Abstract
This study generates a daily temperature and precipitation dataset over Vietnam at a high resolution of 0.1° for the historical period 1980–2005 and the future period 2006–2100 under four representative concentration pathway (RCP) scenarios, namely RCP2.6, RCP4.5, RCP6.0, and RCP8.5. The bias correction (BC) and spatial disaggregation (SD) method is applied to the outputs of 31 global climate models (GCMs) of the Coupled Model Intercomparison Project Phase 5 (CMIP5) to create the new dataset called CMIP5-VN. To guide the BC and SD steps, gridded temperature and precipitation data interpolated from daily observations of 147 and 481 stations respectively are used. Results with the CMIP5-VN show that warming over Vietnam is projected to continue till the end of the 21st century under all four RCPs. The average temperature is projected to increase by 1.3±0.52 °C under RCP2.6 and by 3.85±0.85 °C under RCP8.5 between 2080–2099 and 1986–2005. The future increase is more intense in the northern regions than in the south and higher in summer than in winter. Precipitation is projected to increase by 1.16±7.1% under RCP2.6 and by 4.41±9.2% under RCP8.5. In Central Vietnam, there is a consistent rainfall increase in the future rainy season.
HIGHLIGHTS
This is the first time a complete set of high-resolution temperature and precipitation data for different scenarios has been created for Vietnam.
Besides RCP4.5 and RCP8.5, this is the first time the detailed changes under the RCP2.6 and RCP6.0 scenarios in the sub-regions of Vietnam are assessed.
The warming is more intense in the northern regions than in the south and higher in summer than winter for all RCPs.
Projected temperature changes relative to the baseline period 1986–2005 based on the newly-created CMIP5-VN data for Vietnam. Five-year moving averages are applied. Colored lines show the ensemble means of the models, and colored shaded areas present the uncertainty ranges (±1 standard deviation) for each representative concentration pathway (RCP) scenario. The number of models used for each RCP is shown in brackets. Box plots display the occurrence statistics (the ends of the box are the upper and lower quartiles, the horizontal line inside the box marks the median, and the two horizontal lines outside of the box indicate the 10th and 90th percentile values, respectively) for warming levels at the end of the 21st century.
Graphical Abstract
INTRODUCTION
The Sixth Assessment Report (AR6) of the Intergovernmental Panel on Climate Change (IPCC) indicates that the global average surface temperature during 2010–2019 has increased by ∼1.09 °C compared to 1850–1900 (Arias et al. 2021). The warming is projected to be virtually certain in the future at global and regional scales (Arias et al. 2021). To assess the impacts of global warming and prepare for response measures, future climate information at different spatial scales, including local and regional ones, is needed (Milly et al. 2008; Giorgi & Bi 2009).
The Coupled Model Intercomparison Project (CMIP) is an international experimental protocol coordinated by the World Climate Research Programme (WCRP) for producing and studying the outputs of global climate models (GCMs). Providing the scientific ground for the IPCC Fifth Assessment Report (AR5) (IPCC 2013), the CMIP Phase 5 (CMIP5) provides simulations from GCMs that reflect a joint effort involving many climate research organizations worldwide (Taylor et al. 2012). In the CMIP5 project, historical and future climate simulations under different representative concentration pathway (RCP) scenarios were conducted. The RCP scenarios were developed based on the future estimated radiative forcing, covering a period up to the year 2100 or later (van Vuuren et al. 2011). There are four RCPs consisting of RCP2.6 (a scenario characterized by a low radiative forcing level), RCP4.5 and RCP6.0 (medium stabilization scenarios), and RCP8.5 (a high radiative forcing scenario). The RCPs represent different pathways upon the projected impacts of land use and emission of greenhouse gases (GHGs). Accordingly, RCP2.6 expresses a low GHG concentration scenario in the future with radiative forcing estimated at 2.6 W/m2 by 2100, and so forth with the other RCPs (Moss et al. 2010).
Though CMIP5 outputs can generally reproduce the major climate indicators at global scales, it is still challenging to directly use them to produce relevant climate information at local to regional scales due to their coarse spatial resolution (typically coarser than 100 km) (Taylor et al. 2012; Eyring et al. 2016). This raises concerns about adequately using such products as the driving force in other comprehensive environment-socio-economic assessment models to obtain reliable results to support researchers and policymakers in regional planning. Thus, downscaling methods, which transfer data of GCM from a coarse grid resolution into a much higher spatial resolution, should be applied for limited-area domains.
There are two popular downscaling approaches: dynamical and statistical. For dynamical downscaling, higher-spatial resolution models, which are often known as regional climate models (RCMs), are used. Initial and lateral boundary conditions of an RCM are commonly provided by large-scale GCM products, consisting of wind, temperature, and moisture fields. Although dynamical downscaling has the advantage of well representing the local-scale feedback and dynamical processes (Seaby et al. 2013; Giorgi & Gutowski 2015; Tangang et al. 2020), it is an extremely computationally demanding method; thus, its application for downscaling multiple GCMs and scenarios is limited. For statistical downscaling, it is conducted based on empirical, spatial, and temporal relationships between large-scale and local-scale climate variables (Murphy 1999; Fowler Blenkinsop & Tebaldi 2007). These relationships are assumed to be unchanged with time; thus, they can be used to project future conditions. The primary advantage of statistical downscaling is that it is computationally inexpensive and much faster than dynamical downscaling so that it can be easily applied to generate high-resolution simulations with multiple GCMs. Additionally, statistical downscaling can provide climate information at any specific resolution, so its results can be directly used in climate change impact assessments.
Previous studies showed the skillful performance of different statistical methods (Salathé 2003; Widmann et al. 2003; Maurer & Hidalgo 2008; Noël et al. 2021). Among those methods, the bias correction and spatial disaggregation (BCSD) approach is reliable and effective for downscaling temperature and precipitation data (Wood 2002; Wood et al. 2004). The BCSD has been extensively adopted in various studies in many parts of the world, e.g., in the United States (Rasmussen et al. 2016), South Korea (Eum et al. 2017), China (Xu & Wang 2019), and at the global scale (Zhang et al. 2019).
Located in the tropical region in Southeast Asia, with a long coastline of about 3260 km and an estimated total area of 329,560 km2, Vietnam is one of the countries heavily affected by climate change (MONRE 2009). Therefore, to prepare for short- and long-term action plans to cope with climate change, it is important for Vietnam to establish reliable climate change information. As a result, Vietnam's sea-level rise and climate change (SLRCC) scenarios were released in 2009 and 2012 (MONRE 2009, 2012) and followed up with the updated versions in 2016 and 2021 (MONRE 2016, 2021). In the latest SLRCC scenario, outputs of 16 dynamical downscaling experiments based on five different RCMs and 10 driving GCMs for two scenarios, RCP4.5 and 8.5, were analyzed. It should be noted that the accuracy of an RCM simulation depends on the quality of its driving GCM (Déqué et al. 2012; Diaconescu & Laprise 2013; Tamara et al. 2019); thus, using several GCMs for limited downscaling cases may result in inconsistent future projections. Under the framework of the Coordinated Regional Climate Downscaling Experiment-Southeast Asia (CORDEX-SEA), different dynamical downscaling simulations using seven different RCMs were conducted (Trinh-Tuan et al. 2019; Tangang et al. 2020; Nguyen-Thuy et al. 2021), providing additional projected climate information for Vietnam. However, the number of driving GCMs (11) and scenarios (two) used in CORDEX-SEA is still limited, leading to the possibility that their climate projections might fall outside what would happen in the future. Therefore, our present study is a primary attempt to develop a new comprehensive and high-resolution dataset over Vietnam from a large number of CMIP5 GCMs. The newly-built data, hereafter called CMIP5-VN, consisting of daily mean, maximum, and minimum temperature, and precipitation, are created using the BCSD downscaling method based on a total of 31 CMIP5 GCMs and four RCPs. The new CMIP5-VN dataset is expected to supplement the existing climate change data in Vietnam that were previously obtained by the dynamical downscaling method. In the Results and Discussion section, we also present an application of the CMIP5-VN in assessing future changes in temperature and rainfall over the different climatic sub-regions of Vietnam.
DATA AND METHODS
Study domain
Seven climatic sub-regions of Vietnam. Data from 157 temperature stations and 481 rainfall stations, indicated by purple and red circles, respectively, were used in this study. Topography over Vietnam (shaded, in m) is obtained from the hydrological data and maps based on Shuttle Elevation Derivatives at multiple Scales (HydroSHEDS) (Lehner et al. 2008). Please refer to the online version of this paper to see this figure in color: http://dx.doi.org/10.2166/wcc.2022.144.
Seven climatic sub-regions of Vietnam. Data from 157 temperature stations and 481 rainfall stations, indicated by purple and red circles, respectively, were used in this study. Topography over Vietnam (shaded, in m) is obtained from the hydrological data and maps based on Shuttle Elevation Derivatives at multiple Scales (HydroSHEDS) (Lehner et al. 2008). Please refer to the online version of this paper to see this figure in color: http://dx.doi.org/10.2166/wcc.2022.144.
Daily observation data and associated processing method
In this study, near-surface daily average (T2 m), daily maximum (Tmax), and minimum (Tmin) temperatures and daily rainfall data from 1980 to 2005 are acquired from the Vietnam Meteorological and Hydrological Administration (VMHA). There are 147 and 481 stations for temperature and rainfall, respectively (Figure 1). The station data are interpolated to a 0.1°×0.1° gridded dataset by using the Kriging interpolation technique (Switzer 2014) for temperature and the Sphere map interpolation technique (Willmott et al. 1985) for rainfall. Nguyen-Xuan et al. (2016) showed that the Spheremap interpolation technique has advantages in creating a gridded rainfall dataset over Vietnam in comparison with several other interpolation methods such as Cressman (1959), Inverse Distance Weighted (Shepard 1968), or Kriging (Switzer 2014). Besides, the Kriging technique is an effective interpolation method for continuous spatial variables such as temperature (Wu & Li 2013). The newly-created gridded dataset (hereinafter called OBS) is used in this study to statistically downscale GCM data for Vietnam.
Model data
Daily rainfall (R) and temperatures (T2 m, Tmax, Tmin) from 31 CMIP5 GCMs (Table 1) are obtained via the Earth System Grid Federation portal (ESGF, https://esgf-node.llnl.gov/projects/cmip5/). The present-day simulations for the period 1980–2005 are used as the basis to construct statistical relationships between the high-resolution OBS dataset and the coarse resolution GCMs. Those relationships are further used to statistically downscale the projected GCM variables for the period 2006–2100 under the RCPs 2.6, 4.5, 6.0, and 8.5.
List of 31 CMIP5 GCMs, their resolutions, and RCP availability that are used in this study
No. . | Climate Modeling Group . | CMIP5 Model ID . | Grid resolution (degree) . | Availability . | ||||
---|---|---|---|---|---|---|---|---|
Lon. . | Lat. . | RCP 2.6 . | RCP 4.5 . | RCP 6.0 . | RCP 8.5 . | |||
1 | Commonwealth Scientific and Industrial Research Organization and Bureau of Meteorology, Australia | ACCESS1-0 | 1.25 | 1.875 | * | * | ||
2 | ACCESS1-3 | 1.25 | 1.875 | * | * | |||
3 | Beijing Climate Center, China Meteorological Administration | BCC-CSM1-1 | 2.791 | 2.813 | * | * | * | * |
4 | BCC-CSM1-1-M | 2.791 | 2.813 | * | * | * | ||
5 | College of Global Change and Earth System Science, Beijing Normal University | BNU-ESM | 2.791 | 2.813 | * | * | * | |
6 | Canadian Centre for Climate Modelling and Analysis | CanESM2 | 2.791 | 2.813 | * | * | * | |
7 | National Center for Atmospheric Research | CCSM4 | 0.942 | 1.25 | * | * | * | |
8 | Community Earth System Model Contributors | CESM1-BGC | 0.942 | 1.25 | * | * | ||
9 | CESM1-CAM5 | 0.942 | 1.25 | * | * | * | * | |
10 | Centro Euro-Mediterraneo per I Cambiamenti Climatici | CMCC-CM | 0.748 | 0.75 | * | * | ||
11 | Centre National de Recherches Météorologiques/ Centre Européen de Recherche et Formation Avancée en Calcul Scientifique | CNRM-CM5 | 1.401 | 1.406 | * | * | * | |
12 | Commonwealth Scientific and Industrial Research Organization, Queensland Climate Change Centre of Excellence | CSIRO-Mk3-6-0 | 1.865 | 1.875 | * | * | * | * |
13 | NOAA Geophysical Fluid Dynamics Laboratory | GFDL-CM3 | 2 | 2.5 | * | * | * | |
14 | GFDL-ESM2G | 2.023 | 2 | * | * | * | * | |
15 | NASA Goddard Institute for Space Studies | GISS-E2-H | 2 | 2.5 | * | * | * | * |
16 | GISS-E2-H-CC | 2 | 2.5 | * | * | |||
17 | GISS-E2-R | 2 | 2.5 | * | * | * | * | |
18 | GISS-E2-R-CC | 2 | 2.5 | * | * | |||
19 | Met Office Hadley Centre (additional HadGEM2-ES realizations contributed by Instituto Nacional de Pesquisas Espaciais) | HadGEM2-CC | 1.25 | 1.875 | * | * | ||
20 | HadGEM2-ES | 1.25 | 1.875 | * | * | * | ||
21 | Institut Pierre-Simon Laplace | IPSL-CM5A-LR | 1.897 | 3.75 | * | * | * | |
22 | IPSL-CM5A-MR | 1.268 | 2.5 | * | * | * | * | |
23 | IPSL-CM5B-LR | 1.895 | 3.75 | * | * | |||
24 | Japan Agency for Marine-Earth Science and Technology, Atmosphere and Ocean Research Institute (The University of Tokyo), and National Institute for Environmental Studies | MIROC-ESM | 2.791 | 2.813 | * | * | * | * |
25 | MIROC-ESM-CHEM | 2.791 | 2.813 | * | * | * | ||
26 | Atmosphere and Ocean Research Institute (The University of Tokyo), National Institute for Environmental Studies, and Japan Agency for Marine-Earth Science and Technology | MIROC5 | 1.401 | 1.406 | * | * | * | * |
27 | Max-Planck-Institut für Meteorologie (Max-Planck Institute for Meteorology) | MPI-ESM-LR | 1.865 | 1.875 | * | * | * | |
28 | MPI-ESM-MR | 1.865 | 1.875 | * | * | * | ||
29 | Meteorological Research Institute | MRI-CGCM3 | 1.121 | 1.125 | * | * | * | |
30 | Norwegian Climate Centre | NorESM1-M | 1.895 | 2.5 | * | * | * | * |
31 | NorESM1-ME | 1.895 | 2.5 | * | * | |||
Total number of available models for each RCP | 20 | 31 | 12 | 31 |
No. . | Climate Modeling Group . | CMIP5 Model ID . | Grid resolution (degree) . | Availability . | ||||
---|---|---|---|---|---|---|---|---|
Lon. . | Lat. . | RCP 2.6 . | RCP 4.5 . | RCP 6.0 . | RCP 8.5 . | |||
1 | Commonwealth Scientific and Industrial Research Organization and Bureau of Meteorology, Australia | ACCESS1-0 | 1.25 | 1.875 | * | * | ||
2 | ACCESS1-3 | 1.25 | 1.875 | * | * | |||
3 | Beijing Climate Center, China Meteorological Administration | BCC-CSM1-1 | 2.791 | 2.813 | * | * | * | * |
4 | BCC-CSM1-1-M | 2.791 | 2.813 | * | * | * | ||
5 | College of Global Change and Earth System Science, Beijing Normal University | BNU-ESM | 2.791 | 2.813 | * | * | * | |
6 | Canadian Centre for Climate Modelling and Analysis | CanESM2 | 2.791 | 2.813 | * | * | * | |
7 | National Center for Atmospheric Research | CCSM4 | 0.942 | 1.25 | * | * | * | |
8 | Community Earth System Model Contributors | CESM1-BGC | 0.942 | 1.25 | * | * | ||
9 | CESM1-CAM5 | 0.942 | 1.25 | * | * | * | * | |
10 | Centro Euro-Mediterraneo per I Cambiamenti Climatici | CMCC-CM | 0.748 | 0.75 | * | * | ||
11 | Centre National de Recherches Météorologiques/ Centre Européen de Recherche et Formation Avancée en Calcul Scientifique | CNRM-CM5 | 1.401 | 1.406 | * | * | * | |
12 | Commonwealth Scientific and Industrial Research Organization, Queensland Climate Change Centre of Excellence | CSIRO-Mk3-6-0 | 1.865 | 1.875 | * | * | * | * |
13 | NOAA Geophysical Fluid Dynamics Laboratory | GFDL-CM3 | 2 | 2.5 | * | * | * | |
14 | GFDL-ESM2G | 2.023 | 2 | * | * | * | * | |
15 | NASA Goddard Institute for Space Studies | GISS-E2-H | 2 | 2.5 | * | * | * | * |
16 | GISS-E2-H-CC | 2 | 2.5 | * | * | |||
17 | GISS-E2-R | 2 | 2.5 | * | * | * | * | |
18 | GISS-E2-R-CC | 2 | 2.5 | * | * | |||
19 | Met Office Hadley Centre (additional HadGEM2-ES realizations contributed by Instituto Nacional de Pesquisas Espaciais) | HadGEM2-CC | 1.25 | 1.875 | * | * | ||
20 | HadGEM2-ES | 1.25 | 1.875 | * | * | * | ||
21 | Institut Pierre-Simon Laplace | IPSL-CM5A-LR | 1.897 | 3.75 | * | * | * | |
22 | IPSL-CM5A-MR | 1.268 | 2.5 | * | * | * | * | |
23 | IPSL-CM5B-LR | 1.895 | 3.75 | * | * | |||
24 | Japan Agency for Marine-Earth Science and Technology, Atmosphere and Ocean Research Institute (The University of Tokyo), and National Institute for Environmental Studies | MIROC-ESM | 2.791 | 2.813 | * | * | * | * |
25 | MIROC-ESM-CHEM | 2.791 | 2.813 | * | * | * | ||
26 | Atmosphere and Ocean Research Institute (The University of Tokyo), National Institute for Environmental Studies, and Japan Agency for Marine-Earth Science and Technology | MIROC5 | 1.401 | 1.406 | * | * | * | * |
27 | Max-Planck-Institut für Meteorologie (Max-Planck Institute for Meteorology) | MPI-ESM-LR | 1.865 | 1.875 | * | * | * | |
28 | MPI-ESM-MR | 1.865 | 1.875 | * | * | * | ||
29 | Meteorological Research Institute | MRI-CGCM3 | 1.121 | 1.125 | * | * | * | |
30 | Norwegian Climate Centre | NorESM1-M | 1.895 | 2.5 | * | * | * | * |
31 | NorESM1-ME | 1.895 | 2.5 | * | * | |||
Total number of available models for each RCP | 20 | 31 | 12 | 31 |
* indicates yes.
The BCSD approach
In the second stage, the SD is adopted to spatially translate the BC GCM data from the intermediate resolution of 1°×1° to the targeted high-resolution of 0.1°×0.1°, by implementing the following steps:
Change-factors between the GCM BC data and observations are estimated at the intermediate resolution of 1°×1° for each variable, each model, each month, and each grid cell.
Then, the additive and multiplicative change-factors at the intermediate resolution for the reference period are bilinearly interpolated to the targeted high-resolution;
the high-resolution change-factors are added (for Tmax, Tmin, and T2 m) or multiplied (for R) to the high-resolution observation climatology to construct the high-resolution downscaled BC monthly climatological fields;
finally, the BC GCM daily values of a future month are obtained by randomly taking the same month of the observations in the reference period and additively (for temperature) and multiplicatively (for precipitation) adjusting its daily values to reproduce the monthly BC data.
For training and testing the BCSD procedure, we use the split ratio of 60:40 for the reference period 1980–2005, i.e., the first 16 years 1980–1995 (∼60% data), called the training period, are used for building the BC transfer functions, and the later 10 years 1996–2005 (∼40% data) are used for testing the results. The BCSD-downscaled data obtained for these training and testing periods are called BCSD-CMIP5. Finally, to maximize the construction period of the BCSD approach, following what has been done in previous studies (Reiter et al. 2016; Trinh-Tuan et al. 2019), the total 26 years period from 1980 to 2005 is used to bias correct and guide the SD for all GCM experiments listed in Table 1 for the future period 2006–2100, to obtain the final downscaled dataset for Vietnam, called CMIP5-VN.
The daily CMIP5-VN dataset, which is about 1.08 Terabytes in size, can be downloaded free of charge at the following address: http://remosat.usth.edu.vn/∼thanhnd/Download/CMIP5-VN/.
RESULTS AND DISCUSSION
Performance of the BCSD-CMIP5
Spatial distribution of annual average temperature in Vietnam by OBS and BCSD-ENS. (a, b) The average 1996–2005 temperature by OBS and BCSD-ENS; (c–g) and (h–l) indicate the biases of BSCD-ENS compared to OBS for the testing period 1996–2005 and the training period 1980–1995, respectively. The hatching lines show the regions where over two-thirds of CMIP5 models have the same bias sign with BCSD-ENS.
Spatial distribution of annual average temperature in Vietnam by OBS and BCSD-ENS. (a, b) The average 1996–2005 temperature by OBS and BCSD-ENS; (c–g) and (h–l) indicate the biases of BSCD-ENS compared to OBS for the testing period 1996–2005 and the training period 1980–1995, respectively. The hatching lines show the regions where over two-thirds of CMIP5 models have the same bias sign with BCSD-ENS.
Seasonal temperature biases are generally larger than the annual average bias (Figure 3(c)–(l)). For each season, the temperature bias range is higher in the northern regions (−0.36 to +0.56 °C) than in the southern regions (−0.2 to +0.3 °C), suggesting that the BC accuracy might be sensitive to the magnitude of annual and seasonal ranges of temperature. As an example, the biases of the BCSD-ENS in June–July–August (JJA, Figure 3(f)) and September–October–November (SON, Figure 3(g)) of the testing period 1996–2005 reached more than 0.2 °C in the northern region which is larger than the annual average bias (Figure 3(a)), ranging from −0.1 to 0.1 °C. It is worth noting that over the regions with large BCSD-ENS biases, the BC-GCMs generally show a strong agreement in their bias tendency, i.e., at least two-thirds of the model members have the same bias sign as the BCSD-ENS.
The BCSD-ENS biases in maximum and minimum daily temperature (Supplementary material, Figures A1, A2) are relatively higher than those in average daily temperature for both annual and seasonal averages. This can be due to the fact that the input GCM data of the BCSD processes experience larger biases in representing Tmax and Tmin compared to T2 m. The average (minimum, maximum) T2 m bias of the ensemble GCMs over Vietnam is 1.26 °C (−3.65 °C, 7.61 °C) which is smaller than the bias values of 1.56 (−3.95, 7.64) and 1.99 (−6.50, 9.32) of Tmax and Tmin, respectively (figure not shown). Note that previous studies also indicated large biases of GCMs in addressing daily extreme variables such as Tmax and Tmin (Rana et al. 2014; Tran Anh & Taniguchi 2018; Panjwani et al. 2020). The larger biases in the GCM inputs of the BCSD processes potentially exacerbate the output biases, thus inferring the larger BCSD-ENS biases in Tmax and Tmin.
Spatial distribution of annual average precipitation in Vietnam by OBS and BCSD-ENS. (a, b) The average 1996–2005 precipitation by OBS and BCSD-ENS; (c–g) and (h–l) indicate the biases of BSCD-ENS compared to OBS for the testing period 1996–2005 and the training period 1980–1995, respectively. The hatching lines show the regions where over two-thirds of CMIP5 models have the same bias sign with BCSD-ENS.
Spatial distribution of annual average precipitation in Vietnam by OBS and BCSD-ENS. (a, b) The average 1996–2005 precipitation by OBS and BCSD-ENS; (c–g) and (h–l) indicate the biases of BSCD-ENS compared to OBS for the testing period 1996–2005 and the training period 1980–1995, respectively. The hatching lines show the regions where over two-thirds of CMIP5 models have the same bias sign with BCSD-ENS.
Temporal correlations of the 1996–2005 seasonal cycles between the BCSD-ENS (left) and BIP-ENS (right) with OBS for (a, b) temperature and (c, d) precipitation.
Temporal correlations of the 1996–2005 seasonal cycles between the BCSD-ENS (left) and BIP-ENS (right) with OBS for (a, b) temperature and (c, d) precipitation.
Comparison of seasonal temperature cycles by OBS, BCSD, and BIP for the seven climatic sub-regions for the period 1996–2005. The seasonal cycle is shown for OBS (black), for 31 downscaled individual CMIP5 model members by BCSD (dim red) and BIP (dim blue), and two models ensemble mean, the BCSD-ENS (dark red) and the BIP-ENS (dark blue). Mean square errors (MSE) of the BCSD-ESN and BIP-ENS are also indicated. Please refer to the online version of this paper to see this figure in color: http://dx.doi.org/10.2166/wcc.2022.144.
Comparison of seasonal temperature cycles by OBS, BCSD, and BIP for the seven climatic sub-regions for the period 1996–2005. The seasonal cycle is shown for OBS (black), for 31 downscaled individual CMIP5 model members by BCSD (dim red) and BIP (dim blue), and two models ensemble mean, the BCSD-ENS (dark red) and the BIP-ENS (dark blue). Mean square errors (MSE) of the BCSD-ESN and BIP-ENS are also indicated. Please refer to the online version of this paper to see this figure in color: http://dx.doi.org/10.2166/wcc.2022.144.
Comparison of seasonal precipitation cycles by OBS, BCSD, and BIP for the seven climatic sub-regions for the period 1996–2005. The seasonal cycle is shown for OBS (black), for 31 downscaled individual CMIP5 model members by BCSD (dim red) and BIP (dim blue), and two models ensemble mean, the BCSD-ENS (dark red) and the BIP-ENS (dark blue). Mean square errors (MSE) of the BCSD-ESN and BIP-ENS are also indicated. Please refer to the online version of this paper to see this figure in color: http://dx.doi.org/10.2166/wcc.2022.144.
Comparison of seasonal precipitation cycles by OBS, BCSD, and BIP for the seven climatic sub-regions for the period 1996–2005. The seasonal cycle is shown for OBS (black), for 31 downscaled individual CMIP5 model members by BCSD (dim red) and BIP (dim blue), and two models ensemble mean, the BCSD-ENS (dark red) and the BIP-ENS (dark blue). Mean square errors (MSE) of the BCSD-ESN and BIP-ENS are also indicated. Please refer to the online version of this paper to see this figure in color: http://dx.doi.org/10.2166/wcc.2022.144.
AVs of individual BCSD-CMIP5 and BCSD-ENS for temperature over the BIP-CMIP5 for the 1996–2005 period. The darker saturation of warm/cold color denotes the more skillful/unskillful of the BCSD. The number in the top right corner indicates the percentage of the grids having positive AVs.
AVs of individual BCSD-CMIP5 and BCSD-ENS for temperature over the BIP-CMIP5 for the 1996–2005 period. The darker saturation of warm/cold color denotes the more skillful/unskillful of the BCSD. The number in the top right corner indicates the percentage of the grids having positive AVs.
AVs of individual BCSD-CMIP5 and BCSD-ENS for precipitation over the BIP-CMIP5 for the 1996–2005 period. The darker saturation of warm/cold color denotes the more skillful/unskillful of the BCSD. The number in the top right corner indicates the percentage of the grids having positive AVs.
AVs of individual BCSD-CMIP5 and BCSD-ENS for precipitation over the BIP-CMIP5 for the 1996–2005 period. The darker saturation of warm/cold color denotes the more skillful/unskillful of the BCSD. The number in the top right corner indicates the percentage of the grids having positive AVs.
The positive AV grids dominate most of the Vietnam territory in all BCSD models’ outputs for temperature downscaling, from 94.8% in CSIRO-Mk3-6-0 to 99.3% in NorESM1-M. The BCSD-ENS has 97.3% of grid points with positive AVs. The negative AV range in the models (−1 to 0) is much narrower than the positive range (0–70), suggesting only minor differences between the BCSD and BIP in the locations where the BIP is better than the BCSD (Figure 8).
Over Vietnam, the percentage of positive AV regions for precipitation downscaling in all BCSD-CMIP5 outputs vary from 88.7% in MIROC5 to 97.2% in CMCC-CM (Figure 9). Notably, the BCSD-ENS has 99.9% of grid points with positive AVs. Similar to temperature, the negative AVs vary within the narrow range of (−1 to 0), highlighting minor differences between the BCSD and BIP in those areas (Figure 9).
In summary, this subsection has demonstrated the overwhelming performance of the BCSD downscaling method for temperature and precipitation over Vietnam, even when examining the testing period 1996–2005. For each CMIP5 GCM and the ensemble mean, the BCSD outperforms the simple BIP method over almost all territories of Vietnam.
Future projections
This section presents a preliminary application of the CMIP5-VN in assessing future climate changes over Vietnam and its seven climate sub-regions.
Projected temperature changes relative to the baseline period 1986–2005 based on the CMIP5 GCMs and BCSD-CMIP5 data for global average (left) and Vietnam (right), respectively. Five-year moving averages are applied. Colored lines show the ensemble means of the models and colored shaded areas present the uncertainty ranges (±1 standard deviation) for each RCP. The number of models used for each RCP is shown in brackets. Box plots on the right display the occurrence statistics (quartile, median, 10th, and 90th percentile) for warming levels on the global scale (left boxes) and in Vietnam (right boxes) at the end of the 21st century 2080–2099.
Projected temperature changes relative to the baseline period 1986–2005 based on the CMIP5 GCMs and BCSD-CMIP5 data for global average (left) and Vietnam (right), respectively. Five-year moving averages are applied. Colored lines show the ensemble means of the models and colored shaded areas present the uncertainty ranges (±1 standard deviation) for each RCP. The number of models used for each RCP is shown in brackets. Box plots on the right display the occurrence statistics (quartile, median, 10th, and 90th percentile) for warming levels on the global scale (left boxes) and in Vietnam (right boxes) at the end of the 21st century 2080–2099.
Comparison between the seasonal temperature cycles of the BCSD-ENS and OBS over the seven climatic sub-regions of Vietnam. Black and grey dashed lines, respectively, represent the OBS and BCSD-ENS cycles for 1986–2005, while color lines represent the future 2080–2099 seasonal cycles for each RCP. Average changes between the future 2080–2099 BCSD-ENS and the 1986–2005 OBS with the historical 1986–2005 BCSD-ENS are also displayed. Please refer to the online version of this paper to see this figure in color: http://dx.doi.org/10.2166/wcc.2022.144.
Comparison between the seasonal temperature cycles of the BCSD-ENS and OBS over the seven climatic sub-regions of Vietnam. Black and grey dashed lines, respectively, represent the OBS and BCSD-ENS cycles for 1986–2005, while color lines represent the future 2080–2099 seasonal cycles for each RCP. Average changes between the future 2080–2099 BCSD-ENS and the 1986–2005 OBS with the historical 1986–2005 BCSD-ENS are also displayed. Please refer to the online version of this paper to see this figure in color: http://dx.doi.org/10.2166/wcc.2022.144.
Projected precipitation changes in Vietnam under different RCPs. (a) Colored lines present the 10-year moving average of the BCSD-ENS and colored shaded areas show the uncertainty ranges (±1 standard deviation). Box plots on the right display the occurrence statistics (quartile, median, 10th, and 90th percentile) for precipitation changes at the end of the 21st century 2080–2099. (b–e) Distribution of change patterns between the 2080–2099 and 1986–2005 periods derived by the BCSD-ENS. The hatching lines show the regions where over two-thirds of BCSD-CMIP5 models have the same bias sign as the BCSD-ENS.
Projected precipitation changes in Vietnam under different RCPs. (a) Colored lines present the 10-year moving average of the BCSD-ENS and colored shaded areas show the uncertainty ranges (±1 standard deviation). Box plots on the right display the occurrence statistics (quartile, median, 10th, and 90th percentile) for precipitation changes at the end of the 21st century 2080–2099. (b–e) Distribution of change patterns between the 2080–2099 and 1986–2005 periods derived by the BCSD-ENS. The hatching lines show the regions where over two-thirds of BCSD-CMIP5 models have the same bias sign as the BCSD-ENS.
Precipitation seasonal cycles (left axis, solid lines) under different RCPs and their percentage changes (right axis, dash lines) relative to the 1986–2005 BCSD-ENS over the seven climatic sub-regions of Vietnam. Average changes (%) of the BCSD-ENS between the future 2080–2099 and the baseline 1986–2005 period are also displayed.
Precipitation seasonal cycles (left axis, solid lines) under different RCPs and their percentage changes (right axis, dash lines) relative to the 1986–2005 BCSD-ENS over the seven climatic sub-regions of Vietnam. Average changes (%) of the BCSD-ENS between the future 2080–2099 and the baseline 1986–2005 period are also displayed.
CONCLUSIONS
In this study, the high-resolution (10 km) daily temperature (including near-surface daily average, daily maximum, and minimum temperatures) and precipitation dataset for Vietnam, called CMIP5-VN, has been successfully developed by applying the BCSD method for 31 CMIP5 GCMs for the baseline period 1980–2005 and the future period 2006–2100 under four GHG concentration scenarios RCP2.6, RCP4.5, RCP6.0, and RCP8.5.
With the newly-built CMIP5-VN dataset, future scenarios for Vietnam are not only limited to RCP4.5 and RCP8.5 as shown in previous studies (e.g. MONRE 2016, 2021; Trinh-Tuan et al. 2019) but also available for RCP2.6 and RCP6.0. The CMIP5-VN is recommended for studies on climate change assessment and climate change impacts in Vietnam.
As outputs of CMIP6 GCMs have recently become available on the ESGF, future studies will therefore need to downscale, both statistically and dynamically, the CMIP6 outputs for the region. Furthermore, it should also be noted that the CMIP5 and CMIP6 results may not cover all probabilities that could occur in the future. Thus, a study that applies the probabilistic method suggested by Rasmussen et al. (2016) and Hsiang et al. (2017) could be considered to build a next climate change dataset for Vietnam. This new approach is expected to better depict the tails of the probability distribution and, therefore, could better project the extreme risks that may occur in the future.
AUTHORS’ CONTRIBUTIONS
Q.T.-A., T.N.-D., and E.E. conceptualized the research. Q.T.-A. carried out the downscaling experiments and performed the analysis. L.T.-T. processed and prepared the observation data. Q.T.A. and T.N.-D. drafted the manuscript. All authors discussed the results and commented on the manuscript.
FUNDING
This work belongs to the CLIMATO package of the GEMMES Vietnam project, launched by the French Development Agency (AFD).
DATA AVAILABILITY STATEMENT
All relevant data are included in the paper or its Supplementary Information.
CONFLICT OF INTEREST
The authors declare there is no conflict.