TN, TDN and TPN were measured at St.0 around the river mouth from 2007 to 2009 in the Rausu River basin (Fig. 2). The nitrogen concentration of filtered and non-filtered water samples were analyzed by the cadmium reduction-colorimetric method. Annual TN and annual TDN exports to the ocean were evaluated using the river discharge at St.0 with TDN-discharge and TPN-discharge curves. The TDN-discharge and TPN-discharge curves were produced using ten different peak discharge floods and base flow discharges. As river discharge was not measured during the winter season from January to March, a storage function method was applied to estimate river discharge from 2008 to 2012 (Michael, 1978; Michael et al., 1979). The validity of the storage function method was confirmed through comparison with the observed river discharge from April to December.

To evaluate the contribution of salmon to soil organic matter (SOM), salmon runs were investigated in the Rausu River. Salmon were caught at the river mouth for artificial incubation and release, providing an estimate of the number of salmon caught by the apparatus (Hokkaido National Fisheries Research Institute, Fisheries Research Agency, 2009). The apparatus for catching salmon consisted of lattice fence, which does not obstruct flood flow or completely block the runs of salmon. Therefore, it was necessary to quantify the capture rate of the apparatus in order to estimate the actual volume of salmon runs. Field observations were conducted in the Tokorohoronai River, which is located in the same region of Hokkaido and has a custom to remove its apparatus before and after salmon run seasons, allowing us to monitor the salmon escapement from the apparatus and the salmon run under the open condition at the same place. The capture rate of the apparatus was calculated with numbers of salmons passing the observation point which has a channel section of 3 m in width and 0.2 m in depth, instead of the Rausu River because its river width (about 15 m) is too wide to cover the entire width. We used two infrared cameras (SM-AVIR-602S, Hero Corp., Izumo, Japan) placed 2 m above the river surface and recorded videos in all day to monitor the individual salmon passing this 3 m section. Videos were taken from the 25 to 28 November (before removal of the apparatus) and from the 4 to 7 December (after removal of the apparatus) in 2013. The number of salmon runs was calculated as the differences of the numbers of individuals running to upstream and those to downstream at the observation point. No salmon were captured and tagged for individual identification. There was no influence of rainfall during the observation period.

MDN, such as nitrogen, are generally supplied from the ocean to surface soils in a river basin as SOM, which includes feces of mammals, droppings of birds, and the remains of salmon preyed upon by mammals, birds and insects. To focus on the influence of SOM on particulate nitrogen in the river basin soils, soil particles with diameter of less than 500 µm after rinsing in 1N-HCL solution were used in the analysis. Therefore, it cannot be allowed to evaluate how much TN is exported from the river basin to the ocean. However, TPN export from an entire river basin is revealed to be larger than TDN in the Rausu River basin due to the steep slope (Nakayama et al., 2011). In general, some proportion of the nitrogen is reduced due to denitrification, which indicates the increase in δ15N of soil (Yamada et al., 1996). Wada et al. (1984) demonstrated that denitrification seems to have a small effect on the variation of δ15N of SOM under aerobic conditions close to the groundsurface in the natural forest. Rennie et al. (1976) revealed that isotope ratio of nitrogen is identical with organic nitrogen in the natural forest, which suggests that denitrification does not involve any isotope fractionations. Mckinley et al. (2013) also demonstrates that δ15N of groundsurface soil is in aerobic conditions in forests when water table is not close to the groundsurface. Since our sampling was carried out within 5 cm thickness from the groundsurface and groundsuface soil is not saturated due to the steep slope, sampled SOM is considered to be under aerobic conditions. Therefore, we made an attempt to estimate the contribution of MDN to SOM as a sequel to an accumulation, which directly corresponds to the suspended sediments transporting particulate nutrient through a river to the ocean, by sampling surface soils across the Rausu River basin (Fig. 2). Surface soil samples were taken at 12, 20 and 21 stations in 2008, 2009 and 2012, respectively. In 2008, fewer samples were taken as we did not have permission to sample surface soils in special protection zones. Surface soils were sampled from three different points at each station in a volume of 15cm×15cm×5cm (height × width × depth). Surface soil sampling stations in 2012 are shown in Fig. 2. Since previous studies have revealed that surface soil transport is related to the spatial distribution of surface soil type, land-use type and vegetation (Ishida et al., 2010), the location of each sampling station was selected by dividing the river basin into 21 domains (sub-basin areas) that vary in soil type and vegetation. The spatial distribution of surface soil type is divided into 6 categories. Although the spatial pattern in vegetation is complicated, the vegetation can generally be categorized in terms of altitude. Since Shiretoko is protected as natural World Heritage area, all areas studied are classified as forest and have high vegetation cover. Stable isotope ratios of carbon (δ13C) and nitrogen (δ15N) were measured using a Delta Plus Advantage mass spectrometer (Thermo Electron) coupled with an elemental analyzer (Flash EA 1112, Thermo Electron) at the Port and Airport Research Institute, Japan (Table 1 for δ13C and δ15N of SOM). Stable isotope ratios are expressed in δ notation as the deviation from standards in parts per thousand (‰) according to the following equation: δ13C,δ15N=[Rsample/Rstandard-1] where R=13C/12C or 15N/14N.

Vienna Pee Dee Belemnite and atmospheric nitrogen were used as the isotope standards of carbon and nitrogen, respectively. The analytical precision in the mass spectrometer system based on the standard deviation of the internal reference (L-histidine) replicates was <0.15‰ for both δ13C and δ15N. The contribution of MDN to SOM in surface soils was evaluated by applying a two source mixing model based on stable isotope analysis (SIA) of carbon (δ13C) and nitrogen (δ15N) (Kline et al., 1998; Moore and Semmens, 2008; Hossler and Bauer, 2012). Three soil samples were collected at each sampling station in order to account for small scale variability in SOM (Fig. 2 and Table 1). Salmon tissue isotopes were considered representative of the isotope composition of ocean productivity. To isotopically characterize terrestrial productivity, we considered one terrestrial end-members (sources): Soil Samples exhibiting the Lowest values of δ13C and δ15N (hereafter SSL), and thus assumed to have the highest terrestrial contribution to SOM. SSL was collected close to the top of the mountain, where MDN is not expected to influence isotope values. Representative soil samples collected in the same river basin were chosen because they have isotopically similar characteristics to the target soil samples in this study.

The contribution of MDN to SOM was evaluated using a two sources mixing model based on the measured δ13C and δ15N. The average contribution in the Rausu River basin was computed using each sub-basin area obtained from the Thiessen method. fC_MDN+fC_LDN=1fC_MDNδ13Csalmon+fC_LDNδ13CSSL=δ13CsoilfN_MDN+fN_LDN=1fN_MDNδ15Nsalmon+fN_LDNδ15NSSL=δ15Nsoil where fC_MDN and fC_LDN are the contributions of MDN and land-derived nutrient (LDN) by carbon, δ13Csalmon, δ13CSSL and δ13Csoil are the stable isotope ratios of carbon for salmon, SSL and soil samples, respectively, fN_MDN and fN_LDN are the contributions of MDN and LDN by nitrogen, δ15Nsalmon, δ15NSSL and δ15Nsoil are the stable isotope ratios of nitrogen for salmon, SSL and soil samples, respectively.

As bamboo grass (Sasa senanensis) is the dominant species in the study area, bamboo grass was collected at 13 soil sampling points (St.1, St.2, St.3, St.4, St.7, St.8, St.10, St.11, St.12, St.13, St.14, St.17, and St.21). Furthermore, droppings of sea eagles (Haliaeetus spp.) and feces of brown bear (Ursus arctos), which are typical migratory mammals and birds in Shiretoko, were collected to investigate whether or not they include MDN and thus contribute to SOM. Samples of feces and droppings for SIA analysis offer a major advantage, i.e. no isotopic fractionation and thus ideal to use the stable isotope values as a MDN tracer (Fry, 2006). Chum salmon tissues and droppings of sea eagles were collected at the river mouth and feces of brown bear were collected at St.14. The samples were pre-treated by rinsing with chloroform-methanol solution (2 : 1) prior to SIA, to remove isotopically fractionated metabolites, such Metabolites in the samples were removed by urea and ammonium (Kuwae et al., 2008, 2012).

During 2007 to 2009 the concentration of TDN was observed to be constant, 0.090 mgL-1 (SD 0.022 mgL-1), regardless of the discharge in the Rausu River. In contrast, TPN was revealed to be a function of river discharge (r2=0.88; Eq. 6) (Fig. 3). TPN showed a strong correlation with suspended sediment (SS) concentrations, with SS concentration increasing with increasing river discharge (Fig. 3). TPN was modeled by using our field observation results, discharge and TPN as Eq. (). TPN=0.0032×Q1.771 where Q is the river discharge (m3s-1).

The validity of the storage function method model was confirmed using the observed river discharge from April to September of 2009, which resulted in a Coefficient of Determination (CoD) of 0.61. The reliability of the model has been shown to be high enough for the analysis of river discharge when the CoD is more than 0.6 (Dutta and Nakayama, 2010). Annual mean export of TDN, TPN and TN from 2008 to 2012 were 5210, 14 750 and 19 960 kgyr-1, respectively. Since the size of the Rausu River basin of Shiretoko is 32.5 km2, the annual mean exports of TDN, TPN and TN per unit catchment area were 160, 454 and 614 kgkm-2yr-1, respectively (Table 2). The average concentrations of TDN and TPN from 2008 to 2012 were 0.090 and 0.216 mgL-1, which agrees with a previous study at the site (Nakayama et al., 2011).

The average number of salmon passing the cameras in the Tokorohoronai River during the 4 days while the apparatus for catching salmon was present was 0.49 h-1. The average numbers for 4 days after the apparatus was removed from the river was 0.61 h-1, so the rate of capture of salmon by the apparatus (CS) was estimated as 20 %: (0.61–0.49)/0.61=0.20. Since the field observations were conducted at the end of November and the beginning of December after the peak of salmon runs, floods may damage the apparatus for catching salmon and the estimated capture rate, 20 %, may be underestimated. Therefore, we attempted to apply two different larger rates of capture of salmon, 50 and 80 %, in order to show the possible nutrient re-export from the ocean due to salmon runs.

In the Rausu River of Shiretoko, the annual average numbers of salmon caught by the apparatus at the river mouth were 3075 and 10 580 for chum and pink salmon, respectively, from 2001 to 2009. By assuming that all apparatuses have the same rate of capture, the potential for chum and pink salmon runs can be estimated as 15 375 and 52 900 (CS 20 %), 6150 and 21 160 (CS 50 %), and 3844 and 13 225 (CS 80 %), respectively. The average weight of chum and pink salmon are 3.3 and 2.0 kg,respectively (Makiguchi et al., 2007), with a nitrogen content of about 30.4 gkg-1 (Larkin and Slaney, 1997). Therefore, annual TN potentially transported by chum and pink salmon is estimated to be 1542 and 3216 kgyr-1 (CS 20 %), 617 and 1287 kgyr-1 (CS 50 %), and 386 and 804 kgyr-1 (CS 80 %), respectively. Finally, the annual TN transported by chum and pink salmon per unit catchment area can be estimated as 146 kgkm-2yr-1 (CS 20 %), 59 kgkm-2yr-1 (CS 50 %), and 37 kgkm-2yr-1 (CS 80 %), (SD 19 kgkm-2yr-1), which corresponds to the contribution of TN by salmon, 23.8 % (CS 20 %), 9.5 % (CS 50 %), and 6.0 % (CS 80 %), relative to the annual outflow of TN per unit area (considered to be 100 %) (Table 2).

Both δ13C and δ15N of SOM were lower than those of salmons (Fig. 4). Interestingly, SSL has almost the same value of the mean δ15N of bamboo grass, which may suggest that bamboo grass can be considered to be as LDN. The stable isotope ratios in sea eagle droppings and brown bear feces were higher than LDN, indicating that sea eagles and bears are also one of the transporter of MDN to SOM. In the case of multiple food sources, feces and droppings are likely to be enriched in relatively indigestible food sources, when compared with stomach contents or assimilated materials (Sponheimer et al., 2003; Kuwae et al., 2008). Therefore, in the present study, feces and droppings are likely to be enriched in LDN (e.g., plants) because LDN would be more indigestible than MDN (e.g., fishes). However, such an enrichment does not affect the qualitative investigation, i.e., whether or not feces and droppings include MDN and thus contribute to SOM. Since brown bears are previously thought to be the major terrestrial consumer of spawning salmon, they may impact re-export of nutrient from the ocean across the river basin, such as through release of MDN-rich urine and feces (Hilderbrand et al., 1999). Rennie et al. (1976) demonstrated that δ15N is associated with soil organic matter, which is given as leaf litters, droppings from birds, feces from animals and so on. Also, Wada et al. (1984) revealed that δ15N is almost identical with organic nitrogen in natural forests. Therefore, it is important to understand the influence of sea eagles and bears on nutrient-cycle system. However, from Fig. 4, we cannot quantify the relative contribution of sea eagles and bears to total MDN transport.

The isotopic composition of salmon as representative of oceanic δ15N and δ13C were 10.99 and -20.54, respectively. The δ15N and δ13C of SSL were -3.19 and -29.48, respectively. Therefore, the three year average estimate of the contribution of MDN to SOM for δ15N depending on the choice of terrestrial isotope values was obtained e.g. 22.9 % (SD 3.6 %) by using a two sources mixing model (Fig. 5). As the reference, the three year average estimate of the contribution of MDN to SOM for δ13C was 17.7 % (SD 1.1 %) (Fig. 5). Since annual export of TPN per unit area from the Rausu River basin to the ocean was 454 kgkm-2yr-1, annual re-export of TPN originally derived from the ocean is estimated to be 104 kgkm-2yr-1 (=454kgkm-2yr-1×22.9%) (SD 16kgkm-2yr-1=454kgkm-2yr-1×3.6%) based on the contribution of MDN to SOM (Fig. 5 and Table 2).

Wada et al. (1984) demonstrated that δ15N of SOM in the forest has significant variation in the surface soil, such as about -3 to -2‰ at Jumonji in Chichibu and at Mt. Shigayama, and about 1 to 5 ‰ at Memuro in the eastern Hokkaido. δ15N of about 1 to 5 ‰ at Memuro was obtained in the Hokkaido Agricultural Experimental Station, which is located 10 km from the center of Obihiro city where 150 000 people live. Therefore, the values of about 1 to 5 ‰ at Memuro are expected to include the influence of emission of anthropogenic nitrogen. δ15N of about -3 to -2‰ in surface soil at Jumonji and at Mt. Shigayama may support our assumption that the larger the δ15N is, the higher the contribution of MDN becomes. In order to confirm our assumption, we carried out the similar field observation at the Rusa River basin (Fig. 6). In the Rausu River, only a part of the area is registered as a special protection zone of the Natural World Heritage region, but the whole area of the Rusa River is covered by a special protection zone. The Rusa River basin is thus considered as more protected and natural area as the Natural World Heritage compared to the Rausu River. Therefore, the contribution of MDN is expected to be larger in the Rusa River basin compared to the Rausu River basin (Fig. 2). As a result, the spatial average of δ15N in the Rusa River basin was obtained as 1.1 ‰, which is 1.0 ‰ larger than the Rausu River basin. It may be thus suggested that the higher value of δ15N of SOM in surface soil is associated with the contribution of MDN. However, it should be noted that this value for MDN re-export is estimated without contribution of marine derived TDN and thus should be considered the minimum annual MDN re-export from the viewpoint of TN.