Quantitative impacts of climate change and human activities on runoff in the Huolin River catchment

Water resources are the most precious and irreplaceable natural resources in the world. Water is essential for maintaining human society and natural environmental systems, and it is also a vital link between them (Pedro-Monzonís et al. 2015; Milly et al. 2018; Pokhrel et al. 2021). Recently, it has been established that the uneven distribution of water resources has produced environmental challenges such as regional water shortages, increased dried-up days and frequent droughts (Su et al. 2018; Martin et al. 2020; Amiri & Gocic 2021a). In addition, the conflict between supply and demand for water resources has become more prominent, leading to the vulnerability and uncertainty of ecosystems (Haddeland et al. 2013; Li et al. 2021). For example, China's total freshwater volume is 2.81 × 10¹² m³, ranking sixth in the world. Although the total volume is large, the water resources per capita are only 25% of the world average (Piao et al. 2010). The distribution of water resources in China, a ‘less in the north and more in the south’ pattern, does not match the distribution of population, arable land, minerals, and economic development (Zhang et al. 2015). Under global warming conditions, the frequency and intensity of floods and drought directly linked to water resources gradually increase, and the situation will be more severe in the future, especially in northern China (Sheffield et al. 2012; Dai 2013; Su et al. 2018; Gu et al. 2020). This could adversely influence the expected effects of important projects such as grain production in northern and northeastern China, the South–North Water Transfer Project and the ecological restoration and construction of national watersheds (Jiang 2009; Nakayamaa & Shankman 2013; Xu et al. 2020).

River runoff is an important component of the hydrological cycle, and runoff variations have a considerable impact on water resources (Tao et al. 2003; Yan et al. 2017). According to earlier studies, river runoff in many parts of the world has exhibited a significant decreasing trend (Huang & Zhang 2004; Dai et al. 2009; Martin et al. 2020) and 60% of rivers in the world cease to flow for at least 1 day per year (Messager et al. 2021). It is well accepted that climate change coupled with intensive human activities has directly and indirectly affected runoff (Huntington 2006; Dai et al. 2009; Haddeland et al. 2013). Precipitation is the principal runoff replenishment source, and its frequency, intensity, and distribution directly affect the amount of runoff (Zhang et al. 2011; Amiri & Gocic 2021b). A change in precipitation of 1% will result in a change in runoff of 1.3% in northern China (Lu et al. 2013). Temperature affects runoff through two approaches. Warming can increase runoff through melting of permafrost and glaciers, in cold regions; however, the increase in temperature also enhances catchment evapotranspiration, thus reducing runoff (Cao et al. 2011; Milly et al. 2018; Gocic & Amiri 2021). This is also one of the main reasons for the decrease in runoff in arid regions. Certainly, rainfall–runoff processes are also largely regulated by surface conditions such as vegetation, antecedent soil moisture, and geographic location (Berghuijs et al. 2016; Manoj et al. 2022). Although many studies have emphasized the importance of these factors in hydrological processes, the understanding of these effects is still limited due to regional differences. Land use/land cover (LULC) changes such as afforestation/deforestation, urbanization, and wetland reclamation have led to increased extreme hydrological events. Other human activities, for example, arable land expansion and urbanization, have also added to uncertainties in the vegetation and ecosystem structure changes (Li et al. 2018; Chen et al. 2019). Effective methods of controlling excessive human activity are still lacking. Given these trends, it is critical to quantify the amount of climate and anthropogenic contributions to watershed runoff change and to identify potential driving forces, which are integral to the development of sustainable management strategies.

Recent research has primarily used empirical statistics (Wang et al. 2012; Zhao et al. 2014), elasticity-based approaches (Li & Quiring 2021), and hydrological modeling (Farsi & Mahjouri 2019; Zhang et al. 2020) to examine the effects of climate and human impacts on runoff. Empirical statistical methods facilitate calculations, but the physical mechanisms of hydrological cycle events cannot be well represented (Wu et al. 2017). Elasticity-based approaches require an accurate estimation of the potential evapotranspiration; otherwise, the quantitative results will be adversely affected. The most promising technique for assessing the hydrological response is hydrological modeling, which provides a framework for understanding and predicting the relationships among climate change, human activities, and water resources. The different spatiotemporal scales of the underlying physical processes are represented in the models (Zeng et al. 2015; Wu et al. 2017). However, hydrological models have limitations such as their complex structure, time-consuming construction, and demand for a large number of input datasets. The Soil and Water Assessment Tool (SWAT) is one of the most commonly used distributed hydrological models. It is used to model ecological, hydrological, and environmental processes under diverse climatic and management settings (Gassman et al. 2007; Wu et al. 2013; Hovenga et al. 2016; Wang et al. 2017). The SWAT model has been applied to the analysis of runoff in cold and arid regions in northwestern China, and it has been determined that climate change had a greater impact on runoff than LULC changes, such as in the Tarim River (Li et al. 2021), Heihe River (Yang et al. 2016), and Jing River basins (Yin et al. 2017). In northeastern China, relevant research has revealed that LULC changes had a significant influence on river runoff in the Nenjiang River (Li et al. 2019), Huifa River (Zhang et al. 2012), and Naoli River basin (Liu et al. 2015). Given that climate change and human activities are the primary factors influencing runoff, the local conditions should be taken into consideration.

The Huolin River is a tributary of the Nenjiang River in northeastern China, which has become a typical betrunked river with drastically decreased runoff and an increasing frequency of dried-up days. Based on an examination of the average annual runoff, that the runoff is decreasing in the Huolin River catchment (HRC) (Bian et al. 2004; Dan et al. 2021). Wavelet function analysis also supported the declining trend in runoff and forecasted that the catchment will undergo a larger scale dry period over the next few years (Lu et al. 2005). The multi-timescale variability of runoff has been investigated that the dual actions of flow evolution of natural factors and human disturbances exerted influences of different degrees on the flow signals during different periods (Lu et al. 2006). Changes in the LULC and landscape patterns in the HRC have also garnered a great deal of attention (Lu et al. 2007, 2015; Li & Liu 2011, 2012; Zhang 2014). The relevant results indicate that land use types have exhibited different trends in the HRC, and the wetland landscape has become highly fragmented. Most studies have focused on the hydrologic response to climate change or land use change in the watershed. However, research on hydrological modeling and quantitative attribution is scarce in the study area. Therefore, hydrometeorological and land use data from different periods and hydrological models were used in this study to quantify how these changes affect the hydrology in the HRC. The objectives of this study were (1) to assess the applicability and reliability of the SWAT model in semi-arid regions and (2) to quantify the impacts of climate change and human activities on the hydrological changes in the HRC using the SWAT model. The novelty of this paper lies in the quantitative estimation of the contribution of climate change and human activities to runoff in the HRC. This research improves our understanding of runoff fluctuations in typical inland catchments and provides a reference for water resource management in semi-arid China.

Since the HRC is situated in a semi-arid climate transition zone and a monsoon marginal zone (Dan et al. 2013), it is subject to the effect of the East Asian Summer Monsoon in summer, which brings heat and moisture; while it is subject to the Mongolian High Pressure System in winter, which brings dryness and cold. The average annual temperature of the catchment is 3.1 °C, its annual average precipitation is 385.5 mm, and its annual pan evaporation is 2,032 mm.

The data required to run the SWAT model include meteorological, hydrological, landform, soil property, and LULC data (Table 2) (Gassman et al. 2007; Golmohammadi et al. 2014).

The SWAT requires daily precipitation and maximum/minimum air temperature data as meteorological inputs. The daily observed meteorological data at Bayaertuhushu and Bayanhushu stations were obtained from the Inner Mongolia Meteorological Bureau. The Bayaertuhushu station is located at the headwater of the Kunduleng River (Supplementary figure, Figure S1), which accounts for 70% of the Huolin River runoff and is the most important tributary in the catchment (Sun et al. 2015). The Bayanhushu station is located in a key position where the midstream and downstream sections meet. The meteorological data from these two stations can fully reflect the climatic processes in the catchment. The daily runoff data for Tuliemaodu and Bayanhushu stations, which are the outlets for the upstream and midstream sections, respectively, were obtained from the Inner Mongolia Hydrological Bureau.

Based on the Harmonized World Soil Database (HWSD), the Inner Mongolia soil type map, and the field sampling, the soil type data were established to classify the HRC soil types into nine categories (Supplementary figure, Figure S2). At least 10 samples of each soil type were collected using the random sampling method to measure the soil properties, including the particle-size distribution, bulk density, organic carbon content, and available water capacity. Each sampling was divided into two layers (0–30 cm and 30–100 cm of depth intervals) during the 2 years of field surveys in 2019 and 2020. The mean measured data for the 2 years were used to adjust the soil property parameters in the HWSD to obtain more accurate simulation results.

The LULC maps were obtained via visual interpretation of Landsat images of the catchment. The land use in the HRC was divided into six types, namely arable land, woodland, grassland, water area (including rivers, lakes, and reservoirs), construction land (including mining areas and transportation land), and bare land (Supplementary figure, Table S1). During the 2019 fieldwork, 130 reference samples of various LULC types were gathered. A random sampling design (Hu et al. 2013) was implemented, and 500 random samples in Landsat images were collected based on the LULC classifications for 2018, 2010, 2000, 1990, and 1977. The random samples were combined with visual interpretation of high-resolution Google Earth images taken at the same time, and average user and producer accuracies of 90 and 88% were achieved, respectively. Finally, the arable land, water area, and construction land were all consistently plotted with a high accuracy. However, the mapping of woodland and grassland was often less accurate than other land use types like water or construction land. This could be due to the following reasons. First, some of the woodland samples were probably mixed with the arable land. Second, the similarity between the spectral and phenological characteristics of the mixed woodland and grassland samples resulted in increased mapping errors for these classifications. Third, due to the relatively coarse spatial resolution of the multispectral scanner/thematic mapper (MSS/TM) images, fine-scale land change processes may have been ignored.

The landforms, topography, and physical properties of soil remained relatively stable during the past 59 years and did not change significantly. Thus, we focused on the effects of factors such as precipitation and LULC on the surface runoff in the HRC.

Based on these datasets, the basin was divided into multiple sub-basins and hydrological response units (HRUs) according to the minimum threshold ratios of land use, soil type, and slope (Gassman et al. 2007; Golmohammadi et al. 2014). The HRC was divided into seven sub-basins (Supplementary figure, Figure S3) and 478 HRUs in this study. The SWAT model used the observed monthly discharge data from Tuliemaodu and Bayanhushu hydrological stations during 1960–2018. The periods of 1960–1961, 1962–1982, and 1983–1998 were used as the model warm-up, calibration, and validation periods, respectively.

The observed precipitation and temperature data for 1960–2018 were processed. The data for 1960–1982 were used to calibrate the model, and the data for 1998–2018 were used to validate the model. The first 2 years of each simulation were used as the model warm-up time. We calibrated the SWAT model at the catchment level using the observed river discharge at Tuliemaodu and Bayanhushu stations. Before running the calibration, the sensitivity of the parameters was analyzed using the Latin hypercube one-factor-at-a-time (LH-OAT) method of SWAT (van Griensven et al. 2006). This approach combines the advantages of the global and local sensitivity analysis methods and can efficiently provide a rank ordering of the parameter importance (Sun & Ren 2013). Based on the sensitivity, the top-ranked 12 sensitive parameters (Supplementary figure, Table S2) were optimized using the SUFI2 algorithm in the SWAT-CUP. The results revealed that these parameter designs and model outputs were consistent with the runoff variability in the HRC.

The measured values were higher than the simulated values in the dry season in both the calibration and validation periods, indicating that the simulation accuracy of the SWAT model for partial dry years was low. Except for the mid-1980s when there were no flood peaks in the actual measurements due to the river break, peaks occurred in the simulation. In conclusion, model overestimation or underestimation of runoff due to extreme precipitation existed to varying degrees in hydrological models.

The time series was divided into three periods: 1962–1982, 1983–1998, and 1999–2018, based on the abrupt change points of observed runoff in the HRC (Dan et al. 2021). The baseline period was set as 1962–1982 when human activities were relatively low, and the contributions of human activities and climate change to the change in runoff before and after the abrupt change points were derived (Table 4). The results of the SWAT model analysis revealed that the relative contributions of climate change and human activities were 84.5 and 15.5% during 1983–1998, 36.4 and 63.6% during 1999–2018, respectively. From 1983 to 1998, both the measured and simulated runoff increased by more than three times the average annual runoff compared to the baseline period before 1982, indicating that changes in the climatic factors caused the increase in runoff during this period, and the contribution of climate change to the increase in runoff was 84.5%. The observed mean annual runoff during 1999–2018 was approximately 7% lower than the mean annual runoff during the baseline period, indicating that human activities also caused a decrease in runoff. The contribution of human activities to the reduction in runoff was 63.6% after 2000.

Temperature increases have led to an increase in the rate of evapotranspiration, aggravating the loss of streamflow and the catchment's hydrological cycle (Milly et al. 2018). Therefore, climate was the crucial component of runoff fluctuations and changes in the ecology in the watershed. The annual temperature exhibited a significant increasing trend, with a slope of 0.24 °C per decade (Figure 3(b)), increasing from 3.3 °C during 1960–1986 to 4.1 °C during 1987–2015. Thus, the temperature has risen dramatically since 1986, when an abrupt change occurred. The highest and lowest air temperatures were recorded in 1969 and 2007, with values of 1.58 and 5.51 °C, respectively. The cumulative deviation exhibited an increasing trend during this period, indicating that the temperature was increasing. The results presented here indicate that the HRC has experienced new and more severe dry conditions since 2000.

Agricultural expansion, industrial development, and reservoir construction are examples of anthropogenic interventions in the catchment. These activities have predominantly influenced runoff by changing surface conditions. Since the 1970s, the influence of human activities in the HRC has increased due to the expansion of arable and construction land, while the areas of woodland and grassland have decreased (Supplementary figure, Table S3).

The development and use of water resources, such as dam and reservoir construction, may also have had an impact on runoff (Zhang 2014). The Hangali reservoir in the midstream section is the largest reservoir in the HRC, which covers an area of 38.8 km² and the total water storage of 0.925 × 10⁸ m³. The reservoir was put into operation around 1982, which may help partially explain the abrupt change in the amount of runoff. During the water-abundant season, the reservoir typically holds water for one month every year, and the average water inlet rate of intake sluice is 18 m³/s. The average annual water diversion of the reservoir is 0.46 × 10⁸m³, accounting for 17% of the annual average runoff. The overflow water from the reservoir is dumped back into the main channel of the Huolin River. The midstream impoundment of reservoirs intercepts the river's supply to the wetlands and has a serious negative influence on the ecosystems downstream.

The HRC has abundant forest and mineral resources, but it is extremely lacking in water resources. Climate change played a dominant role in runoff changes in the HRC during the baseline period. However, as human activities intensified, the runoff became significantly influenced by land use changes. In particular, after 2000, the effects of human activities were particularly noticeable. Reservoirs and agricultural irrigation were the major consumption activities that directly reduce runoff in the midstream section. The results revealed that the contribution of human activities to runoff changes in the midstream section was 63.6%, which was less than that in other regions of northeastern China. The results of a previous study pointed out that the contribution of human activities to runoff in the Liao River (Jiang & Wang 2016), Nenjiang River (Zhai & Tao 2017), and Songhua River basins (Wang et al. 2015; Liu et al. 2017; Xin et al. 2019) was generally more than 80%, while the combined contribution of the climatic factor was less than 20%. However, our results are credible because (1) the population density and urbanization level of the study area were low; (2) only 12% of the total arable land was irrigated; and (3) there was only one medium-sized reservoir built in the midstream section. Although the impact of human activities may be intensifying, climate change is crucial to changes in this region, which played a significant role during the baseline period.

The SWAT model in this study can be effectively used to simulate runoff in the HRC, but the simulation accuracy still needs to be improved. We primarily concentrated on the contributions of changes in the climatic and anthropogenic factors rather than on forecasting future runoff variations due to the complicated precipitation–runoff systems and data limitations in the research area. Although it has been widely employed in many studies, this method has some limitations. There is still a lack of applications for modeling long-term groundwater table fluctuations, aquifer storage changes, and water loss through surface evaporation. These variables may even change how the runoff is distributed.

The supply and demand of water resources are increasingly in conflict due to the reduced runoff in the HRC. The warm and dry climate trend in the HRC has affected the river runoff, and the increase in population and the intensity of economic activities have further aggravated the water shortage, which has been particularly prominent in the last 20 years. In this study, the impacts of climate change and human activities on the surface runoff in the HRC in northeast China were assessed using the SWAT model. The main conclusions of this study are as follows.

• The SWAT model performed well in simulating the runoff in the HRC. In the calibration period, the R² and NSE values were 0.72 and 0.7, respectively; in the validation period, they were 0.64 and 0.66, respectively. The SWAT model offers a high degree of viability and application in simulating the monthly runoff in the HRC, which can provide a scientific basis for basin water resource planning.

• Compared with the baseline period in the HRC, the precipitation decreased from 14.27 and 3.1 mm per decade, and the temperature increased 1.58 and 2 °C in human-disturbed period. In addition, the anthropogenic influence in the HRC has increased due to the expansion of the arable land and construction land, and the decrease in woodland and grassland in recent 20 years.

• During 1983–1998, the contributions of climate change and human activities were 84.5 and 15.5%, respectively; while during 1999–2018, they were 36.4 and 63.6%, respectively. Climate change contributed more in the 1980s, while human activities contributed more after 2000. The decrease in runoff in the HRC was the result of a combination of climate change and human activities.

The findings of this study provide a foundation for the sustainable use of water resources in the HRC. The results of this study also offer theoretical support for decision-makers to prevent the adverse effects of climate change in northeastern China. In future research, it is suggested to collect more data and use different LULC simulated models to predict the future climate scenarios and LULC scenarios, respectively.

This article was financially supported by the National Natural Science Foundation of China (41967052), Natural Science Foundation of Inner Mongolia (2019ZD10), and Natural Science Plan of Inner Mongolia (2019GG020).

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

The authors declare there is no conflict.