Use of continuous, real‐time observations and model simulations to achieve autonomous, adaptive sampling of microbial processes with a robotic sampler

摘要

Abstract The Columbia River has a dynamic and fast flushing estuary impacted by strong advection and mixing of riverine and oceanic waters, and high but variable loads of suspended particulate matter. Transient, but recurring water and nutrient fluxes from end‐members impart strong spatial and temporal gradients, contributing to microbiological hotspots that play important ecological roles in the estuary. Investigations of corresponding microbiota require precisely timed samples that are contextualized by physical and biogeochemical data. To accomplish this, we embedded a robotic microbial sampler (Environmental Sample Processor, ESP) within the operations of an interdisciplinary observation and prediction system (Science and Technology University Research Network, SATURN; www.stccmop.org/saturn ). Autonomous, adaptively sampled water collection by the ESP was implemented based on environmental conditions assessed from SATURN physical and biogeochemical sensors. Water was pumped from multiple depths to sensors and the ESP on dry land. If water met user‐defined parameters, ESP sampling was automatically initiated. This strategy was tested during three deployments in summer 2013, during which operational tools for analysis and visualization were used to formulate well‐constrained mission plans by providing estimates of the intensity and timing of oxygen‐depleted ocean water intrusion and estuarine turbidity maxima. This allowed the effective characterization of the impact of these events on selected estuarine microbiota.