1. Introduction
The Yellow River basin (YRB), known as the mother river of the Chinese people, is an important water supply area in northern China that supports and nourishes 100 million people and Chinese civilization (Yang and Liu 2011). However, the climate and environment are diverse and complex in the YRB. It has also been a vulnerable area of water resources in China, especially with regard to floods and droughts (Wu et al. 2018; Wang et al. 2019), water shortages (Xie et al. 2020; Li and Yang 2004), soil erosion (Gao et al. 2019; Peng et al. 2010), ecological decline (Shen et al. 2020), and food security (Mu et al. 2013; Yin et al. 2017). The YRB has a unique physical geographical environment and complicated water problems and is also strongly affected by human activities (Omer et al. 2020; Y. Chen et al. 2020; Yang et al. 2020), attracting more attention and interest (Y. Liu et al. 2020). In 2019, China formally proposed a major national strategy for “ecological conservation and high-quality development of the Yellow River Basin (YB Conservation and Development).” It is the first time the YRB has been mentioned as the height of national governance. Its main goals are to strengthen ecological environment protection; to keep the Yellow River harnessed; to promote efficient use of water resources; and to preserve, inherit, and promote the Yellow River culture. The core element of these goals is water resources. Reasonable prediction of precipitation change in the Yellow River basin is the foundation for dealing with many important climate change problems. For example, it provides essential basic information for flood and drought disaster prevention; provides a relevant policy basis for water shortage control; and provides a basic reference for solving soil erosion, ecological environment improvement, energy and food shortages, and other problems.
Precipitation and runoff are two representative indicators of water (or water resources). Much research has been conducted on these two indicators. The runoff in the YRB (which directly determines the total volume of water resources) showed a decreasing trend, and some reasons for this change could be found from precipitation changes (Zhang et al. 2009; Luo and Wang 2013). From the perspective of the water cycle process, precipitation is the forcing factor and has a decisive influence on the changes in water resources. Precipitation in the YRB had undergone some notable changes based on observed historical data. Zhang et al. (2014), Zhao et al. (2019), and Liang et al. (2015) analyzed the trends of precipitation series at multiple stations in the YRB. They found that the total precipitation in the YRB had a significant decreasing trend. Li et al. (2016) demonstrated that the annual precipitation in the source region of the YRB fluctuated significantly during the past 50 years. The precipitation in spring, summer, and autumn showed increasing trends, while the trend in winter was the opposite. He et al. (2017) and J. Liu et al. (2020) also reached the same conclusion. Another interesting study is that both Wang et al. (2005) and Liu et al. (2019) had paid attention to the changes of the 400-mm isohyet (or the rainfall line) in the YRB. Wang et al. (2005) stated that the 400-mm isohyet tended to move east and south between 1956 and 2000, while Liu et al. (2019) indicated that the 400-mm isohyet was significantly shifted westward and northward in 1961–2015. It proved that the spatial distribution of precipitation in the YRB had indeed changed. Chen and Wang (2019) also found that river connectivity deterioration was related to the decrease in precipitation, leading to water system degradation, sedimentation, and ecological decline. These studies showed significant changes in past precipitation in the YRB, with significant spatial heterogeneity. It affects the environmental elements that directly support our lives to a certain extent.
Besides analyzing historical changes, future precipitation change is also an important issue. It will directly affect the water resources situation and development planning. Global climate models are an effective tool for studying future changes in precipitation, although there are well-known uncertainties. These models simulate the atmospheric circulation under a specific CO2 emission scenario (or a specific socioeconomic development scenario) and output a long sequence of meteorological elements, including the precipitation process. Assessing the precipitation change process in a specific scenario contributes to the formulation of socioeconomic development strategies, which promote future sustainable development, deal with the impact of changing environment and human activities, and handle the relationship between human activities and the environment. Research by Lv et al. (2018) showed that, under the emission scenarios of A1B, A2, and B1, future precipitation in the upper YRB would be higher than that in the historical period, and the spatial distribution of precipitation under the three scenarios was inconsistent. Similarly, Wang et al. (2015) showed that, under A2, B2, and A1B scenarios, the precipitation in the YRB from 2021 to 2050 would increase by 1.28–3.29% relative to 1991–2010. However, Wu et al. (2015) used the RegCM3 high-resolution regional climate model to analyze the average precipitation in the YRB from 2001 to 2030 under the A2 scenario. They found that it would decrease by 2.6% in the future. The research of Li et al. (2012) based on the regional climate model [Providing Regional Climates for Impacts Studies (PRECIS)] showed that the future precipitation in the source area of the YRB would decrease (during the 2010s and 2020s). Based on the Coupled Land Surface and Hydrology Model System (CLHMS) and IPCC AR5 GCMs, Zhu et al. (2016) concluded that under the RCP2.6, RCP4.5, and RCP8.5 scenarios, the future precipitation in the YRB would increase by 2.38, 4.42, and 17.4 mm (10 yr)−1, respectively, but the YRB would remain water deficient. There were specific differences in the analysis results of future precipitation changes in the YRB. It was due to different climate scenarios, climate models, temporal and spatial scales, and even different data processing methods.
The CMIP6 experiment was designed to help understand the response mechanism of the Earth system to climate forcing. It formulated more reasonable different socioeconomic development scenarios, improved system simulation accuracy, and better explained the mechanism of climate change (Eyring et al. 2016). There have been some precipitation-related studies based on CMIP6 models, covering areas such as India (Shrestha et al. 2020), South America (Rivera and Arnould 2020), central Asia (J. Jiang et al. 2020), South Asia (Almazroui et al. 2020), and global land areas (Z. Chen et al. 2020). However, there are few studies on future precipitation in the YRB based on the CMIP6 models. CMIP6 might illustrate a new trend for precipitation change when compared with the current CMIP5 output. With the national strategy of YB Conservation and Development, it is very urgent to know how the precipitation will change in the future Shared Socioeconomic Pathway (SSP) scenarios allowing for the crucial role of water.
In this study, we addressed four quantile-mapping methods (transfer functions) and explored the proper way to correct each model’s projection and the model spread (MS) method to cope with the uncertainty of GCMs based on the CMIP6 output. Then the future precipitation changes in the YRB under different SSP scenarios were analyzed, highlighting more practical support to the future development plan and scientific response strategies to climate change for the YRB.
2. Study area and data
a. Study area
The Yellow River (Fig. 1) is the second-longest river in China and the fifth-largest river globally, with a total length of 5464 km in the mainstream and a drainage area of 79.5 × 104 km2 (the Erdos inner flow area is included). The basin covers the Qinghai–Tibetan Plateau, the Loess Plateau, and the Huang–Huai–Hai Plain. It originates from the Qinghai–Tibetan Plateau flows from west to east through 9 provinces or autonomous regions in Qinghai, Sichuan, Gansu, Ningxia, Inner Mongolia, Shaanxi, Shanxi, Henan, and Shandong. Finally, it flows into the Bohai Sea (Y. Liu et al. 2020). The YRB spans about 1900 km from east to west and about 1100 km from north to south. According to the topography from west to east, the YRB is divided into three parts (Xu et al. 2005). The upper reach of Hekouzhen in Inner Mongolia is a plateau mountain area, where the runoff accounts for about 60% of the whole YRB. The section from Hekouzhen to Taohuayu in Henan Province is the middle YRB, mainly hilly loess areas with severe soil erosion and the leading sediment-yielding area. The reach below Taohuayu is the lower YRB, 3–5 m higher than the ground and so-called a suspended river (Yang et al. 2011). The climate is mainly continental temperate. The winter climate is affected by the monsoon from the northwest, and the summer climate is affected by the monsoon from the southeast. Therefore, the summer is generally warm and humid, and the winter is cold and dry. Simultaneously, the summer precipitation in the YRB accounts for about 70% of the total annual precipitation (Xu et al. 2007; Wang et al. 2007). The altitude of the upper reaches exceeds 4000 m, and the altitude of the middle and lower reaches is 100–2000 m. The average annual precipitation in the basin is about 466 mm, and spatially it is 200–600 mm, but the evaporation of the basin is as high as 700–1800 mm (Zhang et al. 2008; Zhao et al. 2008).
Maps of (top) the YRB (extending from the Bohai in the east to the northern border of China to west central inland China) and three key regions of interest and (bottom) the spatial distribution of annual precipitation. In the inset in the bottom panel, Source is the source region of YRB, Arid is the arid region, and Sediment is the sediment-yielding region.
Citation: Journal of Applied Meteorology and Climatology 61, 10; 10.1175/JAMC-D-22-0022.1
Most basin areas belong to arid or semiarid areas, with a very fragile water resources system. The YRB covers China’s arable land area of approximately 126 × 109 m2, involving more than 114 million people, 9% of China’s population. It is an important water source in the northwest and north China, and it is also an area with significant water shortages in China. In the whole YRB, three regions have attracted attention from all walks of life. They are the source region (the main water-producing areas of the YRB), arid region, and sediment-yielding region (the Loess Plateau on the middle reaches). The annual precipitation in the source region of the Yellow River is about 517 mm, while that in the arid region is only 231 mm (about one-half of YRB) and that in the sediment-yielding region is about 480 mm. Most of the water resources in the YRB are used to support agricultural irrigation, especially in the upper reaches of the Yellow River, where agricultural water consumption accounts for more than 90% of the total water consumption (Kong et al. 2016; Feng et al. 2012). The arid region has the least precipitation in YRB and one of the leading large-scale irrigation areas. Therefore, the precipitation here determines the dependence of agricultural irrigation on upstream water resources and water supply from other areas (Zhang et al. 2014). The unique topography and landscape of the Loess Plateau make the Yellow River with the highest sediment concentration in the world. Studies showed that the sediment yield of the Yellow River is closely related to the precipitation (Dang et al. 2018). In addition to the precipitation of the whole YRB, this study also focuses on the precipitation of three key regions.
b. Observation data and CMIP6 models
A gridded dataset of China’s daily surface precipitation from the China Meteorological Data Service Center (http://data.cma.cn/) with a spatial resolution of 0.5° × 0.5° and a series length of 1961–2014 was collected as the observation series in the YRB. This dataset had a total of 415 grid points in the YRB. This dataset was generated by Zhao et al. (2014) and thoroughly evaluated and checked using cross-validation and error analysis by Zhao and Zhu (2015). This dataset was widely used in climate analysis (Huyan et al. 2017; Xie and Bueh 2015). In general, it had been confirmed that the dataset had less than 10% relative error and less than 10 mm per month when compared with station observations. The dataset was interpolated with thin plate smooth spline method (three times spline) based on over 2400 rain gauge stations over China. The interpolation scheme adopted three variables (longitude, latitude, altitude), and the digital altimetric data was introduced to reduce the effect of elevation changes on interpolation.
This study used the daily precipitation data from the global climate models in the CMIP6 experiments. The data series in 1961–2014 were selected for evaluation, and those from 2015–2100 were selected for projection. Because of differences in each model’s experimental scenario, 12 models were selected for research based on the principle of consistent scenarios (the SSPs are the same) to facilitate follow-up evaluation and analysis. These models were proven to perform well: MRI-ESM2-0, ACCESS-CM2, CNRM-CM6-1, CNRM-ESM2-1, MPI-ESM1-2-HR, FGOALS-g3, ACCESS-ESM1-5, BCC-CSM2-MR, and GFDL-ESM4 ranked in the top 11 of 19 models in the comprehensive evaluation of precipitation simulation ability in YRB (L. Wang et al. 2021). MRI-ESM2-0, MPI-ESM1-2-HR, GFDL-ESM4, ACCESS-CM2, BCC-CSM2-MR, IPSL-CM6A-LR, MIROC6, and CanESM5 had excellent ability to simulate precipitation in China (Yang et al. 2021), and FGOALS-g3, MPI-ESM1-2-HR, MRI-ESM2-0, MIROC6, and CanESM5 performed well in simulations on precipitation in the source region of YRB (Li et al. 2020). The 12 models also participated in the CMIP5 experiments. Detailed information on each model is shown in Table 1. The spatial resolutions of the models were inconsistent with that of the observation, so the bilinear interpolation method was applied to output the model data to 0.5° × 0.5°. We noticed that the variant label of these models was not consistent. The variant label records the forcing conditions for a satisfactory experimental output of the model (https://pcmdi.llnl.gov/CMIP6/Guide/dataUsers.html), but this did not affect our evaluation and application of these models. Some previous studies did not specifically distinguish the variant labels (Almazroui et al. 2020; D. Jiang et al. 2020; Cui et al. 2021; D. Wang et al. 2021; Yang et al. 2021); here, we just kept a clear record.
Basic information of the selected 12 CMIP6 models. The selected SSPs are 126, 245, 370, and 585.
3. Method
a. QM
Quantile mapping (QM) is a correction method based on frequency distribution commonly used for climate data bias correction (Boulard et al. 2017; Cannon et al. 2015; Han et al. 2018). A QM method assumes that the frequency distribution of model and observation should theoretically be the same. The QM method makes the precipitation frequency distribution of the observed data and corrected model data as close as possible by correction, which is achieved by a transfer function. The transfer function’s establishment is based on the quantitative relationship between the quantiles of the model and the observation, and the quantiles correspond to the same cumulative frequency in their respective frequency curves.
QM was usually executed through four steps. First, the two datasets were sorted in ascending order, respectively. These two sets of data are from the observed and model data, and they are precipitation for the same historical period at the same grid point. Second, the empirical frequency of the two datasets was calculated to obtain the two empirical cumulative distribution curves. Third, a set of quantiles were calculated from the empirical frequency curves at a specified set of frequencies. Finally, the above formulas were applied to fit the transfer function parameters between the two sets of quantiles. In the calibration period, we calculated the transfer function parameters for each QM method separately for each model and for each grid. In the validation period, we evaluated the performance of each QM method and selected the most applicable method for each model. For each model, the most applicable QM method was applied at each grid in the future period. In this way, each model was best corrected.
b. Model spread
4. Results
a. Evaluations and corrections
1) The primary deviation of models
The simulation deviation of the average annual precipitation was calculated to investigate each model’s simulation ability on historical precipitation, as shown in Table 2. It can be seen that the deviations of ACCESS-CM2 (ACC), ACCESSES-M1-5 (ACM), BCC-CSM2-MR (BCC), CanESM5 (Can), FGOALS-g3 (FGO), GFDL-ESM4 (GFD), IPSL-CM6A-LR (IPS), MIROC6 (MIR), and MPI-ESM1-2-HR (MPI) were 107.4–376.1 mm in YRB, reaching 23.1%–80.7% of the average annual precipitation (466 mm). Deviations of CNRM-CM6-1 (CNC), CNRM-ESM2-1 (CNE), and MRI-ESM2-0 (MRI) were all about 50 mm, reaching 10.4%–12.6% of the average annual precipitation in the basin. Among all regions, the source region had the greatest deviations (except for Can), and the deviation of all models was between 194.4 and 617.6 mm. In the arid region and the sediment region, the deviation was in −50.6–201.1 mm and −54.0–392.0 mm, respectively.
The simulation deviation relative to the historical annual precipitation (1961–2014).
Figure 2 shows the simulation deviation of historical annual average precipitation calculated on a gridded scale. It visually illustrated each model’s deviation’s spatial distribution characteristics. It is easy to see that all models had larger deviations in the southwest of the basin (mainly overestimated). In contrast, the deviations in the eastern and northern regions of the basin were relatively small. All in all, each model had a large deviation in the historical precipitation simulation, and it is necessary to correct them.
Spatial distribution of simulation deviation to the historical annual average precipitation of each model (1961–2014).
Citation: Journal of Applied Meteorology and Climatology 61, 10; 10.1175/JAMC-D-22-0022.1
2) Performance evaluation of the correction
According to the results in Fig. 2, there was a greater deviation in the local parts than that in the whole YRB. Correcting the models spatially by calculating each grid’s parameter to obtain their respective transfer functions was necessary. The 40 years from 1961 to 2000 were regarded as the calibration period, and the 14 years from 2001 to 2014 were the validation period. This study used monthly data for correction and analysis. Some precipitation data would be corrected to a negative value when it was small enough in some months in the correction process. These negative values were very close to 0, and therefore they were replaced by 0.
The correction effects of different QM methods on each model are listed in Fig. 3. The correction effects of the same method for precipitation vary with different models and regions. The deviation after correction mainly ranged from −83.1 to 174.6 mm. The capabilities of each method in different regions were different, and we finally selected the best methods for each model by a compromise (Table 2). The corrected deviation of each model in every region was −46.4–59.8 mm in each region, indicating that the correction was acceptable.
Deviation of annual average precipitation after correction for the validation period (2001–14). QM1: QM-linear, QM2: QM-exponential, QM2: QM-multiple ratio, and QM4: QM-hybrid. The circle indicates the best method for each model. The rectangle indicates the final method used for each model.
Citation: Journal of Applied Meteorology and Climatology 61, 10; 10.1175/JAMC-D-22-0022.1
Figure 4 showed the spatial distribution of correction effects of different QM methods on each model. Before the correction (Fig. 4a), the spatial pattern of the simulation deviation of each model during the validation period (2001–14) was consistent with the entire historical period (1961–2014; Fig. 2).
Spatial distribution of deviation (a) before and (b) after correction for the validation period (2001–14).
Citation: Journal of Applied Meteorology and Climatology 61, 10; 10.1175/JAMC-D-22-0022.1
To justify the QM, Fig. 5 compares the performance of QM, quantile delta mapping (QDM), and daily bias correction (DBC) for each model, which are commonly used methods for climate model correction (Tong et al. 2021; Chen et al. 2013). The deviations of the annual average precipitation for validation period corrected using the three methods are shown in Fig. 5. The performance of DBC and QM is very similar, while the performance of QMD is poor for a few models. This indicates that among the three BC methods, QM has the better performance and can be used in this study.
Comparison of the calibration capabilities of QM, QDM, and DBC (2001–14).
Citation: Journal of Applied Meteorology and Climatology 61, 10; 10.1175/JAMC-D-22-0022.1
Considering the differences between models, the performance of an ensemble of the 12 models was evaluated for further reducing the deviation and making full use of the capabilities of BC methods. In addition to the equal-weights ensemble method, a weighting scheme that accounts for performance and interdependence (PIW; Knutti et al. 2017) and Bayesian model averaging (BMA; Wang et al. 2017) are considered. The results of PIW and BMA are consistent with the equal-weights ensemble method (see Fig. 6a–c), the equal-weights ensemble can achieve the same or better performance as PIW and BMA. When compared with Fig. 4b, all three ensemble methods improved the underestimation of the source region and the overestimation of the southeast YRB. From the spatial statistics, the deviations in most areas were in −50–50 mm, and the relative deviation was between −10% and 10%. These deviations were even smaller from the whole historical period (Fig. 6d). The ensemble of the models significantly improves the correction effect. In future climate projections, the analyses were based on the equal-weights ensemble.
(a)–(d) Spatial distribution of simulation deviation to historical climatology from the ensemble of the 12 models with the best QM method. For each panel, inset A shows the frequency distribution histogram of deviations and inset B shows the frequency distribution histogram of relative deviations. for both insets A and B, the x axis is deviation (for A) or relative deviation (for B) and the y axis is counts of grids.
Citation: Journal of Applied Meteorology and Climatology 61, 10; 10.1175/JAMC-D-22-0022.1
b. Changes in future precipitation
1) Long-term trends of climatology
The anomaly of annual precipitation under different scenarios is shown in Fig. 7. Under the four SSP scenarios, the future precipitation in all regions was projected to increase. Whether in the whole YBR or key regions, the increasing rates (the linear coefficient) from SSP126 to SSP585 were growing. They were 30–62, 60–103, 84–122, and 134–204 mm (100 yr)−1, respectively. The increasing rate in each region was incremented from SSP126 to SSP585.
Anomaly of future precipitation change under different scenarios (2015–2100) for different regions of interest. The upper and lower boundaries of the shading give the maximum and minimum among models, respectively. The black curve is the historical simulation and future projection of model average; the blue curve is the historical observation. The fitted linear coefficient equations represent the average rate of change of the anomaly.
Citation: Journal of Applied Meteorology and Climatology 61, 10; 10.1175/JAMC-D-22-0022.1
The linear trend of the precipitation changes anomaly (i.e., the linear coefficient in Fig. 7) is shown in Fig. 8. Under the four scenarios, the change rates were all positive numbers, indicating that the future precipitation anomaly changes would increase. The rate of increase was greater in the eastern and southern YRB than in the western and northern parts. Under all scenarios, the rate of increase in the arid region was minimal, and the rate of increase in the source region and sediment-yielding region were about the same.
The change rate of the anomaly in the future precipitation (2015–2100).
Citation: Journal of Applied Meteorology and Climatology 61, 10; 10.1175/JAMC-D-22-0022.1
2) Change of future 30-yr climatology
To understand the future climatology in different periods, we selected two periods (2031–60 and 2061–90) to calculate the 30-yr precipitation, as shown in Fig. 9. In all regions and scenarios, the future precipitation would increase, and the increase in the latter period was greater than that in the previous period. In the first period, the increase under SSP126 and SSP585 was larger than that under SSP245 and SSP 370, while in the latter period, the increase showed an increasing trend from SSP126 to SSP585. In three key regions, the increase in the sediment-yielding region was greatest in both periods. In the first period, the arid region increases more than the source region, while in the latter period, the source region was slightly higher than the arid region.
The future 30-yr precipitation in different regions.
Citation: Journal of Applied Meteorology and Climatology 61, 10; 10.1175/JAMC-D-22-0022.1
Figure 10 shows the spatial pattern of precipitation changes in the two periods. In both periods, the spatial distribution pattern of an anomaly in each scenario was consistent with the change rate of the anomaly of the entire future period (2015–2100; Fig. 8). In the first period, the anomalies under four scenarios were below 100 mm in most parts and below 50 mm in the western YRB. In the second period, the anomalies were higher than the previous period, with most parts above 50 mm and the eastern YRB above 100 mm.
The spatial distribution of future 30-yr precipitation anomaly for (a) 2031–60 and (b) 2061–90.
Citation: Journal of Applied Meteorology and Climatology 61, 10; 10.1175/JAMC-D-22-0022.1
3) Changes in spatial pattern: 400-mm isohyet
As an important geographical dividing line in China, the 400-mm isohyet is the dividing line between semiarid and semihumid areas and the dividing line between agricultural planting, animal husbandry, and vegetation types. Since the 400-mm isohyet in China crosses the center of the YRB (the dividing line between 200 and 400 and 400 and 600 in the bottom panel of Fig. 1), the change of the 400-mm isohyet directly affects the production, life, and ecology of the YRB. On the other hand, the isohyet itself, as an important precipitation characteristic index, can directly reflect the spatial pattern of precipitation. Therefore, the CMIP6 model was used to estimate the change of the 400-mm isohyet and to understand the spatial pattern change of precipitation in YRB.
To understand the capability of the climate models to simulate the 400-mm isohyet, we compared the simulated isohyet with the observed isohyet (Fig. 11). It was easy to see that both the single model (red isohyet) and the model ensemble (green isohyet) could reproduce the historically observed 400-mm isohyet (blue isohyet). However, the model ensemble’s performance was better than any single model. It could coincide with the observed isohyet or be very close in space. The simulation performance of each model was slightly different. For example, the simulation of the northeast YRB of the isohyet in the model ACM and CNE was a bit deviated from the observation, and other models could simulate the observation better. Therefore, we believed that the model ensemble could simulate the 400-mm isohyet or that the models could well reflect the spatial pattern of precipitation.
The simulation of 400-mm isohyet by different models (1961–2014). The blue curve is observation, the green curve is simulation by ensemble, and the red curve is simulation by each model. The former two are identical in each plot.
Citation: Journal of Applied Meteorology and Climatology 61, 10; 10.1175/JAMC-D-22-0022.1
The interannual changes in the location of the observed (blue curve) or the simulated (black curve) isohyet was significant, which would affect our identification of isohyet changes (Fig. 7). This study used interdecadal precipitation to draw the 400-mm isohyet. It could reflect the phase characteristics of the spatial distribution of precipitation and identify its temporal changes.
The 400-mm isohyet would all move to the northwest in the future SSPs scenarios, with the southwest and northeast of the isohyet moving farther than the center (Fig. 12). In terms of temporal scale, the isohyet had the smallest movement in 2031–40 and the most movement in 2081–90 under the four scenarios. The isohyets of other periods were somewhere in between. Among these four scenarios, the largest movement was observed under SSP585, while the other three scenarios showed similar movement.
Interdecadal changes of 400-mm isohyet under different SSPs (2031–90).
Citation: Journal of Applied Meteorology and Climatology 61, 10; 10.1175/JAMC-D-22-0022.1
By carefully observing the isohyets of each decade (it requires some patience), we found that in each scenario, the isohyet had a retreat process in moving into the northwest. The retracement under the four scenarios was in 2061–70, 2061–70, 2051–60, and 2051–60, respectively. After the slight retracement, it continued to move to the northwest. Perhaps this retracement phenomenon had some chance, but on the whole, the isohyet presented a trend of continuous movement to the northwest on a decadal scale.
Although the model had been corrected, the model still had a certain deviation. So, we gave both the observed isohyet and simulated isohyet (black dotted isohyet) as the reference in Fig. 12. The observed isohyet and simulated isohyet here were the same as those in Fig. 11. Therefore, it could be determined that the 400-mm isohyet would move to the northwest in the future period was still valid regardless of whether the reference was the observed isohyet or the simulated isohyet.
Changes in the spatial pattern of precipitation have far-reaching effects, especially in the area where the 400-mm isohyet swings. In future scenarios, as the 400-mm isohyet moved to the northwest, the boundary of the arid region in Fig. 1 (here it refers to the part within the YRB) would also move to the northwest. Based on Figs. 12 and 1, the changing area of the 400-mm isohyet just coincided with the area between the arid region and the sediment-yielding region, which was called the “swing area” here for the time being. The swing area might alternate between wet and dry in the short term, but there would be a tendency to change from dry to wet (relatively wetter than before). The future precipitation in arid regions and swing areas would significantly increase (Figs. 8–10). In YRB, especially in the northern arid region, water resources mainly come from the mainstream of YRB or water supply from other places. Future water resource availability changes in these areas are an important concern, but this needs further study and justification.
c. Uncertainty assessment
Uncertainty in climate projections usually has multiple sources, such as climate natural variation, among-model variability, intramodel uncertainty, the uncertainty of the scenario. This study applied climate models ensemble for future climate projection, the deviation had been corrected and evaluated, and here we mainly consider the among-model uncertainty. As shown in Fig. 7, the range based on all models (the light gray area in 1961–2014) can include the observation (blue line). Therefore, among-model uncertainty can be expressed by the projected range of all models.
1) Relationship between range and MS
In Figs. 3–6, we analyzed the simulation deviation of models to the historical climate in the validation period (2001–14), which was to verify the validity of the adopted correction methods. We analyzed the climate simulation capability of the models for the entire historical period (1961–2014, Table 3 and Fig. 6d). Figure 6d proved the model’s good performance in simulating long-term historical climate. In the whole YRB and three key regions, the simulation deviation was 13.3–18.1 mm, only 2.6%–6.4% of the annual precipitation. Spatially, the simulation deviation was mainly between 0 and 30 mm, which was less than 7% of the annual precipitation. Unlike the results in Fig. 6a, the results in Fig. 6d showed that the models had a better simulation capability on long-term climatology. The model’s capability could also be confirmed from Fig. 11, the 400-mm isohyet based on observation (blue line) and that based on multimodel simulation (green line) were in good consistency. It indicated that the models could reproduce the spatial pattern of historical climatology. In general, the simulation capabilities of climate models were still acceptable, and the uncertainty caused by their deviations was relatively small.
Simulation deviations relative to historical climatology from an ensemble of 12 models (1961–2014).
On the other hand, we used the size of the projected interval to quantify the uncertainty, which was the difference between the maximum and minimum of the 12 model projections (also called the range). The larger the range, the greater the uncertainty, and the smaller the range, the smaller the competition between models. The better the consistency and then the smaller the uncertainty. In comparing Fig. 2 and Fig. 4b, we could see that the range was reduced significantly by correction. To quantify the changes in the ranges between the models, we used the MS index (Fig. 13). When MS was 0, it indicated that the models were the same (no difference), and the larger the MS, the greater the difference. From the spatial distribution of the corrected MS in Fig. 13a, the MS was greater than 25 mm in the source region eastern YRB, while in other areas, it was generally below 20 mm. To investigate the change of MS due to correction, we calculated the reduction of MS by correction. As shown in Fig. 13b, the reduction in the northern YRB was about 65%–75%, and the reduction in most other regions was more than 90%. It showed that the MS was significantly reduced by correction. The difference between the models was reduced, and the consistency between the models improved.
Spatial distribution of MS (a) after correction and (b) its improvement (1961–2014).
Citation: Journal of Applied Meteorology and Climatology 61, 10; 10.1175/JAMC-D-22-0022.1
Although MS is a quantitative indicator, MS cannot intuitively reflect the projection uncertainty. In this regard, we explored the relationship between the MS and range in the historical period. Perhaps the future projection uncertainty could be obtained through the MS. As shown in Fig. 14, we were pleasantly surprised to find that MS and range had an excellent linear relationship, whether on the spatial point scale or the regional scale. We could also learn from the annual MS in Fig. 14 that the range on regions before the correction was mainly 300–800 mm, the range on grids was mainly 100–1500 mm. After correction, the range on regions was mainly 50–300 mm, and that on grids was mainly 0–800 mm. This showed that the projection uncertainty was significantly reduced by correction and showed that the projection uncertainty on the regions was less and more robust than that on grids. When compared with Fig. 13, long-term climate projections were more reliable than short term (both on the annual scale). This could also be confirmed in Fig. 7. In each region, the precipitation simulated by the models (black curve) and the observation (blue curve) was in good agreement in the long run, while there may be a greater deviation in the short term.
The relationship between MS and uncertainty interval (annual from 1961 to 2014). The gray plus signs represent the results of grids, and the blue symbols represent the results of key regions. Each point represents a simulation of annual precipitation for one year, and the results of the four scenarios are aggregated together.
Citation: Journal of Applied Meteorology and Climatology 61, 10; 10.1175/JAMC-D-22-0022.1
2) Uncertainty of future projection
Uncertainty (the upper and lower bounds of the projections) of annual precipitation projections in different regions is shown in Fig. 7. Here we mainly considered the uncertainty of future long-term projections, and the spatial distribution of MS is shown in Fig. 15. Under different SSP scenarios, MS was mainly between 20 and 60 mm on most parts in YRB, the MS on central and western YRB were generally between 20 and 40 mm, and that on partial areas in eastern YRB were higher than 60 mm. The MS in the northern YRB was the largest in SSP585 the smallest in SSP126. In addition, the MS in the eastern YRB was larger than other regions in the four SSPs.
Spatial distribution of MS under different SSPs (2015–2100).
Citation: Journal of Applied Meteorology and Climatology 61, 10; 10.1175/JAMC-D-22-0022.1
Using the linear relationship between MS and range in Fig. 14, we got the uncertainty of long-term precipitation projection, as shown in Fig. 16. It was worth noting that if the future range was directly estimated according to the coefficients in Fig. 14, the spatial distribution characteristics of the obtained range would be completely consistent with Fig. 15. Considering the spatial difference of the relationship between MS and range, we fit the linear relationship on each grid. The spatial distribution characteristics of range in Fig. 16 differed from the MS in Fig. 15, but we believed that the MS in Fig. 16 was more realistic.
Spatial distribution of uncertainty of future projection under different SSPs (2015–2100).
Citation: Journal of Applied Meteorology and Climatology 61, 10; 10.1175/JAMC-D-22-0022.1
In the four SSP scenarios, the range in most areas was between 0 and 50 mm in Fig. 16a, and only the range in some parts of the central and southern YRB exceeded 60 mm. Considering the spatial difference of climatic conditions, we calculated the relative range to the historical annual precipitation (Fig. 16b). The relative range of most areas was below 10%, and the relative range of some parts, such as the source region, northern YRB, was between 10% and 20%. On a spatial scale, the relative range of the source region, the northern part of the arid region, and the local area in the southeast of the sediment-yielding region were larger than the other regions.
5. Discussion
We must admit that the initial deviation of the model was objective and not negligible. Yang et al. (2021) found that the uncorrected models had a simulation deviation of −25%–100% for the annual precipitation in the YRB, which was the same as the results in Table 2 and Fig. 2. D. Jiang et al. (2020) and Cui et al. (2021) came to similar conclusions. These results demonstrated the necessity of correction. However, most studies applied a single correction method to all models (Hong et al. 2021; L. Wang et al. 2021). This study compared the correction ability of four different types of QM methods (or statistical downscaling). Of course, the final correction effect was also related to the model’s performance (i.e., the initial deviation of the model; Fig. 2). The results in Fig. 3 illustrated the applicability difference of correction methods to the models. In the future study, we suggest that appropriate evaluation indexes hat can describe the key climate process should be used according to the research requirements, and the most optimum correction method should be applied to obtain a more reliable projection. Additionally, it was necessary to apply different correction methods to different models in different climate regions to fully exploit the potential of correction methods and models when correcting climate models in large regions.
As well as deviations, uncertainty in climate models was an unavoidable issue. It was a common method to estimate uncertainty by using multimodel projection ranged, and this method was also adopted in this study. In terms of process, the maximum and minimum values in multimodel projection were used as uncertainty range in this study. As shown in Fig. 7, the uncertainty difference between different scenarios was small, while the uncertainty difference between different regions was large. In previous studies, semiquantitative indicators were usually used to quantify the projection uncertainty, such as variance (Zhang and Chen 2021) and model spread (Huang et al. 2013). But this kind of method could not directly give the uncertainty range. Aiming at this limitation, this study tried to convert MS into a quantitative indicator while applying MS to the semiquantitative evaluation of uncertainty. The relationship between MS and deviation, model average, and range were explored during the study, and the results showed that only range and MS had a good relationship (Fig. 14). On the one hand, this work provided a new idea for quantifying uncertainty, to realize the quantitative evaluation of uncertainty by constructing the relationship between the uncertainty evaluation index (usually semiquantitative) and the uncertainty range.
On the other hand, MS describes the competitive relationship (or consistency) between models. It can be seen from the relationship between MS and range that the more competitive the models were, the greater the uncertainty of the projection would be. This fully demonstrated that the differences between models directly affected the reliability of the projection, and more work needs to be done in the future on the theoretical comparison between models. In the study of the ensemble of more models, it is necessary to focus on the maximum and minimum of the projections, such as reducing their weights or eliminating them. It may be expected to improve the reliability of the projection further. Admittedly, both the MS and range can be calculated directly through the models. However, it is really because of the comparison and the excellent relationship between the two that we end up with valuable information. However, this work also had some limitations. Figure 16 showed that scenario uncertainty (with correction) could not be ignored, and the uncertainty range given did not have the probability attribute. It was difficult to clarify the probability distribution of the projection uncertainty. If the probabilistic methods are applied to quantifying uncertainty, more valuable information can be obtained for decision-makers.
6. Conclusions
This study used four quantile-mapping methods to correct the models based on the gauged precipitation data in the Yellow River basin and the precipitation data from 12 CMIP6 models. The evaluation included the long-term trend of precipitation, climatology in different periods in the future, and the changes in the spatial pattern of precipitation. The uncertainty of the projection was quantified by using the relationship between MS and the range of the projection from multimodels. The main conclusions were as follows:
-
The deviation of 12 models in the simulation of precipitation in the YRB was 48.4–376.1 mm. They were relatively large in the southwest of the basin. Four quantile-mapping methods employed had an excellent performance on the annual average precipitation of YRB, reducing the simulation deviation of YRB and three key regions to less than 18.1 mm.
-
The precipitation anomaly under all four scenarios would increase. The increasing trend was more significant in the western and northern parts of the YRB than in other parts. The 400-mm isohyet would continue to move to the northwest in space, and the arid region within the YRB would shrink to the northwest.
-
Using MS to describe the differences between models quantitatively, the corrected MS in historical and future periods had been significantly reduced by 85.9%–94.6%, which significantly improved the reliability of precipitation projection. There was an excellent linear relationship between MS and the projection range, which was used to quantify the uncertainty of projection. The projection range in most YRB would be within 50 mm (about 10% of climatology).
The downscaling method used in this study was a statistical downscaling method. The advantage of this method was that it was convenient to apply and effective, but it lacked a physical mechanism. Although the downscaling effect in this study had been satisfactory, if the dynamic downscaling methods based on physical mechanism were adopted in future studies, more reliable downscaling results should be obtained. On the other hand, this study applied 12 climate models, which gave a certain range in estimated uncertainty. In the next plan, we will apply more climate models and probabilistic methods in the uncertainty assessment of this study to give an uncertainty range with probability distribution properties. Such results are more valuable for decision-makers and policy makers.
Acknowledgments.
This work was supported by grants from the National Key Research and Development Program of China (2018YFC1508104), the National Natural Science Foundation of China (52079079, 51779146, and 51809174), the Natural Science Foundation of Jiangsu Province (BK20191129), and the Natural Science Foundation of Zhejiang Province (2021C03017).
Data availability statement.
Publicly available datasets were analyzed in this study. These data can be found online (daily surface precipitation: http://data.cma.cn/; CMIP6: https://esgf-node.llnl.gov/search/cmip6).
REFERENCES
Almazroui, M., S. Saeed, F. Saeed, M. N. Islam, and M. Ismail, 2020: Projections of precipitation and temperature over the South Asian countries in CMIP6. Earth Syst. Environ., 4, 297–320, https://doi.org/10.1007/s41748-020-00157-7.
Boulard, D., and Coauthors, 2017: Bias correction of dynamically downscaled precipitation to compute soil water deficit for explaining year-to-year variation of tree growth over northeastern France. Agric. For. Meteor., 232, 247–264, https://doi.org/10.1016/j.agrformet.2016.08.021.
Cannon, A. J., S. R. Sobie, and T. Q. Murdock, 2015: Bias correction of GCM precipitation by quantile mapping: How well do methods preserve changes in quantiles and extremes? J. Climate, 28, 6938–6959, https://doi.org/10.1175/JCLI-D-14-00754.1.
Chen, J., F. P. Brissette, D. Chaumont, and M. Braun, 2013: Performance and uncertainty evaluation of empirical downscaling methods in quantifying the climate change impacts on hydrology over two North American river basins. J. Hydrol., 479, 200–214, https://doi.org/10.1016/j.jhydrol.2012.11.062.
Chen, Y., and Y. Wang, 2019: Changes in river connectivity indexes in the lower Yellow River between 1960 and 2015. River Res. Appl., 35, 1377–1386, https://doi.org/10.1002/rra.3420.
Chen, Y., and Coauthors, 2020: Sustainable development in the Yellow River basin: Issues and strategies. J. Clean. Prod., 263, 121223, https://doi.org/10.1016/j.jclepro.2020.121223.
Chen, Z., T. Zhou, L. Zhang, X. Chen, W. Zhang, and J. Jiang, 2020: Global land monsoon precipitation changes in CMIP6 projections. Geophys. Res. Lett., 47, e2019GL086902, https://doi.org/10.1029/2019GL086902.
Cui, T., C. Li, and F. Tian, 2021: Evaluation of temperature and precipitation simulations in CMIP6 models over the Tibetan Plateau. Earth Space Sci., 8, e2020EA001620, https://doi.org/10.1029/2020EA001620.
Dang, S., X. Liu, X. Li, M. Yao, and D. Zhang, 2018: Changes in different classes of precipitation and the impacts on sediment yield in the Hekouzhen-Longmen region of the Yellow River basin, China. Adv. Meteor., 2018, 3537512, https://doi.org/10.1155/2018/3537512.
Eyring, V., S. Bony, G. A. Meehl, C. A. Senior, B. Stevens, R. J. Stouffer, and K. E. Taylor, 2016: Overview of the Coupled Model Intercomparison Project Phase 6 (CMIP6) experimental design and organization. Geosci. Model Dev., 9, 1937–1958, https://doi.org/10.5194/gmd-9-1937-2016.
Feng, K., Y. L. Siu, D. Guan, and K. Hubacek, 2012: Assessing regional virtual water flows and water footprints in the Yellow River basin, China: A consumption based approach. Appl. Geogr., 32, 691–701, https://doi.org/10.1016/j.apgeog.2011.08.004.
Gao, P., Y. Wang, P. Li, G. Zhao, W. Sun, and X. Mu, 2019: Land degradation changes in the Yellow River Delta and its response to the streamflow‐sediment fluxes since 1976. Land Degrad. Dev., 29, 3212–3220, https://doi.org/10.1002/ldr.3049.
Han, Z. Y., Y. Tong, X. J. Gao, and Y. Xu, 2018: Correction based on quantile mapping for temperature simulated by the RegCM4. Climate Change. Res., 14, 331–340, https://doi.org/10.12006/j.issn.1673-1719.2017.156.
He, Y., P. Tian, X. Mu, P. Gao, G. Zhao, F. Wang, and P. Li, 2017: Changes in daily and monthly rainfall in the Middle Yellow River, China. Theor. Appl. Climatol., 129, 139–148, https://doi.org/10.1007/s00704-016-1763-4.
Hong, J., K. Javan, Y. Shin, and J. Park, 2021: Future projections and uncertainty assessment of precipitation extremes in Iran from the CMIP6 ensemble. Atmosphere, 12, 1052, https://doi.org/10.3390/atmos12081052.
Huang, D., J. Zhu, Y. Zhang, and A. Huang, 2013: Uncertainties on the simulated summer precipitation over Eastern China from the CMIP5 models. J. Geophys. Res. Atmos., 118, 9035–9047, https://doi.org/10.1002/jgrd.50695.
Huyan, L., J. Li, S. Zhao, C. Sun, D. Dong, T. Liu, and Y. Zhao, 2017: The impact of layer perturbation potential energy on the East Asian summer monsoon. J. Climate, 30, 7087–7103, https://doi.org/10.1175/JCLI-D-16-0729.1.
Jiang, D., D. Hu, Z. Tian, and X. Lang, 2020: Differences between CMIP6 and CMIP5 models in simulating climate over China and the East Asian monsoon. Adv. Atmos. Sci., 37, 1102–1118, https://doi.org/10.1007/s00376-020-2034-y.
Jiang, J., T. Zhou, X. Chen, and L. Zhang, 2020: Future changes in precipitation over central Asia based on CMIP6 projections. Environ. Res. Lett., 15, 054009, https://doi.org/10.1088/1748-9326/ab7d03.
Knutti, R., J. Sedláček, B. M. Sanderson, R. Lorenz, E. M. Fischer, and V. Eyring, 2017: A climate model projection weighting scheme accounting for performance and interdependence. Geophys. Res. Lett., 44, 1909–1918, https://doi.org/10.1002/2016GL072012.
Kong, D., C. Miao, J. Wu, and Q. Duan, 2016: Impact assessment of climate change and human activities on net runoff in the Yellow River basin from 1951 to 2012. Ecol. Eng., 91, 566–573, https://doi.org/10.1016/j.ecoleng.2016.02.023.
Li, C., and Z. Yang, 2004: Natural runoff changes in the Yellow River basin. J. Geogr. Sci., 14, 427–436, https://doi.org/10.1007/s11442-009-0427-z.
Li, J., Y. Zhao, and Z. Tang, 2020: Projection of future summer precipitation over the Yellow River basin: A moisture budget perspective. Atmosphere, 11, 1307, https://doi.org/10.3390/atmos11121307.
Li, L., H. Shen, S. Dai, J. Xiao, and X. Shi, 2012: Response of runoff to climate change and its future tendency in the source region of Yellow River. J. Geogr. Sci., 22, 431–440, https://doi.org/10.1007/s11442-012-0937-y.
Li, Q., M. Yang, G. Wan, and X. Wang, 2016: Spatial and temporal precipitation variability in the source region of the Yellow River. Environ. Earth Sci., 75, 594, https://doi.org/10.1007/s12665-016-5583-8.
Liang, K., S. Liu, P. Bai, and R. Nie, 2015: The Yellow River basin becomes wetter or drier? The case as indicated by mean precipitation and extremes during 1961–2012. Theor. Appl. Climatol., 119, 701–722, https://doi.org/10.1007/s00704-014-1138-7.
Liu, J., L. Jin, C. Huang, and L. Zhang, 2020: Variations of seasonal precipitation in the Yellow River basin and its relationship to general circulation and SST. Proc. IAHS, 383, 5–12, https://doi.org/10.5194/piahs-383-5-2020.
Liu, Y., Z. Yang, P. Lin, Z. Zheng, and S. Xie, 2020: Comparison and evaluation of multiple land surface products for the water budget in the Yellow River basin. J. Hydrol., 584, 124534, https://doi.org/10.1016/j.jhydrol.2019.124534.
Liu, Y., X. Tang, Z. Sun, J. Zhang, G. Wang, J. Jin, and G. Wang, 2019: Spatiotemporal precipitation variability and potential drivers during 1961–2015 over the Yellow River basin, China. Weather, 74 (S1), S32–S39, https://doi.org/10.1002/wea.3517.
Luo, L., and Z. Wang, 2013: Changes in hourly precipitation may explain the sharp reduction of discharge in the middle reach of the Yellow River after 2000. Front. Environ. Sci. Eng., 7, 756–768, https://doi.org/10.1007/s11783-013-0563-7.
Lv, Z. Y., J. X. Mu, D. H. Yan, and T. L. Qin, 2018: Spatial and temporal variability of precipitation in the context of climate change: A case study of the upper Yellow River basin, China. IOP Conf. Ser. Earth Environ. Sci., 191, 012141, https://doi.org/10.1088/1755-1315/191/1/012141.
Mu, J., Q. Liu, H. F. Gabriel, D. Xu, J. Xu, C. Wu, and H. Ren, 2013: The impacts of climate change on water stress situations in the Yellow River basin, China. Irrig. Drain., 62, 545–558, https://doi.org/10.1002/ird.1784.
Omer, A., N. A. Elagib, Z. Ma, F. Saleem, and A. Mohammed, 2020: Water scarcity in the Yellow River basin under future climate change and human activities. Sci. Total Environ., 749, 141446, https://doi.org/10.1016/j.scitotenv.2020.141446.
Peng, J., S. Chen, and P. Dong, 2010: Temporal variation of sediment load in the Yellow River basin, China, and its impacts on the lower reaches and the river delta. Catena, 83, 135–147, https://doi.org/10.1016/j.catena.2010.08.006.
Rivera, J. A., and G. Arnould, 2020: Evaluation of the ability of CMIP6 models to simulate precipitation over southwestern South America: Climatic features and long-term trends (1901–2014). Atmos. Res., 241, 104953, https://doi.org/10.1016/j.atmosres.2020.104953.
Shen, W., Z. Zheng, Y. Qin, and Y. Li, 2020: Spatiotemporal characteristics and driving force of ecosystem health in an important ecological function region in China. Int. J. Environ. Res. Public Health, 17, 5075, https://doi.org/10.3390/ijerph17145075.
Shrestha, A., M. M. Rahaman, A. Kalra, R. Jogineedi, and P. Maheshwari, 2020: Climatological drought forecasting using bias corrected CMIP6 climate data: A case study for India. Forecasting, 2, 59–84, https://doi.org/10.3390/forecast2020004.
Tong, Y., X. Gao, Z. Han, and Y. Xu, 2017: Bias correction of daily precipitation simulated by RegCM4 model over China. Chin. J. Atmos. Sci., 41, 1156–1166.
Tong, Y., X. Gao, Z. Han, Y. Xu, and F. Giorgi, 2021: Bias correction of temperature and precipitation over China for RCM simulations using the QM and QDM methods. Climate Dyn., 57, 1425–1443, https://doi.org/10.1007/s00382-020-05447-4.
Wang, D., J. Liu, W. Shao, C. Mei, X. Su, and H. Wang, 2021: Comparison of CMIP5 and CMIP6 multi-model ensemble for precipitation downscaling results and observational data: The case of Hanjiang River basin. Atmosphere, 12, 867, https://doi.org/10.3390/atmos12070867.
Wang, F., Z. Wang, H. Yang, Y. Zhao, Z. Zhang, Z. Li, and Z. Hussain, 2019: Copula-based drought analysis using standardized precipitation evapotranspiration index: A case study in the Yellow River basin, China. Water, 11, 1298, https://doi.org/10.3390/w11061298.
Wang, G., and Coauthors, 2015: Impacts of climate change on water resources in the Yellow River basin and identification of global adaptation strategies. Mitigation Adapt. Strategies Global Change, 22, 67–83, https://doi.org/10.1007/s11027-015-9664-x.
Wang, H., D. Yan, D. Qin, and J. Wang, 2005: A study of the spatial shift of 400MM-rainfall contours in the Yellow River basin during recent 50 years. Adv. Earth Sci., 20, 649–655.
Wang, H., Z. Yang, Y. Saito, J. P. Liu, X. Sun, and Y. Wang, 2007: Stepwise decreases of the Huanghe (Yellow River) sediment load (1950–2005): Impacts of climate change and human activities. Global Planet. Change, 57, 331–354, https://doi.org/10.1016/j.gloplacha.2007.01.003.
Wang, L., J. Zhang, Z. Shu, Y. Wang, Z. Bao, C. Liu, X. Zhou, and G. Wang, 2021: Evaluation of the ability of CMIP6 global climate models to simulate precipitation in the Yellow River basin, China. Front. Earth Sci, 9, 751974, https://doi.org/10.3389/feart.2021.751974.
Wang, W. Y., Y. Ding, Q. Shao, J. Xu, X. Jiao, Y. Luo, and Z. Yu, 2017: Bayesian multi-model projection of irrigation requirement and water use efficiency in three typical rice plantation region of China based on CMIP5. Agric. For. Meteor., 232, 89–105, https://doi.org/10.1016/j.agrformet.2016.08.008.
Wu, H., X. Li, and H. Qian, 2018: Detection of anomalies and changes of rainfall in the Yellow River basin, China, through two graphical methods. Water, 10, 15, https://doi.org/10.3390/w10010015.
Wu, Z., H. Xiao, G. Lu, and J. Chen, 2015: Assessment of climate change effects on water resources in the Yellow River basin, China. Adv. Meteor., 2015, 816532>, https://doi.org/10.1155/2015/816532.
Xie, P., L. Zhuo, X. Yang, H. Huang, X. Gao, and P. Wu, 2020: Spatial-temporal variations in blue and green water resources, water footprints and water scarcities in a large river basin: A case for the Yellow River basin. J. Hydrol., 590, 125222, https://doi.org/10.1016/j.jhydrol.2020.125222.
Xie, Z., and C. Bueh, 2015: Different types of cold vortex circulations over Northeast China and their weather impacts. Mon. Wea. Rev., 143, 845–863, https://doi.org/10.1175/MWR-D-14-00192.1.
Xu, Z., K. Takeuchi, H. Ishidaira, and C. Liu, 2005: An overview of water resources in the Yellow River basin. Water Int., 30, 225–238, https://doi.org/10.1080/02508060508691863.
Xu, Z. X., J. Y. Li, and C. M. Liu, 2007: Long-term trend analysis for major climate variables in the Yellow River basin. Hydrol. Processes, 21, 1935–1948, https://doi.org/10.1002/hyp.6405.
Yang, Z., and Q. Liu, 2011: Response of streamflow to climate changes in the Yellow River basin, China. J. Hydrometeor., 12, 1113–1126, https://doi.org/10.1175/JHM-D-10-05004.1.
Yang, Z., Q. Liu, and B. Cui, 2011: Spatial distribution and temporal variation of reference evapotranspiration during 1961–2006 in the Yellow River basin, China. Hydrol. Sci. J., 56, 1015–1026, https://doi.org/10.1080/02626667.2011.590810.
Yang, X., W. Sun, X. Mu, P. Gao, and G. Zhao, 2020: Run-off affected by climate and anthropogenic changes in a large semi-arid river basin. Hydrol. Processes, 34, 1906–1919, https://doi.org/10.1002/hyp.13702.
Yang, X., B. T. Zhou, Y. Xu, and Z.-Y. Han, 2021: CMIP6 evaluation and projection of temperature and precipitation over China. Adv. Atmos. Sci., 38, 817–830, https://doi.org/10.1007/s00376-021-0351-4.
Yin, Y., Q. Tang, X. Liu, and X. Zhang, 2017: Water scarcity under various socio-economic pathways and its potential effects on food production in the Yellow River basin. Hydrol. Earth Syst. Sci., 21, 791–804, https://doi.org/10.5194/hess-21-791-2017.
Zhang, Q., C.-Y. Xu, Z. Zhang, G. Ren, and Y. D. Chen, 2008: Climate change or variability? The case of Yellow River as indicated by extreme maximum and minimum air temperature during 1960–2004. Theor. Appl. Climatol., 93, 35–43, https://doi.org/10.1007/s00704-007-0328-y.
Zhang, Q., C.-Y. Xu, and T. Yang, 2009: Variability of water resource in the Yellow River basin of past 50 years, China. Water Resour. Manage., 23, 1157–1170, https://doi.org/10.1007/s11269-008-9320-2.
Zhang, Q., J. Peng, V. P. Singh, J. Li, and Y. D. Chen, 2014: Spatio-temporal variations of precipitation in arid and semi-arid regions of China: The Yellow River basin as a case study. Global Planet. Change, 114, 38–49, https://doi.org/10.1016/j.gloplacha.2014.01.005.
Zhang, S., and J. Chen, 2021: Uncertainty in projection of climate extremes: A comparison of CMIP5 and CMIP6. J. Meteor. Res., 35, 646–662, https://doi.org/10.1007/s13351-021-1012-3.
Zhao, F. F., Z. X. Xu, J. X. Huang, and J. Y. Li, 2008: Monotonic trend and abrupt changes for major climate variables in the headwater catchment of the Yellow River basin. Hydrol. Processes, 22, 4587–4599, https://doi.org/10.1002/hyp.7063.
Zhao, Y., and J. Zhu, 2015: Assessing quality of grid daily precipitation datasets in China in recent 50 years. Plateau Meteor., 34, 50–58.
Zhao, Y., J. Zhu, and Y. Xu, 2014: Establishment and assessment of the grid precipitation datasets in China for recent 50 years. J. Meteor. Sci., 34, 414–420.
Zhao, Y., X. Xu, W. Huang, Y. Wang, Y. Xu, H. Chen, and Z. Kang, 2019: Trends in observed mean and extreme precipitation within the Yellow River basin, China. Theor. Appl. Climatol., 136, 1387–1396, https://doi.org/10.1007/s00704-018-2568-4.
Zhou, T., and R. Yu, 2006: Twentieth-century surface air temperature over China and the globe simulated by coupled climate models. J. Climate, 19, 5843–5858, https://doi.org/10.1175/JCLI3952.1.
Zhu, Y., Z. Lin, J. Wang, Y. Zhao, and F. He, 2016: Impacts of climate changes on water resources in Yellow River basin, China. Procedia Eng., 154, 687–695, https://doi.org/10.1016/j.proeng.2016.07.570.
