Quantifying the impact of climate change and human activities on the eco-hydrological regimes of the Weihe River Basin, Northwest China

River streamflow and flow regimes determine many fundamental ecological processes in river ecosystems, which in turn affect river ecosystem health (Poff & Zimmerman 2010). With the development of the social economy and population increase, human interventions have caused a massive change in the hydrological cycle of basins and consequently altered the eco-hydrological regimes of rivers to varying degrees, adversely affecting the health of river ecosystems over the past several decades (Zhu et al. 2021; Cui et al. 2022; Jiang et al. 2022; Wang et al. 2022b). River atrophy, sediment deposition, water pollution, and biological diversity are persistent problems (Nienhuis et al. 2020; Zhang et al. 2020; Luo et al. 2021; Wang & Sun 2021). Analysing the degree of change in river eco-hydrological regimes and quantitatively differentiating the impacts of climate change and human activities on the evolution of eco-hydrological regimes are of great scientific importance for river health management with environmental change adaptation.

In recent years, an increasing number of experts have focused on quantifying the evolution of eco-hydrological regimes. The indicators of hydrological alterations (IHAs) with 32 parameters developed by Richter et al. (1996) can be used to quantitatively evaluate the variable characteristics of hydrological regimes and serve as an important foundation for further investigation of hydrological processes. Richter et al. (2003) proposed the range of variation approach (RVA) based on IHA indexes to evaluate the degree of hydrological change and quantitatively set the range of environmental flow change in a natural state. Yang et al. (2012) analysed the hydrological index characteristics of the main channel of the Yellow River in accordance with the RVA method and evaluated the ecological benefits of the river. Lian et al. (2012) used IHA indexes to analyse the impact of human activities such as the construction of locks and dams in the Illinois River, a tributary of the Mississippi River. However, the 33 IHA indexes were too numerous to be easily calculated and analysed, making it challenging to operate and manage water resources in practice. In addition, there is an issue with autocorrelation among the IHAs. Yang et al. (2008) used principal component analysis (PCA) and other techniques to select six indicators as ecologically relevant hydrological indicators (ERHIs) of the Illinois River to simplify the IHA indexes. In addition, Vogel et al. (2007) examined the relationship between dam construction and eco-hydrological regimes by combining flow duration curves (FDCs) with eco-deficit and eco-surplus indicators, a technique that was recently implemented in the Yangtze and Yellow Rivers (Zhang et al. 2016, 2018; Gao et al. 2018). By using the eco-deficit and eco-surplus indicators, the changes in the eco-hydrological regimes of a basin can be observed more intuitively.

In addition, previous research shows that global warming has become an indisputable fact (Ji et al. 2014). With the rapid development of the social economy, it has become impossible to ignore the increasing anthropogenic influence (Jiang et al. 2019). Examples of global environmental problems include river cut-offs (Belletti et al. 2020), ecosystem degradation (Palmer & Ruhi 2019), and a severe shortage of water in meeting ecological water demand (Yan et al. 2022). Under the combined influence of climate change and anthropogenic activities, the characteristics of the water cycle in basins have changed dramatically, especially regarding the evolution of eco-hydrological regimes (Jiang et al. 2021). Therefore, a suitable approach is required to distinguish between the contributions of climate change and anthropogenic activities to eco-hydrological regimes. The simulated-observed comparison method is widely used to quantify the impact of environmental changes on runoff evolution by comparing runoff characteristics under natural and human-influenced conditions (Jiang et al. 2011, 2019). This method requires minimal data and can be used to reconstruct the natural streamflow process with a suitable hydrological model.

The Weihe River, which flows through the arid and semi-arid areas of Northwest China, is the largest tributary of the Yellow River. There is a clear contradiction between the development of water resources and the ecological environment. The ecological water demand is frequently displaced by the national economic water demand (Liu et al. 2022), resulting in a series of environmental problems (Ma et al. 2022). However, most studies on hydrological alterations in the Weihe River basin (WRB) have not used an in-depth approach to improve the redundant IHA indexes, nor have they evaluated the relationship between hydrological changes and ecosystems.

In this study, the effects of climate change and anthropogenic activities on the eco-hydrological regime alterations of rivers (the ERHIs, eco-surplus, and eco-deficit metrics) were quantitatively distinguished using the simulated-observed comparison method to provide a scientific foundation for river health management with adaptation to an environmental change.

Meteorological and hydrological stations are established in the WRB (Figure 1). All meteorological stations possess the daily precipitation, maximum, mean, and minimum air temperatures, and wind speed data from 1 January 1961 to 31 December 2017. These datasets were acquired from the China Meteorological Data Service Centre (http://data.cma.cn/en). Two hydrological stations, Xianyang and Huaxian, with drainage areas of 4.68 × 10⁴ and 10.6 × 10⁴ km², respectively, were selected as the control sites. Both sites had complete daily streamflow records from 1 January 1961 to 31 December 2017. The Yellow River Conservancy Commission supplied the data from hydrological stations. Moreover, three GIS data layers were utilised for variable infiltration capacity (VIC) modelling: the global 3-s Shuttle Radar Topography Mission digital elevation model (DEM) data provided by the U.S. Geological Survey (USGS) (https://lta.cr.usgs.gov/SRTM); the 30-arcsec global DEM data from the USGS website (https://www.usgs.gov/); and the global 1-km land cover classification product developed by the University of Maryland (UMD) (Hansen et al. 2010).

The Mann–Kendall trend test, the Pettitt test, and the double mass curve were applied to the annual runoff data of the WRB from 1961 to 2017 to demonstrate the hydrological process variability. Based on the outcomes of transition point selection, the study period can be divided into the baseline (i.e., less disturbed by human activities) and the human-induced (i.e., more affected by human activities) periods.

The VIC model was used to reconstruct natural runoff in the WRB. VIC is a macroscale distributed hydrological model that aims to balance water and surface energy budgets within the grid-cell (Jiang et al. 2021). It simulates terrestrial surface–atmosphere fluxes of moisture and energy such as evapotranspiration, surface runoff, baseflow, radiative fluxes, turbulent transport fluxes, and sensible heat within the grid-cell. The surface runoff and baseflow components of the gridded runoff are subsequently routed to the basin outlet. The model parameters can be divided into two categories. The first category of parameters is left unadjusted once calculated from the Harmonized World Soil Database and comprises soil porosity θ_s (m³m⁻³), saturated soil potential ψ_s (m), saturated hydraulic conductivity k_sat (ms⁻¹), and exponent B for unsaturated flow, as well as the UMD global land cover classifications (root depth and root fraction). The second category includes user-calibrated parameters (variable infiltration curve parameter b and thickness of different soil layers d) based on the agreement between the simulated and observed hydrographs at the basin outlet (Liang et al. 1994; Wang et al. 2020).

To estimate the relative contributions of climate change and human-induced activities to streamflow changes, the authors assumed that climate change and human activities were the two primary drivers that independently caused streamflow changes in the WRB (Wang et al. 2008). This assumption permits a linear and additive relationship between the relative contributions of climate change and human activities. Based on the change detection, the study period can be divided into baseline and human-induced periods. The VIC model was then used to simulate natural runoff processes in the human-induced period using parameters calibrated in the base period and meteorological forcing variables in the human-induced period. It was assumed that the reconstructed streamflow was free of human disturbances.

In the current study, we employed PCA to remove redundancy and correlation among IHAs, thus improving the RVA method. The PCA method adopts the concept of dimension reduction, and based on the premise that the essence of original data is retained with no or minimal loss of information, the large number of originally related and redundant variables are transformed into a small number of new relevant variables. Each principal component is generated from the linear combination of the original variables and is made up of the new comprehensive variables. The PCA method is described in detail by Cheng et al. (2018).

Richter et al. (1998) proposed the RVA method, which was based on the IHA index system and can quantitatively define the degree of change in a single hydrological index after it has been affected, compare hydrological data before and after alteration and evaluate the degree of change within the hydrological index. For non-parametric statistics, 75 and 25% of the occurrence frequency of indicators are used as the upper and lower limits, respectively (i.e., the RVA threshold), to meet the ecological demand of rivers (Wang et al. 2022a). The IHA index of the year following the transition point was estimated, and the degree of change to the hydrological index was estimated to determine the extent to which hydrological series were affected.

Vogel et al. (2007) derived dimensionless eco-metrics (eco-surplus and eco-deficit) based on FDCs to quantify seasonal and annual ecological runoff changes. An FDC was estimated using daily streamflow data during a time interval of interest; it was used to obtain a measure of the time (in percentage) during which streamflow equals or exceeds a certain value, which can indicate the relationship between runoff and its frequency in a simple and intuitive manner.

Considering that human activities primarily affect the seasonal ecological runoff by altering the annual runoff distribution, the eco-surplus and eco-deficit changes were analysed at annual and seasonal time scales. The annual and seasonal FDCs for each year were calculated using the observed daily flow data of each hydrological station. The streamflow series before the transition point was defined as the natural state, and the annual and seasonal FDCs created using the daily streamflow data before the transition point were set as the 25th and 75th percentile threshold values. If the FDC exceeded 75% of the values for any year and season, the ratio of the area enclosed by the given FDC and 75th percentile FDC to the long-term average annual or seasonal flow was defined as the eco-surplus. This indicated that the hydrological factor exceeded the demand value of the river ecosystem for this factor. If the FDC was lower than 25th percentile, the ratio of the area enclosed by the given FDC and 25th percentile FDC to the long-term average annual or seasonal flow was defined as the eco-deficit. This indicated that the hydrological factor was lower than the demand value of the river ecosystem for this factor (Gao et al. 2012). The eco-surplus and eco-deficit indicators were defined as eco-metrics.

Comparing the FDCs of simulated runoff to that of measured runoff revealed that the observed high flow was lower than that of the simulated data after the transition point in summer. This indicated that, relative to the natural flow regime, the actual flow was significantly reduced in summer. This led to a decrease in the summer eco-surplus, which may be attributed to the construction of numerous water conservation projects based on the Weihe River in recent decades. Changes in FDCs between the two scenarios can be used to estimate the preliminary characteristic differences in annual and seasonal ecological flows, and the change in ecological indicators cannot be alienated from the impact of precipitation. Given the correlation between precipitation anomalies and eco-metrics, this can provide a preliminary foundation for the analysis of eco-metrics caused by climate change.

Analysing the simulated eco-metrics at each time scale, it was found that the eco-surplus was dominant in summer, and the eco-deficit was considerably smaller than the eco-surplus. The eco-deficit and eco-surplus of other seasons and at the annual scales were stable and remained within a relatively narrow range near zero. There was no discernible temporal variation, indicating that the effects of climate change did not have a significant impact on the eco-metrics in the WRB.

Numerous studies have attempted to determine the reasons for runoff fluctuations in the WRB. Liu et al. (2022) showed that climate change exerted specific impacts on runoff reduction in the WRB, but that this impact was minimal. While analysing the impact of climate change on runoff in the WRB, precipitation is often regarded as the driving factor directly related to runoff, as well as the primary source of runoff recharge in the WRB. The annual average precipitation decreased from 563.5 mm during the baseline period to 514.9 mm during the human-induced period; however, the trend analysis revealed no significant decreasing trend for precipitation. Both Guo et al. (2017) and Liu et al. (2022) discovered that precipitation in the WRB did not decrease significantly. Climate change is a slower process than the intense human activity in recent decades. In this study, we used ERHIs and eco-metrics to analyse the changes in eco-hydrological regimes to explore the attributes of river ecology as opposed to runoff only. Similar to the runoff change attribution, the effect of climate change on the alteration of eco-hydrological regimes was limited.

Contrary to the human activities of the past few decades, climate change is a slow-growing process with temporal scales spanning from a hundred to a thousand years (Parmesan & Yohe 2003), thus making it difficult to accurately understand this change using short-term data records. Different results may be obtained when analysing the hydrological response to climate change, depending on the selection of the baseline and study periods.

The attribution analysis revealed that human activities contributed more to the decline of eco-hydrological regimes in the WRB than climate change (Figure 10). Anthropogenic interventions such as land-use changes, water resource consumption, and large-scale water-control projects could all have impacts on regional water resources.

The land in the WRB is mainly cropland and grassland, while cropland acreage continuously diminishes during the human-induced period. A major factor was the Grain-for-Green (GFG, Chen et al. 2007) project that is established by the Chinese government in 1999 which aimed to reduce soil erosion and increase vegetation area by converting sloped cropland into vegetative-friendly land. The substantial increase in vegetation coverage caused by the GFG project could decrease surface runoff, enhance evaporation from soil moisture, and intensify plant transpiration (Li et al. 2016). The expanded vegetation canopy would intercept more precipitation, the majority of which would eventually evaporate into the atmosphere. On the other hand, the continuous advancement of urbanisation could not only increase the impervious area but also weaken the infiltration, resulting in the increase of interception and direct runoff. As a further matter, a portion of the intercepted water will be consumed through evaporation, while the majority of the surface runoff will contribute to urban water circulation by the ‘Rainfall Collection’ (RC) project implemented in 1996 (Gao et al. 2013). According to the statistics, two million cellars have been constructed in recent decades to collect rainfall with an annual water storage capacity of approximately 6.0 × 10⁶ m³ for drinking water supply and irrigation during the dry season. As a result, the water flowing into the river dropped significantly (Gao et al. 2013).

Moreover, vigorous socio-economic development accelerates the consumption of water. The mid-downstream region of the WRB (i.e., the Guanzhong Plain) faced the most rapid economic growth. In recent decades, agricultural irrigation, industrial development, and domestic use have been the primary drivers of rising water demand. In the 1960s, the average annual water consumption was 7.3 × 10⁸ m³, which increased to 19.7 × 10⁸ m³ by the 1990s, representing 46% of the observed streamflow at the Huaxian station (Gao et al. 2013). Ren et al. (2016) reported that irrigation accounted for 71% of total water consumption in the WRB with an average annual amount of 16.7 × 10⁸ m³ and an annual increase ratio of 0.4 × 10⁸ m³ yr⁻¹ during the period from 1991 to 2013. It was evident that agricultural irrigation was one of the most significant water use sectors in the WRB.

In conclusion, human activities impacted the regional eco-hydrological regimes of the WRB in a complex manner. Changes in land-use types could alter eco-hydrological regimes in different ways. For example, afforestation can reduce surface runoff by increasing interception and evapotranspiration, whereas urbanisation can increase surface runoff and weaken infiltration by increasing impervious areas. Moreover, water-control projects, such as reservoirs, could exert a considerable impact on flow convergence by regulating floods and redistributing streamflow. This could subsequently cause seasonal variations in streamflow patterns. In the meantime, agricultural hyper-irrigation consumed substantial volumes of water, and this is also one of the reasons for the reduction of runoff in the WRB.

This study focused on detecting and attributing eco-hydrological regime changes in the WRB based on the VIC modelling framework. However, there were no exact values for the contributions of climate change and human activities to eco-hydrological regime changes. One should keep in mind that this method is limited to fully and objectively describing eco-hydrological regimes in a changing environment because of the complex nature of eco-hydrological regimes and the high non-linearity in the interaction between human society and nature systems. Therefore, it is necessary to evaluate uncertainties in the modelling findings. Uncertainties in this study are considered from three aspects: basic assumptions and associated simplifications, selected hydrological model, and model parameter calibration.

The attribution framework presumed that climate change and human activities can independently affect eco-hydrological regimes in the WRB (Dey & Mishra 2017). Without this basic assumption, one cannot separate the contribution of these two components, because this assumption neglects the inter-correlation among the climate system, hydrological cycle, and land surface. For example, afforestation could indirectly increase evapotranspiration by enlarging canopy interception and enhancing soil and water conservation, thus affecting regional climate conditions. Therefore, this assumption is a simplification needed to solve scientific questions, which inevitably introduces uncertainty.

Hydrological models are an approximation of natural processes. Different models have different structures and thus different regional applicability. Previous studies have shown that the VIC model has good applicability in the northern semi-arid region of China and thus can reduce the uncertainty of hydrological evolution attribution (Ren et al. 2016; Jiang et al. 2019, 2022). However, the VIC model requires multi-source datasets for both model input and a priori parameter estimation, thereby potentially introducing uncertainties arising from these data.

Parameter calibration is another significant source of uncertainty for the VIC model. There are two key main sources of uncertainty in the VIC model calibration. One source of uncertainty is the calibration strategy. Although the manual calibration employed is efficient with the insights of a modeller's priori experience, it is still unclear whether one eventually obtains the global optimum or not (Tolson & Shoemaker 2007). Another source of uncertainty in the VIC is from the conflict among spatially varied parameterisation, simulation accuracy, and computational burden. Considering the computational burden, a regionally uniform calibration strategy was conducted in the WRB.

This study proposed a comprehensive assessment framework to distinguish the relative contributions of climate change and human activities on the eco-hydrological regimes at the daily, monthly, seasonal, and annual timescales. Firstly, the variation trends and abrupt changes of hydrological variables in the WRB between 1961 and 2017 were analysed. Then, seven representative IHAs were selected as ERHIs and used in combination with the eco-deficit and eco-surplus metrics to assess the eco-hydrological regime changes. Finally, the relative contributions of climate change and human activities on eco-hydrological regime changes were explored by using the simulated–observed comparison method. The principal findings of this study can be summarised as follows.

• 1.

  Hydrological variation analysis revealed that the annual precipitation in the WRB did not decrease significantly between 1961 and 2017, whereas the annual streamflow at the Xianyang and Huaxian stations showed a significant decreasing trend. Moreover, abrupt transition points were identified in the streamflow series in 1990, thereby dividing the study period into the baseline (1961–1990) and human-induced (1991–2017) periods. Calibration and validation of the VIC model showed good performance during the baseline period, indicating that this model was appropriately benchmarked.

• 2.

  Seven out of 32 IHA indexes were selected using the PCA method: fall rate, 7-day max, date min, date max, January, December, and high pulse count. These were regarded as ERHIs of the WRB. The trends of these indexes indicated that the flood peak discharge and total runoff were decreasing. Furthermore, the eco-surplus of the two hydrological stations was primarily observable before the transition point, whereas the eco-deficit was observable after it. The eco-deficit at annual and seasonal scales increased, whereas the eco-surplus tended to decrease. This indicated that the river ecological water demand in the WRB was insufficient during the human-induced period.

• 3.

  The attribution analysis revealed that human activities were responsible for more than 80% of the ERHI changes at the Xianyang and Huaxian stations and for more than 60% of the eco-deficit and eco-surplus of these stations. Overall, human activities were the major factors leading to the decline and shifts in the eco-hydrological regime of the WRB.

This work was financially supported by the National Natural Science Foundation of China (U2243203, 51979069); the National Natural Science Foundation of Jiangsu Province, China (BK20211202); and the Research Council of Norway (FRINATEK Project 274310).

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

The authors declare there is no conflict.