On the development of the GRAPES—A new generation of the national operational NWP system in China

Abstract

Numerical weather prediction (NWP) has become one of the most important means for weather forecasts in the world. It also
mirrors a nation’s comprehensive strength in meteorology. In 2000, China Meteorological Administration (CMA) established the
National Innovative Base for Meteorological Numerical Prediction in the Chinese Academy of Meteorological Sciences (CAMS),
to work on developing a new generation of the national operational NWP system—Global/Regional Assimilation and PrEdiction
System (GRAPES), to enhance meteorological services in China in the new century. In recent years, the GRAPES has witnessed
a fast development. The GRAPES has been set up as an integration of the model framework, data assimilation, regional and global
NWP system, which can be commonly used for both operation and research. In this paper, a brief review is made for illustrating
the GRAPES system, including the advanced designs of the GRAPES, its diverse applications in multi-fields, and efficiencies
of the regional and global GRAPES in operational applications based on hindcast results.

Full text

Chinese Science Bulletin
© 2008 SCIENCE IN CHINA PRESS
Springer
www.scichina.com | csb.scichina.com | www.springerlink.com Chinese Science Bulletin | November 2008 | vol. 53 | no. 22 | 3429-3432
ATMOSPHERIC SCIENCES SPECIAL TOPIC
PROGRESS
On the development of the GRAPES――
A new generation of the national operational
NWP system in China
ZHANG RenHe† & SHEN XueShun
State Key Laboratory of Severe Weather, Chinese Academy of Meteorological Sciences, Beijing 100081, China
Numerical weather prediction (NWP) has become one of the most important means for weather fore-
casts in the world. It also mirrors a nation’s comprehensive strength in meteorology. In 2000, China
Meteorological Administration (CMA) established the National Innovative Base for Meteorological
Numerical Prediction in the Chinese Academy of Meteorological Sciences (CAMS), to work on devel-
oping a new generation of the national operational NWP system―― Global/Regional Assimilation and
PrEdiction System (GRAPES), to enhance meteorological services in China in the new century. In re-
cent years, the GRAPES has witnessed a fast development. The GRAPES has been set up as an inte-
gration of the model framework, data assimilation, regional and global NWP system, which can be
commonly used for both operation and research. In this paper, a brief review is made for illustrating
the GRAPES system, including the advanced designs of the GRAPES, its diverse applications in
multi-fields, and efficiencies of the regional and global GRAPES in operational applications based on
hindcast results.
GRAPES, numerical weather prediction, China national operational weather forecasts
China is a country frequently attacked by meteorological
disasters. Meteorological disasters have brought up huge
losses to the nation’s economy, social development, and
people’s lives and properties. Since the 1990s, each year
China would have some 48 million hectares of croplands
attacked by meteorological disasters, with a disaster af-
fected population of about 380 million, and a direct
economic loss worth about RMB 180 billion, or 2.7% as a
proportion of GDP[1]. Raising the accuracy of predicting
severe weathers makes a most direct and effective ap-
proach to mitigate and prevent meteorological disasters.
Since the 1950s, the numerical weather prediction (NWP),
building on the bases of mathematics and physics, has
been one of major accomplishments achieved in the area
of atmospheric sciences[2]. Thanks to more than half a
century development, the NWP has found wide applica-
tions in the world, and has become one of the most im-
portant means in operational weather forecasts.
The development of the NWP depends not only on
the basic researches of the atmospheric sciences, but
also on the space based sounding and retrieval tech-
niques for collecting global atmospheric data, super-
computer technologies that make data processing and
computation more efficient, and telecommunication
technologies that allow a fast dissemination of data. In
this context, a nation’s NWP level mirrors the compre-
hensive strength of these technologies, in addition to the
level of researches on the atmospheric sciences. It also
marks the level of a nation’s meteorological moderniza-
Received June 18, 2008; accepted August 27, 2008
doi: 10.1007/s11434-008-0462-7
†Corresponding author (email: renhe@cams.cma.gov.cn)
Supported by Key Technologies Research & Development Program (Grant No.
2001BA607B), National Key Technology Research and Development Program
(Grant No. 2006BAC02B02) and National Basic Research Program of China (Grant
No. 2004CB418300)

3430 ZHANG RenHe et al. Chinese Science Bulletin | November 2008 | vol. 53 | no. 22 | 3429-3432
tion, and comprehensive strength in the area of meteor-
ology.
China started the national operation of the NWP at
around the end of 1980s and the beginning of 1990s
with the main framework of the NWP models being in-
troduced from the developed countries. The introduced
operational models did not offer needed continuity and
re-development functionality. As a result, China’s NWP
operation deplored a large gap compared with the de-
veloped countries, in a range of indicators including
model’s performance, data assimilation, and accuracy of
prediction, which in turn hindered the improvement of
China’s operational meteorological services, and became
a bottleneck restricting the development of China’s me-
teorological modernization. To realize a leaping devel-
opment in the area, meet the increasing needs of eco-
nomic and social development for meteorological ser-
vices, and to work on a fully modernized operational
NWP system, China Meteorological Administration es-
tablished in October 2000 the National Innovative Base
for Meteorological Numerical Prediction at the Chinese
Academy of Meteorological Sciences, in a move to cre-
ate a proprietary Global/Regional Assimilation and PrE-
diction System (GRAPES) in the new century. Both
Ministry of Science and Technology of China and China
Meteorological Administration nodded their approval to
create projects for the purpose, with further support for
the related basic researches from National Natural Sci-
ence Foundation of China. Recent years have witnessed
a fast development of the GRAPES mainly on the efforts
of Chinese scientists[3 ―5]. The achievements include
model framework, data assimilation, and integrated re-
gional/global numerical forecast system both for re-
search and operation. This issue of the Chinese Science
Bulletin releases 5 GRAPES related papers[6―10], to re-
flect the latest development in the area. The following
presents a brief review of the development of the
GRAPES, illustrating its advanced natures, its diverse
applications in multi-fields, and efficiencies of the re-
gional and global GRAPES in operational applications
based on hindcast results.
1 Advanced natures of the GRAPES
The GRAPES is built using latest achievements and the
experience derived from the development of operational
NWP systems in the world after the late 1990s. The ad-
vanced features of GRAPES are mainly in the following
aspects.
The principle of the intensiveness is applied in the
development of the GRAPES. The GRAPES framework
incorporates the dynamic core of multi-scales, which
can be used as a common base for different application
models. The dynamic core is designed with a range of
options including static or non-static, global or regional,
and optional horizontal or vertical resolutions. An inte-
grated model system capable of diverse forecast applica-
tions is developed. Compared with individual target
based models, it is more desirable for long-term techni-
cal upgrade and development. Modular and standardized
coding is adopted so that it is easy for enlarging the
model functions, allowing scalable and sustainable de-
velopment, reducing the cost of the continuous devel-
opment, and easily transferring research findings into
operational applications. The variational assimilation
system of the GRAPES can assimilate not only the
regular meteorological observational data, but also satel-
lite based direct radiation data, remote sensing data de-
rived from the Doppler weather radar and products from
the remote sensing of the satellite (cloud winds, for ex-
ample). The variational assimilation system in the
GRAPES greatly increases the data that can be used in
the NWP in China, creating a solid ground for raising
the accuracy of the model forecast.
2 Diverse applications
In China, based on the GRAPES, an array of numerical
forecast systems has been developed for different appli-
cations. The GRAPES mesoscale NWP system
(GRAPES-Meso) has been put into the national opera-
tion in 2006 at the National Meteorological Center. The
GRAPES typhoon forecast model (GRAPES-TCM) has
fine performance for predicting route related major ele-
ments of the tropical cyclone (TC), including route type,
intensity, landing process, and sudden changes of the
moving direction and the wind speed of TC[11]. The
GRAPES sandstorm forecast model (GRAPES-SDM)
performs well. It can predict basically the occurrence
and development of sand and dust storms in China[12].
The GRAPES lightening forecast model, nesting a
lightening module in the GRAPES-Meso, has success-
fully simulated several lightening cases occurring in
Beijing. The simulation results show that strong convec-
tive centers and high concentration areas of ice particles
correspond to strong charge centers, in agreement with

ZHANG RenHe et al. Chinese Science Bulletin | November 2008 | vol. 53 | no. 22 | 3429-3432 3431
ATMOSPHERIC SCIENCES SPECIAL TOPIC
PROGRESS
actual observations, demonstrating an enhanced poten-
tial for predicting lightening[13]. The explicit cloud-pre-
cipitation scheme nested in the GRAPES can predict the
distribution of the cloud as well as the cloud structure
and phases in the cloud, which has been used in provid-
ing guidance for weather modification activities in
China[14]. The GRAPES Severe Weather Integrated Tool
(GRAPES-SWIFT) has joined the project of the World
Meteorological Organization WMO/B08FDP/RDP as
the only short-term and nowcasting weather prediction
system developed by Chinese scientists among other
model systems from different countries, providing me-
teorological services for the Beijing Olympic Game in
2008.
3 Hindcast experiments of the GRAPES
NWP system
The GRAPES-Meso system with a 30 km horizontal
resolution was put into operation in 2006 at the National
Meteorological Center. The system was upgraded to a
15 km horizontal resolution in 2007. In order to demon-
strate the forecast effectiveness, Figure 1 shows the ETS
scores for 24-h continuous hindcasts of the precipitation
in the region of whole China for a total year from June
2006 to May 2007. In Figure 1, for comparison a num-
ber of operational models from domestic and overseas
are used to make the precipitation hindcasts. Six models
are employed including GRAPES-Meso (30 km/15 km
horizontal resolution), MM5 regional model (27 km
horizontal resolution) and T213 global model (about 60
km horizontal resolution) used for operational forecasts
in China, Japan’s model (about 20 km horizontal resolu-
tion), and German model (40 km horizontal resolution).
From Figure 1 we can see that, compared with the
version of the 30 km horizontal resolution, it is apparent
the GRAPES-meso model of the 15 km version yields
the better result for the hindcasted light rainfall (<0.1
mm), moderate rainfall (0.1―10 mm), and heavy rain-
fall (10―25 mm), respectively. Compared with other
models, GRAPES-Meso with the 15 km horizontal reso-
lution exhibits a fine systematic strength. The ETS
scores are generally higher than those of other models
except the score for the light rainfall being only a bit
lower than that of its German counterpart, moderate
rainfall a bit lower than the Japanese model, and heavy
rainfall lower than MM5. Comparison results also show
that GRAPES-Meso with the 15 km horizontal resolu-
tion is able to produce 24-h precipitation forecasts in the
region of China equivalent to or even better than those
offered by its advanced international counterparts.
In order to show the forecast effectiveness of the
GRAPES Global Forecast System (GRAPES-GFS), the
hindcasts given by the GRAPES-GFS with the 100 km
horizontal resolution are made continuously for a whole
year from December 2006 to November 2007. Figure 2
gives the anomaly correlation coefficients (ACC) be-
tween the hindcast results of the GRAPES-GFS and the
analyses of the National Center for Environmental Pre-
diction of USA (NCEP). The ACC is calculated based
on the 500 hPa geopotential height fields in the northern
hemisphere in the latitudinal zone of 20°―90°N in the
period of December 2006―November 2007.
From Figure 2 we can see that the GRAPES-GFS has
obviously possessed a fine capability for the prediction.
Figure 1 ETS scores for 24 h hindcasts of precipitations of the light (<0.1 mm, left panel), moderate (0.1―10 mm, middle panel), and heavy (10―
25 mm, right panel) rainfall in China region during the period of June 2006―May 2007 by using different domestic and overseas models.

3432 ZHANG RenHe et al. Chinese Science Bulletin | November 2008 | vol. 53 | no. 22 | 3429-3432
Figure 2 ACC between GRAPES-GFS hindcasts and NCEP analyses
for 500 hPa geopotential height fields in the northern hemisphere (20°―
90°N) during the period of December 2006―November 2007.
The ACC between the GRAPES-GFS hindcasts and
NCEP analyses has reached 0.97, 0.94 and 0.87 for the
GRAPES-GFS hindcast of 24 h, 48 h and 60 h, respec-
tively. When taking 0.6 ACC as a threshold value of the
effectiveness for forecasts, the GRAPES-GFS has sus-
tained the capability of the effective forecast time about
some 6 days, which is basically qualified for an opera-
tional application. As a result, GRAPES-GFS will be put
into a trial operation at the National Meteorological
Center in the near future.
4 Concluding remarks
The GRAPES is a homemade numerical meteorological
prediction system. The satellite data assimilation tech-
nique employed in the system has found a solution to
addressing technical bottlenecks restricting the applica-
tion of the satellite data in numerical weather prediction
in China. The common used multi-scale dynamical core
and static or non-static equilibrium framework have cre-
ated a ground for intensively developing the numerical
model. In addition, the non-static meso-scale forecast
model with sophisticated physical processes has paved
the way for predicting meso-scale severe weathers in
China. The successful establishment of the global pre-
diction system has made China one of the countries that
have mastered the core technologies in building the
global NWP model.
The GRAPES still needs further optimization during
operational applications, and original innovations tai-
lored to special weather in China[15], though it has so far
achieved noticeable progresses. Efforts shall also be
made to optimize the algorithms handling large terrains
in the dynamic framework, especially the topography
around the Tibetan Plateau, which affects weather sig-
nificantly in China. The GRAPES should expand to the
stratosphere and coupling with ocean models, and de-
velop the land process model specially for China’s so-
phisticated surface conditions, and the two-way nested
global and regional model. As far as physical processes
are concerned, the tailored physical parameterized
schemes shall be worked out in the GRAPES to reflect
the characteristics in East Asia where China is located at.
In the data assimilation system, the assimilation of di-
verse observations should be realized, and special atten-
tion shall be paid to utilizing the domestic observational
data. With operation oriented efforts, the GRAPES will
eventually be built into an advanced international NWP
system with Chinese characteristics.
1 China Meteorological Administration. China Severe Weather Atlas
(1961―2006) (in Chinese). Beijing: China Meteorological Press, 2007
2 Chen D H, Xue J S. An overview on recent progresses of the opera-
tional numerical weather prediction models (in Chinese). Acta Meteor
Sinica, 2004, 62(5): 623―633
3 Xue J S, Progress of Chinese numerical prediction in the early new
century (in Chinese). J Appl Meteor Sci, 2006, 17(5): 602―610
4 Chen D H, Shen X S. Recent progress on GRAPES research and ap-
plication (in Chinese). J Appl Meteor Sci, 2006, 17(6): 773―777
5 Xue J, Liu Y. Numerical weather prediction in China in the new cen-
tury-progress, problems and prospects. Adv Atmos Sci, 2007, 24:
1099―1108
6 Chen D H, Xue J S, Yang X S, et al. New generation of multi-scale
NWP system (GRAPES): general scientific design. Chin Sci Bull,
2008, 53(22): 3433―3445
7 Xue J S, Zhuang S Y, Zhu G F, et al. Scientific design and preliminary
results of three-dimensional variational data assimilation system of
GRAPES. Chin Sci Bull, 2008, 53(22): 3446―3457
8 Yang X S, Hu J L, Chen D H, et al. Verification of GRAPES unified
global and regional numerical weather prediction model dynamic core.
Chin Sci Bull, 2008, 53(22): 3458―3464
9 Zhu G F, Xue J S, Zhang H, et al. Direct assimilation of satellite
radiance data in GRAPES variational assimilation system. Chin Sci
Bull, 2008, 53(22): 3465―3469
10 Xu G Q, Chen D H, Xue J S, et al. The program structure designing
and optimizing tests of GRAPES physics. Chin Sci Bull, 2008, 53(22):
3470―3476
11 Zhu Z D, Duan Y H, Chen D H. An analysis of GRAPES-TCM’s op-
erational experiment results (in Chinese). Meteor Mon, 2007, 33(7):
44―54
12 Zhao J H, Li Y H. Summarization of sand-dust prediction based on
application of GRAPES-SDM (in Chinese). Arid Meteor, 2006, 24(1):
7―13
13 Zhang Y J, Zhou X J. Review and progress of lightning research (in
Chinese). J Applied meteor Sci, 2006, 17(6): 829―834
14 Zhang J C, Liu Q J. Analysis of cloud schemes in simulation of short-
term climatic process (in Chinese). Meteor Mon, 2006, 32(7): 3―12
15 Zhang R H. Study on theories and methods for monitoring and pre-
dicting heavy rainfall in South China (in Chinese). China Awards for
Science and Technology, 2005, 1: 74―77