Shift of estuarine type in altered estuaries

To better understand the alteration of the estuarine circulation caused by estuarine dams, four major Korean estuaries were classified by using the Hansen and Rattray stratification–circulation classification scheme. The stratification and circulation parameters were calculated for both discharge and no-discharge periods from the tidally averaged salinity and velocity data obtained from these four Korean estuaries. The estuarine types of the altered Korean estuaries were compared with the previous results for natural estuaries in other countries of similar magnitude in tidal range, water depth, and discharge. This comparison revealed that the estuarine types of the altered Korean estuaries have been shifted from a partially mixed to a well-mixed type (Nakdong River Estuary), from a partially mixed to a coastal bay or a fjord type (Yeongsan River Estuary), and from a well-mixed to a well-mixed type with less tidal modulation (Geum River Estuary). The controlling factors that determined the type in natural estuaries were tide, discharge, and water depth, whereas for altered estuaries, they were the controlled river discharge and water depth. The different estuarine dam gate types with their different modes of operation (surface or bottom discharge) played an important role in the mixing and circulation of the altered estuaries.


Introduction
An estuary is a transitional zone between terrestrial and oceanic environments. Estuaries tend to be areas of high biological productivity and diversity, largely due to the exchange of water and nutrients that occurs in them (Levin et al. 2001). Consequently, estuaries have served as an essential source of habitat and food for living organisms, including humans. The exchange of water masses between estuaries and the coastal ocean is mainly the result of circulation driven by the density difference between freshwater and seawater (Dyer 1973). Classically, this "estuarine circulation" is a balance between the baroclinic pressure gradient and friction and is modified by the river discharge, depth, and wind (Geyer and MacCready 2014;Hansen and Rattray 1966a). The estuarine circulation plays a key role in controlling the transport of mass, such as salt, sediment, and nutrients (Miranda et al. 2012;Pu et al. 2015). Therefore, understanding the characteristics of the estuarine circulation pattern is essential for the sustainable management of estuarine systems (Garel et al. 2009).
A number of parameters have been suggested to describe the estuarine circulation (Dyer 1973;Ippen and Harleman 1961;Nielsen 2009). Among these, Hansen and Rattray (1966b) suggested a classification scheme based on two non-dimensional parameters: the stratification parameter and the circulation parameter. As they include the density-driven buoyancy forcing and velocity shear, these parameters have been used to study predominant forces, such as freshwater discharge and tidal currents (Andutta et al. 2013;Baltazar et al. 2011), flushing time (Kumari and Rao 2009), and the effects of morphological characteristics (Shivaprasad et al. 2013). Based on the two parameters, the stratification-circulation diagram classifies estuaries into four estuarine types: well-mixed, partially mixed, fjord, and salt wedge. Each estuarine type is further divided into mixed-for-stratification parameters smaller than 0.1 (i.e., type 1a) and stratified-for-stratification parameters larger than 0.1 (i.e., type 1b). Although many alternative approaches have been proposed subsequently, Hansen and Rattray's classification scheme has been applied to many estuaries around the world due to its simplicity and wide applicability (Bezerra et al. 2011;Geyer and MacCready 2014;Prandle et al. 2005;Zou et al. 2016).
As many estuaries have undergone intensive human development, anthropogenic effects need to be considered as a major factor affecting estuarine systems. In an "altered" estuary, human activities, which include land reclamation and the construction of dams and weirs, change the river discharge, tidal characteristics, and sediment dynamics (Kennish 2001;Williams et al. 2013). In particular, the construction of an estuarine dam impedes tidal propagation and saltwater intrusion, and results in controlled freshwater discharges into the downstream coastal water. For sustainable management of estuaries, it is important to understand how the characteristics of mixing and stratification are being affected by the anthropogenic alteration of estuaries (Ryu and Chang 1979). Classification of estuaries with parameterization may be helpful in describing the differences and similarities between natural and altered estuarine systems (Engle et al. 2007;U.S. Army Corps of Engineers 2000).
The Republic of Korea has experienced extensive and rapid coastal modifications over the last century, and approximately half of estuaries have been altered by anthropogenic structures, such as estuarine dams, sluice gates, and weirs Williams et al. 2014). Among the four major rivers in Korea, namely the Han, Nakdong, Geum, and Yeongsan rivers, the Han River estuary is the only one that has no estuarine dam. An estuarine dam plays a major role in the reduction of water mixing and the alteration of estuarine circulation (Rhew and Lee 2011). Besides the presence of an estuarine dam itself, the gate type and operation method also impact the characteristics of the discharge and, in turn, the circulation and mixing in the downstream side of the dam. For example, gate opening from the top versus bottom will affect the mixing of freshwater and saltwater. Therefore, the objective of this study is to understand the shift in estuarine circulation and stratification of altered estuaries by classifying the estuarine type of the four major estuaries in Korea using the stratification-circulation classification scheme. Specifically, we aim to determine (i) the classification of the estuarine type of the natural and altered estuaries, (ii) the different circulation characteristics of the altered estuary, and (iii) the factors that contribute to the estuarine type of altered estuaries in comparison to those for natural estuaries. A description of the four Korean estuaries and the data processing used in the study are given in Sect. 2. In Sect. 3, the observed salinity and velocity characteristics of the four estuaries and classification results are presented. The different circulation characteristics of the altered estuaries and the main factors affecting the estuarine type are discussed in Sect. 4. Finally, the main conclusions are presented in Sect. 5.

Study sites, data collection, and analysis
Field data were collected from four Korean estuaries ( Fig. 1): Han River Estuary (HRE), Nakdong River Estuary (NRE), Yeongsan River Estuary (YRE), and Geum River Estuary (GRE). At each estuary, two sets of data were collected: one for low or no-discharge conditions and the other for high-discharge conditions. The field measurement conditions of the four estuaries are given in Table 1.

Han River Estuary
HRE, located on the mid-western coast of the Korean Peninsula (Fig. 1a), is characterized by macro-tides with a maximum range as large as 10 m during spring tides (Park and Lee 2016). The maximum current reaches up to about 2 m s −1 during spring tides (Park et al. 2002). Most of the precipitation in the drainage basin occurs during summer, and thus 70% of the annual freshwater discharge takes place between July and September (Yoon and Woo 2012). Long-term mean freshwater discharge ranges from about 200 m 3 s −1 during the dry season to about 2000 m 3 s −1 during the wet season (Park et al. 2002). The Han river discharges into the coastal area through two channels: the Yumha channel and a channel north of Ganghwa Island (Fig. 1a). The Singok underwater weir is located about 60 km upstream from the river mouth at the northern end of Ganghwa Island (Fig. 1a). Unlike many estuarine dams and slush gates that block water flow entirely and affect the estuarine circulation, water level (Baek and Yim 2011), salinity, and fresh water discharge (Ahn and Lyu 2017) were not affected by the presence of the underwater weir at Jeollyu-ri, located about 15 km downstream of the weir. That is mainly because the freshwater is allowed to flow on the surface, whereas the salt water intrusion upstream of the underwater weir is prohibited. The HRE observations were conducted about 45 km downstream of Jeollyu-ri, and the flow and salinity structure do not appear to be affected by the Singok underwater weir. Thus, the HRE is considered natural at least near the measurement site in this study.
Field measurements were conducted at the Yumha channel during a high-discharge period in 1999 (37°37′53.35″N, 126°32′23.78″E) and a low-discharge period in 2008 (37°36′19.1″N, 126°33′07.3″E). The mean depth of each station was about 8 m in 1999 and 12 m in 2008. In 1999, a downward-looking acoustic Doppler profiler, mounted on the rail of a ship, collected velocity profiles every 15 s continuously and later 40 profiles were burst averaged to make hourly velocity profiles for 7 days on 5-12 August 1999. The blanking distance and bin size were both 1 m. Conductivity, temperature, depth (CTD) casting was conducted to collect vertical profiles of salinity and temperature at hourly intervals. In 2008, an upward-looking 1000 kHz Aquadopp, mounted on a trawl resistant bottom mount (TRBM), collected velocity profiles continuously with a sampling rate of 1 Hz for 22.5 h on 15-16 October 1999. The velocity profiles were averaged over 10 min every 30 min. The blanking distance was 0.4 m, and the bin size was 0.5 m. The CTD casting was concurrently conducted to collect vertical profiles of salinity and temperature at 30 min intervals. Daily freshwater discharge data at the Paldang dam, located 110 km upstream from the estuary mouth, were obtained from the Han River Flood Control Office, and hourly water level data were obtained from the Korea Hydrographic and Oceanographic Administration (KHOA)'s Incheon tidal gauge station.

Nakdong River Estuary
NRE, located in the southeastern part of Korea, has a microtidal range of about 1 m. About 68% of the annual freshwater discharge occurs during the wet season between July and September (Park and Lee 2016). Long-term mean discharge ranges from about 100 m 3 s −1 during the dry season to about 1200 m 3 s −1 during the wet season (Yoon et al. 2008). The NRE dam was completed in 1987 to prevent saltwater intrusion (Williams et al. 2013) and extends across Eulsuk Island (Fig. 1b). East of Eulsuk Island, the dam consists of 10 radial-type floodgates (Fig. 2a). During normal conditions, the gates remain closed and a small amount of freshwater may overflow over the top of the floodgates when the water level inside of the dam becomes higher than the dam. During a high-discharge period, the radial gates open from the bottom, resulting in freshwater discharge through the lower layer of the water column (Fig. 2b). The gates are usually opened only a few hours during low tide to prevent saltwater intrusion. The gates west of Eulsuk Island open only during extreme discharge conditions to alleviate the main floodgates east of Eulsuk Island.
Field measurements were conducted at the channel near the estuarine dam during a low-discharge period in 2010 (35°06′12.74″N, 128°57′10.76″E) and a high-discharge period in 2011 (35°06′10.57″N, 128°57′06.36″E). The mean depth of the station was about 8 m in 2010 and 9 m in 2011. Mounted on a TRBM, an upward-looking 1200 kHz acoustic Doppler current profiler (ADCP) collected 1024 velocity profiles at a sampling rate of 2 Hz every 30 min for 12.5 h on 30 June 2010 during low-discharge conditions, but collected profiles for about a month, 13 July -3 August 2011, during high-discharge conditions. The blanking distance and bin size were 0.22 and 0.1 m, respectively, in 2010, and 0.28 and 0.2 m, respectively, in 2011. Shipboard CTD casting was conducted to collect vertical profiles of salinity and temperature at 30 min intervals for 12.5 h on 30 June 2010 during low discharge and on 16 July 2011 during high discharge. Daily discharge data from the NRE dam were obtained from K-Water, and hourly water level data at the Gaduck Island tidal gauge station were obtained from KHOA.

Yeongsan River Estuary
YRE, located in southwest Korea, is a macrotidal environment with a spring tidal range of about 4.5 m (Park and Lee 2016). About 70% of the annual freshwater discharge occurs between July and September, with the average discharge ranging from about 160 m 3 s −1 during the dry season to about 1400 m 3 s −1 during the wet season. The YRE dam, completed in 1981, is located 6.7 km upstream from the estuary mouth at Goha Island (Fig. 1c) and has eight vertical-lift roller-type floodgates at the southern end of the channel. Unlike the NRE dam, there is no overflow while the gates remain closed. During flood, the gates open from the bottom. However, the large depth difference between upriver (7 m depth) and downriver (20 m depth) of the dam virtually results in surface discharge of freshwater (Fig. 2b). The gates are opened only a few hours during low tide to prevent saltwater intrusion.
Field measurements were conducted at the channel near the YRE dam during a highdischarge period in 2011 (34°47′13.47″N, 126°25′48.88″E) and a low-discharge period in 2012

Geum River Estuary
Located in the mid-western part of Korea, the GRE is a macrotidal environment with mean tidal range of 3.5 m . About 70% of the annual freshwater discharge occurs between June and September (Gunsan Regional Maritime Affairs and Port Office 2013), with the average discharge ranging from about 85 m 3 s −1 during the dry season to about 1000 m 3 s −1 during the wet season Yang 2014). Constructed in 1989 and beginning operation in 1994, the GRE dam is located 12 km upstream from the estuary mouth at Deajuk Island (Fig. 1d). The GRE dam has 20 shell-type roller floodgates at the southern half of the channel. Like Yeongsan, there is no overflow while the gates remain closed. During discharge, the floodgates lift and open from the bottom like Yeongsan, resulting in bottom discharge of freshwater (Fig. 2c). Because of the relatively small freshwater storage upstream of the dam, the gates are usually opened once a day during summer and once every two or three days in winter . Like in other estuaries, the gates are open a few hours during ebb tide to prevent saltwater intrusion.
Field measurements were conducted at the channel near the GRE dam during a lowdischarge period in 2015 (36°00′11.60″N, 126°44′18.92″E) and a high-discharge period in 2016 (36°00′13.69″N, 126°44′13.60″E), and the mean depth was about 6 m in both locations. In 2015, an upward-looking 600 kHz ADCP, mounted on a TRBM, was used to collect 1024 velocity profiles at 2 Hz every 30 min with a blanking distance of 0.5 m and a bin size of 0.25 m for 24 h on 27-28 August. CTD casting was conducted for 22.5 h during the same period. In 2016, an upward-looking 1200 kHz ADCP was used to collect the same number of velocity profiles at 2 Hz every 30 min from 6 August to 3 September. The blanking distance and bin size were 0.38 and 0.25 m, respectively. CTD casting was conducted only for 25 h on 2-3 September 2016. Discharge data from the GRE dam were obtained from the Korea Rural Community Corporation, and hourly water level data at the Janghang port tidal gauge station were obtained from KHOA.

Data analysis
Velocity and salinity data were selected for analysis only when both data overlapped ( Table 1). The instantaneous current velocities were averaged to obtain a burst-mean current velocity. The surface current velocities were reconstructed without applying the customary practice of removing the top 10%-20% of surface data that are degraded by side lobe effects. For the upward-looking Aquadopp and ADCP velocity data, the surface current velocity was reconstructed using the method described in Shin et al. (2014). For the downwardlooking acoustic Doppler profiler data of the HRE in 1999, the surface velocities were estimated by the extrapolation of the velocity data using the law of the wall. However, some of the bottom current data were absent due to the frame height and the blanking distance. The raw salinity profile data were smoothed using weighted linear least squares and a second degree polynomial regression. The smoothing method assigns zero weight to data outside six mean absolute deviations (used "rloess" method in the Matlab proprietary software function of "smooth") to remove the outliers. Some CTD data were absent near the surface and bottom boundaries. Next, the current velocity was rotated on the principal components of current velocity with positive (negative) values indicating seaward (landward) flows, and they were used for subsequent analysis. Kjerfve (1975) emphasized that the use of sigma depth is more appropriate in a macrotidal environment where the ratio of tidal range to mean water depth exceeds 0.3. Thus, the current velocity and salinity data were transformed to a sigma depth of 0.05 intervals from bottom (sigma depth of 0) to surface (sigma depth of 1).
The four variables needed to estimate the two parameters of the stratificationcirculation diagram were estimated: the depth-mean salinity, the salinity difference between bottom and surface, the surface velocity, and the depth-mean velocity. Then, the tidal components were removed by applying a Lanczos low-pass filter with a cutoff period of 36 h for data longer than two tidal cycles (Emery and Thomson 2004). For the data shorter than two tidal cycles, averages over the entire duration of data were used. Finally, the stratification parameter ðδS=SÞ was estimated by the ratio of the tidally averaged salinity difference between bottom and surface (δS) and the tidally averaged depth-mean salinity ðSÞ. The circulation parameter ðu s =UÞ was calculated by the ratio of the magnitude of the tidally averaged surface velocity (u s ) and the tidally averaged depth-mean velocity ðUÞ (Hansen and Rattray 1966b).

Natural macrotidal Han River Estuary
The low-discharge (139-169 m 3 s −1 ) observation was made for 22.5 h during a spring tide in October 2008 in the macrotidal HRE (Fig. 3). Salinity showed distinctive tidal variations from 20 to 30 psu over a tidal cycle with S of 24.5 psu (Figs. 3b and 3d). Although the difference between bottom and surface salinity slightly increased to about 1.5 psu during the ebb, it was usually smaller than 0.5 (Fig. 4a), with δS being about 0.2 psu (Fig. 3e), which resulted in a relatively small stratification parameter ðδS=SÞ of about 0.01 (Fig. 3f). Current velocities were strong at about 1.5 m s −1 with the ebb current slightly larger, and U was 0.12 m s −1 (Figs. 3c, 3g, and 4a). On the other hand, u s was 0.17 m s −1 (Fig. 3h), which resulted in a circulation parameter ðu s =UÞ of about 1.4. Thus, during the low-discharge conditions, the HRE was classified as a type 1a (well-mixed) estuary.
The large discharge (1105-4215 m 3 s −1 ) observation was done over 7 days during a neap to a spring tide in August 1999 (Fig. 5). Salinity exhibited much larger variations from almost 0 to 23 psu (Fig. 5b) because of the large discharge that occurred 5 days before the field observation. The depth-mean salinity fluctuated by more than 15 psu over the tidal cycles, but the tidally averaged depth-mean salinity recovered gradually from about 5 to 10 psu during the observation period ( Figs. 4b and 5d). Although the difference between the bottom and surface salinity remained small during low tide (Fig. 4b), it increased up to 10 psu during the high tide right after the discharge with a gradual decrease over the 7 days of observation (Fig. 5e). The gradual increase of S and the gradual decrease of δS made δS=S decrease from 0.60 to 0.16 from neap to spring. The large discharge increased the current velocities up to 2 m s −1 during the ebb, but U was relatively small in the range between 0.07 and 0.22m s −1 (Figs. 4b, 5c, and 5g). The value of u s was also small in the range between 0.04 and 0.23 m s −1 . The resulting u s =U varied from 0.67 to 1.08, but remained smaller than 2 (type 1), as in low-discharge conditions. During high-discharge conditions, the HRE was classified as type 1b (stratified).

Altered shallow microtidal Nakdong River Estuary
The low-discharge (162 m 3 s −1 ) data were collected for 12.5 h on 30 June 2010 in the shallow microtidal NRE (Fig. 6). While the gates of the estuary dam remained closed, a small amount of freshwater overflow maintained a low surface salinity of about 10 psu (Fig. 6b). At the bottom, the salinity remained about 32 psu. The small amount of discharge (162 m 3 s −1 ), begun at 1600 and lasting for about 20 min, decreased the surface salinity from 17 psu to less than 10 psu, but the bottom salinity remained at about 32 psu. The vertical profile of salinity showed distinct vertical stratification (Fig. 4c). The values of S and δS were 25 and 18 psu, respectively, which resulted in a relatively large δS=S of 0.73 (Figs. 6d-6f). The current velocities were in general weak due to the presence of the estuarine dam, except for the period of discharge during which the peak current velocity reached 0.71 m s −1 (Fig. 6c). The value of U was 0.04 m s −1 (Fig. 4c), and u s was also about 0.04 m s −1 , which resulted in u s =U of about 1 (Figs. 6g-6i). During low-discharge conditions, the NRE was classified as type 1b (stratified).
On 16 July 2011, the high-discharge (2576 m 3 s −1 ) observation was made during a spring tide for 12.5 h (Fig. 7). Because of a large discharge event (∼12 700 m 3 s −1 ) that continued from 9 to 15 July 2011, salinity was virtually zero throughout the water column (Fig. 7b). Fig. 3. Time-series of stratification and circulation parameters for low-discharge conditions for the HRE: (a) water depth, (b) salinity, (c) current velocity (positive, ebb and negative, flood), (d) depth-mean (blue line) and tidally averaged depth-mean salinityS (black dot), (e) difference between the bottom and surface salinity (blue line) and tidally averaged difference of salinity δS (black dot), ( f) stratification parameter, (g) depth-mean (blue line) and tidally averaged depth-mean velocityŪ (black dot), (h) surface velocity (blue line) and tidally averaged surface velocity u s (black dot), and (i) circulation parameter. Fig. 4. Vertical profile of tidally averaged salinity, temperature, and velocity: (a) low-discharge conditions for HRE, (b) high-discharge condition for HRE, (c) low-discharge condition for NRE, (d) high-discharge condition for NRE, (e) low-discharge condition for YRE, (f) high-discharge condition for YRE, (g) low-discharge condition for GRE, and (h) high-discharge condition for GRE. Three lines (solid, dotted, and dashed) in (b) and (f) represent vertical profile averaged over two tidal cycles for different times shown on the figure. Fig. 5. Time-series of stratification and circulation parameters for high-discharge conditions for the HRE: (a) water depth, (b) salinity, (c) current velocity (positive, ebb and negative, flood), (d) depth-mean (blue line) and tidally averaged depth-mean salinityS (black dot), (e) difference between the bottom and surface salinity (blue line) and tidally averaged difference of salinity δS (black dot), ( f) stratification parameter, (g) depth-mean (blue line) and tidally averaged depth-mean velocityŪ (black dot), (h) surface velocity (blue line) and tidally averaged surface velocity u s (black dot), and (i) circulation parameter. Fig. 6. Time-series of stratification and circulation parameters for low-discharge conditions for the NRE: (a) water depth, (b) salinity, (c) current velocity (positive, ebb and negative, flood), (d) depth-mean (blue line) and tidally averaged depth-mean salinityS (black dot), (e) difference between the bottom and surface salinity (blue line) and tidally averaged difference of salinity δS (black dot), ( f) stratification parameter, (g) depth-mean (blue line) and tidally averaged depth-mean velocityŪ (black dot), (h) surface velocity (blue line) and tidally averaged surface velocity u s (black dot), and (i) circulation parameter.
The value of S was very small (∼0.06 psu, Fig. 4d), and δS was virtually zero (∼10 −4 psu), which resulted in a very small δS=S of about 0.001 (Figs. 7d-7f). The currents were directed seaward because of the freshwater discharge throughout the tidal cycle (Fig. 7c), and a Fig. 7. Time-series of stratification and circulation parameters for high-discharge conditions for the NRE: (a) water depth, (b) salinity, (c) current velocity (positive, ebb and negative, flood), (d) depth-mean (blue line) and tidally averaged depth-mean salinityS (black dot), (e) difference between the bottom and surface salinity (blue line) and tidally averaged difference of salinity δS (black dot), ( f) stratification parameter, (g) depth-mean (blue line) and tidally averaged depth-mean velocityŪ (black dot), (h) surface velocity (blue line) and tidally averaged surface velocity u s (black dot), and (i) circulation parameter. maximum current velocity of 0.82 m s −1 was observed during the low tide. The value of U was about 0.42 m s −1 (Fig. 4d), and u s was 0.45 m s −1 , which resulted in u s =U less than 2 (Figs. 7g-7i). During high-discharge conditions, the NRE was classified as type 1a.

Altered deep macrotidal Yeongsan River Estuary
The no-discharge observation at the macrotidal YRE was conducted for 25 h on 18-19 June 2012 (Fig. 8). Prior to the observation, the dam had been closed for a month. Because of the long cessation of discharge, the salinity was as high as 34 psu (Fig. 8b), and the vertical difference of the salinity remained less than 1 psu (Fig. 4e). The values of S and δS were about 33 and 0.5 psu, respectively, which resulted in a small δS=S of about 0.01 (Figs. 8d-8f). The currents were very weak, less than 0.1 m s −1 with minor tidal fluctuations (Fig. 8c), as the flow was blocked by the dam. The value of U was very small (<0.01 m s −1 ; Fig. 4e), and u s was about −0.02 m s −1 , directed landward (Figs. 8f and 8h). The absolute value of u s was taken in calculating u s =U, which was about 12 (>2) because of the very weak U. Thus, during the no-discharge condition, the YRE was classified as a type 2a (partially mixed) estuary. Although the YRE is classified as type 2, it behaved like a bay with salinity being close to 33 psu throughout the water column. This is further discussed in Sect. 4.3.
The high-discharge observation of the YRE was made for about 3 days, 14-17 August 2011 (Fig. 9). There were two discharge events (2812 m 3 s −1 ) on 14 August 2011, and the field observation started about 1 h after the first discharge event (not shown in Fig. 9). The freshwater discharge generated a surface layer above a halocline with a thickness that gradually decreased during the observation period from about 5 m initially (Figs. 4f and 9b). The salinity decreased to about 4 psu and gradually recovered in the surface layer, whereas the salinity remained about 29 psu below the halocline. Although the depthmean salinity rapidly decreased down to 19 psu during the discharge, S was maintained at 26-28 psu (Figs. 4f and 9d). On the other hand, the salinity difference between the surface and bottom layers was as large as 25 psu during the discharge and then decreased to about 5 psu. This trend resulted in δS varying from 21 to 7 psu. Therefore, δS=S varied from 0.8 to 0.3. Current velocities were generally weak, similar to observations during no-discharge conditions, except when the freshwater discharge occurred (Fig. 9c). A seaward peak velocity of 1 m s −1 was observed at the surface layer during the discharge, and thus the depth-mean velocity peaked at about 0.1 m s −1 during the discharge event. After the gate closed, the depth-mean velocity showed weak tidal variation within ±0.05 m s −1 . The value of U was very weak, close to 0, and u s varied between −0.04 and 0.03 m s −1 (Figs. 4f, 9g, and 9h). The resulting u s =U was mostly larger than 2, varying from 0.3 to 33 because of the intermittent, strong freshwater discharge in the surface layer in the deep YRE (Fig. 9i). Therefore, the discharge condition of the YRE exhibited the following transition: type 1btype 2btype 3btype 2btype 1b. These variabilities were caused mainly by the rapid change of the surface velocity u s over relatively stable U at the deep YRE.

Altered shallow macrotidal Geum River Estuary
The low-discharge observation was made for 22.5 h during 27-28 August 2015 in the shallow macrotidal GRE (Fig. 10). A discharge event (1997 m 3 s −1 ) occurred two days before the observation, resulting in salinity varying from 16 to 24 psu. The depth-mean salinity increased from 18 to 23 psu with S of about 21 psu (Figs. 4g and 10d). The difference between the bottom and surface salinity showed a tidal variation, decreasing down to 0 toward the low tide and increasing up to about 2 psu toward the high tide (Fig. 10e). The value of δS was very small at about 1.2 psu, which resulted in relatively small δS=S of about 0.06. Current velocity was moderate in the range of ±0.6 m s −1 with the flood slightly larger than the ebb (Fig. 10c). The value of U was about −0.06 m s −1 (Fig. 4g), and u s was about −0.07 m s −1 (Figs. 10g and 10h). The resulting u s =U was less than 2, and therefore during the lowdischarge condition, the GRE was classified as type 1a (well-mixed). Fig. 8. Time-series of stratification and circulation parameters for low-discharge conditions for the YRE: (a) water depth, (b) salinity, (c) current velocity (positive, ebb and negative, flood), (d) depth-mean (blue line) and tidally averaged depth-mean salinityS (black dot), (e) difference between the bottom and surface salinity (blue line) and tidally averaged difference of salinity δS (black dot), ( f) stratification parameter, (g) depth-mean (blue line) and tidally averaged depth-mean velocityŪ (black dot), (h) surface velocity (blue line) and tidally averaged surface velocity u s (black dot), and (i) circulation parameter.
The relatively high-discharge (2290 m 3 s −1 ) observation was made for 25 h during the spring tide in 2-3 September 2016 (Fig. 11). The salinity exhibited a large variation from 9 to 27 psu, but the depth-mean salinity varied from 21 to over 24 psu with S of 23 psu Fig. 9. Time-series of stratification and circulation parameters for high-discharge conditions for the YRE: (a) water depth, (b) salinity, (c) current velocity (positive, ebb and negative, flood), (d) depth-mean (blue line) and tidally averaged depth-mean salinityS (black dot), (e) difference between the bottom and surface salinity (blue line) and tidally averaged difference of salinity δS (black dot), ( f) stratification parameter, (g) depth-mean (blue line) and tidally averaged depth-mean velocityŪ (black dot), (h) surface velocity (blue line) and tidally averaged surface velocity u s (black dot), and (i) circulation parameter. (Figs. 4h, 11b, and 11d). The vertical difference of salinity decreased from 16 psu to near 0 and δS was about 4 psu, resulting in a relatively large δS=S of 0.18 (Figs. 11e and 11f). The current velocities were moderate with the flood current of about −0.85 m s −1 being stronger than Fig. 10. Time-series of stratification and circulation parameters for low-discharge conditions for the GRE: (a) water depth, (b) salinity, (c) current velocity (positive, ebb and negative, flood), (d) depth-mean (blue line) and tidally averaged depth-mean salinityS (black dot), (e) difference between the bottom and surface salinity (blue line) and tidally averaged difference of salinity δS (black dot), ( f) stratification parameter, (g) depth-mean (blue line) and tidally averaged depth-mean velocityŪ (black dot), (h) surface velocity (blue line) and tidally averaged surface velocity u s (black dot), and (i) circulation parameter.
the ebb current of about 0.5 m s −1 (Fig. 11c). The value of U was −0.04 m s −1 (Fig. 4h), and u s was small at about −0.002 m s −1 , which resulted in u s =U of about 0.04. As a result, during the discharge condition, the GRE was classified as type 1b (stratified). Fig. 11. Time-series of stratification and circulation parameters for high-discharge conditions for the GRE: (a) water depth, (b) salinity, (c) current velocity (positive, ebb and negative, flood), (d) depth-mean (blue line) and tidally averaged depth-mean salinityS (black dot), (e) difference between the bottom and surface salinity (blue line) and tidally averaged difference of salinity δS (black dot), ( f) stratification parameter, (g) depth-mean (blue line) and tidally averaged depth-mean velocityŪ (black dot), (h) surface velocity (blue line) and tidally averaged surface velocity u s (black dot), and (i) circulation parameter.

Do altered estuaries have different circulation characteristics?
In this paper, three altered Korean estuaries (shallow microtidal NRE, deep macrotidal YRE, and shallow macrotidal GRE) and one natural Korean estuary (intermediate depth macrotidal HRE) were classified into estuarine types to examine the circulation characteristics of altered estuaries. To compare the difference of estuarine type between the natural and altered estuaries, the determined estuarine types are shown in Fig. 12 along with the estuarine types of natural estuaries reported by Prandle (1985) for similar environmental conditions, such as water depth, tidal range, and freshwater discharge.
The shallow microtidal NRE can be compared with the James River estuary, Virginia, USA, which has similar water depth and tidal conditions ( Table 2). While the James River estuary is well documented as a partially mixed estuary (type 2b) because of the weak vertical turbulent diffusion related with the microtidal current (Bradshaw and Kuo 1987;Dyer 1973), the altered NRE was classified as a stratified estuary (type 1b) during the lowdischarge condition and as a well-mixed estuary (type 1a) during the high discharge (Fig. 12). This difference is attributed to the decreased bottom residual current and the relatively intensified discharge by the estuarine dam. Before the dam was constructed, the tidal current was in natural balance with the freshwater discharge and waves, and the difference in the residual current between the surface and bottom of the water column was observed in the NRE (Jang and Kim 2006;Ryu and Chang 1979). However, the dam construction decreased wave energy and, while the dam was closed, the bottom residual current became considerably weak despite the surface overflow. During the high episodic discharge, the water column was well mixed. While this does occur in a natural river flood, here it is more likely related to the timing of discharge (discrete and not continuous). If it were continuous, a partially mixed estuarine type might result. The two observations were conducted during similar seasons, with different tidal periods (transition to neap in 2010 and to spring tide in 2011). Consequently, the results indicate that the circulation of NRE has shifted to becoming discharge-dominated, in agreement with previous studies (Jang and Kim 2006;Williams et al. 2015).
The deep macrotidal YRE can be compared to the Columbia River estuary, USA, which has similar deep macrotidal environmental conditions. The natural Columbia River estuary is classified mainly as a partially mixed estuary (type 2b) during both high-and lowdischarge conditions (Table 2 and Fig. 12). On the other hand, the altered YRE was classified as partially mixed without stratification (type 2a) while the gates of the estuary dam were closed. Because there had been no freshwater discharge for a month, the circulation was hard to characterize as "estuarine circulation". The saltwater environment with little density gradient was similar to a coastal bay, and the tidal current was very weak, although YRE is a macrotidal estuary. The reduced tidal current and the stagnant water by the YRE dam are consistent with the numerical research results of Kang (1999). During the highdischarge condition, the short, strong surface discharge caused a large variation of surface velocity relative to the very weak tidally averaged depth-mean velocity, resulting in dramatic variation of the circulation parameter. Thus, the estuary type was changed from type 1b and type 2b to type 3b as the gates were opened and recovered from type 3b to type 2b to type 1b after the gates were closed. Surface jets of freshwater with large variation of velocity agreed well with previous results (Kim et al. 2013;Park et al. 2012;Shin et al. 2014). As a result, the dam caused the circulation of YRE to behave more like a semi-closed coastal bay, and then like a fjord during high discharge. While the Columbia River estuary and YRE are both often type 2b, the episodic discharge at YRE dramatically changes its estuarine classification compared to its natural counterpart. Fig. 12. Stratification-circulation diagram for four Korean estuaries. For comparison, other natural systems are shown with black closed circles (•), which include Columbia River Estuary (Co H and Co L are for high-and lowdischarge conditions, respectively), James River Estuary (Ja), River Mersey Estuary (Me), River Tay Estuary (Ta), River Tees Estuary (Te), and River Thames Estuary (Th). Estuarine types 1-4 indicate well-mixed, partially mixed, fjord type, and salt wedge type estuaries, respectively, and each type is further divided into mixed for δS=S < 0.1 (e.g., type 1a) and stratified for δS=S > 0.1 (e.g., type 1b). Note: u s =U is the ratio of the tidally averaged surface velocity to the tidally averaged depth-averaged velocity; and δS=S is the ratio of the tidally averaged bottom-surface salinity difference to the tidally averaged depth-averaged salinity. a The mean depths indicate the depths at the observation stations for the Korean estuaries and the depths at the mouths for other systems from Hansen and Rattray (1966b) and Prandle (1985).
The shallow macrotidal altered GRE can be compared with the Tay (Scotland), Tees (England), and Thames (England) river estuaries, which are all classified as well-mixed estuaries (type 1; Table 2). While the natural HRE and altered GRE are both classified as well-mixed estuaries (type 1), their circulations are different. The salinity structure of the HRE followed the typical structure of natural estuaries, such as Rivers Tay, Tees, and Thames. That is, the distinct tidal variation of vertically homogeneous salinity is modulated by the vertical mixing of a strong tidal current. On the other hand, a distinct and strong tidal variation was not shown in altered GRE. Before the dam was constructed, strong and vertically homogeneous tidal current profiles were observed in GRE (Oh et al. 1995). However, the dam construction reduced the tidal modulation of salinity and velocity in the altered GRE, although the estuary type (type 1) was unaffected. This was attributed to the frequent but episodic discharges, which prevented it from being vertically mixed and tidally modulated. Hansen and Rattray (1966b) defined the four estuary types with the dominant circulation mechanism. To identify the important factors determining the type of estuarine circulation, Hansen and Rattray (1966b) also introduced two dimensionless quantities: the densimetric Froude number F m = u f /u d and the flow ratio p = u f /u t , where u f is the river discharge per unit cross-sectional area of the estuary; u t is the root mean square tidal current speed; and u d is the densimetric velocity, calculated as u d = ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ffi ðΔρ=ρÞgH p , where Δρ is the density difference between river water and seawater, H is the mean depth of the estuary, and g is acceleration of gravity.

Comparison of factors contributing to estuarine type for natural and altered estuaries
According to the relationship between the dimensionless quantities and the stratificationcirculation diagram ( fig. 4 in Hansen and Rattray 1966b), the type 1 estuaries are where the diffusive flux is dominated by a strong tide (u t ≫ u f ) and the water depth is not deep (F m ≤ 1; u d ≤ u f ). Rivers Tees, Tay, and Thames and Han River estuaries have similar depth and macrotidal characteristics, in agreement with these two environmental characteristics. In particular, the strong tidal modulation of the HRE indicates that the tide is the dominant factor and the discharge is the secondary factor.
Type 2 estuaries have either deep water depth or weak vertical mixing compared to type 1 estuaries (F m > 1; u d > u f ), and both advection and diffusion are important in the circulation. Within the type 2 parameter space, the upper left corner characterizes estuaries in which the advection by river discharge is relatively dominant (u f > u d ) and the lower right corner those in which the tidal mixing is stronger (u f < u d ) (Fig. 12). Columbia, James, and Mersey (England) river estuaries have deep or microtidal characteristics, and the range of the river discharge increases from Mersey to Columbia river estuaries in agreement with this relationship (Fig. 12).
On the other hand, the important factors determining the type of altered estuaries are simpler than those of natural estuaries. In altered estuaries, only the densimetric velocity u d and river discharge velocity u f are important because the effect of the tidal current u t is diminished by the presence of the dam. During no-or low-discharge conditions, the longitudinal difference of salinity may be weak due to the intermittent freshwater discharge, which renders the densimetric velocity unimportant (u d ≤ u f ). In other words, the altered NRE and GRE become diffusion-dominant estuaries (type 1) owing to the freshwater discharge being controlled by the dam. Meanwhile, the densimetric velocity of the altered YRE is relatively strong owing to the deep water depth with a weak freshwater discharge (u d > u f ), and thus the estuary type becomes type 2. This indicates that both advection and diffusion are important in the circulation of the altered YRE.
During the high-discharge period, the controlled discharge is stronger than the natural gravitational convection (u d ≤ u f ) in the NRE and GRE and the estuary types become type 1 again. However, in the deep YRE, only the surface velocity was changed rapidly depending on the controlled river discharge, and the estuary type also rapidly changed from diffusion dominance (type 1) to being dominated by both advection and diffusion (type 2) and then to advection dominance (type 3). This suggests that the controlled freshwater discharge and water depth are the key factors that control the circulation type of altered estuaries.
In addition, the type of floodgate also plays an important role in determining the estuarine stratification. During the low-discharge period, the freshwater overflow increases the vertical salinity gradient in the NRE, while the lift floodgates of the YRE and GRE prevent the freshwater overflow and decrease the vertical salinity gradient. During the highdischarge period, the bottom discharge of NRE and GRE increases the vertical mixing, whereas the surface discharge of the deep YRE increases the stratification by decoupling the surface flow from the depth-mean flow. Therefore, the stratification type (type a or b) depends on the dam's gate operation in altered estuaries.
In this study, we analyzed the factors affecting the type of estuaries by analyzing the short-term data of flow and salinity. However, further studies on the dynamics between longer term discharge and topographic change are needed to clarify the processes and factors of estuarine alteration. While the estuarine dams were constructed, NRE, YRE, and GRE were reduced in size and channelized due to the reclamation of tidal flats Williams et al. 2013Williams et al. , 2014. When the intertidal zone decreased through reclamation, the ebb-dominant GRE transformed to flood-dominant . Moreover, wavedominant estuaries can be shifted to river-dominant ones by the channelization of estuaries by reclamation and dredging (Williams et al. 2013). Analyzing flow dominance coefficient through long-term effluent data over the topographic changes at the Changjiang estuary of China, Dai et al. (2013Dai et al. ( , 2016 also found that the reduction of the intertidal zone decreased the ebb-dominant nature of the estuary. Therefore, it is required to understand more precise factors influencing the estuarine circulation and consequent shift of the estuarine type by considering the construction of estuarine dam and slush gates, as well as the reclamation of tidal flats and the dredging of the channel, which is the topic of a future study.

Limitations of the stratification-circulation diagram
Application of the stratification-circulation classification scheme requires consideration of the possibility that the classification can shift depending on the measurement time as well as the measurement station of an estuary (Bowden and Gilligan 1971;Hansen and Rattray 1966b). Bowden and Gilligan (1971) also pointed out that estuaries are not always in a steady state, as assumed in Hansen and Rattray's theoretical model. In this research, the considerable variation of estuarine type in YRE (Fig. 12) demonstrated the unstable alteration of estuary circulation, especially due to the dam's gate operation. This variation made it difficult to define the estuarine type in a conventional way but was useful to improve our understanding of the effects of estuarine dams on the estuarine circulation. Therefore, the classification can be improved by adding various stations and time periods in altered estuaries.
The stratification-circulation diagram was used in this study to represent the circulation characteristics of altered estuaries. While this widely used scheme is simple and straightforward, it was incapable of representing all estuarine types. For example, the stratificationcirculation classification scheme classified the deep, macrotidal altered YRE as a partially mixed estuary during the no-discharge condition. However, this does not follow the classical physics of estuarine circulation because there had been no discharge for a month.
Instead it was observed that warm ocean water with high salinity flowed in near the surface during the flood and mixed estuarine water flowed out through the middle and bottom layers during the ebb (Fig. 13). Based on this, the altered YRE during no-discharge conditions would more appropriately be classified as an inverse estuary (Valle-Levinson 2010).
Recently, Geyer and MacCready (2014) sug gested a more elaborated classification with several additional types, such as "bay type" and "periodic stratification type". While the stratification-circulation classification is a diagnostic tool to better understand the dominant processes from the observed stratification circulation characteristics (Geyer 2010), the classification scheme suggested by Geyer and MacCready (2014) is a prognostic approach to predict the stratification circulation characteristics by assuming forcing conditions. The prognostic approach usually requires more detailed information about an estuary, such as the bottom boundary stress and buoyancy frequency (Geyer and MacCready 2014;Prandle et al. 2005). Thus, further investigation is required to use a prognostic approach for improved understanding of the circulation characteristics in the various types of altered estuary.

Conclusion
In this study, we have examined the circulation characteristics of altered estuaries using the stratification-circulation classification scheme. The results confirmed that the stratification and circulation characteristics and the resulting estuarine type of altered estuaries were modified in comparison with their natural counterparts. The differences stem mainly from the presence of the estuarine dam, which changes the discharge pattern according to the gate operation. The altered river discharge included no-discharge conditions, weak overflow from a closed radial gate, and intermittent strong discharge from a lift gate. Furthermore, the depth of the estuary was the crucial factor, along with the impact of controlled river discharge, in determining the estuarine type of altered estuaries. We did not consider the effect of altered estuarine circulation on ecosystems, water quality, or the transport of organic and sediment materials. However, greater exposure of estuaries to anthropogenic development is likely to strengthen the alteration of the estuarine circulation. The classification scheme using the stratification-circulation diagram can be useful to understand the effects of these alterations. Moreover, to better classify the diverse estuarine types and predict their stratification circulation characteristics, future studies should use new estuarine classifications that incorporate measurement data of various physical and biological properties, as well as modeling studies of altered estuaries.