Table 1.
Details on NEON sensors, variables collected, units of measurement, associated data-collection intervals, and the NEON data product number for data used in this study.
Table 2.
Details of NEON sites used in this study.
Fig 1.
Box-plots of water-quality data for Arikaree, Caribou and Lewis Run.
Bold lines within boxes represent medians and lower and upper edges of boxes represent the interquartile range (IQR), with whiskers extending to 1.5 times the IQR. Closed circles represent data with values beyond the whiskers. Note, the y-axis for turbidity uses a base-10 log scale.
Fig 2.
The top three plots are examples of diel fluctuations and other trends in water-quality at the study sides, visualised over 5-day windows of representative data, each starting and ending at 10 pm GMT.
Local time in the three top plots starts at 3 pm at Arikaree, 2 pm at Caribou, and 5 pm at Lewis Run. The three bottom plots are examples of flow events, which are indicated by sudden rises in surface water elevation. Cond., conductance; Elev., elevation; Temp., temperature. Note that the scales for the y-axes differ among sites and variables.
Fig 3.
Smooth regressions (solid lines) of the expected change in nitrate concentration (y-axes) when values taken by covariates (x-axis) increased, by site.
More specifically, each plot represents the regressive spline sk(zki) (Eq (1)) of one covariate at one site, i.e. the expected change in nitrate concentration Yi for a one unit increase in the covariate zki. Dashed lines show the standard error estimates. Missing values (gaps in the time series data) are not shown.
Table 3.
Model performance for all three sites, as based on the approximated Akaike Information Criterion (aAIC).
Fig 4.
Variable importance, as the percentage of the total GAMM model deviance, for statistically significant covariates (p < 0.001), by study site.
Total deviance explained by each GAMM was 99% for all three sites.