Abstract:

Salinity is an important water quality parameter that affects ecosystem health and the use of freshwaters for industrial, agricultural, and other beneficial purposes. Although a number of studies have investigated the variability and trends of salinity in rivers and streams, the effects of floods on salinity across a wide range of watersheds have not been determined. Here, we examine this question by utilizing long-term observational records of daily streamflow and specific conductance (SC; a proxy for salinity) in addition to catchment characteristics for 259 United States Geological Survey (USGS) monitoring sites in the contiguous United States spanning a wide range of climatic, geologic and hydrologic conditions. We used a combination of statistical methods, random forest machine learning models, and information-theoretic causal inference algorithms to determine the response of SC to floods and the factors that impact salinity changes within sites (intra-site variability) and across sites (inter-site variability). Our results show that changes to SC during flood events exhibited substantial variability ranging from a 100% decrease to 34% increase relative to the long-term mean. We found that dilution is the prevailing mechanism that decreases SC levels during floods for most sites, but other mechanisms caused an increase of SC for 6.1% (n = 5521) of flood events. Our analysis revealed that antecedent conditions of SC in the few days preceding the flood are the most important factor in explaining intra-site variability. The response of salinity to floods also varied considerably across sites with different characteristics, with a notable effect of urbanization in temperate climates resulting in increased dilution of SC, and mining in arid climates, which adversely increases SC levels. Overall, we find that the combined effect of aridity and anthropogenic factors is of primary importance in determining how salinity responds to floods, and it bears strongly on water quality conditions in a future world — one in which floods are expected to increase in frequency and intensity, concurrent with shifting aridity patterns and increasing urbanization.