This tab displays new modeled fishing effort data created using our latest random forest spatial allocation methodology. This represents our most current approach to mapping global fishing effort using AIS data (industrial) and Sentinel-2/Skylight vessel detections (artisanal). See the About tab for more information.
This tab displays new modeled fishing effort time series created using our latest random forest spatial allocation methodology. This represents our most current approach to mapping temporal trends in global fishing effort using combined industrial and artisanal data. See the About tab for more information.
This tab displays data from Rousseau et al. 2024 representing a previous spatial allocation model and dataset. This provides an independent comparison to the new modeled data shown in the other tabs, using different methodological approaches for mapping global fishing effort. See the About tab for more information.
This app provides an interactive platform for exploring and downloading mapped global fishing effort data. Users can filter by year, country, gear type, vessel length, sector, Exclusive Economic Zone (EEZ), and FAO statistical area using the selection sidebar in each tab.
This latest version of our mapping methodology integrates country-level fishing effort estimates with a statistical spatial allocation model using random forest modeling. The industrial model is built using AIS-derived fishing activity from Global Fishing Watch , while the artisanal model is built using vessel detections from Sentinel-2 from Global Fishing Watch and vessel detections provided by Skylight (via Minderoo Foundation). We combine environmental, economic, and governance variables with the AIS and vessel detections to predict fishing effort globally.
For each fishing country, we trained a two-stage hurdle random forest model to predict the spatial distribution of fishing effort:
By multiplying the predictions from both stages, we obtain the estimated fishing intensity (the proportion of a country's total fishing effort ) in each cell where fishing is predicted to occur. These estimates are then scaled to kW days of fishing effort using total fishing effort values (or number of vessels for artisanal) from Rousseau et al. 2019 (Figure 1).
Figure 1: Schematic overview of the two- stage random forest hurdle modelling and spatial allocation workflow. (A) Observed inputs used to train models, including AIS apparent fishing effort for industrial fleets (2015–2024) satellite-based vessel detections for artisanal fleets (2009–2024) and environmental and governance predictors (e.g., sea-surface temperature, chlorophyll, depth, distance to shore, EEZ/FAO regions). (B) Two-stage hurdle models: Stage 1 predicts fishing presence/absence, and Stage 2 predicts relative effort intensity (proportions), with exclusion layers applied as appropriate (e.g., sea ice; industrial fishing-access constraints; artisanal populated/coastal zones). (C) Model predictions are then used to spatially allocate reported global fishing effort totals to produce annual gridded fishing-effort surfaces (1° × 1°; 1950–2017). Icons adapted from Canva.com.
Mapped effort estimates are provided as nominal fishing effort (kilowatt days) or effective fishing effort (kilowatt days), with a spatial resolution of 1° cell (industrial) and 0.5° cell (artisanal), spanning the years 1950-2017. To estimate effective effort, Rousseau et al. 2019 have assumed a year-on-year increase in technical efficiency of 3.5%, as in Rousseau et al. 2019.
This app and the underlying data was created, and is under continuous development by Gage Clawson, Camilla Novaglio & Julia Blanchard from the Institute for Marine & Antarctic Studies (IMAS), University of Tasmania.
This data is likely not comprehensive and represents modeled outputs only. For example, estimates in Southeast Asia, aside from China, are likely too concentrated for some years (e.g., Indonesia). This is an artifact of insufficient AIS data in this region.
Additionally, users should be aware that historical predictions (1950-2014) may not capture:
The 'Rousseau et al. 2024 data' tab provides access to an independent dataset of mapped global fishing activity from Rousseau et al. 2024 : "A database of mapped global fishing activity 1950–2017". This dataset represents a previous spatial allocation of fishing effort data that is different from the modeled estimates shown in the other tabs.
The Rousseau et al. 2024 dataset offers additional grouping options including functional groups, allowing for detailed analysis of fishing patterns by different fleet characteristics. Users can explore this data by individual flag country, with the same temporal coverage (1950-2017) and location filtering options (EEZ and FAO areas) as the other tabs. Unfortunately, the mapped data (with latitude and longitude points) for the Rousseau data is too large to add to this shiny app, however, this data can be accessed via the IMAS data portal here: https://data.imas.utas.edu.au/attachments/1241a51d-c8c2-4432-aa68-3d2bae142794/
This app has three tabs that allow you to visualise and download fishing effort data:
You can download the data used to create the plots shown in this interactive tool using the 'Download' button included under each tab. Additionally, all model data is available via zenodo and our GitHub repository. As a condition of this tool to access data, you must cite its use: Clawson, S.G., Novaglio, C., & Blanchard J.L. (2025). Global Fishing Effort Model Data and Shiny App: 10.5281/zenodo.15110744.
If you have any ideas on how to improve this app or if you found any issues, you can "create an issue" in our GitHub repository.
For general enquiry we can contact Julia Blanchard at julia.blanchard@utas.edu.au
The development of this app was funded by the Food and Agriculture Organization of the United Nation (FAO) and the Minderoo Foundation. We would also like to acknowledge the use of computing facilities provided by Digital Research Services, IT Services at the University of Tasmania.