Tutorial: AIS in Anchovy Bay
A new management regulation for Anchovy Bay now requires that all trawler fishing boat transmit their position via Automatic Identification System (AIS) while at sea. The AIS signals are automatically picked up by the fisheries management authority, and maps with relative number of hours that trawlers are in the bay are now recorded and used for management purposes. We have permission to access to some of these files, and this tutorial will explain how to use such data.
First, this calls for some changes to our spatial ecosystem model for Anchovy Bay (Anchovy Bay spatial.ewemdb, you can download it from this link) as the effort data is available for two trawler fleet, one operating in the western part of the bay and a second in the eastern part of the bay. Open your model, and do the following,
- Maybe do a Save as, and save it as a new model, e.g., Anchovy Bay AIS
- Add a new fleet as fleet #6: Trawlers East. (Ecopath > Input > Fishery > Fleets > Define fleets)
- Rename your existing fleet #2 to Trawlers West while you are at the Define fleets form.
- On Ecopath > Input > Fishery > Fleets set Effort related cost to 0 for fleet #2 and #6, and Sailing related cost to 80 for fleet #2 and #6. Now that we have AIS data to distribute the effort for the trawlers, we want to follow it as much as possible, and this will help with that.
- On Ecopath > Input > Fishery > Landings, distribute the landings for fleet #2 to be shared equally between fleet #2 and fleet #6, so half of the #2 landings go to the new fleet.
- Open your BayOfAnchovies Ecospace scenario, and go to Ecospace > Input > Ecospace fishery > Habitat fishery, and make fleet #2 and #6 fish in Fish everywhere (all habitats).
Now we need to process the AIS files. They are saved in ASCII format (.asc), and you can download them from this link. ASCII is a spatial raster format, and if you rename the files to .csv they can be read by Excel. It is, however, much easier to do the operations we need in R. The files represent relative number of hours boat are in spatial cell by year for each fleet. We want to express the data as relative fishing cost, so we need to invert the values.
R code for reading AIS ASCII files with relative hours in cells and converting them to relative fishing cost
library(terra) # for spatial operations
#============= Set up directory links and folders =============================
work.dir = dirname(rstudioapi::getActiveDocumentContext()$path)
setwd(work.dir)
# Read .asc files
west = rast(“ab fleet west.asc”)
east = rast(“ab fleet East.asc”)
# these AIS files have values of relative # of hours that boats were in grid cells
# invert to get relative fishing cost
west.c = 1/west
east.c = 1/east
writeRaster(west.c,”ab fleet west cost.asc”,overwrite=T)
writeRaster(east.c,”ab fleet east cost.asc”,overwrite=T)
plot(west.c)
plot(east.c)
- Go to Ecospace > Input > Maps click Sailing cost in the right-hand panel. In the lower right panel, change Fleet to 2. Trawlers West
- Double click where it says Sailing cost at the top right, and a spreadsheet will pop up. Select Import > From ASCII grid and browse to load ab fleet west cost.asc
- Repeat this for fleet #6.
You now have the setup for using the AIS, so all that’s needed is to run your model, check what’s happening to effort for especially the two trawler fleets. Making sense?
But does the predicted effort patterns indeed make sense? Do the two trawler fleets actually fish where the target species (cod and whiting) are? If not, why? You are using actual AIS data, so it does represent where the trawlers have been. Even if it is where they’ve spent time sailing but not necessarily fishing. For instance, the trawlers were allowed to fish where there’s hard bottom/reefs, but they may not actually be fishing there. But, the most likely reason for a discrepancy is?
It’s probably that the model predictions for where the cod and whiting are distributed is flawed, we should check that. When we find discrepancies, that’s when we learn something. There’s nothing better for modelling than finding problems with a model, that’s when you learn something that can be used to improve the model. The more information we get to add to the model, the more constraints are added, and that makes the models better.
Models are like gourmand master chefs, the better ingredients you give them, the better courses they’ll make.