Site description and monitoring strategy

Yudeung Stream, a major tributary of the Geum River in South Korea, is 44 km long and has an approximate drainage area of 289.14 km². Most sections, except for the upstream parts, run through downtown Daejeon City and are constantly disturbed by human activity. We selected 15 study sites that reflect diverse microhabitats from upstream to downstream sections (Fig. 1) and surveyed the stream from early May to late June 2019 over a period of eight weeks before the occurrence of summer monsoon (July to August; Choi & Kim, 2020a) to avoid physical disturbance of fish distributions caused by the monsoon. We randomly selected three sampling points per site based on virtual grids constructed over maps of each study site.

Environmental variables including water temperature, pH, dissolved oxygen (DO), conductivity, turbidity, and water velocity were measured in three quadrats at each of the study sites. We used a DO meter (YSI Model 58; YSI Incorporated, Yellow Springs, OH, USA) to measure temperature and DO, a Fisher Conductivity Meter (YSI model 152; YSI Incorporated) to measure conductivity, and an Orion Model 250A pH meter (Orion Research, Beverly, MA, USA) to measure pH. Turbidity was measured using a turbidimeter (Model 100B; HF Scientific Inc., Ft. Myers, FL, USA). Average water velocity was measured using a portable multi-parameter water checker (Hach Co., Loveland, CO, USA) at six sampling points for each site. Analysis of factors associated with regional variation of fish assemblages was performed using both environmental and spatial explanatory variables.

Fish were caught using a 3.5 m diameter cast net with 7×7 mm mesh size and a 5×5 mm mesh dip net (0.32 m² area). In cast net sampling, the same points along the 100 m riparian zone were sampled in the upper and lower part of the stream at all events. Cast net was cast 10 times at 10 m intervals to yield a total of 20 casts, covering a total water surface area of 61.2 m². Dip net sampling was conducted one hour later along the same stretch of water, with 20 samples covering 6.4 m² water surface area. In study sites dominated by aquatic macrophytes and boulders, we used only dip net sampling. All collected fish were preserved in methanol-formaldehyde solution and were identified to species level according to Kim and Park (2002) and Kim et al. (2005) based on the classification system of Nelson et al. (2016).

Identification of microhabitat patch types

We delineated the habitat and characterized the patches within 200 m upstream and downstream of each study site using a combination of remotely sensed and field-collected data (Kobayashi et al., 2012). Specifically, we used a combination of on-screen digitizing in ArcGIS 10.1 and Arc Pad 8.0 (Environmental Systems Research Institute, Redlands, CA, USA) on a desktop computer. Characterized patches of the study area were obtained from satellite images (Daum Kakao Map in June, 2019; https://map.kakao.com) with 50 cm resolution. Using this approach and guidelines adapted from Holmes and Goebel (2011) and Johansen et al. (2010), we identified and digitized six patch types in the riverine microhabitat as riffle, pool, shallow, boulder, sand/clay, and vegetated (Table 1). Several microhabitat patches that are difficult to distinguish from satellite images were additionally identified through field surveys. Riffles were classified as areas with a steep surface and shallow depth relative to up- and downstream areas, with swift flow and the surface usually broken (white-colored water). Pools were classified as areas with flat, smooth surfaces with greater depth relative to upstream and downstream areas. Shallows were classified as areas of less than 1 m depth, which could overlap with riffles and pools. Boulder classification was used to refer to patches with bottom structure larger than 256 mm. Sand/clay referred to areas with very fine particles of 0.125-0.004 mm. Vegetated areas included those occupied by aquatic macrophytes, namely emergent, floating, and submerged plants. We identified aquatic macrophytes species appearing at each study site.

Data analysis

Environmental variables and microhabitat patch types were log-transformed after assessing their degrees of normality based on the Shapiro-Wilk test. Also, Seventeen species were selected for further analysis. We then used non-metric multidimensional scaling (NMDS) to examine species distribution patterns according to their abundance at each site. NMDS ordination plots were produced based on Euclidean distance, and goodness-of-fit was assessed in terms of loss of stress. Stress value for the two-dimensional solution was 0.186, lower than the generally accepted maximum stress value of <0.2 (Clarke, 1993). Environmental variables and microhabitat patch types were fitted to NMDS ordination axes scores. The significance of the fitted vectors was assessed using 3000 permutations, with P<0.05 considered significant. NMDS ordination was conducted with the R package ‘vegan’ (version 2.5-3; Oksanen et al., 2013).

We also clustered the 15 sites based on environmental variables and microhabitat patch types, allowing the generation of probability values for clusters using bootstrap resampling with 1000 runs. Clustering was carried out using R version 2.5-3 (R Core Team, 2019).

Environmental variables and microhabitat patch types

Environmental variables from each study site reflected the limnological characteristics of a temperate stream (Table 1). Among environmental variables, DO, conductivity, turbidity, and water velocity differed by site (one-way ANOVA, P<0.05). DO was generally highest in upstream regions, with sites 1-3 having <100% DO, and decreased downstream with a range of 47.3 to 95.3%. This reflects water velocity. In contrast, conductivity and turbidity increased downstream. Water temperature and pH did not differ notably among study sites (one-way ANOVA, P>0.05).

Microhabitat patch types differed markedly across study sites (Table 1). Greater areas of pool, sand/clay, and vegetation were present in the downstream regions, whereas riffle and boulder areas dominated in the upstream regions. In streams and rivers, the growth of aquatic macrophytes is closely related to water flow, and downstream areas with relatively lower flow are more suitable for macrophyte growth. The increase in sand/clay areas over boulders in the downstream also contributes to the development of aquatic macrophytes. Plant community composition varied similarly across the study sites. Most of the study sites were dominated by Phragmites communis Trin. and Paspalum distichum L., which were accompanied by a total of seven plant species: Miscanthus sacchariflorus (Maxim.) Hack, Typha orientalis C. Presl, Typha angustifolia L. sensu lato, Paspalum distichum var. indutum Shinners, Ceratophyllum demersum L., Potamogeton malaianus Miq. var. latifolius Nakai ex Mori, and Potamogeton oxyphyllus Miq.

Clustering of the study sites based on environmental similarity and microhabitat patch types yielded three groups with homogeneous native fauna (Fig. 2). These groups reflected the microhabitat characteristics of the studied aquatic ecosystems, implying an underlying biogeographical pattern. Group 1 included sites 14 and 15 as representative of downstream characteristics, and Group 2 comprised sites 10 to 13, representing the middle section. The remaining sites 1-9 in the upstream section constituted Group 3. Based on this, microhabitat characteristics of Yudeung Stream could be clearly distinguished by distinct environmental variables and microhabitat patch types in the upstream to downstream direction.

Effect of fish distribution on microhabitat structure

A total of 2042 individuals belonging to 24 fish species were collected at the 15 study sites (Table 2). Cyprinidae was the dominant family, represented by 14 species, with a relative abundance of 92.1%. They were followed by the Centrarchidae (3.0%), Gobiidae (1.7%), and Centropomidae (1.4%). The remaining families (Odontobutidae, Cobitidae, and Bagridae) represented less than 1%. These distribution characteristics are consistent which reported the dominance of Cyprinidae in freshwater streams (Kim & Park, 2002). On a species level, Zacco platypus dominated with 1015 individuals (ind.) (relative richness, 50%) across study sites, followed by Zacco koreanus (497 ind., 24.4%), Pungtungia herzi (105 ind., 5.2%), and Coreoleuciscus splendidus (80 ind., 3.9%) (Figs. 3, 4). The greatest abundance of fish was found at site 3 with 517 ind., followed by site 5 with 259 ind.

In this study, the fish community structure in Yudeung Stream was very similar to the previous results. Empirical studies suggested that the fish community in Yudeung Stream are dominated by Cyprinidae, and that higher relative richness of Z. platypus, Carassius auratus, and P. herzi (An et al., 2005; Lee, 2001). An et al. (2005) suggested that tolerant species such as Z. platypus and C. auratus were frequently observed in most of streams because of their high resistance to water pollution. In this study, we found that Z. platypus's dominance was similar to the previous results, but Z. koreanus was more abundance than C. auratus. Although the Z. koreanus is a sensitive species, they were mainly distributed as high density at sampling points (site 2 to 7) located upstream section of Yudeung Stream. The prevous studies also suggested that sensitive fish species, such as Rhynchocypris oxycephalus and Zacco temminckii, were frequently observed in the upper parts than the downstream. In contrast, fish species, such as Lepomis macrochirus and Micropterus salmoides, can also be distributed in downstream because they are resistant to water pollution. From the different distribution characteristics of fish according to these types of sensitivity characteristics, we surmise that the environmental gradient (i.e. water quality and habitat environment) from upstream to downstream of Yudeung Stream has changed little from the past (the early 2000s) to the present.

We fitted the 17 dominant fish species to NMDS ordination axes and selected seven factors (one environmental variable and six microhabitat patch types) that were significantly correlated (P<0.05; Fig. 5). The results of NMSD analysis showed that Gnathopogon strigatus, Squalidus gracilis majimae, Z. koreanus, and Z. platypus were associated with riffle and boulder areas, but Iksookimia koreensis, Acheilognathus koreensis, C. splendidus, Sarcocheilichthys nigripinnis morii, and Odontobutis interrupta were more associated with shallow areas. These species were frequently found in the upstream sections of Yudeung Stream (sites 1-9), and the various microhabitat characteristics present in the upstream sections led to different spatial distribution of these fish species. We surmised that Gnathopogon straggagatus, S. gracilis mazimae, Z. Koreanus, Z. platypus are mainly distributed in riffle and boulder areas, because they generally prefer microhabitat with high water velocity and high concentration of DO (Fu et al., 2015; Jang et al., 2008). Rapidly flowing streams tend to have greater atmospheric gas exchange and therefore greater DO (Allan & Castillo, 2007). Also, although fish species such as A. koreensis, C. splendidus, and S. nigripinnis morii have been found in shallow areas, these species do not avoid the habitat environment in riffle and boulder areas. The shallow area is distributed around the riffle and boulder areas, and supported by high concentrations of DO. Therefore, their distribution may be a transient distribution pattern due to the characteristics of fish with strong mobility and may also be related to feeding activity. The food items (e.g., invertebrates) were more abundant in shallow areas with lower water flow rates than riffle areas where water velocity are rapid (Meschiatti et al., 2000). In addition, shallow areas can be highly covered by aquatic macrophytes, which can be used as a habitat for various invertebrates (Williams et al., 2004).

In contrast, Cyprinus carpio, C. auratus, L. macrochirus, and Mi. salmoides were found in the downstream section (sites 14 and 15), associated with pool, sand/clay, and vegetated areas. Cyprinus carpio and C. auratus have relatively large bodies and poor swimming ability, and are not well-suited to rapidly flowing or shallow environments in upstream areas. They are mainly distributed in downstream sections, as well as in areas with limited flow such as wetlands and reservoirs (Hussain et al., 2015). Their distribution is broad due to strong environmental tolerances, such as resistance to drought and pollution (Koehn, 2004; Sun & Liang, 2004). Lepomis macrochirus and M. salmoides are introduced predators that prey indiscriminately on native fish species, zooplankton, and invertebrates (Weyl et al., 2010; Wilson et al., 2011), resulting in high density and exclusion of other species. Due to this, they are managed as ecological disturbance species in South Korea. Interestingly, adult M. salmoides prey on small L. macrochirus, which spends time in vegetated areas to avoid predation (Hossain et al., 2013). The leaves and stems of aquatic macrophytes increase the complexity of habitat structures in the water, and small fish utilize such refugia to avoid predators (Casatti, 2005; Sass et al., 2006). We observed both species at the same sites despite their predator–prey interactions; this suggests that aquatic macrophytes are indirect determinants of M. salmoides habitats. Other small species such as C. carpio and C. auratus also inhabit vegetated areas to avoid M. salmoides predation (Choi & Kim, 2020b; Jin et al., 2019). Therefore, we suggest that aquatic macrophytes are important habitats for fish distributed in downstream areas and their area determines the population size of prey species.

Rhinogobius brunneus, Coreoperca herzi, and P. herzi were mainly found at midstream sites with fluctuating environmental variables and habitat patch types, and their presence did not appear to be determined by key habitat characteristics. They have been previously reported as having a relatively wide range of distribution (Kim et al., 2016). From these points, we considered that these fish species have a broader distribution that does not rely on special habitat characteristics.

Stream management based on fish community distribution

Most of Korea’s streams have been damaged or altered by weirs and embankments, and human activities have increased the inflow of pollutants (Woo et al., 2005). Structures built in the stream sections particularly affect variables such as water flow and slope of habitats, leading to increased nutrient concentration and reduced DO, thereby affecting various organisms (Allan & Castillo, 2007). Fishes respond rapidly to these environmental changes compared with other aquatic organisms and can be used as an 'indicator species' to determine the disturbance gradient in stream ecosystems. In particular, tolerant fish species are sensitive to disturbances such as changes in water quality, so they can continuously monitor the stream environment based on their spatial distribution.

Our results have not found much difference in comparison with the fish communty in Yudeung Stream, from the past data of about two decades ago. Yudeung Stream are supported by distinct gradient of water quality and fish community structre from upstream to downstream from the past to the present. We assumed that through the distribution of sensitive species such as Z. koreanus, the upstream area of Yudeung Stream was from site 1 to 7. The fish communty structure changed rapidly after site 7 of Yudeung Stream. we considered that this change may be caused by physical and chemical disturbances by humans. From the site 7 of Yudeung Stream, the surrounding land cover mainly changes to the urbanized area. This means that not only water pollution caused by inflow of phosphorus and nitrogen, but also physical disturbances such as fishing will occure. From such finding, we speculate that the mid-lower reaches of Yudeung Stream have been continuously affected by human disturbance since the past. Therefore, we recommend minimizing the anthropogenic disturbance in the mid-lower of Yudeung Stream to maintain the typical environmental characteristics of mid-downstream.