Spatial variations of river–groundwater interactions from upstream mountain to midstream oasis and downstream desert in Heihe River basin, China

With changing climate, increasing population and developing economy, water has become increasingly scarce in river basins worldwide, especially in semi-arid and arid areas, and this has triggered many environmental problems such as land desertification and ecosystem degradation (Botes et al. 2003; Marquès et al. 2013; Qin et al. 2014). For sustainable use of water resources, it is important to understand hydrologic processes such as rainfall, surface water and groundwater flow in river basins. Considering the contrast in spatio-temporal scales in which these (i.e., surface water and groundwater) processes operate, these processes have been commonly studied as independent of each other. However, as the two important components of the hydrologic system, rivers and aquifers interact in basic ways: rivers gain water from aquifers through the riverbed, they lose water to aquifers through the riverbed, or they gain in some reaches and lose in others. There is, therefore, a growing concern with river–groundwater interaction in hydrologic studies and basin-scale water resources management.

In recent years, some studies have been carried out to understand the interaction between rivers and groundwater and its driving factors on a basin scale. The work of Boyraz & Kazezyılmaz-Alhan (2014) indicate that the geometric parameters of streams such as slope and flow path of streams are influential in defining stream–groundwater interactions. A theoretical and modeling method was employed by Brunner et al. (2009) to quantify the influence of hydrogeological parameters (e.g., groundwater table depth, the thickness and hydraulic conductivity of the clogging layer) on the river–groundwater connection status. Vegetation plays an important role in the river–groundwater interaction by forming an unsaturated zone between surface water and groundwater and affecting the connection state (Banks et al. 2011; Srivastava et al. 2014). Soil is one of the factors governing the moisture exchange between soil water and groundwater and, consequently, affects the surface water–groundwater interaction (Zhu et al. 2013; Zang et al. 2013). Land use conditions are critical for the hydrology and ecology of river basins and the variations in land use could alter the flow processes of surface water and groundwater (Moiwo & Tao 2014; Ratha & Sarkar 2015). The effect of climate change on river–aquifer interactions is also well documented (Chen et al. 2011; Wang et al. 2011; Zhang et al. 2011; Leterme et al. 2012; Bourgault et al. 2014; Xu et al. 2014). The interactions between a river and an aquifer are governed to a large extent by physiography such as terrain and geology (Radhakrishnan & Elango 2011; Ali et al. 2012). In addition, human activities including agricultural development, urban construction, water withdrawal, drainage of the land surface and modification to river valleys is broad, which affect the interaction of surface water and groundwater (Winter et al. 1998; Zhou et al. 2003; Hu & Jiao 2010; Wang & Hejazi 2011; Wu et al. 2013; Li et al. 2015).

Based on these recent studies, it is indicated that many factors, including natural conditions and human activities, can affect river–groundwater interaction. However, most existing studies focus on surface water–groundwater interaction at field scale or in a relatively small basin. For large river basins that span over a wide range of hydrogeological and climate conditions like the Heihe River basin in Northwest China, very few studies have been done to reveal how the river–groundwater interactions vary in different regions such as mountain, oasis, and desert, and how human activities may affect the river–groundwater interactions in such large-scale river basins (Srivastava et al. 2014).

In view of gradual environmental deterioration and complex hydrologic characteristics of the Heihe River basin, the impact of natural conditions and human activities on the hydrologic cycle has been a great concern for researchers and government recently. Qin et al. (2013) reported the runoff characteristics varying with the forest coverage in the upper Heihe River basin. The effect of climate change on water resources in the Heihe River basin has been investigated by many researchers (Li et al. 2013; Wang et al. 2013). Furthermore, there have been some past studies focusing on the relation between other factors (e.g., land use, terrain, pedologic characteristics, and human activities) and the hydrologic behavior of the Heihe River basin (Wang et al. 2007, 2010; Xi et al. 2010). As far as the river–groundwater interaction is concerned, the previous studies mostly report the interaction between surface water and groundwater in the middle oasis reach or the lower desert reach of the Heihe River basin. A three-dimensional groundwater flow model was established by Hu et al. (2007) to simulate the hydraulic connection between aquifers and rivers in the midstream region of the Heihe River basin and analyze the discharge exchange among springs, rivers, and groundwater. Akiyama et al. (2007) used the isotope data of water collected from the Heihe River to analyze the surface water–groundwater interaction in the lower desert reaches and its seasonal variation. Results show that surface water is available in the lower desert reaches throughout the non-irrigation period while it is often dried up during the irrigation period. Owing to the deficiency of observation data, such as groundwater level and hydrogeologic characteristics, it is very difficult to develop the numerical model to simulate the river–groundwater interaction in the upstream region of the Heihe River basin.

A river constitutes a continuum of physical environments and associated ecological communities (Vannote et al. 1980). The continuum influences the hydrological processes of a river system in which downstream processes are linked to upstream processes. Although previous studies have provided valuable information about the river–groundwater interactions in the middle or lower reaches of the Heihe River, it is necessary to study the hydrologic behavior of all the three regions together to gain insight on spatial variability in river–groundwater interactions within a river basin and identify primary controls over these interactions, which is beneficial to a basin-wide integrated water resources management.

The objective of the study is to investigate the variations of river–groundwater interactions from upstream mountainous region to midstream oasis region and downstream desert region in the Heihe River basin. Geophysical and hydro-meteorological data, as well as agricultural irrigation records, are analyzed to characterize the difference in the three different regions. A river water balance is used to quantify the spatial and temporal exchange between river and groundwater flow in different regions during different seasons. Through the analysis of river–groundwater interactions, some water management recommendations are discussed. Results from the study are expected to provide some reference for water resources research and management in other river basins with complex hydrologic characteristics.

Observed river flow at five hydrological observation stations (see Figure 1) were selected to analyze the hydrological characteristics of the upstream, midstream, and downstream reaches of the Heihe River. The stations include Zhamashike Station, Qilian Station, Yingluoxia Station, Zhengyixia Station, and Shaomaying Station. In the upstream reach, there are two branches of inflows (Zhamashike Station and Qilian Station) and only one outflow at Yingluoxia Station. For midstream and downstream reaches, there is only one inlet and outlet for the other two river segments. Based on the daily discharge data collected from the five stations, the channel water gain analysis was carried out by subtracting the sum of river discharge at inlets (inflow) from the river discharge at outlet (outflow) for the three river segments.

There are very limited data of water diversion available in the downstream desert region, where there are much fewer human activities. Among irrigation zones within the downstream region, water for agricultural irrigation in Dingxin irrigation zone comes from the main channel of the Heihe River, while those in other zones come from tributary streams or reservoirs. Thus, only Dingxin irrigation zone is considered for calculating water diversion from the lower reach of the Heihe River. The crop-cultivation-area ratio values of Dingxin vs the above-mentioned irrigation zones within the midstream region were computed serving as their water-diversion ratio to estimate the amount of water diverted from the downstream river segment of the Heihe River for any particular year of 2004–2008, based on the data set of the Heihe River social economy (Zhao 2011) (download from http://westdc.westgis.ac.cnb). The relationship between monthly water diversions for each irrigation zone in the midstream region was calculated separately and the average of these ratio values was adopted to allocate the yearly water diversion in Dingxin irrigation zone to 12 calendar months.

where Q_sr is surface runoff in m³, RC runoff coefficient representing a ratio of runoff to rainfall, P_t the total precipitation in mm within the region, and S is the tributary drainage area of the river segment in km².

The CL values of the river reaches were obtained by performing geometric calculation on the attribute table of the Heihe River layer in ArcGIS. Since the channel width varies along the river segment, an equidistance measurement method was employed to acquire 20 CW values of a given river segment with the help of Google Earth satellite images overlaid with the layer of measurement locations in kmz format (prepared in ArcGIS). The average of 20 CW values is assigned as the mean width of the corresponding river segment. Daily evaporation data during 2004–2008 were collected from eight meteorological stations, including Yeniugou, Qilian, Sunan, Zhangye, Linze, Gaotai, Jinta, and Dingxin Stations, in as much as the target river segments only run through the Thiessen meteorological zones attached to these stations. Point rainfall/evaporation values at eight observation stations were converted into an average value along a given river segment by multiplying the point precipitation values with the river length ratio values (serving as weights) of their respective Thiessen polygon and then summing up all the weighted values.

The mean slope values along and across the three target reaches, as well as the mean surface slope values for three study regions, are summarized in Table 1. It indicates the gradual decreases in mean slope by 0.55 degrees along and by 17.98 degrees across the main channel of the Heihe River, from upstream to downstream. Additionally, there is a notable decrease from 20.4 to 7.1 degrees in mean surface slope from the upstream to the downstream region of the basin. The steep lateral slope in the upstream mountain region indicates the existence of a hydraulic gradient from watershed to the river, which may drive the groundwater flow into the river in the upstream river reach.

The statistic data of soil type for different reaches of the basin are shown in Table 2. For the upstream region, the area ratio of medium- vs high-permeability soil is approximately 1.5. It rises to 2.5 for the midstream region and declines to 1.1 for the downstream region. The distribution of low-permeability soil accounts for 38.7% of the downstream region, which make it comparatively difficult for rainfall to pass through the soil layer of the downstream region to recharge groundwater aquifers.

Table 3 gives the statistical data of land use in different reaches of the basin. In the upstream region, the area of forest and grass accounts for 75.6%, which is conducive to water conservation. The area ratio values of 15.3% and 1.3% are found for agricultural land and urban area in the midstream region, respectively, where a great deal of surface water is consumed to promote social development. The distribution of high evaporation areas such as desert and bare land accounts for 65.3% of the downstream region, which intensifies water shortage.

Table 4 summarizes the mean annual precipitation and three-period rainfall distribution for the three regions studied during 2004–2008. It can be observed that there is a significant decrease in annual precipitation from the upstream to the downstream region of the basin. Furthermore, rainfall of about 69%–79% is concentrated in the major-irrigation period from June to September and very little rainfall occurs in the minor-irrigation period from December to next March, regardless of the region. Since rainfall has a very close relationship to runoff, the three periods above are employed to investigate the hydrological characteristics of the upstream, midstream, and downstream reaches of the Heihe River.

Table 5 summarizes the mean flow difference between outflow and inflow in the three target segments for different irrigation periods during 2004–2008. It is seen that the mean flow difference between outflow and inflow in the upstream segment decreases greatly from 29.41 to 6.02 m³/s, as time moves from a major-irrigation period to a minor-irrigation period. Since there are few human activities in the upstream mountain area, the upstream flow difference is mainly governed by natural conditions such as meteorology and geophysical settings. In the midstream segment, the mean flow difference increases significantly from −68.16 to 19.35 m³/s, with rainfall decreasing from major-irrigation periods to minor-irrigation periods. It can be explained that the flow difference between outflow and inflow in the midstream segment is greatly influenced by human activities associated with agricultural irrigation in the midstream oasis areas. Compared to the upstream and midstream values, the flow difference between outflow and inflow in the downstream segment varies little ranging from −14.40 to −9.40 m³/s across irrigation periods, due to the drought climate and few human activities.

Area-weighted RC value, region area A_r, and the mean seasonal surface runoff Q_sr were calculated respectively for the upstream, midstream, and downstream regions of the Heihe River basin, as summarized in Table 6. It is indicated that the mean surface runoff in the upstream region is estimated as much as 1.84 × 10⁸ m³, approximately 28.9% of that in the midstream (or downstream) region during major-irrigation periods of 2004–2008, despite the relatively small size of the upstream region (equivalent to 12.1% of the midstream region and 7.4% of the downstream region in area). In medium-irrigation periods, the mean surface runoff is around 0.42 × 10⁸ m³ in the upstream region, which corresponds to 20.9% of that in the midstream region and 25.3% of that in the downstream region. For minor-irrigation periods, the mean surface runoff of 0.07 × 10⁸ m³ in the upstream region is about 8.6% of that in the midstream (or downstream) region.

Table 7 summarizes the mean difference between precipitation and evaporation in the three target river segments for different irrigation periods during 2004–2008. It is found that, for any particular irrigation period, the mean difference between precipitation and evaporation increases significantly along the river channel from the upstream to the downstream. Furthermore, all values of Q_pr minus Q_evp are negative, notwithstanding changes in river segment or irrigation period, which implies that the precipitation is less than evaporation in such an arid basin.

The mean seasonal water diversions from three different river segments of the Heihe River during 2004–2008 are listed in the column ‘Q_wd’ of Table 8. It is shown that the water diversion from the midstream river segment is much more than those from the other two segments. Besides the upstream segment without water diversion, both the midstream and the downstream segments experience a variation of water diversion with the irrigation period. The mean water diversion in major-irrigation periods was almost twice as much as that in medium-irrigation periods in the basin during 2004–2008. In minor-irrigation periods of 2004–2008, there was mean water diversion of 0.45 × 10⁸ m³ and that of 0.04 × 10⁸ m³ approximately, from the midstream and the downstream segments, respectively, to be used to irrigate the winter season crops.

Seasonal flow volume difference (∑Q_up−Q_dn) between the upstream inflow and downstream outflow was calculated by multiplying the difference between inflow and outflow (in m³/s) with the corresponding irrigation period (in seconds). After four terms of (∑Q_up−Q_dn), Q_sr, (Q_pr−Q_evp) and Q_wd are obtained, the river–groundwater exchange Q_gr can be easily quantified according to the channel water balance equation, as shown in Equation (1).

The mean values of (∑Q_up−Q_dn) and Q_gr in the upstream, midstream, and downstream river segments of the Heihe River for different irrigation periods during 2004–2008 are shown in the columns ‘∑Q_up−Q_dn’ and ‘Q_gr’ of Table 8. Results show that groundwater discharge to the river segment occurs during all seasons in the upstream mountain region. In the midstream oasis region, groundwater is recharged by river during the major-irrigation and medium-irrigation periods, while groundwater discharges to the river in minor-irrigation periods. This shows that human activities significantly affect the river–groundwater interaction in the midstream oasis region, which results in less outflow than inflow in the midstream river reach. As a result, it reduces the river inflow to the downstream river reach. In the downstream desert region, the seepage flow of the river to the aquifer is found for all seasons, which is opposite to the river–groundwater exchange in the upstream mountain region.

In the upstream mountain region of the Heihe River basin, an extensive area of forest and grass growing on sandy loam or loam textured soils can help hold moisture in soils or recharge aquifers when a certain amount of precipitation falls in the form of rain or snow. Furthermore, relatively steeper slopes exist along and across the river channel in the mountain area. On the one hand, a river segment with a steep gradient will have a faster flow velocity and, as a result, the river water quickly moves along the channel and hardly percolates into the aquifer through overlying unsaturated materials. On the other hand, due to the high altitude and steep slope of the Qilian mountainous area, abundant groundwater, driven by gravity, moves away from recharge areas in the mountain to the adjacent valley and consequently converted into river water. Thus, the upstream river segment gains water from groundwater (a positive value of Q_gr) regardless of the irrigation period. In addition, a large quantity of precipitation (over 70% of annual precipitation) falls and alpine glaciers melt in summer seasons, which contributes to a rise in the groundwater table in the unconfined aquifer and, consequently, results in more water flowing into the river (Cai et al. 2014a). This can explain why 49.4% of the groundwater discharge to the river is found as about 1.27 × 10⁸ m³ in major-irrigation periods whereas it is 22.2% in medium-irrigation periods and 28.4% in minor-irrigation periods, for the upstream river channel. It is, therefore, indicated that the characteristics of river–groundwater interaction are mainly controlled by geophysical settings and precipitation.

Compared with the upstream region, the midstream oasis region comprises relatively flat terrain and wide river channels, which decreases the velocity of river water. Hence, the slow-moving river water has time to move downwards to groundwater when the groundwater level is lower than the water level in the river. The midstream region is mainly veneered with loam textured soil of medium permeability, which is beneficial to water exchange between the river and groundwater. The high evaporation area (e.g., Gobi, desert, and bare land) and the human activity area (e.g., agricultural and urban land) accounts for 41.85% and 16.65% of the midstream region, respectively. The large-scale groundwater withdrawals for irrigation and domestic use and high evaporation rates give rise to low groundwater-level elevation. The groundwater level may vary due to the seasonal changes, such as agricultural irrigation and precipitation in the farming regions (Chen et al. 2003; Cai et al. 2014b). In major- or medium-irrigation periods, a large amount of groundwater is withdrawn for irrigating crops, which decreases greatly the groundwater level. Meanwhile, due to the precipitation recharge in the upstream mountain area, the river flow increases and, as a result, the water level rises in the midstream river channels. Thus, the groundwater level is lower than the water level in the river channels and, consequently, the river channel loses water to the adjacent groundwater aquifer in the midstream oasis region during regular irrigation months from April to November. In minor-irrigation periods with little precipitation, the runoff from the mountain area decreases significantly, which leads to a very low water level in the channels. However, owing to the irrigation return flows and few groundwater withdrawals, a rise in the groundwater level is maintained during winter months. Hence, the groundwater level is usually higher than the river water level and the river–groundwater interaction can be described generally as groundwater discharging to river channels in the midstream reach of the Heihe River during minor-irrigation periods. Notwithstanding the water flow direction during the river–groundwater exchange varying with irrigation period, the river–groundwater interaction in the midstream region is dominated by human activities, such as groundwater withdrawals and agricultural irrigation, as well as precipitation.

Different to the upstream and midstream regions, the downstream region receives very little precipitation all year round, which makes groundwater difficult to be recharged from rainfall. Furthermore, over 65% of the downstream region is covered by high-evaporation lands such as Gobi, desert, and bare land. Little precipitation and high evaporation results in a continued low groundwater level. As a consequence, the groundwater level is lower than the river water level in the downstream region, whatever the irrigation period. The river and groundwater, therefore, interact through groundwater recharge from the river in the downstream region. In addition, groundwater pumping for agricultural irrigation can affect the water exchange between the river and the groundwater. It can explain why the river–groundwater water exchange in major- and medium-irrigation periods is much greater than that in minor-irrigation periods. The interaction between the river and groundwater is significantly governed by evaporation and human activities, which agrees with the work of Wang et al. (2014). It may also be affected by geological heterogeneity, as discussed by Fleckenstein et al. (2006).

Given the inherent non-linearity and complexity of hydrological processes at any scale (Wu et al. 2014), the research community at times has to rely on simplistic empirical formulae as used in this study. Equations (2)–(4) have been effectively used in hydrology and water resources research and applications. GIS spatial analysis tool has been used to analyze high-resolution GIS data and produce the appropriate runoff coefficient distribution map (Figure 12) based on the table given by Browne (1990). This increases the reliability of the result from the empirical runoff equation. For evaporation and precipitation water volume estimations, no empirical coefficients exist in the equations. Channel width and length are obtained by GIS spatial analysis of high resolution GIS data, which would result in reliable data for estimating precipitation and evaporation water volume by the empirical equations.

Through comparisons among all terms in the balance equation, it is indicated that large water diversion and groundwater recharge from the river channels lead to a reduction of the river flow in the middle and downstream regions and, as a result, the blanking of lower reaches arises frequently and even some river channels have dried up or are drying up. For protecting the ecological environment in the Heihe River basin, particularly in its downstream region, it is suggested that the area of cultivated land should not be expanded further in the basin. Moreover, some ways to save water, such as sprinkling and drip irrigation, should be taken instead of surface flood irrigation. As well as this, it is essential to establish the related computational models to effectively manage the exploitation and utilization of water resources in the Heihe River basin.

Characteristics of river and groundwater interactions in the Heihe River basin have been investigated for the upstream mountain region, midstream oasis region, and downstream desert area. A river water balance equation was modified to represent the natural features and human activities of the Heihe River, and to estimate the water exchange between the river and groundwater. Through the analysis of geophysical characteristics, meteor-hydrological characteristics, and water balance in the river reaches, results show that the tempo-spatial characteristics of river–groundwater exchanges vary geographically along the Heihe River from the upstream mountain region to midstream agriculture area, and the downstream desert region during different seasons. This study intended to study the regional contrast in surface water–groundwater interactions on an annual and seasonal basis. Therefore, the methodology presented here was deemed appropriate to achieve the objectives of the study. Furthermore, given the spatio-temporal scale of the problem in hand, some deviations in the selected empirical coefficients should not effect our findings.

In the upstream mountain region, the river channel gains water from the ground in all seasons during the year, with a groundwater inflow of approximately 2.57 × 10⁸ m³ per year. With the net groundwater inflow to the river, outflow is larger than inflow in the upstream river reach. Physiographic mountain features and precipitation are the major factors controlling the groundwater flow to the river. Steep watershed slopes (about 20 degrees) across the river result in the higher groundwater table in the watershed, and produces the gravity gradient driving groundwater flow into the river.

In the midstream oasis region, the mild lateral slope across the river in the watershed is about 2 degrees. Human activities of water withdrawal during different agricultural irrigation seasons significantly affect the river–groundwater interaction. During the middle-major irrigation period from spring to autumn seasons (April to November), river–groundwater interaction is characterized by the net river water leaks into the ground (about 7.43 × 10⁸ m³ per year), which results in less river outflow than the inflow in the river reach. However, during the minor-irrigation period in the winter (December to March), there is a net groundwater inflow to the river (about 2.24 × 10⁸ m³ per year).

The downstream desert area is characterized by the low precipitation, high evaporation, and near-flat lateral slope of about 1.77 degrees. Opposite to the upstream region, the river–groundwater interaction in the downstream desert area is characterized by the net leakage flow, with the annual net loss about 10.39 × 10⁸ m³, from the river to the ground during all seasons. Generally, several factors contribute to the leakage of river water to the ground in the downstream desert area, including near-flat watershed, poor land cover with high evaporation, and water withdrawal activities. The large water diversion and groundwater recharge from the midstream oasis region result in a notable decrease of river inflow in the downstream desert region. In order to protect the ecological environment against desertification in the downstream desert region, adopting water-saving irrigation techniques in the midstream oasis region is important to reduce the leakage of river flow to the ground so as to increase the river inflow to the downstream desert area.

The study is financially supported by the National Natural Science Foundation of China (No. 91225301), the Scientific Research Foundation for the Returned Overseas Chinese Scholars, State Education Ministry (No. 2013-1792), and the Fundamental Research Funds for the Central Universities. Special thanks go to the Center for Water Research of Beijing University and the Cold and Arid Regions Science Data Center at Lanzhou for providing the data.