Simulation and projection of climate change using CMIP6 Muti-models in the Belt and Road Region

YanRan Lü,Tong Jiang*,YanJun Wang,BuDa Su,JinLong Huang,Hui Tao

1.Institute for Disaster Risk Management/School of Geographical Sciences,Nanjing University of Information Science&Technology,Nanjing,Jiangsu 210044,China

2. State Key Laboratory of Desert and Oasis Ecology, Xinjiang Institute of Ecology and Geography, Chinese Academy of Sciences,Urumqi,Xinjiang 830011,China

ABSTRACT Climate condition over a region is mostly determined by the changes in precipitation, temperature and evaporation as the key climate variables.The countries belong to the Belt and Road region are subjected to face strong changes in future cli‐mate. In this paper, we used five global climate models from the latest Sixth Phase of Coupled Model Intercomparison Project (CMIP6) to evaluate future climate changes under seven combined scenarios of the Shared Socioeconomic Path‐ways and the Representative Concentration Pathways (SSP1-1.9, SSP1-2.6, SSP2-4.5, SSP3-7.0, SSP4-3.4, SSP4-6.0 and SSP5-8.5)across the Belt and Road region.This study focuses on undertaking a climate change assessment in terms of fu‐ture changes in precipitation, air temperature and actual evaporation for the three distinct periods as near-term period(2021?2040),mid-term period(2041?2060)and long-term period(2081?2100).To discern spatial structure,K?ppen?Gei‐ger Climate Classification method has been used in this study.In relative terms,the results indicate an evidence of increas‐ing tendency in all the studied variables,where significant changes are anticipated mostly in the long-term period.In addi‐tion to,though it is projected to increase under all the SSP-RCP scenarios,greater increases will be happened under higher emission scenarios(SSP5-8.5 and SSP3-7.0).For temperature,robust increases in annual mean temperature is found to be 5.2 °C under SSP3-7.0, and highest 7.0 °C under SSP5-8.5 scenario relative to present day. The northern part especially Cold and Polar region will be even more warmer(+6.1°C)in the long-term(2081?2100)period under SSP5-8.5.Similar‐ly, at the end of the twenty-first century, annual mean precipitation is inclined to increase largely with a rate of 2.1% and 2.8%per decade under SSP3-7.0 and SSP5-8.5 respectively.Spatial distribution demonstrates that the largest precipitation increases are to be pronounced in the Polar and Arid regions. Precipitation is projected to increase with response to in‐creasing warming most of the regions. Finally, the actual evaporation is projected to increase significantly with rate of 20.3% under SSP3-7.0 and greatest 27.0% for SSP5-8.5 by the end of the century. It is important to note that the changes in evaporation respond to global mean temperature rise consistently in terms of similar spatial pattern for all the scenarios where stronger increase found in the Cold and Polar regions.The increase in precipitation is overruled by enhanced evapo‐ration over the region. However, this study reveals that the CMIP6 models can simulate temperature better than precipita‐tion over the Belt and Road region. Findings of this study could be the reliable basis for initiating policies against further climate induced impacts in the regional scale.

Keywords:precipitation;temperature;actual evaporation;multi-models CMIP6;SSPs-RCPs;Belt and Road Region

1 Introduction

The concentration of greenhouse gases (mostly CO2, CH4and N2O) in the Earth's atmosphere has been continuing to increase since the industrial revolu‐tion, resulting from diverse anthropogenic activity(IPCC, 2014). The global climate started to become warm significantly from the late 1970s to the early 1980s, eventually the adverse impact of climate change has been perceived worldwide (IPCC, 2013).It has been widely concerned and highly valued by the scientific and international community to cope with changing climate. Therefore, the Intergovern‐mental Panel on Climate Change (IPCC) published its first scientific assessment report in 1990. Afterwards,it has been taken increasing emphasis on assessment activities.According to the IPCC special report on the impacts of global warming of 1.5°C,it is estimated to rise global mean temperature by 0.2±0.1 °C per de‐cade, and has already warmed by 1 °C. On this trend,it will reach 1.5 °C by the year 2030 to 2052 (IPCC SR1.5, 2018). Previous studies have shown that re‐sponse to global warming differs regionally, hence,the changes in temperature and precipitation are spa‐tially different (IPCC, 2014; WMO, 2019). In general significant changes in the climate variables further lead to a greater climate risks.However,scientists and policymakers are now even more urged to assess cli‐mate change at both regional and local scale to have effective policy response.

The Belt and Road Initiative originated in China,but it belongs to the world. It is rooted in the histo‐ry, but oriented toward the future. It focuses on Asia, Europe and Africa, but is open to all partners.To remember, when visiting the Kazakhstan and In‐donesia in September and October of 2013, Chinese President Xi Jinping raised the initiative of jointly building the Silk Road Economic Belt, and the 21st Century Maritime Silk Road which is hereinafter re‐ferred to as the Belt and Road (Wuet al., 2018).Over the past seven years, the Belt and Road Initia‐tive has won positive responses from numerous countries and international organizations as well as attracted worldwide attention. However, previous studies reported that the average number of meteoro‐logical disasters, economic losses and fatalities in the Belt and Road region have accounted for more than half of the global total since 1980, and all have been on the rise (Jianget al., 2020). In order to strengthen disaster risk management and reduce cli‐mate change risks, policymakers urgently need com‐prehensive information on the changing climate as well as future evolution of related extreme events along the Belt and Road region. Therefore, it is of great scientific significance to study the possible evolution rules of future climate change for the cor‐responding mitigation and adaptation measures in the research area.

Climate model is an important tool to understand the history and predict the potential climate change in the future (Wanget al., 2018; Nature, 2019). There‐fore, in the past 25 years, the World Climate Research Program (WCRP) has launched the International Cou‐pled Model Comparison Program (CMIP) from the first phase to the sixth phase, and constructed the most extensive database of climate models output as providing an irreplaceable scientific basis for predict‐ing climate change. Nonetheless, the CMIP5 based studies reported that the annual temperature in the Belt and Road would continue to rise in the future with response to the continuous emission of green‐house gases. In addition to, the amplitude of rising temperature tends to be increased with the enhance‐ment of greenhouse gas emission as extensive in‐crease under higher emission scenario (RCP8.5) is predicted to be exceed 5°C by the end of the 21st cen‐tury(Wanget al.,2020).Likewise,the annual precipi‐tation is projected to increase over most of the region,particularly in the West Asia and North Asia. Further,the number of consecutive dry days is projected to de‐crease in North Asia and East Asia, while increasing in other regions (Zhou BTet al., 2020). It has also been anticipated to witness frequent extreme weather events in Eurasia by the twenty-first century, resulting in huge economic losses and human casualties (Cou‐mou and Rahmstorf, 2012). However, several draw‐backs of CMIP5 have been recognized in projecting climate extremes across the different countries of the Belt and Road region. To date, most studies carried out using CMIP5 models to assess global and regional climate change characteristics, where CMIP6 models have still rarely applied. Therefore, it is imperative to assess how different climate variables will be changed over this region under the new-state-of-the-art CMIP6 scenarios.

The sixth phase of the Coupled Model Intercom‐parison Project (CMIP6) proposed new climate change scenarios combining different shared socioeconomic pathways with recent trends in anthropo‐genic emissions (SSP-RCP), which is called Scenari‐oMIP(O'Neillet al., 2014; Zhao and Luo, 2016). It is expected to yield more reasonable and reliable projec‐tions based on the improved climate models from CMIP6(Eyringet al.,2016).Taking into account,this paper aim to provide a comprehensive climate change overview in terms of spatiotemporal changes in annu‐al temperature, precipitation and actual evaporation under seven SSP-RCP for the three defined future pe‐riods.The result of the study will be considered as the pivotal basis for designing strategies against climate change in the Belt and Road region.

2 Data and methods

2.1 Study area

This study has covered all the countries included in the Belt and Road region. It embrace China, 2 Northeast Asian countries, 5 Central Asian countries,20 West Asian and North African countries, 19 Cen‐tral and Eastern European countries, 11 Southeast Asian countries and 7 South Asian countries. As of 2018, the region had a total population of 4.6 billion,land area of over 57 million square kilometers and GDP of $23 trillion, which is about 62%, 39% and 31% of the world respectively (Lüet al.,2020).Natu‐ral environment of the region varies greatly.The aver‐age temperature in the study area roughly decreases from low latitudes near the equator to middle and high latitudes. The higher temperature (≥30 °C) is re‐corded in the East Asia and the southern part of the Arabian Peninsula. Whereas lower temperature exists in the Northeast Asia and the Tibetan Plateau (Zhanget al., 2019). Further, higher annual precipitation oc‐curs in Southeast Asia and South Asia, where north‐east China, Central Asia and West Asia perceive low‐er annual precipitation. Such uneven spatial distribu‐tion of precipitation further lead to regional differenc‐es in water resources (Wuet al., 2018; Zhanget al.,2019). Importantly, based on the K?ppen-Geiger cli‐mate classification (Peelet al., 2007; Kriticoset al.,2012; Becket al., 2018), the Belt and Road region can be divided into five major-regions:Tropical,Arid,Temperate,Cold and Polar.

Figure 1 The Belt and Road region with K?ppen?Geiger Climate Classification

2.2 Datasets

The Climatic Research Unit (CRU) dataset is based on meteorological stations around the world and covers the land cells of the Earth at a spatial reso‐lution of 0.5°×0.5° (Harriset al., 2014). In this paper,we selected monthly temperature and precipitation data form CRU as observations which involves 27,578 grids in the Belt and Road region.

Multi-model simulations (1995?2014) and projec‐tions(2021?2100)from Coupled Model Intercompari‐son Project Phase 6 (CMIP6) archive are used in this study. We used monthly precipitation, temperature and actual evaporation under seven SSP-RCP (SSP1-1.9, SSP1-2.6, SSP2-4.5, SSP3-7.0, SSP4-3.4, SSP4-6.0 and SSP5-8.5) scenarios from the CMIP6. The CMIP6 scenarios are the combination of Shared So‐cioeconomic Pathways (SSPs) and forcing levels of the Representative Concentration Pathways (SSPRCP).It's based on the SSP,through the common poli‐cy assumptions, the future scenario will be more rea‐sonable. It includes SSP1 (sustainability), SSP2 (mid‐dle of road), SSP3 (regional rivalry), SSP4 (inequali‐ty), SSP5 (fossil-fueled development) scenarios to de‐scribe the possible future worlds,which represents dif‐ferent combinations of mitigation and adaptation chal‐lenges. Only the first realization (r1i1p1f1) from each model is used. The SSP1-1.9 and SSP1-2.6 represents the low end of the range of future forcing pathways.It is anticipated that it will produce a multi-model mean of significantly less than 1.5 ℃and 2 ℃warming by 2100.Similarly,SSP4-3.4 scenario also at the low end of the range of future forcing pathways which reach 3.4 W/m2by 2100.The SSP2-4.5 represents the medi‐um part of the range of future forcing pathways, until 2100 is approximately 4.5 W/m2. Whereas, SSP4-6.0,SSP3-7.0 and SSP5-8.5 are represent the high end of the range of future pathways(O'Neillet al.,2017).

The prime focus of this study to cover all the new scenarios in case of model selection. Up to July 2020,there were only five models available with required variables under all the seven targeted SSP-RCP sce‐narios.Details of the selected GCMs are shown in Ta‐ble 1. Notably, among the five GCMs, there were on‐ly four GCMs (except CNRM-ESM2-1) made avail‐able actual evaporation, whereas precipitation and temperature were found available from all the select‐ed five GCMs.

Table 1 Basic information of the five global climate models(GCMs)

The actual evaporation outputs are interpolated in‐to a common resolution of 0.5°×0.5° by using IDW method. The temperature and precipitation have been bias-corrected against observational gridded data by applying the Equidistant Cumulative Distribution Function (ECDF) method and statistically downscaled to a regular resolution (0.5°) by using the spatial dis‐aggregation (SD) method (Woodet al., 2004; Liet al.,2010;Suet al.,2018).Notably,before bias correc‐tion, results show an underestimation in temperature and an overestimation in precipitation simulation. This overestimation/underestimation is basically raised be‐cause of having large biases in the GCMs. Therefore,to continue further analysis, we applied the bias cor‐rection method to reduce the biases as well as uncer‐tainties (Figure 2).After the bias correction, the mod‐el validation result suggests that CMIP6-GCM per‐forms well in projecting climate variables over the study area.The actual evaporation outputs are interpo‐lated into a common resolution of 0.5°×0.5° by using IDW method. The temperature and precipitation have been bias-corrected against observational gridded data by applying the Equidistant Cumulative Distribution Function (ECDF) method and statistically downscaled to a regular resolution(0.5°)by using the spatial disag‐gregation (SD) method (Woodet al., 2004; Liet al.,2010; Suet al., 2018). Notably, before bias correction,results show an underestimation in temperature and an overestimation in precipitation simulation.This overes‐timation/underestimation is basically raised because of having large biases in the GCMs.Therefore, to contin‐ue further analysis, we applied the bias correction method to reduce the biases as well as uncertainties(Figure 2).After the bias correction, the model valida‐tion result suggests that CMIP6-GCM performs well in projecting climate variables over the study area.

Figure 2 Comparison of multi-year averaged monthly(a)temperature(b)and precipitation in The Belt and Road region for the period of 1995?2014;unprocessed output ensemble mean(CMIP6),processed GCM ensemble mean(Correction)and CRU(observation)

In this study, the period of 1995?2014 in histori‐cal simulation is referred to as baseline period and the periods of 2021?2040,2041?2060 and 2081?2100 un‐der different emission scenarios represents the Nearterm, Mid-term and Long-term accordingly. It is im‐portant to mention that multi-model ensemble average based results are presented in this paper.

3 Results

3.1 Comparison of GCMs simulations with observations

The comparison among unprocessed GCMs out‐put (before bias correction) processed GCMs simula‐tions (after bias correction) and observed (CRU)monthly temperature and precipitation during the 1995?2014 period over the Belt and Road region are shown in Figure 2. The climate models are able to capture the monthly dynamics of regional temperature very well (Figure 2a). It shows that the simulated sea‐sonal mean temperature (after bias correction) holds less bias than unprocessed CMIP6 data.Although the multi-model ensemble mean (Correction) provides mostly negligible underestimation in capturing ob‐served temperature; gap of each month is less than 0.05 ℃(Figure 2a). Similarly, the monthly precipita‐tion distribution shows that GCMs ensemble mean(Correction)is quite good than unprocessed CMIP6 in precipitation simulation over the region (Figure 2b).Though,GCM simulations(both processed and unpro‐cessed)represents overestimation in capturing precipi‐tation but processed GCM ensemble mean discern less overestimation compare to unprocessed. Howev‐er, this underestimation/overestimation is the indica‐tion of still having some uncertainties among latest CMIP6 models output which can be further reduced by applying bias correction method.

Spatially, distribution patterns of annual mean temperature during the 1995?2014 period in The Belt and Road region are quite similar between observed and simulated(Figures 3a,3b).The temperature distri‐bution is more pronounced in the south part (e.g.,Ar‐ab, India, Bangladesh region), while diminish in the north part (e.g., central Russia, Mongolia regions).The spatial correlation coefficient between observed and simulated annual mean temperature is about 0.99.Furthermore, there is a good agreement between ob‐served and simulated annual precipitation during the period of 1995?2014 (Figures 3c, 3d).The annual pre‐cipitation in the Belt and Road region is relatively low(less than 200 mm) in the high-altitude areas (e.g.,arid and polar regions) while, annual precipitation is higher(more than 1,500 mm) in the tropical and cold regions(e.g., Indonesia).The spatial correlation coefficient be‐tween the simulated and observed precipitation is 0.95.From the above results, it can be found that the biascorrected GCMs satisfactorily characterizes the tem‐perature and precipitation in the Belt and Road region in aspects of both temporal and spatial distribution.

Figure 3 Spatial distributions of annual temperature and precipitation in The Belt and Road region for the period of 1995?2014;simulated and observed temperature(a?b);simulated and observed precipitation(c?d)respectively

3.2 Projected changes in annual mean temperature

Temporal changes in annual mean temperature for the period of 1995 ?2100 is shown in Figure 4.The CMIP6 models show an increasing trends in an‐nual mean temperature throughout the 20thand 21thcentury over the Belt and Road region under differ‐ent scenarios (Figure 4). Significant changes can be seen for the long-term period among different scenari‐os. Compared with the baseline climate (1995?2014),projected increase (per decade) for the 2015 ?2100 period will be 0.13 ℃/10a, 0.36 ℃/10a, 0.63 ℃/10a,0.23 ℃/10a, 0.45 ℃/10a, 0.85 ℃/10a for SSP1-2.6,SSP2-4.5,SSP3-7.0,SSP4-3.4,SSP4-6.0 and SSP5-8.5 scenarios respectively. It is important to mention that lowest emission scenario (SSP1-1.9) shows decreasing tendency. Evidently, temperature will rise continuously with the ongoing increase of ra‐diative forcing. The temperature changes are larg‐er for higher emission scenarios, which can be partly attributed to the larger radiative forcing level.

In order to show the temperature changes consid‐ering three different periods, we estimated relative changes for each 20 years period. With relative to 1995 ?2014, warming rate stays about 1.4 °C in the near-term(2021?2040)period under all SSP-RCP sce‐narios. Under lower forcing scenarios (SSP1-1.9 and SSP1-2.6),the increase is about 1.6 °C and 1.9 °C ac‐cordingly in the mid-term (2041?2060) period, which reaches that of 1.2 °C and 2.0 °C respectively in the long-term period (2081?2100). Under medium forcing scenario(SSP2-4.5 and SSP4-3.4),in the mid-term pe‐riod,it will be approximately 2.1°C and 2.3°C respec‐tively,while in the long-term period it goes up to 2.8°C and 3.5 °C. The most robust increase reaches 4.3 °C,5.2 °C and 6.8 °C in long-term (2081?2100) period for higher emission scenarios SSP4-6.0, SSP3-7.0 and SSP5-8.5 respectively.

In spatial distribution (Figure 5), a clear increase in temperature appears in the northern part of the Belt and Road region during 2021?2100 period, especially in the Cold and Polar regions. In these areas, increase will be 1.7 ℃, 1.6 ℃, 1.8 ℃, 1.7 ℃, 1.8 ℃, 1.9 ℃and 2.2 ℃, respectively under SSP1-1.9, SSP1-2.6,SSP2-4.5,SSP3-7.0,SSP4-3.4,SSP4-6.0 and SSP5-8.5 in the near-term period. In the aforementioned areas,temperature is anticipated to increase up to 1.4 ℃and 2.7 ℃in the mid-term and long-term period respec‐tively. The changes in Arid region will be increased up to 1.2 ℃under lower forcing scenarios (SSP1-1.9 and SSP1-2.6), 1.3 ℃for medium forcing scenarios(SSP2-4.5 and SSP4-3.4),and for higher emission sce‐narios (SSP4-6.0, SSP3-7.0 and SSP5-8.5) it will reach up to 1.5 ℃in the near-term period. While, in the Temperate region, it will be increased approxi‐mately 1.1 ℃under different scenarios in near-term.At the hottest region (Tropical region), temperature would be rose very small in amount compared with other regions. In mid-term, annual mean temperature at the northern part of the Belt and Road region will increase by 1.6 °C?3.0 °C. In the long-term period,the increase of the annual mean temperature will in‐crease by 1.2 °C?6.8 °C respectively lower emission to higher emission scenario, especially over the north‐ern part of the Belt and Road region.

Figure 5 Spatial patterns of projected temperature changes(℃)during the near-term(2021?2040),mid-term(2041?2060)and long-term(2081?2100)period relative to the baseline period(1995?2014)in the Belt and Road region under SSP1-1.9,SSP1-2.6,SSP2-4.5,SSP3-7.0,SSP4-3.4,SSP4-6.0 and SSP5-8.5

Table 2 Changes(℃)in temperature for the three defined periods under seven SSP-RCP relative to 1995?2014 period over the five regions of the Belt and Road regions

3.3 Projected changes in annual mean precipitation

Figure 6 shows that the annual precipitation in the Belt and Road region will increase from 2021 to 2100 with a rate of 0.1%/10a,0.6%/10a,1.5%/10a,2.1%/10a,1.0%/10a, 1.9%/10a, 2.8%/10a, respectively under SSP1-1.9, SSP1-2.6, SSP2-4.5, SSP3-7.0, SSP4-3.4,SSP4-6.0 and SSP5-8.5.During the 2021?2060 period,it will reach with a trend of 0.9%/10a for SSP1-1.9,1.6%/10a for SSP1-2.6, 1.5%/10a for SSP2-4.5,1.6%/10a for SSP3-7.0, 1.4%/10a for SSP4-3.4,2.2%/10a SSP4-6.0, and 2.4%/10a for SSP5-8.5.After 2060, trend of annual precipitation shows some differ‐ent features. However, for SSP3-7.0 and SSP5-8.5,annual precipitation will continuously rise until 2060, and then effectively stabilized after 2060 with slope of 1.6%/10a and 2.4%/10a respectively during the period of 2021 ?2060. Which further reaches 2.2%/10a and 3.8%/10a during 2085 ?2100 period accordingly.

Figure 6 Temporal changes of annual precipitation in the Belt and Road region during 1995?2100 period under the scenarios of SSP1-1.9,SSP1-2.6,SSP2-4.5,SSP3-7.0,SSP4-3.4,SSP4-6.0 and SSP5-8.5

In terms of spatial distribution (Figure 7), we used hydrological sensitivity to reflect precipitation change.The sensitivities of precipitation to global mean surface air temperature increase under the seven scenarios are given in Figure 6. Compared to the baseline period(1995?2014),the larger portion of the Belt and Road re‐gion demonstrates to have more precipitation considering response rates of annual precipitation to global warming.For 2021?2100 period,the increasing tendency,over the Belt and Road regions will be 5.7%/℃for SSP1-1.9,8.7%/℃for SSP1-2.6, 4.2%/℃for SSP2-4.5, 4.2%/℃for SSP3-7.0,10.8%/℃for SSP4-3.4,8.3%/℃for SSP4-6.0 and 9.4%/℃for SSP5-8.5. In case of five distinct re‐gions,it will be ?1.5%/℃for Tropical region,20.7%/℃for Arid region,4.1%/℃for Temperate region,8.2%/℃for Cold region,11.1%/℃for Polar region separately.

Figure 7 Multi-model ensemble mean of annual mean precipitation responses(%/℃)to global-mean surface air temperature for the period of the near-term(2021?2040),mid-term(2041?2060)and long-term(2081?2100)relative to the reference period(1995?2014)under SSP1-1.9,SSP1-2.6,SSP2-4.5,SSP3-7.0,SSP4-3.4,SSP4-6.0 and SSP5-8.5

In the near-term period, annual mean precipitation exhibits robust increase with warming level over most part of the Arid region. The most obvious change ap‐pears in SSP1-1.9(35.5%/℃)and SSP1-2.6(47.1%/℃),similarly increase changes are SSP2-4.5 (9.2%/℃),SSP4-3.4(17.5/℃),SSP4-6.0(10.1%/℃)and SSP5-8.5(14.1%/℃), but decrease by ?14.5%/℃in SSP3-7.0.In mid-term period, precipitation decreases with in‐creasing temperature for most of the scenarios. How‐ever, strong increases of area-averaged annual mean precipitation is pronounced with warming level in the long-term period.

Table 3 Changes(%/℃)in hydrological sensitivity for the three defined periods under seven SSP-RCP relative to 1995?2014 period over the five regions of the Belt and Road regions

3.4 Projected changes in annual mean actual evaporation

Temporal changes in annual mean actual evapora‐tion for the period of 1995?2100 is shown in Figure 8. The area average annual actual evaporation over the Belt and Road region was approximately 467.0 mm for the baseline period(1995?2014).The area average annual actual evaporation will increase from 2021 to 2100, with a rate of 0.2%/10a for SSP1-1.9, 0.7%/10a for SSP1-2.6, 1.3%/10a for SSP2-4.5, 1.6%/10a for SSP3-7.0, 0.9%/10a for SSP4-3.4, 1.5%/10a for SSP4-6.0, and 2.2%/10a for SSP5-8.5. For SSP1-1.9 and SSP1-2.6, annual mean actual evaporation will continuously rise until 2060 with slope of 1.2%/10a and 1.8%/10a during 2021?2060 period, and begin to decrease thereafter with rate of ?0.4%/10a and?0.3%/10a during the 2061?2100 period.For SSP4-3.4,actual evaporation will continuously rise until 2070 with slope of 1.3%/10a and then effectively stabilized after 2070(during 2070?2100)with slope of 0.5%/10a.For SSP2-4.5, SSP3-7.0, SSP4-6.0 and SSP5-8.5, ac‐tual evaporation will keep rise continuously with the ongoing increase of radiative forcing.

In terms of three defined time frames,area averaged actual evaporation will be approximately 2.6%/10a and 2.5%/10a under SSP1-1.9 and SSP1-2.6 respectively in the near-term period. Under these lower emission scenarios (SSP1-1.9 and SSP1-2.6) increase reach with a rate of 0.4%/10a and 1.4%/10a in the mid-term,-whereas decrease as ?0.7%/10a and ?0.5%/10a re‐spectively in the long-term period. Under medium forcing scenarios (SSP2-4.5, SSP4-3.4) projected in‐crease will be approximately 1.5%/10a for both the period of near-term and mid-term period. Whereas,relative increase is estimated 0.8%/10a and 0.4%/10a higher in the long-term period than earlier periods. It is a remarkable fact that the increase in evapotranspi‐ration will grow fast under SSP3-7.0, SSP5-8.5 in the long-term period and reach at rate of approximately 2.2%/10a and 2.7%/10a respectively.

In aspect of spatial distribution (Figure 9), most regions of the Belt and Road region are showing a growth trend during the whole period of 2021?2100,especially in the northern part which includes Russia,China, and parts of the Arabia. In different scenarios,projected change will be increased by 7.4%, 6.8%,6.2%,5.9%,6.9%,6.3% and 7.8%,respectively under SSP1-1.9, SSP1-2.6, SSP2-4.5, SSP3-7.0, SSP4-3.4,SSP4-6.0 and SSP5-8.5 in the near-term period.In the mid-term period increase varies 9.8%?13.8% for all the scenarios. In the study area, projected increases will be more significant in the long-term, with rate of 8.3%,12.1%, 16.7%, 20.3%, 14.6%, 18.5% and 27.0%under SSP1-1.9, SSP1-2.6, SSP2-4.5, SSP3-7.0,SSP4-3.4,SSP4-6.0 and SSP5-8.5,respectively.

In Tropical area, spatial changes in the near-term,mid-term and long- term will be almost similar under seven scenarios. Actual evaporation in the Cold re‐gion and Polar region will increase significantly in the long-term, especially under higher emission scenarios SSP4-6.0 SSP3-7.0 and SSP5-8.5 than the other sce‐narios. In Tropical region, there will be a smaller in‐crease in actual evaporation. Among the three future periods, anticipated changes in actual evaporation comparatively greater in the long-term period under SSP3-7.0 and SSP5-8.5 about 7.6%. It is noteworthy,actual evaporation will increase in Arid region under all the scenarios during 2021 ?2100 period except SSP1-1.9.The decrease of actual evaporation estimat‐ed in the Arid region about 2.5%, whereas largest in‐crease in actual evaporation will be significant under SSP5-8.5.

Figure 8 Temporal changes of annual mean actual evaporation in the Belt and Road region during 1995?2100 under the scenarios of SSP1-1.9,SSP1-2.6,SSP2-4.5,SSP3-7.0,SSP4-3.4,SSP4-6.0 and SSP5-8.5

Table 4 Changes(%)in actual evaporation for the three defined periods under seven SSP-RCP relative to 1995?2014 periodover the five regions of the Belt and Road regions

Figure 9 Spatial patterns of actual evaporation for the period of near-term(2021?2040),mid-term(2041?2060)and long-term(2081?2100)under SSP1-1.9,SSP1-2.6,SSP2-4.5,SSP3-7.0,SSP4-3.4,SSP4-6.0 and SSP5-8.5 scenarios relative to the reference period(1995?2014)in the Belt and Road region

4 Discussions

This study focuses on the projected changes in temperature, precipitation and actual evaporation throughout the 21st century.We used five models out‐put under seven SSP-RCP scenarios from the CMIP6 to describe the changes over the Belt and Road region for the three future periods.GCMs performance is an‐alyzed by comparing with ground observations for the period of 1995?2014. The bias correction and multimodel ensemble mean provides sound consistency with observed data, which proves the confidence in projection capability of CMIP6 GCMs over the Belt and Road region. It has great similarities with previ‐ous research results that CMIP6 GCMs demonstrates some added improvements. In this study, based on the ensemble mean of five GCMs, there will be an in‐crease in temperature over The Belt and Road region in the 21st century. Our research result has demon‐strated that with the background of global warming and climate anomalies, the average temperature will continue to increase in the Belt and Road region with the rate of warming being greater at high latitudes than at low latitudes, and with higher emissions sce‐narios. However, there are significant regional differ‐ences and situational dependencies. The warming is mostly pronounced in the northern part of the study area. Importantly, by the mid to long term period,CMIP6 shows warming will be about 2 °C above which is mostly similar as CMIP5 based study that un‐der the highest emissions scenario warming will be higher. The spatial pattern of warming is mostly pro‐nounced in the cold and polar regions.In terms of pre‐cipitation, SSP3-7.0 and SSP5-8.5 shows higher in‐creased by 17.5% and 23.5% respectively relative to 1995?2014 period. Precipitation will increase signifi‐cantly in the West and North Asia. In most of the arid regions, it will be increased up to 1.0?3.0 °C under forcing scenarios. Also, the average annual precipita‐tion showed a strong trend of increase with the in‐crease of temperature. It is noteworthy that the actual evaporation will increase 3.0%?20%in the future.This means that the risk of future droughts will increase due to enhanced actual evaporation caused by global warming (Wanget al.,2020;Zhou Jet al.,2020).The higher emission always associates with both the high‐er precipitation and evapotranspiration over the Belt and Road region. (Zhouet al., 2012). However, this study exhibit that the lower emission corresponds to the higher precipitation and actual evapotranspiration mainly in the mid-21st century. As precipitation in most of the areas is projected to increase resulting from global temperatures rise, which could ease water pressure more significantly but increased actual evap‐oration will make drought worse in some areas.

However, compared to CMIP5, the precipitation,evapotranspiration slightly increased by CMIP6.Simi‐larly, there is a risk of drought in Central Asia, the East Asian monsoon region, and Northern South Asia(Zhouet al., 2020). It should be noted that there are some unavoidable uncertainties in predicting future climate change and risks. In this paper, uncertainties may come from climate models, scenarios, bias cor‐rection, and the interpolation method. Besides, uncer‐tainty can be raised from the design of the SSP-RCP scenarios. Existed uncertainties can be demonstrated by the differences in precipitation, temperature, and actual evaporation changes under different scenarios,especially under the newly developed scenarios.Com‐pared to CMIP5, the CMIP6 models exhibit an im‐provement in simulation against observations, though models tend to show much smaller biases than that among their predecessors (Chenet al., 2020). Never‐theless, a more in-depth analysis is needed to explain the internal sensitivity of the specific model, which needs further study.

5 Conclusions

In this study, we have investigated the changing climate condition over the Belt and Road region in terms of temperature, precipitation and actual evapo‐ration changes using CMIP6 multi-model projections under seven emission scenarios (SSP1-1.9, SSP1-2.6,SSP2-4.5, SSP3-7.0, SSP4-3.4, SSP4-6.0 and SSP5-8.5)for the period of 2021?2100. The major findings are summarized in below:

(1) Except SSP1-1.9 scenario, annual mean tem‐perature is projected to increase continuously by 0.13℃/10a, 0.36℃/10a, 0.63℃/10a, 0.23℃/10a,0.45℃/10a,0.85℃/10a for SSP1-2.6,SSP2-4.5,SSP3-7.0,SSP4-3.4, SSP4-6.0 and SSP5-8.5, respectively. In the near-term period, the spatial pattern of tempera‐ture rise seems to be mostly similar under all the SSP-RCP scenarios. In the mid-term period, spatial distribution is more pronounced in the northern part than the southern part.Particularly in the Cold and Po‐lar regions, increase reaches up to 1.4 ℃and 2.7 ℃in the mid-term and long-term period respectively. It's worth noting that increase reaches highest 6.8 ℃in the long-term (2081?2100) period for SSP5-8.5 scenario.The Tropical and Temperate regions will experience lower temperature compared with other region.

(2) For precipitation changes, multi-model ensem‐ble mean shows an increasing trend across the Belt and Road region for the whole period of 2021?2100,with a rate of 0.1%/10a for SSP1-1.9, 0.6%/10a for SSP1-2.6, 1.5%/10a for SSP3-7.0, 2.1%/10a for SSP2-4.5, 1.0%/10a for SSP4-3.4, 1.9%/10a for SSP4-6.0 and 2.8%/10a for SSP5-8.5 relative to pres‐ent-day. Relatively bigger changes are perceived in the long-term projection under all the seven scenarios,especially for SSP3-7.0 and SSP5-8.5 scenarios secur‐ing 17.5% and 23.5%. In space, we used hydrological sensitivity to reflect precipitation change. Annual mean precipitation exhibits robust increase with en‐hanced warming level over most part of the Arid re‐gion. The Cold and Polar region experience increased precipitation at around 11.0%/℃. In Tropic region,precipitation shows a decreasing trend with increasing temperature.

(3) With regard to baseline period (1995 ?2014),the increasing trend of evaporation will be 0.2%/10a,0.7%/10a, 1.3%/10a, 1.6%/10a, 0.9%/10a, 1.5%/10a and2.2%/10a for SSP1-1.9, SSP1-2.6, SSP2-4.5,SSP3-7.0, SSP4-3.4, SSP4-6.0, SSP5-8.5, respective‐ly. Largest increase in actual evaporation is anticipat‐ed under SSP3-7.0 and SSP5-8.5 scenario in the longterm period. In spatial terms, most regions of the Belt and Road region shows a growth trend during the peri‐od of 2021?2100.In the long-term period,the project‐ed increase in evaporation under high forcing scenari‐os (SSP3-7.0 and SSP5-8.5) will be significantly larg‐er than the other scenarios. It is worth that the higher emission always associates with both the higher pre‐cipitation and evapotranspiration over the Belt and Road region (Zhouet al., 2020). However, this study exhibits that lower emission corresponds to the higher precipitation and actual evapotranspiration mainly in the mid-21st century. The decrease pattern of actual evaporation across the Arid region will be about 2.5%under SSP1-1.9, but increase of actual evaporation will be significantly largest under SSP5-8.5.

Acknowledgments:

This study was cooperatively funded by National Key Research and Development Program of ChinaMOST(2018FY100501)The authors are thankful for the sup‐port by the Postgraduate Research & Practice Innova‐tion Program of Jiangsu Province (KYCX20_0957),High-level Talent Recruitment Program of the Nan‐jing University of Information Science and Technolo‐gy (NUIST), and the Guest Professor Program of the Xinjiang Institute of Ecology and Geography, CAS.The authors would like to thank the World Climate Research Program's working group on coupled model‐ing and European Centre for Medium-Range Weather Forecasts for producing and making available their model output.