Prioritizing Potential Aquaculture Zones on the West Coast of the United States

Background

Marine aquaculture has the potential to play an important role in the global food supply as a more sustainable protein option than land-based meat production. Gentry et al. mapped the potential for marine aquaculture globally based on multiple constraints, including ship traffic, dissolved oxygen, and bottom depth. They found that global seafood demand could be met using less than 0.015% of the global ocean area.

For this analysis, I will determine which Exclusive Economic Zones (EEZ) on the West Coast of the United States are best suited to developing marine aquaculture for several species of Oysters and White Shrimp. Suitable locations will be determined based on a range of suitable sea surface temperatures (SST) and depth values for the species.

Data on species depth and temperature requirements came from SeaLifeBase.

Data on sea surface temperatures came from NOAA’s 5km Daily Global Satellite Sea Surface Temperature Anomaly v3.1.

Bathymetry data came from the General Bathymetric Chart of the Oceans (GEBCO).

Data on Exclusive Economic Zones came from Marineregions.org.

Code

# Transform the CRS of the files to match
bathymetry <- terra::project(bathymetry, crs(sst_stack))

maritime_boundaries <- maritime_boundaries %>% 
  st_transform(crs = st_crs(bathymetry))

# Confirm that the CRS transformed as expected
test_that("The CRS of all data sets match", {
  expect_true(crs(sst_stack) == crs(bathymetry) && crs(bathymetry) == crs(maritime_boundaries))
})

Test passed 🥇

Data Processing

In order to perform this analysis, the data must be processed. Specifically, I must ensure that the Sea Surface Temperature data and Depth data can be combined, since the data have different resolutions, extents, and positions.

Code

# Find the mean SST from 2008-2012 (eg create single raster of average SST...)
sst_average <- mean(sst_stack)

# Convert average SST from Kelvin to Celsius by subtracting 273.15
sst_average_celsius <- sst_average - 273.15

After some initial processing of the SST raster stack, I want to take a quick look at the the bathymetry and SST data, to see what further processing might need to be done.

Code

# Plot bathymetry raster
plot_bathymetry <- ggplot() +
  geom_spatraster(data = bathymetry) +
  scale_fill_viridis_c(name = "Depth") +
  ggtitle("Bathymetry") +
  theme_minimal()

# Plot SST raster
plot_sst <- ggplot() +
  geom_spatraster(data = sst_average_celsius) +
  scale_fill_viridis_c(name = "Temperature (°C)") +
  ggtitle("Average SST (°C)") +
  theme_minimal()

# Combine the plots
plot_bathymetry + plot_sst

I know from the metadata that the resolutions of these raster objects do not match, so I decide to resample the bathymetry data to match the resolution of the SST data, using the nearest neighbor approach. Once the resolutions and CRS of the data matches, I can crop the bathymetry raster to match the extent of the SST raster.

Code

# Resample the bathymetry data to match the resolution of the SST data using the nearest neighbor approach
bathymetry_resampled <- resample(bathymetry, sst_stack, method = "near")

# Verify the results
test_that("The Resolutions of the Rasters match", {
    expect_true(res(bathymetry_resampled)[1] == res(sst_average_celsius)[1]) && 
      (res(bathymetry_resampled)[2] == res(sst_average_celsius)[2])
  })

Test passed 🎊

Code

# Crop the bathymetry raster at last
bathymetry_cropped <- crop(bathymetry_resampled, sst_average_celsius)

Now, to verify that my processing thus far has been successful, I will run a series of tests. One way to know that the rasters match, is if they are able to be successfully stacked together. This code will test for stackability, and then I will finally stack the two together.

Code

test_that("The resolutions of the cropped rasters match", {
  expect_true(all(res(bathymetry_cropped) == res(sst_average_celsius)))
})

Test passed 🌈

Code

test_that("The rasters have the same extent", {
  expect_true(ext(bathymetry_cropped) == ext(sst_average_celsius))
})

Test passed 🥇

Code

test_that("The rasters have the same CRS", {
  expect_true(crs(bathymetry_cropped) == crs(sst_average_celsius))
})

Test passed 😀

Code

# Finally, lets do it! Stack that raster!
bathy_sst <- c(bathymetry_cropped, sst_average_celsius)

Oysters: A “Pearl-fect” Match

After the data is properly prepared and processed, I can begin the analysis to find suitable locations for marine aquaculture. This means finding locations that are suitable in terms of both SST and depth.

I will begin my analysis with oysters. Research has shown that oysters need the following conditions for optimal growth:

sea surface temperature: 11-30°C
depth: 0-70 meters below sea level

With this in mind, I will first reclassify the SST data to only include raster values that are between 11 and 30 degrees Celsius. Using a similar process, I will reclassify the depth data to only include values that are above 70 meters below sea level. Let’s see what that looks like:

Code

# Turn unneeded values to NA
optimal_sst <- sst_average_celsius
optimal_sst[optimal_sst < 11 | optimal_sst > 30] <- NA

# Do it again for the bathymetry data
optimal_bath <- bathymetry_cropped
optimal_bath[optimal_bath < -70 | optimal_bath > 0] <- NA

# Plot optimal bathymetry raster
plot_optimal_bath <- ggplot() +
  geom_spatraster(data = optimal_bath) +
  scale_fill_viridis_c(name = "Depth (m)") +
  ggtitle("Optimal Depth (m) for Oysters") +
  theme_minimal()

# Plot optimal SST raster
plot_optimal_sst <- ggplot() +
  geom_spatraster(data = optimal_sst) +
  scale_fill_viridis_c(name = "Temperature (°C)") +
  ggtitle("Optimal SST (°C) for Oysters") +
  theme_minimal()

# Combine the plots
plot_optimal_bath + plot_optimal_sst

To find suitable locations for oyster marine aquaculture, we’ll need to find locations that are suitable in terms of both SST and depth. Let’s do it with code!

Code

# Use map algebra to find areas of overlap
overlap <- !is.na(optimal_sst) & !is.na(optimal_bath)

big_optimal <- optimal_sst * overlap + optimal_bath * overlap
plot(big_optimal, main = "Area Optimal for Oyster Growth")

The result we get is a map showing a relatively thin line representing the areas of the West Coast of the United States that have the most optimal conditions for oyster aquaculture.

Determine the most suitable EEZ (Exclusive Economic Zone) best suited for developing marine aquaculture for Oysters.

Now, its time to bring in our EEZ data, to determine which EEZs best overlap with our zones of optimal conditions for oysters.

First, lets take a look at the maritime boundaries data that we have.

Code

# Map the EEZs
tm_shape(maritime_boundaries) +
  tm_polygons(col = "rgn",
              palette = "RdYlBu",
              title = "Region Name") +
  tm_layout(main.title = "West Coast EEZs",
            main.title.position = "center",
            legend.outside = TRUE) +
  tm_shape(us_states) +
    tm_polygons(alpha = .20)

Now the question remains: what percent of each EEZ region contains suitable area for oyster cultivation? Using the optimal oyster conditions we already calculated, as well as the EEZ information we have in our dataframe, we can calculate this.

Our final product is a map of visualizations thus far, and a nice table that presents the final numbers from our analysis.

Code

# Let's make a plot!

# Map both raster data sets
final_map <- tm_shape(maritime_boundaries) +
  tm_polygons(col = "rgn",
              palette = "Blues",
              alpha = .60,
              title = "Region Name") +
  tm_layout() +
  tm_shape(us_states) +
    tm_polygons(alpha = .20) +  
   tm_shape(big_optimal) +
     tm_raster(legend.show = TRUE,
               title = "Total Suitable\nArea (km^2)",
               palette = "red",
               breaks = seq(0, 5, by = 5),
               labels = " ") +
     tm_layout(main.title = "EEZs with Optimal\nOyster Habitat",
               main.title.position = "center",
               legend.outside = TRUE)
print(final_map)

Code

# Rasterize the eez dataframe using the 'rgn' column
eez_rast <- terra::rasterize(maritime_boundaries, big_optimal, field = "rgn")

# Create dataframe that displays suitable area of each zone
suitable_area <- terra::expanse(big_optimal, unit = "km", zones = eez_rast)

# Rename columns
suitable_area <- suitable_area %>% 
  rename("suitable_area_km2" = "area",
         "rgn" = "zone")

# Calculate percent of each zone that contains suitable area for oyster aquaculture 
eez_oyster <- maritime_boundaries %>% 
  left_join(suitable_area, by= "rgn") %>% 
  mutate(zone_pct = suitable_area_km2 / area_km2 * 100) %>% 
  select("rgn", "area_km2", "suitable_area_km2", "zone_pct") %>% 
  st_drop_geometry()

# Make a fun table that shows this estimate
final_table <- eez_oyster |>
  gt() |>
  gt_theme_nytimes() |>
  tab_header(title = "EEZ Aquaculture Calculations") |>
  cols_label(
    rgn = "EEZ REgion",
    area_km2 = "Square Kilometers of EEZ",
    suitable_area_km2 = "Square Kilometers of Optimal Area",
    zone_pct = "Percentage of EEZ Suitable for Aquaculture"
  ) |>
  tab_style(
    style = cell_text(weight = "bold"),
    locations = cells_body()
  )

final_table

EEZ Aquaculture Calculations
EEZ REgion	Square Kilometers of EEZ	Square Kilometers of Optimal Area	Percentage of EEZ Suitable for Aquaculture
Oregon	179994.06	1120.146	0.6223238
Northern California	164378.81	194.169	0.1181229
Central California	202738.33	4086.765	2.0157832
Southern California	206860.78	3882.136	1.8766904
Washington	66898.31	2452.143	3.6654784

Aquaculture made “Shrimple”

According to Boyd et. all, one of the most common species of shrimp harvested using aquaculture is Penaeus setiferus, or White Shrimp. According to SeaLifeBase, this shrimp species requires the following conditions for optimal growth:

sea surface temperature: 15.3-27.3°C
depth: 0-119 meters below sea level

I want to run the same analysis for shrimps, that I did for oysters. In order to do that, I will create a function that utilizes the same steps above, but would only require me to input the parameters of the analysis, and result in a comprehensive map and table.

First I will create the function, called optimal_aquaculture:

Code

optimal_aquaculture <- function(sst_data, bathy_data, eez_data, species, min_temp, max_temp, min_depth, max_depth) {
  # Transform the CRS of the files to match
  bathymetry <- terra::project(bathy_data, crs(sst_data))

  maritime_boundaries <- eez_data %>% 
    st_transform(crs = st_crs(bathymetry))

  # Confirm that the CRS transformed as expected
  test_that("The CRS of all data sets match", {
    expect_true(crs(sst_data) == crs(bathymetry) && crs(bathymetry) == crs(maritime_boundaries))
  })

  # Find the mean SST from 2008-2012 (e.g., create single raster of average SST)
  sst_average <- mean(sst_data)

  # Convert average SST from Kelvin to Celsius by subtracting 273.15
  sst_average_celsius <- sst_average - 273.15

  # Resample the bathymetry data to match the resolution of the SST data using the nearest neighbor approach
  bathymetry_resampled <- resample(bathymetry, sst_average_celsius, method = "near")

  test_that("The Resolutions of the Rasters match", {
    expect_true(res(bathymetry_resampled)[1] == res(sst_average_celsius)[1]) && 
      (res(bathymetry_resampled)[2] == res(sst_average_celsius)[2])
  })

  # Crop the bathymetry raster
  bathymetry_cropped <- crop(bathymetry_resampled, sst_average_celsius)

  test_that("The resolutions of the cropped rasters match", {
    expect_true(all(res(bathymetry_cropped) == res(sst_average_celsius)))
  })

  test_that("The rasters have the same extent", {
    expect_true(ext(bathymetry_cropped) == ext(sst_average_celsius))
  })

  test_that("The rasters have the same CRS", {
    expect_true(crs(bathymetry_cropped) == crs(sst_average_celsius))
  })

  # As a final check, stack the rasters
  bathy_sst <- c(bathymetry_cropped, sst_average_celsius)

  # Turn unneeded values to NA
  optimal_sst <- sst_average_celsius
  optimal_sst[optimal_sst < min_temp | optimal_sst > max_temp] <- NA

  # Do it again for the bathymetry data
  optimal_bath <- bathymetry_cropped
  optimal_bath[optimal_bath < max_depth | optimal_bath > min_depth] <- NA

  # Use map algebra to find areas of overlap
  overlap <- !is.na(optimal_sst) & !is.na(optimal_bath)
  big_optimal <- optimal_sst * overlap + optimal_bath * overlap

  # Map both raster data sets
  final_map <- tm_shape(eez_data) +
    tm_polygons(col = "rgn",
                palette = "Blues",
                alpha = .60,
                title = "Region Name") +
    tm_layout() +
    tm_shape(us_states) +
      tm_polygons(alpha = .20) +  
    tm_shape(big_optimal) +
      tm_raster(legend.show = TRUE,
                title = "Total Suitable\nArea (km^2)",
                palette = "red",
                breaks = seq(0, 5, by = 5),
                labels = " ") +
      tm_layout(main.title = paste("EEZs with Optimal\n", species, "\nHabitat"),
                main.title.size = .90,
                main.title.position = "center",
                legend.outside = TRUE)
  print(final_map)

  # Rasterize the EEZ dataframe using the 'rgn' column
  eez_rast <- terra::rasterize(eez_data, big_optimal, field = "rgn")

  # Create a dataframe that displays the suitable area of each zone
  suitable_area <- terra::expanse(big_optimal, unit = "km", zones = eez_rast)

  # Rename columns
  suitable_area <- suitable_area %>% 
    rename("suitable_area_km2" = "area",
           "rgn" = "zone")

  # Calculate the percent of each zone that contains suitable area for oyster aquaculture 
  eez_species <- eez_data %>% 
    left_join(suitable_area, by = "rgn") %>% 
    mutate(zone_pct = suitable_area_km2 / area_km2 * 100) %>% 
    select("rgn", "area_km2", "suitable_area_km2", "zone_pct") %>% 
    st_drop_geometry()

  # Create a table to show this estimate
  final_table <- eez_species |>
    gt() |>
    gt_theme_nytimes() |>
    tab_header(title = paste("EEZ Aquaculture Calculations for", species)) |>
    cols_label(
      rgn = "EEZ Region",
      area_km2 = "Square Kilometers of EEZ",
      suitable_area_km2 = "Square Kilometers of Optimal Area",
      zone_pct = "Percentage of EEZ Suitable for Aquaculture"
    ) |>
    tab_style(
      style = cell_text(weight = "bold"),
      locations = cells_body()
    )

  return(final_table)
}

And then I will use it to run an analysis of optimal conditions for White Shrimp aquaculture:

Code

optimal_aquaculture(sst_data = sst_stack, bathy_data = bathymetry, eez_data = maritime_boundaries, species = "White Shrimp", min_temp = 15.3, max_temp = 27.3, min_depth = 0, max_depth = -119)

Test passed 🎊
Test passed 🌈
Test passed 🌈
Test passed 😸
Test passed 🥇

EEZ Aquaculture Calculations for White Shrimp
EEZ Region	Square Kilometers of EEZ	Square Kilometers of Optimal Area	Percentage of EEZ Suitable for Aquaculture
Oregon	179994.06	NA	NA
Northern California	164378.81	NA	NA
Central California	202738.33	NA	NA
Southern California	206860.78	3292.186	1.591498
Washington	66898.31	NA	NA

While obviously useful to analyzing spatial data for shrimp, this function can be used for any variety of cultivated seafoods, so long as the parameters are known and the data is available. This White Shrimp analysis shows us that optimal conditions for the cultivation of this food can only happen in Southern California. Information such as this can lead to the prioritization and higher likelihood of success for aquaculture planning and projects.

Conclusion

Aquaculture will continue to have a vital role to play in food cultivation, especially in the face of modern climate stressors and human population demands. Analysis such as this one showcase the important role that data science and spatial data analysis have to play in understanding this important food growing practice.

Citations

Boyd, Claude E., and Jason W. Clay. “Shrimp Aquaculture and the Environment.” Scientific American 278, no. 6 (1998): 58–65. http://www.jstor.org/stable/26057855.

Hall, S. J., Delaporte, A., Phillips, M. J., Beveridge, M. & O’Keefe, M. Blue Frontiers: Managing the Environmental Costs of Aquaculture (The WorldFish Center, Penang, Malaysia, 2011).

Gentry, R. R., Froehlich, H. E., Grimm, D., Kareiva, P., Parke, M., Rust, M., Gaines, S. D., & Halpern, B. S. Mapping the global potential for marine aquaculture. Nature Ecology & Evolution, 1, 1317-1324 (2017).

GEBCO Compilation Group (2022) GEBCO_2022 Grid (doi:10.5285/e0f0bb80-ab44-2739-e053-6c86abc0289c).

NOAA Coral Reef Watch. “Daily Global 5km Satellite Sea Surface Temperature Anomaly (Version 3.1, Released August 1, 2018).” Accessed December 3, 2024. https://coralreefwatch.noaa.gov/product/5km/index_5km_ssta.php.

R. Oliver, EDS 223 - Geospatial Analysis and Remote Sensing, Course Notes. 2024. [Online]. Available: https://eds-223-geospatial.github.io/

SeaLifeBase. “Penaeus setiferus (Linnaeus, 1767) - Northern White Shrimp.” Accessed December 3, 2024. https://sealifebase.org.

SeaLifeBase. “Ostrea edulis (Linnaeus, 1758) - Edible Oyster.” Accessed December 3, 2024. https://sealifebase.org.

Acknowledgements

Special shoutout to Jordan Sibley, who helped me debug a particularly fraught piece of code in this analysis.

Code

knitr::include_graphics("data/IMG_20241203_193746.jpg")

Citation

BibTeX citation:

@online{jørgensen2024,
  author = {Jørgensen, Bailey},
  title = {Prioritizing {Potential} {Aquaculture} {Zones} on the {West}
    {Coast} of the {United} {States}},
  date = {2024-12-20},
  url = {https://jorb1.github.io/posts/2024-12-20-aquaculture/},
  langid = {en}
}

For attribution, please cite this work as:

Jørgensen, Bailey. 2024. “Prioritizing Potential Aquaculture Zones on the West Coast of the United States.” December 20, 2024. https://jorb1.github.io/posts/2024-12-20-aquaculture/.