Limits...
Mapping Fishing Effort through AIS Data.

Natale F, Gibin M, Alessandrini A, Vespe M, Paulrud A - PLoS ONE (2015)

Bottom Line: After analysing a large AIS dataset for the period January-August 2014 and covering most of the EU waters, we show that AIS was adopted by around 75% of EU fishing vessels above 15 meters of length.Using the Swedish fleet as a case study, we developed a method to identify fishing activity based on the analysis of individual vessels' speed profiles and produce a high resolution map of fishing effort based on AIS data.The method was validated using detailed logbook data and proved to be sufficiently accurate and computationally efficient to identify fishing grounds and effort in the case of trawlers, which represent the largest portion of the EU fishing fleet above 15 meters of length.

View Article: PubMed Central - PubMed

Affiliation: European Commission Joint Research Centre Institute for the Protection and Security of the Citizen, Ispra, Italy.

ABSTRACT
Several research initiatives have been undertaken to map fishing effort at high spatial resolution using the Vessel Monitoring System (VMS). An alternative to the VMS is represented by the Automatic Identification System (AIS), which in the EU became compulsory in May 2014 for all fishing vessels of length above 15 meters. The aim of this paper is to assess the uptake of the AIS in the EU fishing fleet and the feasibility of producing a map of fishing effort with high spatial and temporal resolution at European scale. After analysing a large AIS dataset for the period January-August 2014 and covering most of the EU waters, we show that AIS was adopted by around 75% of EU fishing vessels above 15 meters of length. Using the Swedish fleet as a case study, we developed a method to identify fishing activity based on the analysis of individual vessels' speed profiles and produce a high resolution map of fishing effort based on AIS data. The method was validated using detailed logbook data and proved to be sufficiently accurate and computationally efficient to identify fishing grounds and effort in the case of trawlers, which represent the largest portion of the EU fishing fleet above 15 meters of length. Issues still to be addressed before extending the exercise to the entire EU fleet are the assessment of coverage levels of the AIS data for all EU waters and the identification of fishing activity in the case of vessels other than trawlers.

No MeSH data available.


Example of speed profile for one vessel and fitting of a bi-modal distribution through the EM algorithm.The confidence lines at 1.5 standard deviation around the first mode indicate the speed thresholds that were used to classify the AIS messages as fishing. The curve and speed profiles were analysed for each vessel resulting in specific classification threshold on the basis of individual fishing behaviour.
© Copyright Policy
Related In: Results  -  Collection

License
getmorefigures.php?uid=PMC4476776&req=5

pone.0130746.g001: Example of speed profile for one vessel and fitting of a bi-modal distribution through the EM algorithm.The confidence lines at 1.5 standard deviation around the first mode indicate the speed thresholds that were used to classify the AIS messages as fishing. The curve and speed profiles were analysed for each vessel resulting in specific classification threshold on the basis of individual fishing behaviour.

Mentions: Fig 1 provides a typical example of speed profiles for one vessel to exemplify the methodology. Assuming that the speed profiles are characterized by only two speed modes is plausible especially in light of our data preparation steps: by omitting the messages with a speed of zero, we are filtering out a component in the distribution of speed values. A fourth component related to searching speed was omitted since it was present in very few vessels. Using an Expectation Maximization (EM) algorithm [26, 27] it is possible to provide estimates of the two distributions' parameters and to assign the observations to a particular model component. The EM algorithm has two steps: expectation and maximization. The Expectation step (E-step) aims to estimate the ownership probability, which in other terms is the expected values of the missing data giving the current model estimate. The Maximization (M-step) instead computes the maximum likelihood model parameters given the observed data and the previously calculated expected value of the missing ones. In the case of mixture models the maximization step results in a weighted regression for each model component and mixing components.


Mapping Fishing Effort through AIS Data.

Natale F, Gibin M, Alessandrini A, Vespe M, Paulrud A - PLoS ONE (2015)

Example of speed profile for one vessel and fitting of a bi-modal distribution through the EM algorithm.The confidence lines at 1.5 standard deviation around the first mode indicate the speed thresholds that were used to classify the AIS messages as fishing. The curve and speed profiles were analysed for each vessel resulting in specific classification threshold on the basis of individual fishing behaviour.
© Copyright Policy
Related In: Results  -  Collection

License
Show All Figures
getmorefigures.php?uid=PMC4476776&req=5

pone.0130746.g001: Example of speed profile for one vessel and fitting of a bi-modal distribution through the EM algorithm.The confidence lines at 1.5 standard deviation around the first mode indicate the speed thresholds that were used to classify the AIS messages as fishing. The curve and speed profiles were analysed for each vessel resulting in specific classification threshold on the basis of individual fishing behaviour.
Mentions: Fig 1 provides a typical example of speed profiles for one vessel to exemplify the methodology. Assuming that the speed profiles are characterized by only two speed modes is plausible especially in light of our data preparation steps: by omitting the messages with a speed of zero, we are filtering out a component in the distribution of speed values. A fourth component related to searching speed was omitted since it was present in very few vessels. Using an Expectation Maximization (EM) algorithm [26, 27] it is possible to provide estimates of the two distributions' parameters and to assign the observations to a particular model component. The EM algorithm has two steps: expectation and maximization. The Expectation step (E-step) aims to estimate the ownership probability, which in other terms is the expected values of the missing data giving the current model estimate. The Maximization (M-step) instead computes the maximum likelihood model parameters given the observed data and the previously calculated expected value of the missing ones. In the case of mixture models the maximization step results in a weighted regression for each model component and mixing components.

Bottom Line: After analysing a large AIS dataset for the period January-August 2014 and covering most of the EU waters, we show that AIS was adopted by around 75% of EU fishing vessels above 15 meters of length.Using the Swedish fleet as a case study, we developed a method to identify fishing activity based on the analysis of individual vessels' speed profiles and produce a high resolution map of fishing effort based on AIS data.The method was validated using detailed logbook data and proved to be sufficiently accurate and computationally efficient to identify fishing grounds and effort in the case of trawlers, which represent the largest portion of the EU fishing fleet above 15 meters of length.

View Article: PubMed Central - PubMed

Affiliation: European Commission Joint Research Centre Institute for the Protection and Security of the Citizen, Ispra, Italy.

ABSTRACT
Several research initiatives have been undertaken to map fishing effort at high spatial resolution using the Vessel Monitoring System (VMS). An alternative to the VMS is represented by the Automatic Identification System (AIS), which in the EU became compulsory in May 2014 for all fishing vessels of length above 15 meters. The aim of this paper is to assess the uptake of the AIS in the EU fishing fleet and the feasibility of producing a map of fishing effort with high spatial and temporal resolution at European scale. After analysing a large AIS dataset for the period January-August 2014 and covering most of the EU waters, we show that AIS was adopted by around 75% of EU fishing vessels above 15 meters of length. Using the Swedish fleet as a case study, we developed a method to identify fishing activity based on the analysis of individual vessels' speed profiles and produce a high resolution map of fishing effort based on AIS data. The method was validated using detailed logbook data and proved to be sufficiently accurate and computationally efficient to identify fishing grounds and effort in the case of trawlers, which represent the largest portion of the EU fishing fleet above 15 meters of length. Issues still to be addressed before extending the exercise to the entire EU fleet are the assessment of coverage levels of the AIS data for all EU waters and the identification of fishing activity in the case of vessels other than trawlers.

No MeSH data available.