Meteorological Controls on Sediment Transfer
Precipitation was excluded from three of four best-fit models due to significant correlations with Q and NTU, suggesting that precipitation has a strong influence on discharge and turbidity. P12at 1750 m asl was a significant predictor in the Chamberlin Creek NTU-based model (Table 3). P12 possibly represents a coarse fraction of suspended sediment eroded and entrained during precipitation-induced high discharges. This would be consistent with nephelometric turbidimeter signals being more sensitive to fine particles (clay and silt) due to their greater light scattering efficiency and particle surface area, thus potentially under-representing the coarse (sand) fraction (Orwin et al., 2010).
We estimated that substantial seasonal sediment delivery in both study creeks was transported over 48 hours (Figure 5). The hydrological response to rainfall during these events may have been enhanced by their timing in mid-July to early-August, when channel-ice and snow were less likely to impede erosion (e.g. Crawford & Stanley, 2014; Irvine-Fynn et al., 2005), and glacier conduits are likely open to flush sediment into the proglacial fluvial system. Additionally, the placement of hydrological stations near stream outlets may have represented the rainfall signal more completely than if they were placed farther upstream (e.g. Irvine-Fynn et al., 2005; Orwin & Smart, 2004b; Willis et al., 1996). The short duration over which the majority of sediment is transferred to Lake Peters is comparable with other rivers that drain the northern Brooks Range, including: the Kuparuk River near Prudhoe Bay where 90% of the annual suspended sediment load was transferred over three days in 2001 (Rember & Trefry, 2004); the Sagavanirktok River, with 88% over 12 days in 2001 (Rember & Trefry, 2004); and the Colville River, with 62% over 13 days in 1961 (Walker & Hudson, 2003), although these studies emphasize melt processes rather than rainfall.
Similar to precipitation variables, temperature variables were excluded from three of four best-fit models due to significant correlations with other predictors. We note that such correlations have not been reported by authors of similar models (Irvine-Fynn et al., 2005; Schiefer et al., 2017). Albeit model limitations, field observations and diurnal signals suggest that temperature-driven melt processes influence sediment transfer in both Carnivore and Chamberlin Creeks. The inclusion of GT2 (1750 m asl) in our Carnivore NTU-based model suggests that melt-processes mobilize a sediment supply less discernible to the turbidity sensor. We note that this model (Table 3) is the first to incorporate ground temperature as a supplementary predictor of SSC, although ground conditions have previously been related to sediment transfer in the Arctic (Favaro & Lamoureux, 2014; Irvine-Fynn et al., 2005; Syvitski, 2002). Supplementing this NTU-based model with ground temperature (using GT2 at 1750 m asl, R2 = 0.68) explains slightly more variability in SSCs than air temperature (using T2 at 1750 m asl, R2 = 0.63), which may relate to ground temperatures better reflecting thaw-related sediment mobilization and/or improved model relations with high-frequency filtering in the ground temperature record (Figure 4).
Although both rainfall and temperature affect discharge and sediment transfer, they do not fully explain the disparities between the 2015 versus 2016 hydrographs (Figure 4). In 2016, Q was elevated for an extended period between mid-June and mid-July in both creeks, but neither precipitation nor temperature were notably higher in 2016. End-of-winter snow water equivalence has been positively related to discharge and sediment transfer in the Arctic (Bogen & Bønses, 2003; Cockburn & Lamoureux, 2008; Forbes & Lamoureux, 2005; Lewkowicz & Wolfe, 1994); however, simple DEM differencing of repeat photogrammetric surfaces calibrated for catchment snow measurement at Lake Peters (Broadman et al., 2019; Nolan, Larsen, & Sturm, 2015) suggested a greater overall snowpack in 2015, thus snow water equivalence does not explain the elevated 2016 Q. Limitations associated with developing continuous discharge time-series from stage could contribute to the seasonal hydrograph differences, although such limitations do not appear great enough to affect average Q, nor the magnitude of sediment yield results (Thurston, 2017). An earlier study on Chamberlin Creek reports an average discharge of 0.65 m3 s-1for 44 days in late summer (between 07/01/1958 and 08/13/1958) (Rainwater & Guy, 1961), which is similar to the mean daily discharges of 0.62 and 0.77 m3 s-1 (2015 and 2016, respectively) for the same days of the year in this study.
We interpret temperature-driven melt processes as secondary to rainfall in controlling sediment yield, because most of the sediment load is transported to Lake Peters during rainfall-induced flood events. Although the same intensity and volume of rainfall does not necessarily equate to the same magnitude of sediment load, rainfall is clearly associated with events that transport the majority of the annual sediment yield to Lake Peters. Rainfall and sediment transfer are strongly correlated from the onset of the open-channel seasons, with no distinct shift between temperature-driven spring snowmelt and late summer rainfall-induced sediment transfer during the period of record in 2015 and 2016, as has been observed in Arctic Canada (Dugan et al., 2009; Lewis et al., 2012; McLaren, 1981). In Arctic Alaska, literature reporting sediment yields (Table 5) has not established the metrological processes driving water and sediment discharge, although hydrograph research provides inference. At Lake Peters, the dual snowmelt and rainfall early in the open-channel-season, and more significant rainfall in July and August, is inconsistent with a regional hydrograph developed for the coastal plain of the Arctic National Wildlife Refuge, which showed most of the total water discharge during the spring freshet (Lyons & Trawicki, 1994) and later work that applies this as an assumption (e.g. Rember & Trefry, 2004). Following observation of an extreme flood event in the Upper Kuparuk River, Kane et al. (2003) suggest that most floods of record in Arctic Alaska are rainfall-generated, especially in smaller catchments; however, our research suggests that seasonal floods are also rainfall-generated. At Lake Peters, significantly greater, and primarily rainfall-driven, discharge volume in 2016 explains the higher annual sediment yield that year than in 2015. In Svalbard and Arctic Canada, metrological processes driving sediment discharge have been researched, in addition to hydrographs. Dominantly rainfall-induced sediment transfer (Bogen & Bønses, 2003; Lamoureux, 2000; Lewis et al., 2012), similar to Lake Peters, and temperature-dominated melt regimes of sediment transfer (Braun, Hardy, Bradley, & Retelle, 2001; Favaro & Lamoureux, 2014; Hardy, 1996; Irvine-Fynn et al., 2005; Moore, Hughen, Miller, & Overpeck, 2001; Smith, Bradley, & Abbott, 2004), have both been reported. Comparison with arctic forecasting studies suggests that relatively small, mountainous arctic catchments (e.g. Lake Peters; and Lewis & Lamoureux, 2010) are more responsive to rainfall than larger, coastal arctic catchments (e.g. Syvitski, 2002).