Search for: All records

Real-time control of stormwater systems can reduce flooding and improve water quality. Current industry real-time control strategies use simple rules based on water quantity parameters at a local scale. However, system-level control methods that also incorporate observations of water quality could provide improved control and performance. Therefore, the objective of this research is to evaluate the impact of local and system-level control approaches on flooding and sediment-related water quality in a stormwater system within the flood-prone coastal city of Norfolk, Virginia, USA. Deep reinforcement learning (RL), an emerging machine learning technique, is used to learn system-level control policies that attempt to balance flood mitigation and treatment of sediment. RL is compared to the conventional stormwater system and two methods of local-scale rule-based control: (i) industry standard predictive rule-based control with a fixed detention time and (ii) rules based on water quality observations. For the studied system, both methods of rule-based control improved water quality compared to the passive system, but increased total system flooding due to uncoordinated releases of stormwater. An RL agent learned controls that maintained target pond levels while reducing total system flooding by 4% compared to the passive system. When pre-trained from the RL agent that learned to reduce flooding, another RL agent was able to learn to decrease TSS export by an average of 52% compared to the passive system and with an average of 5% less flooding than the rule-based control methods. As the complexity of stormwater RTC implementations grows and climate change continues, system-level control approaches such as the RL used here will be needed to help mitigate flooding and protect water quality. 

Numerous studies have documented the linkages between agricultural nitrogen loads and surface water degradation. In contrast, potential water quality improvements due to agricultural best management practices are difficult to detect because of the confounding effect of background nitrate removal rates, as well as the groundwater‐driven delay between land surface action and stream response. To characterize background controls on nitrate removal in two agricultural catchments, we calibrated groundwater travel time distributions with subsurface environmental tracer data to quantify the lag time between historic agricultural inputs and measured baseflow nitrate. We then estimated spatially distributed loading to the water table from nitrate measurements at monitoring wells, using machine learning techniques to extrapolate the loading to unmonitored portions of the catchment to subsequently estimate catchment removal controls. Multiple models agree that in‐stream processes remove as much as 75% of incoming loads for one subcatchment while removing <20% of incoming loads for the other. The use of a spatially variable loading field did not result in meaningfully different optimized parameter estimates or model performance when compared with spatially constant loading derived directly from a county‐scale agricultural nitrogen budget. Although previous studies using individual well measurements have shown that subsurface denitrification due to contact with a reducing argillaceous confining unit plays an important role in nitrate removal, the catchment‐scale contribution of this process is difficult to quantify given the available data. Nonetheless, the study provides a baseline characterization of nitrate transport timescales and removal mechanisms that will support future efforts to detect water quality benefits from ongoing best management practice implementation.