Inferring controls on dissolved oxygen criterion attainment in the Chesapeake Bay

Environmental monitoring programs generate multivariate time series for the assessment of ecosystem health. Recent developments in causal inference offer ways to translate these observational data into networks able to explain gains and losses in the trajectories of indicator variables. Here, we present a case study of this approach using surface water dissolved oxygen (DO) criteria attainment across the Chesapeake Bay. We sought to understand the long-term fluctuations of DO in response to external watershed inputs (freshwater, nutrients, sediments) and internal water properties (water temperature, chlorophyll-a, nutrient concentrations). We found contrasting controls across different regions of the Chesapeake Bay. Summer freshwater and sediment inputs reduced surface water DO criteria attainment in landward regions but elevated attainment in open, mainstem waters. Algal biomass was often positively associated with AD in surface waters, in contrast to deep waters where algae are well-understood to lower oxygen levels. Using estimates of net effects, we also discovered that sediment flowing into many of the mainstem segments during the winter-spring season has had a net negative effect on oxygen levels. These segments have yet to deteriorate, which has masked this risk. This work demonstrates the utility of long-term monitoring programs to better understand and manage complex ecosystems.

Keywords: Chesapeake Bay, Water quality, Dissolved oxygen, Causal inference, Network, Management