Understanding the spatial variability in catchment dynamics: A case study of 107 stream catchments in Victoria, Australia

Abstract

Rivers and streams around the world are being affected by declining water quality. When designing remediation strategies, we must first understand the key factors affecting spatial and temporal variability in stream water quality. As such, the objective of this investigation was to investigate the relationships between in-stream constituent concentrations and streamflow and to understand how these relationships vary across space. We intend to use these findings to add a temporal component into existing statistical models of spatial variability in water quality. Monthly water quality data for total suspended solids (TSS), total phosphorus (TP), filterable reactive phosphorus (FRP), total Kjedahl nitrogen (TKN), nitrate-nitrite (NOx) and electrical conductivity (EC), in addition to streamflow collected between 1994 and 2014 from 107 water quality monitoring sites in Victoria were used for this study. Using these data, we characterized the interaction between constituent concentrations and streamflow in terms of (i) the ratio of the coefficient of variation (CV) of constituent concentrations to the CV of streamflow (CVC/CVQ), and (ii) the slope of the linear regression between the log-transformed constituent concentrations and log-transformed streamflow (the C-Q slope). We then linked the spatial variations in CVC/CVQ and the C-Q slope to catchment characteristics (e.g., land use and climate). We found that the interaction between constituents and streamflow depends significantly on the reactivity of the constituent, and whether the constituent is in the dissolved or particulate state. TSS, TP, TKN, FRP and NOx demonstrated chemodynamic behavior, with the concentrations varying with streamflow (i.e., high CVC/CVQ and large absolute value in C-Q slope). On the other hand, EC demonstrated chemostatic behavior for the selected sites, with low CVC/CVQ values and C-Q slopes. The interaction between streamflow and constituents varied significantly across space. The variability in CVC/CVQ for TSS, nutrients and salts correlated positively with catchment characteristics such as mean catchment slope, average annual rainfall and woodland cover. This could be due to the weaker sources of TSS due to reduced erosion, nutrients due to zero or low application and salts due to high leaching in steeply sloping, vegetated and high rainfall catchments (as they tend to be less disturbed). Lower magnitude and less temporal consistency of constituent sources can lead to greater variability in constituent concentrations relative to streamflow. The spatial variability in C-Q slopes generally did not correlate strongly to catchment characteristics, likely due to the presence of major dams in approximately half of the water quality monitoring sites. However, once these sites were removed, we found that the TSS C-Q slope correlated strongly to average annual rainfall and the mean 7-day low flow. This suggests that there is a stronger positive linear relationship between TSS concentrations and streamflow in catchments with temporally consistent rainfall and streamflow. There were weak correlations between catchment characteristics and the C-Q slopes for nutrients regardless of the exclusion of the water quality monitoring sites with dams. This could be due to the reactive nature of these compounds, leading to less predictable interactions between streamflow and in-stream concentrations. The results of the analysis will be used to develop statistically-based predictive models of spatio-temporal variability in stream water quality.