Seasonal Variations of Dissolved Iron Concentration in Active Layer and Rivers in Permafrost Areas, Russian Far East

Dissolved iron (dFe) in boreal rivers may play an important role in primary production in high-latitude oceans. However, iron behavior in soils and dFe discharge mechanism from soil to the rivers are poorly understood. To better understand iron dynamics on the watershed scale, we observed the seasonal changes in dFe and Dissolved Organic Carbon (DOC) concentrations in the river as well as dFe concentration in soil pore waters in permafrost watershed from May to October. During snowmelt season, high dFe production (1.38–4.70 mg L−1) was observed in surface soil pore waters. Correspondingly, riverine dFe and DOC concentrations increased to 1.10 mg L−1 and 32.3 mg L−1, and both were the highest in the year. After spring floods, riverine dFe and DOC concentrations decreased to 0.15 mg L−1 and 7.62 mg L−1, and dFe concentration in surface soil pore waters also decreased to 0.20–1.28 mg L−1. In late July, riverine dFe and DOC concentrations increased to 0.33 mg L−1 and 23.6 mg L−1 in response to heavy rainfall. In August and September, considerable increases in dFe concentrations (2.00–6.90 mg L−1) were observed in subsurface soil pore waters, probably because infiltrated rainwater developed reducing conditions. This dFe production was confirmed widely in permafrost wetlands in valley areas. Overall, permafrost wetlands in valley areas are hotspots of dFe production and greatly contribute to dFe and DOC discharge to rivers, especially during snowmelt and rainy seasons.


Introduction
Iron is an essential trace element for the growth of all organisms. It plays an important role in in vivo metabolic processes, including photosynthesis, electron transfer, nitrate reduction, and nitrogen fixation [1]. Martin and Fitzwater [2] revealed that iron limits phytoplankton growth in a high-nutrient low-chlorophyll (HNLC) region where the reason for the relatively low phytoplankton productivity despite the abundance of nutrients having long been unclear. Subsequent studies confirmed that

Study Site
A field survey was conducted in the Tyrma region, which is approximately 270 km northwest of Khabarovsk in the Russian Far East. Mean annual air temperature is −1.96 • C and annual precipitation is 654.6 mm. The Tyrma region is situated in a sporadic permafrost area [22]. Three large rivers, namely, the Sutyri River, the Yaurin River, and the Gujal River, join the Tyrma River, and eventually join the Bureya River, which is one of the largest tributaries of the Amur River ( Figure 1). We fixed the Sofron River watershed (50 • 5 42.86 N,132 • 22 1.20" E), which is a tributary of the Gujal River, as the intensive experimental site (Figures 1b and 2a). The Sofron River is a second-order stream with a sub-catchment area of approximately 23 km 2 ( Figure 2). River width is approximately 10 m and central water depth is usually 10-20 cm at the sampling point without rain. Vegetation in the Tyrma region is roughly divided into two types: the forest areas on the ridges and hillslopes are characterized by spruces (Picea ajanensis) and white birches (Betula platyphylla), and the wetlands in the flat valleys are characterized by shrubs, such as bog blueberry (Vaccinium uliginosum), cowberry (Vaccinium vitis-idaea L.), and ledum (Ledum decumbens). In the wetlands, Sphagnum species widely exist on the ground surface and larches (Larix gmelinii var. gmelinii) are scattered. Topsoil layers in the forests and the wetlands are composed of peat soils. In particular, thick peat soil layers are formed in the wetlands due to long-term accumulation of sphagnum biomass. This type of wetland, called Mari, is a typical landscape in the flat valleys in the Tyrma region. Mari can be seen not only in the flat valleys, but also on the ridges and hillslopes with a gentle slope. In the Tyrma region, permafrost generally exists underneath Mari. The vegetation and permafrost distribution in the Sofron basin are consistent with these typical features as described above. (a) nine sampling points for soil pore waters (N1-N9) in the Sofron River and neighboring area. This map is colored by gradient (degree) to clearly show terrain characteristics. Area enclosed with a black line shows the Sofron watershed, and the white arrow denotes the sampling site of the Sofron River. This map was created by the authors based on a 30 m resolution digital elevation model (DEM) provided by the Japan Aerospace Exploration Agency (JAXA); (b) topographic profile through the sampling points for soil pore waters.

Soil Pore Water Sampling and Soil Characteristic Measurement along a Transect
In September 2016, we established a long transect of nine sampling points for pore waters from N1 near the Sofron River to N9 in the adjacent basin ( Figure 2). Permafrost wetland Mari is seen not only at N1-N3 and N9 in valley, but also at N7 on ridge and N8 on gentle hillslope. On the other hand, permafrost does not exist at N5 on steep hillslope and N6 on ridge. We dug 1.8 cm diameter holes using a manual drill (DAIKI, DIK-1721, Konosu city, Saitama, Japan) in the ground of each sampling point and installed ceramic cups with a polyvinyl chloride pipe at 20 cm and 40 cm depths for pore water sampling (N4 has different depths as described below and N5 has only 40 cm depth). In April 2017, we installed the ceramic pipes at 5 cm and 10 cm depths at N4 and 10 cm and 25 cm depths at N6 for soil pore water sampling during snowmelt season. Because 10 cm of snow remained at N4 and N6 on this installation, both points were suitable to observe dFe concentration variations in response to snow melting. The frozen ground at that time existed at 10-15 cm depth at N4 and N6; thus, we used an electric drill to install a ceramic cup at 25 cm depth at N6. All the soil pore water sampling depths at each point are compiled in Table 1. Table 1. Terrain type, sampling depth of soil pore waters, active layer thickness (ALT), and peat layer thickness (PLT) at nine sampling points of the research site in September 2016 (Tyrma, Khabarovsk, Russia). Pore waters at marked (*) depths at N4 and N6 were collected four times in May and twice a month from June to October. Other soil pore waters at nine sampling points were collected twice a month from June to October. At N4 and N6, the soil pore waters (except 20 cm and 40 cm at N6) were collected four times in May. From June, all pore waters at the nine sampling points were collected twice a month until October (see Table 1). The soil pore waters were sucked up overnight by 50 mL disposable syringes (TERUMO, SS-50ESZ, Shibuya ku, Tokyo, Japan) and filtered in situ through 0.45 µm disposable filters made of cellulose acetate (ADVANTEC, DISMIC 25CS045AS, Chiyoda ku, Tokyo, Japan). The filtered waters were preserved in acid-washed 50 mL polypropylene bottles and kept in a refrigerator at 4 • C until analysis.

Point
The peat layer thickness (PLT) and the active layer thickness (ALT) at the nine sampling points were measured in September 2016 by digging a pit down to the permafrost table and examining the soil profile (Table 1). To confirm the ALT data, we checked that the soil temperature of permafrost table was 0 • C. The absence of permafrost at N5 and N6 was judged from the absence of a continuous decline in soil temperatures over 10 cm intervals from the surface. Soils for the analysis of organic carbon content and moisture content were collected in polyethylene freezer bags from 0-10 cm, 10-20 cm, and 20-30 cm depths of the soil profile at the nine sampling points and stored in a freezer until analysis.
When we dug a soil profile at each point in September 2016, we installed geothermal loggers (Onset, U22-001, Bourne, Massachusetts, America) to observe the seasonal thaw dynamics of soil at two depths at N1 (10 cm and 25 cm), three depths at N4 (10 cm, 25 cm, and 50 cm), and three depths at N6 (10 cm, 25 cm, and 50 cm). The soil temperatures were recorded until October 2017.

River Water Sampling and Hydrological Observation
During snowmelt season from late April to mid May, the water samples were collected from the Sofron River twice a week. From June to October, the water samples were collected twice in a month on the same days as soil pore waters. Two hundred milliliters of water was sampled using a disposable syringe (TERUMO, SS-50ESZ, Shibuya ku, Tokyo, Japan) and immediately filtered through 0.45 µm disposable filters made of cellulose acetate (ADVANTEC, DISMIC 25CS045AS, Chiyoda ku, Tokyo, Japan). One hundred milliliters of the filtered water was preserved in an acid-washed propylene bottle for dFe measurement and the other 100 mL was preserved in a propylene bottle for dissolved organic carbon (DOC) measurement, and both were kept in a refrigerator until analysis.
For the observation of water level of the Sofron River, we installed a water pressure logger (Onset, HOBO U-20-001-04, Bourne, Massachusetts, America) at the river bed of the water sampling point and an air pressure logger (Onset, HOBO U-20-001-04, Bourne, Massachusetts, America) nearby in mid June 2017 when the river bed adequately thawed. In addition, daily precipitation data in the Tyrma region from June to October were obtained from the weather site in Russia (http://ru8.rp5.ru/ Weather_archive_in_Tyrma_(Sutyr)).
Water samples were also collected from the Sutyri River (catchment area: 2129 km 2 ), the Tyrma River (6168 km 2 ), and the Yaurin River (3175 km 2 ) ( Figure 1b). The frequency, date, and way of sampling were same as those for the Sofron River.

Chemical Analyses
All the analyses were done in the Institute of Mining in Khabarovsk. The dFe concentrations were measured for soil pore waters and river waters. Prior to Fe analysis, the samples were acidified (~pH 2) with HNO 3 , and Fe concentration was determined with Agilent 7500cx Inductively Coupled Plasma-Mass Spectrometer (ICP-MS) using helium modes to diminish the interferences. The detection limit for Fe by ICP-MS was 10 ng L −1 , and an uncertainty was 5-10%. Standard solutions for Fe analysis were prepared by using Environmental Calibration Standard (Agilent Technologies). In this paper, we define dFe as Fe that passed through 0.45 µm disposable filter in samples. Moreover, for the river water samples, DOC concentrations were also determined with a TOC analyzer (SHIMADZU TOC-LCSH) using the catalytic combustion oxidation method. The detection limit for TOC by TOC analyzer was 0.1 mg L −1 , and an uncertainty was less than 1.5%. Standard solutions for DOC analysis were prepared by using Potassium Hydrogen Phthalate (C 6 H 4 (COOK)(COOH)) (Nacalai tesque). For the data of DOC and dFe concentrations in the Sofron River, simple liner regression analysis and Student's t-test were performed.
Soil organic carbon content was measured by Tyurin's method (wet combustion) based on Bel'chikova (1975) [34]. Weight moisture content was determined from the difference in mass before and after drying soil.

Seasonal Variations in dFe and DOC Concentrations in Sofron River and Large Rivers
The dFe concentration in the Sofron River showed substantial seasonal variation, ranging from 0.10 mg L −1 to 1.10 mg L −1 (Figure 3a). During snowmelt season, dFe concentration increased rapidly within two weeks from 0.27 mg L −1 to the annual highest value of 1.10 mg L −1 . Although it once dropped through June, a relatively high dFe concentration was recorded again in late July and late September. The annual lowest level of 0.10 mg L −1 was noted in October, immediately before soil freezing. DOC concentration in the Sofron River showed a similar seasonal variation to dFe concentration. DOC concentration rapidly increased during snowmelt season from 17.0 mg L −1 to the annual highest value of 32.3 mg L −1 . After that, it increased again in late July and late September, when dFe concentration also increased.
The water level of the Sofron River showed frequent increases accompanied with heavy rainfall of over 10 mm from July to mid August ( Figure 3b). The highest water level of 36 cm was recorded in early August. The increases in dFe and DOC concentrations in late July and late September were associated with the water level risings, that is, the seasonal variations in dFe and DOC concentrations almost synchronized with the change in water level throughout the year. Whereas dFe concentration greatly increased from 0.15 mg L −1 to 1.07 mg L −1 during the period affected by snow melting, the degrees of increases associated with heavy rainfall in late July and late September were relatively small. In the separated periods before and after July, two distinct positive correlations were found between dFe concentration and DOC concentration (r 2 = 0.51, p = 0.056 from April to June, and r 2 = 0.75, p < 0.01 from July to October) (Figure 3c), suggesting a difference in the discharge mechanisms of these two substances among seasons.  The seasonal variations in dFe and DOC concentrations in the large rivers (Sutyri, Tyrma, and Yaurin) were almost consistent with those in the Sofron River throughout the year (Figure 4a,b). During snowmelt season, similar large increases in dFe and DOC concentrations were also observed. The increases in DOC concentrations corresponding to rainfall as seen in the Sofron River were also observed in the large rivers in late July and late September. Although the increases in dFe concentrations during these storm events were not clearly observed in the large rivers, these concentration levels matched those of the Sofron River. In a rainless season (from June to early July and from September to October), however, it appeared that dFe concentrations in the large rivers were generally higher than those in the Sofron River. This difference between the Sofron River and the large rivers was more clearly seen in the seasonal change in dFe/DOC ratio (Figure 4c).

Organic Carbon Contents and Weight Moisture Contents of Soils along a Transect
Organic carbon contents and weight moisture contents of soils at the nine sampling points in the research site are shown in Table 2. The organic carbon contents at the uppermost layer at points located in valley (N1-3 and N9) were 144-478 gC kg −1 , whereas those at points located on hillslope and ridge (N4-8) were 48-358 gC kg −1 , slightly lower than those at valley points. Interestingly, permafrost wetlands Mari in valley had high organic carbon contents of 100-474 gC kg −1 even at 20-30 cm depths, whereas Mari on ridge (N7) and gentle hillslope (N8) had low organic carbon contents of 52 and 23 gC kg −1 , respectively. Although N5 on steep hillslope and N6 on ridge are both no-permafrost points, the organic carbon contents were relatively high at N6 and low at N5. The weight moisture contents showed similar spatial variations to the organic carbon contents. Weight moisture contents were remarkably high, i.e., 92.7-99.8%, at the uppermost layer at points located in valley (N1-3 and N9), but were relatively low, i.e., 49.4-92.4% on hillslope and ridge (N4-8).
In short, wetlands Mari in valley had the highest organic carbon and moisture contents within the watershed. However, Mari on ridge (N7) and gentle hillslope (N8) had relatively low weight moisture contents of 49.4% and 68.1% even at the uppermost layer relative to Mari in valley points. At hillslope and ridge points, weight moisture contents decreased significantly with depths. At N6 that had contained relatively high organic carbon contents among hillslope and ridge points, weight moisture contents were also high.

Seasonal Soil Thaw Dynamics in Different Terrain Types
The seasonal variations in soil temperature at several depths at N1, N4, and N6 are shown in Figure 5. In addition, the thawing date for each depth and the thawing rate, which is calculated simply by dividing the difference in depth by the number of required days for thawing, are shown in Table 3. The thawing of 10 cm depth at N1, which is the closest point to the river in valley, was completed on 24 May, approximately one month later than that at N4 on the boundary between valley and hillslope and that at N6 on ridge. The 25 cm depth at N1 thawed on 15 June, which was also the latest among the same depths.
The 10 cm and 25 cm depths at N4 and N6 showed remarkable rises of soil temperatures after thawing. Soil temperatures as of July 1 for 10 cm and 25 cm were 9.1 • C and 4.2 • C at N4 and 8.5 • C and 3.3 • C at N6, which were considerably higher than those at N1 of 2.4 • C and 0.6 • C. At N6, where permafrost does not exist, soil temperatures always tended to be higher than those at N1 and N4. In fact, the annual highest soil temperatures at N6 (10 cm: 13.1 • C, 25 cm: 9.6 • C, 50 cm: 7.9 • C) were the highest compared with those at N1 (10 cm: 10.9 • C, 25 cm: 6.9 • C) and N4 (10 cm: 11.5 • C, 25 cm: 8.2 • C, 50 cm: 5.2 • C). Moreover, the thawing rate from 25 cm to 50 cm depth at N6 was 1.6 times faster than that for the same depth at N4. Therefore, the seasonal downward thawing of soil from spring to summer was fastest at N6 on ridge, followed by N4 on boundary between valley and hillslope, and the slowest at N1 in valley. This order agreed with the difference in active layer thickness at those points (N1: 64 cm, N4: 102 cm, N6: No permafrost) ( Table 1). Combined with organic carbon content data and weight moisture content data ( Table 2), permafrost wetland Mari in valley area is characterized by an environment with the highest organic matter content and the highest moisture content, and the lowest soil temperature within the watershed.

Seasonal Changes in dFe Concentrations in Soil Pore Waters along a Transect
Seasonal changes in dFe concentration in soil pore waters at N4 and N6, the observations of which only started from snowmelt season, are shown in Figure 6. In early May, when snowmelt was in progress, dFe concentrations in the surface soil pore waters were 1.38-4.70 mg L −1 , clearly higher than those in summer and autumn. These remarkably high dFe concentrations in the soil pore waters were consistent with the increases in dFe concentration in the Sofron River and the large rivers during snowmelt (Figure 4a). From spring to summer, dFe concentrations at N4 (5 cm and 10 cm depths) and N6 (10 cm and 25 cm depths) decreased. Seasonal changes in dFe concentrations in soil pore waters at the transect points where the observation started after snowmelt season are shown in Figure 7. The sampling depths at these points were 20 cm and 40 cm, which were larger than the depths at N4 and N6 observed in snowmelt season. dFe concentrations for both depths at each point except N3, were less than 1.00 mg L −1 for two months from the start of observation. However, from August to September, dFe concentrations at 20 cm and 40 cm depths in permafrost wetland Mari (N1-3 and N7-9) increased rapidly to 2.00-6.90 mg L −1 . In contrast, these increases in dFe concentrations were not observed at N6, where permafrost does not exist. According to the changes in soil temperature ( Figure 5), it appears that the seasonal downward thawing of soil from 20 cm to 40 cm took place in May on ridge and hillslope and in June in valley. Thus, dFe production in these depths occurred two or three months after thawing. After that, these dFe concentrations in the soil pore waters decreased to less than 1.00 mg L −1 in October, when soil started to freeze from the top layer.

dFe Production in Uppermost Soils and Its Discharge into Rivers during Snowmelt Season
From late April to mid May, when snow was melting, dFe and DOC concentrations in the Sofron River rapidly increased, reaching the highest concentration in the year (Figure 3a). In accordance with this, surface soil pore waters also showed the highest dFe concentrations in this period ( Figure 6). These elevated dFe concentrations in the Sofron River and the soil pore waters can be explained by snow melting and the existence of frozen ground at shallow depths. According to the seasonal changes in soil temperature ( Figure 5), frozen ground existed at approximately 0-10 cm depth in valley and 10-25 cm depth in hillslope and ridge. It is known that frozen ground impedes downward infiltration of snowmelt water [35,36], resulting in a waterlogged condition near the soil surface. This was actually the case at N4 and N6 where soil pore waters were collected during this period. In soil saturated with snowmelt water, anoxic conditions could develop through dissolved oxygen consumption by microbial respiration [37,38]. Under such conditions, Fe(III) can be utilized as an electron acceptor by anaerobic microorganisms, and produced Fe(II) is probably complexed with DOC that seeps from ambient organic-rich soils ( Table 2). Although we did not observe DOC concentrations in surface soil pore waters, the positive correlation between DOC and dFe concentrations in the Sofron River during snowmelt season (Figure 3c) suggests that exported dFe is mainly associated with organic compounds, as reported in other boreal regions [39][40][41][42]. The large amount of organically complexed iron produced in this way should be responsible for the high dFe concentrations in the uppermost soil pore waters ( Figure 6) and the Sofron River (Figure 3a).
During snowmelt season, elevated dFe concentrations in surface soil pore waters were observed not only at N4 located on the boundary between valley and hillslope but also at N6 located on the ridge ( Figure 6). In hillslope and ridge without permafrost, soil water usually infiltrates downward easily, and thus the uppermost soil is unlikely to be saturated. However, the seasonally frozen ground at a shallow depth would have led to waterlogging and dFe production in soil pore waters even in these terrains. It is thus expected that dFe production in waterlogged surface soils in this period may occur over a wide range of areas in these terrains where soil freezes and snow falls in winter.
However, in case frozen ground is formed with low water content in winter, snowmelt water may be able to infiltrate it because of discontinuous frozen soil [43,44]. It is therefore likely that, on ridge and hillslope, where the water content before freezing was relatively low (Table 2), snowmelt water probably flowed out while infiltrating frozen ground. Thus, even if snowmelt water on ridge and hillslope once had high dFe concentration, its contribution to the increase in dFe concentration in rivers would be minor. In contrast, in the valley, where the surface soils retained high water content before freezing (Table 2), snowmelt water would have flowed into the rivers over continuous frozen ground without infiltration. Consequently, the valley area most likely contributed to the rapid increase in dFe concentration in the rivers during the snowmelt season.
The rapid increases in dFe and DOC concentrations were also observed in the large rivers during snowmelt season (Figure 4a,b). It is thus suggested that the small watershed dominated by permafrost wetlands Mari like the Sofron river greatly contributes to the terrestrial DOC and dFe transport in this period. Moreover, both DOC and dFe concentrations in the large rivers during snowmelt season had the highest values in the year, similarly to those in the Sofron River, strongly indicating that snowmelt season is the most important period for transporting large amounts of DOC and dFe. Elevated dFe concentrations during snowmelt season were also observed in large tributaries of the Amur River [45,46], and other boreal rivers [39][40][41]47]. To our knowledge, the present study is the first to corroborate this large-scale dFe transport that resulted from the high dFe concentrations in the uppermost soil pore waters.

Seasonal Changes in Iron Behavior in Thawed Active Layer after Snowmelt Season
The dFe concentrations in surface soil pore waters at N4 and N6 declined after snowmelt season ( Figure 6). As soil thawed downward towards summer, the surface soils most likely changed gradually to oxygenated conditions because of the decline of groundwater level. Such a situation that produces Fe(II) in pore waters should be oxidized chemically and/or biologically to insoluble Fe oxides [48], possibly resulting in the declines in dFe concentrations. The early declines in dFe concentrations of soil pore waters at N6 relative to that at N4 are explainable by the faster downward thawing of soil at N6 than N4 from spring to summer ( Figure 5 and Table 2).
The dFe concentrations at 20 cm and 40 cm depths at the transect points, where observation was started in June, were less than 1 mg L −1 until August except N3, but increased markedly to 2-7 mg L −1 from August to September (Figure 7). This abrupt increase in dFe concentration would be associated with heavy rainfall from July to August (Figure 3b). The total amount of rainfall in July and August was 298.4 mm, which is equivalent to 45.6% of the annual average precipitation of 654.6 mm. Abundant rainfall would make subsurface (20-40 cm) soils saturated as permafrost would prevent infiltration of soil water. Under this situation, such electron acceptors as NO 3 − , Mn 4+ , and humic substances would be utilized by being utilized for microbial anaerobic respiration [30,[49][50][51][52]. As a result, microbial iron reduction was likely activated from August to September and large amounts of Fe(II) were produced in soil pore waters at 20 cm and 40 cm depths. Street et al. [53] observed that the redox potential (Eh) of soil in the active layer did not change markedly for one month after thawing but started to decline after the rainy season in summer, in agreement with our interpretation. Moreover, it was reported that iron reduction was facilitated in peat soils with artificial flooding relative to soils without this treatment [29]. It is thus likely that dFe production in soil prevails widely in permafrost wetland Mari after the rainy season. Only N6 did not show the increase in dFe concentrations in late summer, probably because the thawing of seasonally frozen soil at N6 was faster than at the other points in permafrost wetland Mari, leading to infiltration of soil water and an unsaturated condition even after abundant rainfall.
The elevated dFe concentrations in late summer subsequently decreased to approximately 1 mg L −1 towards October. The possible reason for this would be the decline of groundwater level in soils of the active layer in September when there was little heavy rainfall. Soil conditions probably changed to oxidative during this rainless season, and most of the produced Fe(II) in soil pore waters should have been oxidized to insoluble Fe oxyhydroxides by incoming oxygen and/or iron-oxidizing bacteria [54,55]. It was reported that reducible Fe oxyhydrox-ides exist in abundance in soils where oxidizing and reducing conditions are repeated [28,56], and the newly formed Fe oxyhydroxides will be reduced again by iron-reducing bacteria when reducing conditions are developed.

Seasonal Changes in dFe Discharge Mechanism Associated with Downward Shift of Flow Path after Snowmelt Season
After snowmelt season, groundwater table most likely declined as frozen ground thawed downward with rising air temperature. With this hydrological change from spring to summer, both dFe and DOC concentrations in the Sofron River and the large rivers decreased (Figure 4a,b). This agrees with the decline in dFe concentration in surface soil pore waters after snowmelt season ( Figure 6). In addition, because the riverine dFe/DOC ratio also decreased in this period (Figure 4c), the decreases in dFe concentrations in the rivers after snowmelt season probably implies the decrease in discharge of organically complexed iron from the peat soil layer with the downward shift of flow path. This is supported by previous studies reporting that the discharge of DOC derived from humic substances decreased from spring to summer in the boreal region [57][58][59].
In the Sofron River, increases in dFe and DOC concentrations were observed in response to rising water levels due to heavy rainfall in late July and late September (Figure 3a,b). The elevated DOC concentrations were as high as those during snowmelt season, but the degree of these increases in dFe concentrations was less. Although storm events often increase dFe discharge from the upper organic-rich soil in boreal region [60], this result emphasizes the importance of snowmelt flooding to terrestrial dFe transport. The most likely explanation for the smaller increases in riverine dFe concentration due to heavy rainfall compared to snowmelt season is short period of waterlogged conditions in surface soils. It was reported that a high water level after heavy rainfall (~10 mm) lasts for only 1-2 days in permafrost peatland [61], whereas the waterlogged conditions due to snowmelt lasted for at least two weeks in this study. It is therefore suggested that summer rainfall causes the flushing of dFe into river, but, unlike the snowmelt season, it does not contribute to the develop of reducing conditions in surface peat soils. As mentioned in Section 4.2, the summer heavy rainfall probably contributed to the large amount of dFe production in subsurface soil pore waters in late summer ( Figure 7), but, nevertheless, dFe concentration in the Sofron River hardly increased (Figure 3a). This finding is consistent with that of Street et al. [53], i.e., deep soil methane concentration in the active layer increased over time through the season but did not appear to influence stream methane concentration. Possible interpretation for the luck of increase in dFe concentration in the Sofron river in late summer would be a low discharge rate of subsurface soils. According to Quinton et al. [62], the hydraulic conductivity of peat soils below 20 cm depth is approximately less than one tenth of surface peat soils; therefore, even though a large amount of dFe is produced in subsurface soil pore waters, there might be little immediate influence on riverine dFe concentration.
Seasonal changes in dFe and DOC concentrations in the large rivers showed similar patterns to those in the Sofron River throughout the year; DOC variations in particular were exact matches (Figure 4a,b). In addition to snowmelt season, DOC and dFe concentrations in the large rivers during rainfall were also consistent with the elevated concentrations of both in the Sofron River. It is therefore suggested that small watershed with permafrost wetlands Mari like the Sofron River greatly contribute to terrestrial transport of DOC and dFe to the large rivers throughout the year, especially during snowmelt season as well as during the rainy season.
Interestingly, however, dFe concentration and dFe/DOC ratio in the large rivers after September, when there was less rainfall, were higher than those in the Sofron River (Figure 1a,c). The reason for this remains unclear, but likely explanation is the inflow of mineral-rich deep groundwater into overlying river through unfrozen ground (taliks) that commonly lies underneath the large rivers [33,63]. As river discharge decreases in rainless September, the contribution of such groundwater to river water might increase, which could explain the high dFe concentrations and the increases in dFe/DOC ratio in the large rivers in this period. dFe flowing into river in this way is believed to be Fe-bearing organic colloids produced in the hyporheic zone, where Fe(II)-rich groundwater and organic-rich river water are mixed [64,65]. It is therefore suggested that, for small rivers like the Sofron River, such deep ground water may not be important to chemical composition because of no talik. More studies are needed about why dFe concentrations in the large rivers are higher than those in the small river after September because it may be an important mechanism to support terrestrial dFe transport even in rainless periods when dFe concentrations in small rivers are low.

Conclusions
In this study, we observed dFe and DOC concentrations in rivers as well as dFe concentration in the soil pore waters during the water active period (from May to October) to understand the spatial and seasonal iron behavior on the watershed scale. To our knowledge, this study is the first to investigate the seasonal changes in dFe concentration in both river and soil pore water. The results highlight the following conclusions: • In snowmelt season, high dFe production occur in the waterlogged surface soils, which leads to the largest terrestrial dFe transport in the year. • Summer rainfall not only increases in dFe and DOC concentrations in river but probably has the effect of promoting dFe production in subsurface soils of permafrost wetlands in valley area. • Overall, permafrost wetlands in valley areas are important environment in which dFe production occurs in response to seasonal hydrological events (spring snowmelt and summer rainfall) and soil thaw depth, and play a significant role in supplying dFe and DOC to rivers.