GitHub - GlobalFishingWatch/paper-longline-ais-sar-matching: Public repository for the paper Revealing the Global Longline Fleet with SAR.
- ️Mon Apr 04 2022
paper-longline-ais-sar-matching
A Repo for the paper Revealing the Global Longline Fleet with satellite radar.
Kroodsma, D.A., Hochberg, T., Davis, P.B. et al. Revealing the global longline fleet with satellite radar. Sci Rep 12, 21004 (2022). https://doi.org/10.1038/s41598-022-23688-7
Many of the scripts here access the public BigQuery dataset, global-fishing-watch.paper_longline_ais_sar_matching
This repository has the following folders with the following contents (the numbers are just to sort them):
1_Acquisitions
The data obtained from KSAT on SAR detections and scenes, and the code to process the files they provided into csv and BigQuery tables.
2_Evaluation_of_Fishing_Activity
A number of analyses of AIS data in the region
- Number and area of satellite footprints in each region
- map of longline activity with areas of interst
- analysis of fishing activity in these regions from AIS
3_Vessel_Review
We reviewed vessels to be sure of the vessel classification, and to make sure we were correctly classifying which MMSI were gear and which were not. This includes a good deal of manual reviews, and plotting of vessel tracks. In addition, we performed analysis to determine which flag states the vessels were from, what EEZs they were active in, and if they were matched to registries.
4_Probability_Raster_Generation
- Code that produces probability rasters
5_SAR_AIS_Matching
This includes steps that match SAR detections to AIS:
- extrapolationg vessels to the time of the image
- estimating the likelihood a given vessel is within a scene, based on the probability rasters
- scoring each detection to vessel pair using a) a weighted average of the probability rasters before and after and b) an average of the before and after raster
- a ranking method to decide on the most likely matches
6_Evaluating_Matching
- Comparison of multiplying versus averaging scores before and after the scenes
- Visual comparisons of ambiguous matches and low-confidence matches
- Figures for the paper that demonstrate the rasters
- An estimation of how long between the image and the detection.
7_Dark_Vessel_Estimates
Estimate the relationship between vessel size from GFW and SAR, and recall as a function of length. These are combined to produce a probabilistic estimate of the number of non-broadcasting fishing vessels in the region.
8_Area_of_Ocean_to_be_Monitored
Using this technique, how much of the ocean could we monitor? This shows the amount of imagery needed to track 5%, 10%, 20% and 50% of the global pelagic longline fleet. We also estimate the same for for the area that is not covered by Sentinel-1 imagery.
Turning your repo into a module
Run pip install -e . to install the module. This will create a folder titled .egg-info that will allow you to access the code within your folder from outside of that folder by doing import without any need to use paths.