Our analysis of high-frequency sensor data from storms in three reservoir embayments shows that pulses of incoming stormflow can be readily detected, induce significant changes to the physical and chemical conditions, and frequently propagate 800 m or more through reservoir embayments. The chemical effects of stormwater pulses were more pronounced in embayments draining urbanized streams, with the influence of stormwater principally governed by the relative volume of inflowing stormwater relative to the volume of lentic water stored in the reservoir. In contrast, the persistence of thermal stratification of the water column during storms did not depend on the size or intensity of storm. Instead, the probability of destratification during storms depended primarily on the mean air temperature preceding the storm, with destratification more likely when air temperatures were colder.
Conductivity signals of stormwater pulses and propagate further in urbanized sites
The relative volume of stormwater to lentic receiving water is was an important predictor of the magnitude of its effects. We found that a SW:ResVolume of 4.52*10− 6 is needed to attenuate the conductivity signal of stormwater pulses in reservoirs. This means that for a one-inch (2.54 cm) rainstorm in the urbanized Fourth Creek watershed of Knoxville, TN, which yields an approximate total runoff volume of 199,100 m3, would need a reservoir volume of 4.406*1010 m3 (more than the storage capacity of the entire reservoir) to attenuate the physicochemical signal of stormwater.
The landcover of watersheds was related to the characteristics of stormwater effects in receiving reservoirs. At more urbanized sites (FC and TB), the storm conductivity area was similar among buoys for a given storm, suggesting that much of the volume of stormwater runoff can be transmitted large distances through the reservoir. Although the magnitude of the storm conductivity minimum decreases and the area becomes broader and shallower as stormwater pulses move downlake (Fig. 2), the consistency in storm conductivity area among buoys suggests that these urban stormwater pulses travel as coherent slugs through the receiving embayments. In contrast, stormwater pulses observed at the mouth of WB were either partially attenuated before reaching buoys 2 and 3, or followed flowpaths that did not intersect with our sensors. Flashier hydrology in urban as opposed to non-urban streams may facilitate the shunting of stormwater runoff through impoundments. However, it is impossible for us to differentiate whether differences in storm response among our study sites is the result of landcover in the three catchments or arises from the unique morphology of each site (drainage ratio, bathymetry, etc.) or, most likely, the combination of these factors. Additional research on the effects of sotrmwater pulses in receiving reservoirs will help to distinguish the effects of watershed landcover vs. reservoir morphometry.
Residuals of the relationship of storm conductivity area between downlake buoys and the stream mouth were smaller when SW:ResVolume was high, indicating that the volume of stormwater moving past buoys is more similar among buoys during larger vs smaller storms. This suggests that during lager storms these embayments are dominated by directional flow (i.e., act more lotic) and coherent pulses of stormwater are advected through the flooded channel toward the reservoir mainstem.
Seasonal differences in storm-driven destratification
While urbanization can significantly increase water temperature, and particularly stormflow temperature, in streams (Somers et al. 2013), we did not find differences among watersheds in the destratification associated with incoming stormflow pulses. Instead, preceding air temperature, total rainfall, and the volume of stored water were the most important predictors of whether incoming stormflow pulses would result in thermal destratification of the receiving reservoir embayment.
Observed variation in temperature differences between surface and bottom waters may indicate differences in the strength of stratification among seasons. During summer, thermal stratification is strong and may be more resistant to the flow-driven destratification during storms. This, in turn, may suggest different hydrodynamics of incoming flow (e.g. underflow of incoming water beneath warmer, less dense water stored in the reservoir; Thornton et al. 1990) in summer as opposed to winter, when stormflow may be more likely to fully destratify and mix the water column. We also found that storms were more likely to result in destratification when there was less water stored in the reservoir, i.e., when embayments were shallower. Because thermal stratification of the water column (and related changes in deepwater dissolved oxygen concentration) is an important control on processes including greenhouse gas emission (Deemer et al. 2016) and internal nutrient loading (Mortimer 1941), seasonal differences in destratification likelihood have implications for reservoir carbon balance and productivity.
Management of urban watersheds and reservoirs
Stormwater pulses are likely to exert expanding influence on reservoir hydrodynamics and chemistry owing to ongoing changes in land use and management, climatic change, and responses of reservoir management to these drivers. Our analysis points to two main categories of controls on the propagation of the acute effects of stormwater pulses through reservoir embayments: the “signal strength” of incoming storms (runoff volume and intensity of storm) and the “resistance” of the reservoir (volume of stored water and strength of thermal stratification). Ongoing changes in climate, land use, and management responses will alter incoming storms and the resistance of reservoirs (Hayes et al. 2017), with the likely outcome of these changes being increases in the magnitude and spatial footprint of the effects of stormwater pulses. Among the reservoir embayments we studied, pulses of stormwater were larger and propagated further in the reservoir embayments draining catchments with more urban development. Streams draining urbanized watersheds have higher runoff ratios (i.e. a greater proportion of rainfall becomes streamflow) than non-urban streams as a result of higher imperviousness (Dunne and Leopold 1978; Walsh et al. 2005), meaning increased signal strength for storms in urban watersheds. Given the strong influence of SW:ResVolume on propagation of stormwater pulses, we expect urbanization would increase the propagation of stormwater pulses through receiving reservoirs by generating a greater volume of stormwater for a given rainfall.
In our study, SW:ResVolume was a highly important predictor of the propagation of storm signals, and both terms in this ratio are subject to influence by human activity and decision-making. First, stormwater management affects the magnitude of incoming stormwater pulses. Impervious surfaces associated with urbanization increase the runoff ratio and thus the volume of stormwater pulse resulting from a rainstorm. Stormwater control measures that promote infiltration or detention reduce the total volume and intensity of stormwater pulses (e.g., Hopkins et al. 2019), thereby reducing the magnitude and spatial footprint of physicochemical signals of stormwater pulses in receiving waters. The volume of water stored in the reservoir, the denominator of this ratio, is also subject to the outcomes of human decision making. Water level at the dam in these reservoirs varied by more than 2 m over the course of our study, resulting in large differences in the amount of water stored in reservoir embayments and driving considerable variation in SW:ResVolume. As such, reservoir management can be an important driver of propagation or attenuation of incoming storm pulses. The construction, management, and removal of impoundments (including stormwater ponds and detention basins) along an urban stream network will also affect propagation of stormwater signals. The effect of multiple impoundments in a connected network on the size and propagation of stormwater temperature and chemistry signatures deserves further attention.
Finally, climate change will likely affect the characteristics of storms and the responses of reservoirs to them. The amount of precipitation delivered by individual storms is predicted to increase in many parts of the United States (USGCRP 2017). Since storms of larger precipitation volume and greater intensity drive bigger transient changes in reservoir chemistry and higher destratification probability, we may expect more frequent and pronounced lotic conditions in reservoirs. This, in turn, suggests propagation of stormwater and the constituents it carries (e.g. nutrients, pollutants, organic matter) further distances through reservoirs during storms. In contrast, warmer temperatures that result from climate change will likely increase the strength of thermal stratification, decreasing the likelihood of destratification resulting from stormwater pulses. More persistent stratification may facilitate shunting of stormwater, as inflowing water will have less interaction with water of different density (Thornton et al. 1990). Less frequent destratification will also mean less mixing between epi- and hypolimnetic waters, which could generate or exacerbate hypoxic conditions and promote phosphorus availability in the cool, deeper waters. Given increasing urbanization and climate change, the footprint of inflowing stormwater in reservoirs is likely to increase while storm-driven destratification of the water column may be less likely.
Implications for reservoir ecosystem function and services
Incoming stormwater pulses can have significant impacts on the water quality and ecology of receiving reservoirs. Vanni et al. (2006) described the discharge of inflowing streams as the “master variable” determining nutrient limitation status of reservoir phytoplankton, and showed that storms drive a shift from nutrient- to light-limitation. Because stormwater tends to deliver both nutrients and sediment, stormwater pulses can increase the nutrient availability at the same time as reducing light availability. Furthermore, storm pulses may flush phytoplankton from the reservoir or mix them out of the epilimnion as the water column destratifies, reducing competition for nutrients and possibly temporarily increasing the favorability of conditions for denitrification. Storm pulses in reservoirs may also change the structure of phytoplankton communities (Edson and Jones 1988).
Stormwater pulses in reservoirs have implications not only for their ecology but also for their functionality as water resources for human use. In some cases, storms can trigger cyanobacterial blooms that release toxins into drinking water sources (Falconer and Humpage 2005). Dissolved organic carbon concentrations and loads tend to increase at higher flows in many catchments (e.g. Butturini et al., 2006; Dhillon & Inamdar, 2013; Hook & Yeakley, 2005; Inamdar & Mitchell, 2006; Raymond & Saiers, 2010) and can be transported to drinking water reservoirs, interfering with water treatment through the production of disinfection by-products (Leenheer 2004). Fallon et al. (2002) showed that pulses of the herbicides atrazine and deethylatrazine moved through a Kansas reservoir in a distinct pulse following a period of high rainfall. Storms also transport sediment to reservoirs (Knowlton and Jones 1995; Vanni et al. 2006). In addition to interfering with water clarity and the light environment, sediment loads are associated with high loads of pathogens (Kistemann et al. 2002) that may then move through reservoirs in distinct, coherent pulses (Brookes et al. 2005).