Assessing drivers of coastal phytoplankton productivity in northern Marguerite Bay, Antarctica

The coastal ocean of the climatically-sensitive west Antarctic Peninsula is experiencing changes in the physical and (photo)chemical properties that strongly affect the phytoplankton. Consequently, a shift from diatoms, pivotal in the Antarctic food web, to more mobile and smaller flagellates has been observed. We seek to identify the main drivers behind primary production (PP) without any assumptions beforehand to obtain the best possible model of PP. We employed a combination of field measurements and modeling to discern and quantify the influences of variability in physical, (photo)chemical, and biological parameters on PP in northern Marguerite Bay. Field data of high-temporal resolution (November 2013–March 2014) collected at a long-term monitoring site here were combined with estimates of PP derived from photosynthesis-irradiance incubations and modeled using mechanistic and statistical models. Daily PP varied greatly and averaged 1,764 mg C m−2 d−1 with a maximum of 6,908 mg C m−2 d−1 after the melting of sea ice and the likely release of diatoms concentrated therein. A non-assumptive random forest model (RF) with all possibly relevant parameters (MRFmax) showed that variability in PP was best explained by light availability and chlorophyll a followed by physical (temperature, mixed layer depth, and salinity) and chemical (phosphate, total nitrogen, and silicate) water column properties. The predictive power from the relative abundances of diatoms, cryptophytes, and haptophytes (as determined by pigment fingerprinting) to PP was minimal. However, the variability in PP due to changes in species composition was most likely underestimated due to the contrasting strategies of these phytoplankton groups as we observed significant negative relations between PP and the relative abundance of flagellates groups. Our reduced model (MRFmin) showed how light availability, chlorophyll a, and total nitrogen concentrations can be used to obtain the best estimate of PP (R2 = 0.93). The resulting estimates from our models suggest summer PP to have been between 214.4 and 176.1 g C m−2. Through the employment of a modeling technique without any assumptions apart from a representative sampling strategy, we showed and estimated how PP in this climatically sensitive and changing region can best be predicted and described