3. RESULTS
The annual snowmelt pulse typically began in early May, peaked in early June, and had subsided by early July (Figure 2). Seasonal increases in concentrations of DOC and NO3-N appeared to precede the melt pulse by several weeks, typically peaking in April to May and then, especially in the case of NO3-N, already declining well before Q peaked in June. Similar temporal patterns were observed in the historic data collected by the LTER (Supporting Figure 1).
There was considerable interannual variation in the timing and magnitude of the snowmelt pulse among the four study years (Table 1). For example, the centroid of the pulse in 2019 was over 3 weeks later than in 2018, and the water yield for 2019 was over 50% higher than 2018. The historical LTER records showed similar interannual variation (Supporting Table 1).
The annual flux estimates from downsampled sensor data were not substantially different from those made using the whole dataset (Table 2), never off by more than ±3% in any year. This indicates that weekly concentration sampling is likely sufficient for estimating annual fluxes, and that values calculated using the 2018-2021 NEON sensor data can be accurately compared with historical LTER values. We noted that this was not the case if Q was downsampled to a discrete dailymeasurement rather than a daily average . Because of large diel variation in Q, the time of day chosen for the Q measurement has the potential to substantially impact annual flux estimates (by up to ± 30% or more!). Because it exerts such a strong influence, high-frequency Q measurements appear essential for accurately estimating fluxes.
Despite concentrations which were already declining, maximum export of DOC and NO3-N (mass flux shown as gray lines in Figure 2) occurred concurrent with the peak of the melt pulse. Over the 4 years of NEON sensor data, the CVQ was 2.47 while CVDOC was 0.47 and CVNO3-N was 0.61. This made Q variation the dominant control on flux. Flux was highly unequal in time, with a GDOC of 0.87 and GNO3-N of 0.82. Seventy nine percent of the DOC export and 71% of the NO3-N export occurred within just 10% of the time; the peak weeks of snowmelt.
Globally, fitted C-Q relationships suggest DOC was slightly enriched (Log-Log slope = 0.14), though still in the range (slope < ± 0.2; Godsey et al., 2009) which would be considered relatively chemostatic. NO3-N was almost perfectly chemostatic (Log-Log slope = 0.01). In contrast, C-Q relationships for major conservative solutes (Supporting Figure 2) exhibited moderate dilution responses, though with slopes still far from a value of -1 which would signify a perfect dilution response of a fixed mass flux of solute. Coefficients of variation for conservative solutes (e.g. CVNa = 0.64, CVCa = 0.46) were also much less than CVQ , and using SpC as a surrogate for a continuous conservative ion concentration produces a GSpC of 0.75. Thus, despite the weak dilution response, the majority of conservative ion export also occurs during the melt pulse.
The seasonal increase in DOC and NO3-N concentrations which preceded the melt pulse resulted in noticeable clockwise C-Q hysteresis for both solutes over annual time scales (dashed lines in Figure 3a&b). Within these larger annual hysteresis loops are smaller loops generated by individual precipitation events. These include spring ROS, summer rainfall when no snow is present, and autumn/early-winter snow that quickly melts without becoming snowpack. These events produce consistent enrichment of DOC and NO3-N, with fitted LogC-LogQ slopes ranging from 0.15 to 0.35 across individual events. These slopes for individual events are distinctly steeper than those observed at seasonal time scales. Concentration variations typically lag Q during such events (Figure 3c-f), producing counterclockwise C-Q hysteresis loops within the larger clockwise seasonal loop (solid line in Figure 3a&b).
Discharge exhibited significant diel variation during the melt pulse, with peak daily values often 50% higher than the minima for the same day (Figure 4a). These daily pulses, presumably driven by greater daytime melting, typically peak in the late afternoon. DOC and NO3-N, as well as SpC, DO and Temperature also exhibited varying degrees of diurnal variation (Figures 4 & 5). SpC was especially interesting in that during the peak of the melt pulse it showed extremely modest diel variation (<0.2 µS cm-1) despite large diel variation in Q (>100 L s-1). During base flow however, when diel Q variation was orders of magnitude less, diel Spc variation was considerably greater (1-2 µS cm-1).