Using empirical dynamic modeling to assess relationships between atmospheric trace gases and eukaryotic phytoplankton populations in coastal Southern California
Introduction
Phytoplankton can affect the atmospheric mixing ratios of a range of trace gases directly through production and indirectly by contributing biomass that is broken down during senescence and death. It is known that phytoplankton can produce halogenated volatile organic compounds (VOCs), a class of trace gases that are of interest due to their effects on tropospheric chemistry and their ability to destroy ozone when transported to the stratosphere (WMO, 2018). Specifically, a variety of marine phytoplankton and macrophytes have been shown to emit methyl bromide (CH3Br), dibromomethane (CH2Br2), bromoform (CHBr3), methyl chloride (CH3Cl), chloroform (CHCl3), and methyl iodide (CH3I) in the lab (Manley and Dastoor, 1988; Scarratt and Moore, 1996, Scarratt and Moore, 1998; Moore, 2003; Colomb et al., 2008; Brownell et al., 2010; Paul and Pohnert, 2011; Johnson et al., 2015; Lim et al., 2018) and at offshore sites (Baker et al., 2000; Moore and Tokarczyk, 1993; Moore, 2003). While in situ studies relating specific nearshore coastal marine phytoplankton to atmospheric mixing ratios of these trace gases are notably lacking, previous work in the open ocean has linked CH3Br, CH2Br2, CHBr3, CH3Cl, CHCl3, and CH3I to chlorophyll a concentration and phytoplankton type (Arnold et al., 2010).
The inorganic compound carbonyl sulfide (COS) is another trace gas that is linked to phytoplankton. COS is the most abundant sulfur-containing gas in the atmosphere and can indirectly affect climate by contributing to the stratospheric sulfate pool through interactions with UV light and free radicals (Carpenter et al., 2014; Whelan et al., 2018). COS from the marine environment is produced by a combination of sources, including salt marsh vegetation (Commane et al., 2013; Whelan et al., 2013) and the photochemical oxidation of dissolved organic sulfur compounds (Andreae and Ferek, 1992), mainly dimethylsulfide and the organic building block carbon disulfide (Andreae and Crutzen, 1997; Kettle et al., 2002). Thus, COS production is indirectly related to phytoplankton activities and it has also been observed to be consumed at low levels by marine algae (Blezinger et al., 2000).
The aforementioned studies either took place in laboratory settings or are based on relatively few data points during fairly short periods of fieldwork, limiting the projection of these data to other settings or the wider ecosystem. The lack of long-term monitoring has made it difficult to establish clear links between natural populations of phytoplankton and atmospheric trace gases. Long-standing time-series provide a unique framework in which to analyze and statistically relate many different variables in an open system and simultaneously allow for baseline observations, an understanding of natural cycles (e.g., seasonality, upwelling, etc.), and an understanding of anthropogenic change. By studying variables that are not removed from natural ecological interactions or environmental perturbations, conclusions regarding connections between different variables are more accurate.
We took advantage of several long-standing time-series collected from the Ellen Browning Scripps Pier (Scripps Pier), located within the Southern California Bight ecosystem, to establish causal links between naturally occurring and ecologically active (changing due to natural abiotic and biotic pressures) eukaryotic phytoplankton populations and atmospheric trace gases over 8 years. Our overarching goal was to assess how atmospheric mixing ratios of halocarbons and COS may be affected by natural populations of ecologically active marine eukaryotic phytoplankton through time. We did not seek to explain all of the variation observed in any of the trace gas mixing ratios, nor in the phytoplankton populations. Rather, we sought to assess whether changes in groups of specific phytoplankton led to time-lagged changes in trace gases at certain times. Many of the time-series displayed strong nonlinear dynamics (meaning that the time-series could be correlated at some points in time and not others) and so we utilized 1) a self-organizing map (SOM, Kohonen, 2001) to segment and characterize the abiotic portion of the ecosystem state (e.g. water temperature), and 2) empirical dynamic modeling (EDM) with convergent cross mapping (CCM) to assess causal links between individual phytoplankton and trace gases. This latter method has been used for a wide range of applications including identifying drivers of global influenza cases (Deyle et al., 2016) and controls on fish recruitment (Deyle et al., 2018). EDM with CCM has also been used to predict coastal algal blooms in Southern California (McGowan et al., 2017). EDM with CCM is particularly useful to our purposes because it can show causality between nonlinear time-series that may correlate during certain system states (all the ecological and abiotic factors that describe a regime or ecosystem state) but not in others, and can recreate system states without input for all the variables naturally acting on a time-series (Sugihara et al., 2012). Put another way, ecological and environmental interactions may only be present during some time-periods and CCM allowed us to assess causal links between individual phytoplankton taxa and trace gases without data describing every variable present in our system (e.g., terrestrial biomass production of a trace gas, atmospheric transport, etc.). However, the link between phytoplankton and trace gases may not be direct (e.g., trace gases could be indirectly linked to phytoplankton community structure through the microbial degradation of senescent or dead phytoplankton). Therefore, we used CCM to show which populations may have a direct or indirect link to atmospheric trace gas mixing ratios. Finally, while a SOM does not show causality, it can characterize different ecosystem states for the variables that were used and thus allowed us to assess if there were certain abiotic ecosystem states during which phytoplankton and traces gases were causally linked.
We had two broad expectations going into this study. First, because many eukaryotic phytoplankton and trace gases follow seasonal cycles, we expected that annually seasonal phytoplankton would causally link to trace gases that followed corresponding seasonal cycles. Second, we anticipated causal links between stochastic phytoplankton and trace gases with reduced seasonality. Regardless of our expectations, establishing connections between phytoplankton and trace gases (and the ecosystem states that those links exist within) is the first step in assessing how direct and/or indirect phytoplankton production of these trace gases are affected by long-term sustained ecosystem shifts or changes in specific seasonal factors.
Section snippets
Methods
All data were collected off the Scripps Pier at Scripps Institution of Oceanography (SIO; 32.8663° N, 117.2546° W). SIO is located in La Jolla, CA within the often-studied Southern California Bight ecosystem (Di Lorenzo, 2003; Tai and Palenik, 2009; McGowan et al., 2017). Local conditions are influenced by the adjacent California Current ecosystem, with offshore upwelling typically occurring in summer months (Bakun, 1973).
The Advanced Global Atmospheric Gases Experiment (AGAGE) measures a range
Variation in trace gases, environmental variables, and eukaryotic phytoplankton
All of the observed trace gases displayed significant variation over the time-series and were sorted into three groups based on seasonality: 1) visible annual seasonality, 2) visible annual seasonality with an aseasonal feature, and 3) multiple-seasonality (with reliable peaks in multiple seasons) to no seasonality. CH3Cl, CH3I, and COS each displayed strong seasonality with maxima in one season: winter/spring for CH3Cl and summer for CH3I and COS (Fig. 1D, F, G). CH2Br2 and CHBr3 also
Discussion
Through the use of EDM with CCM we were able to show that changes in several species, genera, and types of eukaryotic phytoplankton led to changes in select trace gases in a nearshore Southern California coastal environment. Though ρ values for various phytoplankton affecting trace gases were low, our findings demonstrate that phytoplankton community composition has an impact on atmospheric chemistry in the marine boundary layer. Phytoplankton influence trace gas mixing ratios through direct
Conclusion
This work demonstrated that naturally occurring, ecologically active eukaryotic phytoplankton affect a range of trace gases in coastal Southern California on scales of 0 to 3 weeks during different times of the year. We utilized long-standing in situ observations to capture the effects of naturally occurring environmental or ecological interactions and applied causality testing to assess directionality of relationships. The method successfully identified some causal interactions between coastal
Declaration of Competing Interest
None.
Acknowledgements
The authors wish to specifically acknowledge Kristi Seech and Mary Hilbern for their efforts on phytoplankton cell identification and enumeration, James Fumo for data integration, John A. McGowan for support and leadership over the decades with the McGowan Plankton and Chlorophyll Program (Funding provided by private donors and the MacArthur Foundation) and Southern California Coastal Ocean Observing Harmful Algal Bloom Monitoring Program (NOAA NA16NOS0120022, NA11NOS120029, and NA17RJ1231).
References (81)
- et al.
Emissions of CH3Br, organochlorines, and organoiodines from temperate macroalgae
Chemosph.-Glob. Chang. Sci.
(2001) - et al.
The emergence of Cochlodinium along the California Coast (USA)
Harmful Algae
(2008) Seasonal dynamics of the surface circulation in the Southern California current system
Deep. Res. Part II Top. Stud. Oceanogr.
(2003)- et al.
Ba/ca variations in the modern intertidal bean clam Donax gouldii: an upwelling proxy?
Palaeogeogr. Palaeoclimatol. Palaeoecol.
(2013) - et al.
Decomposition of brominated organic halogens by cultures of marine proteobacteria: Phaeobacter, Roseobacter, and Rhodobacter
Mar. Chem.
(2015) - et al.
Atmospheric methyl chloride
Atmos. Environ.
(1999) - et al.
Brominated methane compounds and isoprene in surface seawater of Sagami Bay: concentrations, fluxes, and relationships with phytoplankton assemblages
Mar. Chem.
(2012) - et al.
Halogen chemistry of the red alga Asparagopsis
Phytochemistry
(1977) Short-term variability in halocarbons in relation to phytoplankton pigments in coastal waters of the central eastern Arabian Sea
Estuar. Coast. Shelf Sci.
(2010)- et al.
Production of methyl chloride and methyl bromide in laboratory cultures of marine phytoplankton
Mar. Chem.
(1996)