| Vessel Monitoring System(VMS)has become a key tool in fisheries law enforcement management as fishing vessels around the world are required to be equipped with Vessel Monitoring System(VMS)equipment,which can effectively monitor the real-time status of fishing vessels,reduce irregularities,ensure the safety of fishing vessels,and play a positive role in fisheries resource management.As VMS continues to develop and popularise,the use of VMS data for monitoring the position of fishing vessels,determining fishing behaviour,and fisheries planning and management has become increasingly widespread.In the current context of limited fisheries data,VMS,as an important source of high-precision,high-accuracy and massive data,has become a hot issue in current fisheries resource assessment research.However,the inability of VMS to record production data,such as fishing vessel catches,makes it potentially inadequate for research in the field of fisheries resource assessment.In this study,to further explore and expand the application of VMS data in fisheries resource assessment,the following research was conducted with two different types of production fishing vessels in the Zhoushan area.Firstly,the operational characteristics of fishing vessels were analysed based on the VMS data of three main vessels in three pairs of double trawlers(Zhepuyu 71319,71528 and 71568)and one canvas spreader stow net fishing vessel(Zhepuyu 71010)in 2021,and a method to determine the state of fishing vessels based on segmentation characteristics was proposed.Secondly,based on the VMS data of fishing vessels and fishing logbook data,a method for predicting the catch of fishing vessels based on VMS data was proposed.Based on the VMS data of fishing vessel fishing states,the fishing effort and temporal and spatial features of fishing vessels per trip were extracted and calculated,and the spatial and temporal distribution patterns of fishing effort were analysed.Six Machine Learning(ML)models(k-Nearest Neighbor[KNN],Random Forest[RF],Support Vector Machine[SVM],Extreme Gradient Boosting[eXtreme Gradient Boosting,XGBoost],Gradient Boosting Decision Tree[GBDT]and Ridge Regression[RR])were used to predict the catch per trip of fishing vessels and the prediction results were evaluated using the Root Mean Square Error(RMSE),the Mean Absolute Error(MAE),the coefficient of determination(R2)and the prediction error(Et)of the catch per trip.The main conclusions were as follows.(1)Threshold range and track characteristics for different states of the double trawler.The fishing state vessel position point had a speed range of 0.5-4 Kn and a heading difference range of-100° to 200°.The sailing state had a speed range of 6-12 Kn and the vessel tracks in a straight line.Slow speed range was 4-8Kn with a heading difference of±50°.At anchor,the speed range was 0-3Kn,the heading difference range was ±360°,and the latitude and longitude range of the track point was less than 30".By calculating the relative displacement distance of the vessel position points in each state of the double trawler,it is concluded that the relative displacement distance of the vessel position points in the anchored state ranges from 0-0.1km,in the fishing state from 0.1-1.0km,in the slow speed state from 1.5-2km and in the navigation state from 2-3km.(2)Threshold ranges for the different states of canvas spreader stow net fishing vessels.The anchored state had a relative displacement distance of less than 0.1km and a duration of more than 3h.The navigation state had a speed higher than 6Kn and a relative displacement distance of more than 2km.The vessel position point between the first navigation and the first anchoring after arrival at the fishing ground was determined as the net shooting state with a speed range of 0-3Kn and a heading difference range of-200° to 200°.The vessel position point between every two anchorings and between the last anchoring and navigation was determined as the net hauling state,with a speed range of 0-4Kn and a heading difference range of-100°-100°.(3)Catch prediction for double trawlers.Two of the three double trawlers were used in turn as the training set to build the ML models,and the other one as the test set.The results suggested that all six ML models exhibited high prediction performance,with coefficients of determination(R2)ranging from 0.66 to 0.95,MAE ranging from 1121.97 to 2302.89,RMSE ranging from 1523.83 to 3127.40 and Et ranging from-0.88 to 1.95.However,the accuracy of catch prediction and the performance of the models varied with different data sets.The ML models with the best prediction performance were XGBoost,GBDT and KNN when Zhepuyu 71319,71528 and 71568 were used as the test set respectively,mainly due to differences in data distribution,model parameters,size of the data set and applicability of machine learning algorithms.(4)Catch prediction for canvas spreader stow net fishing vessels.The data set of three double trawlers was used as the training set to build the ML model,and the canvas spreader stow net fishing vessels data set was used as the test set.The prediction performance of the models was generally weaker due to the different modes of operation of the two types of fishing vessels,with KNN outperforming the other models with R2 of 0.14,MAE of 8377.91,RMSE of 11899.2 and Et of-0.88-1.18.This study was a new attempt in the field of marine fishery resource assessment in China.The study suggested that monitoring the activities of fishing vessels based on VMS and using VMS data information to assess the conditions and utilisation of marine fishery resources can help to accurately determine the time,location and catches of fishing vessel activities,improve the efficiency of fishery production,and contribute to better management of marine fishery resources.The research results will help enrich the fisheries resources assessment methods applicable under data-limited conditions,help improve the reliability of assessment results,and can provide technical references for fisheries management departments to carry out fisheries resources assessment and management and enterprises to record fishing vessel sailing data.The prediction and monitoring of catch in this study can adjust the fishing quota in a timely manner,which is important for the implementation of the quota fishing system.In addition,the development of VMS can strengthen the monitoring,control and surveillance(MCS)of fishing vessels,reduce illegal,unreported and unregulated(IUU)fishing activities,help promote the monitoring and regulation of fishery resources,and assist management departments in conducting fishery resource assessment,and provide some reference and learning for relevant fishery management agencies to formulate scientific and reasonable fishery management policies. |
PDF Full Text Request