Listen

Most previous studies on attributing climate change and human activities to streamflow variation focused only on an annual scale, while few attempts were devoted to shorter time scales, such as seasonal or monthly scales. In this study, a multi-scalar framework for attribution analysis was constructed by integrating hydrological modeling into the Budyko-based decomposition method. Then, the multi-scalar framework was employed in the source region of the Yellow River (SRYR) as a case study. Moreover, the quantification results are compared with the simulation of a monthly water balance model (i.e., abcd model). The main results and findings indicate: (1) The mutation point of the annual streamflow is 1991 in the SRYR, and the monthly ABCD hydrological model achieves good performance in streamflow simulation for the prechange period and postchange period; (2) The contribution of climate change to streamflow variation is 65.99%, 37.50%, and 70.86% for annual, dry season, and wet season, respectively, indicating that climate change is the dominant factor controlling streamflow variation for the annual and wet season, while human activities play a major role in streamflow variation for dry season; (3) The contributions of climate change and human activities show obvious dynamic and seasonal characteristics, which are strongly affected by the transition between dry and wet seasons. Generally, the findings of this study provide valuable references for local reasonable water resource planning and management to make timely and appropriate use of the water supply.

Listen

  • A new multi-scalar framework is proposed for attribution of streamflow variation.

  • The Budyko-based decomposition method is first extended to a monthly scale.

  • The ABCD model is selected for watershed hydrological modeling.

  • The quantification of climate and human impacts is verified by the hydrologic simulation approach.

attribution analysis, budyko-based decomposition method, monthly ABCD model, multi-scalar framework, streamflow
Listen

With global warming and the increasing impact of human activities, the spatial and temporal distribution of water resources in river basins has changed significantly, especially the significant decrease in river streamflow (Jiang et al. 2015; Marisol & Hugo A 2021). Climate change affects streamflow processes through meteorological factors such as precipitation, temperature, and sunshine. (Charles et al. 2000; Piao et al. 2010; Guo et al. 2020). The meteorological factors that can be described quantitatively in hydrology are mainly precipitation and evaporation. The impact of human activities is mainly reflected in land use, water conservancy projects, and direct human access to water (Wang 2014). Studying the impacts of climate change and human activities on streamflow, separating and quantifying their contributions to streamflow variation can help understand the relationship between hydrology and changing environment on the one hand and provide a scientific basis for water resources management on the other hand (Chang et al. 2016).

There are two ways widely used to quantify the impacts of climate change and human activities on streamflow: the hydrologic simulation method and the Budyko-based approach. The hydrologic simulation method separates the relative contributions of climate change and human activities to streamflow variation by comparing the differences between observed and simulated streamflow by hydrological models (Dey & Mishra 2017). Onstad & Jamieson (1970) were the earliest to use the hydrologic simulation to predict the impact of land use or cover change on streamflow change. Since then, an increasing number of scholars have conducted related studies at home and abroad. For example, Chang et al. (2016) analyzed the causes of streamflow variability in Jing River basins using the TOPMODEL and VIC hydrological models, respectively. Narsimlu et al. (2013) used the SWAT hydrological model to distinguish the climate change and human activities in different watershed subdivisions. The hydrological model can accurately describe hydrological variables and their evolution rules. This method has a good physical basis, but there are inevitable uncertainties in model structure and parameters, which would reduce the credibility of the results. Moreover, the data and information required to calibrate hydrological models are often difficult, especially for distributed watershed hydrological models with high data requirements. Many river basins have no data and cannot be simulated by hydrological models (Sivapalan 2003).

In contrast, the Budyko-based approach has strengths in reducing uncertainty due to the relatively low requirements for parameter selection and model construction. Moreover, it has relatively low data requirements, which include the climate factors under the background of the basin and the parameters reflecting the characteristics of the underlying surface of the basin (Liu et al. 2017). Thus, the Budyko hypothesis method is an ideal method for assessing the impacts of climate change and human activities on streamflow variation. Wang et al. (2018) distinguished the contribution rate of climate change, frozen soil degradation, and land type change on annual runoff change in SRYR from 1965 to 2015. The results show that the direct impact of climate change and the impact of frozen ground degradation can explain 55 and 33%, respectively, of the streamflow decrease for the entire SRYR from Period 1 (1965–1989) to Period 2 (1990–2003). Yan et al. (2020) used a Budyko-type equation for separating the impacts of climate and vegetation change on annual streamflow variation in SRYR from 1961 to 2015. However, this method assumes negligible changes in soil water storage and groundwater at annual or multi-year time scales. Therefore, it is impossible to distinguish streamflow variation influences at fine time scales.

To bridge the aforementioned research gaps, a simple and efficient means to quantitatively distinguish the main influencing factors of streamflow variability is urgently needed. Wang & Hejazi (2011) proposed the decomposition method based on a conceptual hydrological model, i.e., the Budyko equation, for quantifying the contributions of climate change and human activities to mean annual streamflow explicitly. This method is intuitive in structure, easy to use, does not require much hydro-meteorological information, and can be used in areas where no information is available. Chen et al. (2013) defined the difference between rainfall and water storage change as effective precipitation to extend the Budyko hypothesis to the seasonal scale. Du et al. (2016) presented a method for explicitly accounting for the water storage variation within the Budyko equation and demonstrated that the method could significantly improve performance in arid regions or over shorter time scales under an unsteady state. Therefore, the variation in water storage should be regarded as an important part of the steady-state assumption of the Budyko hypothesis over shorter time scales. The monthly ABCD model proposed by Thomas & Harold (1981) has only four parameters a, b, c, and d, but explicitly includes the main hydrological processes and can accurately simulate changes in basin water storage. In recent years, the monthly ABCD model has been widely used in the analysis of hydrological processes at annual and monthly time scales (Wang & Zhou 2016; Xin et al. 2019; Li et al. 2020). However, the physical rationale and validity of the ABCD model deserve further applications. How to connect the ABCD model and the Budyko equation is an important issue for understanding hydrological processes across different time scales.

This study aims to quantify the contributions of climate change and human activities to streamflow variation at different time scales. Firstly, the decomposition method based on the Budyko-type equation is extended to the monthly scale by introducing the monthly dryness index and evaporation ratio. The monthly Budyko-based decomposition method has to consider the variation of water storage, but there is generally no observational information on water storage and actual evaporation. To do this, the monthly ABCD model is applied to simulating the monthly water storage and actual evaporation, which are further aggregated to the seasonal and annual time scales to construct the multi-scalar framework for attribution analysis. For the purpose of validation, the proposed multi-scalar framework is employed in the source region of the Yellow River (SRYR) as a case study, and the quantification results are also compared with those derived from hydrologic simulation method based on the monthly ABCD model.

Listen

Study area

Listen
The SRYR refers to the Yellow River basin above the Tangnahai hydrological station and covers an area of about 122,000 km2, approximately 17% of the total Yellow River basin (Figure 1). The SRYR, sometimes also termed ‘the cistern of the Yellower River’, yields 35% of the total annual discharge of the Yellow River (Si et al. 2019). This region is situated in the northeastern Tibetan Plateau, which has an average altitude of over 3,000 meters above sea level. The climate of the SRYR belongs to the typical Qinghai-Tibet Plateau climate system, which is characterized by distinct wet and dry seasons (Xu et al. 2018). Mainly influenced by the southwest monsoon from the Bay of Bengal, more than 70% of the annual precipitation across the SRYR occurs from June to September (Liang et al. 2010). Over the past several decades, the streamflow of the SRYR has exhibited a decreasing trend due to both climate change and human activities (Wang et al. 2018; Ji et al. 2021). Since the SRYR supplies water for millions of people, it is of great significance to attribute the contributions of climate change and human activities to its annual and intra-annual streamflow variation.
Figure 1
Geophysical features of the source region of the Yellow River (SRYR).
View largeDownload slide

Geophysical features of the source region of the Yellow River (SRYR).

Figure 1
Geophysical features of the source region of the Yellow River (SRYR).
View largeDownload slide

Geophysical features of the source region of the Yellow River (SRYR).

Close modal

Datasets

Listen

Daily observed streamflow data of 1956–2018 are from the hydrological station at Tangnaihai in the SRYR. Daily meteorological data of the 1956–2018, including precipitation (P), sunshine duration, and mean air temperature, are provided by the China Meteorological Data Sharing Service System (http://data.cma.cn/). The location of all the stations is shown in Figure 1.

Potential evaporation (Ep) data are derived by the Blaney-Criddle (B-C) equation (Blaney & Criddle 1950). The method involves only one variable (T) and has good applicability in arid and semiarid regions (Zhang et al. 2012):
(1)
where p is the percentage of monthly daytime hours in the total daytime hours, and k is the monthly consumptive use coefficient depending on vegetation types, locations, and the growing seasons. For the high-density natural vegetation in the growth season (May-October), k ranges from 0.5 to 1.2, and the default value is 0.85.

Mean areal precipitation and potential evaporation of the SRYR are calculated using the Thiessen polygon, based on the data from the seven meteorological stations (see Figure 1), which drives the monthly ABCD model and the Budyko-type equation.

Listen

Theoretical framework

Listen
In this study, a multi-scalar framework for attribution analysis was constructed by integrating hydrological modeling into the Budyko-based decomposition method. As shown in Figure 2, the multi-scalar framework for assessing the impacts of climate change and human activities on streamflow variation is developed with the following steps:
  • (1)

    Preparation of monthly meteorological and hydrological data. The monthly actual evaporation (E) is simulated by selected hydrological models (e.g, the ABCD hydrological model is adopted for a case study in the following sections). And then, the variation of water storage (ΔS) can be derived by a water balance equation (ΔS=P-E-Q), where P and Q are monthly precipitation and streamflow, respectively.

  • (2)

    Detection of mutation points of the analyzed streamflow time series. Mutation analysis tools, such as the Mann-Kendall (M-K) and sliding T-tests, are suggested to examine potential sudden variation features of streamflow observations. Thus, the whole study period is divided into two sub-periods (the prechange and postchange periods) before and after the most significant mutation point.

  • (3)

    Computation of evaporation ratio and dryness index for subsequent parameter estimation. The monthly data obtained in step (1) are aggregated to seasonal and annual time scales, and the corresponding evaporation ratio and dryness index, i.e., and for prechange period and and for the postchange period, are then calculated. Here, P1, E1, Ep1, ΔS1 and P2, E2, Ep2, ΔS2 represent precipitation, actual evaporation, potential evaporation and variation of water storage for the prechange and postchange periods, respectively.

  • (4)

    Determination of the parameters in the Budyko-type equation. Specifically, parameters of φ and ω are, respectively, calibrated by and for the prechange period; that is, Budyko curves with different parameters φ and ω are fitted and well established at annual, seasonal (dry season and wet season), and monthly time scales.

  • (5)

    Derivation of the evaporation ratio only impacted by climate change. For the postchange period, the evaporation ratio with only climate change impacts (⁠⁠) is calculated according to the calibrated parameters (φ and ω) in step (4) and the dryness index (⁠⁠) calculated in step (3).

  • (6)

    Calculation principle of Budyko-based decomposition method. Within a Budyko curve, climate change would lead to both horizontal and vertical moving of empirical points, that is, from to along the horizontal direction and from to along the vertical direction, whereas human impacts only cause a vertical alteration of empirical points from to ⁠.

  • (7)
    Attribution analysis through the decomposition scheme. The human-induced change of streamflow (ΔQh) is computed by Equation (2). Then, the climate-induced change of streamflow (ΔQc) is the remaining component calculated by subtracting the human-induced change from the total streamflow variation (ΔQ), as described in Equations (3) and (4).
    (2)
    (3)
    (4)
    where Q1 and Q2 are observed streamflow during the prechange and post change periods, respectively.

  • More theoretical details and the basis of the Budyko-based decomposition method (e.g., how to connect the moving of empirical points in Budyko curves to the quantification of climate and human impacts on streamflow variation) are explained in Section 3.4.

Figure 2
The multi-scalar assessment framework for attribution of streamflow variations to climate change and human activities.
View largeDownload slide

The multi-scalar assessment framework for attribution of streamflow variations to climate change and human activities.

Figure 2
The multi-scalar assessment framework for attribution of streamflow variations to climate change and human activities.
View largeDownload slide

The multi-scalar assessment framework for attribution of streamflow variations to climate change and human activities.

Close modal

Mutation analysis methods

Listen

The M-K analysis method (Mann 1945; Kendall 1990) has been widely used to test abrupt points of hydrological and meteorological elements due to its advantages of no interference from a few outliers and simple calculation. The sliding T-test method is a nonlinear statistical method that can directly judge the data change trend from the curve. Its calculation process is intuitive, and it is also widely used to test the abrupt change points of hydrological and meteorological time series (Du et al. 2019). To determine the mutation year accurately, the M-K method and the sliding T-test method are used together to identify the mutation points of the streamflow time series.

Monthly hydrological modeling

Listen

The ABCD model incorporating snowmelt module

Listen
The monthly ABCD model takes monthly precipitation and potential evaporation as input, captures the dynamics of soil moisture and groundwater, and produces monthly actual evaporation and streamflow as output (Thomas & Harold 1981). The model includes only four parameters (i.e., a, b, c, and d), whose physical meanings are illustrated in Figure 3. These four parameters, as well as the initial soil water storage (between 0 and b), are subject to calibration by a genetic algorithm (Goldberg 2010), and the initial groundwater storage was determined by the method proposed by Alley (1984).
Figure 3
The structure of ABCD model incorporating snow melting module.
View largeDownload slide

The structure of ABCD model incorporating snow melting module.

Figure 3
The structure of ABCD model incorporating snow melting module.
View largeDownload slide

The structure of ABCD model incorporating snow melting module.

Close modal
In addition to these processes considered in the ABCD model, this study embeds the snowmelt module proposed by Xu et al. (1996) into the traditional monthly ABCD model, which can describe snowfall, snow storage, and snowmelt by introducing additional two parameters, a1 and a2. The monthly ABCD model structure considering the snowmelt module is shown in Figure 3. Of these, precipitation can be expressed as the sum of rain and snow:
(5)
where Pt, rt, and SNt are the amount of precipitation, rainfall, and snowfall, in the t period, respectively.
The amount of snowmelt satisfies the snow balance equation for the current time period:
(6)
where Smt is the snow melt during the t period; SPt-1 is the snow storage at the beginning of the t period; SPt is the snow storage at the end of the t period.
The watershed water balance equation considering snowmelt can be expressed as:
(7)
where Et, Qt, and DSt are the actual evaporation, streamflow, and watershed water storage variables for the t period, respectively. Among them, rt and SNt can be expressed as a function of rainfall and temperature, while Smt can be expressed as a function of snow storage and temperature at the beginning of the period.
The performance of the monthly ABCD model is evaluated with the Nash-Sutcliffe efficiency (NSE) and water balance error (WBE) (Li et al. 2020).
(8)
(9)
where Qsim,i and Qobs,i are the streamflow simulation value and observation value at time i, respectively; represents the average value of observed streamflow; n is the length of streamflow time series.

Hydrologic simulation procedure

Listen
In this study, the monthly ABCD model is applied mainly for two purposes. First, to derive the monthly actual evaporation and water storage with two periods, which are further used to calculate the monthly dryness index and evaporation ratio. Second, to validate the proposed multi-scalar attribution analysis framework, the ABCD model is applied for quantifying the climate and human impacts on streamflow variation as a comparative experiment. To this end, based on the monthly ABCD model, a hydrologic simulation method for quantifying the impact of climate change and human activities on streamflow is constructed. And the calculation is as follows:
(10)
(11)
(12)
where ΔQ is the amount of actual total streamflow variation; Q1 and Q2 are observed streamflow during the prechange and postchange periods, respectively; ΔQc is the amount of streamflow variation caused by climate change; Qsim1 and Qsim2 are the simulated streamflowof the prechange and postchange periods; ΔQh is the amount of streamflow variation from human activities.

Decomposition based on extended Budyko-type equation

Listen

Extended Budyko-type equation

Listen
Chen et al. (2013) put forward the definition of effective precipitation, and then the applicability of the Budyko-type equation was extended from the annual time scale to the intra-annual time scales. The effective precipitation refers to precipitation that removes soil water storage and groundwater storage. The extended Budyko-type equation must consider the variation of water storage, but there is no measured value between the variation of water storage and actual evaporation. In this study, the monthly ABCD model was applied for simulating the monthly water storage and actual evaporation, which were aggregated on the seasonal and annual to construct the extended Budyko equation. Present study uses the extended Budyko equation to explore the impacts of the climate change and human activities on streamflow at different time scales. The formula of the extended Budyko equation is shown as following:
(13)
where P, E, and Ep represent the precipitation, actual evaporation, and potential evaporation, respectively; ΔS is the variation of water storage, including soil water and groundwater storage variation, and it may be positive or negative values; E/(P-S) stands for the seasonal evaporation rate; Ep/(P-S) stands for the seasonal dryness index; φ is the parameter representing the lower bound for the seasonal dryness index; ω is the parameter representing the effects of other factors such as vegetation, soil and topography on the partition of precipitation that determines the shape of the Budyko curve. The last two parameters (φ and ω) are obtained by least-squares method, referring to See et al. (2018).

Decomposition method

Listen

Wang & Hejazi (2011) put forward a decomposition method based on the Budyko equation. For a watershed without human impact, this method assumes that if the dryness index (⁠⁠) moves to a drier or wetter condition due to climate change, the evaporation ratio (⁠⁠) would also move to a new position but still follow the original Budyko curve as the prechange period. The assumption is considered reasonable because the shape of Budyko-type curve is largely dependent on the physical properties of the watershed, such as topography, soil properties, and vegetation characteristics (Yang et al. 2007; Wang & Wu 2013).

The schematic of the decomposition method is shown in Figure 4. Under the joint impact of climatic and anthropic factors, point A (⁠⁠, ⁠) in the prechange period will move to point B (⁠⁠, ⁠) in the postchange period. Under the impact of only climate change, point A will move to point C (⁠⁠, ⁠) along the Budyko curve. The climatic conditions of point B and C are consistent. Climate change will cause both horizontal and vertical changes, whereas human activities can only cause changes in the vertical direction. The contributions of climate change and human activities to streamflow variation could be calculated based on the deviations.
Figure 4
Schematic of Budyko-type decomposition method to quantify the climate and human impacts on streamflow variation at different time scales.
View largeDownload slide

Schematic of Budyko-type decomposition method to quantify the climate and human impacts on streamflow variation at different time scales.

Figure 4
Schematic of Budyko-type decomposition method to quantify the climate and human impacts on streamflow variation at different time scales.
View largeDownload slide

Schematic of Budyko-type decomposition method to quantify the climate and human impacts on streamflow variation at different time scales.

Close modal

Seasonality identification

Listen
In this study, wet and dry seasons are determined using a multi-year average monthly dryness index (⁠⁠). The formula to calculate is as follows:
(14)
where and are averaged monthly potential evaporation and precipitation, respectively; is averaged monthly water storage, which is calculated with the equation of monthly water balance (⁠⁠). If the is less than 1, the month is labeled as wet; otherwise, the month belongs to the dry season. According to this method, the wet season is from July to November, and the rest of the months are the dry season.
Listen

Mutation analysis and changes of hydro-meteorological elements

Listen
As shown in Figure 5. Two mutation points (1991 and 1993) were found using the Mann-Kendall method. Combining the sliding T-test method, the significant mutation point of annual streamflow was determined as 1991. The mutation point determined in this study is consistent with the year suggested by Yuan et al. (2015) for the same study region. Then the whole study period was separated into two periods: the prechange period (1956–1991) and the postchange period (1992–2018). The prechange period is considered closer to the natural conditions, while the postchange period is considered a period more impacted by human activities.
Figure 5
The identification of mutation point using (a) M-K and (b) Sliding T-tests for long-term annual streamflow series of 1956–2018.
View largeDownload slide

The identification of mutation point using (a) M-K and (b) Sliding T-tests for long-term annual streamflow series of 1956–2018.

Figure 5
The identification of mutation point using (a) M-K and (b) Sliding T-tests for long-term annual streamflow series of 1956–2018.
View largeDownload slide

The identification of mutation point using (a) M-K and (b) Sliding T-tests for long-term annual streamflow series of 1956–2018.

Close modal

Precipitation, potential evaporation, and streamflow are the key input variables of Budyko's water-energy balance equation. It is significant to explore their characteristics on a time scale for quantitatively revealing the impacts of climate change and human activities on streamflow. This study quantitatively evaluates their relative changes for the prechange and postchange periods using the statistical analysis method, which are shown in Table 1. It is found that the variation of precipitation, potential evaporation, and streamflow differs markedly among different time scales. It should be noted that the variation of streamflow mainly occurred in the wet season (greatest in September, followed by October and August). Additionally, the relatively low increase of precipitation, but the high increase of evaporation and variation of water storage jointly led to the drastic decrease of streamflow in September.

Table 1

Quantitative analysis of the impacts of meteorological and hydrological elements during the different periods (unit: mm)

 
 

Note: P1, Ep1 and Q1 represent the average precipitation, potential evaporation and streamflow in the prechange period, respectively. P2, Ep2 and Q2 represent the average precipitation, potential evaporation and streamflow in the postchange period. ΔP, ΔEp and ΔQ represent their respective change value.

View Large

Performance of streamflow simulation using monthly ABCD model

Listen

According to the analysis results in Section 4.1, the whole study period was separated into two periods: the prechange period (1956–1991) and the postchange period (1992–2018). Then both the prechange and postchange periods are further divided into a calibration period and a validation period. That is, from 1956–1981 (calibration) and 1982–1991 (validation) for the prechange period, and from 1992–2010 (calibration) and 2011–2018 (validation) for the postchange period. The optimized values of model parameters are summarized in Table 2.

Table 2

Optimized parameters of the ABCD model

Periodsab (mm)cd (t−1)
Prechange 0.98 547.87 0.34 0.48 
Postchange 0.99 628.19 0.35 0.61 
Periodsab (mm)cd (t−1)
Prechange 0.98 547.87 0.34 0.48 
Postchange 0.99 628.19 0.35 0.61 
View Large

Figure 6 and Table 3 represent the hydrological simulation process and model performance in the calibration and validation periods of the ABCD model, respectively. The performance of the model is generally satisfactory, especially in the non-flood period. As shown in Table 3, the NSE and WBE range from 0.79 to 0.86 and 0.41 to 1.55, respectively, indicating a fairly satisfactory fitting effect between the simulated and observed values. Therefore, the ABCD model established in this study performs well in the SRYR and lays a good foundation for the Budyko-based decomposition method, as well as for quantifying the relative impacts of climate change and human activities on streamflow variation.
Table 3

Performance of the ABCD model for monthly streamflow simulation

PeriodsNSEWBE (%)NSEWBE (%)
Prechange Calibration (1956–1981) Validation (1982–1991) 
0.84 0.52 0.82 1.34 
Postchange Calibration (1992–2010) Validation (2011–2018) 
0.79 1.55 0.86 0.41 
PeriodsNSEWBE (%)NSEWBE (%)
Prechange Calibration (1956–1981) Validation (1982–1991) 
0.84 0.52 0.82 1.34 
Postchange Calibration (1992–2010) Validation (2011–2018) 
0.79 1.55 0.86 0.41 
View Large
Figure 6
Comparison of observed and simulated streamflow with ABCD model for (a) prechange period, and (b) postchange period.
View largeDownload slide

Comparison of observed and simulated streamflow with ABCD model for (a) prechange period, and (b) postchange period.

Figure 6
Comparison of observed and simulated streamflow with ABCD model for (a) prechange period, and (b) postchange period.
View largeDownload slide

Comparison of observed and simulated streamflow with ABCD model for (a) prechange period, and (b) postchange period.

Close modal

Estimation of parameters upon the Budyko curve

Listen

In order to apply the decomposition method to quantify the impacts of climate change and human activities on streamflow variation, Budyko curves for the prechange period, need to be fitted, i.e. parameters φ and ω in formula (13) need to be determined. In this study, the evaporation ratio and dryness index for the prechange period (1956–1991) are used to estimate the parameters φ and ω at different time scales through the least square method.

Table 4 shows the results of parameters φ and ω of the Budyko formula at different time scales for the prechange period. As can be seen, parameters φ and ω have both similarities and dissimilarities at different time scales. The similarity is reflected in the small difference in the parameter values if the dryness of the basins is close to each other, while the dissimilarity is reflected in the significant difference in the parameter values if the dryness of the basins is different. To be specific, the parameter φ is larger in the dry season while smaller in the wet season. At the same time, the parameter ω is larger in the wet season while smaller in the dry season. Accordingly, it can be concluded that with the increase of dryness degree, parameter φ gradually increases, while parameter ω gradually decreases. This finding is in good agreement with that indicated by Bosshard et al. (2013).

Table 4

Estimated parameters φ and ω of Budyko formula at different time scales for the prechange period

 
 
View Large

The fitting results of the Budyko curve for the prechange period at different time scales are shown in Figure 7. It can be seen that most of the solid blue dots are distributed on the upper side of the fitted curves, indicating a decreasing impact of human activities impact on streamflow at all the time scales for the postchange period (1992–2018). Furthermore, the furher the vertical scatter plot is from the fitted curve, the more significant impacts of human activities are indicated. It can be seen from the fitted curve of different time scales that human activities have less influence on annual streamflow variation. At the seasonal scale, human activities have less impact on streamflow during the dry season, but more impact during the wet season. At the monthly scale, the impact of human activities on streamflow is greater from July to September and smaller from December to February.
Figure 7
Fitting results of (a) annual, (b-c) seasonal, and (d-o) monthly Budyko curves.
View largeDownload slide

Fitting results of (a) annual, (b-c) seasonal, and (d-o) monthly Budyko curves.

Figure 7
Fitting results of (a) annual, (b-c) seasonal, and (d-o) monthly Budyko curves.
View largeDownload slide

Fitting results of (a) annual, (b-c) seasonal, and (d-o) monthly Budyko curves.

Close modal

Quantification of climate and human impacts on streamflow

Listen
Figure 8 illustrates the contributions of climate change and human activities to the variations of annual, seasonal, and monthly streamflow during the postchange period estimated by the multi-scalar framework for attribution analysis. As can be seen, the annual streamflow variation caused by climate change and human activities are −12.60 mm and −6.50 mm, respectively. It can be seen that climate change in the SRYR has a dominant influence on annual streamflow. For intra-annual time scales, climate change and human activities caused streamflow variation of −1.25 mm and −2.09 mm in the dry season and −11.88 mm and −4.06 mm in the wet season, respectively. Notably, for the annual scale and the wet season, climate change has more significant impact on streamflow variation, while in the dry season, streamflow variation is mainly affected by human activities. In addition, the decrease of streamflow in the SRYR mainly occurred in the wet season, especially in September, when the impact of climate change is dominant. In short, climate change could be the main reason for the decrease in annual streamflow in the SRYR.
Figure 8
Contributions of climate change and human activities to the variation of annual, seasonal, and monthly streamflow estimated by Budyko-based decomposition method.
View largeDownload slide

Contributions of climate change and human activities to the variation of annual, seasonal, and monthly streamflow estimated by Budyko-based decomposition method.

Figure 8
Contributions of climate change and human activities to the variation of annual, seasonal, and monthly streamflow estimated by Budyko-based decomposition method.
View largeDownload slide

Contributions of climate change and human activities to the variation of annual, seasonal, and monthly streamflow estimated by Budyko-based decomposition method.

Close modal
With the hydrological simulation method, the attribution consistently results with the multi-scalar framework for attribution analysis (Figure 9). The most important thing is that both climate change and human activities contributed to the decrease of streamflow at different time scales.
Figure 9
Contributions of climate change and human activities to the variation of annual, seasonal, and monthly streamflow estimated by hydrologic simulation.
View largeDownload slide

Contributions of climate change and human activities to the variation of annual, seasonal, and monthly streamflow estimated by hydrologic simulation.

Figure 9
Contributions of climate change and human activities to the variation of annual, seasonal, and monthly streamflow estimated by hydrologic simulation.
View largeDownload slide

Contributions of climate change and human activities to the variation of annual, seasonal, and monthly streamflow estimated by hydrologic simulation.

Close modal
Listen

Correlating streamflow variations to climate factors

Listen
Precipitation and potential evaporation are the main climatic factors determining streamflow. Pearson correlation analysis was applied to further analyze the causes of streamflow variation. As shown in Figure 10, the highest correlation is found between wet season streamflow and wet season precipitation, with a correlation coefficient of 0.535 that passes the test at the significance level of 0.05. In general, streamflow was positively correlated with precipitation and negatively correlated with evaporation, with precipitation playing a major role in both dry and wet seasons, and potential evaporation was more strongly correlated with streamflow in the dry season than in the wet season.
Figure 10
(a) Correlation coefficient matrix and (b) Correlation coefficient test matrix between streamflow, precipitation and potential evaporation in dry and wet seasons.
View largeDownload slide

(a) Correlation coefficient matrix and (b) Correlation coefficient test matrix between streamflow, precipitation and potential evaporation in dry and wet seasons.

Figure 10
(a) Correlation coefficient matrix and (b) Correlation coefficient test matrix between streamflow, precipitation and potential evaporation in dry and wet seasons.
View largeDownload slide

(a) Correlation coefficient matrix and (b) Correlation coefficient test matrix between streamflow, precipitation and potential evaporation in dry and wet seasons.

Close modal

Temperature is another important climatic factor affecting SRYR streamflow variability. The increase in temperature leads to an increase in evaporation, which causes a decrease in streamflow. It also promotes the melting of permafrost and snow and then replenishes the streamflow. The melting of permafrost makes the soil infiltration capacity increase and leads to a decrease in streamflow. It is one of the reasons why the streamflow reduces slightly in winter but greatly in summer and autumn in the SRYR.

Effectiveness and advantages of the multi-scalar framework

Listen

In this study, a multi-scalar framework for attribution analysis was proposed based on the Budyko-based decomposition method and monthly ABCD model. Thus, it has the advantage of the Budyko-based decomposition method that simplifies the model structure and requires only hydro-meteorological elements data without terrain soil, land use, and vegetation data. Furthermore, the proposed multi-scalar framework has the advantage of the hydrologic simulation method of high analytical accuracy, making it possible to analyze the process of intra-annual streamflow at a monthly scale. In theory, the multi-scalar framework for attribution analysis proposed in this study is more comprehensive than conventional the Budyko-based decomposition that can only be applied to multi-annual time scales. This is because the design of traditional Budyko-based decomposition is based on the assumption that the variation of water storage can be neglected at a long-term scale (e.g., far longer than one year). In comparison, the proposed multi-scalar framework for attribution analysis considered the effect of watershed water storage on streamflow variation through introducing monthly dryness index and evaporation ratio. More importantly, the proposed multi-scalar framework for attribution analysis can be readily applied to other basins and regions as well.

Figure 11 also presented the comparisons between the multi-scalar framework and the hydrologic simulation method for streamflow variation due to climate change ΔQc, and due to human activities ΔQh. It is shown that the estimated ΔQc and ΔQh with these two methods are similar. Thus, the cross-corroboration of the attribution analysis between the multi-scalar framework and the hydrologic simulation method indicates that both methods are effective.
Figure 11
Comparison of (a) ΔQc and (b) ΔQh estimated by the Budyko-based multi-scalar framework and the improved ABCD hydrologic model.
View largeDownload slide

Comparison of (a) ΔQc and (b) ΔQh estimated by the Budyko-based multi-scalar framework and the improved ABCD hydrologic model.

Figure 11
Comparison of (a) ΔQc and (b) ΔQh estimated by the Budyko-based multi-scalar framework and the improved ABCD hydrologic model.
View largeDownload slide

Comparison of (a) ΔQc and (b) ΔQh estimated by the Budyko-based multi-scalar framework and the improved ABCD hydrologic model.

Close modal

Uncertainties and limitations

Listen

One issue should be noted that climate change mentioned in the study only refers to the changes in effective precipitation (P-ΔS) and potential evaporation. The multi-scalar framework for attribution analysis cannot separate the impacts of other climate factors. This is because both the direct impact and the indirect impact of climate change on streamflow variation are estimated based on effective precipitation and potential evaporation in this study. However, other climatic characteristics such as seasonality of precipitation, intensity of precipitation, temperature, wind speed, etc. can also influence the water flow production process in the basin.

This study finds a non-negligible role of human activities on the alteration of watershed streamflow at the monthly scale. The impacts of human activities on hydrological processes are in two main ways. One is changing the characteristics of the basin substratum and thus the rainfall-runoff relationship. It has been shown that the current warming and humidifying climatic conditions are favorable to the growth of vegetation in the SRYR (Yan et al. 2020), and the normalized difference vegetation index (NDVI) of SRYR has shown a significant upward trend since 1999. The substantial increase in vegetation cover has also led to a decrease in watershed streamflow. In addition, the construction of large hydraulic projects has altered the natural distribution characteristics of river streamflow on short time scales and increased the area of free water bodies in the watershed, thereby increasing evaporation. The effects of changes in vegetation and reservoir operations on streamflow are not discussed in this study, which would need to be analyzed with a more sophisticated attribution analysis in the following research.

Listen

Quantifying the impacts of climate change and human activities on streamflow changes at multiply intra-annual time scales is essential. In this study, the monthly ABCD model is used to simulate the monthly water storage and actual evaporation in the SRYR, and further to drive the Budyko-based decomposition method. Then, the Budyko-based decomposition method is extended to the monthly scale by introducing the monthly dryness index and evaporation ratio. Thus, a new multi-scalar framework is constructed that can quantify the contribution of climate change and human activities on streamflow. The main findings and conclusions are summarized as follows.

  • (1)

    According to the mutation analysis, the whole study period is separated into the prechange period (1956–1991) and the postchange period (1992–2018). Meanwhile, the monthly ABCD hydrological model has good performance and stability for simulating observed hydrological processes in the two periods, during which the NSE remains above 0.79. More importantly, it performs well in obtaining actual evaporation and water storage data for the SRYR.

  • (2)

    The relative contribution of climate change to streamflow variations is 65.99%, 37.50%, and 70.86% in annual, dry season, and wet season, respectively, indicating that climate change is the dominant factor controlling streamflow variation for the annual and wet season, while human activities play a major role in streamflow variation for dry season. The contributions of climate change and human activities show obvious dynamic and seasonal characteristics.

  • (3)

    This study analyzes the impacts of climate change and human activities on streamflow on seasonal and monthly scales, which can be used to guide the allocation and regulation of water resources in each season (especially in the wet season). The results could provide a reference for theoretically analyzing the driving mechanism of water cycle processes in the changing environment, and are of great significance for rational development, optimal allocation, and sustainable utilization of local water resources in the basin.

Listen

This study is funded by the National Natural Science Foundation of China (Grant Nos. 41730645, 51779070 and 41701022).

Listen

Mingwei Ma: Conceptualization, Data curation, Framework design, Writing – review and editing.

Zhaohang Wang: Methodology, Visualization, Formal analysis, Writing – original draft.

Huijuan Cui: Supervision, Funding acquisition, Writing-review & editing.

Wenchuan Wang: Supervision, Investigation.

Liuyuwei Jiang: Data curation, Format check.

Listen

Data cannot be made publicly available; readers should contact the corresponding author for details.

Listen

The authors declare there is no conflict.

Alley
W. M.
1984
On the treatment of evapotranspiration, soil moisture accounting, and aquifer recharge in monthly water balance models
.
Water Resour. Res.
20
(
8
),
1137
–
1149
.
https://doi.org/10.1029/WR020i008p01137
.
Google Scholar
Crossref
Search ADS
 
Blaney
H. F.
 &
Criddle
W. D.
1950
Determining Water Requirements in Irrigated Areas From Climatological and Irrigation Data
.
US Department of Agriculture, Soil Conservation Service, Teach
, Vol.
96
, p.
48
.
Bosshard
T.
,
Carambia
M.
,
Goergen
K.
,
Kotlarski
S.
,
Krahe
P.
,
Zappa
M.
 &
Schar
C.
2013
Quantifying uncertainty sources in an ensemble of hydrological climate-impact projections
.
Water Resour. Res.
49
(
3
),
1523
–
1536
.
Google Scholar
Crossref
Search ADS
 
Chang
J.
,
Zhang
H.
,
Wang
Y.
 &
Zhu
Y.
2016
Assessing the impact of climate variability and human activities on streamflow variation
.
Hydrol. Earth Syst. Sci.
20
(
4
),
1547
–
1560
.
https://doi.org/10.5194/hess-20-1547-2016
.
Google Scholar
Crossref
Search ADS
 
Charles
J. V.
,
Pamela
G.
,
Joseph
S.
 &
Richard
B. L.
2000
Global water resources: vulnerability from climate change and population growth
.
Science
289
(
5477
),
284
–
288
.
https://doi.org/10.1126/science.289.5477.284
.
Google Scholar
PubMed
 
Chen
X.
,
Alimohammadi
N.
 &
Wang
D.
2013
Modeling interannual variability of seasonal evaporation and storage change based on the extended Budyko framework
.
Water Resour. Res.
49
(
9
),
6067
–
6078
.
https://doi.org/10.1002/wrcr.20493
.
Google Scholar
Crossref
Search ADS
 
Du
C.
,
Sun
F.
,
Yu
J.
,
Liu
X.
 &
Chen
Y.
2016
New interpretation of the role of water balance in an extended Budyko hypothesis in arid regions
.
Hydrol. Earth Syst. Sci.
20
(
1
),
393
–
409
.
https://doi.org/10.5194/hess-20-393-2016
.
Google Scholar
Crossref
Search ADS
 
Du
R.
,
Shang
F.
 &
Ma
N.
2019
Automatic mutation feature identification from well logging curves based on sliding t test algorithm
.
Cluster. Comput.
22
(
6
),
14193
–
14200
.
https://doi.org/10.1007/s10586-018-2267-z
.
Google Scholar
 
Goldberg
D. E.
2010
Genetic Algorithms in Search, Optimization, and Machine Learning
.
Queen's University, Belfast, UK
.
https://doi.org/10.1111/j.1365-2486.2009.02080.x
.
Google Scholar
 
Guo
Y.
,
Huang
S.
,
Huang
Q.
,
Leng
G.
,
Fang
W.
,
Wang
L.
 &
Wang
H.
2020
Propagation thresholds of meteorological drought for triggering hydrological drought at various levels
.
Sci. Total Environ.
712
,
136502
.
https://doi.org/10.1016/j.scitotenv.2020.136502
.
Google Scholar
Crossref
Search ADS
PubMed
 
Ji
G.
,
Wu
L.
,
Wang
L.
,
Yan
D.
 &
Lai
Z.
2021
Attribution analysis of seasonal runoff in the source region of the Yellow River using seasonal budyko hypothesis
.
Land.
10
(
5
),
542
–
542
.
https://doi.org/10.3390/land10050542
.
Google Scholar
Crossref
Search ADS
 
Jiang
C.
,
Xiong
L.
,
Wang
D.
,
Liu
P.
,
Guo
S.
 &
Xu
C. Y.
2015
Separating the impacts of climate change and human activities on runoff using the Budyko-type equations with time-varying parameters
.
J. Hydrol.
522
,
326
–
338
.
https://doi.org/10.1016/j.jhydrol.2014.12.060
.
Google Scholar
Crossref
Search ADS
 
Kendall
M. G.
1990
Rank correlation methods
.
Br. J. Psychol.
25
(
1
),
86
–
91
.
https://doi.org/10.1111/j.2044-8295.1934.tb00727.x
.
Google Scholar
 
Li
Z.
,
Huang
S.
,
Liu
D.
,
Leng
G.
,
Zhou
S.
 &
Huang
Q.
2020
Assessing the effects of climate change and human activities on runoff variations from a seasonal perspective
.
Stoch. Env. Res. Risk A.
34
,
1
–
18
.
https://doi.org/10.1007/s00477-020-01785-1
.
Google Scholar
 
Liang
S.
,
Ge
S.
,
Wan
L.
 &
Zhang
J.
2010
Can climate change cause the Yellow River to dry up?
Water Resour. Res.
46
(
2
),
W02505.1
.
https://doi.org/10.1029/2009WR007971
.
Google Scholar
Crossref
Search ADS
 
Liu
J.
,
Zhang
Q.
,
Singh
V. P.
 &
Shi
P.
2017
Contribution of multiple climatic variables and human activities to streamflow changes across China
.
J. Hydrol.
545
,
145
–
162
.
https://doi.org/10.1016/j.jhydrol.2016.12.016.c
.
Google Scholar
Crossref
Search ADS
 
Mann
H. B.
1945
Nonparametric test against trend
.
Econometrica.
13
(
3
),
245
–
259
.
https://doi.org/10.2307/1907187
.
Google Scholar
Crossref
Search ADS
 
Marisol
D. T.
 &
Hugo A
G. J.
2021
Global analysis of the hydrologic sensitivity to climate variability
.
J. Hydrol.
603
,
126720
.
https://doi.org/10.1016/j.jhydrol.2021.126720
.
Google Scholar
 
Narsimlu
B.
,
Gosain
A. K.
 &
Chahar
B. R.
2013
Assessment of future climate change impacts on water resources of upper sind river basin, India using SWAT model
.
Water Resour. Manag.
27
(
10
),
3647
–
3662
.
https://doi.org/10.1007/s11269-013-0371-7
.
Google Scholar
Crossref
Search ADS
 
Onstad
C. A.
 &
Jamieson
D. G.
1970
Modeling the effect of land use modifications on runoff
.
Water Resour. Res.
6
(
5
),
1287
–
1295
.
https://doi.org/10.1029/WR006i005p01287
.
Google Scholar
Crossref
Search ADS
 
Piao
S.
,
Ciais
P.
,
Huang
Y.
,
Shen
Z.
,
Peng
S.
,
Li
J.
,
Zhou
L.
,
Liu
H.
,
Ma
Y.
 &
Ding
Y.
2010
The impacts of climate change on water resources and agriculture in China
.
Nature
467
(
7311
),
43
–
51
.
https://doi.org/10.1038/nature09364
.
Google Scholar
Crossref
Search ADS
PubMed
 
See
J. J.
,
Jamaian
S. S.
,
Salleh
R. M.
,
Nor
M. E.
 &
Aman
F.
2018
Parameter estimation of Monod model by the Least-Squares method for microalgae Botryococcus Braunii sp
.
J. Phys. Conf. Ser.
995
(
1
),
012026
–
012026
.
https://doi.org/10.1088/1742-6596/995/1/012026
.
Google Scholar
 
Si
Y.
,
Li
X.
,
Yin
D.
,
Li
T.
,
Cai
X.
,
Wei
J.
 &
Wang
G.
2019
Revealing the water-energy-food nexus in the Upper Yellow River Basin through multi-objective optimization for reservoir system
.
Sci. Total Environ.
682
,
1
–
18
.
https://doi.org/10.1016/j.scitotenv.2019.04.427
.
Google Scholar
Crossref
Search ADS
PubMed
 
Sivapalan
M.
2003
Prediction in ungauged basins: a grand challenge for theoretical hydrology
.
Hydrol. Process.
17
(
15
),
3163
–
3170
.
https://doi.org/10.1002/hyp.5155
.
Google Scholar
Crossref
Search ADS
 
Thomas
J.
 &
Harold
A.
1981
Improved Methods for National Water Assessment
.
Water Resour. Concil
,
Washington, D.C
, pp.
1
–
59
.
Google Scholar
 
Wang
X.
2014
Advances in separating effects of climate variability and human activity on stream discharge: an overview
.
Adv. Water Resour.
71
,
209
–
218
.
https://doi.org/10.1016/j.advwatres.2014.06.007
.
Google Scholar
Crossref
Search ADS
 
Wang
D.
 &
Hejazi
M.
2011
Quantifying the relative contribution of the climate and direct human impacts on mean annual streamflow in the contiguous United States
.
Water Resour. Res.
47
(
10
),
W00J12.1
–
W00J12.16
.
https://doi.org/10.1029/2010WR010283
.
Google Scholar
Crossref
Search ADS
 
Wang
X. S.
 &
Zhou
Y.
2016
Shift of annual water balance in the Budyko space for catchments withgroundwater-dependent evapotranspiration
.
Hydrol. Earth Syst. Sci.
20
(
9
),
3673
–
3690
.
https://doi.org/10.5194/hess-20-3673-2016
.
Google Scholar
Crossref
Search ADS
 
Wang
T.
,
Yang
H.
,
Yang
D.
,
Qin
Y.
 &
Wang
Y.
2018
Quantifying the streamflow response to frozen ground degradation in the source region of the Yellow River within the Budyko framework
.
J. Hydrol.
558
,
301
–
313
.
https://doi.org/10.1016/j.jhydrol.2018.01.050
.
Google Scholar
Crossref
Search ADS
 
Xin
Z.
,
Li
Y.
,
Zhang
L.
,
Ding
W.
,
Ye
L.
,
Wu
J.
 &
Zhang
C.
2019
Quantifying the relative contribution of climate and human impacts on seasonal streamflow
.
J. Hydrol.
574
,
936
–
945
.
https://doi.org/10.1016/j.jhydrol.2019.04.095
.
Google Scholar
Crossref
Search ADS
 
Xu
C. Y.
,
Seibert
J.
 &
Halldin
S.
1996
Regional water balance modelling in the NOPEX area: development and application of monthly water balance models
.
J. Hydrol.
180
(
1
),
211
–
236
.
https://doi.org/10.1016/0022-1694(95)02888-9
.
Google Scholar
 
Xu
S.
,
Yu
Z.
,
Yang
C.
,
Ji
X.
 &
Zhang
K.
2018
Trends in evapotranspiration and their responses to climate change and vegetation greening over the upper reaches of the Yellow River Basin
.
Agric. For Meteorol.
263
,
118
–
129
.
https://doi.org/10.1016/j.agrformet.2018.08.010
.
Google Scholar
Crossref
Search ADS
 
Yang
D.
,
Sun
F.
,
Liu
Z.
,
Cong
Z.
,
Ni
G.
 &
Lei
Z.
2007
Analyzing spatial and temporal variability of annual water-energy balance in nonhumid regions of China using the Budyko hypothesis
.
Water Resour. Res.
43
,
W04426
.
Google Scholar
 
Yuan
F.
,
Berndtsson
R.
,
Zhang
L.
 &
Uvo
C. B.
2015
Hydro climatic trend and periodicity for the source region of the Yellow River
.
J. Hydrol. Eng.
20
(
10
),
05015003
–
05015003
.
https://doi.org/10.1061/(ASCE)HE.1943-5584.0001182
.
Google Scholar
Crossref
Search ADS
 
Zhang
X. L.
,
Xiong
L. H.
,
Lin
L.
 &
Long
H. F.
2012
Application of five potential evapotranspiration equations in Hanjiang Basin
.
Arid Land Geogr.
35
(
2
),
229
–
237
.
(In Chinese). https://doi.org/10.1007/s11783-011-0280-z
.
Google Scholar
 
© 2023 The Authors
This is an Open Access article distributed under the terms of the Creative Commons Attribution Licence (CC BY 4.0), which permits copying, adaptation and redistribution, provided the original work is properly cited (http://creativecommons.org/licenses/by/4.0/).