Samples were collected from the Kiso estuary which empties in to the Ise Bay of Japan. In the Kiso estuary, a series of groins has been installed along the right bank in the sections located at 12 - 25 river kilometer (R km) from the river mouth since the Meiji era, which has functioned to control the flood flow and to produce diverse landscapes. As the result, embayments have been formed between two adjacent groins due to sediment accumulation and vegetation invasion.

We set three stations along transverse direction in the subtidal zone of Kiso estuary as follows: St. Right (St. R), St. Middle (St. M) and St. Left (St. L) were respectively located at the right bank side (with groins), the thalweg zone and the left bank side (without groins) in the cross-section of 17 R km (Figure 1 and Figure 2). The details of the sampling designs are listed in Table 1. Potential food sources (SPOMs) were not only collected from 17 R km but also from its upstream (St. U1, U2, U3 and U4) and downstream (St. D1 and D2) area (Table 1).

The salinity, chlorophyll a concentration (Chl. a) were logged with salinometer, turbidity meter and water depth gauges (COMPACT-CT, -CLW and -TD, JFE Advantech Co., Ltd.). The data were gathered every 5 minutes during a series of tidal motion. The logged data (salinity and Chl. a) was pooled to get daily average values and to estimate its mean values in a periodical tidal duration (14 days). Substrate samples for analyzing organic contents and Corbicula bivalves were gathered with the Ekman-Birge grab (area = 0.0225 m²). The SPOM samples were collected by filtering a certain volume of water (25 μm mesh, 2000 L), sealed in a plastic bottle. These collected data are a part of our conventional works [7] .

Benthic organisms were separated from each of the substrate content. Individuals of Corbicula japonica and Corbicula leana were identified and were estimated their numbers and measured their sizes. Then we divided them into each class of the following three groups [7] as L size ≥ 15 mm, 7 mm ≤ M size ≤ 15 mm, and S size ≤ 7 mm. We estimated the density of each size of the bivalves in all stations. To measure the stable isotope ratios of carbon and nitrogen of these individuals, the tissues of them were dried and removed fat by according to the conventional method [8] , then were analyzed with the Flashy EA1112-DELTAV ADVANTAGE ConFloⅣ System (EA-IRMS) (Thermo Fisher Scientific Inc.).

Sampled SPOMs were divided into two parts. One was used to estimate their concentrations by ash free dry mass (AFDM), according to Biggs and Karloy [9] . They were separated from samples by GF/B glass-fiber filters (0.47 mm Whatman Janan Ltd., Tokyo), dried with thermostat at 105˚C for 24 hrs, and measured their dry mass. Their AFDM were estimated by lost contents with a muffle furnace (MPN-310, Shimadzu Rika Co., Ltd.) at 400˚C for 4 hrs. The other was pretreated and was measured stable isotope ratios of carbon and nitrogen [8] similar as the bivalve samples. The δ¹³C and δ¹⁵N are expressed by the standard δ notation as follows:

where X is either ¹³C or ¹⁵N, and R is ¹³C/¹²C for carbon and ¹⁵N/¹⁴N for nitrogen. Pee Dee Belemnite and atmospheric nitrogen were used as the isotope standards for carbon and nitrogen, respectively. Also, we sampled culms or leaves of typical terrestrial plants such as reeds and willows in each of the stations, to measure their stable isotope ratios. Moreover, the Cluster analysis (with Euclidean distance and Ward’s method) was conducted to categorize the SPOM variations with stable isotope ratios.

Student’s t test was conducted to determine whether there were significant differences in physical items between Sts. R and L. Two-way analyses of variance (ANOVA) were used to test the effects of the size (L, M and S) and transverse location (St. R, St. M and St. L) on the density of each of the bivalves. If there were any significant differences in each of the variables, the Scheffe’s F tests were run to confirm where the differences occurred between groups.

The IsoSource (ver. 1.3.1) software [10] was used to determine the percentage of contribution of each of the organic matters to the diet of bivalve using both δ¹³C and δ¹⁵N. All possible combinations of each source contribution (0% - 100%) were examined by small increments (1% - 2%). The incremental parameter was set to 1%, and the tolerance parameter was initially set at 0.1%. If the mixture isotope values were out of bounds, we increased the tolerance parameter by 0.1% until it reached [11] . In adopting these software, we set the hypothetical food source by giving presumptive 0.8‰ trophic shift in the δ¹³C and 3‰ shift in the δ¹³N [12] between potential food sources and the bivalves.

Because most of the features about physical environment conditions have already described in our previous paper [7] , we hereby introduce the items related to the diet conditions. Table 2 shows the concentrations of chlorophyll a, SPOM and salinity in each of the station. According to the Student’s t-test results, both of them showed significantly higher values in St. R compare to those in St. L (P < 0.05). The SPOM at St. R was also somehow higher than that at St. L. It indicated that habitat environment in St. R is considered to be relatively lentic while that in St. L is considered to be lotic, which are contrasting landscapes due to the series of groins.

Figure 3 shows the density distributions with the three size classes of the C. japonica (a) and leana (b) in each

of the stations, as they have been shown in our conventional work [7] . Two-way ANOVA results showed that neither size nor the locations significantly influenced density of the bivalves (P < 0.01). The total density of C. japonica showed lower value in St. R than the other stations. In contrast, C. leana showed much higher density in St. R. Additionally, the C. japonica individuals in St. R were completely composed by the M size ones, however, in Sts. M and L, those were mainly occupied by the S size one.

About the C. leana individuals, their M size ones were main components in each of the station. The density of both of the bivalves showed converse result along the transverse direction.

Figure 4 shows the δ¹³C and δ¹⁵N plots of SPOMs in all of the collected samplings. The δ¹³C ranged from −27.76‰ to 24.24‰, showed a systematic decreased tendency from seaward to landward except some of irregular point such as St. D2 (−27.15‰). In transverse direction, the δ¹³C in the thalweg area (Sts. M and U2), showed relatively higher values than with and without groins area of the same cross-section. In the previous study, marine POM showed significantly higher value of δ¹³C than river and terrestrial POMs [13] . They concluded that the values of δ¹³C were gradually decreasing from seaward to terrestrial side [13] . However, in the current study, considering the different terrain and vegetation distribution between the areas with and without groins, the δ¹³C values of SPOM did not change regularly following the salinity concentration as shown in the conventional report [13] . Additionally, the SPOM showed a wide range of δ¹⁵N values ranged from 1.70‰ to 5.01‰.

Table 3 shows the results of classification by using the cluster analysis with the carbon and nitrogen stable isotope ratios. The sampled SPOMs were categorized into 4 groups based on Euclidean distance. The group I displayed lower δ¹³C values (−27.15‰ - −26.8‰) and more enriched δ¹⁵N values (4.31‰ - 5.01‰) than the other groups. They were similar with the collected samples of reed, P. australis (δ¹³C = −27.29‰, δ15N = 4.95‰). It means that SPOM of group I might derive from detritus of autochthonous terrestrial vegetation. Both of two samples of the type II were collected from the stations near to the left bank without groins (Sts. L and U3), which showed intermediate values of δ¹³C (−26.34‰ - −25.70‰) and relatively lower values of δ¹⁵N (3.00‰ - 3.53‰). The group III was composed of the data from the area with groins (Sts. R and D1) and thalweg area (Sts. M and U2). Their higher salinity values (Table 2) and relatively enriched values of δ¹³C suggested that the type III was mainly affected by high tide intrusion. The δ¹³C-enriched marine organic matter

brought by the tidal current easily intruded or reserved in these. The type IV collected from the upper estuary (St. U4 in 24R km) showed the lowest values of δ¹³C (−27.76‰) and δ¹⁵N (1.7‰), which indicated that it was strongly influenced by terrestrial particulate organic matter (TPOM) [13] .

According to the result of these classifications, the SPOM collected from Sts. D2, L and U4 were selected as the representative end members of group of I, II and IV, respectively. Although Sts. R and M belonged to the same group III, the SPOMs collected from Sts. R or M were selected as the representative end members of type of III according to the proximity principle. In the food source analysis below, these end members are considered to be used for the estimation with the IsoSource [10] .

Figure 5 shows the δ¹³C and δ¹⁵N plots of muscle tissues of C. japonica and leana in each of the stations. The δ¹³C of C. japonica ranged from −26.7‰ to −25.81‰, while those of C. leana showed relatively higher values ranged from −26.2‰ to −25.0‰. The δ¹⁵N of C. japonica showed narrow range (5.94‰ to 7.00‰), while C. leana showed wider and relatively lower values (3.41‰ to 6.68‰). The characteristic of isotope values for these two bivalves suggested that C. japonica preferred the organic matter with depleted-δ¹³C and more enriched-δ¹⁵N SPOM as the food source compared to C. leana in this study area.

Figure 6(a) shows that the δ¹³C and δ¹⁵N plots of SPOM end members and C. japonica individuals with these size variations. Both of the hypothetical diets of each of the bivalves and the food source polygon with the end members are drawn in this figure. Their isotope signatures of hypothetical diets were quite different from those of SPOMs in the same location. It suggested that the neighbor SPOM cannot always be used as the essential food source for C. japonica in each of the locations. As the isotope signatures of bivalves decreasing from St. R, their hypothetical diets were closer to SPOM of St. D2 and St. U4 indicated the contributions of these two food sources increased from Sts. R to L. We also can infer that SPOMs of Sts. D2 and U4 are the necessary diets, because if one of the SPOM of Sts. D2 and U4 were absent, almost all of the isotope signatures of hypothetical diet

would outside the convex polygon boundary of the food sources [11] . The S size C. japonica individuals showed relatively lower δ¹⁵N values in Sts. M and L than the larger individuals, which implied that they might take relatively depleted-δ¹⁵N SPOM was the main diet.

Figure 6(b) shows that the δ¹³C and δ¹⁵N plots of SPOM end members and C. leana with size variations. Both of their hypothetical diets and food source polygon with the end members are drawn in the same figure. The δ¹³C and δ¹⁵N of L and M size individuals of C. leana found in with groin area (St. R) showed the most enriched values (δ¹³C = −25.0‰, δ¹⁵N = 6.68‰; δ¹³C = −25.1‰, δ¹⁵N = 6.60‰, respectively). The isotope signatures of hypothetical diet of L and M size individuals (Figure 6(b), ca. δ¹³C = −25.8‰, δ¹⁵N = 3.64‰) exhibited similar values to St. R (in situ), which suggests that the neighbor SPOM take largest contribution to their food composition. While the δ¹³C and δ¹⁵N of hypothetical diets for M size individuals in Sts. L and M were closer to the SPOMs in Sts.U4 and D2, which indicated they were the important food sources. Additionally, the isotope signatures of hypothetical diets for S size of C. leana displayed similar values of δ¹³C (ca. −26.3‰) in each of the stations, while the δ¹⁵N (0.4‰ - 2.80‰) was relatively lower than the larger individuals (3.40‰ - 3.68‰). The plots of hypothetical diets indicated that C. leana might shift the food sources depending on growth stages; we could not find the food sources for S size individuals in the present study because points of their isotope signatures were out of food sources of polygon.

Figure 7 shows that the food compositions (%) of the SPOM end members for C. japonica (a) and leana (b) in each of the stations. The M size of C. japonica, the SPOM from St. D2 accounted for the largest contribution of the food sources (40.4% - 66.6%), and followed by the SPOM from St. U4 (4.2% - 32.1%) in each of the station. The contributions of these two food sources showed an increased tendency from the area with groins (St. R) to that without groins (St. L). Great utilization of δ¹³C-depleted SPOM (SPOM of Sts. D2 and U4) in each of the stations indicated that the adult of C. japonica might be a selective filter feeder, mainly depended on TPOM. In the area with groins (St. R), water exchange was reduced due to the blocking effect of groins [14] , which limited the material transport between the area with groins (St. R) and thalweg zone (St. M). Consequently, their preferable food sources could be prevented into the semi-closed embayment area (St. R), while in the area without groins (St. L), water exchanging supplied the preferable food sources. Additionally, highest concentration of neighbor SPOM (5.51 mg/L, Table 2) corresponded to the lowest density of C. japonica bivalves (22.2 ind/m², Figure 3) in St. R; the lowest concentration of SPOM in Sts. M and L (1.15 and 0.71 mg/L respectively, Table 2) corresponded to the highest density of C. japonica bivalves (155.5 ind/m², 111.1 ind/m², Figure 3). It also proved that a large amount of δ¹³C-enriched SPOM in St. R might not be as an ideal food source for C. japonica. The SPOM which showed depleted δ¹³C value might be the potential food source for C. japonica such as terrestrial organic matter. These kinds of processes were also observed in the previous study [6] .

Food sources for the S size individuals of C. japonica in Sts. M and L showed relatively lower utilization of SPOM from St. D2 (ca.24%) and higher those from St. U4 (ca.39%), while M size individuals suggested that smaller size of C. japonica tended to digest and assimilate the food sources with relatively lower δ¹⁵N. It was supposed that the smaller size of C. japonica might have weak capacity in assimilating the SPOM derived from nitrogen-polluted water since increased δ¹⁵N value was found in excessive richness of nitrogen loaded water [12] .

Regarding the C. leana, in St. R, the L and M size individuals showed similar food source compositions, mainly depended on the neighbor SPOM which accounted for 60.5% and 53.2% respectively. The M size of C. leana in Sts. M and L showed similar food compositions, mainly depended on the SPOM from St. D2 (44.6%, 40.9% respectively) and St. U4 (28.4%, 29% respectively). The adult C. leana completely shifted food compositions between the area with groins (St. R) and the one without groins (St. L), which indicated that adult C. leana might be an opportunistic filter feeder. In the area with groins (St. R), dense Chl. a with neighbor abundant SPOM (11.72 μg/L, 5.51 mg/L, Table 2) were trapped in this kind of semi-closed area. In this situation, the utilization of the neighbor SPOM was higher. A large quantity of food sources in situ supported approximately 3 times of individuals in Sts. L or M (Figure 3). However, in the area without groins (St. L), due to the bare and relatively open terrain, the primary production might be lower. The Chl. a and SPOM (1.13 μg/L, 0.71 mg/L, Table 2) were insufficient and could not provide enough diet. Under this situation, individuals of C. leana decreased

compared with St. R (Figure 3). However, bare terrain also promoted water mixing process between river flow and intruded marine and made food sources diverse. The conventional study concluded that C. leana was a fresh water bivalve and selected TPOM as the food source [15] . However, in the present research, the adult C. leana was considered as an opportunistic filter feeder as δ¹³C of food source fluctuated from −27.76% to −24.24‰; δ¹⁵N ranged from 1.70‰ to 5.01‰.