UC Santa Cruz

Understanding spatial variability of water quality in estuary systems is important for making monitoring decisions and designing sampling strategies. In San Francisco Bay, the largest estuary system on the west coast of North America, tracking the concentration of suspended materials in water is largely limited to point measurements with the assumption that each point is representative of its surrounding area. In this study, we 1) quantify spatial variability in Suspended Particulate Matter (SPM) concentrations as a proxy for water quality at different spatial scales to contextualize this assumption and 2) demonstrate the potential of satellite and shipboard remote sensing to supplement current monitoring methods. We collected radiometric data from the bow of a research vessel on three dates in 2019 corresponding to satellite overpasses by Sentinel-2. Using our ship-based data, we tracked the location of a low-salinity, high-turbidity zone to show that remote sensing of SPM can inform on physical environmental conditions. We found that features exist that are not picked up by current point sampling, which prompted us to examine how much variability exists at spatial scales between 20 m and 10 km in San Francisco Bay using 10 m resolution Sentinel-2 imagery. We found 23%-80% variability in SPM at the 5km scale (the scale at which point sampling occurs), demonstrating the risk in assuming a single measurement is representative of a 5km area. In addition, current monitoring takes place along a transect within the Bay’s main shipping channel, which we show underestimates the spatial variance of the full bay. Our results suggest that spatial structure and spatial variability in the Bay change seasonally based on freshwater inflow to the Bay, tidal state, and wind speed. We recommend monitoring programs take this into account when designing sampling strategies, and that end-users account for the inherent spatial uncertainty associated with the resolution at which data is collected. This analysis also highlights the applicability of remotely sensed data to augment traditional sampling strategies.