Contents

Leveraging satellite data and pangeo to investigate the role of Gulf Stream frontal dynamics in ocean productivity

Authors

  • Author1 = {“name”: “Patrick Clifton Gray”, “affiliation”: “Duke University Marine Lab”, “email”: “patrick.c.gray@duke.edu”, “orcid”: “0000-0002-8997-5255”}

  • Author2 = {“name”: “David W. Johnston”, “affiliation”: “Duke University Marine Lab”, “email”: “david.johnston@duke.edu”, “orcid”: “0000-0003-2424-036X”}

Table of Contents

In a vessel floating on the Gulf Stream one sees nothing of the current and knows nothing but what experience tells him; but to be anchored in its depths far out of the sight of land, and to see the mighty torrent rushing past at a speed of miles per hour, day after day and day after day, one begins to think that all the wonders of the earth combined can not equal this one river in the ocean.” - J. E. Pillsbury, The Gulf Stream (1891)

Purpose

The Gulf Stream is a dominant physical feature in the North Atlantic, influencing the marine ecosystem across trophic levels, yet the physical-biological interactions along this western boundary current are poorly understood. In this work we analyze Gulf Stream physical and biological patterns over multiple years with particular emphasis on how the current’s front impacts biology across space and time. We ask specifically how phytoplankton patterns along the front correspond to seasonality and wind with speculation on possible nutrient limitations. This work primarily uses satellite-based sea surface temperature (SST), chlorophyll-a (chla), and sea surface height (SSH) data. Using xarray and running in jupyter in the cloud allows efficient analysis across many years of satellite data and holoviz enables interactive visualization and data exploration. Basing this work on a pangeo Docker image allows easy reproducibility and a robust set of packages for scientific computing. This work aims to provide insight into the overall role of Gulf Stream frontal features and dynamics in ocean productivity and increase the ease of access, manipulation, and visualization of these highly dimensional and complex datasets.

Technical contributions

  • collation and visualization of a variety of oceanographic datasets permitting investigation of the Gulf Stream front

  • derivation of the Gulf Stream front location every month for a decade from Sea Surface Height data

  • demonstration of an array of scientific computing tools for oceanography analysis

  • investigation of the physical and biological properties of the Gulf Stream front over a decade

Methodology

The goal of this analysis is to investigate the physical and biological trends and relationships along the Gulf Stream front using SST, wind, and chla data. For comparison we also investigate that same data on the coastal and open ocean side of the front. The first half of the notebook pulls in and visualizes all the different datasets and then derives the front location based on the 0.25m isoline in sea surface height data. The final step in the first half of the notebook is to select data from the main datasets (SST, wind, chla) using these front lines at every time step.

The second half of the notebook then analyzes this along-front data using a variety of visualizations and interprets the results.

Results

Results are discussed in detail throughout the analysis section. Some main takeaways are that for nearly all of the Gulf Stream, from Florida out into the North Atlantic, there is a fall bloom that is comparable or greater than the spring bloom and this is true on both sides of the front.

The northern region (further downstream) has a major increase in chla in the spring - the well studied spring bloom - and the region further south (upstream) around the South Atlantic Bight (SAB) has a relatively large increase throughout the whole winter, starting mid fall and staying high until the spring. The region around Cape Hatteras falls in the middle of these two patterns, but with generally higher chla, particularly on the coastal side, likely due to proximity to the coast and generally large outflows from Pamlico Sound and Chesapeake Bay injecting nutrients into the water. The sustained spikes in chla off Cape Hatteras near the front on the coastal side are likely drivers of the increased biodiversity in that region. It is also worth noting that the northerly section of the front (downstream) has a more notable mid-winter decrease, possibly when light begins to be limiting, and this is not as pronounced in the southerly section of the front (upstream). Though the light difference is not extreme. There are 9.5 hours of sunlight on the winter solstice at 37° N and 10.6 hours at 27° N.

The satellite imagery makes it clear that the Gulf Stream is a major boundary between the coastal and the open ocean ecosystems. A large increase in chla beyond the Gulf Stream front, in the Sargasso Sea, is evident from late fall to early spring. The spring bloom in the North Atlantic is also clear where it peaks in April and May.

The SST data matches what one might expect, decreasing throughout the winter and increasing in summer with the Gulf Stream itself staying as the warmest water throughout all seasons. We can also see a few eddy footprints in SST and SSH even though this data is resampled to monthly time steps.

Overall the analysis generally agrees with the understanding of phytoplankton dynamics in the Atlantic (Siegel et al 2002, Fischer et al 2015) but it doesn’t follow exactly along with the commonly studied bloom in the North Atlantic. Fall production is more intense along the Gulf Stream compared to the North Atlantic, potentially supplied with nutrients from mesoscale upwelling (Pascual et al 2014) and ageostrophic circulation at the front (Levy et al 2018) and supplied with seed populations from further south. This is more in agreement with ideas of winter biomass accumulation despite low light conditions (e.g. Boss and Behrenfeld 2010). Wind patterns are noisy and don’t lead to clear agreement or disagreement with the general assumptions that winter winds drive deeper mixing which inject nutrients supplying the bloom and stronger winds in the spring deepen mixed layers which delay the start of the subpolar spring (Ueyama and Monger 2005, Henson et al 2009), but this is an avenue for further investigation with additional wind datasets.

Funding

Keywords

keywords=[“gulf stream”, “satellite oceanography”, “xarray”, “physical-biological”, “ocean productivity”]

Citation

Gray, P.C., & Johnston D.W., 2021. Leveraging satellite data and pangeo to investigate the role of Gulf Stream frontal dynamics in ocean productivity. Accessed 5/15/2021 at https://github.com/patrickcgray/gulf_stream_productivity_dynamics

Acknowledgements

We acknowledge helpful conversations with many collaborators and mentors as well as the substantial effort of the open source community developing xarray, zarr, holoviz, geopandas, jupyter and the many other libraries we relied on for this analysis.

Setup

Library import

# Data manipulation
import pandas as pd
import geopandas as gpd
import numpy as np
import xarray as xr

# time
import datetime

# downloading data from GDrive
from google_drive_downloader import GoogleDriveDownloader as gdd

# Options for pandas
pd.options.display.max_columns = 50
pd.options.display.max_rows = 30

# Visualizations
import hvplot.xarray
import hvplot.pandas
import matplotlib.pyplot as plt
from matplotlib.colors import LogNorm

# geographic and geometry tools
import cartopy.crs as crs
from shapely.ops import unary_union
from shapely.ops import transform
from shapely.geometry import Polygon
from affine import Affine

# there are many divide by zero warnings that we're going to suppress
# but advised to comment this line out when actively coding
import warnings
warnings.filterwarnings('ignore')