Analysis of Variation Trend and Driving Factors of Baseflow in Typical Yellow River Basins

Abstract

Baseflow is a stable part of streamflow and the main component of streamflow during the dry season. Baseflow plays an important role in the water cycle, and in ecological environment protection of the Yellow River basin (YRB). Taking the Zuli, Kuye, and Tuwei basins, and the Jingle sub-basin as examples, the baseflow was separated using a recursive digital filtering method. The intra-annual, inter-annual, and chronological characteristics of baseflow and the baseflow index (BFI) were analyzed, and driving factors were analyzed from the perspective of climate-change and human-impact factors. The results showed that: (1) The annual baseflow of the basins mainly declined, trending downward in all four test basins, while the BFI increased in two of the basins and remained nearly constant in the other two basins; however, the distributions of baseflow and the BFI were more uniform. (2) The intra-annual patterns for baseflow and the BFI showed changes between earlier and later periods. (3) Precipitation and soil and water conservation measures were the primary driving forces of baseflow change in the basins. The influence of the former weakened while the latter strengthened, and coal mining in the Kuye River and Tuwei River basins also influenced baseflow significantly. (4) When the normalized difference vegetation index (NDVI) < 0.375, the BFI of test watersheds gradually decreased with the increase in the NDVI. When 0.375 < NDVI < 0.65, the BFI of the basins gradually increased with the increase in the NDVI and the underlying surface continued to improve. When NDVI > 0.65, the increase rate of the BFI decreased and the soil and water conservation capacity of the underlying surface tended to be stable.

1. Introduction

“Resources, environment, population, and development” are the primary problems in sustainable-development research in today’s society [1]. As a relatively stable part of streamflow, baseflow is the main source of river streamflow in arid and semi-arid regions [2,3]. The main difference between baseflow and streamflow is regarding the reduction, smoothness, stability, and hysteresis of the flow and cycle [4]. Baseflow generally refers to the streamflow from groundwater or other delayed flows [5], and is the main recharge source of streamflow in the dry season. Maintaining stable baseflow is important for water supply and in ecological environment protection for ensuring production and life in the basin. Meanwhile, baseflow also plays an important role in water-resource security, evaluation, and regulation and management. It is also important for the rainfall–runoff relationship simulation [6,7,8].

Baseflow change mainly depends on the rainfall process, soil and geological indicators, and hydrological characteristics of the basin, and is also affected by the spatial distribution of vegetation and evapotranspiration in the basin; baseflow also increases or decreases depending on human activities [9,10,11]. Mu et al. [12] used a single-parameter digital filtering method to separate baseflow in the Weihe River basin, and found that climate change dominated the surface streamflow variations, while human activities controlled the baseflow variations. Wilby et al. [13] calculated the influence of climate change on baseflow using a large-scale hydrological model and found that the dominant factors of the change in baseflow were different in different periods. Climate change was the dominant factor in the normal and wet periods, while land-use change was the dominant factor in the dry period. Lyu et al. [14] used a single-parameter digital filtering method to segment baseflow in several mountainous watersheds in the North China Plain. Results of the analysis showed that precipitation, evapotranspiration, soil texture, vegetation, and geomorphic factors (i.e., elevation and slope) were the main factors influencing baseflow changes in the summer months, while in the winter months baseflow was associated with the karst geological distribution. Liu et al. [15] used a digital filtering method to segment streamflow in the Huangfuchuan basin and analyzed the periodicity, trend, and mutation of baseflow. They found that the average annual baseflow in the basin showed a significant downward trend, which was mainly affected by climate change and human activities. Zhang et al. [16] were in agreement with Liu et al. Zhang et al. [17] used a digital filtering method to segment baseflow in the upper reaches of the Heihe River and analyzed the variation and influencing factors of baseflow. The average baseflow index (BFI) in the upper reaches of the Heihe River was greater than 0.4, and the change in baseflow was primarily affected by precipitation. Yang et al. [18] used the Eckhardt segment method to separate baseflow in two different geomorphic regions of the middle Yellow River basin. They believed that implementing comprehensive measures such as vegetation restoration, water retention dams, terracing, and other ecological measures could help mitigate the rate of baseflow reduction. Analysis of the variation trend and driving factors of baseflow in the basin can provide references for rational development and utilization of water resources, formulation of soil and water conservation measures, and planning ecological environment protection strategies [19,20].

Baseflow is an important part of water resources in the Yellow River basin. Streamflow in the Yellow River basin is mainly replenished by baseflow during the dry season [21,22]. With the rapid development of the social economy in the Yellow River basin, the difference between supply and demand of water resources has become increasingly prominent. Therefore, it is of great significance to determine the variation in baseflow and its driving factors for studying the water cycle and ecological environment protection in the Yellow River basin. The purpose of the study in this paper was to: (a) use a recursive digital filtering method to separate baseflow at the outlet hydrologic stations of the test basins, (b) characterize changes in baseflow and the baseflow index in the watershed at different time scales, and (c) explore the driving factors of baseflow change from climate-change and human-impact factors.

The novelty of this paper lies in the further in-depth study of the drivers affecting the evolution of river baseflow in typical watersheds of the Loess Plateau in the context of the intensified impacts of climate change and human activities in recent years [23]. It provides a scientific basis for the rational utilization of water resources and better construction of the local ecological environment in the Yellow River basin.

2. Materials and Methods

2.1. Study Area

The Zuli River is a first-grade tributary in the right bank of the upper reaches of the Yellow River. The Zuli River basin (ZLB) is located in the hilly and gully region of the Loess Plateau (104°12′–104°33′ E, 35°18′–36°34′ N). The average annual temperature of the watershed is 6.4° and the elevation ranges from 1500 to 2000 m. The length of the main stream is 220.3 km, with an average slope of 19.4‰. The drainage pattern is innate. The basin has a continental monsoon climate, with more annual precipitation in the south than in the north, mainly in summer, followed by spring and autumn, and the least precipitation in winter [24].

The Kuye River is a first-grade tributary in the right bank of the middle reaches of the Yellow River. The Kuye River basin (KYB) is located in the transition region of three geomorphic types (109°28′–110°52′ E, 38°23′–39°52′ N)—the Maowusu sandy land, the Ordos platform, and the Loess hills. The main-stream length is 241.8 km, with an average slope of 2.58‰. Most of the water system is distributed in the form of branches. The basin has a continental climate and is located in the mid-temperate zone. Precipitation is concentrated from June to September, with heavy rains and an average annual temperature of 7.9 °C [25]. The elevation ranges from 740 to 1500 m.

The Tuwei River is a first-level tributary in the right bank of the middle reaches of the Yellow River. The Tuwei River basin (TWB) is located in the north of the Loess Plateau and the south of the Maowusu Desert (109°45′–110°35′ E, 38°10′–39°10′ N). The length of the main stream is 139.6 km, with an average slope of 3.16‰. The river system is asymmetric. The basin has a continental monsoon climate, with high temperatures in summer and cold and dry winters. Precipitation is concentrated from June to September, with an average annual temperature of 7.9 °C [26]. The elevation ranges from 1000 to 1200 m.

The Jingle hydrology station control basin (hereinafter referred to as the Jingle sub-basin; JLB) is located in the upper reaches of the Fenhe River, the second largest tributary of the Yellow River. In the Jingle sub-basin (38°17′–38°59′ N, 111°42‘–112°26′ E), the main stream length is 83.9 km, and the average slope is 6.7‰. The basin has an arid and semi-arid climate [27]. The average annual temperature of the watershed is 7.9° and the elevation ranges from 1140 to 2021 m.

Table 1. Characteristics of test watersheds.

Basin	Outlet Station	Long. (Degree)	Lat. (Degree)	Area (km2)	Precip. (mm)	Streamflow (mm)	Evap. (mm)	Potential Evap. (mm)
ZLB	Jingyuan	104.67	36.55	10,653	341	13	328	1507
KYB	Wejiachuan	110.75	39.48	8706	387	79	308	1300
TWB	Gaojiachuan	110.48	39.25	3294	422	116	306	1120
JLB	Jingle	111.92	38.35	2799	578	93	485	1267

shows information regarding the typical watersheds, and Figure 1 shows the location map of the typical watersheds.

2.2. Datasets

Hydrologic data were daily runoff data from outlet hydrological stations in typical watersheds, which were obtained from the Yellow River Conservancy Commission. The time ranges of the hydrologic data for the ZLB, KYB, TWB, and JLB were as follows: Period 1: 1957–1987, 1956–1989, 1956–1988, and 1956–1987, respectively, and Period 2: 2006–2014, 2006–2011, 2006–2011, and 2006–2014, respectively. Periodization was based on whether or not the watershed had changed significantly as a result of human activities.

Daily average temperature and daily precipitation data were obtained from the China Meteorological Data Network (http://data.cma.cn/; accessed on 20 March 2023). The time ranges of these parameters for the ZLB, KYB, TWB, and JLB were 1958–2014, 1959–2011, 1959–2011, and 1957–2014, respectively. For the spatial interpolation of meteorological data, the cooperative kriging method was adopted and the elevation factor was considered. Normalized difference vegetation index (NDVI) data from 1981 to 1999 used GIMMS-NDVI data, with a temporal resolution of 15 days and a spatial resolution of 8 km. MODIS-NDVI (MOD13A3) data from 2000 to 2014, with a temporal resolution of 16 days and a spatial resolution of 500 m, was also used. The NDVI was pretreated with geometric precision correction, radiation correction, and atmospheric correction, and the maximum synthesis method was used to reduce the influence of clouds and atmosphere. DEM data had a spatial resolution of 30 m, which was derived from the Geospatial Data Cloud platform (http://www.gscloud.cn; accessed on 25 March 2023).

2.3. Methods

2.3.1. Cooperative Kriging Interpolation

The cooperative kriging interpolation method was used to interpolate the precipitation and air temperature in the basin. The precipitation and air temperature in the basin were calculated by considering the elevation factor. The kriging interpolation method adopts the idea of spatial statistics and considers that the properties of any spatial continuity change are very irregular and cannot be simulated by simple smooth mathematical functions, but can be appropriately described by random surfaces. From the perspective of mathematical abstraction, it is an estimation method for optimal, linear, unbiased interpolation of spatially distributed data [28].

2.3.2. Baseflow Index

The BFI refers to the ratio of baseflow to the runoff in a certain period [29], which can reflect the size of baseflow in the period and reflect the conversion of surface runoff to underground streamflow. The BFI is affected by surface streamflow and baseflow. The calculation formula is as follows:

$$\mathit{BFI}=\frac{{\int }_{t_1}^{t_2}{Q}_{\mathit{Base}}\left(t\right)\mathit{dt}}{{\int }_{t_1}^{{\mathit{t}}_2}{Q}_{\mathit{Total}}\left(t\right)\mathit{dt}}$$

(1)

where $Q_{\mathit{Total}}\left(t\right)$ is the total river streamflow (m³) during the period of $t_1$−$t_2$. and $Q_{\mathit{Base}}\left(t\right)$ is the baseflow (m³) in the period of $t_1$−$t_2$.

2.3.3. Baseflow Separation

The recursive digital filtering method was proposed by Eckhardt in 2005 and contains two filtering parameters [30]. The segmentation equation is as follows:

$$Q_{\mathit{bt}}=\frac{1-{\mathit{BFI}}_{\mathit{max}}}{1-\alpha {\mathit{BFI}}_{\mathit{max}}}{\alpha Q}_{b(t-1)}+\frac{1-\alpha }{1-\alpha {\mathit{BFI}}_{\mathit{max}}}{\mathit{BFI}}_{\mathit{max}}{Q}_t$$

(2)

where $\alpha$ is the constant recession and ${BFI}_{max}$ is the maximum baseflow index. Using this method, Eckhardt recommended ${BFI}_{max}$ = 0.80 in a regularly flowing river dominated by a pore aquifer, ${BFI}_{max}$ = 0.50 in a seasonal river dominated by a pore aquifer, and ${BFI}_{max}$ = 0.25 in a regularly flowing river dominated by a weak pervious layer.

This study calculated the statistical characteristics of the annual BFI of the three parameter schemes of the Eckhardt baseflow separation method (

Table 2. Statistical characteristics of the annual BFI obtained by different parameter schemes.

Period		Period 1				Period 2
Statistics		M	CI	SD	CV	M	CI	SD	CV
ZLB	A	0.354	1.854	0.052	0.146	0.443	1.314	0.036	0.080
	B	0.302	2.402	0.060	0.198	0.434	1.448	0.047	0.108
	C	0.345	1.930	0.053	0.155	0.445	1.335	0.037	0.083
KYB	A	0.408	1.492	0.038	0.093	0.463	1.134	0.021	0.046
	B	0.365	1.677	0.047	0.130	0.426	1.247	0.034	0.079
	C	0.404	1.528	0.041	0.102	0.458	1.151	0.023	0.050
TWB	A	0.492	1.123	0.017	0.034	0.502	1.049	0.009	0.017
	B	0.492	1.224	0.022	0.044	0.501	1.114	0.020	0.041
	C	0.492	1.128	0.017	0.035	0.502	1.056	0.010	0.020
JLB	A	0.480	1.196	0.018	0.037	0.496	1.037	0.006	0.011
	B	0.461	1.253	0.024	0.053	0.483	1.082	0.015	0.030
	C	0.479	1.184	0.017	0.036	0.495	1.033	0.005	0.011

M is the average of the annual BFI; CI is the coefficient of instability; SD is the standard deviation; CV is the coefficient of variation.

). Scheme A: ${BFI}_{max}$ = 0.50, $\alpha$ = 0.925; Scheme B: ${BFI}_{max}$ = 0.50, $\alpha$ = 0.97; and Scheme C: ${BFI}_{max}$ = 0.50, $\alpha$ = 0.935. The extreme ratios and coefficients of variation for Scheme A in the Zuly River, Cave Creek, and Baldy River watersheds were less than those for Schemes B and C, while the extreme ratios and coefficients of variation for Scenario C in the Jingle watershed were less than those for Schemes A and B. Therefore, the optimized parameter schemes according to the results were as follows: Scheme A for the parameter setting of the ZLB, KYB, and TWB, and Scheme C for the parameter setting of the JLB.

2.3.4. NDVI Processing

The NDVI is the best indicator of plant growth status and vegetation spatial distribution density and is linearly correlated with vegetation distribution density [31]. The calculation formula is as follows:

$$NDVI=\frac{NIR-R}{NIR+R}$$

(3)

where NIR and R are the reflectance at the near-infrared band and red band, respectively.

The preprocessing of the NDVI included the conversion of the data format and coordinate system, the clip of test watershed data, the monthly maximum extraction, and the NDVI true value calculation. In order to facilitate the analysis of NDVI changes at different time scales (month and year), the maximum value composite method (MVC) was adopted in this study to synthesize NDVI data to obtain monthly and annual NDVI maximum data [32]. The formula is as follows:

$${NDVI}_i=MAX\left({NDVI}_{ij}\right)$$

(4)

where ${NDVI}_i$ refers to the NDVI maximum value of the ith month of a grid in a year and ${NDVI}_{ij}$ refers to the jth NDVI value of the ith month of the grid.

2.3.5. Correlation Analysis

Correlation analysis was used to study the relationship between driver factors and baseflow [33], and the correlation coefficient (r) was used to describe the degree of correlation among annual baseflow and annual precipitation, annual mean temperature, and the NDVI. The formula for calculating r is as follows [34]:

$$r=\frac{{\sum }_{i=1}^n(x_i-\overline{x})(y_i-\overline{y})}{\sqrt{{\sum }_{i=1}^n{(x_i-\overline{x})}^2{\sum }_{i=1}^n{(y_i-\overline{y})}^2}}$$

(5)

where $x_i$ and $y_i$ refer to the ith data value of the two groups of elements,$\overline{x}$ and $\overline{y}$ are the mean values of the two groups, and n is the sample size of a set of element data. The greater the absolute value of r, the closer the relationship between the two groups of elements and vice versa.

3. Results and Discussion

3.1. Variation Characteristics of Baseflow

Based on daily streamflow separation in typical watersheds, the variation characteristics of the BFI and baseflow in typical watersheds were analyzed on monthly and inter-annual chronological scales to assess the effect of driving factors of baseflow variation.

3.1.1. Intra-Annual Variation Characteristics of Baseflow and BFI

The monthly baseflow and BFI of test watersheds were calculated to obtain the intra-annual change, as shown in Figure 2. Compared with that in Period 1, in Period 2, the intra-annual variation in baseflow and the BFI in each month for typical watersheds was more steady, and the intra-annual distribution of baseflow was greater. The proportion of baseflow in winter and spring (from December to May of the following year) increased, while that in summer and autumn (from June to November) decreased. The changes in the mean monthly baseflow of each basin during the year from Period 1 to Period 2 were: ZLB decreased from 0.3 mm to 0.2 mm, KYB decreased from 2.6 mm to 0.6 mm, TWB decreased from 4.8 mm to 2.5 mm, and JLB increased from 3.5 mm m³ to 4.9 mm. The monthly baseflow in the KYB decreased sharply. The intra-annual change in baseflow between periods 1 and 2 was much larger for the KYB and TWB than for the ZLB and JLB. The KYB and TWB are the two basins influenced by coal mining. Besides, the intra-annual distribution in the KYB and TWB tended to be uniform, indicating that the variation in baseflow in the two basins may be greatly affected by human-impact factors.

3.1.2. Inter-Annual Variation Characteristics of Baseflow and BFI

The annual baseflow and BFI of test watersheds were calculated to obtain the inter-annual variation, as shown in Figure 3. Compared with that in Period 1, the annual baseflow variation in typical catchments in Period 2 was relatively stable, while the inter-annual variation of the BFI was different from one catchment to another. Comparing results from Period 2 to the average for Period 1, the average annual streamflow in Period 2 was lower than in Period 1 for the ZLB, KYB, and TWB, while in the JLB, streamflow increased in Period 2. The decline in streamflow in Period 2 was a continuation of the declining trend during Period 1. The decline in streamflow in the KYB in Period 2 was much larger than in other basins. The average annual baseflow showed the same pattern as annual streamflow with the sharpest decline in the KYB, with the JLB showing the only increase. The BFI increased between periods by 0.11 in the ZLB, by 0.05 in the KYB, and was essentially unchanged between Periods 1 and 2 for the TWB and JLB, as shown in

Table 3. Chronological changes of streamflow, baseflow, and BFI in typical watersheds.

Basin	Elements	Period 1					Period 2	Multi-Year Average
		1950s	1960s	1970s	1980s	Average
ZLB	Average annual streamflow/mm	18.21	12.86	15.02	9.29	12.86	5.82	11.36
	Average annual baseflow/mm	4.88	4.51	3.57	3.38	4.13	2.53	3.66
	Average annual BFI	0.28	0.36	0.35	0.38	0.34	0.45	0.37
KYB	Average annual streamflow/mm	89.48	84.65	83.05	59.15	79.14	14.70	66.16
	Average annual baseflow/mm	33.31	35.03	33.31	23.89	31.36	7.24	27.57
	Average annual BFI	0.39	0.43	0.41	0.40	0.41	0.46	0.42
TWB	Average annual streamflow/mm	125.08	131.15	116.27	91.68	115.97	61.02	105.04
	Average annual baseflow/mm	58.59	64.66	57.07	45.84	56.47	30.66	52.52
	Average annual BFI	0.47	0.49	0.49	0.50	0.49	0.50	0.49
JLB	Average annual streamflow/mm	126.47	103.97	83.96	55.38	92.53	118.61	97.53
	Average annual baseflow/mm	61.81	64.31	39.66	26.80	48.23	58.95	45.73
	Average annual BFI	0.49	0.48	0.47	0.48	0.48	0.50	0.48

. In Period 1, most of the annual streamflow in the basins showed a downward trend. In Period 2, the annual streamflow in the three basins, except the JLB, decreased, while that in the JLB reached the average level during the 1950s and the 1960s. The baseflow and streamflow showed the same trend of chronological variation. The annual mean BFI in Period 2 was greater than that over several years, which indicates that baseflow accounts for an increasing proportion of streamflow in the basins.

3.1.3. Characteristics of Baseflow and BFI

The chronological changes in streamflow, baseflow, and BFI in typical watersheds were analyzed, as shown in Table 3. In Period 1, most of the annual streamflow in the basins showed a downward trend. In Period 2, the annual streamflow in the three basins, except the JLB, decreased, while that in the JLB reached the average level during the 1950s and the 1960s. The baseflow and streamflow showed the same trend of chronological variation. The annual mean BFI in Period 2 was greater than that over several years, which indicates that baseflow accounted for an increasing proportion of streamflow in the basins.

The hydrological process in the Yellow River basin has a close relationship with nature and human activities under the background of global warming and intensification of human activities [35,36], and the importance of analysis and research of baseflow has increased with regard to the imbalance between water-resource supply and demand in the basin. The annual baseflow of the Zuli River, Kuye River, and Tuwei River showed a downward trend, with average annual baseflow of 3.66 mm, 27.57 mm, and 52.52 mm, respectively, and average annual BFI of 0.37, 0.42, and 0.49, respectively. Li and Zhang [37] reported that the baseflow index in the small watershed of the upper Yellow River was 0.69. They also reported that the national average BFI was 0.26, and those of the Northwest River and the Yellow River were 0.57 and 0.43, respectively. The annual baseflow of the Jingle sub-basin first showed a downward trend and then rose in Period 2, with an average annual baseflow of 45.73 mm and an average annual BFI of 0.48. The changes in streamflow and baseflow in the four basins in Periods 1 and 2 were analyzed, and it was found that the different rates of increase or decrease in the two resulted in the differences in the changes in the BFI in different basins. According to the related research, the change in baseflow was influenced by human activities (such as coal mining, soil and water conservation measures, and groundwater development) more than by natural factors such as precipitation, and the magnitude of these influencing factors determines the differences in the change in BFI index and baseflow in different basins. The BFI of each basin differed due to different watershed properties.

3.2. Analysis of Driving Factors of Baseflow Change

3.2.1. Effects of Climate Change on Baseflow

The influence of climate change on baseflow was analyzed from the two aspects of precipitation and air temperature; the variation characteristics of precipitation and air temperature in the basins were analyzed on monthly and inter-annual scales, and their correlation with annual baseflow was also analyzed.

• Intra-annual and inter-annual changes in precipitation and average temperature

Figure 4 shows the intra-annual changes in precipitation and the average temperature in the basins. Compared with that in Period 1, during Period 2, the average monthly temperature in the basins increased, the annual variation of precipitation was relatively small, and the annual distribution was relatively uniform, which corresponded to the annual variation characteristics of the monthly baseflow and BFI discussed in Section 3.1.1. By comparing the annual distribution of precipitation and baseflow, it can be observed that the reduction in baseflow was also affected by other factors, such as the annual distribution of baseflow in the KYB and TWB and did not show a unimodal pattern, as precipitation did. The average annual precipitation in the JLB increased from 438 mm in Period 1 to 463 mm in Period 2, which corresponds to the changes in baseflow discussed in Section 3.1.1 and Section 3.1.2.

The annual precipitation and average temperature of the basins were calculated to obtain the inter-annual variation, as shown in Figure 5. The baseflows of the basins were derived from precipitation; the changes in precipitation abundance and depletion affected the changes in baseflow [35]. As time progressed, the precipitation showed a downward trend (precipitation in the basins outside the Jingle sub-basin decreased by 1.3–1.4. mm/a and that in JLB decreased by 0.43 mm/a), and the fluctuation in inter-annual precipitation decreased. However, the temperature tended to rise (~0.03 °C/a).

II.

Correlation analysis of annual precipitation, annual mean temperature, and annual baseflow

Figure 6 shows the correlation between annual precipitation and annual baseflow during different periods in the basins. In Period 1, the Pearson correlation coefficients between annual precipitation and annual baseflow in the ZLB, KYB, TWB, and JLB were 0.87, 0.77, 0.52, and 0.79, respectively, and the coefficients were 0.34, 0.90, 0.21, and 0.51 in Period 2, respectively. In Period 2, except in the KYB, the correlation between annual precipitation and annual baseflow was significantly reduced, and the influence of precipitation on baseflow weakened. The influence of precipitation on baseflow strengthened in the KYB, owing to the significant reduction of baseflow, causing the baseflow to be more dependent on precipitation.

Figure 7 shows the correlation between annual mean temperature and annual baseflow in the basins during different periods. In Period 1, Pearson correlation coefficients between the annual average temperature and the baseflow in the ZLB, KYB, TWB, and JLB were −0.30, 0.02, 0.14, and −0.15, respectively, and the coefficients were 0.35, −0.08, 0.23, and −0.13 in Period 2, respectively. It is worth noting that baseflow was not correlated with temperature in Period 1, and although the correlation increased after 2006, it was still at a low level. Compared to the correlation coefficients between baseflow and precipitation (Figure 6), the correlation between baseflow and temperature was weaker.

3.2.2. Effects of Human-Impact Factors on Baseflow

In this study, the change in the NDVI was used to reflect the influence of changes in the underlying surface caused by human-impact factors on the baseflow [38]; the influences of soil and water conservation measures and coal mining on baseflow were also analyzed.

Intra-annual and inter-annual variation in the NDVI in typical watersheds can be obtained by NDVI processing, as shown in Figure 8. From 1981 to 2014, the NDVI in the basins increased significantly, and abrupt changes occurred from 2000 to 2002. After that, the growth rate of the NDVI became larger; hence, the vegetation cover of the underlying surface of the basins continued to improve. The annual NDVI of the JLB was greater than those of the other basins in the corresponding period, indicating that the vegetation coverage in the JLB was better during this period.

The correlation between the BFI and NDVI was used to reflect the impact of vegetation cover on baseflow in the basins (as shown in Figure 9). In Period 1, the correlation coefficients of the BFI and NDVI in the ZLB, KYB, TWB, and JLB were −0.498, −0.488, −0.653, and 0.528, respectively, and the correlation coefficients in Period 2 were 0.623, 0.818, 0.324, and 0.475, respectively. The correlation between the BFI and NDVI in the ZLB, KYB, and TWB changed from negative correlation to positive correlation, and the NDVI was around 0.375 during the transition. The correlation between the BFI and NDVI in the JLB weakened, and the NDVI was around 0.65 during the transition. With the increase in the NDVI, the increase rate of the BFI decreased when the NDVI was 0.65. Therefore, the changes in vegetation cover caused by human activities were an important factor affecting baseflow.

Table 4. Soil and water conservation measures in typical watersheds.

Basin	Year	Terrace (km2)	Forest (km2)	Grass (km2)	Silt Dams (km2)
ZLB	1969	67.6	45.1	31.6	0
	1989	810.8	672.4	470.7	50 (seat)
	2006	1817.1	1691.3	1183.9	218 (seat)
KYB	1969	32.9	97.3	51.5	2.4
	1989	67	1004.3	353.1	12.1
	2006	98.3	2638.7	1378.1	49.9
TWB	1969	10.8	77.1	6.1	1.7
	1989	45.5	754.5	28.8	11.1
	2006	82.1	779.3	331.9	24.1
JLB	The main soil and water conservation measures were the harnessing of the Fen River in the last century. The area of forest and grassland increased in abundance.

shows the soil and water conservation measures implemented in the basins. Before the 1970s, only a few soil and water conservation measures were implemented in the Loess Plateau basin. Since the late 1970s and the 1980s, comprehensive treatment of large-scale soil erosion such as planting grass and afforestation has been carried out; since 1999, a project aimed at returning farmland to forest and grassland has been ongoing in the basin [36,39]. On the one hand, soil and water conservation measures improve the soil environment, reduce soil and water loss, and reduce the sediment content in river channels to a certain extent; on the other hand, it intercepts precipitation infiltration to recharge groundwater, and increases baseflow [40]. The average BFI of the ZLB, KYB, TWB, and JLB in Period 1 was 0.35, 0.41, 0.49, and 0.48, respectively, while that of Period 2 was 0.45, 0.46, 0.50, and 0.50, respectively. Although the BFI exhibited a small increase, baseflow still showed a downward trend under the combined influence of groundwater development and utilization measures, water conservancy engineering, and other factors. It is worth mentioning that, although the NDVI has changed considerably since 2000, the BFI has changed much less, especially in the Tuwei and Jingle basins, where the BFI has remained essentially unchanged from the pre-2000 values. This is because changes in the NDVI have mainly altered the infiltration conditions at the ground surface, which reduces overland runoff and other soil factors affecting the baseflow rate changes, and the manifestation of such influencing effects will take a longer time period and these changes in the NDVI have increased evapotranspiration from soil-water storage, which reduces recharge to groundwater resulting in less baseflow. These reductions in stormflow and baseflow occurring together can keep the BFI constant despite the reduction in streamflow.

Groundwater exploitation reduced baseflow by reducing lateral groundwater excretion and baseflow. Groundwater exploitation in the Yellow River basin rose from 9.24 billion m³/a in 1980 to 13.31 billion m³/a in 2003—an increase of 31%.

Table 5. Coal-mining and water conservancy projects in typical watersheds.

Basin	Coal Mining	Water Conservancy Projects
KYB	In 1991, 60 million tons were mined. In 2011, 173 million tons were mined.	In the 1950s, the construction began. In 1988, there were 844 reservoirs and ponds.
TWB	The mine displacement in 1991 was 1.30 × 106 m3. In 2011 it was 3.42 × 107 m3.	By 2010, there were two medium-sized reservoirs and four diversion channels.
ZLB	None	Jinghui diversion irrigation project, the annual water volume was 9.7 million m3 in 1973 and 86.97 million m3 in 2005. There are 21 small reservoirs in upstream.
JLB	None	Almost none.

shows the statistics of coal mining and water conservancy projects in the basins. There was no coal mining activity in the ZLB or JLB; the coal mining in the KYB and TWB had a significant impact on baseflow [25,26]. The aquifer in the KYB and TWB was located above the coal seam. Coal mining dredged the shallow underground aquifer, which led to a continuous decrease in the groundwater level and in groundwater for the replenishment baseflow. Although soil and water conservation measures increased precipitation infiltration to recharge groundwater, they were not enough to replenish groundwater loss. The amount of water used by humans was small and there was no water conservancy project in the JLB; however, the water conservancy projects in the ZLB, KYB, and TWB had an influence on the inter-annual change in baseflow. Water conservancy projects aimed at meeting the industrial, agricultural, and urban domestic water requirements changed the water cycle in the underlying surface of the basin, affected the water balance of the basin, and led to changes in the spatial and temporal distribution of the river baseflow [41].

Precipitation is the main source of baseflow, and its spatial and temporal distribution and temperature directly affect the baseflow of rivers. The human-impact factors change the underlying surface conditions of the basin [33,39], leading to changes in the relationship between precipitation and streamflow, and the streamflow generation and confluence mechanism of the basin [42], thereby significantly impacting the baseflow. The main factors that influenced the change in baseflow in the ZLB were precipitation, soil and water conservation measures, and water conservancy projects. The implementation of soil and water conservation measures increased the forest and grassland area; the improvement in vegetation on the underlying surface was conducive to the increase in baseflow. However, under the combined effect of a decrease in precipitation and an increase in water extraction by water conservancy projects, baseflow in the basin decreased. (1) Precipitation and soil and water-conservation measures affected the spatial and temporal distribution of baseflow in the KYB; however, coal mining was the main reason for the large decline in baseflow (Figure 2). (2) Precipitation and soil and water conservation measures were also the main factors affecting the variation of baseflow in the TWB. The influence of coal mining and water conservancy projects increased, which led to the weakening of the influence of precipitation on baseflow. The intra-annual distribution of baseflow was related to the regulation of reservoirs (Figure 2). (3) Precipitation and soil and water conservation measures were the main factors affecting the change in baseflow in the JLB, while other human-impact factors had little influence. The inter-annual increase in baseflow (Figure 2 and Figure 3) was related to the increase in precipitation during Period 2 and soil and water conservation measures.

The change in basin baseflow was the result of climate change and human activities [33,37]. The influence of precipitation and temperature on baseflow weakened; the increasing influence of human-impact factors significantly changed the underlying surface of the basin; thus, the influence of human activities on baseflow increased. The change in the NDVI was used to reflect the influence of the underlying surface change on the baseflow caused by human-impact factors. The BFI of watersheds gradually decreased with an increase in the NDVI until NDVI < 0.375. The improvement in vegetation coverage on the underlying surface reduced the BFI. When the NDVI was between 0.375 and 0.65, the BFI of the basins gradually increased with the increase in the NDVI. Therefore, with the continuous increase in vegetation coverage on the underlying surface of the basin, the streamflow was reduced. However, as the BFI increased, the proportion of baseflow in the streamflow increased and the soil and water conservation capacity of the underlying surface was enhanced. When the NDVI was greater than 0.65, the rate of increase in the BFI in the watersheds decreased. Therefore, when the vegetation coverage of the underlying surface of the watershed increased to a certain extent, the proportion of baseflow in the streamflow increased slowly, and the soil and water conservation capacity of the underlying surface tended to be stable.

4. Conclusions

The Zuli River basin, Kuye River basin, Tuwei River basin, and Jingle sub-basin were selected as the study areas. The Eckhardt filtering method was used for the baseflow separation of the flow process in the study areas. The intra-annual, inter-annual, and chronology characteristics of baseflow and the BFI were analyzed. The driving factors of baseflow change were analyzed from the perspective of climate-change and human-impact factors. The main conclusions are as follows:

(1)

The baseflow primarily followed a downward trend for most parts of the basins, and the proportion of baseflow in streamflow increased. The intra-annual variations of baseflow and the BFI were steady and the distribution was more uniform; the inter-annual fluctuation tended to be stable. Compared with the decreasing streamflow trend in most rivers, the increase in the BFI indicated that baseflow is becoming increasingly important for the development and utilization of water resources in the Yellow River basin.

(2)

The baseflow was affected by climate change and human activities. The influence of precipitation weakened while the influence of soil and water conservation measures increased; coal mining had a great impact in the Kuye River basin and Tuwei River basin. Under the comprehensive action of precipitation change, coal mining, water conservancy measures, and other human-impact factors, baseflow showed a downward trend.

(3)

When NDVI < 0.375, the baseflow index was negatively correlated with the NDVI; when 0.375 < NDVI < 0.65, the baseflow index was positively correlated with the NDVI; when NDVI > 0.6, the correlation between the baseflow index and the NDVI weakened. The vegetation coverage of the underlying surface of the basin increased to a certain extent; however, the baseflow index did not increase dramatically, and the soil and water conservation capacity of the underlying surface of the basin tended to be stable.

The innovative aspect of our research was the selection of different parameterization schemes for different watersheds and the confirmation that the evolution of baseflow in the experimental watersheds was the result of the combined effects of climate change and human activities.