![Free access]({"type":"image" , "src":"/pb-assets/3rdPartyLogos/FREE_ACCESS_ICON-1465051266437.png"}){kind=icon}
![CrossMark Logo]({"type":"image" , "src":"/templates/jsp/images/CROSSMARK_Color_horizontal.svg"})
ABSTRACT
?
![MathJax Logo]({"type":"image","src":"/templates/jsp/_style2/_tandf/pb2/images/math-jax.gif"})
?ABSTRACT
Headwaters contribute a substantial part of the flow in river networks. However, spatial variations of streamflow generation processes in steep headwaters have not been well studied. In this study, we examined the spatio-temporal variation of streamflow generation processes in a steep 2.98-ha headwater catchment. The time when baseflow of the upstream section exceeded that downstream was coincident with the time when the riparian groundwater switched from downwelling to upwelling. This suggests that upwelling of the riparian groundwater increased considerably in the upstream section during the wet period, producing a shift in the relative size of baseflow between the upstream and downstream sections. The timing of fluctuations among hillslope soil moisture, hillslope groundwater and streamflow reveals that the hillslope contributed to storm flow, but this contribution was limited to the wet period. Overall, these results suggest that streamflow generation has strong spatial variations, even in small, steep headwater catchments.
EDITOR A. Castellarin ASSOCIATE EDITOR X. Chen
1 Introduction
Headwater streams make substantial contributions to the water yield of mid to large size rivers. Alexander et al. (2007) reviewed catchment studies of streams in the northeastern United States and found that first-order streams output approximately 70% of the second-order annual water yield. This contribution of first-order streams dropped only marginally in second- to fourth-order streams. Headwater can also influence downstream water quality, especially during the baseflow period (Uchida et al. 2003, Alexander et al. 2007, Morley et al. 2011). Hence, clarification of the streamflow generation mechanism in headwater catchments is of great importance to water resource management.
Streamflow generation mechanisms have been studied mainly at the hillslope and catchment scales. At hillslope scale, streamflow generation is controlled by various factors such as subsurface structure, surface topography, soil and vegetation. Subsurface flow generated by subsurface structures including bedrock fissures and hollows, which provide preferential pathways, influences storm flow (Anderson and Burt 1978, Freer et al. 2002). Surface topography such as hillslope steepness in mountainous catchments can affect streamflow (Harr 1977, McDonnell 1990). On steep hillslopes, storm flow is generated quickly by rapid flow over the steep slopes (Anderson et al. 1997). Soils dominated by preferential flow paths can control the timing and transfer of mobile water during rainfall events (Weiler and Naef 2003, Weiler and McDonnell 2007). Vegetation types influence precipitation input patterns and soil moisture (Gomi et al. 2010, Burke and Kasahara 2011, Bachmair et al. 2012, Jost et al. 2012). Wainwright et al. (2000) found that compared to grassland, shrubland generated more overland flow, which is responsible for higher overall erosion rates.
As studies were extended to the catchment scale, new factors became dominant, such as large structural elements, landscape organization, shifts of geology, and stream characteristics, which had stronger influences than factors at hillslope scale. Large structural elements such as size of the contributing area and storage were used to predict storm-flow generation (Beven and Kirkby 1979). Landscape organization such as sub-catchment size was correlated with mean residence time of baseflow (McGlynn et al. 2003). The shift of geology from sandstone to granite-gneiss in the midstream reach increased deeper flow paths and discharge water downstream (Payn et al. 2012). Stream characteristics (e.g. river incision, drainage density and hydraulic conductivity) were also found to improve prediction of streamflow generation (Bloomfield et al. 2011).
Bedrock-dominated runoff generation processes that have been identified recently at hillslope scale have yet to be incorporated into catchment-scale hydrological models (McDonnell 2003). This is because catchment-scale studies commonly ignore spatial variability within a watershed (Uchida et al. 2005). Thus, there is a gap in understanding of streamflow generation processes between hillslope and catchment scales. This gap prevents further understanding of the spatial variation of streamflow generation (Payn et al. 2012). Recent studies have drawn attention to closing this gap (Jencso et al. 2009). Spatial patterns in stream discharge along valleys have revealed spatial differences in streamflow generation, which may help link hillslope-scale and catchment-scale processes (Payn et al. 2012). However, there have been few studies focused on seasonality of water yield patterns and related streamflow generation mechanisms (Beven 2006, Payn et al. 2012, Penna et al. 2015). Tracer injection techniques have shown patterns of spatial discharge variability that can be used to understand streamflow generation from hillslope to catchment scale. However, this method consists of one-time only measurement. Accordingly, development of methods for continuous monitoring of spatial variations in streamflow can provide further insight (Botter et al. 2008, Barthold et al. 2010, Payn et al. 2012).
To better understand the relationship between the spatial distribution of water yield and the streamflow generation mechanism, we investigated seasonal streamflow patterns in a small, steep headwater catchment in western Japan. We monitored stream discharge at three locations, the groundwater table and hillslope soil moisture to elucidate the spatial variation of the streamflow generation mechanism. The specific objectives of the study were to (1) identify seasonal patterns in spatial variation of water yield, and (2) explain spatial differences in the streamflow generation mechanism. The results provide useful information for linking hillslope and catchment hydrology in headwater catchments and provide insight into spatio-temporal differences in streamflow generation processes within small, steep headwater catchments.
2 Site description
The study site (Yayama Experimental Catchment, YEC) is a 2.98-ha headwater catchment in Fukuoka Prefecture on Japan’s Kyushu Island. The site is at 33°30ʹN and 130°39ʹE (Fig. 1), and the elevation ranges from 305 to 406 m a.s.l. Mean annual precipitation from 1981 to 2000 in this region is 2098 mm (±387 mm) based on data from Uchino meteorological station (33°32ʹN, 130°38ʹE; 80 m a.s.l.), the nearest meteorological station, which is maintained by the Ministry of Land, Infrastructure, Transport and Tourism. The entire catchment is composed of steep hillslopes and a narrow valley floor. The mean hillslope gradient is 0.81 m/m and mean stream gradient, 0.37 m/m. The valley topography shows that the longitudinal gradient steepens toward the downstream portion of the catchment (Fig. 1). Within the study reach, the substrate changes from sandy in the upstream reach to a bedrock and boulder bed with steep channel gradient in the downstream portion. The channel becomes incised in the section from the midstream to the downstream gauge. From transects measured at each section, the depth of incision is 0.17 m upstream, 0.83 m midstream, and 1.69 m downstream. The geology of the YEC is weathered granite. The thickness of the weathered granite is 13.7 m in the riparian area and 17.5 m on the hillslope, according to a drilling company field survey. Four distinct soil layers within the catchment are classified in the Dixfield-Marlow-Brayton general soil association. Japanese cypress (Chamaecyparis obtusa Sieb. et Zucc.) and Japanese cedar (Cryptomeria japonica D. Don), planted in 1969, cover the catchment. The cypress comprises 67% of all trees, and the cedar accounts for the remaining 33%.
Published online:
20 December 2016Figure 1. Map of the study site in the Yayama Experimental Catchment (YEC): Gup, Gmid and Gdown are the gauges; Hr3.0, Hr20.0 and Hh17.5 are the groundwater wells. Dashed lines are for sub-catchment division. Contour interval is 4 m.
![]({"type":"image","src":"/na101/home/literatum/publisher/tandf/journals/content/thsj20/2017/thsj20.v062.i05/02626667.2016.1266635/20190522/images/medium/thsj_a_1266635_f0001_b.gif"})
Figure 1. Map of the study site in the Yayama Experimental Catchment (YEC): Gup, Gmid and Gdown are the gauges; Hr3.0, Hr20.0 and Hh17.5 are the groundwater wells. Dashed lines are for sub-catchment division. Contour interval is 4 m.
3 Methods
This study was carried out from January to December 2011. It was designed to monitor streamflow at multiple locations in the catchment and to explain spatio-temporal water yield variations using precipitation, groundwater elevation, soil moisture and soil water potential measured on the hillslope and in the riparian zone.
Three streamgauge stations were installed. The upstream gauge (Gup) was immediately downstream of where the stream starts during the dry season. The midstream gauge (Gmid) was above where the stream gradient becomes steep, and the downstream gauge (Gdown) was at the catchment outlet (Fig. 1). The difference in elevation from Gup to Gmid was 21.8 m with a gradient of 0.25 m/m, and 32.9 m with a gradient of 0.55 m/m from Gmid to Gdown. Each station consisted of a V-notch gauge and a Parshall flume. The V-notch gauge was used to monitor baseflow and the Parshall flume, the storm flow. The stage was monitored at 10-min intervals by the V-notch and 5-min intervals by the Parshall flume, using a WT-HR water level logger (TruTrack, Christchurch, New Zealand). Stage sensor readings were checked weekly with visual stage readings from June to July (wet period) and twice per month during the rest of the year to corroborate continuous measurements. Water yield among the three gauge sites was also compared.
Precipitation was recorded at the weather station located 320 m from the centre of the catchment, at an elevation of 390 m. A 0.5-mm tipping bucket raingauge (TK-1; Takeda Keiki, Tokyo) was used, and the data were collected at an interval of 10 min. Long-term precipitation data were acquired from the nearby Uchino meteorological station.
Groundwater level was monitored on the hillslope and in the riparian zone. A hillslope well (Hh17.5) was located upslope and installed to a depth of 17.5 m from the surface (Fig. 1). Two wells of different depths were installed at the same location in the riparian zone to observe the vertical head gradient (VHG). A deep riparian well (Hr20.0) was installed to a depth of 20 m, with a screen present from 4 to 20 m. A shallow riparian well (Hr3.0) was installed to a depth of 3 m, with a screen present from 1 to 3 m. Elevation and horizontal distance between the ground locations of Hh17.5 and Hr20 were 15.2 and 36.6 m, respectively. Water-level fluctuations in each well were recorded with a Hobo U20 water-level data logger (Onset Company, Bourne, MA, USA) at 10-min intervals. Manual measurements of groundwater levels in each well were conducted twice per month to check sensor readings. The VHG between Hr20.0 and Hr3.0 was calculated as: 
(1)
(1)
where Δh1 is the head difference between Hr20.0 and Hr3.0, and l1 is the horizontal distance between them.
The lateral head gradient (LHG) between Hh17.5 and Hr3.0 was calculated as: 
(2)
(2)
where Δh2 is the head difference between Hh17.5 and Hr3.0, and l2 is the horizontal distance between them.
Soil moisture was continuously monitored at three locations in the same hillslope area. The sampling locations were 5 m apart and 14.8 m upslope from Gup (Fig. 1). Each sampling station contained three soil-moisture sensors (EC-5; Decagon Devices Inc., WA, USA) and three tensiometers (DIK-3042; Daiki Rika Kogyo Co., Ltd, Japan) at three depths (10, 30 and 50 cm), which collected data at 1-h intervals. The soil moisture was calibrated using the soil samples excavated from three soil pits on the hillslope where the soil moisture plot was located. The antecedent soil moisture index (ASI) (Haga et al. 2005) was calculated for each storm event as initial storage of the surface soil layer in the catchment, based on the volumetric water content at the measurement points: 
(3)
(3)
where θ is the volumetric soil water content at a depth of 0.1 m (m3/m3) and D is the installation depth (0.1 m).
To identify differences in medians between pairs of water yield data in the same period, the Mann-Whitney U test was used (Iman and Conover 1983). The Kruskal-Wallis test was used to determine any significant overall differences among the three groups of water yield data in the same period (Iman and Conover 1983). This test has the advantage of not requiring normality or equal variances of data.
4 Results and discussion
4.1 Temporal and spatial water yield pattern
The 2011 annual precipitation measured in the YEC was 2469 mm. The mean annual precipitation from 1981 to 2011 at the nearby Uchino weather station was 2098 mm (±387 mm), whereas it was 2632 and 2397 mm in 2010 and 2011, respectively. These data show that the study year and year prior were relatively wet years in the region. Precipitation falls occasionally as snow in January and February, and then melts in early February.
The hydrograph of water yield showed various patterns among the three gauging stations (Fig. 2(a)). We divided the year into three periods based on the hyeto-hydrograph (Fig. 2(a)). The dry period was January to late May when the water yield was low and relatively stable. The wet period was from the first major event recorded in late May to the end of the rainy season in mid-July, when the water yield continuously increased. The dry-down period was from mid-July to December, when the water yield slowly declined.
Published online:
20 December 2016Figure 2. (a) Hydrograph, (b) groundwater and (c) soil moisture for 2011. Vertical dashed lines are for season division. Zero point on the y-scale in (b) represents the ground elevation of the riparian well.
![]({"type":"image","src":"/na101/home/literatum/publisher/tandf/journals/content/thsj20/2017/thsj20.v062.i05/02626667.2016.1266635/20190522/images/medium/thsj_a_1266635_f0002_b.gif"})
Figure 2. (a) Hydrograph, (b) groundwater and (c) soil moisture for 2011. Vertical dashed lines are for season division. Zero point on the y-scale in (b) represents the ground elevation of the riparian well.
During the dry period, the water yield ranked in the order of Gdown, Gmid and Gup. However, values were similar for all three stations, at <0.1 mm/h (Table 1). During the wet period, the water yield increased at each gauge site, and differences of water yield among the three gauges widened. Additionally, the water yield at Gmid did not increase as much as at the other gauges, with 0.185 mm/h on average and the smallest standard deviation (Table 1). Large differences in water yield between Gmid and the other gauges persisted for the remainder of the year. Water yield at Gup showed a slower increase than Gdown at the beginning of the wet period but surpassed that of Gdown in July. During the dry-down period, the difference in water yield between these gauges decreased, but Gup maintained larger values than Gdown to December (Table 1 and Fig. 2(a)). The water yields in the same period from different gauges were significantly different (Kruskal-Wallis test, p = 0.000 for all groups). Also, all combinations of two groups of water yield in the same period had significantly different medians (Mann-Whitney test).
Table 1. Statistical summary of water yield at each gauge in each period (sum, average, standard deviation, interquartile range and maximum-minimum).
The low water yield at Gmid during the wet and dry-down periods may be attributed to the topography at the gauge site. Specifically, Gmid was immediately above where the longitudinal gradient steepened from 0.25 to 0.55 m/m (Fig. 1), and the substrate was sandy gravel with boulders, which is probably highly permeable. Studies of channel morphology and stream–groundwater exchange have revealed downwelling trends upstream of steps (Wondzell et al. 2009), and stronger downwelling trends were associated with greater step size (Kasahara and Wondzell 2003). The slope steepened immediately downstream of Gmid, resulting in a head drop of 23.3 m. This may have had an effect similar to a large step, driving downwelling flow upstream of Gmid. The highly permeable substrate and greater water yield may have accelerated downwelling upstream of Gmid during the wet period, increasing the differences in water yield between Gmid and the other gauges.
The location of Gup was immediately below where the stream started, and Gdown was where bedrock exposure was present. Water yield at the two gauges showed similar annual patterns, but their relative size switched during the wet period. Peak flow during precipitation events was consistently greater at Gdown, but baseflow between precipitation events increased at Gup toward the end of the wet period (Fig. 2(a)).
Several studies have examined spatial differences of water yield in headwater catchments (Jencso et al. 2009, Jencso and McGlynn 2011). For example, Payn et al. (2012) reported a greater downstream water yield during the summer recession period in the Tenderfoot Creek Experimental Forest catchment (USA). In the YEC, the spatial difference in water yield changed seasonally.
4.2 Groundwater contribution during baseflow
Hillslope groundwater showed a steady decline from January, reaching the lowest level at the beginning of May. Hillslope groundwater level rose as the rainy season began, and peaked in July. After the peak, the level gradually declined over the rest of the year, except for a slight increase in response to two typhoon rain events (Fig. 2(b)). The hillslope groundwater table remained within the weathered bedrock layer. Riparian groundwater levels were stable from January to May and began to increase after a rain event on 9 May which produced 226 mm of precipitation. The groundwater levels increased with each large rain event, peaking in mid-July. During the rainy season, the stream expanded upstream to the area where Hr20.0 and Hr3.0 were located, and riparian groundwater levels measured at these wells were above the valley floor surface from 23 June to 31 July. Those levels declined gradually over the remainder of the year, except for a slight increase in response to four typhoon rain events (Fig. 2(b)). During the wet period, the riparian groundwater table could rise above the weathered bedrock layer into the soil.
When groundwater levels on the hillslope and in the riparian zone were compared, the hillslope groundwater level was always higher than the riparian groundwater level, except for a short period in May (Fig. 2(b)). The range of hillslope and riparian groundwater fluctuations also differed, with the shallow riparian groundwater level showing only a 3.2-m range of fluctuation and the hillslope groundwater level a 9.1-m range.
Figure 3 presents LHG values between Hh17.5 and Hr3.0. Positive values indicate that the LHG is from the hillslope to the riparian zone, and negative values indicate the reverse. LHG was positive most of the year, except for 1 day in May. The top of the screening for wells Hr20.0 and Hr3.0 was 4 m and 1 m below ground, respectively, and the groundwater level in both wells was above the screen for most of the year. Thus, we also calculated the VHG between the two riparian wells (Fig. 3). Positive and negative values indicate upwelling and downwelling, respectively. The VHG showed a downwelling trend with relatively stable values from January to May, with fluctuations at the beginning of the wet period. Later in that period, the VHG had a steady upwelling trend, which continued to the end of the year (Fig. 3(a)).
Published online:
20 December 2016Figure 3. (a) Vertical hydraulic gradient (VHG) between Hr20.0 and Hr3.0 (positive values mean riparian groundwater upwelling, and negative values mean riparian groundwater downwelling); and lateral hydraulic gradient (LHG) between Hh17.5 and Hr3.0 (positive value means hillslope groundwater contributing to riparian groundwater, and negative values mean riparian groundwater contributing to hillslope groundwater). (b) Water yield difference between Gup and Gdown (positive values mean Gup > Gdown, negative values mean Gup < Gdown).
![]({"type":"image","src":"/na101/home/literatum/publisher/tandf/journals/content/thsj20/2017/thsj20.v062.i05/02626667.2016.1266635/20190522/images/medium/thsj_a_1266635_f0003_b.gif"})
Figure 3. (a) Vertical hydraulic gradient (VHG) between Hr20.0 and Hr3.0 (positive values mean riparian groundwater upwelling, and negative values mean riparian groundwater downwelling); and lateral hydraulic gradient (LHG) between Hh17.5 and Hr3.0 (positive value means hillslope groundwater contributing to riparian groundwater, and negative values mean riparian groundwater contributing to hillslope groundwater). (b) Water yield difference between Gup and Gdown (positive values mean Gup > Gdown, negative values mean Gup < Gdown).
The LHG was always greater than the VHG, except for an event during May. The catchment has steep hillslopes with a narrow valley floor, and the LHG may represent the hillslope groundwater contribution to streamflow. However, this value was only calculated between two points. Compared with the VHG, the consistently greater LHG may indicate a larger contribution of lateral inflow to the stream, but the spatial variability of hydraulic conductivity in the area is not known. Some studies in zero-order catchments have also reported that lateral inflow dominated streamflow generation (Sidle et al. 2000, Frisbee et al. 2007). Additionally, studies in small granite catchments similar to the YEC have reported that subsurface storm flow through the soil profile can make a dominant contribution to streamflow (Onda et al. 2001, 2006). The wells used to calculate LHG and VHG in our study were in the upstream portion of the catchment (Fig. 1), so the strong lateral inflow may only be applicable to that area. However, considering that the downstream portion of the stream was incised and flowing on exposed bedrock and boulders, a larger contribution of lateral inflow than vertical upwelling may also apply to the downstream portion of the catchment.
The relative size of LHG to VHG does not explain the greater water yield at Gup during the wet and dry-down periods. The VHG values showed that the riparian groundwater changed from downwelling to upwelling, which overlapped when the water yield at Gup became larger than that at Gdown (Fig. 3(b)). During the baseflow period, when VHG indicated downwelling, the water yield was Gup < Gdown, whereas it was Gup > Gdown when VHG indicated upwelling. During events at the beginning of wet period, VHG showed downwelling, and the water yield difference was negative. During events at the end of the wet period and the entire dry-down period, VHG indicated upwelling, while the water yield difference still had negative values. These findings suggest that, in addition to lateral inflow, vertical upwelling in the riparian zone of the zero-order basin supplied water to the upstream reach that was sufficient to switch the water yield balance between the two gauges.
Fluctuations of groundwater level have been linked to stream water yield in other studies. For instance, studies in the Sierra Nevada mountains in California, USA, revealed that fast storm-flow response and extended recession flow were produced by fluctuations in groundwater levels that created saturated areas on hillslopes (McNamara et al. 1998, Yamazaki et al. 2006). Although we could not quantify the contributions of lateral and vertical groundwater inflow to the stream, our results show that fluctuations in groundwater levels partially explain water yield patterns in the YEC.
4.3 Groundwater and subsurface flow contribution during storm flow
Hillslope soil moisture has been used as an indicator of hillslope–stream connectivity and throughflow (Burke and Kasahara 2011, Moore et al. 2011, Penna et al. 2011, Fu et al. 2013). In the present study, we compared hillslope soil moisture to streamflow and hillslope groundwater level to explain water movements during storm flow in the YEC.
The relationship between soil moisture and water yield was examined. The results of a typical event in dry and wet periods are plotted in Figure 4. For the 20–22 March event (dry period), water yield increased and peaked before the hillslope soil moisture, suggesting that the hillslope contribution to storm flow was small (Fig. 4(a), left). Water yield plotted versus soil moisture shows clockwise hysteresis (Fig. 4(b), left). For the event of 30 June–1 July (wet period), water yield reacted more slowly than soil moisture, and peaked after the hillslope soil moisture peak, indicating a potential hillslope contribution to storm flow (Fig. 4(a), right). For this event, the plot of water yield versus soil moisture shows counter-clockwise hysteresis (Fig. 4(b), right).
Published online:
20 December 2016Figure 4. (a) Hydrograph and soil moisture and hillslope groundwater levels during two events in 2011; (b) 10-cm soil moisture plotted with upstream water yield for two events; (c) 10cm soil moisture plotted with hillslope groundwater level for two events. Black dots represent the rising limb; lighter dots represent the falling limb.
![]({"type":"image","src":"/na101/home/literatum/publisher/tandf/journals/content/thsj20/2017/thsj20.v062.i05/02626667.2016.1266635/20190522/images/medium/thsj_a_1266635_f0004_oc.jpg"})
Figure 4. (a) Hydrograph and soil moisture and hillslope groundwater levels during two events in 2011; (b) 10-cm soil moisture plotted with upstream water yield for two events; (c) 10cm soil moisture plotted with hillslope groundwater level for two events. Black dots represent the rising limb; lighter dots represent the falling limb.
All precipitation events with hysteresis of soil moisture and water yield are summarized in Figure 5. The ASI was used together with precipitation. The hysteresis relationship between hillslope soil moisture and water yield showed seasonality. Most events during the dry period had a clockwise hysteresis relationship (i.e. no hillslope contribution), whereas most events during the wet period showed counter-clockwise hysteresis (potential hillslope contribution). During the dry-down period, clockwise hysteresis again became the dominant pattern, except for typhoon events in August which resulted in a clockwise hysteresis relationship. This indicates that subsurface flow can contribute to the stream during typhoons. These results agree with findings of hillslope contributions during rain events in the rainy season only (Onda et al. 2006, Penna et al. 2015). Penna et al. (2015) found that hillslopes delivered water to the stream during events in the rainy season. Onda et al. (2006) reported that shallow subsurface flow could be a major contributor to the stream in a steep granite catchment during the rainy season. A similar change of hillslope–streamflow hysteresis patterns was detected by McGuire and McDonnell (2010), who found hysteresis patterns as a result of increasing wetness conditions.
Published online:
20 December 2016Figure 5. (a) Temporal evolution of rainfall + ASI for the study year. Closed circles represent clockwise hysteresis relationship between soil moisture and water yield; open circles represent counter-clockwise hysteresis relationship between soil moisture and water yield. (b) Soil water head gradient. Vertical dashed lines are for season division.
![]({"type":"image","src":"/na101/home/literatum/publisher/tandf/journals/content/thsj20/2017/thsj20.v062.i05/02626667.2016.1266635/20190522/images/medium/thsj_a_1266635_f0005_b.gif"})
Figure 5. (a) Temporal evolution of rainfall + ASI for the study year. Closed circles represent clockwise hysteresis relationship between soil moisture and water yield; open circles represent counter-clockwise hysteresis relationship between soil moisture and water yield. (b) Soil water head gradient. Vertical dashed lines are for season division.
We also examined the relationship between soil moisture and hillslope groundwater level during storm flow (Fig. 4). Specifically, the relationship between average soil moisture at 10 cm and hillslope groundwater level was compared to address recharge and discharge of hillslope groundwater. We defined the discharge condition as hillslope groundwater level decreasing continuously despite rainfall events. The recharge condition was defined as groundwater level increasing with rainfall events. In mountainous catchments, groundwater in the riparian zone tends to have a different response time from the hillslope groundwater (Seibert et al. 2003). This difference can lead to hysteresis behaviour between hillslope groundwater and runoff (Penna et al. 2010, 2011).
Measurement of water potential in the hillslope soil profile showed that water moved downward for almost the entire study period (Fig. 2(d)), suggesting that the water infiltrated into deeper layers of the soil profile in the YEC. Results from two typical events that were also used for analysis of soil moisture and water yield are plotted in Figure 4(a). The event on 20 March had a continuous decrease of hillslope groundwater level during and after the event, suggesting that hillslope groundwater maintained a discharge trend because the precipitation amount was too small to influence the groundwater level (Fig. 4(a), left). Conversely, the event on 30 June had an increasing hillslope groundwater level with decreasing soil moisture, indicating recharge of hillslope groundwater level (Fig. 4(c), right).
Recharge and discharge during the entire study period are shown in Figure 6. During the dry period, hillslope groundwater showed no response or a slower response and peaked after soil moisture, leading to a clockwise hysteresis relationship between soil moisture and hillslope groundwater. Water was retained in the hillslope, and a disconnected hillslope may not allow water to percolate deep into the hillslope groundwater, causing little or no hillslope groundwater contribution. Conversely, during events in the wet period, hillslope groundwater peaked prior to soil moisture or even continued to increase after the event, producing a counter-clockwise hysteresis relationship. During these events, a state of connection was assumed to be established within the hillslope, and water could percolate quickly into the groundwater. This caused a rapid response of hillslope groundwater, so the hillslope began to release water.
Published online:
20 December 2016Figure 6. Temporal evolution of rainfall + ASI for the study year. Closed circles represent hillslope groundwater discharge conditions, open circles represent hillslope groundwater recharge conditions. Vertical dashed lines are for season division.
![]({"type":"image","src":"/na101/home/literatum/publisher/tandf/journals/content/thsj20/2017/thsj20.v062.i05/02626667.2016.1266635/20190522/images/medium/thsj_a_1266635_f0006_b.gif"})
Figure 6. Temporal evolution of rainfall + ASI for the study year. Closed circles represent hillslope groundwater discharge conditions, open circles represent hillslope groundwater recharge conditions. Vertical dashed lines are for season division.
Seasonal patterns of soil moisture to water yield and soil moisture to groundwater were similar (Figs 5 and 6), indicating hillslope groundwater recharge and subsurface flow during storm flow in the wet period. However, there was a short time difference. At the beginning of the wet period, hillslope groundwater was recharging while the hillslope was disconnected from the groundwater. During the typhoon season, the hillslope groundwater was discharging while hillslope subsurface flow was present.
4.4 Streamflow generation mechanism in the YEC
The studied catchment was small but water yield varied spatially within it. Therefore, relative size of the water yield varied seasonally. In this section, we summarize the results of streamflow, groundwater and soil moisture to elucidate streamflow generation mechanisms in the YEC.
During the dry period, water yield was low, had little variability, and increased slightly downstream. Riparian groundwater was likely a major source of streamflow, and lateral inflow from the riparian zone to the stream channel dominated at all three gauge sites. Precipitation was weak and less frequent, and the response of soil moisture to storm flow was slower than that of streamflow. Precipitation did not translate to an increase in groundwater level, suggesting disconnection of the hillslope from the stream and groundwater.
During the wet period, groundwater level and soil moisture were increased by a large amount of precipitation. In addition to lateral inflow, strong upwelling in the riparian zone of the zero-order basin increased water yield at Gup. The water yield at Gmid did not show as large an increase as at Gup and Gdown, which may be attributable to the sandy permeable substrate and longitudinal gradient in the midstream reach. The upstream water yield exceeded that downstream during the wet period. During storm flow, the response of soil moisture preceded that of streamflow, and this translated to an increase in groundwater level. This suggests that the hillslope was connected to the stream and that subsurface flow may appear during and after the rain event.
During the dry-down period, upstream water yield efficiency remained higher than that downstream, although the water yield declined throughout the catchment. The soil moisture and the groundwater level decreased gradually with a strong response to heavy rainfall events. Upstream riparian groundwater continued upwelling, and subsurface flow was generated during heavy rain from a typhoon.
A lack of downstream soil moisture and groundwater data prevented exploration of more detailed spatial variation of streamflow generation. To generalize the present results, multi-catchment comparison is needed, and the relative influences of geology and topography should be clarified (Jencso and McGlynn 2011, Richardson et al. 2012). However, our results provide valuable insight into the linkage between the seasonal and spatial water yield patterns and spatio-temporal differences in streamflow generation.
4.5 Implications
Our study highlights the importance of groundwater movement in streamflow generation. We emphasize the relationship between water yield patterns and streamflow generation. One typical study of streamflow generation used a lumped model called a “tank model”, which treats the catchment as vertically arranged multiple tanks with several outlets (Sugawara 1961, Sidle et al. 2011). Difficulties often arose when trying to extrapolate their results to larger catchment size.
The relationship between water yield pattern and streamflow generation is likely to be transmitted to other headwater catchments, since headwater catchments typically evolve from a gentle zero-order basin to a relatively steep valley associated with steep, incised channel morphology (Benda et al. 2005). Early studies used multi-gauge set-ups in small catchments to study the storm-flow generation process, transferred to further downstream areas (Ragan 1968, Dunne and Black 1970a, 1970b). However, seasonal variation of water yield was not clarified. In gentle catchments, there would be more groundwater inflow in areas where the regional groundwater table contributed to the stream. It remains in question whether gentle catchments show similar water yield patterns, because the streamflow generation processes may be different.
5 Conclusions
Spatio-temporal variations of streamflow generation in a small, steep headwater catchment were examined. Water yield in the upstream section was less than in the downstream section during the dry and wet periods, but this order changed in the subsequent dry-down period. Baseflow in the upstream section surpassed that downstream at the end of the wet period, which coincided with the time when riparian groundwater switched from downwelling to upwelling. This suggests that upwelling of the riparian groundwater considerably increased the upstream baseflow during the wet period, inducing a shift in relative amount of baseflow between the upstream and downstream sections. Storm flow was consistently greater in the downstream section. However, the contribution source changed seasonally. Hillslope groundwater discharge and hillslope subsurface flow greatly contributed to storm flow, but these contributions only appeared during the wet period. Overall, these results suggest that streamflow generation has a strong spatial variation even in small, steep headwater catchments, which may indicate linkage between hillslope scale and catchment scale hydrological processes. Additional studies using more detailed soil moisture and groundwater data are needed.
| Down | Mid | Up | ||
|---|---|---|---|---|
| Sum (mm) | Dry | 189.1 | 47.0 | 85.1 |
| Wet | 546.9 | 226.9 | 450.2 | |
| Dry-down | 874.3 | 189.6 | 967.7 | |
| Average (mm/h) | Dry | 0.056 | 0.014 | 0.025 |
| Wet | 0.447 | 0.185 | 0.368 | |
| Dry-down | 0.212 | 0.096 | 0.234 | |
| Standard deviation | Dry | 0.057 | 0.019 | 0.016 |
| Wet | 0.244 | 0.105 | 0.227 | |
| Dry-down | 0.107 | 0.068 | 0.116 | |
| Median (mm/h) | Dry | 0.047 | 0.010 | 0.022 |
| Wet | 0.477 | 0.196 | 0.430 | |
| Dry-down | 0.176 | 0.078 | 0.215 | |
| Interquartile range (mm/h) | Dry | 0.024 | 0.018 | 0.021 |
| Wet | 0.368 | 0.179 | 0.377 | |
| Dry-down | 0.101 | 0.090 | 0.122 | |
| Maximum-minimum (mm/h) | Dry | 1.342 | 0.414 | 0.253 |
| Wet | 1.985 | 0.633 | 1.548 | |
| Dry-down | 1.504 | 0.595 | 0.799 |
Acknowledgements
We thank Makiko Tateishi, Sodouangdenh Somsanouk, Xiang Yang, Hiroki Matsuda, Shuji Yunohara and Misako Komatsu at Kyushu University for their assistance with field data collection.
Disclosure statement
No potential conflict of interest was reported by the authors.