5.1. Sources and Fate of Suspended Sediments and Carbonates
 Sediment concentrations and loads for the YR are strongly seasonal and highly variable (Table 2; see also Figure 2), as with most Arctic/subarctic rivers [Holmes et al., 2002]. Annual estimates of SS load for YR-Pilot from 1976 to 96 varied by a factor of 4 [Holmes et al., 2002], while those for YR-Pilot in the USGS 2001–2005 study varied by a factor of 2. Porcupine had the greatest variability in annual SS loads (a factor of >3) while Tanana had the least variability (a factor of <1.5).
 Despite the large variability in annual sediment loads, the 68 ± 19 (SD) Mt a−1 average annual load for 2001–2005 at YR-Pilot agrees well with the 54 Mt a−1 load reported by Brabets et al.  and the 60 Mt a−1 load reported by Holmes et al.  for 1976–1996 using sediment transport curves and USGS sediment data. Interestingly, the BQART model developed by Syvitski and Milliman , applied using the YR-Pilot 2001–2005 average Q, watershed area, relief, average basin temperature, glacial coverage, and lithology, predicts a SS load for YR-Pilot of 68 Mt a−1, in exact agreement with the value estimated by LOADEST. SS loads and yields for YR are much greater than for the major Siberian rivers, owing to the alpine glaciers in the YR Basin that discharge large annual sediment loads, and to dams on other rivers such as the Yenisey that trap sediments [Holmes et al., 2002]. The Mackenzie River carries twice the SS load of the YR [Holmes et al., 2002], but has twice the watershed area, resulting in equivalent yields.
 Seasonally, most of the sediment load was transported during the open-water season (Table 2; see also Figure 2). For Porcupine, 91% of the annual load occurred during spring flush, while for Tanana, 73% occurred during summer-autumn. For the Yukon main stem sites, the relative loads for spring versus summer-autumn were closer to 50/50. The relative seasonal contributions to Q were somewhat different. For Porcupine, 51% of annual Q occurred in spring, while at Tanana, 58% of annual Q occurred in summer-autumn. For the main stem sites, 47–50% of annual Q occurred in summer-autumn, with 34–39% in spring.
 Sediment transport curves can aid in the understanding of SS/Q relations. While typical curves take on a simple power function, deviations can occur which illustrate seasonal differences, or hysteretic relations owing to depletion effects [Meade et al., 1990]. Although transport curve characteristics vary spatially and temporally, the curve for YR-Pilot in 2002 (Figure 4a) is not atypical of the YR main stem sites. From winter base flow conditions, SS concentrations rise with the spring flush, then recede until another pulse of sediment is evident in midsummer. The transport curves for Porcupine and Tanana are less complicated (Figures 4b and 4c).
Figure 4. Sediment transport curves. (a) Yukon River at Pilot Station (YRP), 2002. (b) Porcupine River near Fort Yukon (PR), 2001–2005. (c) Tanana River at Nenana (TR), 2001–2005. Note the different instantaneous discharge scale for Figure 4a.
Download figure to PowerPoint
 The seasonality evident in SS concentrations, loads, and transport curves all point to the different sources of sediment in the YR Basin. These sources are important to consider, as source controls the geochemical properties of the sediment, which in turn controls the magnitude of sediment-associated nutrient and contaminant loads [Walling, 2005]. During spring, which includes ice-out and peak Q, the main sources of sediment include bank erosion and resuspension of previously deposited sediments. As peak Qs recede into summer-autumn, glacial melt becomes the dominant source of suspended sediment. This is evident in the greater summer-autumn SS loads at YR-Eagle and Tanana, the two subbasins with the greatest watershed area covered by glaciers. The seasonality of these sources is illustrated in the transport curve at YR-Pilot (Figure 4a), which shows two distinct inputs in spring and midsummer.
 This seasonality effect is also clear in the sediment mineralogy. Carbonate fractions and loads (Table 2) showed strong maxima in summer-autumn owing to the melting of glaciers in the headwaters of the White River, Yukon Territory. The White has been shown to be the major source of carbonates to the YR by Eberl , who found that suspended sediments in the YR were only 2% carbonate upstream from the confluence with the White; downstream from the confluence carbonate content increased to 14%. Given that virtually all the carbonate measured at YR-Eagle originates from the White, and that Tanana is the only other appreciable source, approximately 88% of carbonates entering the YR originate in the White (Table 2).
 Within basin fate of the suspended sediments and carbonates can be estimated by examining the changes in the modeled loads between stations. The decrease in SS load between YR-Eagle (with the addition of the Porcupine River) and YR-Stevens ranged from 2 to 13 Mt a−1 for 2001–2005 (Ave = 6 Mt a−1, SD = 4 Mt a−1) (Table 2). With the addition of about 1 Mt a−1 from the Chandalar River [Brabets et al., 2000], this suggests that, on average, ∼7 Mt a−1 of sediment was lost between YR-Eagle and YR-Stevens, most likely in the extensive Yukon Flats region. If the sediment load at YR-Stevens is summed with that of Tanana and the Koyukuk River [Brabets et al., 2000], it appears that on average, an additional 4 Mt a−1 was lost between YR-Stevens and YR-Pilot, resulting in an average annual loss of about 11 Mt a−1 between YR-Eagle and YR-Pilot.
 The estimated SS load lost to sediment deposition is in good agreement with that of Brabets et al. , although some additional major sources of sediment loss cannot be taken into account. Riverbank erosion can be substantial on the YR, with as much as 2 m of bank eroded per year [Striegl et al., 2007]. This occurs throughout the open-water season, but may be especially rapid during ice-out, when it is too dangerous to sample for SS. Any eroded sediments deposited in floodplains upstream from the next downriver measurement site are unaccounted for in the total loss calculation. In addition, the Porcupine and Tanana sites are located upstream of their confluences with the YR, and therefore do not include sediment entering the rivers from tributaries located between the measurement sites and the river mouths. This would include the Sheenjek and Black Rivers, tributaries of the Porcupine, and the Nenana and Kantishna Rivers, tributaries of the Tanana. These two latter rivers in particular drain high glaciated regions of the Alaska Range, and likely contribute substantial amounts of SS to the Tanana. Also missing from the sediment export calculation is the contribution by bed load transport. While bed load transport has not been measured in the main stem of the Yukon, measurements in the Tanana suggest that bed load transport accounts for only 1–2% of the suspended sediment load [Burrows et al., 1979]. And finally, the proportion of SS that settles out in the Yukon Delta before reaching the Bering Sea is unknown. While it has been estimated that about half of Mackenzie River sediments are transported beyond the Mackenzie Delta into the Arctic Ocean [Macdonald et al., 1998], YR transport to Norton Sound and the Bering Sea could vary considerably.
 Carbonate loads also indicate a loss moving down the YR in AK (Table 2). Between YR-Eagle and YR-Stevens, the carbonate load decreased 27% on an average annual basis, with a majority of that decrease occurring during summer-autumn. At YR-Pilot, 29% of the carbonate load from YR-Stevens and Tanana was lost annually. These estimates are in agreement with those of Eberl . One gap in understanding the source and fate of carbonate for the entire YR is a result of a lack of SS concentration data for the Donjek River, Yukon Territory. The Donjek River originates in the St. Elias Mountains and flows into the White River downstream from the White River gage. Suspended sediment load estimates for the White [Brabets et al., 2000] therefore do not include the Donjek. Given that the Donjek has an annual Q similar to that of the White (Water Survey of Canada; see http://www.wsc.ec.gc.ca/index_e.cfm?cname = main_e.cfm), and that the carbonate content of Donjek River sediments is almost twice that of White River sediments D. Eberl, USGS, unpublished data, 2005), sediment and carbonate loads from the Donjek are probably substantial. However, any losses of Donjek River carbonates in the reach to YR-Eagle cannot be determined at this time.
 Carbonate losses are due to a combination of deposition and carbonate dissolution [Eberl, 2004; Striegl et al., 2007]. As a first-order approximation, because ∼10% of the SS load at YR-Eagle is lost in Yukon Flats (Table 2), it could also be assumed that ∼10% of the carbonates are also lost owing to sedimentation, with the remainder of the missing carbonate undergoing dissolution. However, Eberl , using a more rigorous approach involving mineralogical analysis of suspended sediments, concluded that approximately 30% of carbonates undergo dissolution between YR-Eagle and YR-Stevens. While this carbonate weathering process contributes little to total DIC (dissolved inorganic carbon) export, it substantially alters the isotopic composition of DIC, and it has occasionally consumed YR CO2 to less than atmospheric equilibrium [Striegl et al., 2007], changing the YR from a source of atmospheric CO2 to a short-term sink.
 On an average annual basis, the YR appears to store millions of metric tons of sediment in floodplains between YR-Eagle and YR-Pilot. It is quite possible that this storage is temporary, however, and that much of this sediment is re-suspended during peak Q every spring. Froese et al. , studying longitudinal valley fill thickness in the middle YR Basin, concluded that this stretch of the river has been in a state of mass balance equilibrium for sediments over the Holocene. The ultimate fate of sediment and carbonate loads remains unclear. While sediment loads at YR-Pilot can be estimated, the extent of sedimentation in the Yukon Delta versus how much sediment reaches the Bering Sea is unknown. This open question has important implications for the ecology of the Bering Sea, as processes in the Delta will dictate the final transport and/or sequestration of sediment-associated organic carbon, carbonates, nutrients, metals, and contaminants.
5.2. Recent Trends From the Tanana River
 While only 5.6% of the Tanana Basin is glaciated, it is not surprising that most of the SS in the Tanana originates from these regions [Chikita et al., 2007]. As glaciers recede under a warming climate, one might expect changes in the SS load of rivers such as Tanana. Using the SS record dating back to 1966 (NWIS; see http://waterdata.usgs.gov/nwis), Tanana SS loads were modeled using LOADEST for the water years (WY) 1966–1970, the earliest 5-year period where data are available, as well as WY 1983–1987, a 5-year period intermediate between the earliest records and those of the major 2001–2005 study.
 There is a significant difference between the three 5-year periods (Figure 5). On an annual basis, cumulative SS yield versus water yield has the lowest slope for 1966–1970, the greatest slope for 1983–1987, and an intermediate slope for 2001–2005 (Figure 5a; p < 0.0001, ANOVA). Seasonally, this trend is most reflected in summer-autumn (Figure 5c; p < 0.0001). This is not unexpected, as most of the SS load from Tanana occurs during this season (Table 2). Spring (Figure 5b) does not show a significant difference between the three periods, while winter (Figure 5d) shows a significant difference between 1966 and 70 and the two later periods.
Figure 5. Cumulative suspended sediment yield versus cumulative water yield, Tanana River at Nenana (TR): (a) annual, (b) spring, (c), summer-autumn, and (d) winter.
Download figure to PowerPoint
 One possible correlate for the computed differences may be climate shifts, often the driving factor for sediment loads after human effects [Syvitski, 2003]. According to Pacific decadal oscillation (PDO) indices, the years 1966–1970 were in a cool phase, the years 1983–1987 were in a warm phase, and the years 2001–2005 appear to encompass a transition between cool and warm (Joint Institute for the Study of Atmosphere and Ocean (JISAO); see http://jisao.washington.edu/pdo). The observed differences in the SS yield/water yield relations (Figure 5a) appear to reflect these PDO phase changes, with 1966–1970 having the least SS yield per given water yield (cool phase), 1983–1987 having the greatest yield relation (warm phase), and 2001–2005 having an intermediate relation (cool to warm transition). Average annual SS yields (Table 3) mirror this trend.
Table 3. Average Annual and Seasonal Water Yields and Suspended Sediment Yields, Tanana River at Nenana ± Standard Deviation
|Years||PDO Phase||Average Spring Water Yield (cm)||Average Summer-Autumn Water Yield (cm)||Average Winter Water Yield (cm)||Average Annual Water Yield (cm)||Average Annual Suspended Sediment Yield (g m2 a−1)|
|1966–1970||cool||9.3 ± 2.8||17.9 ± 4.4||4.5 ± 0.5||31.7 ± 5.6||330 ± 120|
|1983–1987||warm||7.8 ± 1.8||20.4 ± 1.8||5.5 ± 0.5||33.7 ± 2.6||559 ± 93|
|2001–2005||transition||9.3 ± 2.1||20.6 ± 1.3||5.7 ± 0.6||35.6 ± 1.3||503 ± 58|
 In Alaska, the PDO typically manifests itself most strongly in winter, most likely owing to the lack of substantial solar radiation forcing, which results in the advection of warm moist air into the state becoming a major climate driver [Hartmann and Wendler, 2005]. In interior Alaska, wherein the Tanana Basin resides, the 1976 shift between cool and warm phases of the PDO resulted in winter increases in air temperature (+3.1°C), cloudiness (+7%), and snowfall (+21%), although air temperature and total precipitation also increased in summer [Hartmann and Wendler, 2005].
 Seasonal water yields for the Tanana seem to mirror these shifts in precipitation. While not statistically significant, average summer-autumn and winter water yields for Tanana appeared to increase between the cool phase 1966–1970 years and the warm phase 1983–1987 years, while spring water yield seemed to decrease (Table 3). The years 2001–2005 saw an increase in average spring water yield, with no change in average summer-autumn or winter water yields over the 1983–1987 period. A positive change in the sediment/water relation between cool and warm PDO phases for summer-autumn and winter (Figures 5c and 5d) is not surprising, given the concurrent increases in air temperature and precipitation. Unfortunately, lack of long-term data from subbasins such as the White River prohibits an examination of other possible climate shift/sediment yield relations.
5.3. Climate Change Implications
 Warmer temperatures will likely result in continued melting of glaciers and permafrost, changes in Q and precipitation, a longer open-water season, increased fire frequency, and changes in atmospheric CO2 [Serreze et al., 2000]. Each of these factors has the potential to alter future SS loads. Receding glaciers should initially cause more erosion, as glacial drift left behind becomes available for transport [Meade et al., 1990]. Increased SS loads may be sustained even after glaciers disappear, although eventually the loads will decrease as easily transported material decreases. Given the importance of glaciers in the headwater areas of the YR Basin, any continued shrinkage of glaciers could have a large impact on SS loads in the YR.
 While there have been no observed changes in Q for the YR for a >30-year record prior to 2000 [Walvoord and Striegl, 2007; McClelland et al., 2006], net precipitation is expected to increase at high latitudes with continued warming [Cubasch et al., 2001]. Changes in precipitation appear to be the primary driver for changes in Q [McClelland et al., 2004], thus concomitant changes in Q with changes in precipitation would not be unexpected. Given the positive relation between SS yield and water yield (Figures 3 and 5), any increases in precipitation should be accompanied by increases in SS yield. This relation is seen in climate models that predict a 10% increase in SS loads in arctic rivers for every 20% increase in Q [Syvitski, 2002; Gordeev, 2006]. These same models also predict sediment load increases of 22–30% for arctic/subarctic rivers for every 2°C increase in temperature.
 In addition to the total change in precipitation, the timing of any changes in precipitation will also be important. Seasonally, spring sees the greatest increase in SS yield with water yield for all subbasins except perhaps Tanana (Figures 3b and 3c), and thus might be expected to see the greatest increases in SS yield from any increases in spring water yield. However, as 50% of the SS load at YR-Pilot comes from the Tanana, summer-autumn changes in water yield could result in substantial changes in SS loads at YR-Pilot as well.
 Other factors associated with warmer temperatures will also likely increase SS loads. Melting stores of permafrost will expose long-frozen soils to erosion. A longer open-water season will allow more time for bank erosion, and an increased fire frequency will also open up more forest soil to erosional forces. Further, continued increases in atmospheric CO2 concentrations may lead to greater SS loads. The model of Van Blarcum et al.  predicts a 15% increase in precipitation and a 26% increase in runoff for the YR in a doubled CO2 climate. While their model carries some uncertainty owing to low predictability of current precipitation and runoff, their results are consistent with the general prediction that precipitation and runoff would increase in a doubled CO2 climate [Aerts et al., 2006; Miller and Russell, 1992; Stouffer et al., 1989; Mitchell, 1989].
 Finally, linkages between climate variables (i.e., Arctic Oscillation, ENSO, and PDO) and Q and precipitation have been documented [Peterson et al., 2006; Dery and Wood, 2005; Neal et al., 2002; Brabets and Walvoord, 2009], and a possible linkage between SS yield and PDO has been suggested for the Tanana in this study. Thus, an examination of possible linkages should be explored when considering long-term data sets and any potential relation with climate change. However, care must be taken when drawing conclusions on the basis of these linkages, as climate indices are often interrelated [McClelland et al., 2006], and indeed, climate change may be changing the patterns of these indices [Shindell et al., 1999].
 As arctic systems are particularly sensitive to change [Serreze et al., 2000], it is important to understand current fluxes across the terrestrial-aquatic continuum, so that future changes in arctic rivers transport processes might be gauged and predicted. However, the size and direction of trends are not constant across the arctic/subarctic [McClelland et al., 2006]. Therefore it is important to gather as much information as possible, from as many systems as possible, and not rely on broad extrapolations across the pan-Arctic region.