Climate Change Impacts on Runoff in the Fujiang River Basin Based on CMIP6 and SWAT Model

Abstract

Understanding the responses of the hydrological cycle and extreme events to climate change is essential for basin water security. This study systematically assessed climate change impacts on runoff and floods in the Fujiang River basin, which is the main tributary of the upper Yangtze River, China, using the Soil Water Assessment Tool (SWAT) driven by the latest climate simulation of 14 General Circulation Models (GCMs) under two Shared Socioeconomic Pathways (SSP2-4.5 and SSP5-8.5). This study indicates a generally warmer and wetter climate projected in the Fujiang River basin, and correspondingly an overall increase in projected ensemble annual mean runoff, monthly runoff, monthly high flow (Q05), and monthly low flow (Q95) in the periods of 2021–2060 and 2061–2100, with the long-term period being more substantial than that of the near future, especially for SSP5-8.5. However, the projected changes in monthly runoff show a large spread across GCMs, with greater increases mainly occurring in the early rainy season. Most of the GCMs show that projections of Q95 will substantially increase compared to Q05. The intensity and frequency of floods with a 30-year return period are likely to increase, especially under SSP5-8.5. Despite the uncertainties in projected future changes in runoff, these findings highlight the complexity of runoff response to climate change, promoting the need for adaptive water resource management.

1. Introduction

The greenhouse gas (GHG) emissions have caused substantial global warming. The global average surface temperature increased by 0.99 °C in the first two decades of the 21st century (2001–2020) compared with 1850–1900 and will exceed 1.5 and 2.0 °C without a deep reduction in GHG emissions [1]. Anthropogenic climate change has intensified the hydrological cycle with a consequence on physical aspects of water security, and furthermore, has caused extreme weather events with highly impactful floods and droughts [2].

The Yangtze River basin is an important area for China’s economic development, and the upper Yangtze River is also an area with high sensitivity to climate change affected by the East Asian summer monsoon and Indian summer monsoon, as well as the Tibetan Plateau [3]. Studies have shown that the observed runoff generally decreased in the upper Yangtze River and was expected to increase slightly for the Jinshajiang River basin during the past 50 years [4,5,6]. Along the upper Yangtze River, the river discharges in autumn show a more significant decreasing trend, especially on the tributaries of the Minjiang River and Jialingjiang River [7]. The observed changes in annual discharge are mainly contributed by climate change, and human activities have altered the annual discharge distribution during the past 100 years. Many studies have also assessed the climate change impacts on runoff based on various climate change scenarios for the upper Yangtze River. Total annual runoff of the whole upper Yangtze River (above Yichang hydrology station) is projected to decline due to the decreases in precipitation and a potential increase in evapotranspiration, but an increase in sub-regions including the Jialingjiang River basin is predicted during 2010–2099 under the Special Report on Emission Scenarios (SRES) [8,9]. The annual runoff of the upper Yangtze River is projected to decrease by 11.1% due to the significant increase of evapotranspiration caused by the rising temperature, while the precipitation is projected to increase by 4.1% in 2041–2070 under the Representative Concentration Pathway 8.5 (RCP8.5) using 78 climate projections from Phase 5 of the Coupled Model Intercomparison Project (CMIP5) [10]. In contrast, the annual discharge, maximum discharge in flood season, and the daily peak discharge of the upper Yangtze River (above Cuntan hydrology station) are projected to slightly increase with the projected increase in precipitation based on five GCMs and four RCPs [11]. Therefore, there is still great uncertainty regarding the future precipitation projection, runoff simulation, and water balance estimation of the upper Yangtze River, especially in the Jialingjiang River.

The Fujiang River is located on the right bank of the Jialingjiang River, and is the secondary tributary of the upper Yangtze River, providing valuable water in the Chongqing metro area [12]. The Fujiang River flows through the rainstorm area in western Sichuan province and is also prone to major floods [13]. A few studies investigated the changes in observed precipitation and runoff in the Fujiang River basin. Most findings indicate that there is a drier tendency in the Fujiang River basin based on precipitation observations, and the precipitation decreased by 28% and 24% during the entire year and flood season during 1960–2006, respectively. The Xiaoheba hydrology station, which controls about 79% of the drainage of the Fujiang River basin, shows that the annual average runoff decreased by 20% during 1955–2000 [14,15,16]. The simulated future runoff shows a decrease with great uncertainties. The simulated changes in runoff based on the hydrological model and climate change projection from one GCM under SRES A1B, A2, and B2 showed that the decadal runoff in the Fujiang River basin from the 2010s to 2050s will decrease by 0.3~17%, and the magnitude will be larger than 5%, especially after 2030 [17]. The simulated annual runoff shows a reduction of 1.5 °C global warming but with an increase of 2.0 °C in the Fujiang River basin based on five GCMs under four RCPs [18]. These studies, however, have not fully quantified the uncertainties in climate change impact on runoff simulation considering they employed limited GCMs, and furthermore, have not estimated the future change in extreme flow and flood risks.

The comparison of summer precipitation simulation in China by eight climate models from Phase 6 of the Coupled Model Intercomparison Project (CMIP6) and the corresponding eight previous models from CMIP5 showed that the CMIP6 multi-model ensemble (MME) is more skillful than the CMIP5 MME in the simulation of climatological precipitation over eastern China. Most of the CMIP6 models have advantages over precious CMIP5 models in reproducing the interannual anomalous rainfall pattern over eastern China [19]. The evaluation of the ability of 23 climate models from CMIP6 in simulating extreme climate events shows that the MME of CMIP6 models can capture the spatial patterns of precipitation extremes quite well over China [20].

Considering the improvement of CMIP6 models in precipitation and precipitation extremes simulation, this study employs the latest and large-subset climate change scenarios to capture the most plausible range of climate change based on climate simulation derived from 14 GCMs of CMIP6. Moreover, this study explores a more comprehensive climate change impact assessment than previous studies by using metrics covering annual mean runoff, the seasonal pattern of runoff, extreme monthly runoff, and floods to explore the climate change impact on physical water security in the Fujiang River basin. Additionally, future runoff and extreme flow projections are estimated by feeding the climate simulations from each GCM into a calibrated daily hydrological model, and the ranges and uncertainties in runoff projection contributed by GCMs are fully quantified.

2. Materials and Methods

2.1. Study Area

The Fujiang River has a total length of 670 km and a total basin area of 36,400 km². The basin is densely populated with 15.36 million people and consists of heavily developed agriculture and industry sectors. The Fujiang River basin is located in the transition area from the plateau in western Sichuan to the subtropical humid climate zone, with an average annual temperature of 13.5 °C (ranging from 7.8 to 18.1 °C), with an increasing trend of about 0.18 °C/10a, and annual precipitation of 983.9 mm (ranging from 590 to 1230 mm) with a decreasing trend of about 7.2 mm/10a (Figure S1). Xiaoheba hydrological station (106.05° E, 30.1° N) is the drainage control station of the Fujiang River basin (Figure 1). The catchment area above the Xiaoheba hydrological station is about 29,420 km². The annual average runoff depth of the catchment area is about 485 mm, and the annual average discharge is 459 m³/s.

2.2. Available Data

The spatial dataset for parameterization of the hydrological model included DEM, land-use, and soil data. The DEM, generated from a topographic map with a resolution of 1:250,000, was developed by the National Geomatics Center of China. The land-use data for the 2000s, with a spatial resolution of 1000 m, were provided by the National Cryosphere Desert Data Center [21]. The soil data derived from a harmonized world soil database with 0.5° resolution were provided by the Food and Agriculture Organization (FAO) of the United Nations [22].

The daily and monthly discharge data measured by the Xiaoheba hydrological station from 1951 to 2012 were obtained from the Hydrological Bureau of Chongqing. Gridded daily climate data with a spatial resolution of 0.25° × 0.25°, which were obtained from the CN05.1 dataset [23], were used for calibrating the hydrological model and correcting the GCMs’ biases. The CN05.1 dataset was developed from over 2400 climate stations in China from 1961 to 2017 and has been widely used to evaluate climate model performance and assess the impacts of climate change over China.

Climate simulations datasets from 14 GCMs (

Table 1. Model name, resolution, and ECS of 14 CMIP6 GCMs.

No.	Model	Resolution (Latitude × Longitude)	ECS (°C)
1	ACCESS-CM2	1.2° × 1.8°	4.72
2	ACCESS-ESM1-5	1.2° × 1.8°	3.87
3	BCC-CSM2-MR	1.1° × 1.1°	3.16
4	CCCma-CanESM5	2.8° × 2.8°	5.62
5	CNRM-CM6-1	1.4° × 1.4°	4.83
6	CNRM-ESM2-1	1.4° × 1.4°	4.76
7	GFDL-CM4	1.0° × 1.3°	*
8	HadGEM3-GC31-LL	1.3° × 1.9°	5.55
9	INM-CM4-8	1.5° × 2.0°	1.83
10	INM-CM5-0	1.3° × 2.5°	1.92
11	IPSL-CM6A-LR	1.3° × 2.5°	4.56
12	MIROC6	1.4° × 1.4°	2.61
13	MPI-ESM1-2-HR	0.9° × 0.9°	2.98
14	MRI-ESM2-0	1.1° × 1.1°	3.15

Note: * Data absent.

) under two common Shared Socioeconomic Pathway and Representative Concentration Pathways (SSP-RCP) scenarios (SSP2-4.5 and SSP5-8.5) were obtained from the National Climate Center, China Meteorological Administration. The output of these 14 GCMs was originally derived from the latest CMIP6 and further spatial-downscaled and bias-corrected using spatial disaggregation and equal distance cumulative distribution functions [24]. The datasets included daily maximum and minimum temperatures and daily precipitation in China from 1961 to 2100 with a 0.25° × 0.25° resolution. The downscaling method improved the distributions of climate patterns and extreme climate events in China compared with the original simulation [25]. Climate simulations were divided into a historical simulation period covering 1961–2014, and a future projection period under both scenarios of SSP2-4.5 and SSP5-8.5 covering 2015–2100. The equilibrium climate sensitivity (ECS) is an important quantity used to estimate how the climate responds to radiative forcing, and the best estimate is 3 °C in IPCC AR6 [1]. The 14 GCMs included in the dataset with ECS range from 1.83 to 5.62 °C, covering most of the ECS ranges of CMIP6 models [26]. This study used a multi-model ensemble (MME) to represent the general projected changes in climate, runoff, and flood, and further quantified the ranges and the uncertainties contributed by GCMs.

2.3. Methodology

This study followed a top-down assessment framework to build climate change scenarios based on RCP-SSP scenarios and GCMs that were then used to drive calibrated and validated hydrological models to obtain the simulated daily runoff. The study period consists of a baseline period (1971–2010) and two future periods (2021–2060 and 2061–2100). The changes in climate, runoff, and flood between the two future periods relative to the baseline were investigated for different timescales. An overview of the approach to carrying out this investigation is presented in Figure S2.

2.3.1. Application of the Hydrological Model

In this study, the Soil Water Assessment Tool (SWAT) was used to simulate the runoff in the Fujiang River basin. The SWAT is a physically based, semi-distributed hydrological model, which has been widely used for a wide range of environmental conditions. The SWAT model has been successfully used for streamflow simulation and projection in the study area [18] and has been proven to perform as well as other hydrological models in the upper Yangtze River basin [11]. Model parameterization was specified using ArcSWAT, an ArcGIS extension, and an interface for SWAT. The Fujiang River basin was divided into 16 sub-basins based on DEM and 113 HRUs based on land-use and soil characteristics. Using sensitivity analysis procedures embedded in SWAT resulted in the six most sensitive parameters for both daily and monthly timescales (Table S1). The extensive manual calibration was undertaken by manually varying the six most sensitive parameters in SWAT for the Fujiang River basin.

The runoff historical simulations were obtained by feeding the CN05.1 daily climate data from 1961 to 2012 into the SWAT model. The SWAT model was calibrated using the discharge observations from 1961 to 1990 and then validated for 1991–2012. The Nash–Sutcliffe simulation efficiency coefficient (E_ns) [27], coefficient of determination (R²), and percentage bias (PBIAS) were used to evaluate the SWAT model’s performance of runoff simulation.

$$E_{ns}=1-\frac{{{\displaystyle \sum }}_{i=1}^n{\left(Q_{obs,i}-Q_{sim,i}\right)}^2}{{{\displaystyle \sum }}_{i=1}^n{\left(Q_{obs,i}-{\overline{Q}}_{obs}\right)}^2},$$

(1)

$$R^2=\frac{{\left[{{\displaystyle \sum }}_{i=1}^n\left(Q_{obs,i}-{\overline{Q}}_{obs}\right)\left(Q_{sim,i}-{\overline{Q}}_{sim}\right)\right]}^2}{{{\displaystyle \sum }}_{i=1}^n{\left(Q_{obs,i}-{\overline{Q}}_{obs}\right)}^2{{\displaystyle \sum }}_{i=1}^n{\left(Q_{sim,i}-{\overline{Q}}_{sim}\right)}^2},$$

(2)

$$PBIAS=100\times \frac{{{\displaystyle \sum }}_{i=1}^n{Q}_{sim,i}-{{\displaystyle \sum }}_{i=1}^n{Q}_{obs,i}}{{{\displaystyle \sum }}_{i=1}^n{Q}_{obs,i}}.$$

(3)

where, $Q_{sim}$ and $Q_{obs}$ are the simulated and observed monthly runoff, respectively,$\overline{Q}$ is the mean monthly runoff, i is the month, and n is the length of the monthly runoff series in the calibration or validation period.

In general, the model simulation is considered acceptable when the E_ns value is greater than 0.5, while R² should exceed 0.6 and PBIAS should be less than ±20% [28]. Furthermore, the performance of the runoff simulation of the SWAT was also compared by the graphical plots, including monthly and daily discharge time series, which reflects the temporal correlations between simulation and observation in different daily and monthly resolutions.

The simulated discharge results showed that the E_ns at the monthly and daily scales in the calibration and validation periods were consistently greater than 0.55, the PBIAS values were less than 2%, and the R² values were greater than 0.77 (

Table 2. Evaluation of the SWAT model’s performance for monthly and daily runoff simulation in the Fujiang River basin.

Timescale	Period	Ens	PBIAS (%)	R2
Monthly	Calibration (1961–1990)	0.91	1.3	0.92
	Verification (1991–2012)	0.92	0.4	0.92
Daily	Calibration (1961–1990)	0.68	−0.2	0.79
	Verification (1991–2012)	0.55	1.0	0.77

). The statistics for monthly runoff simulation were better than the runoff simulation in daily resolution, especially for the E_ns and R². Both the observed and simulated monthly and daily runoff had a high degree of matching to the observation, as shown by the hydrograph and flow duration curve, respectively (Figure 2). In terms of the flow duration curves, the results showed that the simulated river discharge at the Xiaoheba hydrological station closely matched the frequency distributions of the observed discharge records. In summary, the SWAT model has demonstrated good applicability in simulating the streamflow spatial and temporal characteristics of the Fujiang River basin, and the flow simulation results have indicated that the calibrated SWAT model can be used for runoff projections with future inputs from GCMs.

2.3.2. Projection of Climate Change

Climate changes in the Fujiang River basin were quantified by calculating the difference between a baseline of 1971–2010 at annual and monthly scales and that from two future periods for both mean temperature and precipitation, respectively. The annual difference indicates a general change toward wetting or drying, while monthly variations reflect seasonal changes.

2.3.3. Projection of Climate Change Impacts on Runoff

The annual, monthly, and extreme runoff (i.e., floods) were estimated based on the obtained daily streamflow simulation. The impacts of climate change on annual runoff were first assessed by the change in the mean annual gross amount of runoff in terms of the relative changes between the baseline and two future time horizons of 2021–2060 and 2061–2100 under SSP2-4.5 and SSP5-8.5, respectively. The impacts on monthly runoff were assessed by the change in the mean monthly gross amount of runoff, as well as the extreme monthly high/low flow (Q05/Q95), i.e., the monthly flow exceeding 5% or 95% of the entire period. These changes were measured by both the ensemble mean and ranges of 14 simulations for each future period.

The impacts of climate change on flood events were assessed by the changes in frequency and magnitude of flood events. First, we obtained four flood event sequences based on the annual maximum of a continuous 1-, 3-, 7-, and 15-day moving window. Then, changes in the frequency and magnitude of flood events with the current 30-year return value were estimated for the flood event sequences. Note that a return period of 20–50 years has been widely used in quantifying flood hazards and is also an important criterion for flood management in the upper Yangtze River. The generalized extreme value (GEV) distribution model was further applied to estimate flood frequency with the maximum likelihood method for parameter estimation, and the best fit was determined by the Kolmogorov–Smirnov test. Similar to the investigation of climate change impacts on runoff, the changes in flood magnitude and flood return period were measured by both the ensemble mean and ranges of 14 individual GCM simulations.

3. Results

3.1. Projected Changes in Annual and Monthly Temperature and Precipitation

Figure 3 and

Table 3. Projected changes in the ensemble mean and the ranges of climate and runoff in the Fujiang River basin for 2021–2060 and 2061–2100 under 14 GCMs and 2 SSP-RCPs (relative to 1971–2010).

RCPs	2021–2060
	TMP 1 (°C)	PRE 2 (%, mm)	AMR 3 (%, m3/s)	Q05 4 (%, m3/s)	Q95 5 (%, m3/s)
SSP2-4.5	1.9 [1.1, 2.9] 6	4.3 (42.4) [−2.6, 15.2]	4.4 (20.3) [−6.7, 22.6]	5.4 (68.3) [−14.5, 42.7]	12.3 (9.0) [−7.3, 37.3]
SSP5-8.5	2.5 [1.4, 3.9]	5.3 (52.3) [−2.9, 14.9]	6.0 (27.3) [−9.1, 23.3]	5.5 (68.1) [−26.8, 37.3]	21.1 (15.3) [−0.8, 56.6]
RCPs	2061–2100
SSP2-4.5	3.0 [1.9, 4.3]	9.3 (91.9) [−11.6, 24.0]	13.8 (63.4) (63.4) [−14.6, 36.7]	15.5 (217.2) [−14.5, 60.8]	31.6 (23.0) [−15.3, 69.7]
SSP5-8.5	4.8 [3.2, 6.9]	13.9 (137.5) [−4.7, 36.6]	21.5 (99.0) [−2.7, 61.3]	22.1 (305.4) [−11.0, 74.0]	45.5 (33.1) [−10.3, 94.7]

Notes: ¹ Annual mean temperature, ² annual precipitation, ³ annual mean runoff, ⁴ high flow, ⁵ low flow. ⁶ The bracketed values indicate the minimum and maximum percentages for the corresponding parameters.

display the projected changes in annual mean temperature and annual precipitation in the Fujiang River basin. The projected changes in the ensemble mean of the annual mean temperature show a consistent increase, with the 2061–2100 period being more prominent than that of 2021–2060, especially for the high emission scenario. The projected ensemble mean of the annual mean temperature increases by 1.9 and 2.5 °C in 2021–2060, and by 3.0 and 4.8 °C in 2061–2100, under SSP2-4.5 and SSP5-8.5, respectively, with larger ranges by the 14 GCMs in 2061–2100, especially under SSP5-8.5.

The projected ensemble mean of the annual precipitation increases by 4.3% and 5.3% in 2021–2060, and by 9.3% and 13.9% in 2061–2100, under SSP2-4.5 and SSP5-8.5, respectively. The range of changes in annual precipitation projected by the 14 GCMs is similar in 2021–2060 and larger in 2061–2100. The changes of projected annual precipitation show an increase by most of the GCMs, with the period 2061–2100 being wetter than that of 2021–2060, especially for the high emission scenario.

Figure 4 shows the changes in projected monthly mean temperature and precipitation in the Fujiang River basin. For monthly mean temperature, consistent increases are projected for both time periods across the 14 GCMs. The projected changes in monthly temperature show a substantially warmer January and hotter September in 2021–2060, and warmer January and hotter summer and September in 2061–2100.

The projected changes in ensemble mean monthly precipitation show a general increase. Projected ensemble mean monthly precipitation increases significantly in the dry season from October to May, while monthly increases are smaller in the rainy season from June to September. However, the projected changes in monthly precipitation showed a large spread without agreement even in the sign among most GCMs compared with annual mean precipitation.

3.2. Simulated Changes in Annual and Monthly Runoff

Figure 5 shows the simulated changes in annual runoff in the Fujiang River basin. The simulated changes in the ensemble mean of annual runoff show an increase, with the 2061–2100 period being more substantial than that of 2021–2060, especially for SSP5-8.5. The simulated ensemble mean of annual runoff increases by 4.4% and 6.0% in 2021–2060, and by 13.8% and 21.5% in 2061–2100 under SSP2-4.5 and SSP5-8.5, respectively. The range of changes in annual runoff projected by the 14 GCMs is similar in 2021–2060, and substantial in 2061–2100, especially under SSP5-8.5.

The changes in annual runoff simulated by most GCMs show an increase, especially for the period 2061–2100, with a few GCMs (INM-CM4-8, INM-CM5-0, and MPI-ESM1-2-HR) that simulated a decrease in the annual runoff for both emission scenarios for the period 2021–2060, while only MPI-ESM1-2-HR simulated a decrease in the annual runoff for 2061–2100.

Figure 6 displays the simulated changes in ensemble mean and ranges of monthly average runoff for the periods 2021–2060 and 2061–2100. The simulated ensemble mean of monthly runoff shows a larger range in 2061–2100, especially under SSP5-8.5. There are increases in most months in the two time periods, with considerable differences between months. The months with greater increases mainly occur in the early rainy season from April to May.

The change in projected monthly runoff varies across GCMs, and there is no consistency in the direction of change. The MPI-ESM1-2-HR model projects a significant decrease of runoff in the flood season under SSP2-4.5, resulting in a more moderate runoff distribution pattern. CCCma-CanESM5, HadGEM3-GC31-LL, ACESS-ESM1-5, and ACESS-CM2 models project a significant increase of runoff in April and May under SSP5-8.5, resulting in a shift of runoff distribution to the early flood season.

3.3. Projected Changes in Extreme Monthly Runoff

Figure 7 shows the simulated changes in extreme flow (Q05 and Q95) for the ensemble mean and simulated by 14 GCMs under SSP2-4.5 and SSP5-8.5 in 2 time periods in the Fujiang River basin. The projected increases in the ensemble mean of Q95 are larger than those of Q05. The ensemble mean of Q05 is projected to increase by 5.4% and 5.5% in 2021–2060, and by 15.5% and 22.1% in 2061–2100 under SSP2-4.5 and SSP5-8.5, respectively (Table 3). The ensemble mean of Q95 is projected to increase by 12.3% and 21.1% in 2021–2060 and by 31.6% and 45.5% in 2061–2100 under SSP2-4.5 and SSP5-8.5, respectively. The ranges in simulated changes of Q95 and Q05 are similar by the 14 GCMs. Most GCMs show an increase in projected changes in Q05. A few GCMs show a decrease in projected changes in Q95, such as the GFDL-CM4 model, which shows a consistent decrease.

3.4. Projected Changes in Floods

Table 4. Projected flood frequency with current (1971–2010) 30-year return value of 1-, 3-, 7-, and 15-day floods in the Fujiang River basin for 2021–2060 and 2061–2100 under 14 GCMs and 2 SSP-RCPs.

Scenario	Flood	Return Period (Years)
		2021–2060	2061–2100
SSP2-4.5	1-day	30.6 (4.2, 86.6) 1	31.9 (3.5, 101.6)
	3-day	29.7 (4.1, 109.7)	29.5 (3.3, 106.8)
	7-day	27.5 (3.6, 100.3)	22.2 (3.0, 93.3)
	15-day	58.1 (3.2, 226.8)	21.3 (2.6, 79.4)
SSP5-8.5	1-day	27.2 (6.1, 85.8)	17.0 (2.5, 71.1)
	3-day	25.8 (6.2, 82.7)	14.9 (2.7, 59.4)
	7-day	29.2 (5.1, 135.7)	12.3 (2.5, 49.2)
	15-day	32.1 (4.7, 93.5)	11.3 (2.1, 24.4)

Note: ¹ The bracketed values indicate the minimum and maximum values.

lists the projected return periods with a current 30-year return value of 1-, 3-, 7-, and 15-day floods for 2021–2060 and 2061–2100 under SSP2-4.5 and SSP5-8.5. Judging from the ensemble mean, current 1-in-30-year flood events will occur more frequently under future climate conditions for most cases, except for 1-day floods and 15-day floods in specific periods and scenarios. The flood events will occur the most frequently for 2061–2100 under SSP5-8.5, however, will not change so much for 2021–2060 under the two SSP-RCPs. Uncertainties can be found in Table 4 and Figure 8. The return period of current 1-in-30-year floods will be greater than 30 years for some projected cases, even exceeding 100 years for several cases. On the other hand, it is also shown that the larger the floods, the more frequently they are expected to occur for 2061–2100.

Table 5. Projected changes in flood volume of the 30-year return value of 1-, 3-, 7-, and 15-day floods in the Fujiang River basin for 2021–2060 and 2061–2100 under 14 GCMs and 2 SSP-RCPs (relative to 1971–2010).

Scenario	Flood	Changes in 30-Year Return Value (%)
		2021–2060	2061–2100
SSP2-4.5	1-day	5.6 (−29.4, 52.7) 1	13.6 (−30.7, 67.2)
	3-day	8.3 (−22.5, 8.9)	15.9 (−26.1, 69.1)
	7-day	10.4 (−24.4, 65.0)	22.9 (−21.0, 69.3)
	15-day	7.3 (−22.1, 48.8)	22.4 (−14.3, 81.8)
SSP5-8.5	1-day	13.0 (−22.2, 79.4)	28.8 (−23.7, 118.7)
	3-day	13.6 (−21.7, 72.0)	29.0 (−17.9, 110.1)
	7-day	11.4 (−20.4, 55.3)	34.3 (−11.2, 101.7)
	15-day	8.1 (−19.8, 55.0)	32.0 (3.7, 114.5)

Note: ¹ The bracketed values indicate the minimum and maximum values.

shows projected changes in magnitudes of 30-year return values for flood events under future climate conditions. It is projected that they will increase by 5.6~34.3% in terms of the ensemble mean, and the increase rate will be greater for the long-term than for the near-term, and greater under SSP5-8.5 than under SSP2-4.5. That is, the intensity of floods will be the largest for 2061–2100 under SSP5-8.5. However, uncertainties can be found in Table 5. The 1-in-30-year flood events will double in magnitude for some projected cases for 2061–2100 under SSP5-8.5 but will decrease by more than 20% for other cases for 2021–2060 under SSP2-4.5.

4. Discussion

Global warming has changed the global and regional water cycle, altered the seasonal pattern of precipitation and discharge, and increased the frequency and intensity of extreme precipitation. The projected increase in the intensity of extreme precipitation translates to an increase in the frequency and magnitude of pluvial floods. However, the observed impacts and potential future risks of climate change on the hydrological cycle and hydrological extreme events at the basin scale show spatial heterogeneity and uncertainty. Therefore, it is important to discuss the response of the hydrological cycle to climate change in the Fujiang River basin, which is a sensitive area to climate change.

4.1. Hydrological Cycle to Global Warming

The global mean precipitation increases with global warming by about 2–3% per 1 °C with a model-dependent estimated rate. Summer monsoon precipitation is projected to increase for the East Asian monsoon domain, as the consequences of global warming caused substantial changes in the water cycle at both global and regional scales. In this study, the projected changes in annual precipitation to annual temperature increase show about 5.3% and 3.6% per 1 °C under SSP2-4.5 and SSP5-8.5 in the Fujiang River basin, and the projected changes in annual runoff to annual temperature increase are similar to those of annual precipitation, with more sensitivity, about 9.3% and 6.2% per 1 °C under SSP2-4.5 and SSP5-8.5 in Fujiang River basin, respectively (Figure 9). The high emissions scenario caused substantial global warming, however the hydrological sensitivity is lower than the low emissions scenario.

4.2. Runoff to Changing Climate

The attribution of river flow from 7250 observatories around the world during 1971–2010 indicates direct human influence to have a relatively small impact on the global pattern of streamflow trends [29]; furthermore, the trends in annual streamflow have generally followed observed changes in regional precipitation and temperature rise since the 1950s [2]. Climate change probably has a dominant role in the runoff change of the Jialingjiang River basin, with more than half in the flooding season and 85.5% in the dry season [4]. For the Fujiang River, the changes in runoff are mainly attributed to climate change, especially changes in precipitation, and the decrease in runoff during 1951–2012 could be attributed to precipitation change (71.4%) [16].

In this study, the response of streamflow to precipitation is roughly linear: for each 10% increase in precipitation, streamflow increases by 16% (Figure 10). This is larger than the result by Wang et al. [17], which showed that ±10% changes in precipitation can lead to a 13.6% increase or 13.3% decrease in annual runoff in Fujiang River, and similar to Qin et al. [30], who indicated that with every 10% increase in precipitation, simulated streamflow increased by 15% in the upper Yangtze River. Moreover, the response of runoff to the annual mean temperature increase was negative in this study (resulting in a 1.7% decrease in streamflow). This is the same as the results of Qin et al. [30], which indicated that a temperature increase resulted in a 1.7% decrease in runoff in the Yangtze River, less than the result by Wang et al. [16], which showed that 1 °C warming leads to a 2.7% decrease in streamflow in the Fujiang River basin. However, this is different from the study by Birkinshaw et al. [10], with the temperature increase resulting in a 19% decrease in runoff.

4.3. The Link between Rainfall and Flooding

Global and regional changes in precipitation frequency and intensity have been observed over recent decades. The frequency and magnitude of river floods have changed during 1985–2015, with the occurrence of floods increasing 2.5-fold in the past several decades in northern mid-latitudes based on in situ measurement and satellite monitoring. As very extreme precipitation can become a dominant factor for floods, there can be some correspondence in the changes in very extreme precipitation and floods.

Over China, annual precipitation totals changed little from 1973 to 2016, characteristic with precipitation intensity significantly increasing and the number of days with precipitation significantly decreasing [31]. The in situ analyses have indicated an overall increase in the frequency of river floods from 1960 to 2010 in China. The projected changes from CMIP6 show both maximum consecutive five-day precipitation and the heavy precipitation day increase in southern China [25]. This study indicated a general increase in the frequency and magnitude of floods, especially for prolonged floods for the long-term period under the high emissions scenario. However, the relationship between changes in extreme precipitation and floods should be further quantified. Global warming has accelerated snow and glacier melt in mountainous river basins, which has increased the probability of flooding [32]; however, this study did not analyze the impact of climate change on changes in melting water and its further contribution to streamflow and flooding.

4.4. Uncertainties of This Study

This study showed an increase in the projected ensemble mean of annual runoff in the Fujiang River basin. The result is consistent with that of the projected runoff increase in the upstream of the Yangtze River with a small to moderate increase in precipitation, based on both one regional climate model under SRES A1B [33] and 5 GCMs from CMIP5 under RCPs in [6,11]. However, this is different from the slight projected annual runoff decrease in the Fujiang River basin, with a small decrease in precipitation based on the regional climate model under SRES emission scenarios [17].

Many factors introduce uncertainties in the simulation of climate change impacts on hydrological processes, including but not limited to emission scenarios, climate models and hydrological models, data used for model initial conditions, parameterization, and calibration and validation. In this study, discharge from the control hydrological station (Xiaoheba) of the Fujiang River basin was used for SWAT model calibration and validation. The absence of evaluation for the accuracy of spatial pattern simulation for runoff led to this study’s focus on the assessment of changes in the runoff for the basin and cannot further assess the changes in the runoff spatial pattern.

The uncertainty from GCMs is a larger source of climate change impacts than hydrological models [34]. This study employed 14 GCMs under 2 emission scenarios, which tried to capture a large fraction of the full range of CMIP6 and provided a wide range of climate projections in this study area, allowing for drawing more robust conclusions about uncertainty for the hydrological impact assessment. Comparing the uncertainties constrained by the GCMs’ selection, the average error of annual runoff simulation was less than 5% among the four hydrological models [11]. This study used only one hydrological model (SWAT), which has been proven to perform as well as other hydrological models in the upper Yangtze River [11].

5. Conclusions

Water resource availability and flood risk are the key issues in the Fujiang River basin, which will be affected by climate change. This study highlighted the climate changes and changes of runoff and flood in the Fujiang River basin systematically based on the latest GCMs and emission scenarios and quantified the uncertainties constrained by GCMs and SSPs.

This study indicates a generally warmer and wetter climate projected in the Fujiang River basin compared with the reference period 1971–2010. The projected changes in annual mean temperature show consistent warming by 1.9 and 3.0 °C under SSP2-4.5, and by 2.5 and 4.8 °C under SSP5-8.5 in 2021–2060 and 2061–2100, respectively. The annual precipitation will increase by 4.3% and 9.3% under SSP2-4.5, and by 5.3% and 13.9% under SSP5-8.5.

The simulated changes in annual runoff will increase by 4.5% and 13.7% under SSP2-4.5, and by 5.5% and 21.7% under SSP5-8.5. However, there are widespread differences in the direction of changes across GCMs, especially for projected changes in monthly runoff and Q05. The months with large increases in runoff mainly occur in the early rainy season from April to May. The increase in Q95 points to the potential benefit to increase the supply of water resources in the dry season, while the increase in Q05 applies to the increasing flood risks in the flood season in the Fujiang River basin.

Flood volume and frequency with a current 30-year return period for different durations are likely to increase for the Fujiang River basin, with a high risk of SSP5-8.5 for the long-term period. This implies the increasing trend of extreme hydrological events with associated fluvial and pluvial floods, which will increase the flood prevention pressure. Moreover, these events will reduce water resource availability and challenge water resource management at the basin scale.

For future works, first, the key to projecting future changes to runoff and floods is correctly projecting precipitation linked with summer monsoon under a future climate. Second, the key to improving the understanding of the relationship between climate change and runoff (flood) is the reasonable simulation of the response of the hydrological cycle and water balance to climate change. Additionally, more discharge data from the different hydrological stations across the basin for hydrological model calibration and validation will improve the hydrological modeling accuracy.