4.1. Climate Change Prediction under Different Scenarios
4.1.1. Evaluation and Correction of Climate Model Output
Products from RegCM4.6 driven by HadGEM2-ES have been used to simulate the future climate patterns in northwest China. Pan et al. [
35] found that the temperature bias of HadGEM2-ES is generally within ±2.5 °C in the southeast and in the south during the historical period of 1985–2004. This article has evaluated and revised the temperature output values under the RCP4.5 scenario based on the observed meteorological patterns in the upper Tao River Basin based on the historical period of 2007–2018. The deviation between the simulated value and the measured value is exhibited in
Figure 2. The deviation between the simulated value and the measured value is small in summer and large in winter. The maximum temperature deviation can reach 7.5 °C, but the temperature variation is consistent with the observed value (
Figure 2). According to the relationship between the altitude and temperature in each site, the average observed temperature from 2007 to 2018 is about 0.8 °C. In comparison, the simulated annual average temperature and the observed multiyear average are largely consistent under the different climate change scenarios. Corrections through inverse deductions between the linear equation and the vertical decline rate of the temperature, the average temperature, and the maximum/minimum temperature are adjusted separately. Therefore, the corrected data are more in line with the actual situation of the study area and meet the research needs of future climate change scenarios.
4.1.2. Projection of Future Temperature Change in Upper Taohe River Basin
Under the three greenhouse gas emission scenarios, the average annual temperature in the study area shows a consistent warming trend in the future period (
Figure 3). The average temperature in the future study area would be about 2.83 °C, 3.32 °C, and 4.24 °C under RCP2.6, RCP4.5, and RCP8.5 during 2007–2100, and the change rates may be 0.10 °C/10a, 0.20 °C/10a, and 0.52 °C/10a. The average temperature in the 2080s would be about 0.54 °C, 1.14 °C, and 3.44 °C higher than the average temperature in 1956–1997 [
38], which may be consistent with the global warming trends. Overall, the temperature in the upper Taohe River Basin increases with the increase in emission scenarios.
At the same time, the possible future temperature changes in this basin are also analyzed from the two aspects of maximum and minimum temperatures. The annual average maximum and minimum temperature changes in the upper and middle reaches of the Taohe River under the three greenhouse gas emission scenarios were set for four stages: 2007–2018 and the 2020s (2019–2039), 2050s (2040–2079), and 2080s (2080–2099) (
Figure 4).
The maximum temperature and minimum temperature show increasing trends under the three scenarios. The change in the maximum temperature under the RCP2.6 scenario increases steadily, and the increments in the three stages would be 0.13 °C, 0.44 °C, and 0.44 °C higher than those from 2007 to 2018, which is consistent with the change trends in the average temperature. Under the RCP4.5 scenario, the maximum temperature would increase significantly, and the temperature could be 0.63 °C, 1.02 °C, and 1.42 °C higher than in 2007–2018, under the three scenarios, with a maximum change range of 13–17%. The future maximum temperature shows significant changes under the RCP8.5 scenario, which would increase to 0.49 °C, 1.64 °C, and 3.34 °C, which could be higher than the temperature in 2007–2018, with a maximum range of 33%. The results indicate that the highest temperature in the 21st century would increase gradually.
Compared to the maximum temperature variations, the change range of the minimum temperature in different scenarios is consistent with the maximum temperature in the future. Under the RCP2.6 scenario, the minimum temperature in the three stages would increase to 0.26 °C, 0.61 °C, and 0.60 °C, which would be slightly higher than that in 2007–2018, respectively. The change under the RCP4.5 scenario would be significantly enhanced, and the temperatures of the three stages would increase to 0.71 °C, 1.12 °C, and 1.59 °C, which would be higher than those from 2007 to 2018. Under the RCP8.5 scenario, the minimum temperature changed significantly, the minimum temperature would be 0.64 °C, 1.79 °C, and 3.57 °C higher than those from 2007 to 2018. The results indicate that the minimum temperature will increase gradually in the 21st century and that the minimum temperature change would be more significant than the maximum temperature change, which indicates that the minimum temperature changes make an obvious contribution to future regional warming.
4.1.3. Projection of Future Precipitation Change in Upper Taohe River Basin
According to the precipitation data output by the climate model in the historical period (1985~2015) in the northwest region, the simulation effect in the eastern of Qinghai–Tibet Plateau is poor, which may be due to the influence of monsoon circulation on the Qinghai–Tibet Plateau, resulting in a false high-value precipitation center in the climate model. Compared to the observed precipitation from 2007 to 2018 [
35], it was found that the simulated precipitation from the climate model is similarly overestimated in the eastern part of the middle and upper reaches of the Taohe River. Therefore, the seriously overestimated stations were removed and compared with the precipitation changes observed by all of the stations in the region.
Figure 5 shows the future multiyear precipitation changes in the middle and upper reaches of the Taohe River from 2007 to 2099 and compares the not-removed and removed gridded precipitation. The moving average is a reasonable and practical way to analyze the meteorological data. The 5-year moving average has been selected as a default estimator for the annual survey, partially because it seems easy to understand and compute. The results show that the 5-year moving average curve could be presented as the precipitation change trend.
Under the RCP2.6 scenario, the average precipitation would be about 658 mm in the study area in 2020–2099, demonstrating an insignificant decreasing trend (−3.69 mm/10a), and the average precipitation after excluding abnormal stations would be 620 mm, which is closer to the measured precipitation. The precipitation would fluctuate greatly in the 2030s, while the precipitation would achieve its lowest level in the 2080s. Under the RCP4.5 scenario, the average precipitation in the future would be 677.5 mm, and the average precipitation after excluding abnormal stations would be 638.5 mm, demonstrating an insignificant increasing trend (4.97 mm/10a). The precipitation would be the highest in the 2070s (689 mm) and would reach its lowest value in the 2020s (591.3 mm). The future average precipitation under the RCP8.5 scenario would reach 693.2 mm, and the average precipitation after excluding abnormal stations would be 653.9 mm, demonstrating an increasing trend (12.28 mm/10a). The average precipitation would experience an insignificant change before the 2080s, but significant fluctuations would appear after the 2080s.
4.2. Applicability Evaluation of SWAT Model
The results of hydrological process curves during the calibration period and verification period showed that SWAT model can capture the time and flow of the flood peaks in the three hydrological stations well (
Figure 6). The simulated value in the dry season is also consistent with the basic flow of the basin, but the simulation abilities among the three stations are different. The Luqu, Xiabagou, and Minxian hydrological stations are the main outlets of the source area in the Taohe River. Observational data showed that from 1986 to 2014, the runoff from the Luqu, Xiabagou, and Minxian hydrological stations in the Taohe River Basin showed a significant response to precipitation. From 1986 to 2003, the runoff showed a slight decrease along with precipitation. During 2003 to 2011, the runoff decreased as the precipitation decreased; then, from 2011 to 2014, the runoff increased as the precipitation increased (
Figure 6). If we take 1986 to 2010 as the model calibration period and 2011 to 2014 as the model validation period, the SWAT model can capture the variation trends in the three hydrological stations well. The overall runoff simulation values of the determination coefficient
R2, the Nash efficiency coefficient
NSE, and the relative error
Re of the monthly runoff simulation at regular rates fall within a small uncertainty interval (
Table 5).
Luqu station, which is located at the source of the Taohe River Basin, the simulated performance of this station can be easily observed, with R2, NSE, and Re showing monthly runoff simulation rates of 0.79, 0.89, and 2.39%, respectively. In the verification period, R2, NSE and Re also reached 0.89, 0.95, and −8.9%, respectively. The R2, NSE, and Re of the monthly runoff simulation at the Xiabagou station were 0.77, 0.88, and −3.86%, respectively, and 0.89, 0.96, −5.07% during the verification period. The Minxian station serves as the total outlet of the upper Taohe River Basin.
The results showed that monthly runoff simulation rates of R2, NSE, and Re were 0.83, 0.91, −14.6%, respectively, and during the verification period, R2, NSE, and Re also reached 0.87, 0.94, −8.6%, respectively. The above assessment results indicate that the distributed hydrological model SWAT is feasible to simulate runoff in the middle and upper reaches of Taohe River, which lays a foundation for the subsequent study, which is focus on the response of water resources to climate change in the Taohe River Basin.
The calibration and validation results of three on-site observations showed that the SWAT hydrological model is able to produce an acceptable simulation of runoff at a monthly time step, producing reliable results and meeting the research requirements.
4.3. Projection of Future Runoff Change in the Middle and Upper Reaches of the Tao River
Based on the good application of the SWAT model, the annual runoff changes under the three greenhouse gas emission scenarios in the middle and upper reaches of the Taohe River could be predicted from 2020 to 2099 by inputting the corrected RCP temperatures and eliminating abnormal precipitation data from the grids. As shown in
Figure 7, to compare the long-term runoff changes, this research takes the average runoff trends from 1956 to 2014 as the historical period, allowing the runoff changes in different future periods to be analyzed intuitively.
Under the RCP2.6 scenario, the annual average runoff would be about 30.9 × 108 m3, and the overall change trend is similar to the runoff from 2003 to 2014. Over the whole period, the relative maximum and minimum runoff would alternately appear in the 1930s, while the overall minimum runoff would appear in the mid-1980s and would be as low as 17.1 × 108 m3. Comparably, the overall average runoff would be the highest in the 1950s, with an annual average runoff of 35.2 × 108 m3, and would be the lowest in the 1980s at 24.6 × 108 m3. Drought risk might be estimated in the future. Under the RCP4.5 scenario, the annual average runoff would be about 32.5 × 108 m3, and the overall change trend is 15% lower than during the period of 1956−1985. The highest runoff values might be observed during the 2040s and 2070s. Under the RCP8.5 scenario, the annual average runoff would be about 32.2 × 108 m3, with no significant increasing trend being observed. The maximum runoff would appear in the mid-2080s, and the minimum would appear in the mid-2060s and would be as low as 14.8 × 108 m3.
Table 6 shows the projected runoff changes in each season. Under the RCP2.6 scenario, runoff would decrease significantly in summer, and insignificant changes would be observed during the other seasons. Under the RCP4.5 and RCP8.5 scenarios, the runoff in all seasons would show a fluctuating trend. In general, the future runoff in the upper Taohe River Basin would show a decreasing trend in summer and increasing in autumn. In other seasons, significant fluctuations can be observed the future runoff.