The Role of Ocean Currents in the Temperature Selection of Plankton: Insights from an Individual-Based Model

Abstract

Biogeography studies that correlate the observed distribution of organisms to environmental variables are typically based on local conditions. However, in cases with substantial translocation, like planktonic organisms carried by ocean currents, selection may happen upstream and local environmental factors may not be representative of those that shaped the local population. Here we use an individual-based model of microbes in the global surface ocean to explore this effect for temperature. We simulate up to 25 million individual cells belonging to up to 50 species with different temperature optima. Microbes are moved around the globe based on a hydrodynamic model, and grow and die based on local temperature. We quantify the role of currents using the "advective temperature differential" metric, which is the optimum temperature of the most abundant species from the model with advection minus that from the model without advection. This differential depends on the location and can be up to 4°C. Poleward-flowing currents, like the Gulf Stream, generally experience cooling and the differential is positive. We apply our results to three global datasets. For observations of optimum growth temperature of phytoplankton, accounting for the effect of currents leads to a slightly better agreement with observations, but there is large variability and the improvement is not statistically significant. For observed Prochlorococcus ecotype ratios and metagenome nucleotide divergence, accounting for advection improves the correlation significantly, especially in areas with relatively strong poleward or equatorward currents.

Fig 1. Growth rate vs. temperature.

Optimum temperature (T_opt) = Red: 18, Blue: 21. See S1 File for parameter values.

Fig 2. Time series of model results.

Local temperature (T_loc), optimum temperatures of the most abundant species in simulations with (T_opt(a)) and without(T_opt(na)) advection and historical temperature based on growth (T_hist(g)) and selection (T_hist(s)) rates at two locations in the Gulf Stream (GSS and GSN in Fig 3). Population average growth rate = 0.14 d^-1.

Fig 3. Map of model results.

(A) Advective temperature differential (ΔT_opt) across the global ocean, defined as the difference between optimum temperatures of the most abundant species in simulation with and without advective transport. Population average growth rate = 0.14 d^-1. Values are averages over the 31-year simulation period. Also shown are locations used in Figs 2 and 4 (pink circles and text), isolation locations for phytoplankton in Fig 5 (dark green triangles), Prochlorococcus ecotypes in Fig 6 (light green squares, open symbols are for Gulf Stream (GS) and eastern North Atlantic (ENA) samples, also identified in Fig 6.) and metagenome nucleotide divergence in Fig 7 (medium green circles, samples GS008 and GS366 are labelled). Approx. location of select currents (white arrows and white bold text). GS = Gulf Stream, FC = Falkland Current, CC = Canary Current, KC = Kuroshio Current. (B) Poleward velocity. See S1 File for a larger version of the North Atlantic.

Fig 4. Direction and magnitude of advective temperature differential depends on location and growth rate.

Difference of optimum temperature of the most abundant species in simulations with and without advection (ΔT_opt) for locations in the Gulf Stream (GSN in Fig 3), Bermuda Atlantic Time Series (BATS in Fig 3) and Falkland Current (FC in Fig 3) for different growth rates. Values are averages over the 31-year simulation period.

Fig 5. Model–data comparison for phytoplankton optimum temperatures.

Predicted vs. observed optimum temperatures for the datasets of Thomas et al. [9] and Chen et al. [10]. (A) Prediction is local temperature. (B) Prediction from model without advection. (C) Prediction from model with advection. Numbers on panels are RMSEs (±SD, bootstrap analysis, n = 1,000).

Fig 6. Model–data comparison for Prochlorococcus ecotypes.

(A) Regression of observed log ecotype ratio to local and corrected temperatures. The corrected temperature was calculated as the local temperature (T_loc) plus the advective temperature differential (ΔT_opt) from the atlas (see SI). Open symbols are for Gulf Stream (GS) and eastern North Atlantic (ENA) samples, also identified in Fig 3. (B) Direct prediction of log ecotype ratio using ecotype-specific growth rate vs. temperature functions (see SI). (B1) Observed vs. modeled log ecotype ratio. (B2) Observed and modeled log ecotype ratio vs. time at BATS. Numbers on panel A are R²s (±SD, bootstrap analysis, n = 1,000) and B1 are RMSEs (±SD, bootstrap analysis, n = 1,000). Numbers on panel B2 are parameters of sine curve fit to log ecotype ratios in observed (Data), model with advection (Model(a)) and model without advection (Model(na)). Lag is relative to temperature.

Fig 7. Model–data comparison for metagenome nucleotide divergence.

Regression of average nucleotide divergence (AND) to absolute temperature difference of samples based on (A) local and (B) corrected temperatures. The corrected temperature was calculated as the local temperature (T_loc) plus the advective temperature differential (ΔT_opt) from the atlas (see SI). Population average growth rate = 0.14 d^-1. Numbers on panels are R²s (±SD, bootstrap analysis, n = 1,000).