Historical Evolution and Future Scenario Prediction of Hydrological Drought in the Upper Reaches of Xin’an River

Abstract

Predicting future hydrological drought characteristics can assist relevant departments in taking proactive measures to mitigate drought losses. Based on the SWAT model and the Sixth International Coupled Model Comparison Program, this study employs an improved Mann–Kendall test, cumulative anomaly method, and continuous wavelet transform to investigate future runoff and hydrological drought characteristics in the upper reaches of the Xin’an River under different Shared Socioeconomic Pathways (SSPs). The SSPs scenario consists of three typical paths. SSP126 represents the sustainable development path (low carbon emissions, ecological protection first), SSP245 is the intermediate balance path (equal emphasis on economic growth and environmental protection), and SSP585 is the fossil fuel-intensive path (high emissions, high development intensity). The results indicate that from 2000 to 2020, under the influence of ecological compensation policies, the upper reaches of the Xin’an River transitioned from hydrological drought to hydrological wetness in 2012. Under the three future scenarios, runoff volumes increased by 41.72%, 40.74%, and 40.72% compared to the historical period, respectively, with peak runoff occurring in May, June, and July, alleviating hydrological drought conditions. Under the SSP245 and SSP585 scenarios, drought characteristics were more pronounced, with the number of drought-free months increasing by 21 and 30 months, respectively, compared to the SSP126 scenario, and the number of extremely dry months increased by 9 months and 17 months, respectively. The standard runoff index in the SSP126 scenario exhibits two oscillation cycles of 400 months and 359 months, respectively, while SSP245 and SSP585 both exhibit an oscillation cycle of 835 months. After discussion, it was concluded that ecological compensation policies can improve hydrological drought conditions. Drought characteristics become increasingly pronounced as carbon emissions intensify. This research can provide theoretical references for water allocation and drought prevention in river basins.

1. Introduction

Hydrological drought refers to a natural phenomenon in which the hydrological cycle is adversely affected due to factors such as reduced precipitation, decreased river flow, lowered lake water levels, and reduced groundwater levels over a specific time scale. Global warming is the primary cause of drought. Since the mid-20th century, global temperatures have risen at an average rate of approximately 0.15 °C per decade, while China’s temperature increase has reached 0.26 °C per decade. Data released by the European Union’s Copernicus Climate Change Service indicates that 2024 was the hottest year on record since 1850. The Secretariat of the United Nations Convention to Combat Desertification has pointed out that 15% to 20% of China’s population will face more frequent moderate to severe droughts by the end of this century. According to a report by China’s National Emergency Management Department, drought disasters in 2024 affected nearly 11.49 million people nationwide, with crop-affected areas reaching 1,206,000 hectares and direct economic losses amounting to 8.4 billion yuan. The Chinese government has repeatedly expressed its concern about climate change. In 2013, the Chinese government issued the National Strategy for Adapting to Climate Change. In 2020, the Chinese government announced its carbon emission commitments at the United Nations General Assembly. In 2022, the Chinese government submitted the Progress Report on the Implementation of China’s Nationally Determined Contribution Targets to the United Nations. At the Third Plenary Session of the 20th Central Committee of the Communist Party of China in 2024, it was once again emphasized that China must actively address climate change and steadily advance the goals of carbon peaking and carbon neutrality. The Xin’an River originates in Huangshan City, Anhui Province, flows eastward to Chun’an County, Zhejiang Province, and empties into the Thousand Island Lake Reservoir, providing water and power resources to multiple provinces and cities in the East China region. Hydrological droughts in the upper reaches of the Xin’an River directly impact the power supply and water security of the East China region. Analyzing the runoff and hydrological drought trends in the upper reaches of the Xin’an River is not only crucial for ensuring the water resource supply security of Huangshan City but also contributes to safeguarding the long-term water and power supply of the East China region.

Droughts are generally categorized into meteorological, hydrological, agricultural and socio-economic droughts, depending on their causes. Hydrological droughts are directly related to the availability and utilization of water resources, and are the development and final result of meteorological and agricultural droughts, and may trigger other types of droughts, which in turn have wide and far-reaching socio-economic impacts. Compared to other drought types, hydrological drought has fewer characterization indices. Scholars have long been keen to explore how to quantify the degree of hydrological drought in watersheds [1,2]. Proposed by Shukla and Wood [3] in 2008 based on the Standardized Precipitation Index, the Standardized Runoff Index (SRI) [4] has been applied by researchers [5] to characterize the intensity of hydrological drought in watersheds. To address the climate challenge, the Coupled Model International Comparison Program (CMIP) [6] provides a standard comparative framework for comparing the performance of different climate models on a global scale, which can be used to simulate climate change in the coming decades or centuries, and to help the international community to make decisions on addressing climate change. The latest CMIP6 data is currently open access to researchers via the Earth System Grid Federation (ESGF) starting in 2018. Compared with the previous CMIP, CMIP6 not only maintains the interface with CMIP5, but also introduces the new concept of Shared Socioeconomic Pathways (SSP). In CMIP6, different SSPs are incorporated into the design of climate models in order to better synthesize the impacts of socioeconomic development on climate change [7]. Each SSP is associated with a specific set of social, economic, and energy development scenarios, thus providing more options for climate scientists and policy makers to assess the response of the climate system under different future pathways. With the development and improvement of CMIP, an increasing number of scholars tend to simulate the evolution trend of future runoff based on the meteorological data provided by CMIP and in combination with the SWAT (Soil and Water Assessment Tool) model. For example, Yang [8] assessed the impacts of droughts on the Kelantan River Basin, based on data from CMIP6 and combined with the Soil and Water Assessment Tool (SWAT) and a Deep Learning model, to evaluate the potential impacts of climate change on runoff in the Kelantan River Basin. Lin [9] couples the maximum entropy and the Future Land Use Simulation model to predict the areas prone to waterlogging in the future. Sun [10] combined the SWAT model and the CMIP6 scenario to quantitatively analyze the impact of climate change on runoff. Kong [11] combines CMIP6 data and the SWAT model to optimize the water resource allocation in the Minjiang River Basin. Sivokhip [12] analyzed the study of the CMIP 5 based on data from Regional effects of regional climate change to identify the shifting trends of temperature in runoff parameters. Nagireddy [13] analyzed the trends of water quality and quantity in Nagavalli and Vamsadhara basins based on CMIP 6 and SWAT models.

A review of the existing literature reveals that most scholars, although committed to predicting future runoff trends under different scenarios, tend to stop at the direct prediction of runoff values, and seldom go further than the calculation of standardized runoff indices (SRIs). Although this approach can reveal the changes in runoff itself, it lacks a standardized and comparable quantitative framework to comprehensively assess the severity of hydrological droughts and their potential impacts on future water resources management. Meanwhile, most scholars have used the traditional Mann–Kendall (MK) test to analyze the trend of runoff, but the autocorrelation of time series data may reduce the reliability of the traditional MK test.

Therefore, this study comprehensively assesses the hydrological drought conditions in the upper reaches of the Xin’an River. Based on predicting runoff values under different future scenarios, the SRI is further calculated to eliminate natural variability caused by factors such as climate, geography, and seasonality, enabling runoff data from different years and sub-basins to be compared on the same scale. Additionally, to reduce the autocorrelation of the data, prewhitening processing is applied to make the data more closely resemble independent and identically distributed (i.i.d.) data, thereby making it more suitable for the MK test. Furthermore, cumulative anomaly tests and continuous wavelet transform methods are combined to provide a more comprehensive analysis of the runoff and hydrological drought characteristics of the upper reaches of the Xin’an River. The technical roadmap is shown in Figure 1.

2. Study Area and Methodology

2.1. Overview of the Study Area

As shown in Figure 2, Xin’an River originates from Xiuning County, Huangshan City, Anhui Province, passes through Qiandao Lake and Fuchun River, and finally joins the Qiantang River. According to the geographic location and basin characteristics, the upper reaches of Xin’an River refers to the section from its origin to Qiandao Lake. The upper reaches of Xin’an River are mostly mountainous and hilly, covering an area of about 5000 km². The basin has a subtropical monsoon climate, with an average annual temperature of about 16 °C. Typical vegetation types include evergreen broad-leaved forests and evergreen deciduous broad-leaved mixed forests. The main soil types are yellow loam and mountain yellow-brown loam, with thick soil layers and high gravel content.

2.2. Research Data

The data used to run the SWAT model are shown in

Table 1. Data sources.

Data Type	Data Sources	Data Use
Digital Elevation DEM	Geospatial Data Cloud (https://www.gscloud.cn/, accessed on 10 February 2025)	Extraction of river networks, delineation of sub-basins
Soil type data	HWSD Global Soil Database (https://gaez.fao.org/pages/hwsd, accessed on 10 February 2025)	Construction of a soil database, define the physical properties of the soil
Land use data	Resource Environment and Data Center, Chinese Academy of Sciences (https://www.resdc.cn, accessed on 10 February 2025)	Production of HRUs, affects the processes of surface coverage, evapotranspiration, production, and convergence.
Historical meteorological data	National Weather Science Data Center (https://data.cma.cn/, accessed on 10 February 2025)	Drive the model to operate and directly affect the real-time simulation of the hydrological cycle
Future meteorological data	CMIP6 (https://cds.climate.copernicus.eu/, accessed on 10 February 2025)	Predict future hydrological changes
Hydrological data	ERA5 (https://www.earthdata.nasa.gov/, accessed on 10 February 2025)	Calibration model parameters
NDVI	MOD13A3 Dataset (https://www.earthdata.nasa.gov/, accessed on 10 February 2025)	Analyze the effect of ecological compensation policies

. Among them, the DEM data is downloaded from the geospatial data cloud, and a resolution of 30 m is selected. Soil type data is used to describe the hydrological characteristics of different soils within the watershed, while land use data reflects surface cover conditions and influences hydrological processes in the region. Meteorological data are sourced from the National Meteorological Science Data Center, including daily average wind speed, solar radiation, maximum and minimum temperatures, etc., which are important factors influencing runoff changes. The Jiekou Hydrological Station is located at the junction of the upper and lower reaches of the Xin’an River. Selecting runoff data from the Jiekou Hydrological Station can assess the water volume transported from the upper to the lower reaches of the Xin’an River. However, according to the Taihu Basin Administration Bureau’s Government Information Disclosure Leading Group Office, the Jiekou Water Level Station was only approved by the Ministry of Water Resources in December 2020 to be included in the national basic hydrological station network. As a result, only runoff records from 2022 to 2023 are available, which lack statistical significance. The applicability of Copernicus Climate Data Center (ERA5) data in the Chinese context has been verified by many scholars [14,15], so the runoff data downloaded from ERA5 was chosen to be regarded as measured data for rate-setting the SWAT model parameters.

The predicted meteorological data were obtained from CMIP6, which is the sixth phase of the Global Climate Model Comparison Project (GCMCP), aiming to improve the understanding of the behavior of the climate system and the ability to predict future climate change by comparing and evaluating global climate models. Referring to existing studies, the future meteorological data of SSP126 (low-intensity emission model), SSP245 (medium-emission model), and SSP585 (high-emission model) under the MRI-ESM2-0 climate model, which has a high accuracy of climate simulation for China, are selected. The differences between the models are shown in

Table 2. Parameter sensitivity analysis and rate determination results.

Scenario Name	Descriptive	Radiative Forcing Levels (2100)	Emission Intensity
SSP126	Low-intensity emission models	2.6 W/m2	low
SSP245	Medium Emission Model	4.5 W/m2	moderate
SSP585	High-emission models	8.5 W/m2	high

.

Specifically, SSP126 represents a more sustainable social and economic development pathway. Assuming that more stringent climate policies and sustainable development practices are followed in the future [16], CO₂ emissions decline significantly, reaching net zero by mid-century, with CO₂ emissions declining by about 2.6% by 2100 compared to 2010, average temperature increases within 2 °C, and radiative forcing stabilizing at about 2.6 W/m². SSP245 denotes a moderate social and economic development pathway with moderate response measures to climate change, with CO₂ emissions beginning to decline in 2050, average temperature increase of about 2.7 °C in 2100, and radiative forcing stabilizing at about 4.5 W/m². SSP585 denotes a high-intensity social and economic development pathway accompanied by higher GHG emissions, with CO₂ emissions in 2050 increasing from current levels by approximately doubles, with an average temperature increase of 4.4 °C by the end of the century and a stabilization of the radiative forcing at about 8.5 W/m².

2.3. Research Methodology

2.3.1. SWAT Model

The SWAT model is a hydrological model at the basin scale. Although it requires a large amount of data-driven models and is complex to operate, due to its comprehensive simulation ability of each link in the hydrological cycle and its meticulous consideration of natural factors such as soil, vegetation and terrain, it can accurately capture the reduction and changes in water resources under drought conditions in this study. The simulation process of the SWAT model is mainly divided into two parts: the sub-basin module (or called the land surface water cycle process) and the confluence calculus module (or called the water surface process). The sub-basin module is responsible for simulating flow generation and slope convergence, and controlling the input of water, sediment, nutrients and chemicals in the main river channels of each sub-basin. The confluence calculation module determines the transport movement of substances such as water and sand from the river network to the outlet of the basin. The water balance expression of the SWAT model is as follows:

$$S{W}_t={SW}_0+\sum _{i=1}^t\left(P_d-Q_s-E_a-W_s-W_g\right)$$

(1)

where SW_t is the final soil water content (mm), t is the length of the calculation (days), SW₀ is the initial soil water content (mm), P_d is the daily rainfall (mm), Q_s is the daily surface runoff (mm), E_a is the daily evapotranspiration (mm), W_s is the amount of water (mm), returned to the stream W_g is the soil profile leakage (mm).

In general, the correlation coefficient R² and the Nash efficiency coefficient NES to assess the simulation effect, and the relevant formulas are calculated as follows:

$$R^2={\displaystyle \frac{[{\sum }_{i=1}^n{(Q_0-Q_s){(Q}_p-Q_q)]}^2}{{\sum }_{i=1}^n{(Q_0-Q_S)}^2{\sum }_{i=1}^n{(Q_p-Q_q)}^2}}$$

(2)

$$NES=1-{\displaystyle \frac{{\sum }_{i=1}^n{(Q_0-Q_p)}^2}{{\sum }_{i=1}^n{(Q_0-Q_s)}^2}}$$

(3)

where Q₀ is the measured runoff, m³/s; Q_p, Q_q for the runoff simulation and its average, m³/s; Qs for the average measured runoff, m³/s; n is the number of data.

2.3.2. Standard Runoff Index

Standardized Runoff Index (SRI) is a hydrologically based statistical analysis method to measure the extent to which a river deviates from its natural operating conditions [17]. The SRI index is calculated by the following formula:

$$\mathrm{F}(\mathrm{Y})=\underset{0}{\overset{\mathrm{Y}}{\int }}{\mathrm{f}}_{\mathrm{Y}}\mathrm{d}\mathrm{Y}$$

(4)

$$SRI=\left\{\begin{array}{c}{\displaystyle \frac{\mathrm{I}-({\mathrm{c}}_1-{\mathrm{c}}_2)\mathrm{I}+{\mathrm{c}}_0}{\left[\right({\mathrm{d}}_3\mathrm{I}+{\mathrm{d}}_2)\mathrm{I}+{\mathrm{d}}_1]\mathrm{I}+1}},F>0.5\\ -{\displaystyle \frac{\mathrm{I}-({\mathrm{c}}_1-{\mathrm{c}}_2)\mathrm{I}+{\mathrm{c}}_0}{\left[\right({\mathrm{d}}_3\mathrm{I}+{\mathrm{d}}_2)\mathrm{I}+{\mathrm{d}}_1]\mathrm{I}+1}},F\le 0.5\end{array}\right.$$

(5)

where Y represents the runoff volume under a certain time scale, mm; F is the cumulative distribution probability; C₀ = 2.515517, C₁ = 0.802853, C₂ = 0.010328, d₁ = 1.432788, d₂ = 0.189269, d₃ = 0.001308; I = (lnF − 2) 0.5, combined with the study area actual situation to divide the hydrological drought level standard. The hydrological drought grade standards are shown in

Table 3. Drought classification.

SRI Value	Category
(−∞, −2]	Exceptional drought
(−2, −1.5]	Severe drought
(−1.5, −1]	Moderate drought
(−1, 0]	Mild drought
(0, +∞]	No drought

.

2.3.3. Cumulative Anomaly Method

Cumulative anomaly is a method commonly used to analyze the trend of hydrological time series data, calculating the cumulative sum of the differences between data points and the mean value, and determining the degree of dispersion and the relative trend of the data series. The “inflection points” on the curve can be used to characterize the stages of the data. In this study, the cumulative horizon method can distinguish between “water-rich” and “dry” periods. When the cumulative distance level continues to rise, it means that the data in that period are generally higher than the average value, which is characteristic of “water-abundant period”. On the other hand, a decrease in the cumulative distance level indicates that the data are below the average and a “dry period” has been reached. Specifically, the cumulative level is calculated using the formula:

$$X_j=\sum _{i=1}^j\left(X_i-\overline{X}\right)(i=1,2,\dots ,n)$$

(6)

where $\overline{X}$ is the average of runoff or SRI. X_j is the runoff or SRI in year i. X_j is the cumulative distance level in year j.

2.3.4. Modified Mann–Kendall

The persistence of time series can interfere with the analytical results of the traditional Mann–Kendall test. In order to eliminate the autocorrelation of time series, the Modified Mann-Kendall (MMK) test is used in this study to pre-whiten the runoff data and SRI in advance to analyze the trend of runoff data and SRI. The MMK test method, compared with the MK test method, adds a prewhitening process for the data. Firstly, it standardizes the mean and variance of the original time series to eliminate the dimensional influence. Then, it constructs an independent white noise sequence of the same length as the original sequence (the variance is adjusted according to the first-order autocorrelation coefficient), and superimposes the standardized sequence with the white noise to disrupt the autocorrelation structure of the original sequence. Finally, the original dimension was restored through destandardization, and a new sequence with significantly reduced autocorrelation was obtained.

$$\beta =\mathrm{m}\mathrm{e}\mathrm{d}\mathrm{i}\mathrm{a}({\displaystyle \frac{x_j-x_i}{t_j-t_i}}) ,\forall j>i$$

(7)

where β is the trend of runoff or SRI index under different emission scenarios, t_i and t_j are time series, and x_i and x_j denote the extreme index values at moments i and j. MMK statistic Z with an absolute value greater than 1.96 indicates that it passes the 0.05 significance test.

2.3.5. Continuous Wavelet Transform

Continuous Wavelet Transform (CWT) [18] was proposed by French scientist Morlet in 1980 when he analyzed seismic wave data. It can obtain the representation of a signal at different times and frequencies by convolving it with a wavelet function. By changing the scale of the wavelet function, the information of the signal on the corresponding time scale can be obtained. In this study, CWT is used to identify the start and end time of drought events and severity, as well as periodic and trend changes in runoff data, and to develop appropriate responses. The method is defined as f(t) ∈ L²(R).

$$w_f(a,b)={\left|a\right|}^{{\displaystyle \frac{1}{2}}}\underset{-\mathrm{\infty }}{\overset{+\mathrm{\infty }}{\int }}\mathrm{f}(\mathrm{t})\overline{\Psi }({\displaystyle \frac{t-b}{a}})dt$$

(8)

In Equation (8) w_f (a,b) are the wavelet transform coefficients, a and b are the scale factor and translation factor, respectively, ψ(t) is the basis wavelet, which is the molert wavelet in this paper, t is the time, and f(t) is the original signal.

Meanwhile, red noise is used as the background spectrum to test the continuous wavelet transform energy spectrum. The first-order autoregressive equation (AR1) modeled the test process with the background red noise power defined as

$$p_k=(1-a_2)/|1-a_2{e}^{-2i\pi k}|$$

(9)

In Equation (9) a₂ is the correlation coefficient of the autoregressive equation in the red noise power spectrum, and k is the Fourier frequency exponent, chosen with 95% confidence.

3. Research Results

3.1. Parameterization Results

Based on the DEM image of the upper reaches of Xin’an River after filling the depressions, the boundaries of the entire basin were extracted. According to information such as slope and flow direction, the basin was divided into 35 sub-basins with relatively independent hydrological characteristics. The thresholds for land use, soil and slope were set at 10%, 16% and 23%, respectively, and further divided into 35 sub-basins with a total of 341 hydrological response units. Input the weather data from the meteorological station, set the period from 2000 to 2001 as the preheating period, run the model, and obtain the initial simulation results. To ensure the simulation accuracy, this study calibrated and verified all parameters. Although it increased the workload, it improved the accuracy of the parameters. Referring to relevant literature [19], with the period from 2002 to 2014 as the rate period and the period from 2015 to 2020 as the verification period, a total of 228 sample numbers, using the SUFI-2 algorithm, the simulated runoff values and the measured runoff values were imported into Swat-CUP 2012, and the initial range of SWAT operation parameters was also imported. After 4000 iterations, The optimal parameter Settings are obtained, as shown in

Table 4. Calibration results for sensitive parameters.

Parameter_Name	Physical Meaning	Fitted_Value	Min_Value	Max_Value
1: V__SFTMP.bsn	Snowfall temperature/°C	2.408	1.678	5.000
2: V__SMTMP.bsn	Snowmelt temperature/°C	−4.452	−5.000	−1.655
3: V__SMFMX.bsn	Snowmelt factor for June (21 mm H2O/°C–day)	9.876	6.882	10.000
4: V__SMFMN.bsn	Snowmelt factor for December (21 mm H2O/°C-day)	8.375	6.806	10.000
5: V__TIMP.bsn	Snow temperature lag coefficient	0.538	0.486	0.843
6: V__SURLAG.bsn	Surface runoff hysteresis factor	4.852	1.000	8.320
7: V__TLAPS.sub	Vertical rate of temperature decrease (°C/km)	20.082	17.615	30.685
8: V__SLSUBBSN.hru	Average slope length/m	51.370	23.300	78.474
9: V__HRU_SLP.hru	Average slope/(m∙m−1)	0.005	0.000	0.157
10: V__CANMX.hru	Maximum canopy retention (mm H2O)	2.300	0.000	32.506
11: V__OV_N.hru	Manning’s slope diffuse flow n-value	0.795	0.519	0.800
12: V__ESCO.hru	Soil evaporation compensation factor	0.846	0.516	0.907
13: V__EPCO.bsn	Plant uptake compensation factor	0.114	0.070	0.478
14: R__CN2.mgt	Initial SCS runoff curve number for wet condition II	0.216	0.159	0.500
15: R__BIOMIX.mgt	Biomixing efficiency	−0.213	−0.500	−0.166
16: R__SOL_Z (…).sol	Depth from soil surface to subsoil (mm)	−0.417	−0.500	−0.154
17: R__SOL_BD (…).sol	Soil saturation capacity (mg∙m−1)	−0.189	−0.431	−0.023
18: R__SOL_AWC (…).sol	Effective water capacity of soil layer (mm(H2O)/mm(soil))	0.268	0.228	0.500
19: R__SOL_K (…).sol	Soil saturated hydraulic conductivity (mm∙h−1)	0.547	0.126	0.800
20: R__SOL_ALB (…).sol	Moist soil albedo	−0.123	−0.500	−0.116
21: V__GW_DELAY.gw	Groundwater delay time/d	26.702	0.000	132.683
22: V__ALPHA_BF.gw	Baseflow alpha factor/d	0.152	0.000	0.380
23: V__GWQMN.gw	Depth to which (shallow aquifers produce “base flow” mm)	2538.879	1705.331	3040.610
24: V__GW_REVAP.gw	Re-evaporation coefficient for shallow groundwater	0.063	0.059	0.119
25: V__REVAPMN.gw	Depth of or infiltration of into deeper aquifers (“re-evaporation” shallow aquifers mm)	442.071	363.294	500.000
26: V__RCHRG_DP.gw	Deep aquifer permeability ratio	0.869	0.569	1.000
27: V__CH_N2.rte	Main channel bed Manning’s coefficient	0.017	0.000	0.093
28: V__CH_K2.rte	Effective hydraulic conductivity of the main channel bed (mm∙h−1)	72.612	46.181	88.759

.

The simulation results are shown in Figure 2. Nash coefficient (NSE) and deterministic coefficient (R²) were used to evaluate the fitting effect. It is generally believed that when R² > 0.5 and NSE > 0.5, the runoff fitting degree is more ideal. As can be seen from

Table 5. The results of calibration and verification.

Indicator	Numerical Value
R2	0.76
NSE	0.60

, NES and R2 in the rate period and verification period are above 0.60, and the fitting effect is good, indicating that the SWAT model has high applicability in the upper reaches of Xin’an River.

3.2. Historical Runoff and Drought Characterization

Changes in runoff volume are a direct reflection of hydrologic drought, so when studying the trend of SRI, the changes in runoff volume need to be paid full attention to.

According to the simulation results, it can be found that the runoff from the upper Xin’an River at Jiekou Hydrological Station is concentrated 100–300 m³/s, with an average monthly runoff of 197.35 m³/s, and the peak of the annual runoff is around June. According to the MMK test method to analyze the historical runoff trend changes, the results are shown in

Table 6. MMK test results.

	Runoff			SRI
Scenario SSP	Z-Value	Z-Trend	Sens Slope	Z-Value	Z-Trend	Sens Slope
History	1.1227	1.0000	2.7243	1.3678	1.0000	0.0048
SSP126	1.3218	1.0000	0.0146	1.6029	1.0000	0.0004
SSP245	0.6118	1.0000	0.0057	0.8320	1.0000	0.0002
SSP585	0.6282	1.0000	0.0058	0.8466	1.0000	0.0002

, which shows that the runoff in the upper reaches of Xin’an River is a non-significant during the period of 2000–2020 positive trend. In order to analyze the trend change in runoff in the upper Xin’an River, according to Equation (6), a cumulative distance plot is drawn as shown in Figure 3, and the years from 2002 to 2009 are dry years, from 2009 to 2013 are flat years, and the years from 2013 to 2020 are abundant years.

As shown in Figure 4, the average monthly runoff distribution throughout the year exhibits a consistent unimodal pattern under all three emission scenarios. This single-peak characteristic is evident across the near-term, mid-term, and long-term future periods, as well as over the entire study time series. The number of months with mild, moderate, severe, and exceptional drought were 34, 29, 14, and 2 months, respectively. According to Table 6, the SRI in the upper reaches of the Xin’an River was a non-significant positive trend during 2000–2020. From Figure 5, a 69-month oscillation cycle exists in 2009–2015.

3.3. Runoff Evolution Patterns Under Future Scenarios

To refine the study, the time series of the study was divided into near-term (2025–2050), medium-term (2051–2075), and far-term (2076–2100). The future meteorological data of the three models were downscaled by bilinear interpolation and imported into the weather generator of SWAT to obtain the future runoff data of the upper Xin’an River region, as shown in Figure 6.

Figure 6 shows that for the period 2025–2100, the monthly average runoff under the scenarios SSP126, SSP245, and SSP585 is, respectively, 279.68 m³/s, 277.74 m³/s, and 277.71 m³/s. The increase in monthly average runoff compared to the historical period is 41.72%, 40.74%, and 40.72%, respectively. This is because in future scenarios, the rise in temperature leads to a dual consequence of increased evaporation and increased precipitation. However, the increase in precipitation far exceeds the evaporation consumption, resulting in a significant increase in runoff. In terms of different emission discharge patterns, the monthly mean runoff decreases with the increase in GHG emission concentration, and the monthly mean runoff and its increase in the SSP126 scenario is slightly higher than the other two scenarios, while the average monthly runoff of SSP245 is only 0.03m³/s more than that of the SSP585 scenario. In terms of the degree of dispersion of the runoff, the fluctuation in the monthly mean runoff of the SSP126 scenario is the standard deviation of the SSP126 scenario is 163.39, and the extreme runoff values occur in July 2051 and July 2073, with average monthly runoffs of 1413 m³/s and 1461 m³/s. The fluctuation of the SSP245 and SSP585 scenarios is smaller, with standard deviations of 162.26 and 162.40. The SSP245 and SSP585 scenarios have a lower fluctuation, with standard deviations of 162.26 and 162.40. The extreme runoff values under both scenarios occur in July 2036 and June 2075 and have magnitudes of 1157 m³/s and 1107 m³/s.

From a longitudinal perspective, there are differences in the trends of runoff volume under the three scenarios. According to Table 6, the runoff volume under the three future emission scenarios is a non-significant positive trend. The average monthly runoff volume of SSP126 in the near-term, medium-term, and far-term are 267.95 m³/s, 292.49 m³/s, and 279.07 m³/s, with a spiral upward trend of increasing and then decreasing. The average monthly runoff volume of SSP245 is 274.97 m³/s, 279.05 m³/s, and 279.32 m³/s, showing a continuous upward trend, but the mid-term increase of 1.5% is significantly higher than that of 0.1% in the long term. The average monthly runoff of SSP585 is 274.82 mm³/s, 279.14 mm³/s, 279.28 mm³/s. The trend is basically the same as that of SSP245.

Expanding the analysis from a cross-sectional perspective, the monthly average runoff volume of SSP126 is higher than the other two emission scenarios for 462 months during the period of 2025–2100, accounting for 50.7% of the total number of months. In contrast, although there is a similarity in the trend of monthly mean runoff between the two scenarios SSP245 and SSP585, SSP245 has a higher monthly mean runoff than the SSP585 scenario for 767 months, accounting for 84.1% of the total months.

In order to analyze the annual distribution characteristics of future runoff, the annual average monthly runoff of the upper reaches of Xin’an River was calculated, as shown in Figure 7.

As can be seen from Figure 7, the distribution characteristics of the average monthly runoff within a year under the three emission scenarios in the near term, medium term, long term and the entire research time series are basically similar, showing a unimodal curve distribution. And the peak runoff volume was all in May, June and July. From the perspective of seasonal scale, the runoff volume under the three emission scenarios is mainly concentrated in late spring and early summer, accounting for about 40% of the annual runoff volume.

3.4. Drought Characterization Under Future Scenarios

The simulated runoff data for the three future scenarios were brought into the equation to calculate the standard runoff index for the three future scenarios, and the results are shown in Figure 8.

From Figure 8, it is easy to see that as the concentration of GHG emissions increases, the proportion of drought-free months increases less, but the proportion of extra-dry months increases more. For example, from the perspective of the whole time series, the mean value of the annual-scale SRI of SSP126 is 0.04. The dynamics of the annual-scale SRIs of SSP245 and SSP585 maintain a high degree of similarity, and the mean value is the same as 0.05. According to the statistics of the annual-scale SRIs, it can be seen that during the period of December 2000 to December 2020, the number of drought-free in the SSP126 scenario is 630 months. There were 630 months, accounting for 69.1% of the total number of months. There are 153, 93, 28, and 8 months of mild, moderate, severe, and exceptional drought, respectively. 651 months, or 71.4% of the total number of months, are drought-free under the SSP245 scenario. There were 140, 71, 33, and 17 months of mild, moderate, severe, and exceptional drought, respectively. 660 months, or 72.4% of the total, were drought-free under the SSP585 scenario. There were 148, 77, 39, and 25 months of mild, moderate, severe, and exceptional drought, respectively. In terms of the dispersion of the annual-scale SRI, the SSP126 scenario had the smallest fluctuation in the annual-scale SRI, with a standard deviation of 0.96, and the standard deviation of both SSP245 and SSP585 was 0.98.

According to Table 6, the SRIs for the three future emission scenarios have a non-significant positive trend. When analyzed in the near, medium, and distant future, the average SRIs were −0.20, 0.28, and 0.05, respectively, for SSP126, −0.04, 0.09, and 0.11, respectively, for SSP245, and −0.04, 0.09, and 0.11, respectively, for SSP585. It can be seen that over time, the hydrologic drought under the SSP126 emission scenario, the hydrological aridity was the lowest in the middle future, but then gradually increased in the far future. The hydrological aridity of the SSP245 and SSP585 scenarios, on the other hand, tended to decrease slowly from the near future to the long term.

In order to further analyze the hydrological drought characteristics under three scenarios, the CWT continuous wavelet transform was chosen to analyze the drought distribution cycle in the upper Xin’an River region (SSP126, SSP245, and SSP585), and the continuous wavelet frequencies are shown in Figure 9.

It can be seen from Figure 9 that there are two oscillation periods in the SSP125 scenario. The first cycle period is from 2032 to 2065, lasting approximately 400 months. The second cycle is from 2064 to 2093, lasting approximately 359 months. Since the SRI in the SSP245 and SSP585 scenarios is highly similar, there is not much difference in their continuous wavelet frequency graphs. After passing the 95% confidence level test, the SRI in both the SSP245 and SSP585 scenarios has an 835-month oscillation period, with the period from 2028 to 2098.

4. Discussion

4.1. Discussion on the Evolution Trend of Historical Hydrological Drought

Based on the historical changes in runoff and SRI in the upper reaches of Xin’an River, the turning point from hydrological drought to hydrological wet period in the study area occurred around 2012, and the hydrological wet period ended around 2020. The implementation period of the lateral ecological compensation policy for Xin’an River was from 2012 to 2020, which highly overlapped with the hydrological wet period of the upper reaches of Xin’an River. Compared with the NDVI before and after 2012, the NDVI in 2014 was 0.81 and that in 2010 was 0.80, indicating that ecological compensation measures have increased the vegetation coverage rate. This indicates that the ecological compensation policy of Xin’an River Basin can alleviate the severity of drought in the policy area by increasing the vegetation coverage rate. Compared with previous studies [20] that only focused on the promoting effect of ecological compensation policies on water supply, this study delved deeper into the level of hydrological drought and analyzed the alleviating effect of ecological compensation policies on hydrological drought. In addition, by observing the cumulative deviation curves of historical runoff and SRI in the upper reaches of the Xin’an River, it was found that there was a trend of hydrological drought lasting for about two years around 2017. This is because 2017 was the third pilot stage of the Xin’an River Horizontal ecological compensation policy. The central government’s ecological compensation funds decreased and the regulatory intensity of the central government weakened, resulting in a decline in the implementation effect of the policy. This, in turn, led to the trend of hydrological drought, further verifying that the ecological compensation policy can alleviate the hydrological drought situation in the policy area, which is consistent with the research results of Wang [21]. This is because the ecological compensation funds are specifically used for afforestation, wetland protection and other ecological restoration projects in the upstream areas, enhancing the water retention capacity of the upstream areas. In addition, Anhui and Zhejiang provinces have collaborated on water resource monitoring and allocation to ensure ecological flow during drought periods. In addition, this policy encourages the development of low-water-consuming industries in the upstream areas, reducing water demand.

4.2. Discussion on the Evolution Trend of Hydrological Drought in the Future

Based on changes in the SRI under three future scenarios for the upper reaches of the Xin’an River, the hydrological drought trend is more stable under the SSP126 scenario, and the proportion of extreme drought months is significantly lower than in the other two scenarios. This indicates that as greenhouse gas concentrations increase, extreme weather events in the study area become more frequent, and hydrological drought conditions become increasingly unstable, consistent with the findings of Chen [22]. The primary reason lies in the accelerated evaporation of surface water due to temperature increases caused by carbon emissions, which reduces soil moisture and accelerates surface water loss. Additionally, rising carbon emissions may alter precipitation patterns, affecting precipitation amounts and distribution, triggering extreme precipitation events, and thereby impacting the supply and distribution of surface water resources, exacerbating regional hydrological drought conditions.

5. Conclusions

Based on the meteorological data and DEM data of the upper reaches of Xin’an River, a SWAT model was constructed. The model parameters were adjusted based on the hydrological data from 2000 to 2020. Based on the meteorological data of three scenarios in CMIP6, the daily runoff of the upper reaches of Xin’an River from 2025 to 2100 was predicted. The hydrological drought characteristics of the upper reaches of Xin’an River under the shared path of social economy were analyzed, and the following conclusions were drawn:

(1) The calibration and validation of the SWAT model using ERA5 runoff data yielded a NSE of 0.60 and a coefficient of R² of 0.76, both meeting the acceptable standards for hydrological modeling (NSE > 0.5, R² > 0.5). This outcome not only demonstrates the reliability of ERA5 runoff data within the study area, indicating that discrepancies between the dataset and actual hydrological conditions are within acceptable limits, but also validates the strong applicability of the SWAT model in the upper Xin’an River basin, confirming its effectiveness in simulating fundamental hydrological processes in this region. 2002 to 2009 is a dry period, and 2013 to 2020 is an abundant period, and 2012 is the year when the hydrological drought shifted to hydrological wetness. The SRI of the upper Xin’an River has a 69-month oscillation cycle from 2009 to 2015. Horizontal ecological compensation policies can effectively alleviate hydrological droughts in the Xin’an River basin. It is recommended that the cyclical differences in hydrological droughts be fully considered to ensure that compensation funds are accurately allocated to critical periods. It should be noted that in this study, the ERA5 reanalysis product, as a historical runoff benchmark dataset, has seen its accuracy widely verified and basically meets the analysis requirements for runoff trend changes. However, there are still potential uncertainties. To minimize prediction errors to the greatest extent, it is necessary to continuously monitor and obtain the long-term observed runoff data of Jiekou Hydrological Station in the future, improve the reliability of hydrological drought prediction, and ensure more powerful support for drought risk management and water resources planning in the Xin ‘an River Basin.

(2) The average monthly runoff under the three scenarios in the future will increase compared with the historical period, and the SSP126 scenario has the most significant change, but the highest degree of dispersion. The annual average monthly runoff under the three emission scenarios was concentrated in May, June and July. Therefore, the seasonal allocation of water resources should be strengthened in the future. In the peak of runoff, improve the storage and regulation capacity of the reservoir to ensure sufficient water supply in the dry season. Relying on the Xin’an—Qiandao Lake ecological compensation mechanism, an emergency water diversion channel between Anhui and Zhejiang will be built, and multiple regulating reservoirs will be constructed upstream to meet the emergency needs during the dry season. At the same time, through the cross-basin water transfer project, the excess water resources can be deployed to the water-scarce areas to improve the utilization efficiency of water resources.

(3) The hydrological drought in the three scenarios will be alleviated in the future, but with the increase in emission intensity, the drought-free and drought-intensive months will increase at the same time, and the drought characteristics in the study area will become more obvious. SRI in the SSP126 scenario has two oscillation cycles of 400 months and 359 months, respectively. The continuous wavelet frequency graphs of SSP245 and SSP585 are similar, and both have an oscillation cycle of 835 months. In response to the polarized trend of “no drought—extreme drought” in future hydrological droughts, it is suggested to establish a hierarchical drought resistance system, set four levels of early warning indicators and make dynamic adjustments in combination with seasons, and alleviate the impact of drought through the coordinated efforts of “precise emergency response—cycle adaptation—low-carbon regulation”. During periods of extreme drought, priority should be given to ensuring water supply for residents and implementing agricultural quotas. In periods without drought, groundwater replenishment should be carried out. Simultaneously promote drought-resistant crops and water-saving irrigation technologies, and optimize the irrigation district renovation nodes based on drought cycles of 400 months, 359 months and 835 months. The promotion of water-saving technologies adopts a three-level subsidy mechanism. The basic subsidy is linked to water-saving benefits and carbon sink income. At the same time, a water-saving points system and an incentive system for farmers’ cooperatives are established.