Study on the Impact of Future Climate Change on Extreme Meteorological and Hydrological Elements in the Upper Reaches of the Minjiang River

,


Introduction
Global warming will aggravate the global hydrological cycle and increase global average precipitation and evaporation [1]. Simultaneously, precipitation variability may change, exerting direct efects on evaporation, runof, and soil humidity. Te extreme hydrological events, such as foods and droughts, increase the risk of water disasters, which have become major challenges to human survival [2]. Under the background of climate warming, extreme climate and hydrological events occur frequently in China [3][4][5][6][7][8][9][10]. Te change in extreme climate events and the impact of change in hydrological and water resources caused by extreme climate events on human life and production have attracted increasing attention [8,11]. Research on the changing trend, occurrence mechanism, response, and prediction of hydrological extreme events in river basins under the background of climate change is important to understand. It is signifcant for scientifcally understanding the spatiotemporal evolution law of the land-water cycle under the background of global climate change. It has an important application value in basin food control planning and design, large-scale hydropower development planning and operation management, regional disaster prevention and reduction, ecological environment protection, and sustainable economic and social development.
Te study of the impact of climate change on water resources focuses on the analysis of the water cycle and water resources trends under climate change [12]. Trends in water resources under climate change frst need to provide more accurate climate change scenario data for hydrological simulation. Climate change scenarios are based on several scientifc assumptions and use a long series of historical information to establish a continuous and consistent predictive description of future climate including temperature, precipitation, and other elements. Currently, the GCM (Global Climate Model) output method for developing future climate change scenario design is most widely used, but the basin-scale hydrological models do not match the large grid data of GCMs. So, downscaling of future climate change scenarios is needed [13,14]. Te main types of downscaling methods are statistical downscaling, dynamical downscaling, and combined statistical-dynamical downscaling. Statistical downscaling has been increasingly used to predict future climate scenarios such as precipitation and temperature due to the lower operational resource requirements and higher accuracy [15]. Statistical downscaling refers to the use of historical observations to establish statistical relationships between regional observations and large-scale climate elements and to verify such statistical relationships using other regional observations. Te frequently used statistical downscaling models are ASD (automated statistical downscaling) and SDSM (statistical downscaling model), among which SDSM has less bias and is convenient to use and is one of the most widely used statistical downscaling models [16]. Gulacha et al. [17] assessed that the model and the observed showed a good ft in the Wami-Ruvu River Basin of Tanzania, and the SDSM's R 2 values between raw and model for temperature ranged from 0.42 to 0.98. Tukimat et al. [18] found that the SDSM successfully provided long-term climate pattern at the gauged stations with an R value close to 1.0 in Kuantan River Basin of Malaysia. Dehghan et al. [19] assessed that there is a good agreement between the simulated and the observed precipitation, and R 2 for precipitation is greater than 0.63.
Research shows that the impact of climate change on the hydrological cycle is regional [20][21][22][23][24][25][26][27][28][29][30]. Pandey et al. [20] assessed that there is a high consensus for increase in temperature but higher uncertainty with respect to precipitations in Mahakali of Nepal. Under the projected changes, the average annual streamfow was simulated to increase gradually from the near to far future under both RCPs (Representative Concentration Pathways). Bajracharya et al. [21] found that the extreme projection of a RCP 8.5 scenario shows that the average annual temperature of the basin is expected to increase by more than 4°C in the Kaligandaki Basin of Nepal. Likewise, the average annual precipitation in the basin is projected to increase by as much as 26% during the late century under a RCP 8.5 scenario. Te synergetic efect of an increase in temperature and precipitation shows the aggravated efect on the discharge and water yield with an increase of more than 50% at the outlet of the basin. Fonseca and Santos [22] simulated the potential efects of climate change on the hydrology in the Tâmega River Basin, Northern Portugal, experiencing a Mediterranean climate. Te annual precipitation over the Tâmega River Basin exhibits weakly decreasing trends across the entire future period. On the other hand, temperatures show consistently warming trends throughout the basin for the future period, with a mean warming rate of 0.03°C per year. As a result, the mean annual fow rate decreased at all hydrometric stations by about 0.25 m 3 s −1 per year, with increased fow rates in winter when compared to the historical period but signifcantly lower fow rates in summer. Meaurio et al. [23] assessed the climate change impact on river discharge in the Bay of Biscay, Spain. It was found that trends for extreme fows show an increase in the duration of low fows.
In China, Yu et al. [24] pointed out that while the precipitation in Northern China is decreasing, the change in precipitation exhibits the form of "waterlogging in the south and drought in the north." Te probability of food disaster in the Yangtze River Basin is considerably higher than that in other regions of China. Since the 20 th century, more than 20 foods have occurred in the Yangtze River Basin. Among these, the foods of 1905, 1931, 1954, 1988, 2010, and 2017 were the most severe [24]. Wang et al. [25] reported that the concentrated distribution and high frequency of large-area rainstorms and foods occur in the monsoon region of Eastern China. Meanwhile, in Western China's arid and semiarid regions, disastrous foods are mostly caused by short-term local rainstorms, and small and medium-sized rivers can form high peak fow, causing serious disasters to local regions.
Te Minjiang River Basin is a frst-class tributary in the upper reaches of the Yangtze River. It is located to Southwest China and is in the southeast edge of the Qinghai-Tibet Plateau. Under the background of the large terrain of the Qinghai-Tibet Plateau and the comprehensive infuences of the large, medium, and small complex terrain, the valley is steep, hydropower energy is abundant, and weather and climate disasters, such as rainstorms and mountain torrents, occur frequently. Tis area is sensitive to climate change and extreme hydrometeorological events. Previous studies on the Minjiang River Basin focused on the response of this basin to climate change. For example, Liang et al. [26] found that the temporal variation characteristic of the upper reaches of the Minjiang River is as follows: annual average precipitation exhibits a downward trend due to the reduction in summer precipitation. Meanwhile, the spatial distribution characteristic is that high-altitude areas exhibit an increasing trend, whereas low-altitude areas present a decreasing trend. Te annual runof in the upper reaches of the Minjiang River shows a signifcant downward trend from 1937 to 2018, and it may continue to demonstrate a downward trend in the future. Huang et al. [27] found that the average temperature in the upper reaches of the Minjiang River shows an upward trend, whereas precipitation and annual runof exhibit a downward trend. Te trend of the average temperature has an evident positive correlation with precipitation, particularly in spring and autumn. Te decrease in water infow during spring and the extension of the duration of the low-fow season during autumn will considerably impact irrigation and urban water supply. Huang et al. [28] demonstrated that climate change scenario analysis combined with the Soil and Water Assessment Tool (SWAT) hydrological model can efectively simulate the efect of climate change on runof. Te infuence of precipitation change on runof is greater than that of temperature change. Te impact of temperature change on runof is more evident in dry years than in wet years. Chen et al. [29] showed that the overall temperature in the Minjiang River Basin exhibited a trend of decreasing the number of extreme cold days while increasing the number of extreme warm days. In terms of spatial distribution, the high value of the extreme cold event index in the basin was mostly recorded in the upstream. Te spatial distribution of extreme precipitation indicators in the Minjiang River Basin is highly uneven, as manifested in the high value of extreme precipitation indicators mostly appearing in the middle and lower reaches of the basin. From the analysis of the changing trend, the average characteristics of the Minjiang River Basin range from shortterm to sustained extreme precipitation [30].
In summary, research on the change in extreme climate events in the upper reaches of the Minjiang River mostly used historical data to analyze temperature and precipitation extreme events. Meanwhile, there is less research on future changes in extreme events, particularly hydrological extreme events. In the current study, the hydrological model (i.e., SWAT) is used to simulate runof change in the upper reaches of the Minjiang River under a future climate scenario. Simultaneously, the change characteristics of extreme climate hydrological elements in the upper reaches of the Minjiang River under a future climate scenario are analyzed using extreme climate and runof indices. Te study analyzes the future trends of hydrometeorological element extremes under climate scenarios, providing useful support for theoretical studies of potential drought and food threats in the upper reaches of the Minjiang River, as well as a reference for future water conservancy project design in the upper reaches of the Minjiang River.

Study Area and Data Sources.
Te upper reaches of the Minjiang River Basin are located between 102°59′-104°14′ E and 26°33′-33°16′ N and have a drainage area of about 23,000 km 2 ( Figure 1). Te upper reaches of the Minjiang River are sensitive to climate change and frequent natural disasters due to the high intensity of water resource development, the reduction of forest coverage, the degradation of ecological functions, and serious soil and water losses [29,30].
Te Digital Elevation Model (DEM) data in this study were obtained from the geospatial data cloud (https://www. gscloud.cn/search) with a resolution of 1 km (Figure 2(a)). Land use data were obtained from the International Geosphere-Biosphere Programme (IGBP), which uses the United States Geological Survey (USGS) method to classify land use into 17 categories. Te land use types of the Minjiang River were mainly classifed into 7 categories ( Figure 2(b)). Soil data are from the Harmonized World Soil Database (HWSD) at 1 km resolution and can be downloaded from the Food and Agriculture Organization of the United Nations. SPAW was used to calculate the parameters required for the SWAT soil database, which contains 21 soil types in the assessment area of this study (Figure 2(c) and Table 1).

Methodology
Tis study analyzes the impact of future climate change on extreme hydrometeorological elements in the upper reaches of the Minjiang River. Te statistical downscaling model (SDSM) is selected for future climate scenario analysis, and the SWAT model is used for hydrological simulation. When simulating precipitation, the large-scale climate forecaster is used to frst simulate the probability of precipitation on a particular day and then the amount of precipitation on that rainy day: where W i and W i−1 represent the precipitation probability of days i and i − 1, respectively, and x j is the j-th predictor and regression coefcient. Te occurrence of precipitation is determined by a random number r (0 ≤ r ≤ 1) that follows a uniform distribution. If r ≤ W i , then precipitation will occur on that day. When precipitation occurs on a certain day, a multiple exponential regression function will be used to simulate precipitation on that day: where R i is precipitation on the i-th day, β 0 and β j are the regression coefcients, x j is the prediction factor on the j-th day, and ε i is the error. Te major steps in using SDSM include quality control, downscaling prediction factor screening, model correction, weather generator, and model evaluation. Te original meteorological observation data collected from meteorological stations may be missing. Terefore, quality control should be implemented to identify missing data, outliers, and suspicious incomplete data, improving the quality of model output [6,31].
Referring to Wilby [31], the prediction factors of temperature and precipitation (Table 2) and rainfall station forecast factors are selected.

SWAT Model.
Te SWAT model is a semidistributed watershed hydrological model developed by the United States Department of Agriculture-Agricultural Research Service (USDA-ARS). Te watershed delineation tool in SWAT delineates the whole study basin into several subbasins in accordance with the characteristics of topographic factors and river network distribution. On this basis, hydrological response units are divided in accordance with the land use type, soil type, and slope area threshold of the basin. Runof is calculated separately. Finally, the total runof of an outlet section is obtained through river confuence routing.
In SUFI-2, parameter uncertainty accounts for all sources of uncertainties such as uncertainty in driving variables (e.g., rainfall), conceptual model, parameters, and measured data. Te degree to which all uncertainties are accounted for is quantifed by a measure referred to as the P factor, which is the percentage of measured data bracketed by the 95% prediction uncertainty (95PPU). As all the processes and model inputs such as rainfall and temperature distributions are correctly manifested in the model output (which is measured with some error)-the degree to which we cannot account for the measurements-the model is in error and is hence uncertain in its prediction. Terefore, the percentage of data captured (bracketed) by the prediction uncertainty is a good measure to assess the strength of our uncertainty analysis. Te 95PPU is calculated at the 2.5% and 97.5% levels of the cumulative distribution of an output variable obtained through Latin hypercube sampling, disallowing 5% of the very bad simulations. As all forms of uncertainties are refected in the measured variables (e.g., discharge), the parameter uncertainties generating the 95PPU account for all uncertainties. Reducing the total uncertainty into its various components is highly interesting but quite difcult to do, and to the best of the authors' knowledge, no reliable procedure yet exists.

SWAT Model Development.
In this study, the 1969-1980 is the calibration period and 1981-1987 is the verifcation period. Te Nash-Sutclife model efciency coefcient (NSE) and the coefcient of determination (R 2 ) are selected as indices for evaluating the simulation efect of daily runof in the upper reaches of the Minjiang River. Simultaneously, the simulation efect of the SWAT model on extreme high and low fows is evaluated using the correlation coefcient.

Extreme Climate and Flow Indicators.
Te extreme temperature and precipitation indicators [37] recommended by the World Meteorological Organization are selected in this study. Tese indicators can refect the intensity, frequency, and duration of extreme temperature and precipitation elements [38]. A detailed introduction of these indicators is provided in Table 4. Tis study also selects extreme runof indicators to refect changes in extreme  represent low-fow extreme runof, while Q99, Q95, and Q90 represent extreme high-fow runof [39]. Simultaneously, this study calculates the hydrological frequency of extreme high and low fows under a future climate scenario and the historical period, in which the annual maximum daily fow can represent extreme high fow while the annual minimum monthly fow can represent extreme low fow [38]. however, it lacks regional climate information. Te large-scale and low-resolution variables of CAN-ESM2 simulation are reduced to a regional scale by SDSM. Te analysis shows that the correlation coefcient between the simulated and measured daily maximum and minimum temperatures is about 0.9, while the correlation coefcient between the simulated and measured daily precipitation is between 0.4 and 0.5 (Table 5), indicating that analyzing extreme meteorological elements is feasible by using temperature and precipitation data after SDSM downscaling.

Results and Discussion
Te average absolute error between the simulation and actual measurements during diferent seasons in the verifcation period is calculated (Table 6). Te average absolute error of the highest and lowest temperatures exhibits minimal seasonal variation, and the daily average absolute error of precipitation in summer is higher than that in other seasons. In general, using SDSM for downscaling research in the upper reaches of the Minjiang River is feasible.

Simulation Results of SWAT.
Te SWAT model was used to simulate the daily runof in the upper Minjiang River, and the optimal values of the model parameters are shown in Table 7. Te results showed that R 2 was 0.87 and the NSE was 0.86 in the calibration period, R 2 was 0.79, and the NSE was 0.77 in the validation period, indicating that it is feasible to simulate the runof of the upper Minjiang River using the SWAT model ( Figure 3). From the simulation results, it can be seen that the SWAT model is able to simulate the seasonal distribution characteristics of runof. SWAT simulates runof more accurately in the food season, and the model simulates slightly lower runof in the dry season, which may be related to the greater contribution of high fows in the food season to the simulation error evaluation [7,8].
Te simulated efects of extreme high fows and extreme low fows in the upper Min River from 1969 to 1987 simulated by SWAT were calculated, where the annual maximum daily fows were used to represent extreme high fows, and the annual minimum monthly fows were used to represent extreme low fows. Te simulation results show that the correlation coefcient between the simulated and measured annual maximum 1-day (AM) fow is 0.75, and the correlation coefcient between the simulated and measured minimum monthly fow (IM) is 0.82, indicating that the SWAT model for the upper Minjiang River can be used to carry out studies of extreme runof.

Variation Characteristics of Temperature, Precipitation,
and Runof under Future Scenarios. As shown in Figure 4, the monthly average temperature rises by 2°C to 3°C. Temperature increase is most evident from June to August in summer and December to January in winter. Comparing the three scenarios, temperature rise is most apparent in the RCP8.5 scenario. Precipitation decreases from April to May in spring compared with the historical period, and it increases in other months compared with the historical period, particularly from July to August in summer, with an increase of 27-40 mm. Te increase in runof is most evident from November to April of the following year, while runof from June to July is less than that in the historical period, i.e., more than 10% less than that in the historical period. From June to July in summer, temperature and precipitation increase, whereas runof decreases, indicating that the rising range of precipitation cannot compensate for the impact of the rising evapotranspiration caused by increasing temperature. From April to May in spring, temperature and runof increase,  whereas precipitation decreases, indicating that the impact of temperature increase on the snow-melting process is greater than that on evapotranspiration in spring. Te increase in melting snow amount increases runof. Te decrease in runof from June to July in summer will afect agricultural irrigation water and may lead to agricultural production reduction. It may also afect downstream water diversion and the ecological environment. Table 8 provides the comparison of extreme temperature index values in the upper reaches of the Minjiang River between the historical period and future climate scenarios. Tx10p and Tn10p mostly increase, particularly in the RCP8.5 scenario. Te increase in Tn90p and Tx90p is more evident in the future climate scenario, particularly in the RCP2.6 scenario. Tese fndings show that the relative indicators in the upper reaches of the Minjiang River are largely increasing in the future climate scenario, in which the increase of warm indicators is more evident. Meanwhile, the increase of cold indicators is more apparent in the RCP8.5 scenario, and the increase of warm indicators is more noticeable in RCP2.6.

Extreme Indices (TX n , TN n , TX x , and TN x )
. TX n and TN n in the upper reaches of the basin will rise under the future climate scenario, particularly in RCP8. 5. TX x and TN x will rise in the future climate scenario. Te rise of TX x is most evident in the RCP8.5 scenario, while the rise of TN x is most apparent in the RCP2.6 scenario. Tese fndings show that among the extreme value indices in the upper reaches of the Minjiang River, the cold index decreases, whereas the warm index increases, and the change is more evident in the RCP8.5 scenario. Table 9 presents the comparison of extreme precipitation index values in the upper reaches of the Minjiang River in the historical period and future climate scenario. Compared with the historical period, CWD increased in the future climate scenario, with RCP4.5 increasing most signifcantly in the Maoxian station and RCP8.5 increasing most signifcantly in the other stations. Tese fndings show that continuous precipitation in the upper reaches of the Minjiang River will further increase under the future climate scenario. (R10 mm and R25 mm). Under the future climate scenario, R25 mm largely decreases, particularly in the Songpan, Wenchuan, and Lixian stations. R10 mm is largely rising, particularly in the RCP8.5 scenario. Tis result shows that under the future climate scenario, the frequency of strong precipitation in the upper reaches of the Minjiang River will increase, while the frequency of heavy precipitation will decrease.

Absolute Value Indices (RX1DAY and RX5DAY).
Under the future climate scenario, RX1DAY decreases, particularly in RCP2.6 and RCP4.5, whereas RX5DAY increases, particularly in RCP8.5. Tis result shows that the intensity of continuous heavy precipitation in the upper reaches of the Minjiang River increases, whereas the intensity of short-term heavy precipitation decreases.

Analysis of the Variation Characteristics of the Extreme
Runof Index Value. Table 10 provides the change rate of extreme runof in the upper reaches of the Minjiang River relative to the historical period under the future climate scenario. Under the future climate scenario, AM in the upper reaches of the Minjiang River exhibits a downward trend, i.e., a reduction of 32%-42%. Among these, the decrease in Heishui station is the most evident, followed by that in Zipingpu station. When diferent future climate scenarios are compared, RCP4.5 presents the most noticeable downward trend, followed by RCP2.6. Q10, Q5, and Q1, which represent the extreme runof of low fow and exhibit increasing trends, indicating that the risk of drought in the upper reaches of the Minjiang River will be weakened under diferent discharge scenarios in the future. Q90 and Q95 of Heishui and Shaba stations show an increasing trend, whereas Q99 shows a decreasing trend. Q90 of Zipingpu station presents a weak upward trend (i.e., less than 5%), and Q95 and Q99 exhibit a downward trend. Tese results show that the overall food risk in the upper reaches of the Minjiang River is weakened. However, the spatial distribution in the upper reaches is diferent, and food risk in the upstream source area demonstrates an increasing trend. Figure 5 shows the changes in Q90, Q95, and Q99 at Zipingpu station from May to October under the historical period and future climate scenario. Te overall characteristics of Q90, Q95, and Q99 that refect the peak runof are as follows: they decrease from June to July, decrease weakly in May, and increase in other months. Combined with the runof wet season in the upper reaches of the Minjiang River from May to October, extreme food events from June to July  Hydrological frequency is analyzed using annual maximum daily runof under the future climate scenario and the historical period. As indicated in Figure 6 and Table 11, the Monte Carlo simulation [28] shows that the optimal frequency distribution line of RCP2.6 and RCP4.5 in the historical period is generalized logic distribution (GLO), while that of RCP8.5 is Wakeby distribution.
By using the optimal frequency distribution, the annual maximum daily runof under diferent discharge scenarios is calculated to be reduced compared with that of the historical period. Te annual maximum daily runof with a 20-year return period under RCP2.6, RCP4.5, and RCP8.5 discharge scenarios is reduced by 47.3%, 45.9%, and 43.3%, respectively, compared with Table 5: Correlation coefcients of measured and simulated daily maximum and minimum temperature and precipitation during the verifcation period.

Station
Maximum temperature Minimum temperature Precipitation Maoxian 0.9 * * * 0.9 * * * 0.4 * * Songpan 0.9 * * * 0.8 * * * 0.4 * * Wenchuan 0.9 * * * 0.9 * * * 0.5 * * Lixian 0.9 * * * 0.9 * * * 0.5 * * Heishui 0.9 * * * 0.9 * * * 0.5 * * * * indicates passing the 95% signifcance test; * * * indicates passing the 99% signifcance test.  r_ means an existing parameter value is multiplied by (1þ given value); v_ means the existing parameter value is to be replaced by given value; the detailed description of the parameters was introduced as technical report [36,40]. 10 Advances in Meteorology    Advances in Meteorology   Advances in Meteorology 13 that of the historical period. Meanwhile, the annual maximum daily runof with a 100-year return period is reduced by 54.8%, 52.2%, and 50.6%, respectively, compared with that of the historical period. Tis result shows that the risk of food under diferent discharge scenarios is lower than that in the historical period. Hydrological frequency is analyzed using annual minimum monthly runof under the future climate scenario and the historical period. As indicated in Figure 7 and Table 12, Monte Carlo simulation [28] is used to determine that the optimal frequency distribution line of RCP2.6 and RCP8.5 in the historical period is generalized logistic distribution (GLO), while that of RCP4.5 is generalized extreme value distribution (GEV).
Under diferent discharge scenarios, the annual minimum monthly runof increases compared to historical period. Under RCP2.6, RCP4.5, and RCP8.5 discharge scenarios, the annual minimum monthly runof with a 20-year return period increases by 30.9%, 28.8%, and 45.2%, respectively, compared with that of the historical period. Meanwhile, the annual minimum monthly runof with a 100-year return period increases by 26.6%, 14.1%, and 47.0%, respectively, compared with that of the historical period. Terefore, the risk of drought in the upper reaches of the Minjiang River is reduced under diferent emission scenarios.

Results and Discussion
7.1. Discussion. Te research shows that precipitation from April to May in spring in the upper reaches of the Minjiang River is lower than that in the historical period, and the duration of the low-fow season is prolonged due to the reduction of water infow in spring, which will afect       irrigation and urban water supply. In the future, temperature and precipitation will increase, whereas runof will decrease from June to July in summer. Meanwhile, precipitation will decrease and runof will increase from April to May in spring. Several previous studies have shown that runof change exhibits a signifcant positive correlation with precipitation, and the efect of temperature on runof is more evident in dry years than in wet years [27]. Tis fnding indicates that the efects of temperature on snow melting in spring and evapotranspiration in summer will further increase in the upper reaches of the Minjiang River under future climate change.
Under the future climate scenario, the cold index frequency of extreme temperature increases and intensity decreases. In accordance with the fact that extreme cold events in the historical period mostly occur in the upper reaches [19], extreme cold events in the upper reaches of the Minjiang River will further increase in the future. Te intensity of continuous heavy precipitation under the future climate scenario will increase, and this fnding is consistent with the overall change trend of the Minjiang River Basin in the historical period [20]. It indicates that continuous extreme precipitation events in the upper reaches of the Minjiang River will further increase in the future.
Te research shows that the occurrence of food events is typically related to the occurrence of short-term extreme heavy precipitation, and it is similar to the results of previous studies [7]. In the future scenario, the intensity and frequency of short-term heavy precipitation decrease, reducing the risk of extreme food. Te research shows that drought risk is related to precipitation during the dry season. In the future climate scenario, an increase in precipitation during the dry season in the upper reaches of the Minjiang River reduces the risk of drought in that area.
At present, research on the impact of climate change on industries is mostly based on climate and hydrological models, including some uncertainties [8,11], such as the simplifcation of physical processes by climate and hydrological models, the introduction of parametric processes by climate models, and future climate scenario assumptions. In consideration of the aforementioned characteristics, the current study selects three scenarios (low, medium, and high) in the typical concentration path recommended in the ffth assessment report of the Intergovernmental Panel on Climate Change (IPCC) that can refect the degree of radiation force under diferent emission reduction strategies in the future. In the next step, multiple climate models will be considered for uncertainty analysis.

Results.
Under the future climate scenario in the upper reaches of the Minjiang River, the frequency and intensity of the extreme temperature warming index will increase while those of the cold index will increase and weaken.
Under the future climate scenario in the upper reaches of the Minjiang River, the duration of precipitation, the intensity of continuous heavy precipitation, and the frequency of strong precipitation increase, whereas the intensity of short-term heavy precipitation and the frequency of heavy precipitation decrease.
Monte Carlo simulation is used to determine that for the RCP2.6 and RCP4.5 climate scenarios under the historical period, the optimal frequency distribution line of annual maximum daily fow is GLO, while that for the RCP8.5 climate scenario is Wakeby distribution. By using the optimal frequency distribution line, the annual maximum daily fow under diferent climate scenarios is calculated to be lower than that in the historical period, indicating that the risk of food in the upper reaches of the Minjiang River under diferent climate scenarios is lower than that in the historical period.
Monte Carlo simulation is used to determine that for the RCP2.6 and RCP8.5 climate scenarios under the historical period, the optimal frequency distribution line of annual minimum monthly fow is GLO, while that for RCP4.5 is GEV. By using the optimal distribution line, the annual minimum monthly fow under diferent climate scenarios is determined to increase compared with that in the historical period, indicating that the risk of drought in the upper reaches of the Minjiang River under diferent climate scenarios is reduced.
Te indicators that represent low fow runof exhibit an increasing trend, indicating that drought risk in the upper reaches of the Minjiang River is reduced. Te indicators that represent high water runof mostly demonstrate an increasing trend at Heishui and Shaba stations, while those at Zipingpu station present a decreasing trend, indicating that the overall food risk in the upper reaches of the Minjiang River is reduced. Meanwhile, spatial distribution in the upper reaches is diferent, and thus food risk in the upstream source area still exhibits an increasing trend.
Te distribution characteristics of the indicators that represent high water runof in the year show that extreme runof in the upper reaches of the Minjiang River will present an upward trend from August to October in the future. Particular attention should be given to the increase in autumn food risk in the upper reaches of the Minjiang River.

Data Availability
Te data used in this paper were provided by the State Meteorological Administration of China. Relevant station data are available at https://data.cma.cn/.