Download as pdf or txt
Download as pdf or txt
You are on page 1of 16

Atmospheric Research 282 (2023) 106522

Contents lists available at ScienceDirect

Atmospheric Research
journal homepage: www.elsevier.com/locate/atmosres

Evaluation of CMIP6 GCMs performance to simulate precipitation over


Southeast Asia
Sittichai Pimonsree a, Suchada Kamworapan b, Shabbir H. Gheewala c, d,
Amornpong Thongbhakdi e, Kritana Prueksakorn b, *
a
Atmospheric Pollution and Climate Change Research Unit, School of Energy and Environment, University of Phayao, Phayao 56000, Thailand
b
Faculty of Environment and Resource Studies, Mahidol University, Nakhon Phathom 73170, Thailand
c
Joint Graduate School of Energy and Environment (JGSEE), King Mongkut’s University of Technology Thonburi (KMUTT), Bangkok 10140, Thailand
d
Center of Excellence on Energy Technology and Environment (CEE), Ministry of Higher Education, Science, Research and Innovation, Bangkok, Thailand
e
SCGP, 1 Siam Cement Road, Bangkok 10800, Thailand

A R T I C L E I N F O A B S T R A C T

Keywords: The performance for simulating precipitation of 27 global climate models (GCMs) participating in the Coupled
Global climate models Model Intercomparison Project Phase 6 (CMIP6) is evaluated over the Southeast Asia (SEA) region and nine SEA
CMIP6 countries for the period 1975–2014 by comparing observation data from the Global Precipitation Climatology
Precipitation
Centre (GPCC). The performance ranking of the GCMs was conducted with the ranking scores method using eight
Model ranking
Southeast Asia
performance metrics that cover both spatial and temporal patterns. Compared to GPCC, the results show that
some GCMs in CMIP6 reasonably capture the spatial precipitation patterns, with more in maritime and less in
mainland. However, the output of most models presents considerable overestimation (wet bias) in Thailand,
Cambodia, central Myanmar and maritime countries, and underestimation (dry bias) in Indonesia, Lao, northern
Vietnam and western Myanmar. In addition, the findings illustrate that many models can reproduce the annual
cycle shape and inter-annual variability which are consistent with GPCC; however, only 2 out of 27 models can
detect increasing trends such as GPCC in every study domain. By model ranking, the best models vary from area
domain to area domain. TaiESM1 performs best among the 27 GCMs over SEA region as well as Thailand. The
model that has the best performance in most counties, i.e., Cambodia, Indonesia, Lao, and Vietnam is EC-Earth3-
Veg-LR. EC-Earth3 is the best model in Brunei, Malaysia while CESM2-FV2 is the best model in Myanmar and the
Philippines. It is also discovered that the mean ensemble of all GCMs has limited skills for all study domains. The
results of this study can be used to support selection of the suitable models for simulating precipitation in specific
study domains.

1. Introduction rainfall variability (Tan and Pereira, 2010). According to Santhanasamy


et al. (2021), the SEA region faced a number of disasters between 1971
Precipitation is one of the most important factors influencing the and 2020, including 687 floods, 450 tropical cyclones, 114 landslides,
existence of humans, plants, animals and the natural environment. and 21 droughts and forest fires. Furthermore, the SEA region was
Variations in precipitation contribute to changes in magnitude and affected by major disasters such as Cyclone Nargis in 2008 and Typhoon
spatiotemporal patterns (Yazdandoost et al., 2021). Southeast Asia Haiyan (also called Typhoon Yolanda) in 2013 (Santhanasamy et al.,
(SEA) is one of the regions most affected by natural disasters related to 2021). These precipitation-related natural disasters have caused

Abbreviations: APHRODITE, Asian Precipitation – Highly Resolved Observational Data Integration Towards Evaluation; CMIP, Coupled Model Intercomparison
Project; CMIP5, Coupled Model Intercomparison Project Phase 5; CMIP6, Coupled Model Intercomparison Project Phase 6; CRU, Climatic Research Unit; CV, Co­
efficient of Variation; ESGF, Earth System Grid Federation; GCMs, Global climate models; GPCC, Global Precipitation Climatology Centre; MA, Mean Annual; MBE,
Mean Bias Error; mm, Millimeter; r, Correlation Coefficient; RMSE, Root Mean Square Error; RS, Ranking Scores; rs, Spatial Correlation Coefficient; rt, Temporal
Correlation Coefficient; SEA, Southeast Asia; SeasonAmp, Mean Seasonal Cycle Amplitude; SST, Sea Surface Temperature; SWSA, Southwestern South America; ◦ E,
Degrees East; ◦ N, Degrees North; ◦ S, Degrees South; 27- MODEL ENSEMBLE, Multi-Model Ensemble of 27 Models.
* Corresponding author.
E-mail address: kritana.pru@mahidol.ac.th (K. Prueksakorn).

https://doi.org/10.1016/j.atmosres.2022.106522
Received 27 August 2022; Received in revised form 13 November 2022; Accepted 16 November 2022
Available online 19 November 2022
0169-8095/© 2022 Elsevier B.V. All rights reserved.
S. Pimonsree et al. Atmospheric Research 282 (2023) 106522

significant environmental, economic, and human losses (Alam et al., Iran, showed noticeable differences between the different climate areas.
2018). To be prepared for fluctuating climate change and future natural Another study by Desmet and Ngo-Duc (2022) showed that monsoon
disasters, reliable and efficient tools for analyzing trends and changes in winds and climatic events cause variations in different SEA regions,
precipitation are needed. which is why the region is divided into 3 sub-domains: Indochina, the
Global climate models (GCMs), also known as general circulation Maritime Continent and the Philippines. The result of their study
models, are a useful tool widely applied to study and understand the showed that the magnitudes of annual cycles of each sub-domain
Earth’s climate system in the past, present, and future (Guo et al., 2021; significantly contribute to the higher or lower estimated rainfall in
Yazdandoost et al., 2021). GCMs are numerical models representing each sub-domain in different periods. There is however a small number
processes in the atmosphere, ocean, cryosphere, land surface and the of studies evaluating the efficiency of CMIP6 GCM precipitation simu­
interactions among these components (Randall et al., 2007) which are lation in the SEA region or the sub-domain partitioning in the SEA region
based on physical, chemical, and biological principles (Goosse et al., (Iqbal et al., 2021; Desmet and Ngo-Duc, 2022). In addition, previous
2010). In 1995, the Coupled Model Intercomparison Project (CMIP) was studies have identified different precipitation trends for countries in
established to collect GCMs and define the scope and parameters to SEA. An increasing trend in precipitation was observed in Vietnam and
standardize all GCMs (Meehl et al., 2005; Eyring et al., 2016). CMIP has the Huai Luang watershed in Thailand, while a decreasing trend in
continued to evolve to this day and is in its 6th phase or CMIP6 (Eyring rainfall was apparent in Cambodia (Phuong et al., 2019; Piman et al.,
et al., 2016). GCMs under CMIP6 have been improved in several areas 2016; Thoeun, 2015). Therefore, an investigation on CMIP6 GCMs that
compared to the previous phases (CMIP 1–5) (Eyring et al., 2016) for can provide the most accurate rainfall simulation result for each country
example; spatial resolution, physical parameterization, and simulation should be conducted.
of biogeochemical cycles (Hajima et al., 2020; Jiang et al., 2020; Ayugi Recently, CMIP6 GCMs released new simulation data (the time of
et al., 2021). Additionally, calculation efficiency has also been improved starting this research on 31 November 2021) which have not been
in the current phase to avoid coding errors in the parameterization of entirely evaluated for the SEA region (Iqbal et al., 2021; Desmet and
clouds, convection, and turbulence (Hajima et al., 2020). Moreover, new Ngo-Duc, 2022). The objective of this research is to compare perfor­
features, such as land and ocean nitrogen cycles, ice sheet component, mance in simulating precipitation between newly released GCMs and
and atmospheric radiation and aerosol parameterizations have been formerly studied GCMs with good performance. Not only were the
added in the CMIP6 GCMs (Eyring et al., 2019; Mauritsen et al., 2019; previous evaluations of CMIP6 GCMs in SEA restricted to the mainland
Kamruzzaman et al., 2021; Ayugi et al., 2021). Eyring et al. (2016) re­ SEA, but they also lacked characteristics, seasonal variations, and
ported that these improvements have made the GCMs in CMIP6 more trends. Therefore, the scope of this study covers both mainland and
efficient in terms of understanding the role of clouds in climate sensi­ islands of SEA and evaluates the spatiotemporal precipitation pattern,
tivity and changes in the cryosphere due to climate change and short- seasonal amplitude, fluctuation and trends. The evaluation of CMIP6
term forecast. For example, compared to CMIP5, CMIP6 GCMs have GCMs performance to simulate precipitation over SEA and countries in
shown improvement in predicting the Indian monsoon (Gusain et al., SEA provides information in deciding which model is suitable for them
2020) and precipitation indices over China (Zhu et al., 2020). for further application of SEA impact assessment and planning for
Global climate simulation developers have continued to publish climate change response effectively.
GCM results under CMIP6 since 2017 (Stouffer et al., 2017; Cui et al.,
2021), and there are currently >70 results of CMIP6 GCMs (ESGF the 2. Methodology
Earth System Grid Federation, 2021). The performance of each CMIP6
GCM varies due to model structure, parameterization, initial condition 2.1. Research flow
setting, and differences in climate and topography (Guo et al., 2021;
Wang et al., 2021). The study by Desmet and Ngo-Duc (2022) showed The framework of this study consists of 3 steps as shown in Fig. 1:
that despite using the same simulation model, the evaluation results of first, data are collected from 27 GCMs in CMIP6 and Global Precipitation
precipitation simulations varied from country to country; this was partly Climatology Centre (GPCC). All GCMs are then regridded to GPCC grid
due to the differences in factors affecting precipitation (e.g., season, resolution before separating study domains into SEA domain and each
distance to the ocean, geographical characteristic, monsoon influence, country-specific domain. Second, 27 GCMs CMIP6 are evaluated on their
mountain ranges, ocean currents), all of which contribute to the effi­ performance to simulate precipitation in each study-domain by
ciency of simulation models (Baede et al., 2001). For this reason, CMIP6 comparing with GPCC through statistics. From the result, GCMs will
GCMs have been tested in different areas: Globe (Chen et al., 2020), then be ranked based on their simulation performance based on results
China (Jiang et al., 2020; Xin et al., 2020; Wang et al., 2021; Zhang of all performance values. Last, the performance results of GCMs from
et al., 2022), Central Asia (Guo et al., 2021), Iran (Yazdandoost et al., each criterion are compared to assess the compatability between GCMs
2021), East Asia (Ayugi et al., 2021; Yan et al., 2021), Southwestern and GPCC. The result of the study provides performance ranking among
South America (Rivera and Arnould, 2020), North Africa (Babaousmail 27 GCMs and most suitable of model group for simulation of precipi­
et al., 2021), Tibetan Plateau (Cui et al., 2021), United States (Almazroui tation in SEA-domain and each specific country domain.
et al., 2020), and SEA (Iqbal et al., 2021; Desmet and Ngo-Duc, 2022).
However, Yazdandoost et al. (2021) reported that model performance 2.2. Study area
may vary due to area scope and geography. For example, Wang et al.
(2021) discovered that CMIP6 GCMs overestimated precipitation in In this study, SEA domain evaluation scope is between latitude 13◦ S
upstream and underestimated it in the middle of the Yellow River Basin. to 29.5◦ N and longitude 90◦ E to 143◦ E as shown in Fig. 2. The main
Therefore, in addition to evaluating the efficiency of the model within portion of this domain is located in the tropics and some in the subtro­
the area, some studies analyzed a specific area of the study domain. For pics (Yuen and Kong, 2009), covering a total land area of 4.48 million
example, the location of the Andes Mountain affected the balance of km2 (Gupta, 2010). The geography of SEA consists of mainland and
precipitation volume in the eastern and western parts of the South­ islands. The topography of the area consists of high hills, plateaus, and
western South America (SWSA); therefore, Rivera and Arnould (2020) mountains (placed in the north and northwest ward of the region and
divided the study area into the North SWSA and South SWSA subareas. including some islands in Indonesia and Philippines), river plains
The result of the study showed that in most of the simulations, the values (mostly in the mainland), coastal plains (commonly found in countries
of mean bias error and the root mean square error (RMSE) are higher in with coastline), and islands (Iqbal et al., 2021; Santhanasamy et al.,
the South SWSA area than in the North SWSA area. The study of Yaz­ 2021). Precipitation in the country located in the mainland is mainly
dandoost et al. (2021), which used CMIP6 GCMs in 5 climate areas of influenced by the Asian summer monsoon, while that in the island area

2
S. Pimonsree et al. Atmospheric Research 282 (2023) 106522

Fig. 1. Research framework of this study (RS method: Ranking score method).

for specific country.

2.3. CMIP6 models and observation dataset

Since current release of CMIP6 GCMs models have not been fully
evaluated for SEA, this research, therefore, selects recently released
models (before 31 November 2021) to compare with the 5 highest
performing CMIP6 GCMs previously used in SEA region study (Iqbal
et al., 2021; Desmet and Ngo-Duc, 2022). Hence, 27 GCMs in CMIP6
(Table 1) are used in this study area. The monthly precipitation data
from historical simulations of all GCMs are obtained from the Earth
System Grid Federation (ESGF) data portal (ESGF the Earth System Grid
Federation, 2021). Since this study also aims to evaluate the ability of
model in simulating precipitation trend, the chosen time is 1975–2014
during which there was a significant increase in precipitation (IPCC,
2021). The precipitation simulation of all 27 CMIP6 GCMs and multi-
model ensemble of 27 models, called 27- MODEL ENSEMBLE (Eq. (1))
is assessed against the reference dataset.

1∑ n
27 − MODEL ENSEMBLE = GCMi (1)
n i=1

Fig. 2. Topography (elevation in meters) of the study area. where GCMi is the simulation output of the ith GCM, and n is the total
number of GCMs.
is influenced by the Australian summer monsoon and tropical cyclone Observation data from Global Precipitation Climatology Centre
(Desmet and Ngo-Duc, 2022). With regards to geography, topography, (GPCC) are used as the reference dataset in this study. This is a monthly
and monsoon influence, there is likely significant distinction in precip­ global land-surface precipitation dataset collected from ~80,000 sta­
itation from country to country (Rivera and Arnould, 2020; Desmet and tions worldwide that cover the period of 1901–2016. GPCC was selected
Ngo-Duc, 2022). For these reasons, in order to better evaluate the per­ according to better performance in reproducing the spatial distribution
formance of precipitation simulation, SEA domains are divided into nine of rainfall climatology and their inter-annual variation over the country
countries: Brunei, Cambodia, Indonesia, Lao, Malaysia, Myanmar, in SEA than other grid datasets (CRU and APHRODITE) (Torsri et al.,
Philippines, Thailand, and Vietnam with the exception of Singapore due 2022; Faikrua et al., 2020). GPCC has a spatial resolution of 0.5 degrees
to grid area data). The evaluation results in countries provide decisive (Schneider et al., 2018). Because the spatial resolution of each CMIP6
conclusion in the selection of appropriate GCM to simulate precipitation GCM is different (Table 1), all models are regridded to GPCC grid res­
olution (0.5◦ × 0.5◦ ) using bilinear interpolation before evaluating the

3
S. Pimonsree et al. Atmospheric Research 282 (2023) 106522

Table 1 GCM simulation and GPCC. The lower that values of RMSE, the higher is
List of the 27 CMIP6 GCMs used in this study. the performance of the model (Rivera and Arnould, 2020). The rs is
No. Models Institute Resolution calculated to measure the similarity of the gridded spatial patterns of
(long*lat) GCM and GPCC; rs value is between − 1 and 1. According to the criteria
1. ACCESS-CM2 Australian Community Climate and 1.88◦ × 1.25◦ of Hinkle et al. (2003), r ≥ ±0.91 = very high correlation, ±0.71 ≤ r ≤
Earth System Simulator ±0.90 = high correlation, ±0.51 ≤ r ≤ ±0.70 = moderate correlation,
2. AWI-CM-1-1- Alfred Wegener Institute, Germany 0.94◦ × 0.94◦ ±0.31 ≤ r ≤ ±0.50 = weak correlation, and 0 ≤ r ≤ ±0.30 = very weak
MR correlation. These criteria are also used to indicate the degree of spatial
3. AWI-ESM-1- Alfred Wegener Institute, Germany 1.88◦ × 1.88◦
1-LR
correlation between GCM and observation over Tropical South America
4. CESM2 National Center for Atmospheric 1.25◦ × 0.94◦ in the former study (Dias and Reboita, 2021).
Research, USA The 2nd group of performance indicators focuses on spatially aver­
5. CESM2-FV2 National Center for Atmospheric 2.50◦ × 1.88◦ aged time series, namely Temporal Correlation Coefficient (rt), Coeffi­
Research, USA
cient of variation (CV), Trend, and Mean Seasonal Cycle Amplitude
6. CMCC-CM2- Fondazione Centro Euro-Mediterraneo 1.25◦ × 0.94◦
HR4 sui Cambiamenti Climatici, Italy (SeasonAmp). For this group, there is computed mean precipitation for
7. CMCC-CM2- Fondazione Centro Euro-Mediterraneo 1.25◦ × 0.94◦ each year before computing statistics metrics (except SeasonAmp that
SR5 sui Cambiamenti Climatici, Italy computes mean monthly precipitation). The rt is calculated to judge the
8. CMCC-ESM2 Fondazione Centro Euro-Mediterraneo 1.25◦ × 0.94◦ temporal correlation between the GCM and GPCC; rt at or close to 1
sui Cambiamenti Climatici, Italy
9. E3SM-1-1- E3SM-Project 1.00◦ × 1.00◦
implies the best performance or perfect correlation. While the CV de­
ECA termines the relative variance of precipitation in term of average per­
10. E3SM-1-0 Lawrence Livermore National 1.00◦ × 1.00◦ centage per area, higher value indicating greater variance. The degree of
Laboratory, USA variability of rainfall can be classified into: CV < 20 = less variability,
11. EC-Earth3 EC-Earth-Consortium, Europe 0.70◦ × 0.70◦
20 < CV < 30 = moderate variability, CV > 30 = high variability
12. EC-Earth3- EC-Earth-Consortium, Europe 0.70◦ × 0.70◦
AerChem (Mohammed and Yimam, 2022). Trend is used as an indicator to
13. EC-Earth3-CC EC-Earth-Consortium, Europe 0.70◦ × 0.70◦ determine the movement direction (increase/decrease) of precipitation.
14. EC-Earth3- EC-Earth-Consortium, Europe 0.70◦ × 0.70◦ The positive direction of the calculated slope indicates increasing trend
Veg and vice versa (Azam et al., 2018). As the case of SeasonAmp, the first
15. EC-Earth3- EC-Earth-Consortium, Europe 1.13◦ × 1.13◦
step computes the mean annual cycle of monthly precipitation to find
Veg-LR
16. GFDL-ESM4 NOAA Geophysical Fluid Dynamics 1.25◦ × 1.00◦ the wettest and driest month. After that, the spatial difference between
Laboratory, USA those months is computed as a percentage of mean annual precipitation
17. GISS-E2-1-G Goddard Institute for Space Studies, USA 2.50◦ × 2.00◦ before averaging for all pixels. The SeasonAmp value indicates the
18. GISS-E2-2-H Goddard Institute for Space Studies, USA 2.50◦ × 2.00◦
severity of changes in precipitation. Higher values indicate greater
19. IITM-ESM Centre for Climate Change Research, 1.88◦ × 1.91◦
Indian Institute of Tropical Meteorology, changes of precipitation seasonally.
India
20. IPSL-CM5A2- Institut Pierre Simon Laplace, France 3.75◦ × 1.88◦ 2.5. GCMs ranking
INCA
21. IPSL-CM6A- Institut Pierre Simon Laplace, France 2.50◦ × 1.26◦
27 GCMs from CMIP6 and 27-MODEL ENSEMBLE performances are
LR-INCA
22. KIOST-ESM Korea Institute of Ocean Science and 1.88◦ × 1.88◦ ranked on the precipitation simulation using ranking scores (RS)
Technology, Republic of Korea method, based on 8 indicators. RS is a widely used method in perfor­
23. MRI-ESM2-0 Meteorological Research Institute, Japan 1.13◦ × 1.13◦ mance ranking of GCMs (Rupp et al., 2013; Ruan et al., 2018; Jia et al.,
24. NorCPM1 NorESM Climate Modeling Consortium, 2.50◦ × 1.88◦
2019). RS uses a 3-step ranking process; firstly, absolute error of the
Norway
25. NorESM2-LM NorESM Climate Modeling Consortium, 2.50◦ × 1.88◦
statistical values between GCM and GPCC (ErrorGCM, M) is calculated
Norway using Eq. (2), where GPCCM and GCMM are the evaluation result of the
26. SAM0- Seoul National University, Republic of 1.25◦ × ~0.9◦ performance metric M of GPCC and GCM, respectively. However, in the
UNICON Korea case of r, GPCCM is set as 1. The lower the ErrorGCM, M value, the higher
27. TaiESM1 Research Center for Environmental 1.25◦ × 0.94◦
performance of GCM is indicated (Ruan et al., 2018). Second step is to
Changes, Taiwan, China
calculate relative error of GCMs in each performance metrics M (REGCM,
M) (Eq. (3)). The min and max are the lowest and highest values from all
model performance. GCM respectively. Finally, the total relative error (TREGCM) value is
derived from total sum of relative error by all performance metrics (Eq.
2.4. Performance metrics (3)) where N equals 8.
ErrorGCM,M = ∣ GPCCM − GCMM ∣ (2)
For model performance validation, precipitation outputs of GCM and
GPCC in SEA domain and country-specific domain over the period ( )
ErrorGCM,M − min ErrorGCM,M
1975–2014 are assessed using eight performance metrics, as listed in REGCM,M = ( ) ( ) (3)
max ErrorGCM,M − min ErrorGCM,M
Table 2. These indicators can be categorized into 2 main groups.
The 1st group of performance indicators focuses on evaluating

M
spatial pattern of precipitation, namely Mean annual (MA), Mean bias TREGCM = REGCM,M (4)
errors (MBE), RMSE and Spatial Correlation Coefficient (rs). The MA is N=1

used to present the spatial pattern of the annual precipitation between


The performance ranking of each GCM is based on total relative error
GCMs against the GPCC (Abadi et al., 2018; Almazroui et al., 2020),
values. The lower total relative error of the GCM indicates a better
while the spatial bias direction between them is assessed by the MBE
performance in simulating precipitation closest to the data provided by
(Karim et al., 2020; Rivera and Arnould, 2020; Han et al., 2022; Yaz­
GPCC.
dandoost et al., 2021). A negative or positive direction of MBE indicates
where the precipitation value of GCM is underestimated or over­
estimated as compared to GPCC, respectively (Abbasian et al., 2019).
Whereas RMSE is used in this study to measure range of error between

4
S. Pimonsree et al. Atmospheric Research 282 (2023) 106522

Table 2
Statistical equations of the performance metrics used in this study.
Group Metrics Equation Unit Range
∑N
I MA y=1 Py
mm ∞
N
MBE MAGCM − MAGPCC mm − ∞ to ∞
√̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅
RMSE ∑n ( )2 mm 0 to ∞
p=1 MAGCMp − MAGPCCp
RMSE =
n
∑∑ ( )( )
rs i j MAGCMij − MAGCM MAGPCCij − MAGPCC
NA − 1 to 1
√̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅
(∑ ∑ ( )2 )(∑ ∑ ( )2 )
i j MAGCMij − MAGCM i j MAGPCCij − MAGPCC
∑ ( )( )
II rt y MAGCMy − MAGCM MAGPCCy − MAGPCC
NA − 1 to 1
√̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅
(∑ ( )2 )(∑ ( )2 )
t MAGCMy − MAGCM t MAGPCCy − MAGPCC
)
CV ( SD % ∞
x100
MA
Trend Y = a + bx mm -∞ to ∞
∑ ∑ ∑ ∑
MAy Ty 2 − Ty Ty MAy
when b = ∑ 2 ( ∑ )2
N* Ty − Ty
∑ ∑ ∑
N* Ty MAy − Ty MAy
a = ∑ ( ∑ )2
N* Ty 2 − Ty
SeasonAmp ((Monwettest − Mondriest)/MA) x 100 % − ∞ to ∞

Where Py is annual precipitation for year y and N is total number of the years; MAGCM is mean annual of GCM; MAGPCC is GPCC data; MAGCMp is the mean annual of the
GCM for a pixel p. MAGPCCp is the mean annual of the GPCC for a pixel p. n is total number of the pixels; MAGCMij is mean annual precipitation of GCM at row i and
column j; MAGPCCij is mean annual precipitation of GPCC at row i and column j. MAGCM is the average of mean annual precipitation of GCM. MAGPCC is the average of
mean annual precipitation of GPCC; MAGCMy is mean annual precipitation of GCM for year y; MAGPCCy is mean annual precipitation of GPCC for year y; MAGCM is the
average of mean annual precipitation of GCM; MAGPCC is the average of mean annual precipitation of GPCC; Monwettest is mean warmest month of model for 40-years
period; Mondriest is mean coldest month of model for 40-years period; SD is standard deviation; Y is forecast for period x, a is the intercept; b is slope; x is the time
period; Ty is the time series studied in year y. MAy is the annual mean for year y.

3. Results and discussion According to Fig. 3, typical GCMs simulate similarly spatial pattern
with high precipitation in maritime area and low precipitation in
3.1. Spatial pattern of precipitation simulations mainland SEA but in the detailed analysis, each model has different
accuracy to simulate different spatial distribution and volume of pre­
The performances in simulating spatial pattern of precipitation of cipitation. The variations of precipitation in the SEA region depend on
GPCC, 27-MODEL ENSEMBLE, and each of the GCMs were evaluated many factors, especially the occurrence of convective rain over the
under the spatial distribution of mean annual precipitation during equator which has highly complex change causing difficulty to set pa­
1975–2014 as shown in Fig. 3. Model performance in SEA domain and rameters of cumulus scheme in simulating climate in each simulation
country domain are shown in Table A1-A10 in Supplementary Material. (Ngo-Duc et al., 2017; Zhou and Han, 2018; Doutreloup et al., 2019).
GPCC data presented high changes in precipitation amount in SEA re­ However, there were several GCMs which showed spatial distribution in
gion with the values ranging between 654.75 and 7083.70 mm. The SEA similar to GPCC such as; 27-MODEL ENSEMBLE, CESM2-FV2,
average precipitation for 1975–2014 of GPCC in SEA domain is 2389.22 NorESM2-LM and families of EC-Earth3, with low relative error value
mm. where Brunei is the country with the highest amount of precipi­ as seen in RMSE (Fig. 4). At the same time, some GCMs such as; AWI-
tation in the region, with MA value at 3450.25 mm. Whereas Thailand ESM-1-1-LR, and IPSL-CM6A-LR-INCA did not simulate well as notice­
was found to have the lowest MA value in the region, at 1538.50 mm. able from the spatial distribution of the simulated precipitation that are
The area average of MA values in SEA of all GCMs and 27-MODEL not similar to the pattern of GPCC and from the RMSE values which were
ENSEMBLE is in the range of 1335.74 - 3346.93 mm. The major fac­ as high as 1519.42 mm and 2491.07 mm respectively. When taking into
tors contributing to the changes of precipitation in SEA include monsoon consideration the value of MBE in Fig. 5, most simulation have over­
influence, low air pressure/monsoon trough, and the storm path (Des­ estimation for countries located in maritime as well as those in Thailand,
met and Ngo-Duc, 2022). Compared from Figs. 2 and 3, it can be noticed Cambodia and central Myanmar. Underestimation was seen in IITM-
that coastal areas or mountain ranges are the areas with highest pre­ ESM and KIOST-ESM in some simulations especially in Indonesia. Un­
cipitation. For example, in western area of mainland Myanmar; pre­ derestimations are also found in Lao, the northern Vietnam and the
cipitation is greatly influenced by the Southwest monsoon from the western area of Myanmar in several models which are related with low
Indian ocean. The eastern area of mainland precipitation is greatly performance for simulation of the precipitation influenced by southwest
influenced by Northeastern monsoon from South China sea passing monsoon. When considering the results of spatial precipitation simula­
through mountain ranges in Vietnam causing heavy precipitation in tion in Fig. 3 with the topography map shown in Fig. 2, it was found that
Vietnam, Lao, and Cambodia. The southern region of Thailand and several models showed low performance in the mountainous area. This
Malaysia is influenced by Southwest and Northeast monsoons, and those had also been reflected in previous studies where GCM over predicted
countries located in the maritime area precipitation are influenced by precipitation in mountainous areas, e.g. in the Tibetan Plateau, China
passing of Australian summer monsoon and the typhoon path occurred (Su et al., 2013), in the Andes, between Argentina and Chile (Rivera and
in the Pacific Ocean (Desmet and Ngo-Duc, 2022). Furthermore, the Arnould, 2020), in the southern Qinling Mountains, China (Wang et al.,
changes in precipitation volume in SEA is also influenced by the 2021), and in the Tianshan Mountain, between Central Asia and the
topography (Iqbal et al., 2021) since precipitation have been found low northwestern part of China (Guo et al., 2021). This can be due to limited
in flat area (middle of mainland) and high in mountain ranges (i.e., simulation performance in mountainous areas which have complex
northern area of Myanmar, Papua, and Borneo islands) which showed a circulation, possibly generating precipitation from orographic lifting
correlation with precipitation-elevation (Guo et al., 2021). (Adam et al., 2006). At the same time, the limited availability of

5
S. Pimonsree et al. Atmospheric Research 282 (2023) 106522

Fig. 3. Spatial distribution of mean annual precipitation (MA) of GPCC, 27-MODEL EENSEMBLE and 27 GCMs over the SEA land surface during 1975–2014.

measurement data in mountainous areas also posed inaccuracy in pre­ domain as shown in Fig. 4. Furthermore, it was also noticed that simu­
cipitation simulation since gridded data building of initial conditions lation models from the same developers tend to exhibit similar results in
required sufficient data covering spatial variation and uniform distri­ precipitation pattern and volume for example, EC-Earth models group
bution of measurement (Su et al., 2013; Guo et al., 2021). Furthermore, and GISS models group. It can be hypothesized that simulation models
the simulation performance of precipitation in high mountainous areas made by the same developer apply similar basic configurations resulting
may be low because some GCMs with low spatial resolutions were not in similar simulation results (Döscher et al., 2022). Considering the
able to simulate precipitation variations in the mountainous regions difference between the models and observation based on RMSE; 27-
(Yazdandoost et al., 2021). For example, GISS-E2–1-G and GISS-E2–2-H MODEL ENSEMBLE, CESM2-FV2, and EC-Earth3-CC perform well in
exhibit high wet biases (>3000 mm) in high mountain area especially the SEA domain, with values of 662.78 mm, 704.10 mm, and 715.82
north of Myanmar as shown in Fig. 5. The results of this study show that mm, respectively. The best performance considering the RMSE in
even with the advancement of current CMIP6 GCMs, precipitation Brunei, Cambodia, Indonesia, Lao, Malaysia, Myanmar, Philippines,
simulation in mountainous regions still posed challenges. Additionally, Thailand, and Vietnam are CMCC-CM2-SR5, EC-Earth3-CC, 27-GCM-
the findings also shown that some simulation models tended to under­ ENSEMBLE, EC-Earth3-Veg-LR, 27-GCM-ENSEM, ACCESS-CM2,
estimate annual precipitation such as in the case of IITM-ESM and CESM2, EC-Earth3-Veg, and EC-Earth3-Veg-LR with RMSE values of
KIOST-ESM which revealed significant dry bias in almost all regions. 60.26 mm, 379.04 mm, 680.06 mm, 367.43 mm, 454.93 mm, 753.32
This results in both simulation models being ranked as the top 5 worst mm, 771.67 mm, 348.35 mm, and 399.64 mm, respectively.
with the highest RSME value in SEA-domain and all specific country In addition, the study also considers the consistency in spatial

6
S. Pimonsree et al. Atmospheric Research 282 (2023) 106522

Fig. 4. RMSE values of mean annual precipitation in 1975–2014 of 27-MODEL ENSEMBLE and models based on GPCC.

distribution between simulation models and GPCC data set using spatial 0.63, respectively. From the values of rs, it can be noticed that lower
correlation coefficient (rs). The results shown in Table A.1. in the Sup­ model performance to simulate spatial variation was found in the
plementary Material found the simulation models that best simulate maritime area compared to that in mainland SEA. In Brunei case, even
spatial variation of rainfall are 27- GCM-ENSEMBLE and ACCESS-CM2 though all simulation models underwent spatial resolution downscaling
with the rs values of 0.7, and most of simulation models in family of prior to the evaluation, the data is sparse since there are only 2 grid cells
EC-Earth3 has the rs values of 0.67. The best performance considering data availability within the domain. The insufficient amount of datasets
the rs in Cambodia, Indonesia, Lao, Malaysia, Myanmar, Philippines, available is not suitable for evaluating correlation coefficient (Aggarwal
Thailand, and Vietnam are 3 GCMs in EC-Earth3 family (EC-Earth3- and Ranganathan, 2016); therefore, rs of Brunei domain was not
AerChem/EC-Earth3-CC/EC-Earth3-Veg-LR), ACCESS-CM2, EC-Earth3- evaluated.
AerChem, CMCC-ESM2, EC-Earth3-AerChem, CMCC-ESM2, EC-Earth3-
Veg, EC-Earth3 with rs of 0.84, 0.57, 0.64, 0.72, 0.76, 0.25, 0.75 and

7
S. Pimonsree et al. Atmospheric Research 282 (2023) 106522

Fig. 5. Spatial distribution of mean annual precipitation biases (MBE) of 27-MODEL EENSEMBLE and 27 GCMs over the SEA land surface during 1975–2014.

3.2. Temporal assessment of precipitation simulations domains for all models and GPCC (1st and 3rd columns). The study on
the simulation performance of annual precipitation changes showed that
Since the volume and fluctuation of precipitation varies for each the values of the temporal correlation coefficient (rt) range from − 0.25
country within the SEA region, the evaluation of the performance was to 0.40 in the SEA, − 0.40 to 0.28 in the Brunei, − 0.20 to 0.38 in the
based on the ability to detect fluctuation of precipitation on temporal Cambodia, − 0.24 to 0.33 in the Indonesia, − 0.40 to 0.32 in the Lao,
basis. The analysis found that the best model performance in terms of − 0.23 to 0.40 in the Malaysia, − 0.38 to 0.54 in the Myanmar, − 0.26 to
magnitudes of MA over SEA is for IPSL-CM5A2-INCA (MA = 2392.39 0.26 in the Philippines, − 0.23 to 0.19 in the Thailand, and − 0.24 to
mm), with a difference from GPCC of only 3.17 mm. The models that 0.35 in the Vietnam (Fig. 7a). The inter-annual cycle of precipitation
best perform for MA in Brunei, Cambodia, Indonesia, Lao, Malaysia, from almost all the models has no significant correlation with the GPCC
Myanmar, Philippines, Thailand, and Vietnam are EC-Earth3-Veg-LR, for all domains (significance level < 0.05). The result showed that most
E3SM-1-0, GISS-E2–1-G, AWI-ESM-1-1-LR, ACCESS-CM2, SAM0- models in this study did not perform well in simulating the inter-annual
UNICON, NorCPM1, EC-Earth3-Veg-LR, and CESM2 with the mean change of precipitation within SEA-domain and each specific country
annual precipitation of 3441.08 mm, 1892.56 mm, 2903.30 mm, domain.
1791.57 mm, 2930.50 mm, 2060.45 mm, 2443.23 mm, 1520.32 mm, The degree of inter-annual variability precipitation can be observed
and 1939.29 mm, respectively. The differences from GPCC were 9.18 the value of CV in Fig. 7b. The CV of GPCC in SEA domain is 7.16%,
mm, 4.58 mm, 45.95 mm, 65.04 mm, 3.34 mm, 11.81 mm, 4.37 mm, while that in Philippines is highest (13.87%) and lowest in Cambodia
18.18 mm, and 26.37 mm, respectively. Fig. 6 shows the time series of (8.26%). Major factors contributing to the high fluctuation of precipi­
annual precipitation during 1975–2014 over SEA domain and 9 country tation in the SEA region for Philippines are monsoon variation, El Nino/

8
S. Pimonsree et al. Atmospheric Research 282 (2023) 106522

Fig. 6. Time series of mean annual of precipitation (1st and 3rd columns) and mean annual cycle of monthly precipitation (2nd and 4th columns) during the
period 1975–2014.

Southern Oscillation and cyclonic anomalies (Akasaka, 2010; World Philippines. In all domains, comparisons of CV among GCMs reveal that
Bank and GFDRR, 2011; Kubota et al., 2017; Olaguera et al., 2018; there are no noticeable differences between simulation models. CV
Bagtasa, 2020). Furthermore, Philippines is geographically one of the values of most simulation models are closer and consistent with GPCC;
most typhoon-prone counties (Bohra-Mishra et al., 2017) resulting in for example, for the SEA domain, the CV value of GPCC is 7.16 whereas
Philippines being affected by typhoon 6–9 times a year (Strobl, 2019). the simulation models values ranged between 3.42 and 9.99 with the
The severity of typhoons also varied from year to year which is one of exception of the 27 model ensemble, which has a value of 1.38. In the
the factors influencing high fluctuation of precipitation in the case of performance evaluation for SEA domain simulation, it was found

9
S. Pimonsree et al. Atmospheric Research 282 (2023) 106522

Fig. 7. Values of rt (a), CV (b), Trend (c), and SeasonAmp (d) for years 1975–2014.

10
S. Pimonsree et al. Atmospheric Research 282 (2023) 106522

that CMCC-ESM2 has the variance value similar to GPCC, at the rate October to January causing large volume of rainfall during the time (wet
difference of only 0.02%. The best-performed models that can reproduce season) and dry season during June to August. However, it was found
the inter-annual variability of precipitation in Brunei, Cambodia, that precipitation in Philippines was quite different from other coun­
Indonesia, Lao, Malaysia, Myanmar, Philippines, Thailand, and Vietnam tries, with the Southwest monsoon causing significant spike in May and
are NorCPM1, CMCC-CM2-SR5, CMCC-CM2-SR5, CMCC-ESM2, CMCC- continuing to influence the rainfall onwards until November before
ESM2, ACCESS-CM2, KIOST-ESM, EC-Earth3-CC, and IPSL-CM6A-LR- decreasing in December. Furthermore, due to Philippines’s location in
INCA. CMCCs were able to remarkably simulate precipitation vari­ the typhoon belt, the occurrences of typhoon causing a high rainfall
ability in the study wells at five out of ten study domains. In addition, it level are generally in July to October (de Viana, 2014). It is noted that
is observed from Fig. 7b that 27- GCM ENSEMBLE has the furthest only a few models that can catch the increase of rainfall level in October
agreement with the CV of GPCC in all domains. Considering the change for Philippines’s case. The results of the average precipitation cycle are
tendencies of annual precipitation from 1975 to 2014, GPCC shows an used to calculate SeasonAmp to measure precipitation differences in
increasing trend in precipitation in all domains (Fig. 7c). SEA domain each area between highest rainfall and lowest rainfall month (Fig. 7d).
showed an increasing volume of precipitation at 5.82 mm/ year which is Over SEA, all simulations perfomed quite well with SeasonAmp values
consistent with frequency and severity of floods related to an increase in ranged between 5.53% (AWI-ESM-1-1-LR) to 9.69% (KIOST-ESM).
the number of heavy rains and tropical storms in the past decades (Torti, SeasonAmp in Fig. 7d indicates noticeable difference between countries
2012). The value of precipitation in each country domain in the region lated in mainland and maritime. Precipitation change in Cambodia, Lao,
illustrated increasing trend ranging from 1.64 mm/year to 11.74 mm/ Myanmar, Thailand, and Vietnam were noticeably more for SeasonAmp
year, being lowest in Myanmar and highest in Philippines. The SST as compared to those in Brunei, Indonesia, Malaysia, and Philippines.
anomalies, severity of monsoon, and the fluctuation of low pressure/ The figure also showed that all models were able to simulate the Sea­
monsoon trough are important factors in increasing precipitation within sonAmp of precipitation changes in consistent with GPCC. GISS-E21-G
the region (Dado and Takahashi, 2017; Trinh-Tuan et al., 2019; Taka­ performed best in SEA, with SeasonAmp value of 6.53%. The best per­
hashi et al., 2020; Fakaruddin et al., 2022; Tomkratoke and Sirisup, formance in Brunei, Cambodia, Indonesia, Lao, Malaysia, Myanmar,
2022). In the study by Lasco et al. (2009), the increment of precipitation Philippines, Thailand, and Vietnam are with SAM0-UNICON, GISS-
in Philippines has a high relation to frequency and severity of tropical E2–2-H, GISS-E2–2-H, CMCC-ESM2, EC-Earth3-CC, CMCC-CM2-SR5,
cyclone occurrence in the Western Pacific, where approximately 532 SAM0-UNICON, TaiESM1, and CMCC-CM2-HR4, with an SeasonAmp
tropical cyclones have occurred between 1975 and 2004. Of all the value of 5.24%, 17.53%, 5.11%, 19.60%, 5.80%, 18.39%, 6.90%,
models, only ACCESS-CM2 and AWI-ESM-1-1-LR can capture the posi­ 16.84%, and 14.21%, respectively.
tive trends like with GPCC in every study domain, while the other
models rather fail to catch the wetting trends, especially EC-Earth3- 3.3. Model ranking
AerChem. In terms of trend magnitude over SEA domain, AWI-ESM-1-
1-LR showed changing trend of precipitation between 1975 and 2014 Based on the results from Sections 3.1 and 3.2, it can be seen that
closest to the observed trend. In the case of other domain, the results each simulation performed differently on the different metrics. No single
(Fig. 7c) reveal that IPSL-CM6A-LR-INCA, SAM0-UNICON, IPSL-CM6A- model can perform best on all metrics. For example, in the SEA domain,
LR-INCA, KIOST-ESM, EC-Earth3, TaiESM1, AWI-ESM-1-1-LR, ACCESS- ACCESS-CM2 is the best model for rs; however, it is ranked third for
CM2, and NorCPM1 had trend values nearest to GPCC in Brunei, Trend and eleventh for MA. Therefore, multiple performance metrics are
Cambodia, Indonesia, Lao, Malaysia, Myanmar, Philippines, Thailand, usually applied in simulation assessment (Jia et al., 2019; Iqbal et al.,
and Vietnam, respectively. The simulation of precipitation trends is 2021; Desmet and Ngo-Duc, 2022; Wang et al., 2021). The different
therefore a challenging task in developing more accurate simulations for results from each metric must be summed up to provide the overall
the SEA region. performance of each simulation using the RS method. Fig. 8 shows the
Previous research has shown that monthly precipitation climatology relative error values of each metric of each model in the SEA-domain
varies by country/area (Rivera and Arnould, 2020; Guo et al., 2021; Li and each specific country domain. The relative error value of the
et al., 2021; Yazdandoost et al., 2021; Desmet and Ngo-Duc, 2022). For metric presents the gap value between GCM and GPCC, which ranged
this reason, this research also verifies performance of all models in between 0 and 1. Therefore, the GCMs which have the relative error
simulating the monthly precipitation in all study domains. Fig. 6 (2nd value closest to 0 can indicate higher performance in that metric; on the
and 4th columns) show mean seasonal cycles of precipitation averaged other hand, GCMs with lower performance have relative error value
over the SEA domain and 9 country domains for the period 1975–2014. closer to 1. Finally, GCMs performance can be ranked by summarizing
GPCC shows that monthly precipitation over SEA ranges from 133.99 all relative error values derived from all indicators, the lower value
mm in February to 254.10 mm in August. It was discovered that volume indicating better simulation performance with the model values closer
of precipitation increases slightly from March to August before to GPCC.
decreasing from September to December. Overall, it can be seen that all The results of performance ranking are shown in Fig. 8 from best to
models in SEA were able to create seasonal cycles consistent with GPCC worst performing (least to most total relative error) as seen from top to
data where as many as 15 models including the 27-GCM-ENSEMBLE bottom. For the SEA domain, the lowest total relative error score shows
were able to detect minimum and maximum precipitation month in that TaiESM1 (1.25) is clearly top-ranked and is followed by EC-Earth3
February and August, respectively. However, there were also several (1.40), and ACCESS-CM2 (1.55), respectively. The result of this study is
models which have failed to generate the shape of the seasonal cycle consistent with previous CMIP6 study by Iqbal et al. (2021) and Desmet
where their simulation results of monthly precipitation were too low and Ngo-Duc (2022) which had reported that EC-Earth3 and ACCESS-
when compared to GPCC (e.g., IITM-ESM and KIOST-ESM) and notice­ CM2 were among top ranking model in precipitation simulation
ably too high (e.g., IPSL-CM6A-LR-INCA). The seasonal cycle of all areas assessment in SEA. Since TaiESM1 (ranked 1 in this study) is a newly
indicates significant relations between volume of precipitation in each published simulation model, its performance has not been evaluated in
country with the monsoon occurrences (Moon and Ha, 2017). Data SEA in Iqbal et al. (2021) and Desmet and Ngo-Duc (2022). Therefore,
analysis revealed that seasonal patterns of precipitation volumes in this study helps update the top ranking precipitation simulation model
mainland of Cambodia, Lao, Myanmar, Thailand, and Vietnam were in SEA where TaiESM1 is ranked first. The models from the institute of
similar, with heavy rainfall from May to September (Fig. 6), consistent EC-Earth-Consortium (EC-Earth3 and EC-Earth3-Veg-LR) were shown as
with the Southwest monsoon. It is not the same with the countries best performers for the domains of Brunei, Cambodia, Indonesia, Lao,
located in maritime areas (with the exception of Philippines) showing Malaysia and Vietnam with the total relative error value <1.24. While
the simulation results related to the Australian summer monsoon during the models with relatively highest performance to simulate precipitation

11
S. Pimonsree et al. Atmospheric Research 282 (2023) 106522

Fig. 8. Rank of 27 CMIP6 models and 27-MODEL ENSEMBLE based on relative error calculated by the summation of 8 performance metrics for (a) SEA, (b) Brunei,
(c) Cambodia, (d) Indonesia, (e) Lao, (f) Malaysia, (g) Myanmar, (h) Philippines, (i) Thailand, and (j) Vietnam. Number inside each metric is the ranking number. [In
Brunei’s case, rs is not evaluated; therefore, there is no result value for rs for the Brunei domain. It is defined as NaN.]

in Myanmar, Philippines, and Thailand are CESM2-FV2, CESM2-FV2, outstandingly and comparably, as these models are often found in the
and TaiESM1, respectively. Even though TaiESM1 is the best model for high-performance category even in subequatorial climates. Only EC-
SEA domain, EC-Earth3 and EC-Earth3-Veg-LR also performed better in Earth3-AerChem performed worse compared to the models developed
different country domains. Model ranking also indicated that KIOST- by the same organization, ranked 15th, 22nd, 15th, 18th, 21st, and 15th
ESM and AWI-ESM-1-1-LR showed limited performance for precipita­ in Cambodia, Lao, Myanmar, the Philippines, Thailand, and Vietnam,
tion simulation in the regional level, with total relative error higher than respectively. All the EC-Earth models in this study have basic configu­
other models in almost all study domain cases. The simple mean rations derived from EC-Earth3 and higher grid resolution than the other
ensemble based on 27 models has mediocre performance in SEA and the datasets, and thus generally performed better than the other models.
countries in SEA. This is probably because the 27-MODEL ENSEMBLE is However, what set EC-Earth3-Veg-LR apart from the developing group is
an average result of good and bad performing models. Therefore, many that it has an additional component for simulating interactive vegetation
studies have chosen the 3–10 best performing models for computing at low resolution, as well as additional new physical parameterizations
ensembles (Ahmadalipour et al., 2017; Hussain et al., 2018; Khan et al., and improved tuning parameters (Döscher et al., 2022). These could be
2018; Xuan et al., 2017; Guo et al., 2021). the reasons why EC-Earth3-Veg-LR performs better than EC-Earth3, EC-
Arranged by climate zones, most of the study areas are in the equa­ Earth3-Veg, EC-Earth3-CC and EC-Earth3-AerChem including other
torial and subequatorial climatic zones. Fig. 3 revealed that the perfor­ models in simulating precipitation in Cambodia, Lao, Thailand, and
mance of CMIP6 differs notably in different climatic zones. The mean Vietnam. At the same time, it was found that the best models for
annual precipitation for most areas in Cambodia, Lao, Myanmar, the simulating precipitation in Myanmar and the Philippines differ from
Philippines, Thailand, and Vietnam, which are in the subequatorial those of other countries in the same climate zone. The best model for
climate zone, is between 1000 and 2000 mm, which is markedly simulating precipitation for both Myanmar and the Philippines is
different from Brunei, Indonesia, and Malaysia, which are in the equa­ CESM2-FV2. Looking at the location of Myanmar and the Philippines, it
torial climate zone and have a mean annual precipitation of 2000 to is found that both countries are more affected by stronger monsoons
6000 mm. The model ranking by all metrics in Fig. 8 shows that coun­ than other countries in the same climate zone and they are located in the
tries in the same climate zone or with similar climate factors show re­ passage area of oceanic storms (Desmet and Ngo-Duc, 2022). As a result,
sults in the same direction. The model ranking results clearly show that rainfall in both countries is quantitatively consistent and can be as high
the model group of EC-Earth-Consortium is among the most effective as 4500 mm in some areas (Fig. 3). Moreover, the evaluation results of
models for simulating rainfall in subequatorial climates, especially EC- model performance in this study clearly show that the best performing
Earth3-Veg-LR, which is ranked first in Cambodia, Lao, Thailand, and models (top 5) in Myanmar and the Philippines are quite different from
Vietnam and third in Myanmar. It was also found that three models, those in the same climate zone. So, alghough EC-Earth3-Veg-LR is able
namely EC-Earth3, EC-Earth3-Veg and EC-Earth3-CC, performed to simulate precipitation well in most of the countries in the

12
S. Pimonsree et al. Atmospheric Research 282 (2023) 106522

subequatorial climates, CESM2-FV2 is able to detect monsoon-


influenced precipitation better than other models in Myanmar and the
Philippines. When considering model performance in equatorial cli­
mates, EC-Earth3 proved to be highly effective in simulating precipita­
tion with results ranked 1st in Brunei and Malaysia and 2nd in
Indonesia. The mountainous terrain and abundant agricultural land in
Indonesia might be the reason that EC-Earth3-Veg-LR, with further de­
velopments and particular improvements in interactive vegetation at
low resolution, is the most efficient. Overall, many countries within the
same climate zone show a well- consistent model group. However, this
research also indicates that there are other factors affecting performance
evaluations in different countries, leading to different results even in the
same climate zones.

3.4. Multi-model ensemble mean of top 5 models

Multi-model ensemble of the top 5 in the SEA domain is calculated


for the test of its accuracy compared to other single models in the SEA
region. Table 3 and Fig. 9 show that the ensemble mean of the top 5 can
further reduce the simulation errors of the model’s precipitation in
terms of mean, bias direction, error, and correlation compared to the
best model in SEA (TaiESM1). Although the multi-model ensemble
method can improve this simulation performance in many aspects, it
also has limitations. Table 3 illustrates the effects of averaging methods
for multiple model, especially in terms of variance. It is evident that the
5-GCM-ENSEMBLE’s CV is significantly reduced, which differs greatly
from GPCC compared to TaiESM1 and 27-GCM-ENSEMBLE.
Eyring et al. (2019) reports that CMIP6 GCMs improves the process
of physical parameterization and higher spatial resolution; there are also
more components of Earth system processes than in CMIP5. However,
performance comparisons with previous GCMs can help validate capa­ Fig. 9. Rank of 27 CMIP6 models, 5-MODEL ENSEMBLE, and 27-MODEL
ENSEMBLE based on relative error calculated by the summation of 8 perfor­
bilities and increase confidence in the selection of the CMIP6 model,
mance metrics for SEA.
which has been evaluated as suitable for SEA precipitation simulations
(ACCESS-CM2, CMCC-CM2-SR5, EC-Earth3, EC-Earth3-Veg-LR, and
TaiESM1). Kamworapan and Surussavadee (2019) evaluated the effi­ covering both spatial and temporal variation. The results of all perfor­
ciency of 40 CMIP5 GCMs to find suitable models for simulating climate mance metrics were tabulated to rank the overall performance of each
in the SEA region, and found that CESM1-BGC, CCSM4, CNRM-CM5–2, GCM using RS method.
CESM1-FASTCHEM, and CESM-CAM5 are a family of high-performance The results of this study can be summarized as follows: First, some
models for precipitation simulation. Therefore, these five best models models performed well in simulating spatial distribution of precipitation
from Kamworapan and Surussavadee (2019) are selected in this study to in SEA region and country domains, with more precipitation level in
determine the effectiveness of precipitation simulation. The spatial maritime and less in mainland. However, most models tended to have
distribution of mean annual precipitation bias (MBE) of the multi-model wet biases in countries located in maritime countries, Thailand, central
ensemble of selected CMIP5 and CMIP6 GCMs during the period Myanmar and Cambodia, while they tend to have dry biases in
1960–1999 is shown in Fig. 10. It can be observed that the spatial bias Indonesia, Lao, northern Vietnam and the western area of Myanmar. In
patterns of the two phases are similar and in the same direction. How­ addition, there were limitations in precipitation simulation in moun­
ever, CMIP5 showed a large dry bias in Indonesia, especially in Sumatra tainous areas in several models of CMIP6 with similar results to previous
and Borneo. In contrast, the wet bias in the eastern Borneo Islands region phases. Moreover, there are many GCM datasets with different grid
in CMIP6 was slightly higher than in CMIP5. Overall, it appears that the resolutions. The uncertainty might exist due to re-gridding GCMs to be
improvement in the CMIP6 model contributes to CMIP6 capturing the the same grid resolution as the observation dataset. These results,
precipitation patterns in the SEA area better than CMIP5 with lower indicate that rainfall simulation in high mountain range is still a key
MBE and RMSE and higher rs (Table 4). challenge for model developers. In the SEA domain, 15 out of the 27
CMIP6 GCMs and the 27-MODEL ENSEMBLE showed a moderate degree
of spatial correlation between simulation models and GPCC (rs > 0.6).
4. Conclusions
This also shows that at least one of the models in Cambodia, Malaysia,
Myanmar, and Thailand has a high spatial correlation (rs > 0.7), while
This research evaluated the performances of 27 CMIP6 GCMs in
none of the models shows rs >0.6 for Indonesia and Philippines. Second,
simulating the precipitation of the SEA region and nine countries therein
most CMIP6 models perform well in terms of inter-annual variability of
by comparing with the GPCC gridded datasets for the period of 1975 to
precipitation during 1975–2014 in all domains, with CV very close to
2014. GCMs’ performance was evaluated by using performance metrics

Table 3
MA, MBE, RMSE, rs, rt, CV, Trend, and SeasonAmp values for years 1975–2014 in SEA.
Models MA (mm) MBE (mm) RMSE (mm) rs rt CV (%) Trend (mm) SeasonAmp (%)

GPCC 2389.22 0 0 1 1 7.16 5.82 6.53


TaiESM1 2313.59 − 75.62 784.67 0.59 0.21 7.39 1.30 7.14
5- GCM-ENSEMBLE 2384.23 − 4.99 649.69 0.72 0.44 3.14 1.20 7.57

13
S. Pimonsree et al. Atmospheric Research 282 (2023) 106522

Fig. 10. Spatial distribution of mean annual precipitation biases (MBE) of selected (a) CMIP5 and (b) CMIP6 over the SEA land surface during 1960–1999.

CRediT authorship contribution statement


Table 4
Values of MBE, RMSE and rs of selected CMIP5 and CMIP6 GCMs.
Sittichai Pimonsree: Conceptualization, Methodology, Validation,
MBE (mm) RMSE (mm) rs Formal analysis, Writing – original draft, Supervision. Suchada Kam­
CMIP5 − 68.54 742.55 0.59 worapan: Conceptualization, Methodology, Software, Validation,
CMIP6 49.08 668.35 0.68 Formal analysis, Investigation, Data curation, Visualization. Shabbir H.
Gheewala: Writing – review & editing, Supervision. Amornpong
Thongbhakdi: Writing – review & editing. Kritana Prueksakorn:
GPCC. Only 27-MODEL-ENSEMBLE was not able to simulate variation in
Conceptualization, Methodology, Validation, Resources, Writing – re­
precipitation during the study as there was the furthest agreement with
view & editing, Supervision, Project administration, Funding
the CV of GPCC. The analysis of observations found that the precipita­
acquisition.
tion in the SEA region and 9 countries produces an increasing trend.
Only 2 out of 27 models can capture the positive trends like with GPCC
in every study domain, while the other models rather fail to catch the
Declaration of Competing Interest
wetting trends. For all study domains, most models can simulate the
seasonal cycle of precipitation corresponding to GPCC. The magnitude
The authors declare that they have no known competing financial
of seasonal variation of precipitation for countries located in mainland
interests or personal relationships that could have appeared to influence
and maritime were largely different, with fairly high SeasonAmp values
the work reported in this paper.
in Cambodia, Lao, Myanmar, Thailand and Vietnam and fairly low
SeasonAmp values in Brunei, Indonesia, Malaysia, and Philippines.
Data availability
Lastly, this research also pointed out that the best GCM in precipitation
simulation of each study-domain have varied results. The result showed
We have shared the link to the data used in this study in the reference
that TaiESM1 is the top performer in simulating precipitation in SEA-
list.
domain followed by EC-Earth3 and ACCESS-CM2, respectively. Even
though TaiESM1 was ranked as a top performer for SEA domain, it is not
the best performing model for each country’s domains of the SEA region Acknowledgement
(except Thailand). EC-Earth3 and EC-Earth3-Veg-LR from the institute
of EC-Earth-Consortium perform best in Brunei, Cambodia, Indonesia, This research project is supported by Faculty of Environment and
Lao, Malaysia and Vietnam, while CESM2-FV2 is the highest perfor­ Resource Studies, Mahidol University. The authors also acknowledge the
mance to simulate precipitation in Myanmar and Philippines. Addi­ climate modelling groups for making their model output and the World
tionally, it was found that 27- MODEL ENSEMBLE based on all models in Climate Research Programme’s Working Group on Coupled Modelling
this study has ordinary performance in both SEA and the country in SEA. that established the Coupled Model Intercomparison Project, which is a
Accurate data of climate change from GCMs are a crucial input in project to collect all GCMs. We are grateful to the Earth System Grid
assessing the impact of climate change. This study provides useful in­ Federation (ESGF) for archiving and sharing the dataset CMIP6 GCMs.
formation in the selection of GCMs for climate change applications in
SEA region and each country domains in SEA. The simulation of pre­ Appendix A. Supplementary data
cipitation in tropical area is quite difficult due to complexities of tem­
poral and spatial characteristics. Hence, the simulation of precipitation Supplementary data to this article can be found online at https://doi.
in SEA still poses challenges for model developers. The uncertainty in org/10.1016/j.atmosres.2022.106522.
this work also provide feedback information for improvement of climate
change model for SEA in the future. For further studies, the best model in References
this study is crucial in using it to be a reference for future precipitation
research to investigate precipitation variation and trends in the SEA in Abadi, A.M., Oglesby, R., Rowe, C., Mawalagedara, R., 2018. Evaluation of GCMs
historical simulations of monthly and seasonal climatology over Bolivia. Clim. Dyn.
the 21st century for developing and creating adaptation and mitigation 51, 733–754. https://doi.org/10.1007/s00382-017-3952-y.
strategies for disaster risk management and sustainable water resource Abbasian, M., Moghim, S., Abrishamchi, A., 2019. Performance of the general circulation
management. models in simulating temperature and precipitation over Iran. Theor. Appl. Climatol.
135, 1465–1483. https://doi.org/10.1007/s00704-018-2456-y.
Adam, J.C., Clark, E.A., Lettenmaier, D.P., Wood, E.F., 2006. Correction of global
precipitation products for orographic effects. J. Clim. 19, 15–38. https://doi.org/
10.1175/jcli3604.1.

14
S. Pimonsree et al. Atmospheric Research 282 (2023) 106522

Aggarwal, R., Ranganathan, P., 2016. Common pitfalls in statistical analysis: the use of Goosse, H., Barriat, P.Y., Loutre, M.F., Zunz, V., 2010. Introduction to Climate Dynamics
correlation techniques. Perspect Clin. Res. 7, 187. https://doi.org/10.4103/2229- and Climate Modeling. Centre de recherche sur la Terre et le climat Georges
3485.192046. Lemaître-UCLouvain. Cambridge University Press, Cambridge, UK.
Ahmadalipour, A., Rana, A., Moradkhani, H., Sharma, A., 2017. Multi-criteria evaluation Guo, H., Bao, A., Chen, T., Zheng, G., Wang, Y., Jiang, L., De Maeyer, P., 2021.
of CMIP5 GCMs for climate change impact analysis. Theor. Appl. Climatol. 128, Assessment of CMIP6 in simulating precipitation over Arid Central Asia. Atmos. Res.
71–87. https://doi.org/10.1007/s00704-015-1695-4. 252 (105451), 1–15. https://doi.org/10.1016/j.atmosres.2021.105451.
Akasaka, I., 2010. Interannual variations in seasonal march of rainfall in the Philippines. Gupta, S., 2010. Synthesis report on ten ASEAN countries disaster risks assessment. In:
Int. J. Climatol. 30, 1301–1314. https://doi.org/10.1002/joc.1975. ASEAN Disaster Risk Management Initiative. United Nations International Strategy
Alam, M., Lee, J., Sawhney, P., 2018. Status of Climate Change Adaptation in Asia and for Disaster Reduction (UNISDR)/Washington, DC, USA: World Bank, Geneva,
the Pacific. Springer, New York, NY, USA, pp. 27–40. Switzerland. https://doi.org/10.13140/RG.2.2.32014.36160.
Almazroui, M., Saeed, S., Saeed, F., Islam, M.N., Ismail, M., 2020. Projections of Gusain, A., Ghosh, S., Karmakar, S., 2020. Added value of CMIP6 over CMIP5 models in
precipitation and temperature over the South Asian countries in CMIP6. Earth Syst. simulating Indian summer monsoon rainfall. Atmos. Res. 232, 104680 https://doi.
Environ. 4, 297–320. https://doi.org/10.1007/s41748-020-00157-7. org/10.1016/j.atmosres.2019.104680.
Ayugi, B., Zhihong, J., Zhu, H., Ngoma, H., Babaousmail, H., Rizwan, K., Dike, V., 2021. Hajima, T., Watanabe, M., Yamamoto, A., Tatebe, H., Noguchi, M.A., Abe, M.,
Comparison of CMIP6 and CMIP5 models in simulating mean and extreme Ohgaito, R., Ito, A., Yamazaki, D., Okajima, H., Ito, A., Takata, K., Ogochi, K.,
precipitation over East Africa. Int. J. Climatol. 41, 6474–6496. https://doi.org/ Watanabe, S., Kawamiya, M., 2020. Development of the MIROC-ES2L earth system
10.1002/joc.7207. model and the evaluation of biogeochemical processes and feedbacks. Geosci. Model
Azam, M., Maeng, S.J., Kim, H.S., Lee, S.W., Lee, J.E., 2018. Spatial and temporal trend Dev. 13, 2197–2244. https://doi.org/10.5194/gmd-13-2197-2020.
analysis of precipitation and drought in South Korea. Water. 10, 765. https://doi. Han, Y., Zhang, M.Z., Xu, Z., Guo, W., 2022. Assessing the performance of 33 CMIP6
org/10.3390/w10060765. models in simulating the large-scale environmental fields of tropical cyclones. Clim.
Babaousmail, H., Hou, R., Ayugi, B., Ojara, M., Ngoma, H., Karim, R., Rajasekar, A., Dyn. 1-16 https://doi.org/10.1007/s00382-021-05986-4.
Ongoma, V., 2021. Evaluation of the performance of CMIP6 models in reproducing Hinkle, D.E., Wiersma, W., Jurs, S.G., 2003. Applied Statistics for the Behavioral
rainfall patterns over North Africa. Atmosphere. 12, 475. https://doi.org/10.3390/ Sciences, 5th ed. Houghton Mifflin, Boston, p. 756.
atmos12040475. Hussain, M., Yusof, K.W., Mustafa, M.R.U., Mahmood, R., Jia, S., 2018. Evaluation of
Baede, A.P.M., Ahlonsou, E., Ding, Y., Schimel, D., Bolin, B., Pollonais, S., 2001. Climate CMIP5 models for projection of future precipitation change in Bornean tropical
change 2001: the scientific basis, contribution from Working Group I to the Third rainforests. Theor. Appl. Climatol. 134, 423–440. https://doi.org/10.1007/s00704-
Assessment report of the intergovernmental panel on climate change. In: 017-2284-5.
Houghton, J.T., Ding, Y., Griggs, D.J., Noguer, M., van der Linden, P.J., Dai, K., IPCC, 2021. Climate Change 2021: The Physical Science Basis. In: Masson-Delmotte, V.,
Maskell, K., Johnson, C.A. (Eds.), The Climate System: An Overview. Cambridge Zhai, P., Pirani, A., Connors, S.L., Péan, C., Berger, S., Caud, N., Chen, Y.,
University Press, Cambridge, United Kingdom. Goldfarb, L., Gomis, M.I., Huang, M., Leitzell, K., Lonnoy, E., Matthews, J.B.R.,
Bagtasa, G., 2020. Influence of Madden–Julian oscillation on the intraseasonal variability Maycock, T.K., Waterfield, T., Yelekçi, O., Yu, R., Zhou, B. (Eds.), Contribution of
of summer and winter monsoon rainfall in the Philippines. J. Clim. 33, 9581–9594. Working Group I to the Sixth Assessment Report of the Intergovernmental Panel on
https://doi.org/10.1175/JCLI-D-20-0305.1. Climate Change. Cambridge University Press, Cambridge, United Kingdom and New
Bohra-Mishra, P., Oppenheimer, M., Cai, R., Feng, S., Licker, R., 2017. Climate variability York, NY, USA, p. 2391. https://doi.org/10.1017/9781009157896.
and migration in the Philippines. Popul. Environ. 38, 286–308. https://doi.org/ Iqbal, Z., Shahid, S., Ahmed, K., Ismail, T., Ziarh, G.F., Chung, E.S., Wang, X., 2021.
10.1007/s11111-016-0263-x. Evaluation of CMIP6 GCM rainfall in mainland Southeast Asia. Atmos. Res. 254,
Chen, Z., Zhou, T., Zhang, L., Chen, X., Zhang, W., Jiang, J., 2020. Global land monsoon 105525 https://doi.org/10.1016/j.atmosres.2021.105525.
precipitation changes in CMIP6 projections. Geophys. Res. Lett. 47, e2019GL086902 Jia, K., Ruan, Y., Yang, Y., Zhang, C., 2019. Assessing the performance of CMIP5 global
https://doi.org/10.1029/2019GL086902. climate models for simulating future precipitation change in the Tibetan Plateau.
Cui, T., Li, C., Tian, F., 2021. Evaluation of temperature and precipitation simulations in Water. 11, 1771. https://doi.org/10.3390/w11091771.
CMIP6 models over the Tibetan Plateau. Earth Space Sci. 8, e2020EA001620 https:// Jiang, D., Hu, D., Tian, Z., Lang, X., 2020. Differences between CMIP6 and CMIP5 models
doi.org/10.1029/2020EA001620. in simulating climate over China and the East Asian monsoon. Adv. Atmos. Sci. 37,
Dado, J.M.B., Takahashi, H.G., 2017. Potential impact of sea surface temperature on 1102–1118. https://doi.org/10.1007/s00376-020-2034-y.
rainfall over the western Philippines. Prog. Earth Planet. Sci. 4, 1–12. https://doi. Kamruzzaman, M., Shahid, S., Islam, A.R.M., Hwang, S., Cho, J., Zaman, M.A.U.,
org/10.1186/s40645-017-0137-6. Ahmed, M., Rahman, Md.M., Hossain, M.B., 2021. Comparison of CMIP6 and CMIP5
de Viana, A.V., 2014. The Philippines’ typhoon alley: the historic bagyos of the model performance in simulating historical precipitation and temperature in
Philippines and their impact. Jurnal Kajian Wilayah. 5, 184–216. http://jkw.psdr.lip Bangladesh: a preliminary study. Theor. Appl. Climatol. 145, 1385–1406. https://
i.go.id/index.php/jkw/article/view/262/136. doi.org/10.1007/s00704-021-03691-0.
Desmet, Q., Ngo-Duc, T., 2022. A novel method for ranking CMIP6 global climate models Kamworapan, S., Surussavadee, C., 2019. Evaluation of CMIP5 global climate models for
over the southeast Asian region. Int. J. Climatol. 42, 97–117. https://doi.org/ simulating climatological temperature and precipitation for Southeast Asia. Adv.
10.1002/joc.7234. Meteorol. 2019, 1–18. https://doi.org/10.1155/2019/1067365.
Dias, C.G., Reboita, M.S., 2021. Assessment of CMIP6 Simulations over Tropical South Karim, R., Tan, G., Ayugi, B., Babaousmail, H., Liu, F., 2020. Evaluation of historical
America. Braz. J.Phys. Geogr. 14, 1282–1295. https://doi.org/10.26848/rbgf.v14.3. CMIP6 model simulations of seasonal mean temperature over Pakistan during 1970-
Döscher, R., Acosta, M., Alessandri, A., Anthoni, P., Arneth, A., Arsouze, T., 2014. Atmosphere. 11, 1005. https://doi.org/10.3390/atmos11091005.
Bergmann, T., Bernadello, R., Bousetta, S., Caron, L.-P., Carver, G., Castrillo, M., Khan, N., Shahid, S., Ahmed, K., Ismail, T., Nawaz, N., Son, M., 2018. Performance
Catalano, F., Cvijanovic, I., Davini, P., Dekker, E., Doblas-Reyes, F.J., Docquier, D., Assessment of General Circulation Model in Simulating Daily Precipitation and
Echevarria, P., Fladrich, U., Fuentes-Franco, R., Gröger, M.V., Hardenberg, J., Temperature using Multiple Gridded Datasets. Water. 10, 1793. https://doi.org/
Hieronymus, J., Karami, M.P., Keskinen, J.P., Koenigk, T., Makkonen, R., 10.3390/w10121793.
Massonnet, F., Ménégoz, M., Miller, P.A., Moreno-Chamarro, E., Nieradzik, L., van Kubota, H., Shirooka, R., Matsumoto, J., Cayanan, E.O., Hilario, F.D., 2017. Tropical
Noije, T., Nolan, P., O’Donnell, D., Ollinaho, P., van den Oord, G., Ortega, P., cyclone influence on the long-term variability of Philippine summer monsoon onset.
Prims, O.T., Ramos, A., Reerink, T., Rousset, C., Ruprich-Robert, Y., Le Sager, P., Prog. Earth Planet. Sci. 4, 1–12. https://doi.org/10.1186/s40645-017-0138-5.
Schmith, T., Schrödner, R., Serva, F., Sicardi, V., Sloth Madsen, M., Smith, B., Lasco, R.D., Pulhin, F.B., Jaranilla-Sanchez, P.A., Delfino, R.J.P., Gerpacio, R., Garcia, K.,
Tian, T., Tourigny, E., Uotila, P., Vancoppenolle, M., Wang, S., Wårlind, D., 2009. Mainstreaming adaptation in developing countries: the case of the Philippines.
Willén, U., Wyser, K., Yang, S., Yepes-Arbós, X., Zhang, Q., 2022. The EC-Earth3 Clim. Dev. 1, 130–146. https://doi.org/10.3763/cdev.2009.0009.
Earth system model for the coupled model intercomparison project 6. Geosci. Model Li, Y., Yan, D., Peng, H., Xiao, S., 2021. Evaluation of precipitation in CMIP6 over the
Dev. 15, 2973–3020. https://doi.org/10.5194/gmd-15-2973-2022. Yangtze River Basin. Atmos. Res. 253, 105406 https://doi.org/10.1016/j.
Doutreloup, S., Wyard, C., Amory, C., Kittel, C., Erpicum, M., Fettweis, X., 2019. atmosres.2020.105406.
Sensitivity to convective schemes on precipitation simulated by the regional climate Mauritsen, T., Bader, J., Becker, T., Behrens, J., Bittner, M., Brokopf, R., Brovkin, V.,
model MAR over Belgium (1987–2017). Atmosphere. 10, 34. https://doi.org/ Claussen, M., Crueger, T., Esch, M., Fast, I., Fiedler, S., Flaschner, D., Gayler, V.,
10.3390/atmos10010034. Giorgetta, M., Goll, D., Haak, H., Hagemann, S., Hedernann, C., Hohenegger, C.,
ESGF the Earth System Grid Federation, 2021. https://esgf-node.llnl.gov/search/cmip6/ Ilyina, T., Jahns, T., Jimenez de la Cuesta, D., Jungclaus, J., Kleinen, T., Kloster, S.,
(accessed 15 December 2021). Kracher, D., Stefan, K., Kleberg, D., Lasslop, G., Kornblueh, L., Marotzke, J.,
Eyring, V., Bony, S., Meehl, G.A., Senior, C., Stevens, B., Stouffer, R.J., Taylor, K.E., 2016. Matei, D., Meraner, K., Mikolajewicz, U., Modali, K., Mobis, B., Muller, W., Nabel, J.,
Overview of the coupled model intercomparison project phase 6 (CMIP6) Nam, C., Notz, D., Nyawira, S., Paulsen, H., Peters, K., Pincus, R., Pohlmann, H.,
experimental design and organisation. Geosci. Model Dev. 9, 1937–1958. https:// Pongratz, J., Popp, M., Raddatz, T., Rast, S., Redler, R., Reick, C., Rohrscheneider, T.,
doi.org/10.5194/gmd-9-1937-2016. Schemann, V., Schimidt, H., Schnur, R., Schulzweida, U., Six, K., Stein, L.,
Eyring, V., Cox, P.M., Flato, G.M., Gleckler, P.J., Abramowitz, G., Caldwell, P., Stemmler, I., Stevens, B., von Storch, J., Tian, F., Voigt, A., Vrese, P., Wieners, K.,
Collins, W.D., Gier, B.K., Hall, A.D., Hoffman, F.M., Hurtt, G.C., 2019. Taking Wilkenskjeld, S., Winkler, A., Roeckner, E., 2019. Developments in the MPI-M earth
climate model evaluation to the next level. Nat. Clim. Chang. 9, 102–110. https:// system model version 1.2 (MPI-ESM1. 2) and its response to increasing CO2. J. Adv.
doi.org/10.1038/s41558-018-0355-y. Model. Earth Syst. 11, 998–1038. https://doi.org/10.1029/2018MS001400.
Faikrua, A., Pimonsree, S., Wang, L., Limsakul, A., Singhruck, P., Dong, Z., 2020. Decadal Meehl, G.A., Covey, C., McAvaney, B., Latif, M., Stouffer, R.J., 2005. Overview of the
increase of the summer precipitation in Thailand after the mid-1990s. Clim. Dyn. 55, coupled model intercomparison project. Bull. Am. Meteorol. Soc. 86, 89–96. https://
3253–3267. https://doi.org/10.1007/s00382-020-05443-8. doi.org/10.1175/BAMS-86-1-89.
Fakaruddin, F.J., Nawai, N.A., Abllah, M., Tangang, F., Juneng, L., 2022. Climatological Mohammed, J.A., Yimam, Z.A., 2022. Spatiotemporal variability and trend analysis of
Features of Squall Line at the Borneo Coastline during Southwest Monsoon. rainfall in Beshilo sub-basin, Upper Blue Nile (Abbay) Basin of Ethiopia. Arab. J.
Atmosphere 13, 116. https://doi.org/10.3390/atmos13010116. Geosci. 15, 1387. https://doi.org/10.1007/s12517-022-10666-6.

15
S. Pimonsree et al. Atmospheric Research 282 (2023) 106522

Moon, S., Ha, K.J., 2017. Temperature and precipitation in the context of the annual warming in a high-resolution global nonhydrostatic model. J. Clim. 33, 8147–8164.
cycle over Asia: model evaluation and future change. Asia-Pacific J. Atmos. Sci. 53, https://doi.org/10.1175/JCLI-D-19-0824.1.
229–242. https://doi.org/10.1007/s13143-017-0024-5. Tan, C.T., Pereira, J.J., 2010. Climate change adaptation: an overview of Southeast Asia.
Ngo-Duc, T., Tangang, F.T., Santisirisomboon, J., Cruz, F., Trinh-Tuan, L., Nguyen- Asian J. Environ. Disaster Manage. 2, 371–395. https://doi.org/10.3850/
Xuan, T., Phan-Van, T., Juneng, L., Narisma, G., Singhruck, P., Gunawan, D., S1793924011000812.
Aldrian, E., 2017. Performance evaluation of RegCM4 in simulating extreme rainfall Thoeun, H.C., 2015. Observed and projected changes in temperature and rainfall in
and temperature indices over the CORDEX-Southeast Asia region. Int. J. Climatol. Cambodia. Weather Clim. Extrem. 7, 61–71. https://doi.org/10.1016/j.
37, 1634–1647. https://doi.org/10.1002/joc.4803. wace.2015.02.001.
Olaguera, L.M., Matsumoto, J., Kubota, H., Inoue, T., Cayanan, E.O., Hilario, F.D., 2018. Tomkratoke, S., Sirisup, S., 2022. Influence and variability of monsoon trough and front
Abrupt climate shift in the mature rainy season of the Philippines in the mid-1990s. on rainfall in Thailand. Int. J. Climatol. 42, 619–634. https://doi.org/10.1002/
Atmosphere. 9, 350. https://doi.org/10.3390/atmos9090350. joc.7263.
Phuong, D.N.D., Linh, V.T., Nhat, T.T., Dung, H.M., Loi, N.K., 2019. Spatiotemporal Torsri, K., Lin, Z., Dike, V.N., Thodsan, T., Wongsaming, P., 2022. Evaluation of spatial-
variability of annual and seasonal rainfall time series in Ho Chi Minh city, Vietnam. temporal characteristics of rainfall variations over Thailand inferred from different
J. Water Clim. Change. 10, 658–670. https://doi.org/10.2166/wcc.2018.115. gridded datasets. Water. 14, 1359. https://doi.org/10.3390/W14091359.
Piman, T., Pawattana, C., Vansarochana, A., Aekakkararungroj, A., Hormwichian, R., Torti, J., 2012. Floods in Southeast Asia: a health priority. J. Glob. Health 2, 020304.
2016. Analysis of historical changes in rainfall in Huai Luang watershed, Thailand. https://doi.org/10.7939/R3VM42Z8K.
Int. J. Technol. 7, 1155–1162. https://doi.org/10.14716/ijtech.v7i7.4709. Trinh-Tuan, L., Konduru, R.T., Inoue, T., Ngo-Duc, T., Matsumoto, J., 2019. Autumn
Randall, D.A., Wood, R.A., Bony, S., Colman, R., Fichefet, T., Fyfe, J., Kattsov, V., Rainfall Increasing Trend in South Central Vietnam and its Association with Changes
Pitman, A., Shukla, J., Srinivasan, J., Stouffer, R.J., Sumi, A., Taylor, K.E., 2007. in Vietnam’s East Sea Surface Temperature, 54. Geographical Reports of Tokyo
Cilmate models and their evaluation. In: Solomon, S., Qin, D., Manning, M., Chen, Z., Metropolitan University, pp. 11–22.
Marquis, M., Averyt, K.B., Tignor, M., Miller, H.L. (Eds.), Climate Change 2007: The Wang, L., Zhang, J., Shu, Z., Wang, Y., Bao, Z., Liu, C., Zhou, X., Wang, G., 2021.
Physical Science Basis; Contribution of Working Group I to the Fourth Assessment Evaluation of the ability of CMIP6 Global climate models to simulate precipitation in
Report of the Intergovernmental Panel on Climate Change. Cambridge University the Yellow River Basin, China. Front. Earth Sci. 9 (751974), 1–12. https://doi.org/
Press, Cambridge, UK; New York, NY, USA. 10.3389/feart.2021.751974.
Rivera, J.A., Arnould, G., 2020. Evaluation of the ability of CMIP6 models to simulate World Bank and GFDRR, 2011. Vulnerability, Risk Reduction, and Adaptation to Climate
precipitation over Southwestern South America: climatic features and long-term Change – Senegal. World Bank and GFDRR, Washington, DC.
trends (1901–2014). Atmos. Res. 241, 104953 https://doi.org/10.1016/j. Xin, X., Wu, T., Zhang, J., Yao, J., Fang, Y., 2020. Comparison of CMIP6 and CMIP5
atmosres.2020.104953. simulations of precipitation in China and the East Asian summer monsoon. Int. J.
Ruan, Y., Yao, Z., Wang, R., Liu, Z., 2018. Ranking of CMIP5 GCM skills in simulating Climatol. 40, 6423–6440. https://doi.org/10.1002/joc.6590.
observed precipitation over the lower Mekong Basin, using an improved score-based Xuan, W., Ma, C., Kang, L., Gu, H., Pan, S., Xu, Y.-P., 2017. Evaluating historical
method. Water. 10, 1868. https://doi.org/10.3390/w10121868. simulations of CMIP5 GCMs for key climatic variables in Zhejiang Province, China.
Rupp, D.E., Abatzoglou, J.T., Hegewisch, K.C., Mote, P.W., 2013. Evaluation of CMIP5 Theor. Appl. Climatol. 128, 207–222. https://doi.org/10.1007/s00704-015-1704-7.
20th century climate simulations for the Pacific Northwest USA. J. Geophys. Res. Yan, Y., Zhu, C., Liu, B., Jiang, S., 2021. Annual cycle of East Asian precipitation
Atmos. 118, 10884–10906. https://doi.org/10.1002/jgrd.50843. simulated by CMIP6 models. Atmosphere. 12, 24. https://doi.org/10.3390/
Santhanasamy, C., Arcos González, P., Castro Delgado, R., 2021. (Un) natural disasters in atmos12010024.
Southeast Asia. Emerg. Disast. Rep. 8, 1–52. https://doi.org/10.13140/ Yazdandoost, F., Moradian, S., Izadi, A., Aghakouchak, A., 2021. Evaluation of CMIP6
RG.2.2.26315.92963. precipitation simulations across different climatic zones: uncertainty and model
Schneider, U., Becker, A., Finger, P., Meyer-Christoffer, A., Ziese, M., 2018. GPCC full intercomparison. Atmos. Res. 250, 105369 https://doi.org/10.1016/j.
data monthly product version 2018 at 0.5◦ : monthly land-surface precipitation from atmosres.2020.105369.
rain-gauges built on GTS-based and historical data. In: Global Precipitation Yuen, B., Kong, L., 2009. Climate change and urban planning in Southeast Asia. In: S.A.P.
Climatology Centre. https://doi.org/10.5676/DWD_GPCC/FD_M_V2018_050. I.E.N.S. 2.3. http://journals.openedition.org/sapiens/881.
Stouffer, R.J., Eyring, V., Meehl, G.A., Bony, S., Senior, C., Stevens, B., Taylor, K.E., 2017. Zhang, X., Hua, L., Jiang, D., 2022. Assessment of CMIP6 model performance for
CMIP5 Scientific Gaps and Recommendations for CMIP6. Bull. Am. Meteorol. Soc. temperature and precipitation in Xinjiang, China. Atmos. Ocean. Sci. Lett. 15,
98, 95–105. https://doi.org/10.1175/BAMS-D-15-00013.1. 100128 https://doi.org/10.1016/j.aosl.2021.100128.
Strobl, E., 2019. The impact of typhoons on economic activity in the Philippines: Zhou, W.C., Han, Z.Y., 2018. Assessing CMIP5 climate simulations and objective
evidence from nightlight intensity. In: ADB Economics Working Paper Series. selection of models over the Yellow River Basin. J. Meteorol. Environ. 34, 42–55.
https://doi.org/10.22617/WPS190278-2. https://doi.org/10.3969/j.issn.1673-503X.2018.06.005.
Su, F., Duan, X., Chen, D., Hao, Z., Cuo, L., 2013. Evaluation of the global climate models Zhu, H., Jiang, Z., Li, J., Li, W., Sun, C., Li, L., 2020. Does CMIP6 inspire more confidence
in the CMIP5 over the Tibetan Plateau. J. Clim. 26, 3187–3208. https://doi.org/ in simulating climate extremes over China? Adv. Atmos. Sci. 37, 1119–1132.
10.1175/JCLI-D-12-00321.1. https://doi.org/10.1007/s00376-020-9289-1.
Takahashi, H.G., Kamizawa, N., Nasuno, T., Yamada, Y., Kodama, C., Sugimoto, S.,
Satoh, M., 2020. Response of the Asian summer monsoon precipitation to global

16

You might also like