An improved method for interpretation of riverine concentration‐discharge relationships indicates long‐term shifts in reservoir sediment trapping

Abstract

Derived from river monitoring data, concentration‐discharge (C‐Q) relationships are powerful indicators of export dynamics. Proper interpretation of such relationships can be made complex, however, if the ln(C)‐ln(Q) relationships are nonlinear or if the relationships change over time, season, or discharge. Methods of addressing these issues by “binning” data can introduce artifacts that obscure underlying interactions among time, discharge, and season. Here we illustrate these issues and propose an alternative method that uses the regression coefficients of the recently developed “Weighted Regressions on Time, Discharge, and Season” model for examining C‐Q relationships in long‐term, discretely sampled data for various water‐quality constituents, including their uncertainties. The method is applied to sediment concentration data from Susquehanna River at Conowingo Dam, Maryland, to illustrate how the coefficients can be accessed and presented in ways that provide additional insights toward the interpretation of river water‐quality data, which reaffirms the recently documented decadal‐scale decline in reservoir trapping performance.

Concentration versus streamflow trends of major ions and tritium in headwater streams as indicators of changing water stores

“…Variations in solute concentrations with streamflow (discharge) in rivers and streams contain vital information on residence times, processes of solute generation, biogeochemical processes, and sources of stream water; thus, documenting the controls on solute geochemistry is vital to understanding catchment functioning (Ameli et al, ; Bieroza, Heathwaite, Bechmann, Kyllmar, & Jordan, ; Bouchez et al, ; Clow & Mast, ; Creed et al, ; Diamond & Cohen, ; Duncan, Band, & Groffman, ; Godsey, Kirchner, & Clow, ; Guan et al, ; Hunsaker & Johnson, ; Ibarra, Moon, Caves, Chamberlain, & Maher, ; Kim, Dietrich, Thurnhoffer, Bishop, & Fung, ; Koger, Newman, & Goering, ; Uhlenbrook & Hoeg, ; Wlostowski, Gooseff, McKnight, & Lyons, ; Zhang, Harman, & Ball, ). There are several potential sources of the water in streams and rivers that may have significantly different geochemistry (Cartwright, Gilfedder, & Hofmann, ; Cook, ; Gonzales, Nonner, Heijkers, & Uhlenbrook, ; Nathan & McMahon, ; Yu & Schwartz, ).…”

Dynamic linear models to explore time‐varying suspended sediment‐discharge rating curves

“…However, as noted in Zhang et al . [], this approach requires a subjective choice of bin intervals, which can impart discontinuities in the rating curves that are more artifacts of bin selection than true shifts in sediment yield. To circumvent this issue, Zhang et al .…”

“…To circumvent this issue, Zhang et al . [] employed a new rating curve method ‐ weighted regression on time, discharge, and season (WRTDS) – first proposed by Hirsch et al . [] in the context of nutrient loading.…”

Time‐varying, nonlinear suspended sediment rating curves to characterize trends in water quality: An application to the Upper Hudson and Mohawk Rivers, New York

“…A common approach to explore time variations in rating curves is to fit the model using nonoverlapping bins of data (usually year by year) and then examine how the parameters change across fitting intervals (Herndon et al, 2015;Warrick, 2015). This approach has been criticized for the arbitrary selection of bin size that can influence the timescales of change (e.g., daily, monthly, interannual) reflected in the parameter variations (Zhang, Harman, & Ball, 2016). It is also complicated by potential nonlinearity in the underlying rating curve relationship, which is a common feature worldwide (Moatar, Abbott, Minaudo, Curie, & Pinay, 2017) and may be difficult to adequately characterize with a simple parametric curve (McDonnell et al, 2007;Warrick, 2015).…”

“…WRTDS has been utilized for three primary applications: data infilling (Chanat, Moyer, Blomquist, Hyer, & Langland, 2016;Moyer, Hirsch, & Hyer, 2012;Pellerin et al, 2014;Zhang & Ball, 2017), flow-normalized trend analysis of concentrations or loads (Medalie, Hirsch, & Archfield, 2012;Moyer et al, 2012;Zhang, Brady, Boynton, & Ball, 2015), and the analysis of change over time in rating curve relationships (Zhang, 2018;Zhang et al, 2016). For the last application, one challenge that can arise but has not been previously discussed is the interaction between the intercept and coefficient on time.…”