The FLR project Objectives To develop a platform for quantitative - PowerPoint PPT Presentation

The FLR project • Objectives ◦ To develop a platform for quantitative work in fisheries Fisheries modelling in R: the FLR (Fisheries Library in R) biology, assessment and management based on R. project ◦ To encourage open and transparent collaboration in fisheries research. P. Grosjean, R. Hillary, E. Jardim, L. T. Kell, I. Mosqueira, ◦ To introduce new tools and procedures already in use J. J. Poos, R. Scott in other fields. ◦ To improve upon the quality of the scientific work carried out for fisheries management. • Research and management applications • The project UseR2006 – p. 1 UseR2006 – p. 2 The FLR project The FLR project • Objectives • Objectives • Research and management applications • Research and management applications ◦ Support for data collection and analysis of sampling • The project design issues ◦ http://flr-project.org ◦ Exploratory data analysis, data aggregation and error ◦ A small team in charge of FLCore, general design and checking package release ◦ Stock assessment and estimation of stock status ◦ EU-funded research projects indicators ◦ Simulation testing of management scenarios • The project UseR2006 – p. 3 UseR2006 – p. 4

The FLQuant class The FLQuant class • Basic “building block” of FLCore, holds most fisheries data North Sea Plaice, area 4 (biological, technological, economic) 13 14 15 1 2 3 4 1 2 3 4 5 • A five dimensional array (soon to be 6D) 1.5 0.5 • Dimensions: quant, year, unit, season, area, (iter) 1960 1980 2000 1960 1980 2000 1960 1980 2000 9 10 11 12 0 2 4 6 8 10 • Attribute: units 2 4 6 8 5 10 15 10 5 catch−at−age 0 0 1960 1980 2000 1960 1980 2000 1960 1980 2000 1960 1980 2000 5 6 7 8 60 100 0 20 40 60 80 0 510 20 30 50 100150 20 1960 1980 2000 1960 1980 2000 1960 1980 2000 1960 1980 2000 1 2 3 4 0 2 4 6 8 50 150 250 020 60 100 150 50 1960 1980 2000 1960 1980 2000 1960 1980 2000 1960 1980 2000 UseR2006 – p. 5 UseR2006 – p. 6 FLCore: classes FLCore: classes • Fully designed around S4 classes • Example: FLStock ◦ Inheritance provides good extensibility (FLAssess) ◦ Method overloading reduces command set for interactive use and simplifies modular development (assess) • C++ classes ◦ FLCore classes have been replicated in C++ to use with R headers ◦ To help integrating legacy code and speed up slow calculations • Accesor and replacement functions automatically generated at package compile time UseR2006 – p. 7 UseR2006 – p. 8

FLCore: methods Other packages • Extensive use of lattice to deal with plots of • FLEDA: exploratory data analysis, lattice plots, multi-dimensional data • FLBayes: Bayesian fisheries models, McMC S4 class • Minimum set of methods required for new classes • FLAssess + FLXSA: Stock assessment using VPA methods ◦ show, summary, plot • FLOE: Observation error ◦ window • FLOM: Fisheries Operating Model conditioned on • A number of new generic methods covering common age-structured assessment results operations • FLEcon • Overloading of many S3 methods in R base UseR2006 – p. 9 UseR2006 – p. 10 Stock assessment with FLAssess Management Strategy Evaluation • Stock assessment is a fundamental task in fisheries • Computer simulation of stock, fishery, advice and science. management systems • Separate implementations of sometimes similar methods • Exploration of uncertainties and their impact on require and return input and output files in different formats management • Data available as FLR objects can be input to a range of • Comparison of complex models and simpler management assessment methods rules under varipous scenarios • Output diagnostics and standard plots are available with • Objective is the design of management procedures robust the same syntax for different assessment models to present and future uncertainties • ICES advice system requires yearly evaluation of stock • Pioneered by the development of the Revised status and trends Management Procedure of the IWC ◦ Operational Model of the fishery system (stock & fleet) • Exploration of data and results is limited by time constraints ◦ Data collection and stock assessment and the difficulty of moving data between incompatible software ◦ Harvest Control Rule for management decision making ◦ Interaction through Bayesian Belief Networks UseR2006 – p. 11 UseR2006 – p. 12

Management Strategy Evaluation Future developments • Fisheries operating models: Northern Hake (Garcia, D., • Increase its adoption on various fora (ICES, Tuna Mosqueira, I.) Commissions) • Packages in development ◦ Cluster and grid computation ◦ Storage of FLR objects in SQL databases • File format based on XML and StatDataML • Implementation of unit testing • Variability in SSB, F and TAC due to uncertainty in recruitment and indices of abundance UseR2006 – p. 13 UseR2006 – p. 14