Big Ships, Little Plankton.
Rerouting Shipping Vessels to Avoid Zooplankton
What are zooplankton?
Zooplankton refers collectively to the species making up the small (<2cm) animal component of the planktonic community. They can be either adult or larval staged organisms.
Zooplankton diversity is immense, with over 10,000 species classified (Stromberg et al., 2009). These species are important for transferring energy from the photosynthesizing phytoplankton to the larger secondary consumers (EPA, 2021).
left: larval sea star by Richard Kirby
Why should we care?
Plankton specialists around the world have been stressing the impact of declining plankton communities within the 21st century (see this , and this , and even this ). One article from CBC (Adey, 2018) quoted a 50% decline in overall plankton biomass on Canada's East coast in recent years; backed up by numerous other global trends over the last two decades (Piontkovski & Castellani, 2009; Wells et al., 2021). The environmental mechanisms for zooplankton declines have been well identified by others within academia (Wells et al., 2021) however, recent work suggests climate variables may not be the full story...
left: adult brine shrimp by Dan Olsen
Shipping & zooplankton:
Figure 1: 2018 Global Shipping Routes ( link )
Ninety percent of our global resources are transported by ship (OECD, 2021) with around 55,000 actively trading merchant ships worldwide (Roberts, 2021). Like cars, ships follow network routes to get from port to port. Along these routes, the high ship densities threaten marine wildlife. Among the various ship-caused threats, noise pollution (Erbe et al., 2019) and collisions (Shoeman et al., 2020) have consistently been proven to impact large mammal survival. Recent work has now shown that zooplankton are similarly affected, with specific noise effects eliminating 64% of individuals within a kilometer radius (McCauley et al., 2017) and localized mechanical effects eliminating up to 80% of individuals (Sawant et al., 2008).
No work has been done in an attempt to protect zooplankton from shipping vessels using spatial methods. Most of the existing work to mitigate shipping impacts focus on marine mammals and identifying high-risk areas for those species (see this and this ). However, even among the mammal research, there is limited work on designing shipping routes that avoid these species.
How will this project help?
This project's main objective was to provide preliminary work towards the design of optimal shipping routes that avoid important marine wildlife. Therefore, the following goals needed to be achieved in order to provide valuable knowledge on this topic:
First Goal
Make use of long-term data to identify areas of abundant zooplankton.
Second Goal
Determine the spatial relationships between zooplankton abundance and shipping density.
Third Goal
Determine optimal shipping routes between major ports that avoid the identified zooplankton hotspots.
Study Area:
This project will be looking at the ocean off the East Coast of the United States and Canada, from about Virginia to Nova Scotia (right). This area was primarily chosen because of zooplankton data availability. A 5 km2 grid was created for this area to function as the resolution of the analyses.
The Northeast US Continental Shelf (NEUS) runs through the study area and extends up to 200 km offshore (NOAA Fisheries, 2021). Continental shelves have high zooplankton abundance (Strömberg et al., 2009), and so the study area was designed as a buffer of 250km offshore.
Part 1: Finding Areas of Abundant Zooplankton
The zooplankton abundance data came from the Ecosystem Monitoring Program (EcoMon) surveys run by the NOAA from 1977-2017 (US DOC/NOAA/NMFS, 2017). The entire dataset was used to find long-term consistent regions of high zooplankton abundance using chlorophyll-a concentration (right) as a predictor variable.
Chlorophyll-a data was sourced from NASA (NASA Ocean Biology, 2014). Chlorophyll is used as a proxy for phytoplankton abundance, which is the main food source of zooplankton. Previous work has proven that accurate predictions of zooplankton abundance using chlorophyll-a data are possible using regression methods (González et al., 2011).
To predict zooplankton abundance over the gaps in EcoMon's data, the EBK regression method of interpolation was used. As previously mentioned, chlorophyll concentration was used as the explanatory variable in the interpolation.
The map on the right shows the final EBK predictions of zooplankton abundance. Two main characteristics of these abundance predictions are important to remember:
- Much like the chlorophyll layer, abundance predictions are larger in the Northern region of the study area.
- Unlike the chlorophyll layer, abundance predictions are low near the shorelines.
Part 2: Spatial Relationships
The purpose of exploring relationships between zooplankton abundance and shipping density is to determine whether there is a statistically significant reason for avoiding zooplankton within the study area.
Shipping density data from 2009-2013 was sourced from Halpern et al. (2015). This data is a simple linear average of the layers created by Walbridge's (2013) density algorithm (see his thesis for the hardcore details). The map on the right shows the average shipping density values, summarized within each 5km2 grid of the study area. Take note of the common shipping areas on the map.
Once the shipping data was finally contained, the zooplankton abundance predictions were summarized within the same 5 km2 grids and reclassed into 10 categories using geometric intervals. This made it so that my computer could handle the relationship analyses.
The map on the right shows the results from the Local Bivariate Relationship tool using the summarized zooplankton abundance and shipping density layers. This tool determined the shape of the abundance-density relationship in each 5 km2 grid cell by running a regression with 50 of the grid cell's surrounding neighbours (ESRI, 2021).
The results for the different relationship types were surprising. Negative and convex relationships were expected to be much higher (see this for each relationship shape) however, reasons for the lack of such results may include:
- interpolation error confounded results
- the resolution of the analysis masked localized effects
- there may be another variable influencing abundance that does not relate to shipping density
Because the results had no clear meaning, any significant negative linear or concave results were deemed to be adequate reasoning for the design of new shipping routes.
To reiterate: the goal of exploring these relationships was to assess the need to avoid zooplankton within the study area. It is important to note that the results of this step did not affect the shipping route design in the next step.
Part 3: Optimal Shipping Routes
Optimal routes that avoided zooplankton were done using the Least Cost Path tool. This tool ultimately finds the shortest path between two points with the smallest accumulated cost (ESRI, 2021).
Creating least-cost paths requires us to assign the cost of traveling across the individual cells of the study area; we'll call the result of this cost assignment the 'cost surface'. Layers incorporated in the cost surface need to be reclassed to a common scale so that cells with larger scaled values are 'more costly' to travel through. The cost surface that was created incorporated three reclassed layers:
- The zooplankton abundance layer, previously reclassed into 10 categories using geometric intervals. Cells with a scaled value = 10 are the highest abundance cells.
- The shipping density layer, further reclassed into 10 categories using geometric intervals. Cells with a scaled value = 10 are the lowest density cells.
- A newly created buffer layer (left), with a 25 km offshore buffer region reclassed to a value of 10 and a 250 km offshore buffer region reclassed to a value of 1
Reclassing the layers in this way incentivized traveling through cells that are:
- regions with low zooplankton abundance
- commonly shipped in the past
- offshore
Each reclassed layer was then converted to a 5 km2 resolution raster. These rasters were used to create the final cost surface.
On the left, you will see the general-purpose cost surface that was used for every least-cost path to follow. The higher the 'cost', the less incentive there is to travel through that cell.
This cost surface was created by combining the zooplankton abundance, shipping density, and buffer rasters. 70% of this combination represents the zooplankton raster, and 30% represents the shipping and buffer rasters equally.
Two primary characteristics of this map will be important to remember:
- It is less optimal to travel through the Northern region and the Northern shoreline has higher cost values than further offshore.
- It is more optimal to travel through the Southern region and the Southern shoreline has lower cost values than further offshore.
The cost surface that you saw on the previous map had to be used to create a cost distance surface for each pair of shipping ports. This step essentially just accounts for distance in the 'cost' of traveling between ports. Once a cost distance surface is made, it is simply inputted into the Least Cost Path tool which will output a polyline route between the specified start and stop points.
The map on the left shows the final least-cost polylines created. These will be discussed later however, take note of the paths' distance from shore.
Some other important characteristics of the least-cost paths are:
- paths were only created between large shipping ports (see this for the classification system)
- some ports were slightly outside of the study area, and so a nearby location was instead used as the start/stop point
- some ports were very close to one another, and so only one path was created for the neighboring ports
What can we say about these findings?
The results from this project show a glimpse of the potential for designing routes that are both feasible for the shipping industry and more beneficial for the marine life of the traveled waters. Let's dive a bit deeper into the optimal routes created:
(if you just want to see the common characteristics between all routes created, click here )
Part 1 of 5: Port of Halifax
Least-Cost Paths from the Halifax Port
Some notable characteristics of the routes from the port of Halifax include:
- all routes are at least 50 km offshore from Nova Scotia
- routes typically follow a 'J' shape to their respective endpoints
- route to Boston travels through a relatively high proportion of abundant zooplankton regions
- route to New York/Brooklyn follows very closely to the shoreline
- some resolution issues where the route travels over a peninsula
Part 2 of 5: Port of Boston
Least-Cost Paths from the Boston Port
Some notable characteristics of the routes from the port of Boston include:
- southbound routes initially travel extremely close to the shoreline
- route to Halifax travels through a relatively high proportion of abundant zooplankton regions
- routes to Halifax and Norfolk/Baltimore remain far offshore
- some resolution issues where the route travels over a peninsula
Part 3 of 5: Ports of New York & Brooklyn
Least-Cost Paths from the New York/Brooklyn Ports
Some notable characteristics of the routes from the ports of New York and Brooklyn include:
- all follow extremely close to the shoreline
- overall avoidance of the offshore water regions
- route to Chester pass through a relatively low proportion of high abundant zooplankton regions
- some resolution issues where the route travels over a peninsula
Part 4 of 5: Ports of Chester & Philadelphia
Least-Cost Paths from the Chester/Philadelphia Ports
Some notable characteristics of the routes from the ports of Chester and Philadelphia include:
- route to Boston does not follow as closely to the shoreline as other previous maps
- route to Halifax travels in a 'J' shape
- route to Boston travels through a relatively high proportion of abundant zooplankton regions
- some resolution issues where the route travels over a peninsula
Part 5 of 5: Ports of Norfolk & Baltimore
Least-Cost Paths from the Norfolk/Baltimore Ports
Some notable characteristics of the routes from the ports of Norfolk and Baltimore include:
- route to Halifax is very linear
- half of the routes staying close to shoreline and half of routes traveling more offshore
- route to Boston travelling through a relatively large proportion of abundant zooplankton regions
- some resolution issues where the route travels over a peninsula
Take Home Messages:
1.
Routes to/from more northern ports traveled through proportionally more highly abundant zooplankton regions than the mid-southern ports.
2.
As the distance between the ports decreased, routes traveling on the shoreline became more common.
3.
There were consistent resolution errors in the creation of the routes that caused travel across islands or peninsulas.
Limitations:
This project's main objective was to provide preliminary work towards the design of optimal shipping routes that avoid important marine wildlife. As such, there were some major limitations that should be revisited in future work on the topic.
Accessibility of shipping data:
The lack of shipping route information was very apparent when looking for data sources and most information was exclusive to the shipping companies who collected the data. Had this been different, more parameters of the Cost Distance tool could have been specified and the optimal shipping routes could have been compared to preexisting shipping routes. The extent to which this type of comparison can be done is shown below using the shipping density layer.
Comparison of Optimal Shipping Routes and Shipping Density
Also relating to the shipping data was my own lack of common knowledge about the shipping industry. I attempted to consider the feasibility of the routes while developing the cost surface however, I was unaware of the major motivations for the choice of major shipping routes.
Bias in the zooplankton interpolation:
Because the zooplankton data were point features, there were large gaps along the study area that limited the accuracy of the interpolation. In reality, the ships pulled plankton nets up to 650 ft below sea level to capture the samples rather than sampling at a single location for each survey.
This would have affected the bivariate relationships between zooplankton abundance and shipping density, and there may have been more than the 6% of features with significant negative linear relationships had the zooplankton data been of higher quality.
Overall project scope:
There were three major scope limitations for this project:
- the shipping routes did not take into account other marine species that are similarly affected by ships
- the computer used for this project limited the resolution of the analyses which affected the applicability and accuracy of the results
- there was a bias in the shipping data towards cargo shipping and the results say little about other shipping vessel types
Conclusion:
Zooplankton makes up a large proportion of the ocean's marine life and many zooplankton species are important in various/all stages of their lives. This project gives the preliminary work for designing optimal shipping routes to avoid zooplankton and identifies key limitations in the methods used. The lessons learned from this project can help guide further work on this topic for both zooplankton and other marine wildlife. The potential for multi-species approaches using similar methods is immense and the data we have available now, even just from NOAA's EcoMon (see below), could allow for more sustainable and data-driven decisions for our oceans.
Northeast Surveys: The Fisheries We Count On
If I've kept you around this long and you're still interested in this topic...
- Visit this great resource for more information on zooplankton in the NorthEast US Continental Shelf (NEUS) and this resource that describes EcoMon's zooplankton monitoring in more detail!
- Check out this article on how AIS is improving its data accessibility to provide better data-driven shipping routes!
- Here's another data-driven approach that the shipping industry is using to protect marine wildlife!
“Intelligence takes chance with limited data in an arena where mistakes are not only possible but also necessary." (Liet-Kynes in Frank Herbert's Dune)