith% o#
Cood$erlet et al. (,-:-) and is less
sensiti&e to clouds and %oisture' gi&
ing a %uch larger co&erage.
8t (as disco&ered in 7arch ,--.
that the ==788 (ind speeds assi%i
lated #or the period July ,-:/" 5e
ce%$er ,--,' co%puted #ro% the
cli%ate ==788 data$ase' did not con
tain a trans#or%ation #ro% "antenna"
te%perature to "$rightness" te%
perature. A preli%inary e&aluation
esti%ated that this error created ,!"
% (ind speeds (ith a positi&e $ias
o# a$out 4 % ="8. This $ias resulted
in an increase in sur#ace #lu3es
o# .W",!W. ?ith the corrected
$rightness te%perature ==788 data'
the ju%p is %uch s%aller. Lo(
e&er' the &ery large &olu%e o# the
original SSM/- radiance data' and
e&en o# the reduced SSM/- radi
ance data archi&ed and *uality con
trolled $y <. ?ent2 (,--' NA=A'
personal co%%unication)' results
in a &ery signi#icant slo(do(n o#
the speed o# the reanalysis process
ing. <or this reason (e ha&e de
cided not to use ==718 (inds in the
Estimated /re>i*i tation 5mmD day6 -rom Reanalysis
'ul!+u& 8N49
:71
SA1
=71
30N
C71
ION
E6
IDS
4!=
3%S
405
50S
0!=
Di--eren>e
'ul!+u& 8N44?8N49
SON
@71
=71
$ON
20N
ION
EA
lOS
205
$OS
40S
50S
60S~~~~------~-----!------~------~~~~
o :7E 8C7E 847 8C7W :7W 7
<8C. 0a. Esti%ation o# a&erage precipitation during July"August ,-::' July"
August
,-:/' and their di##erence' accu%ulated in the 0"h #orecasts o# the reanalysis. Contour
lines at 4' ' :' ,4' and ,0 %% day9J.
#irst phase o# the reanalysis. The second phase (ill
include the use o# all ==788"deri&ed products.
c. Sensiti5ity of monthly means
@e#ore the reanalysis $egan' the i%pact on %onthly
%ean #ields o# changes in the %odel used #or the #irst
guess (as e3a%ined. These changes included the e#
#ect o# hori2ontal and &ertical resolution (T04 and
T,40) and di##erent con&ection sche%es. =alstein
(,--;) also e3a%ined the e##ect o# hori2ontal resolu"
Bulletin o- the American Meteorological Society 4,,
tion during 7ay ,--4. The results
indicated that upper"le&el di&er
gent #lo(' precipitation' and
strato spheric (inds (ere %ost
sensiti&e to changes in the NCEP
analysis #orecast syste%. The
large"scale pattern o# upper"le&el
di&ergent #lo( (the scales
represented $y the &elocity
potential) appeared to $e #airly
ro$ust in the Tropics) ho( e&er'
the %agnitude o# the upper le&el
di&ergent #lo( in the Tropics and
the s%aller"scale #eatures are still
poorly de#ined $y a %odern
state"o#"the"art analysis syste%.
d. Precipitation and soil moisture
Precipitation and soil %oisture
ha&e C classi#ications' (hich
%eans that data o# these types are
not assi%ilated $ut rather are de
ri&ed co%pletely #ro% the %odel
0" h #orecasts. <igures 0a'$
depict precipitation #ro% $oth
reanalysis and #ro% a dataset
containing sat ellite"deri&ed
rain#all esti%ates o&er the
oceans (=pencer ,--;) #ro% the
7=B and rain gauge data o&er
land. Precipitation %aps are
presented #or the July"August
%eans o# the years ,-:/ and
,-::' and soil %oisture %aps #or
August
,-:/ and ,-:: are included.
?e chose to co%pare the
results $et(een ,-:/ and
,-::$ecause o# the large
precipitation shi#ts that (ere
o$ser&ed in %any i%portant
regions o# the Tropics' associated
(ith the transition $et(een "(ar%
episode" E8 Nino"=outhern 6scil
lation (EN=6) conditions (Ras
%usson and Carpenter ,-:;) and
"cold episode" conditions (=hu+la
and Paolino ,-:;) in the tropical
Paci#ic during the ,-:0":: period.
Estimated /re>i*itation 5mmDdoy6 -rom MSE
' ul !+u& 8N44
SA1 .
40N
30N
C71
86N
EA
lOS
205
305
405
.!=
Estimated /re>i*itation 5mmDday6 -rom MSE
'ul!+u& 8N49
SON
40N
30N
C71
ION
EG
IDS
20S
2
=7@ ",
r
!,
.aill'F<<'r<.O?)3iIisPad
50&~~~~~~~II~iI~~II~~~~~~~
Di--eren>e
'ul!+u& 8NS4?8N49
SON
=71 .
'A1
C71
86N
EA
l6=
205
305
405
.!=
:7E 8C7E 847 8C7W 0!,J, a
<8C.0$. As in <ig. 0a $ut esti%ated #ro% 7=B and rain gauges.
==Ts (ere %ore than ;SC higher o&er %uch o# the
tropical Paci#ic during the (ar% e&ent co%pared to
the cold e&ent' (hich had a large i%pact on the pat
tern o# tropical con&ection and su$se*uent latent
heat release. Ropele(s+i and Lalpert (,-:/' ,-:-)
ha&e sho(n that precipitation tends to $e less (%ore)
than
456
nor%al o&er 8ndia and the surrounding
ocean during
(ar% (cold) episodes' and they pro&ide
e&idence that the Paci#ic 8TCR is
displaced south(ard during (ar%
episode conditions relati&e to cold
episode conditions. The precipitation
patterns o&er 8ndia and the Paci#ic 8TCR
region (ere su$se*uently docu%ented
#or these
Vol. //' o.!" March #$$%
20N
SS. and Estimated Soil Moisture
+u& 8N4
SS. and Estimated Soil Mois ure
, +u& 8N49
50N~~--------~~~~--~~~~~~
&'
(0
$O
86N "-,]r'
EG
)OS
205
30S
!=OOOO
.
6
!
0
=
S
"H"9UYOU.499OOOOOOOO""))))4OYO
o
o i--eren >e
+u& 8N44?8N49
*ON
SON
!,J,
278)8
20N
i6N
EA
87@
205
;!=
405
@7S
50S-~~~~~~~~~~~--~~~~~~~+,~
7.=
!';.
!.;
!.4.
7.C
7.8@
0!4
7.2@
7.2
7.C@
7.C
0.15
7.74
7.78
"!.!,
"!.!:
rain gauge dataset and (ith the
studies %entioned a$o&e' o&er
$oth the 8ndia region and the
Paci#ic 8TCR. ?hile the a%pli
tude o# the di##erences are con
sidera$ly less than those o# the
7=Blrain gauge esti%ates in
the tropical Paci#ic' there are
also large di##erences in %agni
tude a%ong independent satel
lite esti%ates o# rain#all such as
those $ased on 7=B' in#rared'
and ==788 data. The 7=B esti"
%ates' li+e in#rared algorith%s'
tend to o&eresti%ate rain rates
and their geographical e3tent (P.
Ar+in ,--.' personal co%%uni
cation). The reanalysis contains
s%aller scales than the 7=B'
(hich is pro$a$ly too s%ooth.
The soil %oisture changes
(<ig. 0c) sho( that in the re
analysis 8ndia (as (etter and
North A%erica (as generally
drier in ,-:: than in ,-:/' as
o$ser&ed. The 7=B1raingauge
esti%ates (<ig. 0$) suggest that
Central A%erica had %ore rain
in ,-:: than in ,-:/ $ut
that %ost o# the rest o# =outh
A%erica (as drier irl,-::. This
is also true in the reanalysis $ut
(ith (ea+er a%plitudes.The
re analysis underesti%ates the
in tensity o# the drought in
the south and east o# the
Bnited =tates during ,-::.
6&erall' the soil %oisture
generally appears to $e reason
0IG. 0c. =oil %oisture (relati&e to total #ield capacity' 4!! c%) and sea sur#ace
te%peratures a&eraged #or August ,-::' August ,-:/' and their di##erence.
a$le and does not sho( a long
ter% tendency to dri#t into
e3cessi&ely dry or (et regi%es
(<ig. /a)' e&en (ithout use o#
speci#ic e&ents $y Jano(ia+ and Ar+in (,--,). The
rain#all patterns $et(een the Paci#ic (ar% and cold
episodes that are represented in the 7=B1rain gauge
dataset descri$ed a$o&e are also consistent (ith the
studies %entioned a$o&e.
The di##erence in the pattern o# reanalysis precipi
tation $et(een the northern su%%ers o# ,-:/ and
,-:: co%pare (ell (ith those o$ser&ed $y the
7=B8
Bulletin of the American Meteorological
Society
sur#ace data $ut (ith a s%all nudging to(ard cli%a
tology. <igures /$'c sho( the %a3i%u% and %ini
%u% %onthly soil %oisture content as a percentage
o# the #ield capacity (4.!! %) and the %onth o# oc
currence (only e&ery third gridpoint arro( is
plotted). As e3pected' the maximum soil %oisture
generally occurs at the end o# the (inter in
%idlatitudes and at the end o# the %onsoonal regi%e
in the Tropics. The
457
,4 ] i'" [[[[[].[.,. 1..J[9J9[ ] I." i.!'
...!.!.... .... ......,/.
!
-,.".' "'!'
.
.
..
.
.
..
'
. .
.
.
..
.
.
(a)
4/
' (J] , ' ("] 8OJ
I I I D I I I D I
%ini%u% occurs generally a#ter the su%%er in the
%idlatitudes and $e#ore the %onsoon season in the
4 J] ] I#"$, )+ 8 tr:= I >l>?" ?
O 8 '# 8 I%& >r ?" J8 /r ?" ?
,
4,
,Q D, )Q Q D, )Q Q ,I, Q I /- >? ? /I
J]]1T ( ? '% ( I 8# O ] Jp O I J0
,: *J >"7 " I Q.8 I ' " I ',
",I
LR: .33 )Q' J
. h.@ 66 " ? R I'/' -. ' #
,.
1/ . ,' / ".: DD Q .." Ii ], !.')
I I I ( I
Tropics.
A *uantitati&e co%parison o# the reanalysis precipi
tation ano%alies o&er the Bnited =tates ((ith respect
to the ."yr %ean) (ith the %onthly precipitation
ano%alies esti%ated $y the NC5C cli%ate di&isions
sho(s a correlation pattern o# the ano%alies o# a$out
!W"0!W' so%e(hat higher in the (inter than in the
- NNN"""" NNNNNUNNN1"" N. """""""""""" "".""U. .'."""""""""""""""" .' U'1 "'...U"1 NN"""".....U"""""""""""c'' .....%!....
s
.
u%%er.
..........
.D
.D
MDr?????,?<>1..3.,.?).
?3
OOO9Ooa
N............3 ......
A13..I...
4
<igure : sho(s daily precipitation rates #or 7ay
,-:.":- o&er the Bnited =tates in the NCEP1NCAR
reanalysis and in the o$ser&ations' <igs. :a and :$'
respecti&ely. The corresponding standard de&iations
'EL '+1 'Et.
,-:0
,
,
-:
,
/
'EL
'+1 'El
,-:a
'+1 'EL
,-:- o# the daily %ean precipitation rates
(ithin 7ay ,-:.":- are also sho(n #or
the reanalysis (<ig. :c) and #or the o$
ser&ations (<ig. :d). The o$ser&ations
(ere o$tained #ro% the hourly precipi
tation data$ase co%piled $y the Tech
ni*ues 5e&elop%ent Ka$oratory o# the
National ?eather =er&ice (N?=) and
contain a$out ;!! N?= sites and 4.!!
cooperati&e stations. This data (ere
gridded on a 4..S grid (I. Ki ,--' per
sonal co%%unication). A co%parison o#
<igs. :a and :$ sho(s that in the south
eastern Bnited =tates the reanalysis pre
cipitation is larger than o$ser&ed $y a
#actor o# al%ost 4. As pre&iously %en
tioned' this is due to a regional spindo(n
o# the %odel' (hich' $eing slightly drier
and colder than the at%osphere' tends to
rain out incre%ents o# %oisture reintro
duced $y the analysis. Lo(e&er' the
daily &aria$ility o# the precipitation
analysis co%pares *uite (ell (ith the
station &aria$ility (<igs. :c'd).
<8C./. (a) E&olution o# the soil %oisture content a&eraged o&er se&eral
regions o# the (orld during the #irst . years o# reanalysis. The units are
percentages o# the total #ield capacity (4!! ern). ($) A&erage %a3i%u% soil
%oisture content esti%ated #ro% / years o# reanalysis. The arro(s indicate the
%onth at (hich the soil %oisture is' on the a&erage' %a3i%u%. (c) As in ($)
$ut #or the %ini%u% soil %oisture content.
4,/
e. .uasi:(iennial oscillation and the
stratospheric analysis
The operational NCEP glo$al data as
si%ilation syste% had poor resolution
in the stratosphere until July ,--;'
(hen the &ertical resolution (as
increased #ro% ,: to 4: le&els' and the
top %odel le&els (as %o&ed up to 4./
hPa' changes that (ere also
incorporated into the re analysis.
<igure -a sho(s a .!"hPa
Lo&%oeller diagra% (longitude"ti%e)
o# the 2onal &elocity at .SN to .!= #or
the
Vol. //' No.;' March #$$%
=71 .J,
a
$
<8C. /d. 5i##erence $et(een %a3i%u% and %ini%u% soil %oisture.
point in the reanalysis near Canton 8s
land. The characteristics o# the cross sec
tion are si%ilar to those sho(n $y Reed
and Rogers (,-04) #or the Canton 8sland
data. Their hand analysis o# the station
ra(insonde data also has a do(n(ard
propagation o# the phase' (ith a #aster
change #ro% easterlies to (esterlies than
the re&erse. Co%parisons o# the reanaly
sis near =ingapore also sho( that there
is good agree%ent (ith the ra(insonde
data' indicating that the analysis syste%
is a$le to assi%ilate (ell the data e&en
in the upper"e*uatorial stratosphere.
7ore generally' co%parisons o# the
stratospheric reanalysis (ith the o##"line
stratospheric analysis per#or%ed $y
NCEP (<inger et al. ,--;) sho(s
operational C5A= #or the . years ,-:.":- (denoted
cli%ate diagnostics data $ase or C55@). This (as
the highest %andatory pressure le&el a&aila$le in
the C5A= at the ti%e. <igure -$ sho(s the sa%e ."
yr plot #or the reanalysis. The *uasi"$iennial
oscillation is &ery clear in the reanalysis and
essentially a$sent in the operational analyses.
<igure ,! sho(s a log pressure"ti%e cross section
#ro% ,!! to ,! hPa #or a
&ery good agree%ent (=."K. Iang ,--' personal
co%%unication).
f -mpact of the 3''< o(ser5ing system
A study (as %ade to assess the i%pact that the in
troduction o# the #ull satellite o$ser&ing syste% (ill
ha&e on the reanalysis (7o et al. ,--.). T(o sets o#
analyses and #orecasts (ere %ade (ith and (ithout
=6N
m<S<<<FFFF<<<<<)iFII<<s<<FFFltIIFF<<FF<<<<<<<?3<??9?<F<<<<ryIsFF3
=.41
46N
==1

=71
4
.-PF
N#JtOtlJOriOt99OOOO*O
24 1 -4-546-7808-.-
'oS1
F%1
'<C1
271
28N
@
)*
7
N
1
.
3
-t
t09i-)II333IrFS33<<<<??D3e<<'
=:1
%1
=C1 33)8....L..S,3N3h.L"<??l
38N
2:1
34N
32N
271
C41
)6N~~ ~~~~ ~~ ~
~~~~~ ~
2+N #"""r"""O.J...X..'J9X"""r"U'U.'.........K......'"""OO'....""""r" ""J"........'.""""8
125W 120W )88@W 88I'W 87@W 877W N@W TlAW 4@W 47W 9@W 978)8 @@8)8
SON
=41
=)S1 -l,)=L.Pt,
%1
=C1 8tlr33h!5HIHH
=71 <
38N
':1
2=1
2C1
271
28N
C:1.3 3333 33 3 333 3 3
<8C. :. 5aily %ean precipitation rates (%% day9J) #or 7ay ,-:.":- o&er the Bnited =tates in (a) the NCEP8NCAR reanalysis
and in ($) the o$ser&ations. =tandard de&iation o# the daily %ean precipitation rates (%% day9J) (ithin 7ay ,-:.":- in (c) the
NCEP1 NCAR reanalysis and (d) the o$ser&ations. Contour inter&al is , %% dayJ' greater than , %% day"J is shaded.
Bulletin of the American Meteorological Society
459
(a)
($)
REanl< K@1?@SU +ve @7 m" onal Wind
JAN,-:. .O.... 78AB:::::.CC:.:!:.:D::>:.:CC:,:::CC:==.C:==.C::.C:::::OOJ9))""))))))"J,8"999.O.".0YYY99t
JBK,-:.
JBK,-::
JAN,-:-
JBK,-:-
8C7E 847 120W :7W
Lon&itude
:7E 8C7E 847
Lon&itude
<8C. -. a) Lo&%oeller diagra% (longitude"ti%e) #or the 2onal (ind co%ponent at .! hPa at the e*uator (.SN".S=)' in the operational
NCEP Clo$al 5ata Assi%ilation =yste% (#ro% the cli%ate diagnostics data$ase). ($) As in (a) $ut #or the reanalysis 2onal (inds.
the use o# satellite data (=AT and N6=AT) (ithin
the data assi%ilation. The resulting i%pact is s%aller
than that o$tained in pre&ious satellite i%pact studies
%ade using data #ro% the <CCE (,-/-) e3peri%ent'
re#lect ing the e##ect o# i%pro&e%ents that ha&e
ta+en place in the glo$al analysis sche%e and the
%odel. 6&er all' the results are &ery encouraging'
indicating that a long reanalysis should $e use#ul
e&en $e#ore ,-/-' (hen the <CCE satellite"
o$ser&ing syste% (as esta$ lished9 in the NL' the
analyses o# $oth pri%ary &ari a$les and eddy #lu3es
are $asically una##ected $y the satellite data' and
e&en in the =L a large co%ponent o# $oth the
%onthly and the daily ano%alies can $e captured in
the a$sence o# the satellite data.
<igure .a o# 7o et al. (,--.) sho(ed the 2onal
a&erage o# the s*uare o# the correlation $et(een the
N6=AT and =AT daily analyses. 8t indicated that the
N6=AT analysis e3plains close to l66Wo# the daily
5ariance o# the =AT geopotential height analysis in
the NL e3tratropics' $et(een /!W and -!W in the
460
Tropics' %ore than -!W in the %idlatitudes o# the
=L' and $et(een !W and :!W in the Antarctic
region. <igure .$ sho(ed the s*uare correlation o#
the 2on ally asy%%etric stationary (%onthly
a&eraged) eddies de#ined $y the t(o analyses. The
co%parison sug gested that N6=AT captures o&er
-!W o# the 2onal 5ariance o# %onthly %ean
stationary (a&es o# the =AT analyses in %ost o# the
Tropics and =L do(n to
0!o=' (hereas in the NL e3tratropics the agree%ent
is once again close to l66W.?ith respect to the $ias
o# the 2onally a&eraged &alues' the agree%ent
$et(een =AT and N6=AT is generally good' e3cept
a$o&e 4!! hPa and in the polar regions. 6$&iously'
the di##er ences increase #or %ore sensiti&e
*uantities' such as *uadratic #lu3es and their
di&ergence. Typically' the relati&e di##erences
$et(een %eridional #lu3es o# 2onal %o%entu% or
heat esti%ated $y the =AT and the N6=AT are less
than ,!W in the NL e3tratropics and less than 4!W
in the =L %idlatitudes' $ut they can $e as large as o#
order one in the Tropics' the strato"
Vol. //' o.!" March /--0
sphere' and south o# 0!!=. =atellite data did not i%
pact su$stantially the esti%ated precipitation #ields.
g. Comparisons )ith other operational
analyses
<inally' (e co%pare the NCEP operational Clo$al
5ata Assi%ilation =yste% in use during ,--4
(T,401
,: le&els and Kuo con&ection) (ith parallel runs us
ing the ne( si%pli#ied Ara+a(a"=chu$ert sche%e
(Pan and ?u ,--) and the T0414:"le&el syste%
adopted in the reanalysis. ?e also co%pared the
NCEP and se&eral other operational analyses. These
di##erences are pro$a$ly the $est (ay to esti%ate the
precision o# the resulting analyses gi&en si%ilar o$
ser&ational data$ases and' there#ore' are representa
ti&e o# the ro$ustness o# the NCEP reanalysis #ields.
?e de#ine "internal analysis di##erences" as the
r%s di##erence $et(een %onthly %eans co%puted
(ith NCEP syste%s using di##erent %odels.
"E3ternal analysis di##erences" are the r%s
di##erences $et(een NCEPJs %onthly %ean
analysis and those o# other operational syste%s.
The internal di##erences re#lect the sensiti&ityto
the
#irst guess used in the analysis and are an esti%ate
o# the uncertainty in the %onthly %ean analysis
o# the NCEP syste%. <or the Northern
Le%isphere (4!!":!!N)' the internal di##erences are
a$out ; %at
:.! hPa and 0 % at .!!"4!! hPa. 8n the =outhern
Le%isphere the internal di##erences are .':' ,.' and
;! % at :.!' .!!' ;!!' and 4!!' respecti&ely' re#lect
ing the %uch higher uncertainty introduced $y the
lac+ o# ra(insonde data. The e3ternal di##erences
$et(een the NCEP analysis and the BK76 analysis
are a$out
,4' /' -' and ,4 % at :.!' .!!' ;!!' and 4!! hPa'
re
specti&ely. The larger &alues at :.! hPa re#lect the
un certainty introduced $y di##erent terrains and
e3trapolations $elo( the sur#ace. 8n the =outhern
Le%isphere (4!!":!!=)' the di##erences are a$out
4!'
,4',.' and 4. %respecti&ely. Co%parisons (ith
other operational syste%s (ere si%ilar.
<or the %onthly %ean (ind analysis' in the NL
the internal r%s di##erences in $oth the 2onal
and %eridional co%ponent are a$out !. % ="8 at :.!
hPa and !./ % ="8 at 4!! hPa. The e3ternal r%s
di##erences are a$out , % ="8 at :.! hPa and ,.4 % ="8
at 4!! hPa. 8n the Tropics (4!!="4!!N)' the internal
r%s di##er ences #or the 2onal (ind analysis are !./
% ="8 at :.! hPa and 4 % ="8 at 4!! hPa. The
e3ternal r%s di##er ence $et(een the NCEP and the
BK76 operational analysis in ,--4 (as 4 % ="8 at
:.! hPa and 4.. % ="8 at 4!! hPa. <or the %eridional
co%ponent' the r%s di##erences (ere a$out ;!W
s%aller. 8n the =outhern
Bulletin of the American Meteorological
Society
0IG. ,!. Kog pressure"ti%e cross section o# %onthly 2onal
(inds #ro% the ,-:.":- reanalysis #or a point near Canton
8sland #ro% ,!! to ,! hPa.
Le%isphere' the internal di##erences at :.! hPa (ere
a$out !.: % ="8' the e3ternal di##erences (ere ,.;
% ="8' and the %eridional (ind r%s di##erences (ere
a$out 4!W s%aller. At 4!! hPa' $oth the internal and
e3ternal r%s di##erences (ere a$out ,.: % ="8 #or ,
and ,.4 % ="8 #or -.
These #igures indicate the precision (ith (hich
%ode% analysis syste%s can deter%ine %onthly
%ean %eteorological #ields and can $e regarded as a
lo(er esti%ate o# the accuracy (ith (hich such
#ields can $e deter%ined. =ince all the analyses used
e%ploy &ery si%ilar data$ases' it is li+ely that the
true error is larger than the di##erences $et(een the
di##erent analyses' since errors due to data gaps and
%easure %ent errors (ould $e si%ilar in the di##erent
analysis syste%s.
10. Summary
NCAR and NCEP ha&e colla$orated to create a
&ery long reanalysis using a #ro2en' state"o#"the"art
glo$al data assi%ilationsyste% and a data$ase as
co% plete as possi$le. Changes in the o$ser&ing
syste%s can still produce percei&ed changes in the
analy2ed cli%ate' $ut this pro$le% is approached $y
producing parallel reanalyses (at least , year long)
(ith and (ith out using the ne( o$ser&ing syste%
#or the period i%%ediately a#ter its introduction.
The syste% has $een designed (ith ad&anced *ual
ity control and %onitoring syste%s and can produce
4-1
, %onth o# reanalysis per "cloc+" day on a CRA I
YMP/8 superco%puter. 5i##erent types o# output ar
chi&es are $eing created #or di##erent user needs' in
cluding a *uic+"loo+ C5"R67 (one per year)
archi&e (ith the %ost #re*uently used at%ospheric
#ields' as (ell as sur#ace' top"o#"the"at%osphere'and
isentropic #ields.
The output &aria$les ha&e $een classi#ied into
#our
classes' depending on the degree to (hich they are
in#luenced $y the o$ser&ations and1or the %odel.
Bs ers are cautioned that C &aria$les (such as
precipita tion and sur#ace #lu3es) are co%pletely
deter%ined$y the %odel' #orced $y the data
assi%ilation to re%ain close to the
at%osphere.Ne&ertheless'a co%parisono# these
&aria$les(ith di##erenttypes o# o$ser&ationsand
cli%atologies sho( generally use#ul in#or%ation on
ti%escales #ro% a #e( days to interannual
&aria$ility.
Ac/no)ledgments. ?e are &ery grate#ul to the 5irector o#
NCEP' Ron 7cPherson' #or his continued enthusiastic support
#or this project. ?ithout the generous and signi#icant help pro
&ided $y John 5er$er' 5a&id Parrish' Ken Ca%pana' Ro$ert
Cru%$ine' Lua"lu Pan' @ert Kat2' Jordan Alpert' Alan @asist'
7i+e Lalpert' and 5on Carrett #ro% NCEP' Ki"Ioung Ki%
(K7A)' and =te&e ?orley and Chi"<an =hih #ro% NCAR' this
project could not ha&e $een carried out. Ceorge 7urphy and
=arah Roy ha&e supported our co%puter resources needs. The
consultation (ith Cray Research analysts Ron @ag$y' Ceorge
Dander$erg' 7ary 7cCann' and =. ?ang has also $een &ery
help#ul. 5a&id Par+er and Nic+ Raynor (BK76) ha&e gener
ously o##ered the pre",-:4 sea sur#ace te%perature analysis.
@rian 5oty (C6KA) created and enhanced the graphical pac+
age CrA5=' (hich (as crucial #or this project' and 7i+e
<iorino (PC758) de&eloped essential diagnostic so#t(are'
including porting CrA5= to the Cray syste%. Re3 Ci$son and
Per Kall$erg (EC7?<) and =ieg#ried =chu$ert and 5a&id
Ka%ich .NA/A( CKA) ha&e pro&ided us (ith o$ser&ations to #ill
data gaps. =e& eral countries' including Japan' China'
Argentina' @ra2il' and others ha&e contri$uted data not
a&aila$le in real ti%e on CT= especially #or this project. Pro#.
E. Kung (Bni&ersity o# 7issouri) colla$orated (ith the Chinese
7eteorological Agency in reco& ering early Chinese
ra(insonde data. The C6A5= (sur#ace %arine) dataset has
%any participants' including =cott ?oodru## (N6AAlERK)'
=te&e ?orley (NCAR)' Joe El%s (NC5C)' and @o$ Keeley
(Canada17arine En&iron%ental 5ata =er&ice). 6ther +ey
NCAR participants in data preparation are @o$ 5attore
(rao$s' aircra#t)' ?il$ur =pangler (upper air)' Cregg ?alters
(upper air)' 88ana =tern (sur#ace)' Roy @arnes (sur#ace' T5,;)'
and Joey Co%eau (T6I=). 5ic+ 5a&is (NC5C) has helped
(ith sur#ace data and consulting on old ra(insonde data' and
John Kan2ante (C<5K) (ith the preparation o# old rao$
dataset T5.. People in %any countries and la$oratories also
deser&e credit #or data preparation. Key participants are the o$
ser&ers around the (orld (ho (or+ day and night to ta+e the
o$ser&ations that %a+e these analyses possi$le.
The project has $een supported since its inception $y the
N6AA 6##ice #or Clo$al Progra%s and $y the National
?eather
462
=er&ice (NCEP and C<5K)' and $y the National =cience <oun
dation (NCAR). ?ithoutthis generous support (e (ould not
ha&e $een a$le to de&elop this project. ?e are particularly
grate#ul to 7ichael Coughlan and Ken 7ooney #or their
support and guid ance.
The ad&isory co%%ittee' chaired $y Julia Nogues"Paegle #ro%
,-:- to ,--; and $y A$raha% 6ort #ro% ,--; to the present'
has $een a continuous source o# ad&ice and co%#ort (hen
pro$le%s arose. 7ar+ Cane' Julia Nogues"Paegle' and 7ilt
Lale% origi nally suggested per#or%ing a &ery long reanalysis'
and J. =hu+la spearheaded such a project throughout the
research co%%unity. ?e are grate#ul to the% and to the other
%e%$ers o# the panel9 7aurice @lac+%on' 5onald Johnson'
Per Kall$erg' 5a&id =alstein' =ieg#ried =chu$ert' John
Kan2ante' and Ja%es Lurrell.
Pro#essor I. 7. Na&on and t(o anony%ous re&ie(ers %ade
&ery help#ul suggestions and corrected %any errors in the origi
nal %anuscript.
A))eni( A7 NCEP/NCAR reanalysis
co')rehensi$e o!t)!t $aria2les
The output &aria$les are classi#ied into #our
catego ries' depending on the relati&e in#luence o#
the o$ser &ational data and the %odel on the gridded
&aria$le. An A indicates that the analysis &aria$le is
strongly in#luenced $y o$ser&ed data and' hence' it
is in the %ost relia$le class (e.g.' upper"air
te%perature and (ind). The designation@ indicates
that' although there are o$ser&ational data that
directly a##ects the &alue o# the &aria$le' the %odel
also has a &ery strong in #luence on the analysis
&alue (e.g.' hu%idity and sur #ace te%perature). The
letter C indicates that there are no o$ser&ations
directly a##ecting the &aria$le' so that it is deri&ed
solely #ro% the %odel #ields #orced $y the data
assi%ilationto re%ain close to the at%osphere (e.g.'
clouds and precipitation). <inally' the letter 5
represents a #ield that is #i3ed #ro% cli%atological
&alues and does not depend on the %odel (e.g.' plant
resistance'land"sea %as+). This appendi3 contains
the co%plete classi#ication o# &aria$les. Although
this classi#ication is necessarily so%e(hat
su$jecti&e' the user should e3ercise caution in
interpreting the results o# the reanalysis' especially
#or &aria$les classi#ied in categories @and C. 8n
addition to this rule o# thu%$' the user should +eep
in %ind that *uadratic &aria$les (e.g.' +inetic energy'
transport o# (ater &apor) are in general less relia$le
than the co%ponents #ro% (hich they (ere
co%puted.
a. Standard '4-B output
86 PRE==BRE9 PRE==BRE C66R58NATE 6BTPBT
Regular (4..! lat 3 4..! long) grid. All #ields are
instantaneous &alues at a gi&en ti%e9
Vol. //' o.!" March #$$%
Class Field type C Cloud #orcing net solar #lu3 at the top o# the
A
A
A
Ceopotential height (gp%)
u (ind (% ="8) at ,/ le&els
& (ind (% ="8) at ,/ le&els
at ,/ le&els
C
at%osphere (? rrr,6
Cloud #orcing net solar #lu3 at the sur#ace
(? rrr?6
A Te%perature (K) at ,/ le&els
C
Cloud #orcing net solar #lu3 #or total at%o"
@ Pressure &ertical &elocity (Pa ="8) at ,4
spheric colu%n (? m)6
le&els
C
Con&ecti&e precipitation rate (+g nrT ="8)
@ Relati&e hu%idity (W) at : le&els
C
Clear s+y do(n(ard long (a&e #lu3 at the
A A$solute &orticity (="8) at ,/ le&els
sur#ace (? %")
A u (ind o# the lo(est ;!"hPa layer (% ="8)
C
Clear s+y do(n(ard solar #lu3 at the sur#ace
A
& (ind o# the lo(est ;!"hPa layer (% ="8)
(? rn)6
@ Te%perature o# the lo(est ;!"hPa layer (K)
C
Clear s+y up(ard long (a&e #lu3 at the top
@ Relati&e hu%idity o# the lo(est ;! hPa (W)
o# the at%osphere (? rrr")
@ Pressure at the sur#ace (Pa)
C
Clear s+y up(ard solar #lu3 at the top o#
@ Precipita$le (ater (+g rrr,6
at%osphere (? rrr,6
@ Relati&e hu%idity o# the total at%ospheric
C
Clear s+y up(ard solar #lu3 at the sur#ace
colu%n (W)
(? nr?6
A Te%perature at the tropopause (K)
C
Cloud (or+ #unction (J Kg"8)
A Pressure at the tropopause (Pa)
C 5o(n(ard long(a&e radiation #lu3 at the
A u (ind at the tropopause (% ="8)
sur#ace (? %"4)
A & (ind at the tropopause (% ="8)
C 5o(n(ard solar radiation #lu3 at the top o#
A Dertical speed shear at the tropopause (, ="8)
the at%osphere (? rrr,6
@ =ur#ace li#ted inde3 (K)
C 5o(n(ard solar radiation #lu3 at the sur#ace
@ "@est" ( layer) li#ted inde3 (K)
(? m,6
A Te%perature at the %a3i%u% (ind le&el (K)
C Cround heat #lu3 (? %")
A Pressure at the %a3i%u% (ind le&el (Pa)
5 8ce concentration (ice Y ,) no ice Y !) (,1!)
A u (ind at the %a3i%u% (ind le&el (% ="8)
5 Kand"sea %as+ 58 Y land) ! Y sea) (integer)
A & (ind at the %a3i%u% (ind le&el (% ="8)
C Katent heat #lu3 (? %")
5 Ceopotential height at the sur#ace (gp%)
C Near 8R $ea% do(n(ard solar #lu3 at the
A Pressure reduced to 7=K (Pa)
sur#ace (? %")
@ Relati&e hu%idity in three sig%a layers9
C Near 8R di##use do(n(ard solar #lu3 at the
!."!./4'!./4"!.-'!.",.! (W)
sur#ace (? rrr,6
@ Potential te%perature at the lo(est sig%a
C Potential e&aporation rate (? %Y)
le&el (K)
C Precipitation rate (+g %T ="8)
@ Te%perature at the lo(est sig%a le&el (K)
C Pressure at high"cloud top (Pa)
@ Pressure &ertical &elocity at the lo(est
C Pressure at high"cloud $ase (Pa)
sig%a le&el (Pa ="8)
C Pressure at %iddle"cloud top (Pa)
@ Relati&e hu%idity at the lo(est sig%a le&el
C Pressure at %iddle"cloud $ase (Pa)
(W)
C Pressure at lo("cloud top (Pa)
@ u (ind at the lo(est sig%a le&el (% ="8)
C Pressure at lo("cloud $ase (Pa)
@ & (ind at the lo(est sig%a le&el (% ="8)
C
C
Pressure at the sur#ace (Pa)
Runo## (+g nrT per 0"h inter&al)
4) CR@459T?6"
587EN=86N5A8KACN6=T<8C8KE
5 =ur#ace roughness (%)
Class Field type
C Near$y %odel le&el o# high"cloud top
(integer)
C Cloud #orcing net long(a&e #lu3 at the top C Near$y %odel le&el o# high"cloud $ase
o# at%osphere (? rrr,6 (integer)
C Cloud #orcing net long(a&e #lu3 at the C Near$y %odel le&el o# %iddle"cloud top
C
sur#ace (? %
Cloud #orcing
")
net long (a&e #lu3 #or total C
(integer)
Near$y %odel le&el o# %iddle"cloud $ase
at%ospheric colu%n (? rrrI6 (integer)
Bulletin o- the American Meteorological Society 463
e Near$y %odel
(integer)
le&el o# lo("cloud $ase e Karge"scale condensation heating rate
(K ="8)
e =ensi$le heat #lu3 (? nr") e Kong(a&e radiati&e heating rate (K ="8)
e Dolu%etric soil %oisture content (#raction) e =hallo( con&ecti&e heating rate (K ="8)
(t(o layers) e =hallo( con&ecti&e %oistening rate
; =peci#ic hu%idity at 4 % (+g +g ") F+g (+g s"ltlG
e Near$y %odel le&el o# lo("cloud top e 5eep con&ecti&e %oistening rate
(integer) F+g (+g s)6<)U
e Total cloud co&er o# high"cloud layer (W) e =olar radiati&e heating rate (K ="8)
e Total cloud co&er o# %iddle"cloud layer (W) e Dertical di##usion heating rate (K ="8)
e Total cloud co&er o# lo("cloud layer (W) e Dertical di##usion %oistening rate
; 7a3i%u% te%perature at 4 % (K) F+g (+g s"lt,G
; 7ini%u% te%perature at 4 % (K) e Dertical di##usion 2onal acceleration
ture) (K) e Dertical di##usion %eridional acceleration
e Te%perature o# the soil layer (three layers) F% (s s"8tlG
(K)
; Te%perature at 4 % (K) ) =8C7A
e Te%perature o# high"cloud top (K) Caussian grid (,-43 -) on 4: %odel le&els or sur"
e Te%perature o# lo("cloud top (K) #ace. All #ields are instantaneous &alues at a speci#ied
e Ronal gra&ity (a&e stress (N m)6
; Ronal co%ponent o# %o%entu% #lu3
Class Field tye
(N %Y) A Relati&e &orticity (4: le&els) ( ="8)
; u (ind at ,! % (% ="8) ; 5i&ergence (4: le&els) ( ="8)
e Bp(ard long(a&e radiation #lu3 at the top A Te%perature (4: le&els) (K)
o# the at%osphere (? %") ; =peci#ic hu%idity (4: le&els) (+g +gJ)
C Bp(ard long(a&e radiation #lu3 at the A 0 gradient o# log pressure (sur#ace) ( nr,6
sur#ace (? %") A y gradient o# log pressure (sur#ace) ( m)6
e Bp(ard solar radiation #lu3 at the top o# the A u (ind (4: le&els) (% ="8)
at%osphere (? m)6 A
& (ind (4: le&els) (% ="8)
C Bp(ard solar radiation #lu3 at the sur#ace A Pressure (sur#ace) (Pa)
(? %") A Ceopotential height (sur#ace) (gp%)
C 7eridional gra&ity (a&e stress (N m)6 A 0 gradient o# height (sur#ace) (% m)6
C Disi$le $ea% do(n(ard solar #lu3 at the A y gradient o# height (sur#ace) (% m)6
C
sur#ace (? m)6
Disi$le di##use do(n(ard solar #lu3 at the .) 8=EN98=ENTR6P8CC66R58NAT6EBTPBT
C
sur#ace (? m)6
7eridional co%ponent o# %o%entu% #lu3
Caussian grid (,-43 -) %ost on ,! isentropic le&"
els. All #ields are instantaneous &alues at a speci#ied
(N m,6 ti%e9
;
C
& (ind at ,! % (% ="8)
?ater e*ui&alent o# accu%ulated sno(
Class Field tye
depth (+g m,6 A Potential te%perature (sur#ace) (K)
A Te%perature (K)
;) GR;2D< TLREE"
587EN=86N5A8AKCN6=T<8C8KE
Caussian grid (,-4 3 -) on 4: %odel le&els. All
#ields are a&erage o# 0"h integration starting #ro% a
gi&en ti%eJ.
Class Field tye
C +eep con&ecti&e heating rate (K :",)
A
A
;
;
A
;
;
u (ind (% ="8)
& (ind (% ="8)
Pressure &ertical &elocity (Pa ="8)
Relati&e hu%idity (W)
7ontgo%ery strea% #unction (%T ="8)
@runt"Daisala #re*uency s*uared ( ="4)
Potential &orticity F%" (s +g"8)"8G
4-4
Vol. //' o.!" March #$$%
h. *ther non:'4-B output files
C Con&ecti&e rain area co&erage (W)
R6NAK58ACN6=T<8C8KE(@8NARI) @ =ur#ace pressure (hPa)
A&erage o&er -!o"0!o=' 0!o";!o=' ;!o=";!oN' C =ur#ace s+in te%perature (K)
;!S""0!oN' 0!o"-!oN' and glo$al. Bn%ar+ed #ields are C =oil (etness (e%)
instantaneous &alues at a gi&en ti%e. "A &" indicates C =no( depth (%)
a&erage during the 0"h integration9 C
,!"c%"deep soil te%perature (K)
C .!"c%"deep soil te%perature (K)
Class Field tye
5
.!!"c%"deep soil te%perature (K)
A u co%ponent o# (ind (% ="8) at 4: %odel C =ur#ace net short(a&e #lu3 (? %") (A &)
le&els C =ur#ace net long(a&e #lu3 (? %") (A &)
A & co%ponent o# (ind (% ="8) at 4: %odel @ Relati&e hu%idity at the lo(est %odel le&el
le&els (W)
A Dirtual te%perature (K) at 4: %odel le&els @ Dirtual te%perature at the lo(est %odel
@ =peci#ic hu%idity (g g"8) at 4: %odel le&els le&el (K)
@ =*uared &orticity 58 ="4)at 4: %odel le&els @ Te%perature at the lo(est %odel le&el (K)
C =*uared di&ergence 58 ="4)at 4: %odel @ =peci#ic hu%idity at the lo(est %odel le&el
le&els (K)
@ Pressure &ertical &elocity (Pa ="8) at 4: 5 =ur#ace roughness (%)
%odel le&els 5 Kand"sea sea"ice %as+ (integer)
A Te%perature (K) at 4: %odel le&els C Ronal acceleration $y gra&ity (a&e drag
@ Relati&e hu%idity (W) at 4: %odel le&els
(% ="4)(A&)
@ Kinetic energy (%T ="4)at 4: %odel le&els C 7eridional acceleration $y gra&ity (a&e
C Con&ecti&e heating (K ="8) at 4: %odel (% ="4)(A&)
le&els (A&) @ =ur#ace tor*ue Fg (rnT ="4)",G(A&)
C Karge"scale heating (K ="8) at 4: %odel C Cra&ity (a&e drag tor*ue Fg (rrrJ ="4)"8G(A&)
le&els (A&) @ 7ountain tor*ue Fg (rrrJ ="4)",G(A&)
C =hallo( con&ection heating (K sol) at 4: @ Total angular %o%entu% (%T ="8)
%odel le&els (A&) @ Planetary angular %o%entu% (%T ="8)
C Dertical di##usion heating (K ="8) at 4:
%odel le&els (A&) c. 4estart files 8(inary9
C Con&ecti&e %oistening Fg (g ="8)"8G at 4: =pectral (4: %odel le&els) or Caussian grid
(,-4 %odel le&els (A&) 3 -). All #ields are instantaneous &alues at a
speci"
C =hallo( con&ection %oistening Fg (g ="8)"8G #ied ti%e.
at 4: %odel le&els (A&)
C Dertical di##usion %oistening Fg (g ="8)"8G at
,) =lC7A=PECTRACK6E<<8C8EN<8TKE
4: %odel le&els (A&)
C Ronal accel $y &ertical di##usion (% ="4)at Class Field tye
4: %odel le&els (A &) 5 =ur#ace geopotential
C 7eridional acceleration $y &ertical di##usion @ Natural log o# sur#ace pressure
(% ="4)at 4: %odel le&els
A Dirtual te%perature
C =hort(a&e radiation heating (K sol) at 4: @ 5i&ergence
%odel le&els (A&) + Dorticity
C Kong(a&e radiation heating (K sol) at 4: @ =peci#ic hu%idity
%odel le&els (A &)
C Total precipitation (+g %"R) (A &)
4)
=BR<AC<E8KE(6NCAB==8ANCR85) C Con&ecti&e precipitation (+g m?6 (A &)
C =ensi$le heat #lu3 (( %"R) (A&)
Class Field tye
%0
Katent heat #lu3 (( %") (A&) C Earth sur#ace te%perature (K)
H
Ronal stress (dyn %Y) (A&) C =oil %oisture le&el , (W &olu%e)
<H
7eridional stress (dyn m)6 (A&) C =oil %oisture le&el 4 (W &olu%e)
Rain area co&erage (W) C =no( depth (%)
B.Bno- the American Meteorological Society
465
C =oil te%perature le&el 8 (K)
C =oil te%perature le&el 4 (K)
C =oil te%perature le&el ; (K)
5 =ur#ace roughness length (%)
C Con&ecti&e cloud co&er (W)
C Con&ecti&e cloud $otto% height (sig%a)
C Con&ecti&e cloud top height (sig%a)
C Al$edo (#raction)
C =no(1icelland %as+
5 7ini%u% sto%atal resistance (s %"I6
C Canopy (ater content (%)
C Ratio o# l6"% and lo(est sig%a le&el
(inds
gridded &aria$le (see appendi3 A #or a co%plete
clas si#ication). The user should e3ercise caution in
inter preting the results o# the reanalysis' especially
#or &aria$les classi#ied into categories @and C.
0000 an 1"00 9Te Analyses Esti'ate si:e
u" &' te%perature at :.!' .!!' 4!! hPa ,,
7@
Ceopotential height
,!!!' :.!' /!!' .!!' ;!!' 4!! hPa 4-
7@ (sea le&el pressure can $e deri&ed
#ro% the a$o&e #ields)
(#raction)
A))eni( B7 #!t)!t le$els
Standard pressure le5els 8hPa9:
,!!! -4. :.! /!! 0!! .!! !! ;!! 4.!
,!! ,.! ,!! /! .! ;! 4! ,!
-sentropic surfaces 8E9:
=ur#ace pressure
6%ega at .!! %$
Precipita$le (ater
Te%perature at 4 %
=peci#ic hu%idity at 4 %
u" &at ,!%
RL at .!! and 4!! hPa
Total #or !!!! and
,4!! BTe analyses
4, 7@
,!7@
,!7@
,:7@
,7@
4-7@
,;7@
4.:7@yr,
0.! ..! .! !! ;.! ;;! ;,. ;!! 4-! 4:! 4/! +aily a$era0e analyses
Ronal' %eridional (ind stress
,-7@
Sigma le5els:
!.--.! !.-:4, !.-0 !.-4. !.-,.- !.::;:
!.:.: !.:!, !./.!: !.0-; !.0;4- !..0:,
!..!,/ !.;./ !.;/4! !.;,4. !.4.:4 !.4,!,
!.,0:4 !.,;40 !.,!4: !.!/:4 !.!.:! !.!,:
!.!4:: !.!,:; !.!,!, !.!!4/
A))eni( C7 N;C/NCAR reanalysis
o!t)!t on C+<R#;
Net shortllong(a&e #lu3 at sur#ace
Precipitation
Katent1sensi$le heat #lu3
7odel6KR
5o(n(ard short(a&e #lu3 at sur#ace
6utgoing short(a&e #lu3 at top
.min, T%a3 (4"h period)
,7@
:7@
,07@
/7@
/7@
/7@
,/7@
=o%e o# the reanalysis products (ill $e
distri$uted
=+in te%perature (includes ==T) - 7@
on C5"R67. Currently' t(o types o# C5"R67s
are $eing planned. The #irst (ould contain
reanalysis products #or a single year (one C5"R67
per year).
=no( (li*uid (ater e*ui&alent)
Total #or daily a&eraged #ields
,;7@
,,/7@yr,
The second type (ould $e produced a#ter a$out ,!
years and (ould contain ti%e series o# relati&ely
#e( &aria$les. ?e $elie&e this is an e##icient (ay to
sat
0000 =isentro)ic> an 1"00 =stratos)heric>
analyses
Leight and te%perature at
is#y the re*uire%ents o# %ost %e%$ers o# the
%eteo
rological co%%unity' %any o# (ho% (ere
consulted in the preparation o# the output list. The
#ollo(ing is the plan #or the #irst type o# C5"R67.
Note that the output &aria$les should $e classi#ied
into #our categories' depending on the relati&e in#lu
ence o# the o$ser&ational data and the %odel on the
,O6' .! and 4! hPa
,, &at ,!!'.!' and 4! hPa
Potential &orticity on
three - sur#aces (;,.' ;;!' .! K)
,, & on three theta
sur#aces (;,.' ;;!' .! K)
4-7@
4:7@
,.7@
;!7@
466
Vol. //' o.!" March
6##iceNote ;::
(CR8@ta$leo# localde#initions'docu%entation)
7iscellaneous
Totaldocu%entation&olu%e ;7@
7=B
NA=A
%icro(a&e sounding unit
National Aeronautics and =pace
Ad%inistration
NCAR National Center #or At%ospheric
Research
NC5C National Cli%ate 5ata Center
Esti%ated total &olu%e 0/! 7@ yr",
NCEP National Centers #or En&iron%ental
7odeling (#or%erly N7C)
NE=58= National En&iron%ental =atellite' 5ata
+**endiV D< List o- a>r nyms and 8n#or%ation =er&ice
Te%perature on three theta
C5A=
sur#aces (;,.' ;;!'.! K)
,7@
C<5K
Total #or isentropic and
stratospheric analyses ,,/7@yr"8
C8==T
All cross sections
CKA
(%onthly a&eraged) 4 7@ yr"8
C7=
CrA5=
7onthly %eans' &ariances'
CR8@
and co&ariances ,;: 7@ yrT
CT=
L8R=
6$ser&ed 6KR . 7@yr"8
8PCC
CrA5= control and inde3 #iles 47@yr"8
J7A
K7A
5ocu%entation
7E5=
Kalnayet al. (,--;(ithupdates)
Clo$al 5ata Assi%ilation =yste%
Ceophysical <luid 5yna%ics
Ka$oratory
Clo$al 8ce and =ea =ur#ace Te%perature
dataset
Coddard Ka$oratory #or At%ospheres
Ceostationary 7eteorological =atellite
Crid Analysis and 5isplay =yste%
gridded $inary #or%at
Clo$al Teleco%%unications =yste%
Ligh"Resolution 8n#rared =ounder
8ntergo&ern%ental Panel on Cli%ate
Change
Japan 7eteorological Agency
Korean 7eteorological Agency
7arine En&iron%ental 5ata =er&ice
(Canada)
=o#t(areto read gri$ (pC"
CrA5='(gri$)
07@
NL Northern Le%isphere
ADLRR Ad&anced Dery Ligh Resolution N7C
Radio%eter
@B<R $inary uni&ersal #or%at representation N6AA
C5A= Cli%ate 5ata Assi%ilation =yste%
C55@ Cli%ate diagnostics data$ase N=<
C6A5= Co%prehensi&e 6cean"At%osphere N?=
5ata =et 6A
C6KA Center #or 6cean"Kand"At%osphere 6CP
CPC Cli%ate Prediction Center (e3"C5C) 78
CMC co%ple3 *uality control 6KR
57A decision %a+ing algorith% 68MC
56E 5epart%ent o# Energy M@6
57=P 5e#ense 7eteorological =atellite MC
Progra% =i@
EC7?< European Centre #or 7ediu%"Range =L
?eather <orecasts =8R=
EN=6 E8 Nino"=outhern 6scillation ==8
ERK En&iron%ental Research Ka$oratories
<CCE <irst CARP Clo$al E3peri%ent (,-/-) ==718
<TP #ile trans#er protocol =77R
CARP Clo$al At%ospheric Research Progra%
CATE CARP At%ospheric Tropical ==T
E3peri%ent ==B
Bulletin of the American Meteorological Society
National 7eteorological Center (no(
NCEP)
National 6ceanic and At%ospheric
Ad%inistration
National =cience <oundation
National ?eather =er&ice
opti%al a&eraging
6##ice o# Clo$al Progra%s
opti%al interpolation
outgoing long(a&e radiation
68"$ased Muality Control
*uasi"$iennial oscillation
*uality control
=i%ple @iosphere 7odel
=outhern Le%isphere
=atellite 8n#rared =pectro%eter
spectral statistical interpolation (also
+no(n as a ;5 DAR sche%e)
=pecial =ounding 7icro(a&e18%ager
=canning 7ultichannel 7icro(a&e
Radio%eter
sea sur#ace te%perature
stratospheric sounding unit
0/
T04
T8R6=
Triangular 04"(a&es truncation
Tele&ision 8n#rared 6$ser&ation
=atellite
?ith the e3ception o# the #irst se&en' these are single
le&el #ields.
T6D=
T?ERKE
T8R6="N 6perational Dertical =ounder
Tropospheric ?ind Earth Radio
Kocation E3peri%ent
R
Ceopotential height (gp%)>
E
u (ind 5m ="8)>
W
v (ind (% ="8)>
BK76 Bnited Kingdo% 7eteorological 6##ice
T
Te%perature (K)>
B=A< Bnited =tates Air <orce
W
Pressure &ertical &elocity (Pa ="8)> (.!!
DTPR Dertical Te%perature and Pressure hPa only)
Radio%eter
RL
Relati&e hu%idity (W) ($elo( ;!! hPa)
?76 ? orld 7eteorological 6rgani2ation M =peci#ic hu%idity (+g $&)6 at :.!' /!!'
Q@T $athyther%ograph and .!! hPa' #or daily #ields only
P?AT Precipita$le (ater (+g %"4)>
7=KP Pressure reduced to 7=K (Pa)>
A))eni( E7 Content o& the NCEP/NCAR
Reanalysis Cli'atolo0y A;S C+<R#;
CPRATE
C=5K<
Con&ecti&e precipitation rate (+g nr" ="8)>
Clear s+y do(n(ard long(a&e #lu3
(? rrr?6
The enclosed C5"R67' the #irst e&er included (ith
the Bulletin of the American Meteorological Society"
includes #our types o# #iles9 cli%atologies (,;"year a&
erage %onthly #ields)' %onthly #ields (#or each o#the
,; years)' selected daily #ields #or ,--;' and selected
o$ser&ed #ields. All the #ields ha&e $een interpolated
to a uni#or% latitude"longitude 4..! resolution grid
(, $y /; grid points). The ,/ pressure le&els are
,!!!'-4.':.!'/!!'0!!'.!!'!!';!!'4.!'4!!'
,.!' ,!!' /!' .!' ;!' 4!' and ,! hPa #or the cli%atol
ogy and %onthly %ean #ields' and a su$set o# #i&e le&
els (:.!' /!!' .!!' 4!!' and ;! hPa) #or daily &alues.
There are other single"le&el #ields (e.g.' precipitation))
C=5=< Clear s+y do(n(ard short(a&e #lu3
(? rrr,6
C=BK< Clear s+y up(ard long(a&e #lu3 5W %"4)
C=B=<T6A Clear s+y up(ard short(a&e #lu3 at
top o# at%osphere (? rrr?6
C=B=<=<C Clear s+y up(ard short(a&e #lu3 at
the sur#ace (? rrr,6
5K ?R<=<C 5o(n(ard long(a&e radiation #lu3
at the sur#ace (? nr?6
5=?R<T6A 5o(n(ard short(a&e radiation #lu3
at top o# at%osphere (? %"4)>
5=?R<=<C 5o(n(ard short(a&e radiation #lu3
at the sur#ace (? %"4)>
isentropic potential &orticity (8PD) on ,, isentropic
le&els (0.!' ..!'.!'!!';.!' ;;!' ;,.' ;!!' 4-!'
4:!' and 4/! K) #or the %onthly #ields) and three se
lected le&els (.!' ;;!' and ;,. K) #or the daily #ields.
@ecause o# the hori2ontal and &ertical interpolation'
it is reco%%ended that these #ields not $e used #or
$udget studies' (hich generally re*uire access to the
original data. R' B' D' T' and 7=KP are o# type A
(analysis &aria$le is strongly in#luenced $y o$ser&ed
8CEC
KLT<K
PRATE
RBN6<<
=<CR
=LT<K
=68K?,!
8ce concentration (iceY ,) no iceY6)
(l16)
Katent heat #lu3 (? %"4)>
Total precipitation rate (+g mJ ="8)
Runo## (+g %"4)> =ur#ace
roughness (%) =ensi$le heat
#lu3 (? %"4)>
Dolu%etric soil %oisture content
,!"% layer (#raction)>
data)) ?' RL' M' P?AT' Bl6' D,!' T47' and 8PD
can $e considered o# type @ (although there are o$
ser&ational data that directly a##ects the &alue o# the
=68K ?4!! Dolu%etric soil %oisture content at
4!!"% layer
(#raction) >
&aria$le' the %odel also has a &ery strong in#luence
on the analysis &alue)) %ost other &aria$les are o# type
C (indicating that there are no o$ser&ations directly
a##ecting the &aria$le' so that it is deri&ed solely #ro%
the %odel #ields #orced $y the data assi%ilation to re
%ain close to the at%osphere.
The #ollo(ing #ields are included in the %onthly
and cli%atological #iledirectories. The > indicates they
are also included in the daily (,--;) directory #ile.
468
M47
LCK5C6D
7CK5C6D
KCK5C6D
T=<C
T47
BC?5
=peci#ic hu%idity at 4 % a$o&e
ground (+g +g9J)"
Ligh"cloud co&er (W)
7iddle"cloud co&er (W)
Ko("cloud co&er (W)
=+in te%perature (K)>
Te%perature at 4 % a$o&e ground
(K)>
Ronal gra&ity (a&e stress (N %"4)>
Vol. //' No.;' March #$$%
B<KQ Ronal co%ponent o# %o%entu%
#lu3 (N m?4)>
7easure%ents o# 6KR are o$tained #ro% the Ad
&anced Dery Ligh Resolution Radio%eter (A DLRR)
B8!7 u (ind at ,! % a$o&e ground
5m ="8)>
a$oard the N6AA polar or$iting spacecra#t (Cru$er
and Krueger ,-:). The data units are ? %"4 and each
BK ?R<T6A Bp(ard long(a&e radiation #lu3
at top o# the at%osphere 6KR
(? m?4)>
&alue represents the areal a&erage 6KR #lu3 #or a 4..S
V 4..S "$o3." The o$ser&ations o# 6KR during the
,-/-"- ti%e period that are included on this C5
BK ?R<=<C Bp(ard long(a&e radiation #lu3 at R67 are e3clusi&ely #ro% the "a#ternoon" satellite)
the sur#ace (? %"4)> that is' one (hich o$ser&es at the E*uator near !4;!1
B=?R<T6A Bp(ard short(a&e radiation #lu3 at ,;! K=T. It should $e noted that considera$le o$
top o# the at%osphere (? %"4)> ser&ing ti%e dri#t occurs during the li#eti%e o# the
B=?R<=<C Bp(ard short(a&e radiation #lu3 at a#ternoon polar or$iting satellites' and o$ser&ing ti%es
the sur#ace (? rrr") can $e up to . hours later to(ard the end o# a satellites
DC?5 7eridional gra&ity (a&e stress
51 m?C)>
li#eti%e co%pared to the initial launch o$ser&ing ti%e.
The Qie"Ar+in precipitation analysis (Qie and
D<KQ 7eridional co%ponent o# %o%en
tu% #lu3 (N %"4)>
Dl67 1 (ind at ,! % a$o&e ground (% ="8)>
The #ollo(ing co%ponents o# the heat and %ois
ture $udget are only a&aila$le #or the ,;"year cli%a
tology (,/ pressure le&els)9
KRCLR Karge"scale condensation heating rate
5K ="8)
Ar+in ,--0) is deri&ed in a t(o"stage process #ro%
%onthly rain gauge o$ser&ations and se&eral esti%ates
$ased on satellite data. <irst' the satellite esti%ates are
co%$ined using a (eighted a&erage (here the (eights
are proportional to the esti%ated errors o# the &arious
esti%ates. This (eighted a&erage is then %erged (ith
an analysis o# gauge o$ser&ations o&er land and (ith
o$ser&ations #ro% atoll gauges o&er the ocean. 8n gen
eral' gauge &alues are used (here&er a&aila$le.
The %erged precipitation dataset #or ,-/-",--4'
CNDLR
=LALR
D5<LR
=?LR
K?LR
5eep con&ecti&e heating rate (K ="8)
=hallo( con&ecti&e heating rate (K ="8)
Dertical di##usion heating rate (K ="8)
=hort(a&e radiati&e heating rate (K ="8)
Kong(a&e radiati&e heating rate (K ="8)
prepared $y J."K. E. =che%%' (as generated $y co%
$ining o$ser&ed %onthly total precipitation data #ro%
the (orld sur#ace station cli%atology (=pengler and
Jenne ,--!) #ro% NCAR and esti%ated oceanic pre
cipitation #ro% the 7=B %easure%ents (=pencer
,--;). The station data (ere interpolated to a resolu
The isentropic potential &orticity is a&aila$le in the
%onthly %ean (,, le&els) and ,--; daily (; le&els)9
8PD 8sentropic potential &orticity 5rn) =", +g).
<i3ed #ields (type 5) are included in a separate #ile.
tion o# 4..S longitude"latitude $y a&eraging station
&alues (ithin a 4!!"+% radius (ith (eights propor
tional to the in&erse o# s*uare distance (=che%% et
al. ,--4). An atte%pt (as %ade to control the *uality
o# the dataset $y re%o&ing station data reporting to
tal precipitation a%ounts o&er ,!!! %%. The 7=B
esti%ates (ere screened #or sea ice conta%ination $y
6R6C
7A=K
6rography 5m6
Kand"sea %as+ 58 #or land' ! #or sea).
re%o&ing data (ith %onthly totals greater than -!!
%% in regions pole(ard o# .!S latitude.
The Clo$al Precipitation Cli%atology Project'
The o$ser&ed #ields included are (hich is ad%inistered $y the Clo$al Energy and ?a
ter cycle E3peri%ent' has produced a %onthly %ean
6@=-;!KR 5aily &alues o# outgoing long(a&e 4..S gridded precipitation dataset #or the period July
radiation #or ,--; (? rnI6 ,-:/"5ece%$er ,-- (5ece%$er ,-:/ is %issing).
6@=7N6KR 7onthly %eans outgoing This dataset has $een produced $y $lending gauge and
long(a&e radiation (? m)6 in#rared and %icro(a&e satellite esti%ates o# precipi
NCEPRA8N Qie"Ar+in esti%ated rain#all rates
(%% ="8)
tation. ?hile the instantaneous %icro(a&e"$ased pre
cipitation esti%ates are %ore accurate than 8R"$ased
7RC5RA8N =che%% esti%ated rain#all rates esti%ates' the %icro(a&e esti%ates su##er #ro% re
5min ="8). duced te%poral sa%pling (t(ice daily) relati&e to the
Bulletin o- the American Meteorological Society
469
8R (eight ti%es daily) due to the polar or$it o# the
spacecra#t that house the ==788 instru%ents (%ost o#
the 8R data are #ro% geostationary satellites). Thus'
an adjust%ent procedure has $een de&eloped that at
te%pts to %eld the strengths o# these t(o esti%ates
that is' increased accuracy #ro% the %icro(a&e
co%$ined (ith $etter te%poral sa%pling #ro% the 8R.
The adjust%ent procedure is an adaptation o# ear
lier (or+ $y Lu##%an et al. (,--.)' and consists o#
steps that #irst re%o&e the $iases in in#rared esti%ates
$y adjusting to coincident %icro(a&e esti%ates o#
precipitation. The %icro(a&e esti%ates are o$tained
#ro% the SSM/- instru%ent a$oard the 5e#ense 7e
teorological =atellite Progra% series o# satellites and
utili2e a scattering %odel #or esti%ates o&er land
and an e%ission %odel #or o&er ocean esti%ates.
The #i nal analysis step adjusts the %erged satellite
data to the gauge o$ser&ations and co%$ines the%
using (eights that depend on the esti%ated local
error o# each #ield. The gauge data are analyses #ro%
the Clo $al Precipitation Cli%atology Centre and
re#lect ap pro3i%ately 0/!! gauges that ha&e $een
care#ully *uality controlled.
This is a ne( dataset and (e re*uest that users
pro
&ide co%%ents a$out it to Arnold Cru$er' %anager
o# the CPCP' at agru$erAor$it.nesdis.noaa.go&. <or
%ore in#or%ation a$out the Clo$al Precipitation
Cli %atology Project' see the CPCP ho%e page on
the ?orld ?ide ?e$9 http911or$it"
net.nesdis.noaa.go&1 2343/.
Re&erences
Ca%pana K. A' I."T. Lou' K. E. 7itchell' =."K. Iang' and R.
Cullather' ,--9 8%pro&ed diagnostic cloud para%eteri2ation
in N7CJs glo$al %odel. Preprints' #Fth Con, on umerical
GeatherPrediction"Portland6R' A%er.7eteor.=oc.' ;4""
;4..
Chelliah' 7.' ,--9 An assess%ent o# the %onthly %ean at%o
spheric #ields #ro% the NMCINCAR Reanalysis Project.
Proc.
#$th Annual Climate +iagnostics Gor/shop" College Par+'
75' B.=. 5ept. o# Co%%erce' 4;!""4;;.
Collins' ?. C.' and L. =. Candin' ,--!9 Co%prehensi&e hydro
static *uality control at the National 7eteorological Center.
Mon. Gea. 4e5." ,,:'4/.""4/0/.
""' and ""' ,--49 Co%ple3 *uality control o# ra(insonde
heights and te%peratures (CMCLT) at the National
7eteoro logical Center. N7C 6##ice Note ;-!' ;! pp.
FA&aila$le #ro% N6AAlNCEP' .4!! Auth Rd.' ?ashington'
5C' 4!4;;.G
5er$er' J. 5.' and A Rosatti' ,-:-. A glo$al oceanic data
assi%i
lation syste%. #. Phys. *ceanogr." ,-' ,;;;",;/.
""' 5. <. Parrish' and =. J. Kord' ,--,9 The ne( glo$al
opera tional analysis syste% at the National 7eteorological
Center. Gea. 3orecasting" 0'.;:"./.
470
4.1

You might also like