The increasing risk of future simultaneous droughts over the Yangtze River basin based on CMIP6 models

Drought projection is critical for water resource planning and management, as well as disaster prevention and mitigation. As a strategic national water source for China, the Yangtze River Basin (YRB) plays a vital role in the connectivity of rivers and economic development, flowing through 11 provincial administrative regions and is injected into the East China Sea, with a total length of 6,397 km. The watershed covers an area of 1.8 million square kilometers, accounting for about 1/5 of China's total land area. However, frequent droughts have caused water shortages in the YRB in recent years. Based on observed meteorological and hydrological data, the CMIP6 model and SPEI (standardized precipitation evapotranspiration index) drought models were used to elucidate the risk of future simultaneous droughts in the upper and mid-lower reaches of the YRB from 2015 to 2100. SRI has been used based on SWAT model to study the transfer process of meteorological drought to hydrological drought. The results indicated that, (1) The average of 10 CMIP6 models showed a good verification of historical precipitation and temperature for drought predictions. The MMK and Sen’s slope demonstrated consistency for historical and future droughts in the YRB. From a historical perspective (1961–2019), the middle reaches of the YRB experienced intensifying drought frequency with the highest total drought (Moderate and above drought events) frequency (> 17%); (2) In the future (2020–2100), the higher emission signifies higher moderate and total drought frequency, intensity, and scope of the YRB in FF, lower in NF. The ratio of autumn severe and extreme droughts would increase in mid-twenty-first century; (3) Severe drought risk encounters were projected in the upper and meanwhile in the middle-lower reaches in YRB, especially in the 2030–2040 period. Under all three scenarios, severe droughts occurred more frequently with SPEI close to − 2. The middle-lower reaches of the YRB are forecast to witness the largest scope and highest intensity of drought under the SSP1-2.6 scenario.; (4) The future runoff in the YRB during the dry period varied less, but in May and June during the main flood season the runoff under SSP1-2.6 would be the largest. Maximum decrease in runoff in the mid-lower reaches under the SSP2-4.5 scenario would be 2045, reaching 13.9%. Extreme flooding events and extreme meteorological droughts would happen accompanying with hydrological droughts would occur more frequently and severely under different scenarios. More attention and improved strategies should be brought to bear to address future simultaneous droughts in the upper and mid-lower YRB.


Introduction
Drought is an important natural hazard in many regions around the world, and there are significant concerns that climate change will increase the frequency or severity of drought events in the future (Cook et al. 2020). According to the International Disaster Database (EM-DAT, 2021), drought accounts for 59% economic losses induced by climate-related extremes (Li et al. 2021a, b, c). However, Debates on whether future droughts will increase or decrease on a global scale have been taking place for a long time. Drought is anticipated to increase in frequency and severity in the future due to climate change, primarily as a consequence of decreased regional precipitation, but also because of increasing evaporation driven by global warming (Sheffield andWood 2008, Lavell et al. 2012). Climate change is forecast to exacerbate drought frequency and severity over much of the planet (Dai 2011(Dai , 2013, Extended author information available on the last page of the article while some studies (e.g., (Sheffield et al. (2012)) have suggested that there has been little change in drought events over the last 60 years, taking into account changes in available energy, humidity and wind speed and thus responds incorrectly to global warming. The main rationale for these controversies is the inadequate simulation capacities of climate models and significant uncertainties in the prediction of future climate. Thus, under these circumstances the accuracies of drought projections still require considerable improvements.
The most prominent index of meteorological drought in the United States is the Palmer Drought Severity Index (PDSI). The PDSI was created with the intent of ''measuring the cumulative departure of moisture supply'' (Palmer 1965). However, the PDSI has spatial simplicity; thus, it cannot accurately describe characteristic large-scale drought changes (Maity et al. 2016). McKee et al. (1993) proposed the standardized precipitation index (SPI index) in 1993 to reflect the hydrological gain and loss at different time scales. However, the influence of temperature is ignored in the calculation, leading to a certain error in the description of drought events by the SPI index. The SPEI considers monthly precipitation, surface and potential evapotranspiration against the background of global warming (Gao et al. 2017). SPEI combines the advantages of PDSI index which takes into account the influence of evapotranspiration on drought. SPEI based on precipitation and potential evapotranspiration, combines the sensitivity of PDSI to changes in evaporation demand, caused by temperature fluctuations and trends (Vicente-Serrano et al. 2012). SPEI can conveniently monitor short-and long-term drought using selected timescales (Zhao et al. 2017a(Zhao et al. , 2017b. SPEI assesses the climatic water balance (precipitation minus potential evapotranspiration) over multiple time scales (Zhao et al. 2017a(Zhao et al. , 2017b. SPEI also considers multiple time scales of SPI, making SPEI more convenient and scientific in analyzing meteorological drought characteristics in drought-prone areas at spatiotemporal scales. In terms of long-term drought forecasting, various hydrometeorological data outputs of GCMs (Global Climate Models) from different phases of the CMIP (Coupled Multi-model Inter-comparison Project) have been employed (Sheffield and Wood 2008;Zhai et al. 2020). Studies using CMIP5 simulations have advanced our understanding of regionally heterogeneous hydroclimatic responses to global warming (Cook et al. 2014). Currently, CMIP6 (Sixth Phase) promises to fill the scientific gaps of the previous CMIP phases because of the common standards, coordination, infrastructure, and documentation that will facilitate the distribution of model outputs and the characterization of the model ensemble (Eyring et al. 2016). Previous investigations revealed that CMIP6 was more reliable and stable than CMIP5 (Chen et al. 2020;Nie et al. 2020). This is due to CMIP6 adopting more sensible shared socioeconomic pathways (SSPs) than CMIP5 through the exchange of policy assumptions to function in tandem with RCPs (Ritchie and Dowlatabadi 2017).
The YRB is a strategic water resource in China, and one of the most prosperously developed and densely populated regions in China (Li et al. 2021a, b, c). The main climate type in the YRB region is subtropical monsoon, where over the last 60 years the drought rate has changed greatly from year to year. As the water source for the South-North Water Transfer, the occurrence of drought in the YRB has profound impacts on China's economic development and security of the national water network, where climatic and geological conditions are extremely complex. Some previous studies have examined future drought characteristics in the YRB. For instance, previous studies show that many drought projection have been conducted in short time. Wang (2019) used the Modified Mann-Kendall (MMK) trend test to analyze the trend of SPEI values in the YRB over a three-month scale. The scale were concentrated in large scales (Song et al. 2021;Ma et al. 2022;Yu et al. 2022). _ENREF_45 Wang et al. (2021) assessed global drought characteristics using multiple indicators based on eleven CMIP6 models. Based on CMIP6 simulations, several drought predictions have been conducted at both regional and global scales (Ukkola et al. 2020;Zhai et al. 2020). CMIP6 has added improvements in multiple aspects, for example, in capturing the spatiotemporal pattern of monsoon over Indian landmass (Gusain et al. 2020). The multi-year mean of drought intensity in the future projection period is not much different from the historical period (Liu et al. 2020). Zhai et al. (2020) estimated that future average drought duration and frequency in Northwestern South Asia are expected to increase significantly based on the CMIP6 multi-model product. Simulations of downscaled and MME (Multi model ensemble) precipitation and temperature revealed improved performance and higher stability (Zhang et al. 2021). Most previous studies targeted on the drought time but ignored other important drought characteristics (e.g., duration, severity). It remains complex to employ global models for precise long term regional drought projection research. The areas with frequent droughts are primarily located in the upper reaches, while high-intensity droughts mostly occur in the middle and lower reaches of the Jinsha River. Also, seasonal drought encounters in the upper and middle and lower reaches of the Yangtze River are relatively scarce. Consequently, there is lack of research in-depth of the drought frequency, spatial, and temporal distribution in the YRB divisions, especially based on CMIP6 results. Taking all the above into consideration, this study was mostly based on CMIP6 models to evaluate the frequency distribution, and spatial and temporal patterns of drought predictions. These analyses focused on three research aspects, (1) Whether the frequency, intensity, and duration of drought in the divisions increase especially under global warming in the long-term projection? (2) What will the near future seasonal severe and extreme drought trend be in the YRB under CMIP6 simulation? (3) Will future drought encounter events occur more prominently and simultaneously in the upper and captured at the same time in the middle-lower reaches of the YRB under the SSPs scenarios? The framework if this research is as follows (Fig. 1).

Study area and data
The YRB refers to a vast region through which the mainstream and tributaries of the Yangtze River flow through the hinterland of China, spanning the three major economic zones of Western, Central, and Eastern China (Fig. 2). Although the YRB has abundant water resources, the spatial and temporal distribution of precipitation in the basin is very uneven. The average annual precipitation ranges from less than 300 mm in the west to over 2400 mm in the east, where [ 60% of the rainfall is concentrated in summer.
This study established 1961-2019 as the historical period, 2020-2100 as the future period. 138 meteorological stations in the YRB were selected. Since the establishments of the meteorological points varied, the meteorological data for 59 years (January 1961-December 2019) were uniformly selected, and stations with missing meteorological data that accounted for more than 5% of the time series were excluded. Missing values were supplemented by linear interpolation. The Global SPEI Monthly Dataset was at a spatial resolution of 0.5°9 0.5°(http://spei.csic.es/data base.html).

Methodology
For this study, the SPEI was employed to classify drought grading standards. Grade classification standards refer to  Table 2. The SPEI index was calculated based on two different potential evapotranspiration formulas due to different parameter acquisition in the CMIP6 data (Vicente-Serrano et al. 2010). The PET based on the Penman formula (PM) was to calculate the SPEI for the historical period. The PET based on the Thornthwaite formula (Tho) was used to calculate the future SPEI index (Thornthwaite 1948): Tho: 2.3 PM: Record the difference between monthly precipitation and potential evapotranspiration as D i : Normalize the D k n and get the SPEI value: In the above normalization process, Vicente-Serrao et al. compared the Log-logistic distribution, Pearson III distribution, Log normal distribution and Generalized extreme value distribution with the sequence D k n , and the results showed that the Log-logistic distribution was the best fit to the sequence D k n (Thornthwaite 1948). Drought frequency describes the frequency of drought occurrence at a station or grid point during the study per-iod, and the larger the value, the higher the frequency of drought occurrence, calculated as follows: where n is the total number of years, that the drought occurred at the site, and N is the total length of the data series of the site, i.e. the total number of time.
The drought station sub-ratio describes the extent of drought impact at a certain point in time when drought occurs in the study area, and the larger the value indicates the greater the extent of drought impact, calculated as follows: The Mann-Kendall (MK) test is often used to evaluate trends of long series of hydrological, meteorological, and other data (Mann 1945). Modified Mann-Kendall (MMK) was proposed by a modified MK test (Hamed and Rao 1998) to test the significance of SPEI sequences for different time periods and scenarios. Below are the calculation steps:for a time series X ¼ x 1 ; x 2 ; Á Á Á ; x n f g , The period statistic S is: where sgn is the sign function, the time lag correlation coefficient r i of the original time series X corresponding to the rank is calculated, and if r i passes the significance test at a given significance level a, the variance varðSÞ is calculated according to the following formula: i¼1 ðn À iÞðn À i À 1Þðn À i À 2Þr i , the statistic Z in the MMK trend test can be obtained, and the calculation formula is as follows: Sen's slope estimator is a simple nonparametric estimator (Sen 1968). Sen ? Mann-Kendall combination can improve the accuracy of judging climate trend changes (Li et al. 2015). The slope between all points S i of a given time series is calculated, X ¼ x 1 ; x 2 ; Á Á Á ; x n f g : In this study, to increase the applicability of the simulation results in the research area, Bias Corrected Spatial Disaggregation (BCSD) (Li et al. 2010) was used. The data of each mode was resampled to a resolution of 0.5°9 0.5°. The Bias Correction was as follows: X is the variable value, F is the cumulative probability distribution function, o À c represents the actual values in the historical period, m À c represents the simulation values in the historical period, m À p represents the simulation values in the future period, and X mÀp;adj is the correction result of the simulated value in the future period.
The Taylor diagram was adopted to appraise the simulation of all the bias-corrected CMIP6 climate patterns, which provided a concise statistical summary of how well the patterns resembled each other in terms of their centered RMS (root-mean-square) difference, standard deviation, and correlation (Taylor 2001): Measured data mean square error: Mode data mean square error: Correlation coefficient: Root mean square error: where O n is the value in the sequence of observations, O is the mean value of the sequence of observations, X n is the value in the sequence of pattern data, and X is the mean value of the pattern data sequence. The EOF method (Empirical orthogonal function decomposition), was initially proposed by statistician Pearson in 1902. Lorenz integrated it into atmospheric science research in the mid-1950's. The basic principle was to decompose the field containing m spatial points over time, and subsequently to extract several major independent and orthogonal spatially distributed and time-varying characteristic patterns from the elemental variable fields.
The EOF decomposition gives: V is a function of spatial coordinates only, and is the eigenvector of the covariance array quantity, which is the eigenvalue of the explained variance corresponding to the covariance array; T is the time function, which is uniquely determined by V and X.
SAWT (Soil and Water Assessment Tool) model is based on physical mechanism and can consider various factors such as topography, vegetation, weather, land use, etc. The model is favored by many domestic and foreign workers in runoff simulation and water resources evaluation and management because of its open source, easy access to input variables, strong physical mechanism, and the ability to simulate long series of watersheds, etc.
The model portrays the hydrological cycle process in two main parts, the first part is the calculation of surface confluence, and the second part is the calculation of river network confluence. The water balance equations in the model are as follows: SW t is the water content at the end moment of soil water, SW 0 denotes the water content at the beginning moment of soil water, t denotes the time period of the current hydrological cycle process, R day;i denotes the amount of precipitation on dayi, Q surf;i denotes the amount of surface production on day i, E a;i denotes the total amount of evapotranspiration on dayi, W seep;i denotes the amount of infiltration on dayi, and Q gw;i denotes the amount of groundwater remitted to the river network on dayi.
The coefficient R 2 and the NS (Nash-Sutcliffe coefficient) coefficient evaluation of the model efficiency are used. The evaluation is carried out, where R 2 is used to evaluate the degree of agreement between simulated and measured runoff. NS coefficient is used to evaluate the model efficiency (Panagopoulos et al. 2012). The formula for calculating R 2 and the NS coefficient are communicated as follows: Q is the flow rate: Q is the average value of the flow rate, o is the measured value, s is the simulated value, i means the ith data that the closer the R 2 to 1, the more credible the simulation result is, usually R 2 [0.5 is considered that the result meets the requirements. The NS coefficient is also considered the closer to 1, the more efficient the model is; when 0.36 B NSB0.75; it indicates that the simulation result is credible, when NS [0.75, it indicates that the simulation When NS [0.75, the simulation results are credible; general model simulation requires NSC0.5 The SRI index is calculated by selecting the probability density function that is adapted to the runoff volume within a certain period of time, and then performing standard normalization to obtain the SRI index for the corresponding period of time. Assume that the runoff volume during the selected time period satisfies the T-distribution probability density function f(x): where b and c are shape parameters, respectively, which can be obtained by the maximum likelihood method, and the cumulative probability of runoff x for a given time period.
The SRI values can be obtained by normalizing the T distribution:

Validation of precipitation and temperature simulation based on CMIP6 model in the YRB
The validity of drought forecasts was verified based on the precipitation and temperature observations from 1961 to 2019 (Fig. 3). MMK trend test significance level is taken as a ¼ 0:05. That is, when Z À 1:96, it becomes a significant decrease and when Z ! À 1:96, it becomes a significant increase. Sen's slope reflects the slope of the series, that is, a positive value of Sen's slope is an increase, and vice versa is a decrease. The spatial distribution of annual precipitation exhibited a decreasing trend from the southeast to northwest. The low precipitation center was situated in the source region of the Yangtze River, where less than 400 mm of precipitation falls annually, on average. Sen's slope estimator's conclusion for the spatial distribution of precipitation showed a ''increasing-decreasing-increasing'' pattern from west to east, which was in good agreement with the MMK trend test. Moreover, the increased rate in the upper reaches of the YRB was less than that (between 0 and 2) in the lower reaches. At the intersection of Sichuan, Chongqing, and Guizhou, the precipitation had the lowest Sen's slope (-2.98). However, various meteorological stations in the lower and upper reaches of the YRB indicated a large rise in precipitation, while the middle region showed a significant drop through the MMK test (Fig. 3a). The spatial distribution of the annual mean temperature in the YRB also showed a decreasing trend from southeast to northwest. The highest yearly average temperature was situated in the southern portion of Jiangxi, reaching 20.12°C. The lowest yearly average annual temperature center, located in the source area of the Yangtze River, was \ -5°C. These two methods revealed an increasing temperature trend. The growth rate distribution in the YRB was slow in the middle and rapid in the surrounding areas. The Sen's slope was highest at the source of the Yangtze River, which was [ 0.04, and the lowest was 0.009 in Chongqing. Approximately 87% of the meteorological stations in the study area passed the MMK significance test, indicating that the temperature increased significantly.
To enhance the precision and scientific rigor of the CMIP6 model simulation for meteorological drought analysis in the YRB over the coming years, multi-year annual mean precipitation and temperature correlation coefficients for each model have been used. 10 models of CMIP6 and the average MME of that were combined to evaluate the simulation capacity from A to K. The correlation coefficient between each model data in the precipitation Taylor diagram was slightly smaller, ranging between 0.8 and 0.9. Among the 10 CMIP6 model data, the MIRO mode by dot H showed the highest correlation, smallest standard deviation for precipitation simulation; however, MME data represented by dot K showed the best performance. This meant that compared to other CMIP6 modes, the MME data was the best and was improved by 0.1 to more than 0.9. (Fig. 4). Same applies to temperature Taylor diagram. Therefore, we performed the future meteorological drought simulations based on the MME data (Fig. 5).
Correspondingly, we estimated the precipitation anomalies (PA) and temperature anomalies (TA) for the 10 CMIP6 models in the Shared Socioeconomic Pathways SSP1-2.6, SSP2-4.5, and SSP5-8.5 to perform spatiotemporal drought analyses. Overall, there was reasonable agreement between the observations and CMIP6 estimates from 1961 to 2014. The PA fluctuated more for all three scenarios, while the TA was continuously increased under the scenarios with higher emissions (Figs. 6a, b).

Characteristics of drought evolution in the YRB over the last 60 years
The MMK test was performed on the drought occurrence frequency events based on SPEI-12, which is calculated by monthly precipitation, maximum temperature, minimum temperature, 2 m wind speed, and sunlight hours to obtain the SPEI index for the 12-month time scale from 1961 to 2019 in the study area. It was used to investigate the spatial distribution of droughts. The overall drought frequency range in the YRB was 16.2-18.8% (Fig. 7a). Despite having a frequency range from 9.2 to 12.3%, moderate drought showed a similar spatial distribution as did total drought. The highest drought frequency occurred at the junction of the upper and middle reaches. In other regions,  the frequency of drought was similarly distributed. The severe drought frequency varied from 4.4 to 6.3% (Fig. 7c). In contrast to the remainder of the region, where there was only a differential of * 0.6%, we discovered that the junction of Yunnan-Guizhou-Sichuan experienced the highest frequency of extreme drought, at more than 1.8% (Fig. 7d). By further extending the evaluation, we obtained consistent results through Sen's slope in the variable trends of annual, spring, summer, and autumn drought. It was clear that the annual drought generally presented a 'mitigateintensify-mitigate' spatial distribution trend from west to east. Sen's slope was less than zero for the majority of the YRB, showed a trend of increasing annual drought. The Sen's slope, which was -0.001 in the southern portion of Shaanxi, was the smallest. Among them, the MMK significance test revealed that interannual drought was noticeably exacerbated at 14 meteorological stations. Both the source and the lower reaches of the YRB exhibited a trend of decreasing drought. Most of the lower reaches of the YRB showed Sen's slopes of [ 0.005, and the highest value was 0.009. In Northern part, the Sen's slope in most areas of the YRB source was [ 0.003, the extent of drought mitigation was slightly less than that in the lower reaches of the Yangtze River. The spatial distribution of the spring drought illustrated the trend that, ''the western border of the upper YRB became wet, while the east of it became dry'' (Fig. 8b). Comparable to the spring drought, the broad spatial distribution tendency of the autumn drought was similar. The Shanghai region and Central Jiangxi region showed a slowing trend, which was the most noticeable contrast from the spring drought. Furthermore, with a Sen's slope of less than -0.02, the intersection of Yunnan, Guizhou, and Sichuan had the strongest intensification. The strongest intensification was observed in the Southern Shanxi region, which showed a reading of -0.27, while the summer drought had the opposite spatial change to the spring drought. Thus, it is likely that severe drought will affect the majority of Western Yuanjiang. The central portion of Yunnan showed the strongest intensification of summer drought (Fig. 8c). Figure 9 clearly reveals that based on MME data calculation, the ratio of SPEI-12 series over the next 81 years (2020 to 2100) for future drought stations across different seasons and years. The increasing or decreasing trends of the spring drought station ratio under various conditions was the lowest among all types. Further, changes in the drought station ratio for different seasons, and interannually, were found to be synchronous under the same shared socioeconomic pathway scenarios. The ratio of various drought stations was reduced dramatically under SSP1-2.6. Both the ratio of spring and summer drought stations were steady at * 25% and 13%, respectively, before and after 2050. The autumn drought station ratio trended downward, whereas a global drought event in 2071 may occur as high as 82.9%. In contrast, the annual drought station ratio displayed a consistent downward trend and was less than 10% toward the end of the twenty-first century. Under SSP2-4.5, the seasonal and the annual ratio of drought stations did not fluctuate greatly. Under SSP5-8.5, the ratio of drought stations revealed that it will increase significantly, among which all the summer, autumn, and annual ratio of drought stations indicated that there will be continuous regional and global drought events at the end of the twenty-first century. Moreover, it appeared that as radiative forcing scenarios increased, the ratio of severe and extreme drought increased dramatically for various seasonal and annual drought events. In spring, the ratio over 20% were 3 years in the severe drought, 20.92%, 21.22%, 21.51% in the year 2055, 2065 and 2100, respectively under SSP1-2.6. While under SSP2-4.5 and SSP5-8.5, the year was extended to 4 years and the ratio was increased to 25.22%, 24.33%, 24. 93%, 21.22% in 2061, 2067, 2092 and 2093, and 24.33%, 27.33%, 28.19%, 35.46% in 2084, 2086, 2096 and 2099, respectively. The largest ratio of extreme drought in spring in 21.81% and followed by only 8.9% under SSP1-2.6. However, it was 36.36% under SSP5-8.5 and followed by 30.86%. The conditions in summer and autumn were    We simulated the spatial distribution of various drought frequency levels under different future scenarios following validation (Fig. 10). Here, two future periods were extracted, i.e., near future (NF, 2021-2050) and far future (FF, 2070(FF, -2100. We focused primarily on the evolution and spatial distribution of extreme and total drought frequency. Figure 10 shows that under SSP1-2.6 (2021-2050), the frequency of extreme droughts will range Fig. 9 Ratio of drought stations in under SSP1-2.6, SSP2-4.5 and SSP5-8.5 scenarios from 2020 to 2100 from 0.8 to 4.4%. The occurrence of extreme drought frequencies in the upper and middle reaches of the Han River basin were highest, exceeding 4%. The second highest frequency was 3%, which was concentrated in the Dongting Lake area. Central Jiangsu and Southern Jiangxi were the lowest at 1%. The frequency of intense droughts ranged from 0.04 to 3.61% from 2021 to 2050 under SSP2-4.5. The frequency spatial distribution was observed to steadily decrease from west to east. The frequency of extreme drought was greater than 2% west of the Qinghai-Tibet Plateau's southeastern edge, which might exceed 3% in some areas. The boundary between Chongqing and Guizhou had the second highest frequency. However, except for the west of the southeastern margin of the Qinghai-Tibet Plateau and the boundary between Chongqing and Guizhou, the frequency of extreme drought increases ranged from 0.14 to 2.9% from 2071 to 2100. The highest extreme in the research area was at the confluence upper and middle reaches where it was * 2.5%. Correspondingly, the northern region at the upper reaches had the highest frequency of intense droughts from 2021 to 2050 under SSP5-8.5. However, from 2071 to 2100, the frequency of drought grew considerably in other locations, ranging from 0.4 to 6.6%, with the exception of a small decrease in the upper reaches.

Projections of drought in the YRB
Internal variability (e.g., natural variability independent of external forcing), presents a third major source of uncertainty for climate change projections (Deser et al. 2012), which must be accounted for when assessing changes in drought events (Trenberth et al. 2014). We analyzed the total drought frequency under SSP1-2.6 from 2021 to 2050, when the spatial distribution gradually decreased from the center of the upper reaches to the surrounding areas. Under SSP2-4.5, the drought frequency occurring from 2071 to 2100 increased and decreased across the entire basin, typically occurring between from 5.5 to 30.2%. Compared with 2021-2050 (at 15%) it remained stable at * 10%, while the total drought occurring in the southeast edge of the Tibet Plateau increased by * 5%. Under SSP5-8.5, non-climate change policy interventions, population expansion, the slow development of scientific and technological development, and the exacerbation of fossil energy consumption caused greenhouse gas emissions to increase. Drought at the junction of upper and middle reaches increased to varying degrees, while the frequency of total drought was from between 3.6 and 45.96%. At the southeastern edge of the Tibet Plateau the frequency of total drought evidently increased even more than 45% (Fig. 11). We also achieved the prediction of spatial distribution trends under different future scenarios using MMK significant test and Sen's slope. The SPEI values passed the a ¼ 0:05 test. When the statistic Z À 1:96, showing significant drought, while under the SSP2-4.5 and SSP5-8.5 scenarios, the statistic Z ! À 1:96 and the midstream and downstream showed significant wet trend. Here, we listed Fig. 11 Evolution of spatial distribution of total drought frequency under different future scenarios only future annual drought distribution variation trends, but they had the following features. The drought grid number expanded the most with greater radiative forcing, which meant that annual droughts were more susceptible to radiative shifts than seasonal droughts. The middle reaches of the YRB had propensities to become extremely moist under various radiative situations and seasonal conditions. During the annual drought changes, the western area of the upper reaches of the YRB suffered significant drought intensification (Fig. 12).
The temporal future drought distribution is illustrated in Table 3. Generally, in the NF, the probability of the occurrence of drought first decreased then increased as the emission scenario increased. While in the FF, the probability of the occurrence of drought increased and deepened with correspondingly intensification as the emission scenario increased.
In Table 3, under SSP1-2.6, the probability of extreme drought in the NF was higher than in the FF. The probability of severe drought increased by 1.65% and 3.63% under SSP1-2.6 and SSP5-8.5 in the NF and FF, respectively, which were second to the moderate drought increases (from 10.11 to 10.21%) under SSP1-2.6, to 13.42% and 15.08% under SSP5-8.5 in the NF and FF, respectively. The probability of mild drought outweighed by 17%. Results under SSP2-4.5 were compatible with those under SSP5-8.5.

Risk of future drought events in upper and mid-lower reaches of the YRB
We initially analyzed the variations of SPEI-12 in the upper and middle-lower reaches of the YRB during the historical period. Upstream flows were primarily responsible for the water entering the downstream, where the occurrence of drought affected the inflow. From a historical perspective, upstream droughts exhibited stable fluctuations. From 1992 to the end of 2010, the upstream drought index fluctuated significantly and the SPEI index in the midstream was * -1; however, the downstream SPEI index was mostly close to -1.5 during this period. In the future, under the SSP1-2.6 scenario (2028-2035), most of SPEI indices showed negative values close between -1 and 0, while downstream, the SPEI indices were lower than -1.5, and even close to -2, with 2080 to 2090 being very harsh. Under the higher SSP2-4.5, when the negative SPEI values in the downstream were large, SPEI in the upstream in 2036-2045 were more intense, even close to -2, which represented severe drought shown in Table 2. A corresponding drought, between -0.5 and -1 also occurred in the lower reaches, but to a lesser extent than in the upper reaches. Moreover, under higher SSP5-8.5 emissions, when the upstream SPEI fluctuated in the range of -1, the downstream SPEI not only decreased to -2, but it also occurred over many years and more frequently. From 2055 to 2060, the upstream and downstream SPEI changes were inconsistent, although they were more consistent under the three different scenarios until 2030 (Fig. 13). Basin-wide droughts were further analyzed by EOF, where the contribution margins of the first spatial mode (EOF1) and second spatial mode (EOF2) were 33. respectively. Spatial values showed a negative to positive trend from upstream to downstream in EOF1 under SSP1-2.6 (Fig. 14a). The basin presented an overall negative spatial distribution to a greater degree in the upper and middle reaches (EOF1 \ -3), followed by the lower reaches under SSP1-2.6, which were consistent in reflecting basinwide drought. The distribution of the SPEI across the EOF2 region showed a consistent pattern, with positive values ([ 0.1) that were mostly concentrated in Yunnan Province. Conversely, under SSP2-4.5, the SPEI showed a regressive distribution from upstream to downstream. A saddleshaped distribution was shown in the middle-lower parts, and the low-value centre was located in Hubei Province. Under SSP5-8.5, the spatial distribution patterns were similar to that of the SSP1-2.6, with a higher spatial density consistency.
Multplied with the spatial mode (EOF), time coefficients (PCA) were selected from 2028 to 2040 under both SSP1-2.6 and SSP2-4.5 (2030-2045) under SSP5-8.5 (Fig. 15) to present the years and duration of simultaneous basin-wide drought for the Yangtze River. The duration of PCA1 increased under the high scenario and the intensity was enhanced. EOF1 showed a larger area of positive values (Fig. 14e), combined PCA1 positive values (Fig. 15e), with the duration of simultaneous drought being extended. The results of SSP5-8.5 were often more noticeable than those of SSP2-4.5 (Figs. 14c, 15c). The results for the three SSP scenarios showed bounded distinctions.
To further research on the spatial and temporal evolution of upper and mid-lower reaches drought encounters problems and provide a scientific basis for disaster prevention and mitigation decisions with respect to national or regional level policy in China. We collected DEM data, soil type, land use data related to the construction of the SWAT model, and information about the selected hydrological stations. CMIP6 data were used as input data for the SWAT model to perform runoff simulations at the Cuntan, Hankou, and Datong stations to analyze and predict the upper and mid-lower reaches of the YRB under different future scenarios.
We applied the NS (Nash-Sutcliffe coefficient) and R 2 Coefficient of determination in the upstream Cuntan station, midstream Hankou and downstream Datong on the YRB. The NS coefficients of each hydrological station during the validation period are 0.64, 0.74, and 0.76, which are above 0.6. The NS coefficient of Datong station was also the highest among the three selected stations, and the R 2 of each station is 0.83, 0.84, and 0.84. It demonstrated that the NS coefficient of the simulation results in each station all over 0.5 during the calibration and validation period and verified to be reliable. Meanwhile, the NS showed a decreasing trend from the upper reaches to the lower reaches of the YRB, which indicated that the simulation effects: Downstream of the YRB [ midstream of the YRB [ Upstream of the YRB (Fig. 16, Table 4).
In Fig. 17, we found that the total annual runoff in the upper, middle and lower reaches of the YRB would be increasing from upstream to downstream in the future, and the annual runoff changes at each station were highly consistent and similar. The runoff at Cuntan, Hankou, and Datong stations under each scenario had an increasing trend, and the runoff at each station continued to increase at the end of the twenty-first century under SSP585 scenario, while the runoff at each station decreases at the end of the twenty-first century under SSP245 and SSP216 scenarios. The annual runoff fluctuations were relatively small before 2070 under different scenarios, and the annual runoff fluctuations tended to become larger after 2070 under all scenarios. The maximum percentage difference of the adjacent annual runoff at the three stations was less than 18% for each scenario before 2070, while the percentage interpolation of the runoff from 2091 to 2092 at the Chase station under SSP585 scenario after 2070 was up to 28.5%. The specific runoff values representing the up and midlower encounter problems after 2070 can be found in Table 5.
Moreover, to further study on the spatial and temporal transfer pattern and of meteorological drought to hydrological drought in the YRB, SRI based on SWAT model to simulate the runoff values were calculated for identifying the occurrence of extreme hydrological drought events under different future scenarios, Fig. 18 showed the sequence of SRI values for the upper, mid-lower reaches of the Yangtze River basin in 2020-2100 period, there is a trend of becoming flooded in all upper, mid-lower reaches of the Yangtze River basin under different future scenarios. The trend of SRI changes in the upper, mid-lower reaches were similar, and the fluctuation of SRI before 2060 was smaller than that after 2060, which indicated that extreme hydrological droughts and extreme flooding events occur more frequently and more severely in the future under different scenarios, and had more socio-economic impacts. Figure 18 showed that extreme flooding would occur in the YRB in 2071 under the SSP245 scenario, with the severity of flooding in the middle and lower reaches of the basin being slightly higher than that in the upper reaches, and a more severe extreme hydrological drought would occur in the basin in 2060 under the SSP126 scenario.

Discussion
We demonstrated that CMIP6 could reasonably predict the future spatial distribution of drought and extreme drought events in the YRB. According to Xu et al. (2022)_ENREF_45, both the CMIP5 and CMIP6 model MME could sensibly duplicate the climate means, and the biases were generally lower in the CMIP6 models than the c Upper and lower reaches under the SSP5-8.5 scenario CMIP5 models. The CMIP6 models exhibited proficiency in simulating the extreme precipitation and temperature indices over the YRB. However, the capacity of CMIP6 in the upper and middle reaches of the YRB was still insufficiently captured. Additionally, the magnitude of drought frequency responses in the CMIP6 projections depended on the region and season, which was consistent with earlier studies (Cook et al. 2020) that emphasized drought metrics. Scenario uncertainty outweighed model uncertainty from mid-century onwards (Hawkins and Sutton 2011), which would experience more droughts occurring in the Southeast Asian region under SSP1-2.6, SSP2-4.5, and SSP5-8.5 (Li et al. 2021a, b, c;Zeng et al. 2022). Overall, our results aligned with those previous studies based on CMIP6 model projections. However, attributions of seasonal drought changes were not mentioned in this research. Our analysis will focus further on drought from seasonable characteristics.
The MME technique has been considered as a beneficial strategy for enhancing weather and climate forecasting to reduce uncertainties (Feng et al. 2011;Sun et al. 2015). The Taylor plots study of prediction and temperature precision indicated that the dependability of MME was higher than that of other methods. The northwest-southeast fluctuation can be adequately reproduced using the CMIP6 MME model. While considerably better than any one model, the data trend that emerged via MME projections still had sizable deviations, which were not thorough explainable. We used these data to estimate the frequency and spatial distribution of drought. (Jiang et al. 2022) suggested that quantitative future climate change records with high credibility generated by robust GCMs merged datasets from CMIP6 were scarce. This indicated that more possibilities needed to be added to the model projections for future droughts, and the impact factors on regional downscaling need to be further investigated.  -4.5 (2028-2040), e first spatial mode and f second spatial mode under SSP5-8.5 (2030SSP5-8.5 ( -2045 SPEI-12 was the main assessment method for drought variations, as SPEI values over different timescales reflect complex and long-term variations in drought conditions. In our study, different time scales of SPEI were not compared. SPEI-3. SPEI-6 and SPEI-12 these different time scales of drought factors will be more detailed for seasonal and upstream and downstream droughts in the basin. The drought intensity, distribution trend, and duration were also well demonstrated in this study, particularly for the prediction of the simultaneous occurrences of YRB drought events under increased emissions scenarios. Other impacts on SPEI-12 were not addressed. (Li et al. 2021a, b, c) calculated and analyzed the correlation coefficients between the SPEI-12 and ENSO indices of each station. According to SPEI, the middle reaches of the LMRB are forecast to have the greatest increases in drought duration and intensity in the dry season, while the upper Lancang and middle Mekong rivers are projected to experience longer drought durations and greater drought intensity in the wet season (Dong et al. 2022). We simulated the ratio of drought stations for different future periods under the different scenarios and found that the duration of consecutive droughts increased. Distinguishing wet and dry seasons will provide more reflections for thought for subsequent studies. Moreover, uneven drought conditions exist in the YRB, North Atlantic Oscillation (NAO) and Northern Oscillation Index (NOI) can be utilized as early warning indicators for monitoring drought events (Huang et al. 2018). And the lack of circulation influence factors in this study will be added to the subsequent related studies. Fig. 15 Empirical orthogonal decomposition of SPEI of the entire Yangtze basin a first temporal mode and b second temporal mode under SSP1-2.6 (2028-2040), c first temporal mode and d second temporal mode under SSP2-4.5 (2028SSP2-4.5 ( -2040, e first temporal mode and f second temporal mode under SSP5-8.5 (2030SSP5-8.5 ( -2045 Additionally, this study showed that the lower reaches of the YRB in Hubei and Jiangsu are predicted to experience continuous prolonged droughts from 2030 to 2040. Emission scenario increments will intensify simultaneous droughts across the entire YRB. However, the impacts of human activities were not considered separately. Dam building and urban construction in the upper and middle reaches of the YRB these factors on drought have not been considered in this study. The construction of the Three Gorges Dam induced more humidity to the climate of the lower reaches of the basin (Li et al. 2021a, b, c), which was contrary to our findings that the intensity of future drought events in the lower YRB will be enhanced. This also provides ideas for subsequent investigations. The impacts of anthropogenic activities on hydrological processes continue to change, which results in disordered stochastic changes in streamflows and droughts (Zhang et al. 2018). Considering the unbalanced distribution of hydraulic engineering facilities and spatially heterogeneous changes in future drought events, we provided a warning background of drought for other future development changes in the Yangtze River Basin, which is useful for strategic planning development in different regions of the Yangtze River Basin. But specific quantitative analysis on practices and implications with respect to national or regional level policy in China is lack in this study, additional studies will be required to explore how influencing factors precisely impact regional variations in droughts for the YRB region and specific quantitative analysis to national policies under future climate changes.
Uncertainties exist in climate simulations and future climate change projections, which have often been a limiting factor, particularly at local and regional scales (Knutti and Sedláček 2013). Consequently, future climate change projections should be interpreted with caution. Ensemble

Conclusions
For this study, we explored future drought projections using SPEI for the YRB, based on the MME derived from the average of 10 CMIP6 models and historical observations data under three climate change scenarios. We emphasized spatiotemporal patterns and comparative variations of drought events in the upper, and middle-lower reaches of the YRB, resulting in the following conclusions: (1) The average of 10 CMIP6 models showed a good verification of historical precipitation and temperature for drought predictions. The MME model accurately replicated historical drought spatiotemporal changes. The MMK and Sen's slope demonstrated consistency for historical droughts. From a historical perspective, the middle reaches of the YRB experienced intensifying drought frequency. (2) In the future (2020-2100), the higher emission signifies higher moderate and total drought frequency, intensity, and scope of the YRB in FF, lower in NF. The ratio of autumn severe and extreme droughts would increase in mid-twenty-first century. Compared with the SSP1-2.6 scenario, under SSP5-8.5 scenario, the probability of the occurrence of moderate drought increased from 10.11 to 15.08%, and extreme drought events varied from 5.10 to 8.46%. In the NF, the total drought frequency was relatively high under the SSP1-2.6 scenario, remaining in the range of from 15 to 25%, while under the SSP5-8.5 scenario, the total drought frequency was higher in the FF, with a frequency of over 35% in the Sichuan basin. The frequency of extreme droughts will remain at * 4%-5% under the SSP5-8.5 scenario.
(3) Severe drought risk encounters were projected in the upper and meanwhile in the middle-lower reaches in YRB, especially in the 2030-2040 period. Under all three scenarios, severe droughts occurred more frequently with SPEI close to -2. Under the SSP5-8.5 scenario, there was a basin-wide distribution drought and a trend of increasing drought from the upper to the lower reaches. The middle reaches indicated a saddle-shaped distribution, which attained a maximum in the lower reaches, while Sichuan Basin was the sub-peak. (4) The future runoff in the YRB during the dry period varied less, but in May and June during the main flood season the runoff under SSP1-2.6 would be the largest. The mid-lower reaches under the SSP2-4.5 scenario would experience the greatest runoff reduction during the occurrence of region-wide drought, maximum decrease in runoff would be 2045, reaching 13.9%. Extreme flooding events and extreme meteorological droughts would happen accompanying with hydrological droughts would occur more frequently and severely under different scenarios.