Variation in Runoff, Suspended Sediment Load, and Their Inter-Relationships in Response to Climate Change and Anthropogenic Activities Over the Last 60 Years: A Case Study of the Upper Fenhe River Basin, China

The variation of river runoff (Q) and suspended sediment load (S), in addition to their influential factors, is an important area of focus in hydrological sciences. Here, Mann–Kendall tests and double mass curve analyses were used with hydrometric data from four hydrological stations in the upper Fenhe River Basin in China to evaluate temporal trends in annual Q (Qa) and S (Sa) values between 1955 and 2015. Based on the observed inflection points, three distinct periods were identified, namely, 1955–early 1980s (period I), the early 1980s–1996 (period II), and 1996–2015 (period III). The Qa and Sa values for the four stations, except for Qa values for the Jingle station, significantly decreased over the study period, with average reduction rates of 0.68–1.07 mm·km−2·a−1 and 9.24–54.39 t·km−2·a−1, respectively. Decreased rainfall, implementation of soil and water conservation program, and reservoir construction were primarily responsible for decreased Qa and Sa values for the three stations during period II, while the first two factors led to variation in Qa and Sa for the Jingle station during the same period. During period III, the Qa, Sa, and Qa–Sa relationships for the four stations were intensively affected by increased anthropogenic activities, including water diversion, cross-basin water transfers, soil and water conservation measures, revegetation efforts, and sand excavation. Further, the Qa, Sa, and Qa–Sa relationships at the Lancun station were affected by the construction of the No. 2 Fenhe Reservoir. Effective water use and supply strategies should be implemented in the future for the upper Fenhe River Basin.


Introduction
Climate change and increased anthropogenic activities in recent decades have introduced altered dynamics in river systems, including frequent occurrences of discontinuous flow [1], sharp decreases in suspended sediment loads (S) [2], increasingly strained water supplies [3], and deteriorated river ecologies [4]. Understanding these dynamics and their underlying factors have become areas of intense focus in hydrological research. Indeed, investigations of the variations in runoff (Q) and

Study Area
The upper Fenhe River is about 217.3 km in length, with the Lancun hydrological station at the terminus. The river basin covers about 7705 km 2 within the Ningwu, Jingle, Lanxian, Loufan, and Yangqu counties, in addition to Gujiao city and the Jiancaoping District of Taiyuan city, which is the capital city of the Shanxi Province [29]. The area is characterized by a temperate continental monsoon climate. The annual precipitation varies from 213 to 798 mm, more than 80% of which falls between June and September. The annual average temperature is 8.5-11.8 • C. The major topography of the basin is loess hilly-gully, rocky mountains, and riparian zone. Particularly, the loess hilly-gully area, which forms a severe erosion area in the basin with an average erosion modulus of 6700 ton·km −2 ·a −1 , contributes about 70% of the erosion in the basin with less than 40% of the area [30]. The zonal vegetation is characterized as a temperate sub-humid deciduous broad-leaved forest [31]. The dominant plant species of the area include Quercus wutaishansea Mary, Pinus tabuliformis, Hippophae rhamnoides, and others. Agricultural cultivation was previously the primary source of economic income in the basin, where croplands accounted for more than 50% of the basin area at one point [32]. Following the implementation of China's reform and opening-up policy in 1978, coal mining industries began to dominate economies in Ningwu county, Lanxian county, Gujiao city, and Taiyuan city, with some croplands and forestlands being converted to construction lands. Construction lands substantially expanded after 2000 with the rapid development of the economy and urbanization across the whole basin [33]. Concomitantly, vegetation coverage increased in the basin due to the implementation of the national level "Soil and Water Conservation" program for 20 years, beginning in 1988. Further, the initiation of the "Natural Forest Protection" and "Grain for Green" programs in the late 1990s and 2003, respectively, particularly promoted the restoration of native environments [33,34].
Hydrological stations and rainfall gauge stations were established by the Chinese Water Conservancy Department on the main stream of the upper Fenhe River and its major tributaries (e.g., the Lanhe River, which is the largest tributary of the upper Fenhe River) in the 1950s to monitor variations in hydrological characteristics. The No. 1 and No. 2 Fenhe reservoirs are present on the main stream of the upper Fenhe River and were completed in 1961 and 2000, respectively. The reservoirs have capacities of 7.12 × 10 8 and 1.33 × 10 8 m 3 , respectively, and are used for flood protection, water supplies, and electricity. In addition, the Lancheng and Hamashen reservoirs were built on the Lanhe River tributary in 1973 and 1975, respectively, and feature storage capacities of 6.0 × 10 6 and 6.24 × 10 6 m 3 , respectively [29]. Furthermore, a large-scale, cross-basin water transfer project, the Yellow River Water Transfer Project, was initiated in October 2003, primarily to alleviate intensified water scarcities in Taiyuan city, and exhibits an annual water transfer capacity of 640 million m 3 . The project transfers water from the Wanjiazhai Reservoir of the Yellow River via tunnels, aqueducts, pipelines, and a section of the upper Fenhe River channel to the No. 1 Fenhe Reservoir. The waters then supplement the Taiyuan city water supplies via underground pipelines from the reservoir [35].

Data and Methods
Annual Q (Q a ) and S (S a ) data were collected from four hydrological stations in the upper Fenhe River Basin, representing the longest available time series  to investigate temporal variation in Q a and S a values over those 60 years. The four hydrological stations included three stations on the main stream of the upper Fenhe River (Jingle, Zhaishang, and Lancun; referred to as JL, ZS, and LC, respectively) and one station (Shangjingyou, SJY) at the terminus of the Lanhe River.
Annual precipitation (P a ) data from the four hydrological stations, in addition to six other rainfall gauge stations (Dongzhai, Ninghuabao, Suopo, Puming, Miyu, and Chakou), from 1955 to 2015 were collected to explore the Q a and S a values for JL, SJY, ZS, and LC in response to changes in the average P a values (P ma ) for individual catchments. All of the above data were collected from the Yellow River Water Conservancy Commission. The P ma values for the JL, SJY, ZS, and LC catchments (abbreviated as P maJL for the JL catchment, for example) were calculated using the arithmetic mean of P a for stations and rainfall gauges that were located within the catchments.
The annual average air temperature data for the Taiyuan meteorological station (T TY ) from 1955 to 2015 were only collected due to the lack of such long-term time-series data at other meteorological stations. The T TY data were collected from the Shanxi Meteorological Information Center to explore variations in Q a and S a values at the nearby LC station in response to air temperature changes. Variations of P ma for JL, SJY, ZS, and LC catchments and T TY during the different periods are provided in Table 1. Additional details for the above stations are provided in Figure 1. The annual average air temperature data for the Taiyuan meteorological station (TTY) from 1955 to 2015 were only collected due to the lack of such long-term time-series data at other meteorological stations. The TTY data were collected from the Shanxi Meteorological Information Center to explore variations in Qa and Sa values at the nearby LC station in response to air temperature changes. Variations of Pma for JL, SJY, ZS, and LC catchments and TTY during the different periods are provided in Table 1. Additional details for the above stations are provided in Figure 1.   Monthly normalized difference vegetation index (NDVI) datasets, including the Global Inventory Modeling and Mapping Studies (GIMMS) NDVI3g dataset and the Terra Moderate Resolution Imaging Spectroradiometer (MODIS, NASA, Washington DC, USA) Vegetation Indices (MOD13Q1) dataset, were collected to analyze the relationships between vegetation coverage and Q a or S a values over the study period. The GIMMS NDVI3g dataset was derived from the NOAA/AVHRR land dataset at a spatial resolution of 8 km in 15-day intervals from 1982 to 2015. The MODIS MOD13Q1 dataset exhibited a 250 m spatial resolution at 16-day intervals from 2000 to 2015, and was derived from the Earth Observing System of NASA [36]. Monthly NDVI values for the two datasets were first obtained using the maximum value composite method, which minimizes cloud contamination, atmospheric effects, and solar zenith angle effects [37]. The yearly NDVI value of each pixel was then calculated by computing the maximum values for the 12 monthly NDVI values. Finally, the annual regional values of GIMMS NDVI (NDVI 8km ) and MODIS NDVI (NDVI 250m ) for the JL, SJY, ZS, and LC catchments were separately calculated by computing the average yearly NDVI values for all grid cells within those catchments. Additionally, the annual floor space data for buildings under construction in Taiyuan city (FSBC TY ) were collected for the years from 1996 to 2015 from the Shanxi Province Statistical Yearbook [38] to better understand the use of sand excavation and explore the associated impacts on S a .
Autocorrelation was evaluated for the calculated metrics before analyzing the monotonic trends of the Q a and S a time series for the JL, SJY, ZS, and LC stations, in addition to their respective P ma , NDVI 8km , and NDVI 250m values for the four respective catchments. Furthermore, autocorrelation was evaluated for the T TY , Q aFR , and D aFR metrics [39]. The monotonic trends of the above variables were determined using nonparametric Mann-Kendall tests [40,41]. The magnitudes of the trends for the respective variables were indicated using Sen's slopes [42]. The inflection points of the variables were identified based on the separate intersection of U F and U B curves from the Mann-Kendall tests within the confidence interval of [−0.01, 0.01] [43,44]. In addition, abrupt inflection points for Q a and S a values for the four stations were detected from the variation in slope for straight lines that were fitted to double mass curves of Q a -P ma and S a -P ma [45]. The abrupt inflection points obtained using the double mass curve method were used as nodes to divide periods (see Section 3.3) and further investigate changes in Q a , S a , and the Q a -S a relationships, in addition to their potential influential factors. The Q a -S a relationships within different periods were analyzed using simple linear regression analyses. Linear regression models were tested using F tests to evaluate the statistical differences in relationships. Correlations between Q a or S a and influential indicators (i.e., P ma , NDVI 8km , NDVI 250m , T TY , Q aFR , and D aFR ) were analyzed using Pearson correlation analysis.

Temporal Variation in Q a and S a Based on Mann-Kendall Tests
The temporal variation in Q a and S a values for the JL, SJY, ZS, and LC stations from 1955 to 2015 is shown in Figure 2. The Q a variables, excluding Q aJL , exhibited highly significant decreasing trends over the 60 years of the study, with average reduction rates of 0.68-1.07 mm·km −2 ·a −1 (p < 0.001) ( Table 2). S a values for the above four stations also exhibited significant decreasing trends over the same period, with average reduction rates of 9.24-54.39 t·km −2 ·a −1 (p < 0.001 or p < 0.01). The most significant decreases in the Q a and S a rates were observed at LC and SJY, respectively. The inflection points for Q a were identified in 1983 for SJY, 1980 for ZS, and 1981 for LC, while the inflection points for S a occurred in 2004 for JL, 1997 for SJY, 1991 for ZS, and 1994 for LC ( Figure 3).  Z: Z value from the Mann−Kendall tests; β: Sen's slope represents the average change rate; ns: no significant change; ** significant at p < 0.01; *** significant at p < 0.001.

Temporal Variation of Qa−Pma and Sa−Pma Based on Double Mass Curves
Cumulative Qa and Sa values were plotted against cumulative Pma values for the four stations ( Figure 4). Except for the double mass curve of QaJL−PmaJL, which manifested as a straight line without segmentation, the other double mass curves for SaJL−PmaJL, QaSJY−PmaSJY, and SaSJY−PmaSJY unanimously trended rightwards with abrupt inflection points occurring in 1982 and 1996. The double mass curves of Qa−Pma and Sa−Pma for ZS and LC, respectively, varied nearly synchronously, with the same abrupt inflection points occurring in 1980 and 1996, respectively, in the cumulative Qa−Pma curves

Temporal Variation of Q a -P ma and S a -P ma Based on Double Mass Curves
Cumulative Q a and S a values were plotted against cumulative P ma values for the four stations ( Figure 4). Except for the double mass curve of Q aJL −P maJL , which manifested as a straight line without segmentation, the other double mass curves for S aJL −P maJL , Q aSJY −P maSJY , and S aSJY −P maSJY unanimously trended rightwards with abrupt inflection points occurring in 1982 and 1996. The double mass curves of Q a −P ma and S a −P ma for ZS and LC, respectively, varied nearly synchronously, with the same abrupt inflection points occurring in 1980 and 1996, respectively, in the cumulative Q a −P ma curves and for the cumulative S a −P ma curves in 1959, 1980, and 1996.  Dotted red arrows denote inflection point years during the study period using double mass curves.

Changes in Average Q a , S a , and the Q a -S a Relationships during Three Separate Study Periods
Three periods were used to investigate changes in the Q a , S a , and the Q a -S a relationships, with 1982 and 1996 as the nodes of Q aSJY and S aSJY , and 1980 and 1996 as the nodes of Q a and S a for ZS and LC. To facilitate comparisons among the four stations, the Q aJL and S aJL series were divided into three periods, with 1982 and 1996 as the nodes. Specifically, the three periods for Q a and S a values of JL and SJY were defined as 1955-1982 (period I), 1983-1996 (period II), and 1997-2015 (period III). Likewise, the three periods for Q a and S a values for ZS and LC were defined as 1955-1980 (period I), 1981-1996 (period II), and 1997-2015 (period III). The reductions in average Q a for the four stations during periods II and III compared to period I were 21.77-49.46% and 7.25-70.35%, respectively ( Table 3). The level of reduction in the average S a was greater than that of Q a for the four stations, with average S a values decreasing in periods II and III by 45.03-76.62% and 88.08-99.04%, respectively, when compared to period I.
The Q a -S a relationships for JL, SJY, ZS, and LC during the three periods are shown in Figure 5. The simple linear regression lines, equations, and determination coefficients (R 2 ) for all relationships are presented simultaneously for ease of interpretation, although some were not statistically significant. The Q a -S a relationships for the four stations during periods I and II fit linear functions well, as indicated by the significant R 2 values (p < 0.01). All of the regression lines for period II were below those of period I, indicating that the value of S a decreased given the same values of Q a under such conditions. The slope of the Q aJL -S aJL relationship regression line was negative during period III, contrasting with the positive slopes observed during the first two periods. Extremely low R 2 values observed for SJY, ZS, and LC regressions during period III indicated the lack of significant correlations between Q a and S a values for the three stations during this period.

Temporal Variation in Potential Influential Variables
Trend-based analyses were conducted for representative environmental indicators in the study area, including Pma, NDVI8km, and NDVI250m metrics for the four catchments, in addition to the other environmental metrics TTY, QaFR, and DaFR. Significant changes were not observed for PmaJL, PmaSJY, PmaZS, and PmaLC from 1955 to 2015 (Table 4)

Temporal Variation in Potential Influential Variables
Trend-based analyses were conducted for representative environmental indicators in the study area, including P ma , NDVI 8km , and NDVI 250m metrics for the four catchments, in addition to the other environmental metrics T TY , Q aFR , and D aFR . Significant changes were not observed for P maJL , P maSJY , P maZS , and P maLC from 1955 to 2015 (Table 4). However, average P maJL , P maSJY , P maZS , and P maLC values for period II (484.  Z: Z value from the Mann−Kendall tests; β: Sen's slope represents the average change rate; ns: no significant change; * significant at p < 0.05; ** significant at p < 0.01; *** significant at p < 0.001. (a) (b)

Pearson's Correlations between Qa, Sa, and Potential Influencing Variables
The Pearson's correlation coefficients among Qa, Sa, and the potential influential factors for the JL, SJY, ZS, and LC stations were investigated for different periods (Table 5). Significant and positive correlations were observed between Qa and Sa values for the four stations over the first two periods (p < 0.01 or p < 0.05). In contrast, significant negative correlations for QaJL−SaJL (p < 0.05) and no significant correlations in the Qa−Sa relationships for the SJY, ZS, and LC stations were observed during period III. Significant and positive correlations were observed for the Qa−Pma values of JL during the three periods (p < 0.01 or p < 0.05) and for SJY during the first two periods (p < 0.01). In  Z: Z value from the Mann−Kendall tests; β: Sen's slope represents the average change rate; ns: no significant change; * significant at p < 0.05; ** significant at p < 0.01; *** significant at p < 0.001.

Pearson's Correlations between Qa, Sa, and Potential Influencing Variables
The Pearson's correlation coefficients among Qa, Sa, and the potential influential factors for the JL, SJY, ZS, and LC stations were investigated for different periods (Table 5). Significant and positive correlations were observed between Qa and Sa values for the four stations over the first two periods (p < 0.01 or p < 0.05). In contrast, significant negative correlations for QaJL−SaJL (p < 0.05) and no significant correlations in the Qa−Sa relationships for the SJY, ZS, and LC stations were observed during period III. Significant and positive correlations were observed for the Qa−Pma values of JL during the three periods (p < 0.01 or p < 0.05) and for SJY during the first two periods (p < 0.01). In

Pearson's Correlations between Q a , S a , and Potential Influencing Variables
The Pearson's correlation coefficients among Q a , S a , and the potential influential factors for the JL, SJY, ZS, and LC stations were investigated for different periods (Table 5). Significant and positive correlations were observed between Q a and S a values for the four stations over the first two periods (p < 0.01 or p < 0.05). In contrast, significant negative correlations for Q aJL -S aJL (p < 0.05) and no significant correlations in the Q a -S a relationships for the SJY, ZS, and LC stations were observed during period III. Significant and positive correlations were observed for the Q a −P ma values of JL during the three periods (p < 0.01 or p < 0.05) and for SJY during the first two periods (p < 0.01). In contrast, significant correlations were not observed for Q aZS −P maZS and Q aLC -P maLC during the three periods. Significant positive correlations were observed for the S a −P ma values of the four stations during the first two periods (p < 0.01 or p < 0.05), except for S aZS −P maZS during period I. Strongly positive and significant correlations were observed between Q a for the four stations and Q aFR during the three periods (p < 0.01 or p < 0.05). Likewise, significant positive correlations were also observed for S a −D aFR values of the four stations during the first two periods (p < 0.01). Significant positive correlations were observed for Q a −NDVI 8km for the JL and ZS stations during period II (p < 0.05). Likewise, significant positive correlations were observed for JL, SJY, and ZS during period III (p < 0.01 or p < 0.05), in addition to similar relationships for Q a −NDVI 250m for all four stations during period III (p < 0.01). In contrast, significant negative correlations were observed for S a −NDVI 8km values of JL and ZS, and for S a −NDVI 250m of LC over the same period (p < 0.05).

Discussion
The Mann-Kendall tests indicated that the Q a values for the SJY, ZS, and LC stations dramatically decreased, beginning in the early 1980s (Table 2 and Figure 3). This was further evinced from obvious rightward shifts of the double mass curves for Q a −P ma for the three stations around 1980 (Figure 4). A similar result was obtained by Zhang et al. [44] for the Q a series of the Fenhe River from 1957 to 2010. Average P maSJY values during period II were lower than those of period I (Figure 6a), and significant positive correlations were observed for Q aSJY −P maSJY during the two periods (p < 0.01) ( Table 5), indicating that the decreased P maSJY values during period II could lead to reduced Q aSJY values. Similarly, significant correlations in the S aSJY −P maSJY and Q aSJY −S aSJY relationships (p < 0.01) indicated that decreased P maSJY and Q aSJY during period II could have weakened sediment generation and carrying capacity, further reducing S aSJY . In addition to rainfall, other factors affected Q aSJY and S aSJY during period II, as indicated by the obvious rightward deflection of the cumulative Q aSJY −P maSJY and S aSJY −P maSJY curves around 1982. Qin [34] reported that the large-scale "Soil and Water Conservation" program was carried out in the upper Fenhe River Basin after 1988. Niu and Niu [46] further confirmed that 39 new check dams were built in the SJY catchment between 1983 and 1994, 33 of which were built after 1988, while only two reservoirs (the Lancheng and Hamashen reservoirs) and three check dams were built during period I. These check dams and reservoirs cumulatively controlled 50% of the drainage area in the loess hilly-gully areas of the SJY catchment, thereby effectively intercepting runoff and sediment, and dramatically reducing Q aSJY and S aSJY during period II. The reduced rainfall and increased soil and water conservation measures were also responsible for the decreased Q aJL and S aJL values observed for the same period. The Q aJL and S aJL values had comparatively lower decreases than Q aSJY and S aSJY during period II, respectively, when compared with the values for period I (Table 3). These differences may be related to different topographies among the two catchments. The SJY catchment is dominated by loess hilly-gully features that are more prone to soil erosion and more affected by soil and water conservation measures. In addition, the construction of two reservoirs in the SJY catchment could have also contributed to these observed differences.
Minimal P ma metric values were also observed for the ZS and LC catchments in the 1980s (Table 1 and Figure 6a), when the "Soil and Water Conservation" program was conducted in the two catchments after 1988. However, in contrast to the significant correlations in the Q aSJY −P maSJY and Q aJL −P maJL relationships observed during periods I and II (p < 0.01), significant correlations in the Q aZS −P maZS and Q aLC −P maLC relationships were not observed during the two periods. Likewise, significant correlations were not observed for the S aZS −P maZS relationships during period I. These observations could be attributed to the regulation of runoff and sediment-intercepting effects of the No. 1 Fenhe Reservoir that is upstream of ZS. This supposition is supported by significant correlations in the Q aZS −Q aFR and Q aLC −Q aFR relationships during the three periods (p < 0.01), as well as the synchronous appearance of the first inflection points for the Q a data series of ZS, LC, and FR in 1980, and that of in the cumulative S aZS −P maZS and S aLC −P maLC curves in 1959 due to construction of the reservoir. The former occurred later than the latter, which was mainly due to differential effects of the reservoir on the regulation of gradual flows and complete sediment deposition. The reservoir mainly supported farmland irrigation but has continually supplied water for industrial activities in Gujiao city within the ZS catchment and the Taiyuan city downstream of the LC station since the 1980s [47]. Five large underground coal mines and associated coal preparation plants were successively built in Gujiao city during period II. By the end of this period, the annual coal output reached 16.5 million tons and the total industrial output value of the city increased 36.3 times over that of 1978 when China's reform and opening-up policy was implemented [48]. Taiyuan city is an important industrial city in northern China and harbors industries with high water consumption (e.g., metallurgical, coal-fired, coking, and chemical industries) that accounted for 68.81% of the city's industrial output value in the early 1990s [49]. Dramatic decreases in Q aZS and Q aLC over period II were the result of industrial development in the two cities and the reservoir filling, which led to the increased water diversion from rivers, coupled with reduced rainfall, reduced incoming flows upstream, and the implementation of soil and water conservation measures.
Moreover, extensive goafs from Gujiao coal mining activities could have intensified the infiltration of flows in leakage areas, thereby resulting in decreased runoff [50], as was observed by Cravotta et al. [51] in the upper Schuylkill River of Pennsylvania. These factors also contributed to decreased S aZS and S aLC values, as indicated by the second abrupt inflection of cumulative S aZS −P maZS and S aLC −P maLC curves in 1980 and the significant correlations in the Q a -S a relationships for the two stations over this period (p < 0.05 or p < 0.01).
The abrupt inflection points of cumulative Q a −P ma curves for the SJY, ZS, and LC stations and those of the cumulative S a −P ma curves for the four stations coincidently occurred in 1996. Previous hydrological records [52] indicated that catastrophic flood events occurred in the upper Fenhe River Basin in 1996 due to an extreme precipitation event when rainfall lasted for 34 hours and 314 mm of accumulated rainfall was measured at the rainfall center. A persistent five-year drought subsequently occurred during 1997-2001, with an average annual P ma value for the five years of 280 mm, representing 78% of the average P ma for 1955-2015. These observations indicate that high magnitude flood events and climatic fluctuations are strongly coupled to fluxes in river discharge and sediment deposition [53]. However, significant correlations were not observed for the Q a −P ma relationships for the three stations during this period, except for Q aJL −P maJL of JL, nor for S a −P ma at the four stations, or for Q aLC −T TY and S aLC −T TY . These results could be primarily attributed to complex human activities that weaken the roles of climate change on the above values. As discussed earlier, 20 successive years of the "Soil and Water Conservation" program took place in the basin. By the end of 2007, 30341 hm 2 of terraced fields, 131933 hm 2 of afforestation and grass, and 118 check dams had been developed in the upper reaches of the No. 1 Fenhe Reservoir [34]. These data provide some evidence for the effects of soil and water conservation measures on reduced Q a and S a values for JL and SJY during period III. Further evidence of this is given by the dramatic reduction in D aFR during this period (Figure 7a). The afforestation effects of the "Soil and Water Conservation," "Natural Forest Protection," and "Grain for Green" programs were confirmed by the significantly increased NDVI 8km and NDVI 250m values for the four catchments (p < 0.05 or p < 0.01 or p < 0.001). Moreover, the effects of sediment interception were represented by negative correlations for the S aJL −NDVI 8kmJL , S aZS −NDVI 8kmZS , and S aLC −NDVI 250mLC relationships during this period (p < 0.05).
Booming industrial development and associated economies, along with urbanization, led to a continuous increase in water demand, thereby resulting in increased water diversion from rivers and dramatic decreases in runoff. The direct or indirect discharge of industrial wastewater and domestic sewage into rivers also led to the deterioration of water quality [47]. Consequently, severe water scarcity in industrial, agricultural, and domestic water occurred in Taiyuan city in the 1990s [54]. Concomitantly, over-exploitation of groundwater intensified, with the area of the deep groundwater depression cone near the lower reaches of LC station expanding by 44.8 km 2 from 1990 to 2000 and central water levels dropping by 8.83 m [55]. As a result, the Yellow River Water Transfer Project was initiated in the basin, beginning in 2003, to primarily alleviate water scarcity in Taiyuan city. From 2003 to 2010, a total of 523 million m 3 of Yellow River water had been supplied to Taiyuan city, and the annual water supply reached 320 million m 3 by 2015 [56]. In addition, rapid urbanization development resulted in the development of a real estate industry. This was especially evident in the rapid development after 1998 upon implementation of China's urban housing system reform, as indicated by the increased FSBC TY during 1996−2015 (Figure 8). The result caused an increased demand for river sand, thereby contributing to increased sand excavation along the river due to the lack of any policy forbidding sand excavation during that time, and decreased S a values at the four stations. The above-mentioned factors profoundly influenced the spatiotemporal dynamics of rainfall-driven runoff generation and sediment delivery in the upper Fenhe River Basin, resulting in fundamentally changed Q a -S a relationships at the four stations in period III ( Figure 5). Similar results were reported by Xu [45] for the coarser sediment production area in the Middle Yellow River Basin. The comparably low reduction rate of Q aJL in period III compared with that in period I (when compared to the other three stations) was due to the transferred Yellow River water that largely supplemented Q aJL when passing through the river course upstream of the JL station. Conversely, the comparatively higher reduction of S aLC in period III than S aZS , when compared with values for period I, was caused by the construction of the No. 2 Fenhe Reservoir, which retained most of the sediments downstream of ZS after its commission in 2000.

Summary and Conclusions
In this study, temporal variation in Qa and Sa values, as well as the Qa−Sa relationships, were (2) Decreased rainfall, soil and water conservation measures, and the construction of reservoirs were the primary factors underlying the decreased Qa and Sa values for SJY, ZS, and LC during period II. The first two processes were responsible for the variation in QaJL and SaJL during period II. Following 1997, dramatically decreased Qa and Sa values, in addition to fundamental changes in the Qa−Sa relationships for the four stations, were primarily caused by increased anthropogenic activities, including water diversion, a cross-basin water transfer project, soil and water conservation measures, revegetation efforts, and sand excavation. In addition, construction of the No. 2 Fenhe Reservoir was also responsible for variation in QaLC, SaLC, and the QaLC−SaLC relationship. In total, the influence of anthropogenic activities on Q and S was greater than that of climate change over the last 60 years in the upper Fenhe River Basin.
(3) The above results suggest that effective strategies should be stressed, including promoting the efficiency of water use, strengthening water resource management, etc., in order to adjust increasingly severe differences in water supplies and demands, as well as changed flow-sediment relationship in the upper Fenhe River Basin. In addition, the long-term impact of underground coal mining and a large-scale, cross-basin water transfer project on the environments of the upper Fenhe River should be an area of further concern and investigation.

Summary and Conclusions
In this study, temporal variation in Q a and S a values, as well as the Q a -S a relationships, were investigated for four stations (JL, SJY, ZS, and LC) in the upper Fenhe River Basin during the period encompassing 1955-2015. The primary results showed that: (1) Q a values for SJY, ZS, and LC dramatically decreased with average reduction rates of 0.68−1.07 mm·km −2 ·a −1 , and inflection points occurring around 1980, as shown using Mann-Kendall tests. The S a values at the four stations also significantly decreased, with average reduction rates of 9.24−54.39 t·km −2 ·a −1 . The inflection points of S a occurred in 2004 for JL, 1997 for SJY, 1991 for ZS, and 1994 for LC. (2) Decreased rainfall, soil and water conservation measures, and the construction of reservoirs were the primary factors underlying the decreased Q a and S a values for SJY, ZS, and LC during period II. The first two processes were responsible for the variation in Q aJL and S aJL during period II. Following 1997, dramatically decreased Q a and S a values, in addition to fundamental changes in the Q a -S a relationships for the four stations, were primarily caused by increased anthropogenic activities, including water diversion, a cross-basin water transfer project, soil and water conservation measures, revegetation efforts, and sand excavation. In addition, construction of the No. 2 Fenhe Reservoir was also responsible for variation in Q aLC , S aLC , and the Q aLC −S aLC relationship. In total, the influence of anthropogenic activities on Q and S was greater than that of climate change over the last 60 years in the upper Fenhe River Basin. (3) The above results suggest that effective strategies should be stressed, including promoting the efficiency of water use, strengthening water resource management, etc., in order to adjust increasingly severe differences in water supplies and demands, as well as changed flow-sediment relationship in the upper Fenhe River Basin. In addition, the long-term impact of underground coal mining and a large-scale, cross-basin water transfer project on the environments of the upper Fenhe River should be an area of further concern and investigation.