Attribution of runoff variation to climate and human-driven changes in the transition zone between the Qinling Mountains and the Loess Plateau under vegetation greening

The Fifth Assessment Report (AR5) of the Intergovernmental Panel on Climate Change (IPCC) noted that the global average land temperature and ocean surface temperature increased by 0.85 °C from 1885 to 2012 (Field & Barros 2014; IPCC 2019), implying that global warming and climate change were becoming indisputable realities. Climate change affects patterns of water vapor transmission (Papalexiou & Montanari 2019) and has accelerated runoff variation in various areas worldwide (Milly et al. 2005; Kay 2021; Lian et al. 2021). The rapid development of human society and increase in human activities (such as continuous improvements in water and soil conservation programs, excessive land exploitation, increasing water withdrawal, and large-scale construction of water control projects) have had major impacts on the hydrological cycle (Latrubesse et al. 2017; Yang et al. 2020; Zheng et al. 2021), and climate change has exacerbated these effects and led to a significant decline in streamflow (Jiang et al. 2015; Khazaei et al. 2019). Vegetation coverage has also changed significantly in agricultural areas in Asia in recent years (Fu 2003; Tuo et al. 2018), which affects the regional energy balance and global material flows. Higher vegetation can also alter regional hydrological processes. Human intervention is an important contributor to global vegetation coverage and is estimated to have driven more than 60% of observed changes in global vegetation coverage (Song et al. 2018). All of the above changes have increased the need for water resource adaptive management (Dey & Mishra 2017; Rogger et al. 2017; Li et al. 2021). Attribution of runoff variation to climate and human-driven changes is critically important for the development of adaptation strategies and policies for the planning and management of water resources.

Several studies have focused on the attribution of runoff variation, and this has motivated interest in characterizing the relative importance of the effects of climate change and human activities on runoff variation. A variety of approaches have been used to conduct this research, such as hydrological simulations based on physical mechanisms, coupled water–energy balance theory based on the Budyko hypothesis, and double-mass curves. The results of this research have shown that the individual impacts of climate change and human activities vary among regions. Tan et al. (2020) investigated the relationships of baseflow with precipitation, temperature, and four other indicators at 2,374 streamflow gauges worldwide from 1970 to 2016 and found that precipitation and terrestrial water storages played a dominant role in affecting baseflow changes globally, which made contributions of 64.8 and 20.2%, respectively. Luo et al. (2020) explored the effects of vegetation and climate change in global arid, semi-arid, and sub-humid regions based on the Budyko framework. They found that the increase in vegetation coverage represented by the satellite-derived leaf area index (L) was responsible for 48.0, 33.2, and 20.2% of the reduction in runoff in three typologies, which had an effect greater than that of climate change. Ban et al. (2020) built four land surface (hydrological) models to explore the asymmetrical responses of streamflow changes in the warm season and cool season in the Western United States. They found that increasing temperature would lead to a runoff reduction from 2.0 to 7.5% under 3 °C warm season warming and from 2.2 to 4.7% under 3 °C cool season warming. Hosseini-Moghari et al. (2020) employed the WaterGAP Global Hydrology Model (WGHM) to evaluate the relative contribution of human activities and climate change to lake water loss and found that 39–43% of the loss was driven by human activities and 57–61% of lake water loss was driven by climate change. Liu et al. (2019a) estimated the relative contributions of precipitation, potential evapotranspiration, and vegetation change to runoff variation in southwestern Australia using the theoretical framework of the Budyko curve. They found that the effects of climate and vegetation were equal (contributions of 45.4 and 47.4%, respectively). Champagne et al. (2020) simulated the future streamflow in four watersheds in southern Ontario using a precipitation runoff modeling system (PRMS) hydrological model under the RCP8.5 scenario. They predicted that a small rise in streamflow during January–February would happen in Big Creek (18 ± 8.7%), the Grand River (30.5 ± 10.8%), the Thames River (29.8 ± 10.4%), and the Credit River (31.2 ± 13.3%). In addition, Wang et al. (2013) used a simple ecohydrological approach, an elasticity differential analysis, and the MIKE-SHE model to estimate the impacts of land use and climate change on runoff reduction in northern China. The results indicated that the contributions of climate and land-use changes were 44 and 56% during 1980–1989 and 51 and 49% during 2000–2008, respectively. Liang et al. (2015) analyzed the effect of ecological restoration on the hydrological process in China's Loess Plateau using the Budyko framework. They found that the dominant contributor to runoff decline was ecological restoration (68%), which was greater than the impact of climate change (32%). Ge et al. (2020) evaluated the influence of revegetation on the hydrological cycle in the Loess Plateau and found that increases in evapotranspiration have caused a decline in streamflow and soil moisture since the launch of the Grain for Green Program (GGP).

Although many studies have evaluated the contributions of climate change and human activities to streamflow change in arid and semi-arid areas (Cui & Graf 2009; Zhai & Tao 2017; Chen et al. 2020; Samimi et al. 2020; Wu et al. 2020a), few studies have proposed a framework for attribution analysis of the variation in the boundary belt of various geographical units, such as the transition zone between the Qinling Mountains and the Loess Plateau in China. Additionally, most previous studies have considered the impacts of climate change and human activities overall; by contrast, few studies have examined the effects of specific factors or actions falling under the broader scope of climate change and human activities. Here, a novel framework to quantitatively estimate the response of streamflow to changes in meteorological factors and various human interventions under vegetation greening in the transition zone between the Qinling Mountains and the Loess Plateau was developed. Segment identification was implemented across the entire study period, and comparisons between or among multiple sub-periods permitted the quantitative estimation of the relative contribution of various factors to driving runoff variation. The contributions of runoff variation were calculated under changing climate indices (e.g., precipitation and temperature) and human interference indices (e.g., vegetation greening, direct water withdrawal, rainwater cellars, and reservoir impoundment).

This paper was organized as follows. In section 2, the overview of the study area, data resources, and changes in the recorded data series were introduced. In section 3, the Hydrological Model of L′École de Technologie Supérieure (HMETS)-based framework for quantifying the attribution of runoff variation, including vegetation greening, was presented. In section 4, the contributions of all factors contributing to runoff variation in the different sub-periods were stated. The effects of increased vegetation greening and temperature in the transition zone were discussed. Finally, the main conclusions were presented in section 5.

The study area is located in the ecotone of the Loess Plateau and the Qinling Mountains (34 °15′–35 °22′ N, 104 °2′–104 °57′ E), high in the west and low in the east, and spans the 7,210.09 km² catchment controlled by the Wushan gauge of the main Wei River (Figure 1). It has a typical semi-humid continental monsoon climate with an annual average precipitation of 518 mm (Xue et al. 2017; Jin et al. 2020; Yu et al. 2020). The distribution of precipitation throughout the year is uneven, with over 70% of precipitation falling from July to October. This generates approximately 4.8 × 10⁸ m³ of runoff annually. The regional average temperature is 7.8–13.5 °C; summers are hot, and winters are cold and dry. The study region occurs in the source area of the Wei River Basin, including three types of landforms, i.e., the Loess Plateau hilly region, the west Qinling Mountains, and the valley regions. This shows the uniqueness of the region and implies the complexity of watershed hydrological processes. Due to its location at the source of the Wei River, human activities were not very strong in the study area, though these activities have been increasing in the last few years (Lu et al. 2019; Wu et al. 2020b; Song et al. 2021). This offers the possibility to separate the effects of direct and indirect human activities on changes in runoff. Currently, GGP-induced vegetation greening has occurred in this region. As a large-scale indirect human activity, it could affect the regional hydrological process and even threaten the water security of the Wei River Basin. Thus, it is meaningful to understand the impact of vegetation greening on regional surface water resources because it can provide essential information for curbing the increasing water demand and ensuring ongoing water security in the transition zone between the Qinling Mountains and the Loess Plateau.

The HMETS requires daily runoff, precipitation (liquid and solid), and maximum and minimum temperature time series data. The daily observed streamflow data at the Wushan gauge from 1976 to 2015 were used and obtained from hydrological manuals published by the Hydrological Bureaus of the Yellow River Conservancy Commission, China. In addition, three meteorological stations (Lintao, Huajiling, and Minxian stations) are present near the study area. The National Meteorological Information Center was the source of daily meteorological data at all three stations, and provided the records of daily precipitation (liquid and solid) and maximum and minimum temperatures from 1976 to 2015. The areal meteorological data were obtained using the Thiessen polygon method. Furthermore, the normalized difference vegetation index (NDVI) was used to characterize changes in vegetation coverage during 1982–2015. The NDVI time series data were obtained from the GIMMS NDVI3 g (v1) dataset. Direct water withdrawal data were obtained from the annual water resources reports for Gansu Province (2008–2015), including the records in water consumption for agricultural irrigation, forestry, livestock, industry, public supply, domestic, and ecological environment. Data on rainwater cellars and reservoir impoundments were collected from the Water Conservancy Bureau of Gansu Province, including the distribution and water yield of rainwater cellars and reservoirs. The trends in areal precipitation, temperature, the NDVI, and streamflow are shown in Figure 2. The precipitation time series in the study area exhibited a weak decreasing trend, whereas the temperature time series exhibited a sharp increasing trend. The runoff time series had a significant decreasing trend in recent years. The NDVI increased rapidly after 2010, which means that increases in regional-scale vegetation cover have occurred.

A hydrological model was used to simulate the streamflow in the transition zone. We used the HMETS, which was more effective for conducting simulations in the study area. The HMETS, developed by Jean-Luc Martel, is a lumped-conceptual model based on MATLAB (Martel et al. 2017). This model can simulate hydrological processes in the water exchange process between vadose and saturated zones. The basic hydrological processes (e.g., evapotranspiration, infiltration, snow accumulation, melting, and refreezing processes) were calculated using four computation units. The model was run using 21 parameters, including 10 snowmelt model parameters, a real evapotranspiration parameter, six subsurface parameters, and four units of hydrograph parameters. These parameters are shown in Table 1.

Model calibration and validation. The Shuffled Complex Evolution-University of Arizona (SCE-UA) (Duan et al. 1992) and the dynamically dimensioned search (DDS) algorithm (Tolson & Shoemaker 2007) are typically used for model calibration. In this study, we used the SCE-UA to calibrate the hydrological model.

A novel framework for the attribution analysis of runoff variation, including vegetation greening, was developed. This framework makes use of five main elements: variation characteristics, driving factors analysis, segment identification, hydrological modeling, and attribution of runoff variation. The steps of the framework are detailed below (Figure 3):

Variation characteristics. Variation characteristics of the meteorological and hydrological variable series, including precipitation, temperature, streamflow, and the NDVI, were determined using trend and change-point analysis. Mann–Kendall (Mann 1945; Kendall 1948; Kendall & Stuart 1977; Liu & Lin 2003) and Pettitt tests (Pettitt 1979) were used to detect changes in the climatic and hydrological time series for segment identification. The Mann–Kendall test was used to assess the significance of trends in the time series. The Pettitt test was used in the hydro-climatological time series to detect abrupt changes in the mean of the variable.

Driving factors analysis. Climate change and human activities are the major factors affecting hydrological processes and causing changes in streamflow. Climate change consisted of precipitation and temperature variability in this framework, and both precipitation and temperature variability are major components of the hydrological cycle. Human activities fell into two categories, and they affected runoff changes directly and indirectly, respectively. Direct human impact mainly came from direct water withdrawal, rainwater cellar impoundment, and reservoir impoundment. Among them, direct water withdrawal referred to seven variables, as the water consumption for agricultural irrigation, forestry, livestock, industry, public supply, domestic, and ecological environment. Whereas, vegetation greening triggered by the large-scale ecological restoration programs contributed the indirect influence, and impacted the streamflow by means of disturbing the runoff generation in the study area. This framework permitted the specific impacts of climate change and human activities on streamflow to be determined in the identified sub-periods.

Segment identification. Segment identification was implemented to quantitatively estimate the effects of various factors and activities through comparison between or among multiple sub-periods. Therefore, we divided the observation period into different sub-periods based on the trend and change-point analysis. Hence, the whole period was divided into several sub-periods. Each sub-period was characterized by dominant influencing factors. Climate change and human activities are complex; consequently, the effects of individual factors can be determined when one is dominant.

Hydrological modeling. HMETS, an effective hydrological model, was used to quantify the impacts of climate variability and human activities on the hydrological cycle. The model requires representative hydro-climatological data. Thus, streamflow data under the influence of climate change and human activities were acquired.

Attribution of runoff variation. Hydrological modeling was carried out using the stationary temperature and nonstationary precipitation time series, and the output can be defined as precipitation–variation–impacted runoff (PIR). The deviations from simulated natural runoff (using the stationary temperature and precipitation) refer to the impacts of precipitation variability on streamflow. When nonstationary temperature and precipitation data were considered together, the model came up with the precipitation–temperature–variation–impacted runoff (PTIR), and the individual impact of temperature change was determined by comparing the PIR and PTIR. Furthermore, the difference between simulated PTIR and observed records can be considered to be caused by human intervention. As mentioned above, human interventions mainly manifest as direct water withdrawal, rainwater cellars, reservoir impoundment, and the indirect water loss triggered by vegetation greening. Through the water consumption and water storage data provided by the Water Conservancy Bureau and the annual water resource reports of Gansu Province, this framework would quantify the contributions of the direct human influences. Also, the remaining part in human impacts belongs to the contribution of vegetation greening in the study area.

The time series of precipitation, temperature, streamflow, and the NDVI in the typical regional area are shown in Figure 2. Increasing trends were observed for temperature and the NDVI. Decreasing trends were observed for precipitation and streamflow. The non-parametric Mann–Kendall test and Pettitt test were used to detect temporal trends and change-points in the hydro-climatological time series.

Variation characteristics of the precipitation, temperature, and the NDVI time series were determined (Figure 4). The precipitation, temperature, and streamflow time series for the last 40 years, and the NDVI time series for the last 30 years were analyzed. Increases in the temperature (Z = 4.71, p < 0.05) and the NDVI (Z = 1.32, p < 0.1) time series were observed; however, decreases were observed in precipitation (Z = −0.86, p > 0.1) and streamflow (Z = −4.10, p < 0.05). The timing of the change-points identified for the precipitation, temperature, streamflow, and the NDVI time series occurred in 1996, 1996, 1993, and 2007, respectively. Furthermore, some studies have shown that the revegetation in the study area started in the early 21st century due to the implementation of large-scale ecological restoration programs in the late 1990s (Feng et al. 2016; Zuo et al. 2016; Liu et al. 2019b; Han et al. 2020; Shao et al. 2021). Figure 4 shows the maximum NDVI increased from 0.8710 to 0.9218 during 1982–2015, and the minimum increased from 0.2707 to 0.4950. Spatially, the vegetation greening was higher in the southwest and lower in the northeast. The greening phenomenon was especially pronounced after 2010, which is consistent with the results in change-point analysis.

Trend analysis was conducted, and abrupt changes were identified. The results indicate precipitation as a means of climate change impacts, has gradually decreased since 1985, and streamflow has changed correspondingly in 1993. The temperature increased after 1996, and it induced an increase in evaporation and runoff variation.

Human activities, including direct water withdrawal, rainwater cellars, reservoir impoundments, and vegetation greening, also alter regional hydrological processes. Generally, direct water withdrawal had a stable change, while rainwater cellars and reservoir impoundment fluctuated greatly among the different periods by the influence of the occurrence time of heavy precipitation and irrigation requirement. Vegetation greening affected the runoff as an indirect human-driven factor. As mentioned above, the large-scale ecological restoration program has achieved success at the beginning of the early 21st century. Figure 5 shows the spatial distribution of the NDVI in the Wushan catchment during different periods and reveals that vegetation coverage improved from 1982 to 2015. The rapid increase in the NDVI may progressively drive some subsequent complications, e.g., higher evaporation and runoff reductions, and further pose severe challenges to regional water resources management after 2010. From the above analysis, it can be seen that there was a non-linear relationship between climate change and runoff variation in the study area, and hence the effects of human activities require further analysis.

The contributions of climate change and human activities in the Wushan catchment were determined in four periods (Period 1: 1976–1984, Period 2: 1985–1995, Period 3: 1996–2007, and Period 4: 2008–2015) which were segmented according to the results of trend and change-point analysis. Period 1 (1976–1984) was considered as the baseline period with the stationary hydro-meteorological time series, namely a near natural period, spanning over a decade. It can provide a reference value for other contribution calculations. Periods 2–4 demonstrated the joint disturbance periods by climate change and human activities, in which six sub-periods were selected depending on the representations of contribution characteristics in different stages.

Period 2 (1985–1995) showed the time when the runoff was affected by precipitation variability and direct water withdrawal, in which there is a reference period (1986–1992, low precipitation variability + direct water withdrawal, P1) and a representative period (1993–1995, high precipitation variability + direct water withdrawal, P2).

Period 3 (1996–2007) refereed to a period where precipitation variability, temperature variability, and human activities influenced the runoff together, including a reference period (1997–2000, precipitation variability + temperature variability + human activities, P3) and a representative period (2003–2006, precipitation fluctuation variability + temperature variability + human activities, P4).

Period 4 (2008–2015) was a complex impact period when the runoff was affected by climate change, direct and indirect human activities together. It can be divided into a reference period (2008–2010, precipitation variability + temperature variability + low vegetation greening + direct human activities, P5) and a representative period (2011–2015, precipitation variability + temperature variability + high vegetation greening + direct human activities, P6).

The segment identification and results are shown in Figure 6.

The relationship between precipitation and streamflow was examined in the pre-analysis of the Wushan catchment. The correlation coefficient was only 0.321, which indicated that the precipitation–streamflow relationship was weak. Thus, the hydro-meteorological time series used in the hydrological model was averaged over 20 days to more robustly estimate the precipitation–streamflow relationship. Two periods were selected for this analysis based on the trend and change-point analysis: the calibration period (1976–1982) and the validation period (1983–1984). The HMETS was calibrated and validated for the Wushan gauge against the streamflow from 1976 to 1984. The daily hydrographs for the calibration and validation period are presented in Figure 7. Statistics show the NSE value in the calibration was 0.73, and the R² and RE values were 0.73 and 2.40%, respectively, at the daily scale. In the validation, the NSE, R², and RE values were 0.79, 0.82, and 12.91%, respectively. From this, it is seen that the performance of HMETS was acceptable for streamflow simulation with precipitation and temperature changes in this study area.

Comparisons between observations and simulations are valuable for characterizing the impact of human interventions on runoff variation. In this study, we used the hydrological model to simulate the effect of climate change (precipitation and temperature), and reflected the human impacts on the runoff via the differences between observed and HMETS-simulated streamflow. Figure 8 shows the Pearson type III (P-III) curves and boxplots of observations and simulations.

The difference value of statistical features in the box plot shows the impacts of human activities on runoff change (including the land surface changes and direct human activities). From the box plot, the minimum, median, maximum values, and 25, 75% quantiles of observed streamflow had a significant decline under the influence of human activities. The influence on streamflow in the wet season, moderately wet season, and the median season was slight and less than 25%, while streamflow in the dry season and moderately dry season declined obviously with the change rate between 42 and 66%. It can be seen from two frequency curves that the impact of human activities on streamflow in the wet season was relatively limited, but had a significant impact on the dry season streamflow, which was of a small magnitude and largely increased the probability of water supply security destruction in the dry season. In summary, the government should pay more attention to drought damage in the dry season.

Observed and simulated values with the different frequencies were calculated according to the fitted P-III curve, and shown in Table 2. Amplitudes of variations refer to human impacts in this table. It is found from the variations that human activities drove the runoff reduction at all frequency values. Their influence increased from the wet season to the dry season. The once-a-century flow exhibited the largest change with a range of 83.73%, approaching zero. This indicates that the probability of water supply security destruction is higher in extreme dry years.

The above results indicate that human activities have played an important role in runoff reduction during the entire period. In this section, the contributions of climate change and human activities to runoff variation were quantified using the HMETS-based segment identification analysis framework (Figure 9) as described in section 3.2. The vegetation greening impacts were calculated by stripping the effects of other human-driven factors. Table 3 shows the results of the attribution analysis.

The results of the attribution analysis indicated that the impact of climate change decreased during the entire study period. Precipitation was the dominant factor governing runoff variation and exhibited a periodically decreasing trend. Temperature variability promoted runoff reduction, the influence of which increased gradually and became more significant when precipitation was high.

The impact of human activities has increased significantly during the study period. There were four main types of human activities affecting runoff: water withdrawal, vegetation greening, reservoir impoundment, and rainwater cellars. All these factors of human activities can have considerable impacts on runoff variation. Direct water withdrawal was the major factor contributing to variation in the runoff, and its contribution was relatively stable during the entire period. The effect of rainwater cellars was related to the implementation of manna programs, namely rainwater catchment and utilization project. Rainwater cellars had large impacts on regional hydrology, as they collected rainwater from some small catchment areas. The influence of reservoir impoundment on runoff reduction was comparatively weaker, while vegetation greening made a significant contribution to runoff decline, largely increasing after 2007.

Specifically, the primary causes of runoff reduction were precipitation variability and direct water withdrawal during Period 2 (Figure 10). The precipitation time series exhibited a fluctuating trend in Period 2, and it was positively correlated with runoff change. Comparison of P1 and P2 revealed that precipitation plays a dominant role in runoff variation. Human activities began to affect the hydrological cycle during this period.

Period 3 was used to parse the effects of precipitation variability, temperature variability, and human activities on observed streamflow. This period was divided into two sub-periods as P3 and P4. The aim was to explore how precipitation and rainwater-harvesting policies affected runoff generation. Large-scale rainwater harvesting implemented in manna programs began during Period 3. These programs, which involved the construction of rainwater cellars, tanks, and other facilities, were initiated to meet the needs of livestock production and drought-resistant crop production in water-limited regions. The relevant programs were completed in 2005. The area of rainwater harvesting reached 384,000 ha in Gansu Province at the peak. After 2011, the scale of these programs decreased rapidly. The area of rainwater harvesting in Gansu Province was only 295,590 ha in 2015. From a functional point of view, the rainwater-harvesting systems were to provide drinking water for humans and livestock from 1997 to 2005, and irrigation was the only use of the rainwater-harvesting cellars because of the development of drinking water programs after 2005. Thus, the impact of rainwater cellars stemmed from the implementation of manna programs, fluctuating due to the changes in precipitation and function in the study period. The precipitation had an upward trend periodically in P4. Correspondingly, the rainwater cellars stored more water yield under the effects of precipitation. This part of water yield was supposed to generate runoff without rainwater cellars, further affecting the runoff generation process, especially when precipitation was abundant. The attribution analysis shows the rainwater cellars had a greater effect on runoff variation in P4 and it is very significant in the rainy season. This was the reason why less streamflow was generated when precipitation was higher in P4. Positive precipitation–streamflow feedback first occurred in P4. High precipitation caused more runoff generation during this period. As some reservoirs were built in P3, reservoir impoundment became an important factor driving runoff variation.

The temperature increased steadily and made a greater contribution to runoff reduction in Period 4. High precipitation led to an increase in runoff in P6. The influence of vegetation greening became increasingly significant with the implementation of large-scale ecological restoration programs in Period 4. The statistical results indicated the effects of vegetation greening exhibited an increase from 11.73 to 25.95% after 2007 along with the improvement of catchment vegetation coverage. Vegetation greening alters soil properties, intercepts precipitation, and affects surface roughness, manifesting as indirect influences on the hydrological cycle. In addition, production–orientation change of rainwater harvesting (from drinking water to irrigation) began to be implemented in the manna programs during Period 4, and the area of rainwater harvesting gradually decreased and reached a minimum in 2015, which cause the effect of rainwater cellars decreasing during Period 4.

In conclusion, the impacts of climate change and human activities on runoff variation have increased markedly over the study period. The influence of climate change on runoff was not only caused by precipitation change in the catchment. The temperature change was becoming the key point that significantly affected the runoff evolution development. Especially when precipitation was in the decreasing trend, the contribution of temperature increasing was not ignored. Therefore, it was urgent to improve in adaptive resource management, promoting the water budget balance to hydrothermal coupling balance. The sustainable utilization of water resources was important to achieve with the all-sided management of the water cycle.

At the same time, the influence of human activities on streamflow changes was gradually becoming dominant. Vegetation coverage change, as one of most intuitive phenomena, has had a great influence on streamflow evolution recently. As is well known, vegetation greening is closely related to the implementation of large-scale ecological restoration programs by means of soil and water conservation measures (such as conversions of farmland to woodland and grassland). Vegetation greening alleviates the problem of land degradation, and the increases in vegetation improve soil conservation, water regulation and carbon sequestration. However, newly planted vegetation requires water to grow and promotes higher evapotranspiration, which implies a potential increase in vegetative water consumption. When water is limited, vegetation greening as an indirect human-driven factor could trigger runoff reduction and even diminish the amount of water available for direct human activities. Many reports have pointed out that water shortage issues may potentially be exacerbated in the future, and might have adverse socio-economic consequences. Hence, ensuring that water resources are used sustainably is a major challenge. Policymakers should be concerned with the water demand of vegetation greening in this transition area. More strategies for balancing revegetation and human water demand should be considered to promote the sustainable management of ecological restoration programs and safeguard water demand in the socio-economic system. A major future task will involve balancing the demand for water among various users in the transition zone, such as the ecosystems, farmers, and population.

In this study, a novel framework for the attribution analysis of runoff variation, including vegetation greening, was studied. The contributions of climate change, vegetation greening, and human activities to the hydrological cycle were quantified during sub-periods based on the lumped-conceptual HMETS in the transition zone between the Qinling Mountains and the Loess Plateau. The main conclusions are detailed below.

Climate change and human activities alternately played major roles in affecting runoff variation during the study period in the transition zone between the Qinling Mountains and the Loess Plateau. Climate change was the dominant factor driving runoff variation before 2000. Abundant precipitation increased streamflow, whereas low amounts of precipitation resulted in runoff reduction. The rising temperature always triggered the runoff declining, especially significant when precipitation was high. Under the greening environment, the impacts of temperature were higher and that of precipitation was reducing. Therefore, it is considered that attention should be paid to the increasing temperature in future water resource management. The contribution of human activities increased significantly after 2000, triggering the greater runoff reductions. Therein, vegetation greening altered runoff indirectly, causing a gradual decrease of runoff. Also, the contribution rate increased gradually from 11.73 to 25.95% during 2007–2015. Thus, it is recommended that policymakers should be concerned with the water demand of vegetation greening in this transition area, and adaptation strategies and policies for effectively balancing the ecological conservation and social demands are in need of development under vegetation greening.

The work described in this paper was supported financially by the National Natural Science Foundation of China (51979005) and the Natural Science Basic Research Program of Shaanxi (2020JM-250).

All relevant data are included in the paper or its Supplementary Information.