Research activities in Earth System Modelling. Report No. 55. WCRP Report No. 08/2025

Resea ch
Ac i i ies
in Ea h Sys em
Modelling
WGNE Blue Book
Repo No. 55
WCRP Repo No. 08/2025
No embe 2025
© 2025, Wo ld Me eo ological O ganiza ion
Bibliog aphic in o ma ion
This epo should be ci ed as
“Wo king G oup on Nume ical Expe imen a ion (WGNE), 2025. Resea ch ac i i ies in
Ea h sys em modelling
.
Repo No. 55. WCRP Repo No. 08/2025. WMO, Gene a“
Abou WGNE
The Wo king G oup on Nume ical Expe imen a ion (WGNE) o os e collabo a i e de-
elopmen o models o he Ea h sys em (design, implemen a ion, e o diagnosis and
model e ision) ac oss he ull ange o empo al and spa ial scales. WGNE con ibu es
o he mission o he WCRP co e p ojec ESMO.
Abou ESMO
The Ea h Sys em Modelling and Obse a ions (ESMO) co e p ojec coo dina es, ad-
ances, and acili a es all modelling, da a assimila ion and obse a ional ac i i ies wi hin
WCRP. Websi e: wc p-esmo.o g
Edi o s
Bimochan Ni aula (ESMO IPO) and membe s o he Wo king G oup on Nume ical Ex-
pe imen a ion.
Design co e
Sa a Pasquale o (ESMO IPO)
Copy igh no ice
This epo is published by he Ea h Sys em Modelling and Obse a ions (ESMO) P oj-
ec unde a C ea i e Commons A ibu ion 3.0 IGO License (CC BY 3.0 IGO, www.c e-
a i ecommons.o g/licenses/by/3.0/igo) and he eunde made a ailable o euse o any
pu pose, subjec o he license’s e ms, including p ope a ibu ion.
NOTE
The designa ions employed and he p esen a ion o ma e ial in his publica ion do no
imply he exp ession o any opinion s a us o any wha soe e on he pa o he Sec e a -
ia o he Wo ld Me eo ological O ganiza ion conce ning he legal coun y, e i o y, ci y
o a ea, o o i s au ho i ies, o conce ning he delimi a ion o i s on ie s o bounda ies.
Edi o ial no e: This epo has been p oduced wi hou edi o ial e ision by he WMO
sec e a ia . I is no an o icial WMO publica ion and i s dis ibu ion in his o m does no
imply endo semen by he O ganiza ion o he ideas exp essed.
No es om he
Edi o s
A
F om he Edi o ,
Ea h Sys em Modeling and Nume ical Expe imen a ion a e ad ancing a a signi ican
a e. The use o newe and eme ging echnologies is mo ing he ield o wa d, ei he di-
ec ly h ough no el me hodologies o h ough associa ed g ow h in exis ing a eas such
as coupling, o cing esponse, pa ame e iza ion, and e i ica ion. We a e pleased o p es-
en a collec ion o a icles desc ibing such de elopmen s and esea ch in his publica ion.
This publica ion, “Resea ch Ac i i ies in Ea h Sys em Modelling,” o en e e ed o as
he “WGNE Bluebook,” has been a longs anding ac i i y ca ied ou by he WCRP Wo k-
ing G oup on Nume ical Expe imen a ion (WGNE) since 1970. When WGNE joined he
WCRP co e p ojec Ea h Sys em Modelling and Obse a ion (ESMO), i was decided ha
he ESMO IPO would assume edi o ial and publica ion esponsibili y o he Bluebook.
A e a success ul i s yea , we a e pleased o publish he second WGNE Bluebook unde
ESMO’s s ewa dship and con inue o ul ill he publica ion’s o iginal in en ion: o p o ide
an accessible a enue o in o ma ion exchange among modeling esea che s.
Rema kably, his yea ma ks he 40 h anni e sa y o WGNE’s o ma ion as a wo king
g oup, ini ially unde he join supe ision o he Commission o A mosphe ic Sciences
and he Wo ld Clima e Resea ch P og amme and la e epo ing o he WMO Resea ch
Boa d. The echnological p og ess and ans o ma ions o e hese ou decades a e im-
mense and his publica ion s ands as a es amen o he wo k o housands o scien is s
and esea che s who con inuously ad ance he ield.
P oducing his publica ion equi ed he con ibu ions o many indi iduals. I would like o
hank he panel membe s o WGNE and D . Fanny Adlo o hei ca e ul e iew o he
a icles. Special hanks go o Sa a Pasquale o o he excellen wo k on he design and
o ma ing o he inal book. Las bu no leas , I ex end my g a i ude o all he au ho s who
con ibu ed o his collec ion. Th ough he esea ch published in his olume, hei wo k
will in o m and inspi e many o he s o pu sue u he esea ch and de elopmen in Ea h
Sys em Modelling.
Since ely,
D . Bimochan Ni aula
Scien i ic O ice ,
WCRP Ea h Sys em Modelling and Obse a ions (ESMO) In e na ional P ojec O ice
Table o
Con en s
B
Ti le Au ho s Page
Numbe
Sec ion 1 - De elopmen and s udies o coupled models
and Ea h Sys em Models
7
Gene ic Co-p ocessing o Ea h Sys em Models: An Im-
plemen a ion wi hin he Uni ied Fo ecas Sys em Wea h-
e Model (UFS-WM)
U. Tu uncoglu 8
De elopmen o High-Resolu ion Regional P ojec ions o
Indian Ocean Chlo ophyll-a
A. P. Joshi, P. K.
Ghoshal, and K.
Chak abo y
10
A Compa a i e Analysis o Bias-co ec ion Me hods o
CMIP6 A mosphe ic S a e Va iables O e he Indian
Ocean
A. P. Joshi, S. Bhaga ,
P. K. Ghoshal, and K.
Chak abo y
12
Upg ade o he JMA Sub-Seasonal and Seasonal Ensem-
ble P edic ion Sys em (JMA/MRI-CPS4)
Y. Kubo, K. Ochi, J.
Chiba, T. Yoshida, T.
Takaku a, R. Sekiguchi,
Y. Adachi, M. Deushi, S.
Hi aha a
14
The con ec ion pe mi ing con igu a ion o he Regional
Ea h Sys em Model RegCM-ES: simula ion o he sea-
sonal p ecipi a ion o e he No he n Ad ia ic Region
M. Reale, G. Giuliani,
F. Gio dano, M. Ga cia
Valdecasas-Ojeda, L.
Va gas-Heinz , S. Que -
in, E. Coppola, C. So-
lido o, S. Salon
16
Sec ion 2 - Global and egional clima e models: e-
sponse o o cing, impac s udies, sub-seasonal and
seasonal o ecas ing
18
Assessing he Indian Ocean Dipole P edic abili y in
MMCFS 1: Role o Subsu ace Ocean Dynamics
A. Alone, A. S i as a a 19
Clima e-D i en Va iabili y o G oundwa e Recha ge in
A id and Semi-A id A ica: A Regional Modeling Pe spec-
i e
Wael F. Galal 21
S eam low Simula ion using WRF-Hyd o Model o e he
Cau e y Ri e Basin
A. Soni, A. S i as a a 23
Assessing P ecipi a ion Clima ology Using a High-Resolu-
ion Coupled Regional Ocean–A mosphe e Model (RSM-
ROMS)
C. B. Jayasanka , Vasu-
bandhu Mis a
25
Sec ion 3 - Ad ances in o ecas / NWP models: case
s udies, p edic abili y, ensembles
27
E alua ion o WRF Model Pe o mance in Simula ing
T ack and Rain all o Supe Cyclones Gonu (2007), and
Amphan (2020)
SH.Babji , K.Nagalaksh-
mi
28
Ope a ional KIM4.0 Upg ade o Global Medium-Range
P edic ion
E.-H. Lee, J. Kim, I.-H.
Kwon, S. Bae, M.-S. Koo,
K.-H. Seol, I.-H. Cho, H.-
J. Choi
30
Upg ade o JMA's Ope a ional Global Nume ical Wea he
P edic ion Sys em
M. Kawaguchi, T.
Kinami, Y. Ku oki, N.
Shimokawa, K. Su -
ou, H. Yoneha a, H.
Yoshimu a
32
In oduc ion o S ochas ic Humidi y P o ile o Con ec-
i e pa ame iza ion (SHPC) me hod in JMA’s Global
Ensemble P edic ion Sys em
Y. O a 34
Upg ade o JMA’s Global Ensemble P edic ion Sys em Y. O a, K. Ochi, J. Chiba,
H. Oashi, T. Takaku a
36
Ope a ional Use o AMSU-A and ATMS Window Channels
in JMA’s NWP Sys ems
T. U a a, H. Shimizu 38
Implemen ing Isocho ic-Isoba ic T ans o ma ions in he
MCV Model's Dynamic-Physical Coupling
D. Chen, X. L. Li, X. S.
Shen
40
Nume ical expe imen s on snow pa icle cha ac e is ics
in Fu ano, Japan
A. Hashimo o 42
Sec ion 4 - Pa ame e iza ion o physical p ocesses in
Ea h Sys em Models o hei componen s
44
The impac o he mixing leng h cons ain on opical
cyclone simula ions o HAFS
W.Wang, B. Liu, Z.
Zhang
45
A Simpli ied Lake Model o Seasonal P edic ion Y. Adachi, K. Ochi, S.
Hi aha a, Y. Kubo, R.
Sekiguchi
47
Sec ion 5- Fo ecas e i ica ion; no el me hodologies o
diagnose and measu e sys ema ic e o s
49
Powe Spec a Ve i ica ion o Machine Lea ning Wea he
P edic ion
Ba ba a Casa i, Leo
Sepa o ic, Syed Z.
Husain
50
Sec ion 6 - Uncoupled and coupled da a assimila ion o
in eg a ed ea h sys em analysis and p edic ion; me h-
odology and da a impac sensi i i y s udies
52

Upda e o he Radia i e T ans e Model o RTTOV-13 in
JMA's NWP Sys ems
H. Shimizu, A. Ando, N.
Kamekawa, K. Kondo,
N. Kusano, H. Mu a a,
M. Toyokawa, T. U a a
53
Upda ing Su ace Humidi y Obse a ion E o in JMA's
Regional NWP Sys ems
R. Sai o, R. Toguchi 55
Sec ion 7 - De elopmen s in ocean, sea-ice, and wa e
modeling
57
MOM6-CICE6 Fo ecas ing using he Uni ied Fo ecas
Sys em-Wea he Model F amewo k
Z. Ga a o, J. Cum-
mings, A. Wallc a , A.
Bozec, E. Chassigne ,
H.C. Kim, S. Akella, D.
I edell, D. Dukho skoy,
A. Meh a, N. Ba on
58
Upg ade o he Real-Time Ocean Fo ecas Sys em o
e sion 2.5
S. Akella, Z. Ga a o, D.
I edell, J. Cummings, A.
Meh a
60
Sec ion 8 - Reanalysis da ase s and s a is ical pos -p o-
cessing
62
Compa ison o annual a e age clima ology o key ocean-
og aphic pa ame e s in he Bay o Bengal and A abian
Sea
Gopik ishna N, Suja a
K. Mandke, Susmi ha
Joseph
63
Annual cycle o su ace salini y, SST, p ecipi a ion, MLD
and BLT: Compa ison be ween Bay o Bengal and A abi-
an Sea
Gopik ishna N, Suja a
K. Mandke, Sushmi a
Joseph
65
Con en ional Obse a ion Reanalysis (CORe) Da ase s L. Zhang, W. Ebisuzaki,
A. Kuma , and W. Wang
67
Sec ion 9 - Nume ical/compu a ional echniques and
model esolu ion, physics-dynamics and physics-physics
c oss-componen coupling
69
De elopmen o Regional Sho - ange p edic ion Sys-
ems Based on KIM
E.-H. Lee, J. Kim, H.
Cho, S. Y. Bae, K.-H.
Seol
70
Sec ion 10 - Machine lea ning and AI in wea he p edic-
ion and clima e modeling
72
Assessmen o Regional Acidi ica ion and i s Ampli ude
Va iabili y in he Indian Ocean Using an Imp o ed ML-
Based Su ace pCO2 Da a P oduc
A. P. Joshi, P. K.
Ghoshal, and K.
Chak abo y
73
High esolu ion ain all p edic ions o e he Indian egion
h ough hyb id in eg a ion o dynamical and a i icial
in elligence models
D. T i edi, S. Pa naik, O.
Sha ma, N. B. Puhan
75
An SPI-SVM App oach o Ea ly Wa ning o Rain all Ex-
emes: Demons a ion o e a Regional Tes bed
A. Mahapa a, G. P.
Samal
77
Applica ion o a i icial in elligence models in imp o ing
he in ensi y o ecas o T opical Cyclones o e No h
Indian Ocean Basin
D. T i edi, O. Sha ma,
S. Pa naik, H. Kuma , S.
Bansal, N. B. Puhan,
79
De elopmen and
s udies o coupled
models and Ea h
Sys em Models
1
7
Gene ic Co-p ocessing o Ea h Sys em Models: An Implemen a ion wi hin he
Uni ied Fo ecas Sys em Wea he Model (UFS-WM)
U uk Tu uncoglu
Na ional Cen e o A mosphe ic Resea ch, Clima e and Global Dynamics Labo a o y
Email: u uncu@uca .edu
1. In oduc ion
As he complexi y o he ea h sys em modeling applica ions inc eases signi ican ly in e ms o ep esen ed
physical p ocesses and hei nonlinea in e ac ions, ou abili y o p ocess and in e ac wi h he as amoun o
da a p oduced by hose applica ions is undamen ally changed and pushed he communi y o de elop new
and no el da a p ocessing app oaches such as co-p ocessing and in si u isualiza ion. The Uni ied Fo ecas
Sys em Wea he Model (UFS-WM; Wo hen e al., 2024) is one o he eal examples o such a complex,
mul i-componen modeling sys em ha aims o eplace exis ing NOAA’s ope a ional model sui e and
applica ions by uni ying di e en modeling applica ions and con igu a ions. The main objec i e o his s udy
is o inc ease in e ope abili y be ween Ea h sys em models and no el da a-p ocessing ools by p o iding an
easy- o-use, e icien , and s anda dized modeling en i onmen o in e ac ing wi h la ge da a. The newly
de eloped da a p ocessing componen (GeoGa e) in e ac s wi h any ESMF/NUOPC-based model componen
o b ing co-p ocessing capabili y and allow gaining insigh om da a lowing om mul iple Ea h sys em
model componen s and using i o make imely, da a-d i en decisions. The NUOPC Laye mainly de ines
con en ions and a se o gene ic componen s o building coupled models using he Ea h Sys em Modeling
F amewo k (ESMF; Theu ich e al., 2016). The new componen is ini ially in eg a ed wi h he UFS Coas al
modeling sys em and cu en ly es ing o suppo model de elopmen e o s as a pa o he exis ing
eg ession es ing (RT) amewo k. In addi ion, i s plugin-based design enables i o ex end i s unc ionali y
such as In e ac ing wi h di e en p og amming languages (i.e. Py hon) o p ocess he da a, and ha ing online
in e ac ion wi h AL/ML and da a p ocessing ools while he model is unning.
2. Resul s
The GeoGa e is designed o connec ESMF/NUOPC-complian Ea h sys em model componen s o ha ness
he da a p o ided by each componen (Fig. 1a). The GeoGa e co-p ocessing componen ac s like a egula
physical model componen , bu i is specialized o la ge-scale da a in e ac ion. I can p ocess da a lowing
om se e al model componen s o p oduce da a ia in e ac ing wi h AI/ML-based models o d i e o he
physical model componen s in a hyb id coupled modeling con igu a ion. In his design, he GeoGa e
co-p ocessing componen mi o s he expo s a e (ESMF's building block o s o e ields p o ided by he
componen ) o each connec ed componen (Fig. 1a) as i s impo s a e. Then, i c ea es a sepa a e da a
channel (Condui nodes; Ha ison e al., 2022) o each indi idual connec ion (e.g., ATM, OCN, ICE, and
WAV, as shown in Fig. 1b). Then, GeoGa e passes da a o he desi ed p ede ined plugin (I/O, Py hon, o
Pa aView Ca alys ) o in e ac wi h he da a. To demons a e i s capabili ies, he newly de eloped GeoGa e
componen is in eg a ed wi h he ESMF/NUOPC-based UFS-WM. The UFS-WM is a ully coupled ea h
sys em model o sho - and medium- ange esea ch and ope a ional o ecas s. Fo demons a ion, he
mul i-componen ully coupled UFS-WM con igu a ion (cpld_con ol_ge s, including FV3, MOM6, CICE,
and WW3 model componen s) is used. This con igu a ion showcases he Pa aView Ca alys ( 2) plugin and
8
CPS4 akes o e he ole o backing he issuance o One-mon h Fo ecas s om he Global Ensemble P edic ion
Sys em (GEPS). Ope a ional p edic ion will un e e y day wi h i e membe s as pe CPS3, bu on Tuesdays
and Wednesdays he ensemble is enhanced o 25 membe s up o he one-mon h lead. Simila ly, i e-membe
e- o ecas s a e enhanced o 13 membe s up o he one-mon h lead.
3. Ve i ica ion Based on Re- o ecas Expe imen a ion (1991 – 2020)
Figu e 1 shows SST biases. CPS4 educes cold biases in he eas e n Indian Ocean (IO) and mid-la i udes o
bo eal summe and in he eas e n Equa o ial Paci ic due o enhancemen o he cloud and cumulus con ec ion
scheme. Figu e 2 shows a composi e o ou going longwa e adia ion (OLR) lux s a ing om an ini ial da e
when he con ec i ely ac i e phase o he Madden-Julian Oscilla ion (MJO) is p esen in he IO. In ela ion o
GEPS, CPS4 ep oduces he slow eas wa d p og ession o he MJO as well as CPS3.
Re e ences
Adachi, Y. e al, 2025: A simpli ied lake model o seasonal p edic ion. WGNE Res. Ac i . Ea h Sys. Model., submi ed.
Deushi, M. and K. Shiba a, 2011: De elopmen o a Me eo ological Resea ch Ins i u e Chemis y-Clima e Model e sion 2 o he
s udy o oposphe ic and s a osphe ic chemis y. Pap. Me eo ol. Geophys., 62, 1–46.
Hi aha a, S. e al, 2023: Japan Me eo ological Agency/Me eo ological Resea ch Ins i u e Coupled P edic ion Sys em e sion 3
(JMA/MRI-CPS3). J. Me eo . Soc. Japan, 101, 149–169.
Kawai, H. e al, 2017: In e p e a ion o Fac o s Con olling Low Cloud Co e and Low Cloud Feedback Using a Uni ied P edic i e
Index. J. Clima e, 30, 9119-9131.
O a, Y., 2025: In oduc ion o S ochas ic Humidi y P o ile o Con ec i e pa ame iza ion (SHPC) me hod in JMA’s Global Ensemble
P edic ion Sys em. WGNE Res. Ac i . Ea h Sys. Model., submi ed.
Sakamo o, K. e al, 2023: Re e ence manual o he Me eo ological Resea ch Ins i u e Communi y Ocean Model e sion 5
(MRI.COM 5). Tech. Rep. MRI, 87.
Smi h, R. N. B., 1990: A scheme o p edic ing laye clouds and hei wa e con en in a gene al ci cula ion model. Qua . J. Roy.
Me eo . Soc., 116, 435-460.
Yoneha a, H. e al, 2020: Upg ade o JMA’s Ope a ional Global Model. WGNE Res. Ac i . Ea h Sys. Model., 50, 6.19-6.20.
Figu e 1: SST biases o (a, d) CPS4, (b, e) CPS3, and (c, ) ela ed di e ences. (a – c) JJA and (d – ) DJF wi h 1-mon h p edic ion
lead.
Figu e 2: Longi ude- ime composi e o OLR anomalies a e aged a 5°S – 5°N o p edic ions s a ing wi h he MJO’s con ec i ely
ac i e phase in he IO, based on 5 membe s o each ini ial da e in 305 cases.
15

The con ec ion pe mi ing con igu a ion o he Regional Ea h Sys em Model
RegCM-ES: simula ion o he seasonal p ecipi a ion o e he No he n Ad ia ic
Region
M.Reale (1,*), G. Giuliani (2,3), F. Gio dano(2,1), M. Ga cia Valdecasas-Ojeda(4), L. Va gas-Heinz (2,3), S.
Que in(1), E. Coppola(3), C. Solido o(1), S. Salon(1)
(1) Na ional Ins i u e o Oceanog aphy and Applied Geophysics-OGS, T ies e, I aly, (2) Uni e si y o
T ies e, T ies e, I aly, (3) Abdus Salam ICTP, T ies e, I aly, (4) Uni e si y o G anada, G anada, Spain
*Email: [email p o ec ed]
In oduc ion
The No he n Ad ia ic egion is cha ac e ized by a complex opog aphy wi h a p esence o se e al chain
moun ains ( ec angula box in ATM in Figu e 1, le panel), a ela i ely la alley wi h a complex i e
ne wo k (Ri e in Figu e 1, le panel), signi ican ai -sea in e ac ions which d i es deep wa e o ma ions
p ocesses in he No he n Ad ia ic Sea (OCE in Figu e 1, le panel) as well as se e e hunde s o ms o e he
land. He e we desc ibe he con ec ion pe mi ing con igu a ion o he Regional Ea h Sys em Model RegCM-
ES (Si z e al., 2017; Reale e al., 2020) and show i s added alue in simula ing he seasonal p ecipi a ion o e
he egion.
The Regional Ea h Sys em Model RegCM-ES
The RegCM-ES (Figu e 1, igh panel) is composed by he Regional Clima e Model 5
(RegCM5/CLM5/CLMU, Gio gi e al., 2023) wi h a ho izon al esolu ion o 3 km and 41 e ical le els
(con ec ion pe mi ing, ATM), he ocean module MITgcm (OCE, Ma shall e al., 1997a,b) wi h an ho izon al
esolu ion o app oxima ely 700 m and 59 e ical le els (non-hyd os a ic), and (iii) a i e discha ge module
CHyM (RIVER, Coppola e al, 2007) wi h an ho izon al esolu ion o app oxima ely 1 km. The un o each
componen o he sys em and he exchanges o he ields is managed by a d i e based on he Ea h Sys em
Modeling F amewo k (ESMF). The model uses as ini ial and bounda y condi ions he ECWMF ERA5
eanalysis o ATM (He sbach e al., 2020) and Cope nicus Ma ine Se ice Physical Reanalysis o OCE
(Escudie e al., 2021). The model has been un in e alua ion mode s a ing om Augus ,1s 1987 onwa ds.
Figu e 1
Resul s
Figu e 2 shows he simula ed o al p ecipi a ion (in mm) in win e (DJF, a), sp ing (MAM, c), summe (JJA,
e) and au umn (SON, g) in he EURO4AM da ase (Iso a e al., 2014). Panel b, d, , h shows he minimum
biases (con ou lines) be ween he simula ed alues and he obse a ions, while in shaded colo s he added
16
alues (AV) de ined as 100* abs(BiasRegCM5)-abs(BiasRegCM-ES)/abs(BiasRegCM5). When AV>0 he
coupled model imp o es he simula ion o he p ecipi a ion. I is clea ha he coupled sys em imp o es he
simula ion o he p ecipi a ion o e he domain mainly in SON ( ha is he we es season o e he domain) in
he Po Valley and a lee o chain moun ains (Figu e 2h). This imp o emen is likely linked o a be e
ep esen a ion in he coupled sys em o he ai -sea in e ac ion and anspo o mois u e om he sea o he
inland.
Figu e 2
Re e ences
Coppola, E., Tomase i, B., Ma io i, L., Ve decchia, M., & Viscon i, G. (2007). Cellula au oma a algo i hms o d ainage ne wo k ex ac ion and
ain all da a assimila ion. Hyd ological Sciences Jou nal, 52(3), 579–592. h ps://doi.o g/10.1623/hysj.52.3.579
Escudie , R., Clemen i, E., Cipollone, A., Pis oia, J., D udi, M., G andi, A., & Pina di, N. (2021). A high esolu ion eanalysis o he Medi e anean
Sea. F on ie s in Ea h Science, 9, 702285
Gio gi, F., Coppola, E., Giuliani, G., Cia lo`, J. M., Pichelli, E., Noghe o o, R., e al. (2023). The i h-gene a ion egional clima e modeling sys em,
RegCM5: Desc ip ion and illus a i e examples a pa ame e ized con ec ion and con ec ion-pe mi ing esolu ions. Jou nal o Geophysical Resea ch:
A mosphe es, 128, e2022JD038199. h ps://doi.o g/10.1029/2022JD038199
He sbach, H., Bell, B., Be is o d, B., Hi aha a, S., Ho anyi, A., Munoz-Saba e , J., e al. (2020). The ERA5 global eanalysis. Qua e ly Jou nal o he
Royal Me eo ological Socie y, 146(730), 1999–2049. h ps://doi.o g/10.1002/qj.3803
Iso a, F. A., F ei, C., Weilguni, V., Pe cec Tadic, M., Lassegues, P., Rudol , B., e al. (2014). The clima e o daily p ecipi a ion in he Alps: De elopmen
and analysis o a high- esolu ion g id da ase om pan-alpine ain-gauge da a. In e na ional Jou nal o Clima ology, 34(5), 1657–1675.
h ps://doi.o g/10.1002/joc.3794
Ma shall, J., Adc o , A., Hill, C., Pe elman, L., & Heisey, C. (1997). A ini e. olume, incomp essible Na ie S okes model o s udies o he ocean on
pa allel compu e s. Jou nal o Geophysical Resea ch, 102(C3), 5753–5766. h ps://doi.o g/10.1029/96JC02775
Ma shall, J., Hill, C., Pe elman, L., Heisey, C., & Adc o , A. (1997). Hyd os a ic, quasi.hyd os a ic and nonhyd os a ic ocean modeling. Jou nal o
Geophysical Resea ch, 102(C3), 5733–5752. h ps://doi.o g/10.1029/96JC02776
Reale, M., Gio gi, F., Solido o, C., Di Biagio, V., Di San e, F., Ma io i, L., Fa ne i R., Sannino G. (2020).” The Regional Ea h Sys em Model RegCM-
ES: E alua ion o he Medi e anean clima e and ma ine biogeochemis y” Jou nal o Ad ances in Modeling Ea h Sys ems,12, e2019MS001812.
h ps://doi.o g/10.1029/2019MS001812
Si z, L. E., Di San e, F., Fa ne i, R., Fuen es-F anco, R., Coppola, E., Ma io i, L., Reale, M., Sannino, G., Ba ei o, M., Noghe o o, R., Giuliani, G.,
G a ino, G., Solido o, C., Cossa ini, G. and Gio gi, F. (2017), “Desc ip ion and e alua ion o he Ea h Sys em Regional Clima e Model (RegCM-ES)”
J. Ad . Model. Ea h Sys . doi:10.1002/2017MS000933
17
Global and egional
clima e models:
esponse o o cing,
impac s udies,
subseasonal and
seasonal o ecas ing
2
18
Assessing he Indian Ocean Dipole P edic abili y in MMCFS 1: Role
o Subsu ace Ocean Dynamics
Ashish Alone1,*, Anku S i as a a1
1Indian Ins i u e o T opical Me eo ology, Minis y o Ea h Sciences, D . Homi Bhabha Road, Pashan, Pune,
Maha ash a-411008, India
Email*: [email p o ec ed]
In oduc ion
Indian Ocean Diapole (IOD) plays an impo an ole in he opical Indian Ocean clima e a iabili y, cha ac e ized
by a zonal g adien o sea su ace empe a u e (SST) anomalies be ween he wes e n (WEIO) and eas e n equa o ial
Indian Ocean (EEIO). The posi i e IOD e en is ma ked by anomalously wa m WEIO and coole EEIO which esul s
in abo e-no mal ain all o Eas A ica and d ough s o e Indonesia and Aus alia, while s ongly modula ing he
Indian summe monsoon. Con e sely, a nega i e IOD e en wi h coole WEIO and wa me EEIO exe s opposi e
impac s by lipping he ain all and d ough pa e ns. Wi h he s ong in luence o IOD on ain all, ag icul u e, and
socio-economic ac i i ies ac oss he Indo-Paci ic im, accu a e p edic ion o IOD e en s is a key equi emen o
egional clima e se ices.
P edic ing IOD emains a challenge due o i s s ong coupling wi h bo h ocean and a mosphe e p ocesses. The
seasonal cycle o he he mocline - he bounda y be ween wa me su ace wa e and coole deep wa e , along wi h he
ole o deepe ocean cu en s, and he eedback p ocesses collec i ely play a ole in IOD e olu ion. Model IOD o e-
cas o en encoun e s biases, especially in simula ing he mocline dep h, SST g adien s and ain all eleconnec ions,
limi ing he skill o IOD o ecas s. P e ious s udies highligh ed ha coupled o ecas sys ems end o show excessi e
ampli ude in simula ed IOD e en s, exhibi ing a weake link be ween su ace and subsu ace ocean p ocesses. S udies
ha e shown e o s in he ep esen a ion o clima e pa e n like El Ni˜
no-Sou he n Oscilla ion (ENSO)–IOD in e ac-
ions, he eby educing eliabili y a longe lead imes. The Monsoon Mission Coupled Fo ecas Sys em e sion 1
(MMCFS 1) has been widely used o seasonal p edic ion o monsoon a iabili y. While MMCFS 1 has demon-
s a ed skill in simula ing la ge-scale clima e modes such as ENSO, i s pe o mance in o ecas ing IOD e en s has no
been ex ensi ely e alua ed.
This s udy e alua es he seasonal p edic i e capabili y o MMCFS 1 o IOD du ing he pe iod 1982–2016. S udy
uses diagnos ics me hods ha combine su ace SST anomalies, subsu ace ocean he mocline dep h, and coupled
eedback mechanisms. Analysis examines he model’s skill in ep oducing he Dipole Mode Index (DMI), he leading
modes o SST a iabili y, he ole o subsu ace anomalies, and he co ela ions be ween obse ed and simula ed ields.
Wi h he compa a i e analysis o model and obse ed da a he s udy aims o iden i y key s eng hs and limi a ions
o he model in ep esen ing IOD dynamics also he IOD e olu ion, wi h s ong seasonal dependence be ween (June-
Augus ) JJA onse and (Sep embe -No embe ) SON o imp o ing seasonal o ecas s.
Model and Da a
MMCFS 1 is a ully coupled ocean–a mosphe e gene al ci cula ion model de eloped o ex ended ange and seasonal
p edic ion unde he Indian Monsoon Mission ini ia i e. The a mosphe ic componen is based on NCEP CFS 2, wi h
a ho izon al esolu ion o T382 ( 38 km). The ocean componen is based on he MOM4p0d ocean model, wi h a
ho izon al esolu ion o 0.25° ×0.25° nea he equa o . Fo his s udy, we use May Ini ial Condi ions (ICs) o gene a e
seasonal o ecas s o June–Augus (JJA) and Sep embe –No embe (SON) seasons o he pe iod 1982–2016. These
lead imes cap u e he e olu ion o IOD e en s du ing hei de eloping (JJA) and ma u e (SON) phases.
S udy uses obse a ions and eanalysis da ase s essen ial o e alua ing he simula ions om he MMCFS 1. The
ERA5 eanalysis daily sea su ace empe a u e (SST), SST anomalies, and U and V wind componen s a a 0.25°
esolu ion, co e ing he pe iod om 1982 o 2016 a e used. Fo he obse ed he mocline dep h, he da a comes
om he Global Ocean Da a Assimila ion Sys em (GODAS) mon hly empe a u e p o iles, whe e he mocline dep h
is de ined as he dep h o he 20°C iso he m. In he model, he mocline dep h is ex ac ed om MMCFS 1 subsu ace
empe a u e ields by iden i ying he dep h whe e he empe a u e c osses 20°C. This in e pola ion is ca ied ou o
e e y g id poin and ime s ep, bo h in he model and he obse ed empe a u e p o iles.
Resul s
In Figu e 1, he esul highligh s he abili y o MMCFS 1 May IC o cap u e he in e annual a iabili y o he Indian
Ocean Dipole (IOD). The JJAS DMI de ended ime se ies (Fig. 1a) shows ha he model success ully ep oduces
mos posi i e and nega i e IOD e en s du ing 1982-2016, compa ed o obse a ions. Al hough MMCFS 1 sligh ly
19
Figu e 1: IOD Diagnos ics om MMCFS 1 May IC and Obse a ions o 1982–2016: (a) De ended JJAS: DMI
Time Se ies, (b) EOF Modes o JJAS: SST Anomalies, and (c) SST–Wind Co ela ion Pa e ns du ing Posi i e IOD
E en s.
unde es ima es he ampli ude o ex eme e en s, he o e all co ela ion be ween he modeled and obse ed DMI
emains signi ican , indica ing a skilled seasonal o ecas ing abili y om he ini ial May condi ion.
The spa ial pa e ns o SST a iabili y de i ed om he EOF analysis (Fig. 1b) u he demons a e he model’s
abili y o ep oduce he canonical IOD mode. The i s EOF mode in MMCFS 1 shows a dipole s uc u e wi h wa m
anomalies o e he wes e n equa o ial Indian Ocean (WEIO) and cool anomalies o e he eas e n equa o ial Indian
Ocean (EEIO), closely esembling he obse ed mode. The SST-wind co ela ions (Fig. 1c) also show consis ency
wi h he physical p ocesses o he de elopmen o he IOD, wi h enhanced eas e ly wind anomalies o e he cen al
equa o ial Indian Ocean coinciding wi h he cooling o SST in he EEIO, sugges ing ha he model cap u es he
coupled ai –sea eedbacks.
The esul s in Figu e 2 ocus on he ela ionship be ween SST and he mocline dep h (D20), a key d i e o
IOD e olu ion. Panel (a) shows ha he co ela ion be ween obse ed SST and obse ed D20 is high in he EEIO
du ing JJA and SON, con i ming ha shallow he mocline condi ions a ou cold SST anomalies du ing posi i e
IOD e en s. MMCFS 1 eplica es his ela ionship, hough wi h sligh ly weake co ela ions. Panel. I signi ies he
shallow he mocline condi ions a e closely linked o coole SSTs, a key mechanism d i ing posi i e IOD e en s due
o nega i e co ela ions EEIO. The MMCFS 1 model cap u es his ela ionship easonably well, sugges ing ha while
he model ep oduces he essen ial he mocline–SST eedback, i may unde es ima e i s in ensi y. Panel (b) u he
illus a es a clea linea ela ionship be ween de ended SST anomalies and D20 anomalies, wi h highe IOD index
alues co esponding o s onge nega i e D20 anomalies in he EEIO. This highligh s he impo ance o accu a ely
simula ing he mocline dep h a iabili y o imp o e IOD p edic ion skill. Fu he analysis using MMCFS 2 is planned
o e alua e whe he ad ancemen s in model physics and esolu ion can educe coupled biases, imp o e he mocline
ep esen a ion, and enhance he seasonal p edic abili y o he IOD.
Figu e 2: Rela ionship Be ween SST and D20 o IOD Diagnos ics: (a) Co ela ion o Obse ed SST wi h Obse ed
and Model D20 du ing JJA and SON (1982–2016), and (b) Sca e plo s o SST and D20 Anomalies wi h IOD Index
o JJA and SON (De ended).
20

Clima e-D i en Va iabili y o G oundwa e Recha ge in A id and
Semi-A id A ica: A Regional Modeling Pe spec i e
Wael F. Galal
Geology Depa men , Facul y o Science, Assiu Uni e si y
Email: wael a hi70@ho mail.com
In oduc ion
In A ica, G oundwa e is an essen ial wa e sou ce o domes ic, ag icul u al, and indus ial
ac i i ies in a id and semi-a id egions. Howe e , g oundwa e echa ge in hese egions is
ex emely sensi i e o clima e a iabili y and change. Limi ed p ecipi a ion, high
e apo anspi a ion, and in equen ain all lead o complex and a iable echa ge sys ems.
P edic ing g oundwa e echa ge in hei changing clima e condi ions is c i ical o planning and
managing wa e esou ces. Ad ances in coupled Regional Clima e Models (RCMs) and
hyd ological models pe mi s an assessmen and p ojec ion o egional g oundwa e echa ge
a iabili y. None heless, challenges emain in clima e p ojec ion downscaling, es ima ing echa ge
in da a-spa se egions, and ep esen a ion o land-a mosphe e in e ac ions.
Me hodological O e iew
Recha ge a iabili y in a id and semi-a id A ica is assessed using RCMs (e.g. CORDEX A ica)
o GCMs, downscaled o use in hyd ological models such as SWAT and Wa e GAP, and bias-
co ec ed o emo e model e o s (Ali e al., 2022). The coupled models analyze spa io empo al
echa ge a iabili y as a unc ion o soils, land use and wa e inpu da a, and alida ed using
medioc e sa elli e da a (e.g. GRACE, SMAP) o compa able da a in some cases.
Clima e Fo cing and Response
Fo ecas ed changes in he in ensi y, du a ion, and in e al o ain all ha e a di ec e ec on echa ge
e en s. In his ega d, inc eased empe a u es aise e apo anspi a ion, and subsequen ne echa ge,
e en wi h in equen ex ao dina y ain all e en s, d ops. Resea ch shows ha echa ge ac oss
semi-a id A ica is closely linked o a ia ions in seasonal p ecipi a ion associa ed wi h la ge
clima e sys ems such as he El Niño Sou he n Oscilla ion (ENSO) and Indian Ocean Dipole (Taylo
e al., 2018). Recha ge ends o occu in sho bu s s a e in ense s o ms, which suppo s
in es iga ing sub-seasonal and seasonal clima e o ecas s o see how he p incipal ain all and
echa ge sys ems ollow a simila beha io and pa e n. The unce ain y in an RCM's p ojec ed
hunde s o ms and con ec i e ain pe o mance de ac s om o e all con idence in u u e echa ge
es ima es.
Impac S udies
Recha ge ins abili y endange s sus ainable wa e supplies and jeopa dizes ecosys ems and
li elihoods, especially hose o a me s using shallow aqui e s. Reduc ions in wa e able le els
ha e e e sed good ac ions aken o p o ec we lands and oases. Unplanned echa ge can cause
unin ended consequences and un a el du ing d ough s and wa e sca ci y c ises. Scien is s call o
immedia e ac ion ega ding g oundwa e managemen , ad oca ing o he adop ion o managed
aqui e echa ge (MAR) and clima e adap i e policies (Zhang e al., 2021).
21
Sub-seasonal and Seasonal Fo ecas ing
Regional clima e models can easonably iden i y season ape u e p ecipi a ion o ecas alues, bu
epea edly s uggle o p o ide an accu a e o ecas o g oundwa e echa ge, in ol ing p ecipi a ion
bias, p ecipi a ion iming, weak soil mois u e eedback, and p oblema ic inco po a ion o
g oundwa e modelling physically. Th oughou A ica, s andalone, eal- ime ime se ies
hyd oclima ic da a (e.g. soil mois u e, ege a ion indices, e apo anspi a ion) was success ully
employed in egional echa ge o ecas ing in selec sho - e m o ecas s. Whe e o om he e leads
o o ecas ing, modelling and u iliza ion o seasonal clima e p edic ions combined wi h land su ace
models o u he imp o e ye o be imminen d ough ea ly wa ning sys ems.
Conclusion
Signi ican obs acles o sus ainable wa e esou ce managemen in en i e A ica’s a id and semi-
a id egions due o unce ain y ega ding clima e-induced g oundwa e echa ge a e ongoing.
Though ecen p og ess using egional clima e models has e ealed some ac o s in luencing
echa ge dynamics, challenges emain o sub-seasonal and seasonal a iabili y. Remedia ion o
hese knowledge gaps will in ol e a mo e composi e app oach ha ex ends obse a ional ne wo ks,
downscaling app oaches, and he in eg a ion o socio-hyd ological p ocesses. Resea che s mus
also in es in hyb id modelling amewo ks ha combine p ocess-based unde s anding wi h da a-
d i en analysis o imp o e ac ionable g oundwa e echa ge o ecas s, a necessa y ac ion o
imp o e wa e secu i y o he mos ulne able popula ions in A ica.
Re e ences
● Ali, R., e al, (2022). Clima e change impac s on g oundwa e echa ge in A ica. Science o
he To al En i onmen , 159765. h ps://doi.o g/10.1016/j.sci o en .2022.159765
● E a is o, J., & Sa enije, H. H. G. (2022). Linking seasonal clima e o ecas s o g oundwa e
echa ge. Jou nal o Hyd ology: Regional S udies, 101076.
h ps://doi.o g/10.1016/j.ej h.2022.101076
● Taylo , R. G., e al, (2018). G oundwa e and clima e in A ica. Hyd ogeology Jou nal, 26,
1081-1094. h ps://doi.o g/10.1007/s10040-018-1898-8
● Zhang, Y., e al, (2021). Managed aqui e echa ge unde clima e change. Jou nal o
Hyd ology, 126602. h ps://doi.o g/10.1016/j.jhyd ol.2021.126602
22
S eam low Simula ion using WRF-Hyd o Model o e he Cau e y Ri e Basin
Aa i Soni and Anku S i as a a
Indian Ins i u e o T opical Me eo ology, Minis y o Ea h Sciences, D . Homi Bhabha Road, Pashan, Pune,
Maha ash a-411008, India
Email: aa i.soni@ opme . es.in
Backg ound
Unde s anding he hyd ological cycle is c ucial o managing wa e esou ces and assessing isks associa ed
wi h wa e , such as d ough and loods. (Liu e al., 2008). In his s udy he la ge-scale hyd ologic modeling
sys em WRF-Hyd o is used o assess he s eam low simula ion. This modeling amewo k was de eloped by
he Na ional Cen e o A mosphe ic Resea ch (NCAR) o simula e he mo emen o wa e h ough he land
su ace and i e channels (Gochis e al., 2018). WRF-Hyd o is he hyd ologic ex ension o Wea he Resea ch
Fo ecas ing (WRF) model, which is a nex gene a ion nume ical wea he p edic ion sys em. The model can be
un in wo modes s andalone and coupled wi h he WRF a mosphe ic model o unde s and eedback be ween
land and a mosphe e. I inco po a es key hyd ological p ocesses such as su ace and subsu ace uno , channel
ou ing, and land-a mosphe e in e ac ions, making i sui able o la ge-scale hyd ologic applica ion (i.e.,
s eam low o ecas ing, lood modeling, and wa e esou ces analysis). To es ima e land su ace luxes such
as sensible, la en hea , and ne adia ion in he 1-dimensional column land su ace model, WRF-Hyd o
p o ides a ious physics op ions. The land su ace model also es ima es soil mois u e dynamics in ou
di e en laye s, canopy in e cep ion, canopy anspi a ion and ain all pa i ioning a land in o su ace uno
and in il a ion. In he p esen s udy, s andalone WRF-Hyd o model is se -up wi h one domain o e he en i e
Cau e y basin bounda y (Figu e 1a). The pa en domain co e s he Sou he n pa o India wi h a spa ial
esolu ion o 5 km.
Figu e 1(a) S udy a ea showing ese oi s a ions ma ked as black do s om ups eam o downs eam
(1=Kudige, 2=Kollegal, 3=Bilungundulu, 4=U achiko ai, 5=Kodumudi, 6=Musu i) loca ed in sub-basins.
(b) Timese ies o daily s eam low es ima es (m3/s) o e six s a ions in he Cau e y Ri e basin o 2003.
Model Se up and P ep ocessing
In his s udy, GLDAS e sion 2.1 da a se s a e used as o cing da a du ing he pe iod o 2001-2022 in he
WRF-Hyd o model, which is a ailable wi h a spa ial esolu ion o 0.25 and a empo al esolu ion o 3 hou s.
GLDAS da a p o ided by he Na ional Ae onau ics and Space Adminis a ion (NASA) Godda d Space Fligh
S eamo de
(b)
(a)
23
Cen e (GSFC) Hyd ological Sciences Labo a o y (HSL) a e he combina ion o sa elli e and g ound-based
obse a ional da a p oduc s (Rodell e al., 2004). Me eo ological o cings o he model un, 3-hou ly da ase
a e p epa ed ai empe a u e, sho wa e and longwa e adia ion, su ace p essu e, speci ic humidi y, nea -
su ace U and V wind speed componen s and p ecipi a ion. To c ea e hyd ological g ids o he NOAH-MP
module, we use digi al ele a ion model (DEM) as e and o ecas poin s. In addi ion o me eo ological o cing
da ase s, he model also u ilizes soil p ope ies and gene al in o ma ion is a ailable a he NCAR. The spa ial
soil a iabili y da a a e p epa ed using R u ili y sc ip p o ided by he NCAR. Fu he mo e, s eam low in si u
da a is ob ained o 6 s a ions in he Cau e y Ri e Basin om he Cen al Wa e Commission (CWC), India.
Daily ain all obse a ions a 0.25° spa ial esolu ion a e acqui ed om India Me eo ology Depa men (IMD).
Fo se ing up he hyd ologic model domain, high- esolu ion e ain da a such as low di ec ion, low
accumula ion, and s eam ne wo ks de i ed om Digi al Ele a ion Model (DEM), land use land co e , soil
ype, and wa e shed bounda ies and s eamo de , a e p ocessed and esampled o ma ch he ou ing g id. To
c ea e he wa e shed delinea ion bounda ies, he la i ude and longi ude o each s a ion loca ed wi hin he basin
bounda y a e used. Fu he , he delinea ion ool iden i ies he ups eam a ea ha con ibu es uno o he
discha ge loca ion, he eby de ining he wa e shed bounda y. This con i ms ha he modeled wa e shed aligns
wi h he obse ed loca ion o accu a e s eam low compa ison and calib a ion.
Resul s
Figu e 1b shows he s eam low compa ison be ween WRF-Hyd o-simula ed s eam low (blue line) and
obse ed s eam low (g een line) ac oss six s a ions o he yea 2003. Resul s show ha ac oss all s a ions,
especially a he downs eam si es, he model consis en ly o e es ima es s eam low, wi h peak lows and
base lows bo h appea ing highe han obse ed alues. Lowe Cau e y Basin co e s mo e han 50% o o al
a ea o he basin o he o al a ea o he basin, which lies mainly in he Tamil Nadu s a e and ecei es ain all
mainly om he No heas Monsoon (Oc obe o Decembe ). In con as , uppe Cau e y basin loca ed in
Ka na aka, is p ima ily dependen on he Sou hwes Monsoon du ing June o Sep embe . The spa ial a ia ion
in ain all pa e ns impac s he hyd ological esponse o he basin. The ime se ies o s eam low obse a ion
du ing Oc obe o Decembe illus a es ha s eam low is compa a i ely highe a he s a ion p esen in he
lowe basin due o he ain all con ibu ion om he No heas Monsoon. The o e es ima ion in model
simula ion sugges s insu icien in il a ion o s o age in he model, o uncalib a ed ou ing pa ame e s ha
lead o excessi e uno . The model will be calib a ed o imp o e he accu acy o he s eam low p edic ions
and will be ul ima ely u ilized o subseasonal o seasonal (S2S) s eam low p edic ion using o cings de i ed
om ope a ional S2S o ecas s.
Re e ences
Rodell, M., House , P. R., Jambo , U. E. A., Go schalck, J., Mi chell, K., Meng, C. J., ... & Toll, D. (2004).
The global land da a assimila ion sys em. Bulle in o he Ame ican Me eo ological socie y, 85(3), 381-394.
Liu, C., Wang, Z., Zheng, H., Zhang, L., & Wu, X. (2008). De elopmen o hyd o-in o ma ic modelling sys em
and i s applica ion. Science in China Se ies E: Technological Sciences, 51(4), 456-466.
Gochis, D. J., Ba lage, M., Dugge , A., Fi zGe ald, K., Ka s en, L., McAllis e , M., e al. (2018). The WRF-
Hyd o modeling sys em echnical desc ip ion, e sion 0. In NCAR Technical No e (p. 107).
24
3. E alua ion and o ecas pe o mance
KIM4.0 demons a es o e all supe io pe o mance in global simula ions, as summa ized in he sco eca d in
Fig. 1. Speci ically, he e is an app oxima ely ~3.5% educ ion in e o o +5-day geopo en ial heigh o ecas s
a 500 hPa o e he no he n ex a opics, based on he mon hly a e age. Sligh deg ada ion is obse ed in he
sou he n hemisphe e, likely due o he disca ding o some ope a ional physics op imiza ions, which will be
add essed in u u e upg ades. In e ms o p ecipi a ion, KIM4.0 shows no able imp o emen s, pa icula ly in
simula ing s ong signals associa ed wi h hea y ain all e en s. This imp o emen is pa ially e lec ed in he
s a is ical p ecipi a ion skill sco es shown in Fig. 1b, whe e he new e sion ou pe o ms KIM3.9 ac oss all
p ecipi a ion h esholds. The enhanced simula ion pe o mance is especially e iden in e ain- ollowing
ea u es o e he Ko ean Peninsula, such as snow all along eas e n moun ain egions (no shown).
Cu en ly, he KMA plans o upg ade he nex e sion, ocusing on enhancemen s in da a assimila ion and
physics e isions. KIAPS is also de eloping a nex -gene a ion KIM wi h new con igu a ions, such as ~km
scale sho - ange p edic ion sys ems o egional domains and ex ended-p edic ion sys ems wi h coupled
models. The e ad ancemen s a e scheduled o elease a e he comple ion o he KIAPS’ second p ojec in
2026.
Figu e 1: Pe o mance e alua ion o KIM4.0 e sus KIM3.9. (a) sco eca d compa ing global RMSE agains
analysis o Janua y 2023 (00 UTC), wi h g een indica ing imp o emen and ed indica ing deg ada ion. (b)
F equency bias and equi able h ea sco e o 24-h accumula ed p ecipi a ion (+3-day p edic ion) o e Sou h
Ko ea and Asia, e i ied agains gauge obse a ions in Augus 2022.
Re e ences
Han, J.-Y. (2025). Impac o di e en scale-awa e cumulus pa ame e iza ions on p ecipi a ion o ecas s o e Ko ea.
A mos. Res., 317, 107990. doi:10.1016/j.a mos es.2025.107990.
Lee, E.-H., K.-H. Seol, H.-J. Pa k, S. Cho, K.-H. Cho, G. Lee, J. Jung, J. Lee, E. Lee, I.-J. Choi, J. Jang (2025) A
Fo ecas Ve i ica ion/E alua ion Sys em o De elopmen o he Global Wea he P edic ion Sys em: Ko ean
In eg a ed Model Analysis/E alua ion Tool (KAT), A mosphe e. Ko ean Me eo ological Socie y, 35(3), 395-412,
doi:/10.14191/A mos.2025.35.3.395 (in Ko ean wi h English Abs ac )
31

Upg ade o JMA’s Ope a ional Global Nume ical Wea he
P edic ion Sys em
KAWAGUCHI Masashi1, KINAMI Teppei1, KUROKI Yukihi o1, SHIMOKAWA Nao umi1, SUTOU Kouhei1,
YONEHARA Hi oshi1, and YOSHIMURA Hi omasa2
1Nume ical P edic ion Di ision, Japan Me eo ological Agency
2Me eo ological Resea ch Ins i u e, Japan Me eo ological Agency
Email: m.kaw[email p o ec ed]
In oduc ion
In Ma ch 2025, he Japan Me eo ological Agency (JMA) upg aded i s ope a ional global nume ical wea he p edic ion
sys em by in oducing a e ised e sion o i s Global Spec al Model (GSM). The e ision in ol ed e inemen s o
pa alleliza ion and da a s uc u e o educe compu a ional esou ce usage, and pa ame ized adia ion and land su ace
p ocesses, he eby imp o ing o ecas ing o e he p e ious e sion (Yoneha a e al. 2023), pa icula ly o he mid-
and high la i udes o he No he n Hemisphe e. This epo ou lines indi idual componen s o he upg ade and ela ed
e i ica ion esul s.
Main Upda es
Pa alleliza ion and Da a S uc u e
The wo-dimensional decomposi ion me hod adop ed o pa alleliza ion among MPI p ocesses was enhanced. The
p e ious GSM used di e en domain decomposi ions o semi-Lag angian ad ec ion scheme calcula ion and o he
g id-poin based calcula ions such as physical pa ame iza ion (JMA 2025). In he upda ed model, all g id-poin based
p ocesses a e compu ed in he same domain wi h ewe MPI communica ions. In associa ion, he decomposi ion
app oach in he da a assimila ion sys em was also enhanced. These upda es led o a educ ion in compu a ional
esou ces such as memo y usage. Addi ionally, he s uc u e o g id-poin model da a in he GSM was upg aded,
enabling mo e lexible adap a ion o compu ing a chi ec u e cha ac e is ics.
Radia ion
The globally uni o m clima ological annual mean used o ca bon dioxide concen a ion was upda ed om 396.0
ppm ( he 2013 obse a ion alue) o 417.9 ppm ( he 2022 alue)(WMO 2023). This educed empe a u e biases
in he s a osphe e h ough s eng hened cooling due o inc eased longwa e adia ion emissions. Radia ion scheme
calcula ions we e also op imized.
Land Su ace
The Lea A ea Index (LAI) mon hly clima ology was upda ed o a g idded e sion, as opposed o he p e ious a e age
o e opical, empe a e and bo eal la i udinal zones o each ege a ion ype. The upda ed da a a e de i ed om a
mo e ecen MODIS p oduc (Myneni e al. 2002) ea u ing highe esolu ion and g ea e accu acy han he e sion
used in he p e ious model (Yan e al. 2016 a, b). The upda e enables mo e ealis ic ep esen a ion o LAI dis ibu ion,
leading o a educ ion in empe a u e and ela i e humidi y biases in he lowe oposphe e h ough enhanced o ecas
accu acy o sensible and la en hea luxes.
Ve i ica ion esul s
The e ec s o pa alleliza ion enhancemen we e e alua ed h ough wo expe imen s un om a single ini ial condi ion.
Figu e 1 shows educed memo y usage o 132-hou o ecas s and 4D-Va .
The wo pe o mance e alua ion expe imen s (TEST o he upda ed sys em and CNTL o he p e ious one) co e ed
July o Sep embe 2023 (summe ) and Decembe o Feb ua y 2023/2024 (win e ). Figu e 2 shows e ical p o iles o
mean e o s (ME) in empe a u e o CNTL and TEST a e aged o e summe . Cold biases in he lowe oposphe e
and wa m biases in he s a osphe e we e educed, mainly due o he upda e o LAI mon hly clima ology and ca bon
dioxide concen a ion clima ology, espec i ely.
32
Re e ences
Japan Me eo ological Agency, 2025: Ou line o The Ope a ional Nume ical Wea he P edic ion a JMA. Japan Me e-
o ological Agency, Tokyo, Japan.
Myneni, R. B., S. Ho man, Y. Knyazikhin, J. L. P i e e, J. Glassy, Y. Tian, Y. Wang, X. Song, Y. Zhang, G. R.
Smi h, A. Lo sch, M. F iedl, J. T. Mo ise e, P. Vo a a, R. R. Nemani, and S. W. Running, 2002: Global p oduc s
o ege a ion lea a ea and ac ion abso bed PAR om yea one o MODIS da a. Remo e Sens. En i on.,83,
214–231.
WMO, 2023: The s a e o G eenhouse Gases in he A mosphe e Based on Global Obse a ions h ough 2022. WMO
G eenhouse Gas Bulle in,19, 1–11.
Yan, K., T. Pa k, G. Yan, C. Chen, B. Yang, Z. Liu, R. R. Nemani, Y. Knyazikhin, and R. B. Myneni, 2016a: E alua ion
o MODIS LAI/FPAR P oduc Collec ion 6. Pa 1: Consis ency and Imp o emen s. Remo e Sens.,8, 359.
Yan, K., T. Pa k, G. Yan, Z. Liu, B. Yang, C. Chen, R. R. Nemani, Y. Knyazikhin, and R. B. Myneni, 2016b:
E alua ion o MODIS LAI/FPAR P oduc Collec ion 6. Pa 2: Valida ion and In e compa ison. Remo e Sens.,8,
460.
Yoneha a, H., Y. Ku oki, M. Ujiie, C. Ma sukawa, T. Kanehama, R. Nagasawa, K. Ochi, M. Higuchi, Y. Ichikawa, R.
Sekiguchi, and S. Hi aha a, 2023: Upg ade o JMA’s Ope a ional Global Nume ical Wea he P edic ion Sys em.
WGNE Res. Ac i . Ea h Sys. Modell., 6.15-6.16.
Figu e 1: Maximum pe -node memo y usage [MiB] o
jobs (132-hou o ecas s (FcE , Tq959) and 4D-Va (Fd ,
Tl319)) among all compu ing nodes. Blue: CNTL; ed:
TEST.
Figu e 2: Ve ical p o iles o mean e o s o empe a-
u e [K] agains adiosonde obse a ions in he 20 – 90°N
egion o he summe expe imen . Blue: CNTL; ed:
TEST. Lines ep esen ME a di e en o ecas lead imes
o days 1 o 11.
33
In oduc ion o S ochas ic Humidi y P o ile o Con ec i e pa ame iza ion
(SHPC) me hod in JMA’s Global Ensemble P edic ion Sys em
OTA Yoichi o
Japan Me eo ological Agency
Email: yoichi o-o a@me .kishou.go.jp
1. In oduc ion
The Japan Me eo ological Agency (JMA) in oduced a new model ensemble scheme called he S ochas ic
Humidi y P o ile o Con ec i e pa ame iza ion (SHPC) in o i s Global Ensemble P edic ion Sys em (GEPS)
in Ma ch 2025. The SHPC pe u bs humidi y p o ile inpu o con ec i e pa ame iza ions o ep esen
unce ain ies in con ec i e ac i i y. I mi iga es he issue o unde -dispe si eness in he opics and makes i
possible o educe he excessi e ampli ude o ini ial pe u ba ions made om he Singula Vec o (SV) me hod
in he opics.
2. Fo mula ion and Se ings
Based on Tompkins and Be ne (2008), he SHPC pe u bs ela i e humidi y p o ile inpu o con ec i e
pa ame iza ion. Pe u ba ions a e gene a ed om andom pa e ns wi h ho izon al and empo al co ela ions
based on he combina ion o sphe ical ha monic unc ions such as he S ochas ic Pe u ba ion o Physics
Tendency (SPPT) scheme. The pe u ba ion ampli ude is a i s maximum a he lowes model le el, and is
exponen ially educed wi h he loga i hm o e ical p essu e le els. Figu e 1 shows an example o andom
pa e ns used o he SHPC. In he expe imen s desc ibed below, he maximum wa e numbe , empo al
co ela ion scale and e ical e- olding scale o pe u ba ions a e se as 20, 72 hou s and 0.8 (in loga i hm wi h
p essu e le els), espec i ely. The pe u ba ion s anda d de ia ion a he lowes model le el is app oxima ely
4%, and pe u ba ions a e unca ed a +/-10%. Pe u ba ion ampli ude is adjus ed so as no o induce
supe sa u a ion and nega i e humidi y in he inpu humidi y p o ile.
3. Expe imen s
To e i y sys em pe o mance o medium- ange o ecas s wi h lead imes o up o 11 days, e ospec i e
expe imen s co e ing pe iods o one mon h o mo e o summe 2021 (12UTC ini ials om 21 July 2021 o
11 Sep embe 2021) and win e 2021/22 (12UTC ini ials om 21 Decembe 2021 o 11 Feb ua y 2022) we e
conduc ed. The CNTL expe imen in ol ed he use o he ope a ional GEPS as o Ma ch 2024, wi h he numbe
o ensemble membe s educed o 13. In he TEST expe imen , he SHPC was applied o CNTL and he
ampli ude o ini ial pe u ba ions made om SV in he opics was educed o 60%. Figu e 2 shows he sp ead
o zonal wind speed a 250 hPa (U250) no malized using he ensemble mean RMSE in he opics (20°S –
20°N). The sp ead was educed by up o h ee o ou days and inc eased beyond ou days. The ini ial sp ead
educ ion obse ed was due o he lowe ini ial pe u ba ion ampli ude. The SHPC mi iga ed he unde -
dispe si eness o medium- ange o ecas s in he opics. Figu e 3 shows Con inuous Ranked P obabili y Sco es
(CRPSs) o U250 o he opics. P obabilis ic o ecas skill was enhanced wi h be e sp ead-skill ela ionships.
The e ec o he ex a- opics was smalle han ha o he opics (no shown).
Re e ences
Tompkins, A. M. and J. Be ne (2008). A s ochas ic con ec i e app oach o accoun o model unce ain y due o
un esol ed humidi y a iabili y. J. Geophys. Res. (113), D18101. DOI: 10.1029/2007JD009284.
34
Figu e 1: Example o a andom pa e n used o he SHPC.
Figu e 2: Sp ead o zonal wind speed a 250 hPa (U250) no malized using he ensemble mean RMSE o he
opics (20°S – 20°N) o summe 2021 (le ) and win e 2021/22 ( igh ). G een and ed lines show CNTL and
TEST, espec i ely.
Figu e 3: As pe Figu e 2, bu o CRPS agains analysis (uni : m/s). Pu ple lines and yellow iangles show CRPS
change a ios ((TEST-CNTL)/CNTL; %, igh axis) and s a is ically signi ican enhancemen a a 95% con idence
le el wi h boo s ap applica ion, espec i ely.
35
Upg ade o JMA’s Global Ensemble P edic ion Sys em
OTA Yoichi o*, OCHI Ken a, CHIBA Jo a o, OASHI Hi oaki, and TAKAKURA Toshina i
Japan Me eo ological Agency
Email: yoichi o-o a@me .kishou.go.jp
1. In oduc ion
The Japan Me eo ological Agency (JMA) upg aded i s Global Ensemble P edic ion Sys em (Global EPS) on
Ma ch 18 2025 o inco po a e a new model ensemble scheme, educed ampli ude o ini ial pe u ba ions in he
opics, ecen Global Spec al Model (GSM) de elopmen s and e ised sea su ace empe a u e (SST)
pe u ba ions.
2. Majo Upda es
(1) New model ensemble scheme
A new model ensemble scheme called he S ochas ic Humidi y P o ile o Con ec i e pa ame iza ion
(SHPC, O a (2025)) was in oduced o ep esen unce ain y in con ec i e ac i i y and mi iga e unde -
dispe si eness in he opics.
(2) Reduced ampli ude o ini ial pe u ba ions in he opics
The ampli ude o ini ial pe u ba ions made om Singula Vec o (SV) da a o he opics was educed o
60% o he p e ious le el.
(3) Recen GSM de elopmen s
The o ecas model was upg aded o a low- esolu ion e sion o he la es GSM (Kawaguchi e al. 2025).
The equi ed compu a ional esou ces a e educed wi h enhancemen s in MPI-based pa alleliza ion and
da a ans e .
(4) Re ised SST pe u ba ions
SST pe u ba ions we e e ised so ha he ensemble mean SST coincides wi h he unpe u bed SST o all
lead imes.
3. Ve i ica ion
To e i y sys em pe o mance o medium- ange o ecas s wi h lead imes o up o 11 days, e ospec i e
expe imen s co e ing pe iods o h ee mon hs o mo e in summe 2023 and win e 2023/24 we e conduc ed.
The esul s showed enhanced sp ead-skill ela ionships and con inuous anked p obabili y sco es (CRPSs),
especially in he opics (20°S – 20°N). Figu e 1 shows he sp ead no malized by he ensemble mean RMSE,
and CRPSs o 250 hPa zonal wind speed (U250) o he opics in summe . B ie skill sco es o p ecipi a ion
o ecas s in Japan we e also enhanced o summe (no shown).
Hindcas expe imen s we e also conduc ed o he 30-yea pe iod om 1991 o 2020 wi h a mosphe ic ini ial
condi ions om he la es Japanese eanalysis (JRA-3Q; Kosaka e al. 2024). Reduced ampli ude o ini ial
pe u ba ions and implemen a ion o he SHPC mi iga e excessi e and insu icien ensemble sp eads,
espec i ely (Figu e 2). MJO o ecas skill was almos neu al o all lead imes (no shown).
36

Figu e 1: Sp ead no malized by he ensemble mean RMSE (le ) and CRPS ( igh ; uni : m/s) o zonal wind speed
a 250 hPa (U250) in he opics (20°S – 20°N) o summe 2023. G een and ed lines ep esen e i ica ion esul s
o he p e ious (CNTL) and new (TEST) Global EPS. The pu ple line and yellow iangles in he panel on he
igh ep esen sco e change a ios ([TEST-CNTL]/CNTL, igh axis; uni : %) and s a is ically signi ican
enhancemen s a a 95% con idence le el wi h boo s ap applica ion, espec i ely.
Figu e 2: Ensemble sp ead (le ; uni : x106 m2/s) and sp ead no malized by he ensemble mean RMSE ( igh )
o eloci y po en ial a 200 hPa in he opics (20°S – 20°N) o win e 1991 – 2020. Black and ed lines
ep esen e i ica ion esul s o he p e ious (CNTL) and new (TEST) Global EPS, espec i ely.
Re e ences
Kawaguchi, M., T. Kinami, Y. Ku oki, N. Shimokawa, K. Su ou, H. Yoneha a, and H. Yoshimu a (2025). Upg ade
o JMA’s Ope a ional Global Nume ical Wea he P edic ion Sys em. Res. Ac i . Ea h Sys. Modell. (55), submi ed.
Kosaka, Y., S. Kobayashi, Y. Ha ada, C. Kobayashi, H. Naoe, K. Yoshimo o, M. Ha ada, N. Go o, J. Chiba, K.
Miyaoka, R. Sekiguchi, M. Deushi, H. Kamaho i, T. Nakaegawa, T. Y. Tanaka, T. Tokuhi o, Y. Sa o, Y. Ma sushi a,
and K. Onogi (2024). The JRA-3Q eanalysis. J. Me eo . Soc. Japan, 102, 49-109,
h ps://doi.o g/10.2151/jmsj.2024-004.
O a, Y. (2025). In oduc ion o S ochas ic Humidi y P o ile o Con ec i e pa ame iza ion (SHPC) me hod in
JMA’s Global Ensemble P edic ion Sys em. Res. Ac i . Ea h Sys. Modell. (55), submi ed.
37
Ope a ional Use o AMSU-A and ATMS Window Channels in JMA’s NWP
Sys ems
URATA Tomoya and SHIMIZU Hi oyuki
Japan Me eo ological Agency
Email: [email p o ec ed]
1. In oduc ion
Radiance da a collec ed by sa elli e-based mic owa e sounde s mainly p o ide in o ma ion on he e ical
dis ibu ion o a mosphe ic empe a u e and humidi y, as assimila ed in JMA’s Global Analysis (GA),
Meso-scale Analysis (MA) and Local Analysis (LA) sys ems. AMSU-A and ATMS mic owa e sounde s
a e equipped wi h window channels ha ha e highe ansmi ance in he a mosphe e han sounding
channels and a e sensi i e o humidi y a lowe - oposphe ic le els. These channels a e used only o
Quali y Con ol (QC), bu a e conside ed capable o con ibu ing o mo e accu a e wea he o ecas by
enhancing analysis humidi y ields.
2. Me hodology
2.1. Channel Selec ion
In his s udy, window channels wi h equencies o 23.8 GHz (AMSU-A/ch1 and ATMS/ch1) and 31.4
GHz (AMSU-A/ch2, ATMS/ch2) we e assimila ed. 89 GHz channels (AMSU-A/ch15 and ATMS/ch16)
a e also sensi i e o humidi y a lowe - oposphe ic le els, bu we e no assimila ed because o hei highe
sensi i i y o ice clouds and p ecipi a ion, which complica es QC p ocessing.
2.2. QC
A clea -sky me hod was applied in which only da a una ec ed by clouds o p ecipi a ion a e assimila ed.
The same cloud de ec ion scheme used o channels sensi i e o empe a u e a lowe - oposphe ic le els
(Okamo o e al., 2005) using e ie ed cloud liquid wa e and sca e ing index alues was implemen ed.
Only da a om a eas o e oceans a e assimila ed in conside a ion o he g ea e unce ain y o su ace
emissi i y and empe a u e in he i s guess om a eas o e land and sea ice. Da a om a eas nea land
o sea ice a e no used in o de o a oid con amina ion o such su aces wi hin he ocean-based obse a ion
oo p in . Da a a he edge o scans whe e he obse a ion oo p in is la ge a e also excluded o a oid
con amina ion.
2.3. Obse a ion E o s
Obse a ion e o s a e de ined o each sa elli e based on he s anda d de ia ion o O-B (Obse a ion minus
Backg ound) calcula ed om 15 days o GA s a is ics (Table 1). These alues a e in la ed in analysis
sys ems because in e -channel and spa ial obse a ion e o co ela ions a e no conside ed in assimila ion.
E o in la ion ac o s a e se o ma ch hose o mic owa e image channels whose equencies a e close
o hose o he window channels (3.0 o GA, 4.0 o MA and 6.0 o LA).
3. NWP E ec s and Resul s
Nume ical expe imen s we e conduc ed o e alua e he e ec s o window channels on he global,
meso-scale and local NWP sys ems. He e, he esul s o expe imen s in ol ing he meso-scale NWP
sys em o July 2023 and Janua y 2024 a e epo ed.
Window channel assimila ion enhanced consis ency be ween he i s guess and humidi y-sensi i e
obse a ions such as AMSR2 and he humidi y channels o ATMS and C IS (Fig. 1), which implies
be e i s -guess accu acy. The alse ala m and miss a ios compa ed o ada / aingauge-analyzed
p ecipi a ion da a we e educed o ela i ely weak p ecipi a ion o ecas s up o 5 mm/3 h o he
summe pe iod. Se e al case s udies also indica ed ha addi ional in o ma ion on lowe - oposphe ic
38
humidi y p o ided by he window channels o AMSU-A and ATMS nea p ecipi a ion a eas
con ibu ed o be e o ecas accu acy (Fig. 2). Simila enhancemen s we e obse ed in he global
and local NWP sys ems.
Based on hese esul s, assimila ion o he window channels o AMSU-A and ATMS was adop ed in MA
and LA in Ma ch 2025, and is scheduled o adop ion in GA in au umn 2025.
Sa elli e/Senso
ch1
ch2
Me op-C/AMSU-A
2.80
2.40
NOAA-15/AMSU-A
3.00
2.60
NOAA-18/AMSU-A
2.80
2.40
NOAA-19/AMSU-A
2.80
2.40
Suomi-NPP/ATMS
2.80
2.20
NOAA-20/ATMS
3.00
2.70
NOAA-21/ATMS
3.00
2.70
Re e ences
Okamo o, K., Kazumo i, M., and Owada, H. (2005). The assimila ion o ATOVS adiances in he JMA
Global Analysis Sys em. J. Me eo . Soc. Japan, 83, 201-217. h ps://doi.o g/10.2151/jmsj.83.201
Table 1: Window channel obse a ion
e o se ings [K]
Fig. 1: Ra io o change [%] in he s anda d de ia ion o O-
B esul ing om window channel assimila ion in MA. (a)
Mic owa e image s: GMI, AMSR2 and SSMIS. (b) C IS
hype spec al in a ed sounde . The ed and g een lines
ep esen expe imen esul s o July 2023 and Janua y
2024, espec i ely.
Fig. 2: 3-hou cumula i e p ecipi a ion [mm]. (a) Fo ecas wi hou window channels. (b) Fo ecas wi h
window channels. (c) Fo ecas di e ence esul ing om window channel assimila ion: (b) – (a). (d)
Rada / aingauge-analyzed p ecipi a ion da a. The o ecas ime is 3 hou s om 15 UTC on 7 h July 2023.
39
Implemen ing Isocho ic-Isoba ic T ans o ma ions in he MCV Model's
Dynamic-Physical Coupling
Dong Chen 1, Xingliang Li 1*, Xueshun Shen 1
1 CMA Ea h Sys em Modeling and P edic ion Cen e (CEMC), Beijing 100081, China
*Email: [email protected] .cn, che[email p o ec ed]o .cn
1. In oduc ion
Dynamic-physical coupling in a mosphe ic models enables accu a e cha ac e iza ion o
mul i-scale a mosphe ic in e ac ions h ough he in eg a ion o dynamic and physical
p ocesses. The design o he coupling s a egy di ec ly in luences ene gy conse a ion, he
e iciency o physical eedbacks, and he coo dina ion o mul i-scale in e ac ions. The
accu acy o coupling p ocesses u he a ec s he enhancemen o simula ion and he s abili y
o un ime pe o mance (G oss e al., 2018). Beljaa s e al. (2004) demons a ed ha
imp o ed nume ical coupling mechanisms wi hin dynamic-physical in e ac ions lead o a
educ ion in p edic ion e o s. The nex - gene a ion egional/global uni ied nonhyd os a ic
Mul i-momen Cons ained ini e Volume (MCV) model by CMA (Li e al., 2013; Chen e al.,
2014; 2023), achie es e icien dynamic-physical coupling ia inno a i e amewo k design.
Howe e , cu en MCV model s ill exhibi s inhe en biases in isocho ic-isoba ic coo dina e
he modynamic con e sion wi hin he coupling p ocess, which may esul in imbalances in
empe a u e endencies, biases in p ecipi a ion simula ions, and educed p edic abili y sco es.
To add ess his issue, his s udy aims o implemen isocho ic-isoba ic ans o ma ions wi hin
he dynamic-physical coupling s a egy o he MCV model.
2. Me hod and esul s
2.1 Isocho ic-isoba ic coupling me hod
The physics-dynamics coupling in MCV model di e s om many exis ing a mosphe ic
models in ha i is pe o med a cons an olume (densi y) a he han cons an p essu e.
Howe e , he physical pa ame e iza ion schemes is buil a p essu e-based coo dina es.
The modynamically, his co esponds o he dynamic co e being associa ed wi h an isocho ic
p ocess and he physical p ocesses wi h an isoba ic p ocess. Du ing dynamics-physics
coupling, he upda e o empe a u e endencies mus inco po a e he con e sion be ween hese
wo dis inc he modynamic p ocesses o ensu e hea conse a ion. P e iously, he MCV
model’s coupling scheme lacked conside a ion o his con e sion, esul ing in biases in he
po en ial empe a u e ed back o he dynamic co e ollowing physical p ocess compu a ions.
This, in u n, led o sys ema ic unde es ima ion o o al and g id-scale p ecipi a ion in opical
egions. To add ess his issue, we edesigned he empe a u e con e sion based on ene gy
conse a ion and he modynamic p inciples. Remembe ha o al hea mus be s ic ly
conse ed, hen he dynamical empe a u e endency ha ed back om physics becomes
∆Qdyn=∆Qphy ⟹∆Tdyn=cp
c m
∆Tphy (1)
whe e Q ep esen s he hea pe mass, Tis empe a u e, cpand c ma e speci ic capaci y a
cons an p essu e and mois speci ic capaci y a cons an olume, espec i ely.
2.2 Resul s and analysis
Case simula ions implemen ing he dynamics-physics con e sion o mula ion (1)
demons a e ha , unde iden ical esolu ion and ini ial condi ions, o al p ecipi a ion and
g id-scale p ecipi a ion exhibi enhanced physical easonableness (Fig. 1 a), wi h biases in
empe a u e and humidi y h oughou he a mosphe ic column signi ican ly educed ( igu e
40
A Simpli ied Lake Model o Seasonal P edic ion
ADACHI Yukimasa, OCHI Ken a, HIRAHARA Shoji, KUBO Yu a o, SEKIGUCHI Ryohei
Me eo ological Resea ch Ins i u e / Japan Me eo ological Agency, Tsukuba, Japan
Email: yadachi@m i-jma.go.jp
1. Abs ac
The ep oducibili y o in e annual a ia ions in he lowe a mosphe e and he condi ions o land su aces wi h
which i in e ac s a e impo an in seasonal p edic ion. To enhance ela ed da a o e p e ious JMA seasonal
p edic ion sys ems (which elied on mon hly clima ology o lake- ela ed a iables such as su ace wa e
empe a u e and ice concen a ion), a lake model was in oduced in o he sys em (CPS3; Hi aha a e al. 2023)
and modi ied in he nex i e a ion (CPS4; Kubo e al. 2025). As a iabili y in ai empe a u e abo e lakes is
high in win e and closely associa ed wi h ice p esence, pa icula emphasis is placed on he ea men o lake
ice. Acco dingly, he lake model p edic s only lake su ace condi ions and lake ice a iables equi ed as he
lowe bounda y o he a mosphe ic model, and is highly e icien in ope a ional usage. Since he e is no en i e-
lake a ge ing, lake dep h da a a e no equi ed; only mon hly clima ology o su ace empe a u es (which a e
based on sa elli e obse a ion and he e o e conside ed highly accu a e) a e needed. The model helps o make
da a on he ampli ude o in e annual a ia ions in su ace ai empe a u e o e lakes mo e ealis ic.
2. Lake Model O e iew
Based on e ical column he modynamics, he model p edic s lake ice o ma ion and a ia ions in wa e
empe a u e h oughou wa e phase changes and hea ans e among wa e , ice, and snow. I consis s o h ee
laye s o lake wa e , ou laye s o lake ice and one laye o snow on lake ice. Wa e , ice and snow densi ies
a e cons an , and he wa e empe a u e in he bo om ( hi d) laye is se as a cons an annual-mean clima ology
alue because lowe wa e empe a u e changes li le h oughou he yea . The ene gy o wa e ’s uppe wo
laye s, lake ice and snow on lake ice is conse ed, bu he a iabili y o lake wa e mass is no aken in o
accoun . The hickness o lake wa e laye is cons an . The he mal di usion coe icien be ween he i s and
second laye s is se o be globally uni o m as he in e se o an a bi a y elaxa ion ime. This ime be ween he
second and hi d laye s depends on he obse ed ampli ude o he seasonal cycle in su ace empe a u e, wi h
sho e elaxa ion imes associa ed wi h shallow lake condi ions. The seasonal ampli ude is de ined as he
di e ence be ween he maximum and minimum clima ology mon hly mean su ace empe a u e based on
MODIS/Te a Land Su ace Tempe a u e da a (MOD11C3_ 006; Wan, Z. e al, 2015). The bo om wa e
empe a u e is se o a p opo ional alue be ween he annual mean o he obse ed su ace empe a u e and
he empe a u e whe e he densi y o wa e eaches i s maximum (3.98°C), as bo om wa e empe a u e is
close o he annual minimum in non- ozen lakes and he maximum wa e densi y empe a u e in ozen lakes.
The model uses a ixed wa e densi y, and he annual mean ( a he han he annual minimum) is applied. Thus,
i ac s in a pendulum-like manne , unlike he he mos a beha io obse ed a ound he annual minimum alue,
o 3.98°C in na u e. The numbe o lake ice laye s is cons an , and he uppe and lowe limi s o ice hickness
a e se indi idually o keep he uppe laye s hinne han he lowe ones. Lake ice concen a ion and hickness
a y independen ly du ing eezing and mel ing, espec i ely, enabling ep esen a ion o bo h wide/ hin and
na ow/ hick condi ions.
47

3. CPS4 Enhancemen s
In CPS3, a lowe limi o 0°C was se o he obse ed mon hly mean empe a u e o calcula ion o annual
mean and seasonal ampli udes. This limi is emo ed in CPS4, esul ing in la ge lake ice indica ions due o
he lowe annual mean empe a u e and la ge seasonal ampli ude in ozen lakes. Bo om wa e empe a u e
has also been modi ied o he a e age o he annual mean and annual minimum, esul ing in highe wa e
empe a u e indica ions in ice- ee lakes such as he Caspian Sea.
4. Resul s
Figu e 1 shows seasonal changes in in e annual su ace ai empe a u e a ia ion on he G ea Lakes.
Va iabili y is la ge o No . – Ma . in eanalysis (JRA-55; Kobayashi e al. 2015). Applica ion o he lake
model enables app op ia e lake ice ep esen a ion, wi h ampli udes o in e annual a ia ion much close o
hose o JRA-55 han hose o clima ology lake ice. Figu e 2 shows seasonal su ace ai empe a u e biases
ela i e o eanalysis (JRA-3Q; Kosaka e al. 2024) and he e ec s o model enhancemen . Wa m biases o
No h Ame ica and Sibe ia in bo eal win e and cold biases in he Caspian Sea a e educed.
Re e ences
Hi aha a, S. e al, 2023: Japan Me eo ological Agency/Me eo ological Resea ch Ins i u e Coupled P edic ion Sys em e sion 3
(JMA/MRI-CPS3). J. Me eo . Soc. Japan, 101, 149–169.
Kobayashi, S. e al, 2015: The JRA-55 eanalysis: Gene al speci ica ions and basic cha ac e is ics. J. Me eo . Soc. Japan, 93, 5–18.
Kosaka, Y. e al, 2024: The JRA-3Q eanalysis. J. Me eo . Soc. Japan, 102, 49–109, doi:10.2151/jmsj.2024-004.
Kubo, Y. e al, 2025: Upg ade o he JMA Sub-Seasonal and Seasonal Ensemble P edic ion Sys em (JMA/MRI-CPS4). WGNE Res.
Ac i . Ea h Sys. Model., submi ed.
Wan, Z. e al, 2015: MOD11C3 MODIS/Te a Land Su ace Tempe a u e/Emissi i y Mon hly L3 Global 0.05Deg CMG V006. NASA
Land P ocesses Dis ibu ed Ac i e A chi e Cen e , doi:10.5067/MODIS/MOD11C3.006.
(a)
(b)
Figu e 1: Mon hly in e annual
a ia ion o su ace ai empe a u e
on he G ea Lakes in (a)
clima ology lake ice and (b) he lake
model. Colo ed lines ep esen
model esul s o ini ial da a, and
black lines ep esen JRA-55.
(a)
(c)
(b)
(d)
Figu e 2: (a)
Clima ological mean
ields o su ace ai
empe a u es (con ou s)
and biases ela i e o
JRA-3Q (shading) wi h
he cu en lake model,
and (b) e ec s o
enhancemen s in CPS4
o bo eal summe . (c)
and (d): as pe (a) and (b)
bu o bo eal win e .
48
Fo ecas e i ica ion;
no el me hodologies
o diagnose and
measu e sys ema ic
e o s
5
49
Powe Spec a Ve i ica ion o Machine Lea ning Wea he P edic ion
Ba ba a Casa i, Leo Sepa o ic, Syed Z. Husain
Me eo ological Resea ch Di ision, En i onmen and Clima e Change Canada
Email: [email p o ec ed]
E icien ye meaning ul e i ica ion o compa ing machine lea ning wea he p edic ion (MLWP) agains adi ional physics-
based nume ical wea he p edic ion (NWP) is c ucial o scien i ic c edibili y and ope a ional up ake. T adi ional summa y
s a is ics, like MSE, bias, a iance o co ela ions alone, can be hedged by smoo he o ecas s o clima ology-induced alse skill
(Hamill and Ju as, 2006). Mo eo e , adi ional me ics can mask unphysical beha io s and canceling e o s. Powe spec a
diagnos ics ha e been ecen ly le e aged, o enable scale-awa e physically meaning ul e alua ion o bo h MLWP and NWP
models (e.g. Husain e al, 2025; Rodwell e al, 2025). This no e wishes o highligh some o he key ea u es o he powe -spec a
diagnos ic app oaches, wi h ecommenda ions o ope a ional e i ica ion p ac ices.
The diagnos ic capabili ies o he powe -spec a e i ica ion s a is ics a e illus a ed o global o ecas s om he En i onmen
and Clima e Change Canada’s (ECCC) ope a ional NWP-based Global De e minis ic P edic ion Sys em (GDPS) and wo o he
ECCC expe imen al sys ems: he GEML (Global En i onmen al eMuLa o ) MLWP model, and he hyb id NWP-MLWP sys em
based on Spec al Nudging (GDPS-SN). P edic ions om all hese sys ems a e e i ied agains he ope a ional GDPS analysis.
Figu e 1: (Le ) Fo ecas ac i i y a io as a unc ion o lead ime and ( igh ) 120-h spec al ac i i y a io o 500-hPa geopo en ial
heigh anomalies o (black) he GDPS analysis, (blue) GDPS, ( ed) GDPS-SN and (g ay) GEML, a e aged o e 60 cases o
JFM 2022.
Clima ology and anomalies: A mosphe ic ields om he global o ecas s and e i ying analysis a e sepa a ed in o ansien
anomalies and s a iona y clima ology, by sub ac ing he ERA5 30-yea clima ology. This enables ocusing on wea he
phenomena and emo ing a i icial skill due o pe sis en clima ological ea u es.
Powe Spec al Decomposi ion: G idded a mosphe ic ields a e decomposed spec ally using unca ed sphe ical ha monic
expansions (e.g., up o n=800 o ~0.25˚ esolu ion). This spec al ep esen a ion enables e alua ion as a unc ion o o al
Figu e 2: Spec al ac i i y a io o 72-h sc een-le el
empe a u e o GDPS (blue) and GEML ( ed), a e aged
o e 62 cases o JJA 2022. The a io is shown o he
s a iona y JJA 2022 sample clima ology (do ed),
ansien anomalies (dashed) and he o al/in eg al
sc een-le el empe a u e ields (solid).
50
wa enumbe (n), co esponding o physical scales (L = 2πR/n). Such scale decomposi ion is c i ical o de e mining he o ecas
"e ec i e esolu ion" and diagnosing compensa ing e o s a di e en scales. Fu he mo e, spec al decomposi ion pe mi s he
assessmen o bias, e o , and skill a di e en scales sepa a ely (e.g. Casa i e al, 2023).
In e i ica ion, bias and accu acy/skill a e key and complemen a y a ibu es o he o ecas quali y: he bias compa es o ecas -
e sus-obse a ion ma ginal dis ibu ions, whe eas he accu acy pe ains o he o ecas -obse a ion join dis ibu ion (Jolli e
and S ephenson, 2012). In he powe spec a amewo k, hese a ibu es can be assessed by compu ing o ecas and analysis
a iances and hei co a iance.
Ac i i y and cohe ence: he ac i i y, de ined as he s anda d de ia ion o he anomalies, is associa ed wi h he amoun (and
in ensi y) o ac i e wea he (Ben Bouallègue e al, 2024). The o ecas e sus analysis ac i i y a io is a measu e o he ma ginal
bias in he o ecas - e sus-analysis a iabili y, o o he ampli ude e o o he sphe ical ha monics. The anomaly co ela ion
(named cohe ence o he powe spec a) can p o ide a complemen a y measu e o accu acy/skill (o phase e o in spec al
space). When e alua ed o he spec al componen s, ac i i y a io and cohe ence can assess scale-dependen bias/ampli ude
e o s and accu acy/skill o each spa ial scale, sepa a ely. While in his sho no e we illus a e solely he ac i i y a io, measu es
o bo h bias and accu acy/skill should always be e alua ed in andem, o p o ide a mo e comple e assessmen o o ecas
pe o mance (e.g. Figu e 2 o Husain e al, 2025).
Figu e 1 shows ha he seemingly almos neu al bias o GEML when he ac i i y a io is e alua ed on he anomaly ields only
(le ), is a ec ed by unphysical compensa ing e ec s on meso-α and meso-β scales, which a e de ec ed solely when using powe -
spec al diagnos ics ( igh ). Figu e 2 shows ha he spec al ac i i y a io cap u es GEML a iance de iciency o mesoscales,
ypical o he o e ly smoo h ML model ou pu s. Mo eo e , he s a iona y e sus ansien signal sepa a ion e eals ha GEML
esol es well he clima ological (s a iona y) a iabili y, bu ha his compensa es o a la ge nega i e bias associa ed wi h he
ansien anomalies. This is an expec ed beha io o any MLWP model ained o minimize MSE, which leads o ine-scale
smoo hing o educe he “double penal y” e ec (Subich e al, 2025).
In conclusion, o ope a ional e i ica ion p ac ices we ecommend: i) compu ing bo h ac i i y a io and anomaly co ela ion o
assess ma ginal bias/ampli ude e o and o ecas skill/accu acy; ii) sepa a e s a iona y clima ology om ansien anomalies, o
dis inguish skill in ep oducing clima e e sus ac ual wea he ; iii) apply spec al decomposi ion o analyze sepa a ely he dis inc
model beha io s a di e en scales, o de ec unphysical e o compensa ion, and o assess he “e ec i e esolu ion”, pa icula ly
o de e minis ic ML models ha a e p one o smoo hing.
Re e ences
Ben Bouallègue, Z., M.C. A. Cla e, L. Magnusson, E. Gascón, M. Maie -Ge be , M. Janoušek, M. Rodwell, F. Pinaul , J. S. D amsch,S.T. K.
Lang, Baudouin Raoul , F. Rabie , M. Che allie , I. Sandu, P. Dueben, M. Chan y, F. Pappenbe ge , 2024: “The Rise o Da a-D i en Wea he
Fo ecas ing”. Bulle in o he Ame ican Me eo ological Socie y, E864-E883, h ps://doi.o g/10.1175/BAMS-D-23-0162.1
Casa i, B., C. Lussana, and A. C espi, 2023: “Scale-sepa a ion diagnos ics and he symme ic bounded e iciency o he in e compa ison o
p ecipi a ion eanalyses”. In . J. Clima ol., 43, 2287–2304, h ps://doi.o g/10.1002/joc.7975
Hamill, T., and Ju as J. , 2006: “Measu ing o ecas skill: Is i eal skill o is i a ying clima ology?” Qua . J. Roy. Me eo . Soc., 132 , 2905–
2923, h ps://doi.o g/10.1256/qj.06.25
Husain, S.Z., L. Sepa o ic, J-F Ca on, R. Aide , M. Buehne , S. Chambe land, E. Lapalme, R. McTagga -Cowan, C. Subich, P.A. Vaillancou ,
J. Yang, A. Zad a. 2025: “Le e aging Da a-D i en Wea he Models o Imp o ing Nume ical Wea he P edic ion Skill h ough La ge-Scale
Spec al Nudging”. Wea he & Fo ecas ing 40, n. 9: 1749-1771, h ps://doi.o g/10.1175/WAF-D-24-0139.1
Jolli e, I.T. and S ephenson, D.B. (2012) Fo ecas Ve i ica ion: A P ac i ione ’s Guide in A mosphe ic Science. 2nd Edi ion, Wiley-Blackwell,
Ox o d, ISBN:9781119960003Rodwell, M. J., M. C. A. Cla e, S.-J. Lock, K. Loni z, and M. Che allie , 2025: “ Powe Spec a o Physics-
Based and Da a-D i en Ensembles.” Me . Apps. 32, no. 5: e70071. h ps://doi.o g/10.1002/me .70071.
Rodwell, M. J., M. C. A. Cla e, S.-J. Lock, K. Loni z, and M. Che allie , 2025: “ Powe Spec a o Physics-Based and Da a-D i en
Ensembles.” Me . Apps. 32, no. 5: e70071. h ps://doi.o g/10.1002/me .70071.
Subich, C., Husain, S.Z., Sepa o ic, L., and Yang, J. (2025): Fixing he double penal y in da a-d i en wea he o ecas ing h ough a modi ied
sphe ical ha monic loss unc ion. h ps://a xi .o g/abs/2501.19374
51
Uncoupled and
coupled da a
assimila ion o
in eg a ed ea h
sys em analysis
and p edic ion;
me hodology and da a
impac sensi i i y
s udies
6
52

Upda e o he Radia i e T ans e Model o RTTOV-13 in JMA's NWP Sys ems
SHIMIZU Hi oyuki, ANDO Aki a, KAMEKAWA No io, KONDO Keiichi, KUSANO Nao o,
MURATA Hidehiko, TOYOKAWA Masakazu, URATA Tomoya
Japan Me eo ological Agency
E-mail: shimizu_[email p o ec ed].go.jp
1. In oduc ion
Va ious ypes o sa elli e adiance da a a e assimila ed o de e mine accu a e ini ial condi ions o
nume ical wea he p edic ion (NWP) models. The RTTOV as adia i e ans e model (Ey e 1991),
de eloped by he EUMETSAT Sa elli e Applica ion Facili y on Nume ical Wea he P edic ion (NWP
SAF), is used as he obse a ion ope a o o assimila ing sa elli e adiance da a in JMA’s NWP sys ems
(Global, meso-scale and local NWP). As a i s s ep in upda ing RTTOV10.2 (Saunde s e al. 2012) o
RTTOV-13 (Saunde s e al. 2020), he minimal changes necessa y o ensu e secu e implemen a ion wi h
no signi ican impac on NWP accu acy (such as changes o modules and ele an ile o ma s) we e made
in 2022. The op ical dep h coe icien iles ( e e ed o he e simply as “coe icien iles”) and sea su ace
emissi i y models in RTTOV we e subsequen ly upda ed o he new e sions, and ele an quali y con ol
p ocedu es we e adjus ed acco dingly. This epo b ie ly summa izes he e ec s o he changes on NWP
sys ems.
2. Modi ica ion o Quali y Con ol P ocedu es
The coe icien iles o senso s al eady in use we e upda ed o Ve sion 13 o he han o C IS, which
showed no enhancemen om coe icien ile upda ing.
Sea su ace emissi i y models (FASTEM-6 (Kazumo i and English 2015) o mic owa e obse a ion
and IREMIS o in a ed obse a ion) a e now a ailable in RTTOV13. FASTEM-6 models he dependence
o sea su ace emissi i y on ela i e wind di ec ion (RWD) mo e app op ia ely. Global analysis indica ed
smalle O-B (obse a ion minus backg ound; i.e., he di e ence be ween obse ed and calcula ed
b igh ness empe a u e) biases in mic owa e image s dependen on RWD. Howe e , biases dependen on
su ace wind speed we e la ge o su ace-sensi i e channels o AMSU-A and ATMS. Acco dingly,
su ace wind speed was added as a p edic o in a ia ional bias co ec ion (Va BC) o ch4 and ch5 o
AMSU-A and ch6 o ATMS. FASTEM-6 was no adop ed in meso-scale and local analysis because no
enhancemen was obse ed in ela ion o a FASTEM upda e om e sion 4 o 6 and addi ion o su ace
wind speed as a p edic o in Va BC.
The s a is ical cha ac e is ics o calcula ed b igh ness empe a u e changed as a esul o he upda es o
he coe icien iles and he sea su ace emissi i y models desc ibed abo e. The s a ic scan bias co ec ion
alue and he p ecipi a ion sc eening pa ame e s o AMSU-A, which we e es ima ed in ad ance om
s a is ics on obse ed and calcula ed b igh ness empe a u e, we e upda ed o he new calcula ion
cha ac e is ics. Re ie ed cloud wa e amoun s o AMSU-A quali y con ol a e also la ge han be o e,
mainly due o he upda ing o scan bias co ec ion alues. As a esul , he amoun o da a conside ed o
ep esen cloud inc eased, and he amoun o da a used dec eased. In e e ence o O-B s a is ics, he
h eshold o cloud wa e con en used o cloud de ec ion was inc eased om 100 o 120 g/m2 o make he
numbe o da a used compa able o ha be o e he upda e.
3. E ec s o Upda ing Coe icien Files and Sea Su ace Emissi i y Models on NWP
Sys ems
Da a assimila ion expe imen s we e conduc ed o e alua e he e ec s o upda ing he coe icien iles
and sea su ace emissi i y models o he new e sions in JMA’s NWP sys ems. This epo de ails he
esul s o expe imen s in he global NWP sys em o he pe iods om June o Oc obe 2023 and No embe
2023 o Ma ch 2024. The con ol expe imen s (CNTL) had he same con igu a ion as JMA’s ope a ional
global NWP sys em as o Ma ch 2025, and he es expe imen s (TEST) we e as pe CNTL excep o he
modi ica ions desc ibed in Sec ion 2.
Figu e 1 shows changes in he s anda d de ia ion and mean o FG depa u es (ano he exp ession o O-
B) o adiosonde empe a u e agains CNTL. Those o TEST we e close o obse a ions han CNTL,
which indica es an enhanced i s guess o he empe a u e ield. Meanwhile, nega i e biases in 500 hPa
53
geopo en ial heigh agains adiosonde obse a ions o e he opics in CNTL showed a sligh ly inc eased
endency in TEST (Figu e 2). This change co esponded closely o ha in he O-B o ATMS ch6 (no
shown), which was likely he esul o a bug ix a ec ing senso s ha ha e channels in luenced by he 184
GHz ozone abso p ion line (Saunde s e al. 2017). The esul s o p ecipi a ion o ecas s and opical
cyclone ack and in ensi y o ecas s we e almos neu al. Expe imen s wi h meso-scale and local NWP
sys ems showed no signi ican changes in o ecas accu acy.
4. Summa y
Upda ing o he coe icien iles and sea su ace emissi i y models o RTTOV o he new e sion in
JMA’s NWP sys ems showed an almos neu al e ec on NWP, while s a osphe ic empe a u e ields
we e enhanced in he global NWP sys em. The modi ica ion was implemen ed in he meso-scale and local
NWP sys ems in Ma ch 2025, and is scheduled o implemen a ion in he global NWP sys em in au umn
2025.
Figu e 1: (a) Ra ios o change [%] in he s anda d de ia ion o FG depa u es o adiosonde empe a u e
obse a ion. E o ba s show con idence le els o 95%. (b) Mean o FG depa u es [K] o adiosonde
empe a u e obse a ion. Solid and do ed lines ep esen he esul s o TEST and CNTL expe imen s. Red
and g een lines ep esen he esul s o summe and win e expe imen s.
Figu e 2: Mean e o o 500 hPa geopo en ial heigh [gpm] in 1-day o ecas ing agains adiosonde obse a ions.
(a) ep esen s CNTL esul s, and (b) ep esen s di e ences be ween TEST and CNTL.
Re e ences
Ey e, J. R., 1991: A as adia i e ans e model o sa elli e sounding sys ems. ECMWF Tech. Memo., 176,
28pp.
Kazumo i, M. and S. J. English, 2015: Use o he ocean su ace wind di ec ion signal in mic owa e adiance
assimila ion. Qua . J. Roy. Me eo . Soc., 141, 1354–1375, doi:10.1002/qj.2445.
Saunde s, R., J. Hocking, P. Raye , M. Ma ica di, A. Gee , N. Bo mann, P. B unel, F. Ka bou, and F. Ai es,
2012: RTTOV-10 science and alida ion epo . Tech. ep., EUMETSAT NWP SAF, 31 pp. h ps://nwp-
sa .eume sa .in /oldsi e/deli e ables/ m/ m_ o 10.h ml.
Saunde s, R., J. Hocking, D. Rundle, P. Raye , S. Ha emann, M. Ma ica di, A. Gee , C. Lupu, P. B unel, J.
Vido , 2017: RTTOV-12 science and alida ion epo . Tech. Rep., EUMETSAT NWP SAF, URL
h ps://nwp-sa .eume sa .in /si e/download/documen a ion/ m/docs_ o 12/ o 12_s .pd .
Saunde s, R., J. Hocking, E. Tu ne , S. Ha emann, A. Gee , C. Lupu, J. Vido , P. Chambon, C. Köpken-Wa s,
L. Scheck, O. S ille , C. S ump , E. Bo bas, 2020: RTTOV-13 science and alida ion epo . Tech. ep., EU
METSAT NWP SAF, URL h ps://nwp-sa .eume sa .in /si e/download/documen a ion/ m/docs_ o 13/ o
13_s .pd .
54
Upda ing Su ace Humidi y Obse a ion E o in JMA's Regional NWP
Sys ems
SAITO Riku, TOGUCHI Ryo
Japan Me eo ological Agency
Email: iku.s[email p o ec ed]p
1. In oduc ion
Wa e apo in he lowe oposphe e has a signi ican in luence on he occu ence and de elopmen o
o en ial ain all caused by s a iona y linea mesoscale con ec i e sys ems. Acco dingly, assimila ion o
obse a ion da a ha cap u e he in low o such apo is c i ical o op imal o en ial- ain o ecas ing. In his
con ex , he Japan Me eo ological Agency (JMA) began p og essi ely ins alling hyg ome e s a i s
Au oma ed Me eo ological Da a Acquisi ion Sys em (AMeDAS) acili ies on a na ional basis in Ma ch 2020.
JMA assimila es sc een-le el ela i e humidi y as obse ed by hese hyg ome e s in o i s egional NWP
sys ems (mesoscale and local; JMA, 2025). As s a is ical examina ion o hese hyg ome e s indica ed a
endency o educed obse a ion accu acy unde high-humidi y condi ions, JMA applied modi ica ion o
inc ease he obse a ion e o o su ace speci ic humidi y unde hese condi ions o i s egional NWP
sys ems in Feb ua y 2025.
This epo desc ibes indings on hyg ome e da a quali y and obse a ion e o cha ac e is ics, along
wi h ela ed e ec s on egional NWP sys ems.
2. Hyg ome e Da a Quali y
The Ro onic MP-102H and Vaisala HMP155 hyg ome e s deployed a AMeDAS s a ions di e in
measu emen cha ac e is ics due o s uc u al a ia ions, such as hea e unc ionali y and dis inc in e nal
p ocessing speci ica ions. S a is ical compa ison o humidi y da a om hese ins umen s showed ha he MP-
102H equen ly eco ds 100% ela i e humidi y (RH), whe eas he HMP155 a ely does so (Fig. 1). This is
o en seen wi h he MP-102H a s a ions nea ice paddies o o he consis en ly humid a eas. In con as , he
HMP155 a ely exceeds 97% RH, e en du ing ain all o simila condi ions. In d y condi ions, such as sunny
day ime pe iods, bo h ins umen s show simila eadings. These di e ences sugges ha he da a cha ac e is ics
o AMeDAS hyg ome e s a y signi ican ly in high-humidi y condi ions, indica ing la ge obse a ion
unce ain ies.
3. Upda e o Obse a ion E o Se ings
As AMeDAS hyg ome e s s ill p o ide use ul da a despi e lowe accu acy in humid condi ions, i is
deemed app op ia e o u ilize hese da a while accoun ing o he sho comings. Acco dingly, a me hod is
applied o inc ease he obse a ion e o o su ace speci ic humidi y when RH exceeds 90%.
The e o is cu en ly se as 0.7 – 0.75 g/kg (equi alen o app oxima ely 3.6 – 3.8% RH a 25°C) o he
egional NWP sys em. In he mesoscale NWP sys em, he s anda d de ia ion o clima ological o ecas e o
o he i s model laye o e land du ing summe is app oxima ely 5% RH. Conside ing he la ge obse a ion
e o in high-humidi y condi ions and he ep esen a i eness e o o wa e apo , he obse a ion e o a
100% RH was se a h ee imes he cu en alue – oughly wice he model o ecas e o . The obse a ion
e o inc eases in linea p og ession om he cu en alue a 90% o h ee imes he cu en alue a 100%.
55
4. E ec s on Analysis and P edic ion
To e alua e he e ec s o he upda ed obse a ion e o se ings on egional NWP sys ems, nume ical
expe imen s we e conduc ed based on he sys em as o Ma ch 2023 (CNTL) wi h upda ed obse a ion e o
se ings (TEST) o June 26 – July 25, 2023. In he mesoscale NWP sys em, inc easing he obse a ion e o
in high-humidi y condi ions mi iga ed he ini ial bias o su ace speci ic humidi y and educed oo mean
squa e e o s (RMSEs) (Fig. 2). In p ecipi a ion p edic ion, alse ala ms o ligh ain dec eased, leading o
enhanced equi able h ea sco es (ETS). In he local NWP sys em, he ini ial d y bias o su ace speci ic
humidi y was mi iga ed, while e ec s on p ecipi a ion and o he elemen s we e small.
5. Summa y
The s udy sugges ed ha JMA’s AMeDAS hyg ome e s exhibi lowe accu acy in high humidi y. By
inc easing he obse a ion e o se ings o such condi ions, accu acy o su ace speci ic humidi y and
p ecipi a ion o ecas s in he mesoscale NWP sys em was enhanced. JMA implemen ed hese changes in i s
egional NWP sys ems in Feb ua y 2025.
Figu e 1: Obse a ion equency dis ibu ion o ela i e humidi y by ins umen ype based on s a is ics om Ma ch
10 2023 o Feb ua y 29 2024. (a) Ro onic MP-102H (157 s a ions), (b) Vaisala HMP155 (208 s a ions).
Figu e 2: Mean e o s and RMSEs o obse a ion o speci ic humidi y [g/kg] o e he expe imen pe iod o June
26 – July 25 2023 in he mesoscale NWP sys em. Blue lines: CNTL; ed lines: TEST.
Re e ences
Japan Me eo ological Agency, 2025: Da a Assimila ion Sys ems, OUTLINE OF THE OPERATIONAL
NUMERICAL WEATHER PREDICTION AT THE JAPAN METEOROLOGICAL AGENCY, 13-56.
56
Compa ison o annual a e age clima ology o key oceanog aphic
pa ame e s in he Bay o Bengal and A abian Sea
N. Gopik ishna1, Suja a K. Mandke1, *, Susmi ha Joseph1
1 Indian Ins i u e o T opical Me eo ology, Minis y o ea h Sciences, India.
Email: *[email p o ec ed]
1. In oduc ion
The Bay o Bengal (BoB) and A abian Sea (AS) a e adjacen sub-basins o he no h Indian Ocean (NIO), wi h
ma kedly di e en uppe ocean s a i ica ion due o con as ing eshwa e inpu s and wind-d i en mixing (5). The
he mohaline s uc u e o he NIO plays a pi o al ole in egula ing ai –sea hea and mois u e exchanges, which in u n
in luence he onse , in ensi y, and a iabili y o he Indian monsoon (4, 5). To enhance ou unde s anding o uppe -
ocean s abili y, e ical mixing p ocesses, and hei coupling wi h a mosphe ic ci cula ion, a comp ehensi e
clima ology o pa ame e s such as sea su ace empe a u e (SST), salini y, mixed laye dep h (MLD), iso he mal laye
dep h (ILD), and ba ie laye hickness (BLT) o e he NIO is an essen ial p e equisi e o he modelling communi y
in de eloping ocean–a mosphe e coupled models. In he p esen s udy, we compa e he annual clima ological
he mohaline s uc u e o BoB and AS.
2. Da a and Me hodology
Mon hly g idded SST and salini y da a om he ARMOR3D da ase (2) (Cope nicus Ma ine Se ice) a e used. The
MLD is di ec ly ob ained, while he ILD is calcula ed as he dep h whe e empe a u e dec eases by 0.2 °C om he
10 m e e ence dep h (Boye e al., 2004). The BLT is hen compu ed as (ILD − MLD) (1&6). Annual clima ology is
es ima ed by a e aging he annual means o he a iable o e he pe iod 1993–2022.
3. Resul s and Discussion
A compa a i e desc ip ion o he annual clima ology o SST, salini y, MLD, ILD, and BLT, conside ing hei e ical,
la i udinal, and longi udinal a ia ions o e he BoB and AS is p o ided below.
3.1 Ve ical T–S S uc u e (Fig. 1) he annual clima ology o su ace salini y is lowe in BOB han AS, as
eshwa e inpu h ough ain all and i e discha ge ma kedly exceeds e apo a ion in BOB. In con as , he
e apo a ion exceeds p ecipi a ion and i e discha ge is limi ed in AS. The salini y inc eases wi h dep h, showing a
apid ise in he uppe 100 me e s in BOB. In con as , a slow inc ease o a subsu ace maximum a ~75 me e s, hen a
e y g adual dec ease below in AS. The annual empe a u e clima ology is sligh ly wa me in BOB han AS a he
su ace, g adually dec easing wi h dep h in bo h basins. This e ical T-S s uc u e leads o a highe B un –Väisälä
equency in he BoB due o high su ace laye s a i ica ion, suppo ing he de elopmen o a s able ba ie laye (BL),
ha inhibi s e ical mixing (3,6). The annual clima ology o BOB and AS highligh s ollowing di e ences
(co esponding alues o AS a e p esen ed in pa en heses): BLT: ~20 m (8m); MLD: ~20 m (29.6m); ILD: ~35 m
(29.5m).
3.2 La i udinal a ia ion (Fig. 2) zonal a e age o e he BoB and AS depic s SST’s a e wa me and su ace
salini y is lowe in BOB han AS, no h o 20ºN, whe eas, BoB SST’s a e coole han he AS be ween 0°-20°N. The
MLD emains shallowe in he BoB han in he AS, while he ILD ex ends deepe in bo h basins, excep om 0°-4°N.
No ably, nea he equa o (0-5°N), MLD is deepe in BOB han he AS due o weake eshwa e s a i ica ion and
s onge wind-d i en mixing (3, 4, 6, and 8). BLT exhibi s s ong la i udinal a ia ion in BOB, exceeding 20 m
be ween 15–20°N due o eshwa e inpu ha enhances s a i ica ion, supp esses mixing, and esul s in a shallow
MLD. Con e sely, lowe eshwa e and s onge mixing in AS main ain a shallow BLT (0-10 m) ac oss mos
la i udes, simila o ea lie indings (1, 6). Howe e , no h o 20°N in he BoB, bo h he ILD and BLT app oach ze o
due o limi ed ba hyme ic dep h in nea -coas al egions, whe e mixing domina es o e s a i ica ion. The la i udinal
g adien o SST, BLT, ILD and su ace salini y a e much s onge no h o 20°N in BOB han AS, due la ge in low o
eshwa e om p ecipi a ion and i e uno .
3.3 Longi udinal a ia ion (Fig. 3) he me idional-a e aged a ia ion illus a e ha SST’s in BOB a e
wa me and su ace salini y is less han in AS o e all longi udes. SST’s g adually inc eases om wes o eas in BoB
63

(5), while in AS, SST dec eases om wes o eas in he wes e n (40°-50°E) and eas e n (72°-78°E) basin, while
inc eases om wes o eas in he cen al 50°-72°E basin. The wes e n AS (nea 45°-50°E) is no ably coole han he
eas e n, due o s ong winds causing upwelling in he summe monsoon (5). Su ace salini y dec eases eas wa d in
BOB owing o eshwa e inpu (4, 8). The AS is cha ac e ized by high su ace salini y h oughou , wi h sligh ly
highe salini y in he eas han in he wes . The BoB is cha ec e ized by a dis inc pa e n: coole (wa me ) SSTs in he
wes e n (eas e n) BoB co espond o highe (lowe ) salini y - a signa u e o eshwa e -induced s a i ica ion. The ILD
also deepens eas wa d, bu a ia ion wi h longi ude is less han he MLD, esul ing in a p onounced BL (BLT up o
~20-30 m) in he BoB and a e y shallow BLT in he AS (6).
4. Conclusion
In he e ical T–S s uc u e, he BoB has lowe su ace salini y and highe su ace empe a u e han he AS, wi h
salini y inc easing and empe a u e dec easing mo e sha ply wi h dep h in he BoB. Ac oss mos la i udes, he MLD is
shallowe and he BLT is hicke in he BOB han in he AS, associa ed wi h lowe su ace salini y ha enhances
s a i ica ion and supp esses e ical mixing. The BOB exhibi s a p onounced eas -wes con as , wi h lowe SST and
highe salini y in he wes han eas . This zonal asymme y in he he mal-haline s uc u e p omo es he o ma ion o a
subs an ially hicke BL (20-30 m) in he BoB. In con as , he AS emains well-mixed wi h a consis en ly shallow
BLT ac oss all longi udes, lacking a p onounced BL. These esul s p o ide an upda ed baseline o unde s anding
su ace ocean s a i ica ion in he NIO.
Re e ences
1. Boye , T. P., e al. (2004). Linea ends in salini y o he wo ld ocean, 1955–1998. Geophysical Resea ch Le e s.
2. Guinehu , S., D ecou , J. P., Le T aon, P. Y., & La nicol, G. (2012). Valida ion o la ge‐scale high‐ esolu ion sea
su ace empe a u e ields om ARMOR3D. Jou nal o A mosphe ic and Oceanic Technology.
3. Lukas, R., & Linds om, E. (1991). The mixed laye o he wes e n equa o ial Paci ic Ocean. Jou nal o
Geophysical Resea ch.
4. Rao, R. R., & Si akuma , R. (2003). Seasonal a iabili y o sea su ace salini y and sal budge o he mixed laye
o he no h Indian Ocean. Jou nal o Geophysical Resea ch: Oceans.
5. Shenoi, S. S. C., Shanka , D., & She ye, S. R. (2002). Di e ences in hea budge s o he nea -su ace A abian Sea
and Bay o Bengal: Implica ions o he summe monsoon. Jou nal o Geophysical Resea ch: Oceans.
6. Sp in all, J., & Tomczak, M. (1992). E idence o he ba ie laye in he su ace laye o he opics. Jou nal o
Geophysical Resea ch: Oceans.
7. Thada hil, P., Ghosh, A. K., Ramesh Kuma , M. R., & Ra ichand an, M. (2007). Su ace laye empe a u e
in e sion in he Bay o Bengal du ing win e . Geophysical Resea ch Le e s.
8. Vinayachand an, P. N., Mu y, V. S. N., & Ramesh Babu, V. (2002). Obse a ions o ba ie laye o ma ion in he
Bay o Bengal du ing summe monsoon. Jou nal o Geophysical Resea ch: Oceans.
Fig. 1: Annual clima ological e ical
a ia ion o empe a u e, Salini y, MLD,
ILD, and BLT a e aged o e BoB
(solid) and AS (dashed).
Fig. 2: Annual clima ological
zonal a ia ion o SST, su ace
Salini y, MLD, ILD, and BLT
a e aged o e he BoB (solid) and
AS (dashed).
Fig. 3: Annual clima ological
me idional a ia ion o SST,
Salini y, MLD, ILD, and BLT,
a e aged o e he BoB (solid) and
AS (dashed).
64
Annual cycle o su ace salini y, SST, p ecipi a ion, MLD and BLT: Compa ison be ween Bay
o Bengal and A abian Sea
N. Gopik ishna1, Suja a K. Mandke1, *, Sushmi a Joseph1
1 Indian Ins i u e o T opical Me eo ology, Minis y o ea h Sciences, India.
Email: *[email p o ec ed]
1. In oduc ion
Seasonal a ia ions in he No h Indian Ocean (NIO) a e d i en by he in e play o sea su ace
salini y (SSS), sea su ace empe a u e (SST), p ecipi a ion, and mixed laye dynamics, which
oge he shape ai –sea in e ac ions and clima e eedbacks. The Bay o Bengal (BoB) and A abian
Sea (AS), hough pa o he same monsoon sys em, espond di e en ly o hese d i e s because o
hei con as ing eshwa e inpu s, e apo a ion–p ecipi a ion balance, and wind-d i en mixing
(Shenoi e al., 2002). These p ocesses c i ically in luence uppe -ocean s a i ica ion, mixed laye
dep h (MLD), and ba ie laye hickness (BLT), which a e cen al o monsoon a iabili y and
egional clima e (Rao & Si akuma , 2003; de Boye Mon égu e al., 2007). This s udy examines he
clima ological annual cycle o key oceanic a iables ac oss BoB and AS, he wo sub- egions o he
NIO, using high- esolu ion sa elli e and eanalysis da ase s, wi h emphasis on he seasonal e olu ion
o MLD and BLT.
2. Da a and Me hodology
The weekly clima ology is cons uc ed using he ollowing da ase s:
SST: Daily 0.25° × 0.25° NOAA OISST 2.1, P ecipi a ion: Daily 1° × 1° Global P ecipi a ion
Clima ology P ojec (GPCP). Salini y and MLD: Weekly ARMOR3D Le el 4 eanalysis
(Cope nicus Ma ine Se ice), and BLT: Compu ed as iso he mal laye dep h (ILD) − MLD,
ollowing he Boye e al. (2004). ILD is es ima ed as he dep h a which empe a u e dec eases by
0.2 °C om he 10 m e e ence dep h. Two egions a e de ined o spa ial a e aging: 1. BoB: - 78°–
100°E, 0°–25°N, 2. AS: - 40°–78°E, 0°–25°N. Each da ase is i s a e aged in o weekly means o
each yea . Weekly clima ology o each o he 53 weeks is hen es ima ed by a e aging weekly
means ac oss he en i e pe iod om 1993-2023.
3. Resul s and Discussion
The annual cycle o SSS, SST, p ecipi a ion, MLD and BLT o e BoB and AS, based on he weekly
clima ology as illus a ed in Figu e 1, a e compa ed and discussed below.
SSS: - In he BoB, SSS peaks du ing he p e-monsoon season, and hen declines sha ply as he
summe monsoon in ensi ies due o hea y ain all and i e discha ge (Shenoi e al. 2002), eaching
a minimum in he pos -monsoon. The AS emains mo e saline han he BoB yea - ound. The
minimum SSS in AS is du ing he p e-monsoon, hen inc eases h ough he monsoon season as
e apo a ion exceeds p ecipi a ion, peaking in he pos -monsoon when p ecipi a ion subsides and
pe sis s ill win e .
SST: In he BoB, SST has annual minimum du ing win e and maximum in p e-monsoon. Du ing
he monsoon, SST declines g adually due o educed sola insola ion om inc eased cloud co e ,
enhanced la en hea loss om e apo a ion, and su ace cooling om hea y p ecipi a ion (Rao &
Si akuma , 2003), and wa m SST >28°C pe sis s ill pos monsoon. E apo a ion a es in he BoB a e
lowe han in he AS, and he shallow MLD limi s e ical mixing (Shenoi e al., 2002), acili a ing
ela i ely wa m and s able SSTs h ough he pos -monsoon. In he AS, SST has wo peaks, one
p ima y peak in he p e-monsoon and seconda y peak du ing pos monsoon, ollowed by minimum
in win e . S onge winds du ing he summe monsoon season cause deepe mixing, and subs an ially
highe e apo a ion a es in AS han in he BoB, esul ing in g ea e la en hea loss and a mo e
p onounced cooling (Shenoi e al 2002).
P ecipi a ion: - The p ecipi a ion inc eases sha ply wi h he onse o he summe monsoon in la e
Ap il o i s week o May and main ained ill pos monsoon in bo h BoB and AS. SSTs abo e
con ec i e h eshold (>28 °C) enhance e apo a ion causing inc eased a mosphe ic mois u e, he eby
c ea ing a ou able condi ions o con ec ion and ain all (Gadgil, 2003). In he BoB, low SSS due
o eshwa e inpu s eng hens su ace s a i ica ion, and shallow MLD which in u n main ains hese
65
wa m SSTs du ing pos monsoon seasons, esul ing in mo e p ecipi a ion (Sp in all & Tomczak,
1992). A hick BLT ein o ces his s abili y by apping hea and mois u e in he uppe laye ,
p o iding sus ained ene gy o deep con ec ion and hea y p ecipi a ion (Sp in all & Tomczak, 1992;
Thada hil e al., 2007). Thus, p ecipi a ion is mo e in BoB han AS h oughou he yea .
MLD: - In he BoB, he MLD is shallowe han in he AS h oughou he yea . The seasonal
e olu ion o he MLD in BoB is cha ac e ized by shoaling in win e and pos monsoon, a i s
minimum in he p e-monsoon and subsequen deepening in he monsoon. Howe e , he concu en
inc ease in BLT indica es ha eshwa e s a i ica ion om ain all and i e discha ge keeps he
e ec i e MLD shallow (Sp in all & Tomczak, 1992; de Boye Mon égu 2007). In he AS, he MLD
is deepes in win e due o con ec i e o e u ning
and su ace cooling, shoals in he p e-monsoon,
and hen deepens again du ing he monsoon unde
wind-d i en mixing (Shenoi e al 2002). This deep
mixing en ains coole , sal ie subsu ace wa e ,
lowe ing SSTs and educing a mosphe ic mois u e
a ailabili y o ain all (Scho e al., 2009).
BLT: - In he BoB, BLT e ol es seasonally wi h
a sp ing minimum (Ap il–May), la e-win e
maximum (Janua y–Feb ua y), and a summe –
ea ly win e ansi ion (June–Decembe ). A hick
BLT insula es he su ace om coole he mocline
wa e s, sus aining wa m SSTs and enhancing
monsoon ain all (Thada hil e al., 2007). BLT
emains e y shallow h oughou he yea in AS
han BoB. The BLT in AS ea u es annual
maxima du ing la e win e and emains shallow es
o he yea ; he shallow BLT is insu icien o
signi ican ly al e SST o p ecipi a ion (Shenoi e
al., 2002). The seasonal cycle o BLT is d i en by
o cing mechanism o MLD and ILD a iabili y (ILD is con olled by Ekman pumping, ne su ace
hea ing and p opaga ing long wa es. MLD is o ced by ans e o u bulen luxes o hea , mass, and
momen um and su ace ci cula ions) (Thada hil e al., 2007).
Conclusion
The seasonal e olu ion o mos pa ame e s is b oadly simila ac oss he wo basins, bu he e is a
la ge di e ence in he seasonal cycle o SSS, d i en by eshwa e inpu and su ace ci cula ion
(Shenoi e al., 2002). In he BoB, low SSS combined wi h a shallow MLD and a hick BLT esul in
s ong uppe -ocean s a i ica ion ha aps hea and mois u e. This s abili y main ains SSTs abo e
he con ec i e h eshold, suppo ing in ense and p olonged p ecipi a ion. In con as , he AS is
cha ac e ized by highe salini y, a deepe MLD, a hinne BLT, and s onge wind-d i en mixing
(Thada hil e al., 2007), which oge he p omo e coole SSTs and consequen ly weake p ecipi a ion.
The s a k con as be ween he s a i ied BoB and he well-mixed AS unde sco es he c i ical ole o
uppe -ocean p ocesses in modula ing monsoon p ecipi a ion.
Re e ences
1. Boye , T. P., e al. (2004). Linea ends in salini y o he wo ld ocean, 1955–1998. GRL, 31(1), L01303.
2. De Boye Mon égu , C., e al. (2007). Simula ed seasonal and in e annual a iabili y o mixed laye dep hs in
he opical Indian Ocean. JGR: Oceans, 112(C6).
3. Gadgil, S. (2003). The Indian monsoon and i s a iabili y. Annual Re iew o Ea h and Plane a y Sciences, 31,
429–467.
4. Shenoi, S. S. C., Shanka , D., & She ye, S. R. (2002). Di e ences in hea budge s o he nea -su ace A abian
Sea and Bay o Bengal: Implica ions o he summe monsoon. JGR: Oceans, 107(C7).
5. Sp in all, J., & Tomczak, M. (1992). E idence o he ba ie laye in he su ace laye o he opics. JGR:
Oceans, 97(C5).
6. Thada hil, P., e al. (2007). Su ace laye empe a u e in e sion in he Bay o Bengal du ing win e . GRL, 34(3).
Fig 1: Annual cycle o SSS, SST, P ecipi a ion,
MLD and BLT a e aged o e BoB and AS,
depic ed using weekly clima ology (53 weeks)
66
Con en ional Obse a ion Reanalysis (CORe) Da ase s
Li Zhang2, Wesley Ebisuzaki1, A un Kuma 2, and Wanqiu Wang1
1NOAA/NWS/NCEP/CPC, 2ERT a NOAA/NWS/NCEP/CPC
Email:[email protected]
In oduc ion
The Con en ional Obse a ion Reanalysis (CORe, Ebisuzaki e al., 2020) is planned o become an ope a ional
sys em a he Na ional Cen e s o En i onmen al P edic ion (NCEP) in he calenda yea o 2025. I is in ended o
eplace he long- unning NCEP/NCAR Reanalysis (Kalnay e al., 1996), which is cu en ly used by he Clima e
P edic ion Cen e (CPC) a NCEP o eal- ime clima e moni o ing. CORe is a global a mosphe ic eanalysis
co e ing he pe iod om 1950 o he p esen . I ’s designed speci ically o ge be e long- e m consis ency and ends
by assimila ing only con en ional (in si u) obse a ions and A mosphe ic Mo ion Vec o s (AMVs). This ocus
suppo s CPC’s clima e moni o ing ac i i ies.
Da a Assimila ion Sys em
CORe uses an Ensemble Kalman Fil e , speci ically he Local Ensemble T ans o m Kalman Fil e (LETKF). The
sys em has an ensemble o 80 membe s and uses inc emen al upda es. Fo he o ecas model, CORe uses a C128,
64 le el cubed sphe e model (FV3GFS, 15) wi h an app oxima e esolu ion o 0.7 deg ees. A unique ea u e o
CORe among eanalyses is i s use o a non-hyd os a ic model, which may imp o e he a mosphe ic ep esen a ion
o ce ain physical p ocesses such as b eaking g a i y wa es.
Resul s
The op panel o Figu e 1 shows global empe a u e anomaly om CORe as a unc ion o p essu e (1000 hPa o
10 hPa, y-axis) and yea (x-axis), ela i e o he 1991-2020 clima ology. The oposphe e exhibi s a g adual
wa ming end, while he s a osphe e shows a g adual cooling end. The wa m spikes in he s a osphe e
co espond o e ec s o olcanic e up ions which inc ease he s a osphe ic ae osols. The middle panel is he same
as he op panel excep o he JRA-3Q eanalysis (Kosaka e al, 2024), and he bo om panel o ERA-5 (He sbach
e al., 2020). All h ee eanalyses display qui e simila ends, wi h JRA-3Q exhibi ing sligh ly smalle s a osphe ic
wa ming om he olcanoes. These compa isons demons a e a high le el o ag eemen among he la es eanalyses.
Figu e 1. Global empe a u e anomalies o CORe,
JRA-3Q and ERA-5.
Figu e 2. 10S-10N p ecipi a ion anomalies o DYNAMO
pe iod o obse a ions, R1, R2, CFSR and CORe.
67
Figu e 2 displays he 10°S-10°N p ecipi a ion anomalies du ing he Dynamics o he MJO (DYNAMO) campaign
pe iod om (a) obse a ions (Xie e al., 2017), (b) NCEP/NCAR Reanalysis (R1), (c) NCEP/DOE Reanalysis (R2,
Kanami su e al., 2002), (d) Clima e Fo ecas Sys em Reanalysis (CFSR, Saha e al., 2010) and (e) CORe. No ably,
CORe, despi e no using sa elli e adiance da a, success ully cap u es he Madden-Julian Oscilla ion (MJO)
p ecipi a ion signal o e he Indian and Wes e n Paci ic Oceans, while he olde R1 and R2 sys ems, despi e
assimila ing sa elli e e ie als, show weake MJO signals.
Da ase s
The p ima y CORe da a a chi e con ains ensemble means o he a mosphe ic and land su ace s a e. The analyses
a e on p essu e-le el, model le els and special laye s. These da a a e in GRIB e sion 2 (g ib2) o ma on a 512x256
Gaussian g id which is he da a assimila ion g id. The seconda y le el a chi e includes ensemble membe s and
s a is ics as well as he obse a ion inc emen s and a subse o model es a iles. These iles a e in g ib2, NEMSIO
o Ne CDF o ma s.
Cloud A chi e
The main dis ibu ion o CORe analyses will be ia he NOAA Open Da a Dissemina ion (NODD) cloud
pla o m. The ull a chi e will e en ually co e Janua y 1950 o p esen wi h a 2-3 day la ency. As o Sep embe
2025, he a chi e includes da a om Janua y 1950 o Decembe 2021. A ailable p oduc s include 3-hou ly, daily
and mon hly ensemble mean analyses.
The a chi e is expec ed o expand a e CORe becomes ope a ional. These da ase s a e in g ib2 o ma , and
con e sion o o he g ids can be easily done using p og ams, such as wg ib2, and pyg ib.
Re e ences
Ebisuzaki W., and coau ho s (2020). A Con en ional Obse a ion Reanalysis (CORe) o Clima e Moni o ing.
Science and Technology In usion Clima e Bulle in, 44 h NOAA Annual Clima e Diagnos ics and P edic ion
Wo kshop, Du ham, NC 22-24 Oc obe , 2019, h ps://doi.o g/10.25923/ 4qa-ae63.
He sbach, H., and coau ho s (2020). The ERA5 Global Reanalysis. Qua e ly Jou nal o he Royal Me eo ological
Socie y 146 (730), h ps://doi.o g/ 10.1002/qj.3803.
Kalnay, E., and coau ho s (1996). The NCEP/NCAR 40-Yea Reanalysis P ojec . Bull. Ame . Me eo . Soc. 77(3),
437-472, h ps://doi.o g/10.1175/1520-0477(1996)077<0437:TNYRP>2.0.CO;2
Kanami usu, M., and coau ho s (2002). NCEP-DOE AMIP-II Reanalysis (R-2). Bull. Ame . Me eo . Soc. 83(11),
1631-1644, h ps://doi.o g/10.1175/BAMS-83-11-1631.
Kosaka, Y., and coau ho s (2024). The JRA-3Q Reanalysis. J. Me eo . Soc. Japan (102), 49-109.
h ps://doi.o g/10.2151/jmsj.2024-004.
Saha, S., and coau ho s (2010). The NCEP Clima e Fo ecas Sys em Reanalysis. Bull. Ame . Me eo . Soc. (91),
1015–1057, h ps://doi.o g/10.1175/2010BAMS3001.1.
Xie, P., and coau ho s (2017). Rep ocessed, Bias-Co ec ed CMORPH Global High-Resolu ion P ecipi a ion
Es ima es om 1998. Jou nal o Hyd ome eo ology (18), 1617-1641, h ps://doi.o g/10.1175/JHM-D-16-0168.1
68

Nume ical/
compu a ional
echniques and
model esolu ion,
physics-dynamics and
physics-physics c oss-
componen coupling
9
69
De elopmen o Regional Sho - ange P edic ion Sys ems Based on KIM
Eun-Hee Lee, Junghan Kim, Heeje Cho, Soo Ya Bae and Kyung-Hee Seol
Ko ea Ins i u e o A mosphe ic P edic ion Sys ems, Seoul, Sou h Ko ea
Email: [email p o ec ed]
1. De elopmen o sho - ange p edic ion sys ems based on KIM
The Ko ean In eg a ed Model (KIM) was o iginally designed o he global medium- ange o ecas . Since i s
adop ion as he ope a ional wea he p edic ion model o he Ko ea Me eo ological Adminis a ion (KMA), he
Ko ea Ins i u e o A mosphe ic P edic ion Sys ems (KIAPS) has expanded i s applica ions o suppo seamless
p edic ion ac oss scales. A ecen ad ancemen is he de elopmen o sho - ange p edic ion sys ems by
implemen ing wo model unc ionali ies: he a iable- esolu ion g id sys em and he limi ed-a ea model (LAM)
con igu a ion. These sho - ange p edic ion sys ems sha e co e componen s wi h he global uni o m- esolu ion
KIM, including he non-hyd os a ic dynamic co e which u ilizes he spec al elemen me hod on a cubed-
sphe e g id and he scale-awa e physics.
1.1 Va iable Resolu ion g id
The a iable- esolu ion e sion o KIM was de eloped using he Schmid ans o ma ion, wi h inc easing
ho izon al esolu ions a ound he a ge a ea. The cubed-sphe e amewo k was ex ended h ough an upda ed
ans o ma ion ma ix, and he ime s ep and nume ical iscosi y coe icien s we e adjus ed o main ain
consis en nume ical beha io ac oss esolu ions. The deg ee o g id sys em de o ma ion is con olled by a
s e ching ac o , S, ela i e o a e e ence uni o m- esolu ion g id (S=1). Because explici ime in eg a ion
schemes a e limi ed by he smalles g id spacing, he ime s ep is p opo ionally educed by a ac o o 1/S.
E alua ion esul s indica e ha he a iable- esolu ion KIM p o ides o ecas skill compa able o high-
esolu ion uni o m-g id simula ions wi hin he a ge ed domain and signi ican ly ou pe o ms lowe - esolu ion
simula ions. Figu e 1 illus a es a ious g id con igu a ions o s e ched global g ids wi h di e en S ac o s.
1.2 Limi ed A ea Model
The limi ed-a ea ( egional) e sion o KIM was de eloped by cons uc ing a modelling amewo k ha allows
he model o ope a e on a single panel o he cubed sphe e. Because he LAM con igu a ion can be employed
wi h he a iable- esolu ion g id sys em, ou “single-panel” app oach is able o na ow he LAM domain by
se ing he s e ching ac o S>1. Figu e 1 illus a es he domains wi h di e en S alues. La e al bounda y
condi ions a e speci ied a all Gauss-Loba o-Legend e (GLL) poin s o he ou e mos elemen s and a e de i ed
h ough spa ial and empo al in e pola ion om coa se - esolu ion global KIM simula ions. To enhance
simula ion s abili y, a elaxa ion (o nudging) e m is applied nea he domain bounda ies. This e m g adually
blends he egional model solu ion owa d he backg ound ield and ollows he same unc ional o m used in
WRF and MPAS.
2. Semi- eal ime ope a ion and pe o mance
The la es upg ade o KIAPS KIM includes u ili ies o wo egional-scale p edic ion sys ems. KIAPS has
ini ia ed he in e nal semi- eal- ime es ing o wo sho - ange p edic ion sys ems: (1) a a iable- esolu ion
global g id a ge ing a 3 km esolu ion o e Eas Asia (NE1=576, S=2.5), and (2) a limi ed-a ea g id a ge ing
1 Numbe o elemen s in each panel o he cubed-sphe e. Each elemen has 3×3 GLL g id poin s o all model
con igu a ions men ioned he e.
70
a 1 km esolu ion o e he Ko ean Peninsula (NE=768, S=5.0). These sys ems ope a e in pa allel wi h he
global medium- ange o ecas sys em wi h uni o m 8-km esolu ion g id (NE=576) which p o ides ini ial and
bounda y condi ions o he wo sho - ange p edic ion sys ems. All h ee sys ems sha e he same physics,
suppo ing scale-awa e ea u es like con ec ion schemes and g a i y wa e d ag pa ame e iza ions. No e ha
all con igu a ions use he same e ical disc e iza ion and esolu ion. The wo sho - ange p edic ion sys ems
do no ha e da a assimila ion sys ems, bu he 8 km backg ound ields p o ide obse a ional in o ma ion.
P elimina y e alua ions du ing he summe monsoon season show ha he e y-high- esolu ion sys ems
cap u e in ense p ecipi a ion signals o hea y ain all e en s, closely ma ching obse a ions. This highligh s
he po en ial o egional ~km esolu ion sys ems o eliable o ecas ing and seamless p edic ions using KIM
(Fig. 2). These de elopmen s a e se o be in eg a ed in o ope a ional sys ems a KMA ollowing he
comple ion o he second KIAPS p ojec (2020–2026).
Figu e 1: Va ious g id sys ems in KIM: Global uni o m (le ), s e ched (cen e ), and limi ed-a ea g id ( igh )
Figu e 2: 6-h accumula ed p ecipi a ion (+12-h p edic ion) simula ed by a ious KIM con igu a ions a 0000 UTC
on July 17 2025. Resul s include simula ions om he uni o m global 8 km g id (NE576), he global a iable-
esolu ion model (NE576S2.5), and he 1 km LAM (NE767S5.0) alongside gauge obse a ion (le mos ).
71
Machine lea ning
and AI in wea he
p edic ion and clima e
modeling
10
72
Applica ion o a i icial in elligence models in imp o ing he in ensi y
o ecas o T opical Cyclones o e No h Indian Ocean Basin
1Dhananjay T i edi, 1Om ee Sha ma, 1Sandeep Pa naik*, 2Heman Kuma , 2Sama h Bansal, 2Nilad i
Biha i Puhan
1 School o Ea h, Ocean, and Clima e Science, Indian Ins i u e o Technology, Odisha, India
2School o Elec ical Sciences, Indian Ins i u e o Technology, Odisha, India
Email: [email p o ec ed]
1.T opical Cyclone In ensi y
1.1 Es ima ion
T opical cyclones (TC) pose se ious isks and in lic majo p ope y and human damage;
comp ehending a TC's a ious s ages is c ucial o assessing i s e ec s. I has been mo e
challenging bu essen ial in ecen yea s o c ea e e ec i e, high-pe o ming echniques o
cyclonic s o m p edic ion, mi iga ion, and coexis ence. In gene al, i is ex emely di icul o
ope a ional agencies o es ima e he in ensi y o TC using physics-based dynamical models,
and he si ua ion is highly challenging i a TC goes h ough a Rapid In ensi ica ion (RI) phase.
We design a no el a en ion mechanism o cyclone in ensi y es ima ion based on he
pe cep ual Mach band e ec ha acili a es boos ing and supp ession o edges in he inpu
image. This is he i s wo k o es ablish a use ul connec ion be ween he pe cep ual Mach band
e ec and he adap i e ea u e weighing mechanism in insic o he a en ion model deployed
wi hin a deep con olu ional neu al ne wo k (CNN). The Mach band a en ion model (MBAM)
aims o ampli y o supp ess he p ominence o ea u e loca ions wi hin he con olu ional
ea u e space by le e aging a en ion weigh s. We ained and es ed ou model using INSAT-
3DR sa elli e images and India Me eo ological Depa men (IMD) bes in ensi y a chi e
da ase s.The p oposed model shows he Mean Absolu e E o (MAE) o 0.354 (k s) and Roo
Mean Squa e E o o 0.415 k s in wind speed es ima ion. In he cyclone se e i y classi ica ion,
he model shows accu acy o 0.9944, ecall o 0.9943, and F1-sco e o 0.9938 shows he
excellence o model pe o mance in cyclone’s in ensi y es ima ion o e NIO egion (Bansal e
al., 2024).
Fu he , he model has been es ed o RI o TCs in NIO basins. The es ing o he RI TCs (i.e.,
Bulbul, Maha, and Nilo a ) has been ca ied ou using a ious a en ion mechanisms, such as
Mach Band and CBAM, wi h a basic model as ResNe 18.The esul s sugges ed ha he
MBAM3 shows he leas (2.87 k s) and CBAM4 shows he highes (3.39 k s) RMSE o all
TCs. Fu he , he MBAM3 model is used o RI es ima ion o he TC. The MBAM model
cap u es he RI phase o he TCs accu a ely wi h an RMSE o 1.65 k s o Bulbul, 2.63 k s o
Maha, and 1.68 k s o Nilo a (Sha ma e al., 2025a).
79

Figu e 1. Ac ual (obs) and Es ima ed (model) in ensi y plo o he TC Nilo a (Sha ma e al. 2025).
1.2 P edic ion
We ha e designed a used deep lea ning (DL) ne wo k comp ised o a Vision T ans o me
(ViT) and Con olu ional Block A en ion Model (CBAM) wi h ResNe , known as HASTVi,
o cyclone in ensi y o ecas wi h a lead ime o up o 24 hou s. The model has been ained
and es ed using in a ed sa elli e images ob ained om he Indian Space Resea ch
O ganiza ion (ISRO) and he India Me eo ological Depa men (IMD) bes es ima es o TC
in ensi y. A o al o 38 TCs we e used o ain and es he model, ou o which 35 TCs (7274
images) we e used o aining and 3 TCs (694 images) we e used o es ing pu poses. The
h ee mos de as a ing TCs, iz. Amphan (2020), Fani (2019), and Tauk ae (2021) we e used
o es ing pu poses due o hei immense socio economic impac . The p oposed ision
HASTVi model shows he Mean absolu e e o (MAE) o 2.90 k s, 3.14 k s, and 5.38 k s, o
3h o ecas , which is be e han he s a e-o - he-a Con olu ional Long Sho Te m Memo y
(Con LSTM) model, ha ing an MAE o 15.35 k s, 18.99 k s, and 29.90 k s o Tauka e, Fani,
and Amphan, espec i ely. In addi ion, he HASTVi model has been es ed o he apid
in ensi ica ion phase o he TC o e he NIO basin (Sha ma e al., 2025b). These indings ha e
di ec implica ions o imp o ing he TC ea ly wa ning sys ems o e he NIO basins.
Re e ences
Bansal, S., Puhan, N. B., & Pa naik, S. (2024). No el Pe cep ual Mach Band-Based Deep
A en ion Ne wo k o Cyclone In ensi y Es ima ion. IEEE T ansac ions on Ins umen a ion
and Measu emen , 73, 1-11.
Sha ma, O., Kuma , H., T i edi, D., Goswami, N., Pa naik, S., & Puhan, N. B. (2025a).
Es ima ion o he apid in ensi ica ion o opical cyclones o e he No h Indian Ocean using
a en ion-based deep lea ning models. Na u al Haza ds, 1-16.
Sha ma, O., T i edi, D., Kuma , H., Pa naik, S., & Puhan, N. B. (2025b). Enhancing
in ensi ica ion o ecas o opical cyclones in he No h Indian Ocean Basin using used Vision
ans o me and Con olu ional block a en ion wi h ResNe model. IEEE T ansac ions on
Geoscience and Remo e Sensing.
80
WGNE
Wo king G oup on
Nume ical Expe imen a ion
ESMO
Ea h Sys em Modelling
and Obse a ions
WCRP
Wo ld Clima e
Resea ch P og amme
WORLD
METEOROLOGICAL
ORGANIZATION