Dissolved oxygen from 20 lake observatories: Changing drivers from minutes to months

Background/Question/Methods Dissolved  oxygen (DO) time dynamics in lakes reflect a complex suite of physical, chemical, and biological controls. Some controls respond directly to exogenous drivers. For example, changing weather patterns can alter lake hydrology and lake mixing, which can disrupt the mass balance of DO and its distribution within the water column. Other controls are more endogenous in nature and reflect, for example, changing biomass in primary producers that alter DO concentrations through metabolic pathways. Long-term time series of DO contain evidence of the relative importance of internal and external control of lakes. How does that balance of control shift over broad time scales? What lake attributes determine how DO responds to exogenous drivers? We answer these questions by analyzing sensor network data from 20 lakes that are part of the Global Lake Ecological Observatory Network. We use a model aggregation technique, bootstrap aggregating (or “bagging”), to search simultaneously model and parameter space within a time series family of models. Predictor variables include sensor and manually sampled limnological variables as well as meteorological variables.

Results/Conclusions Larger lakes had DO signals with much less noise at shorter time scales, with shorter term variation driven by irradiance-mediated metabolism. Smaller lakes tended to show strong responses to external perturbations, such as wind events corresponding to weather fronts. DO in smaller lakes also showed evidence of internal waves at the sub-diel scale. In lakes with pigment sensors that indicate phytoplankton biomass, there was correspondence between pigments and DO at diel scales. However, DO at scales of weeks did not correspond to pigments, suggesting other processes are in control. In many lakes, no variables explained DO dynamics at moderate time scales of days to weeks. Clearly, control over observed variance in DO depends on time scale and lake characteristics. Scale-dependent interpretation can provide insights into the circumstances under which physical or chemical processes control observations, and how those processes respond to exogenous drivers, such as weather events.