Freshwater Creel Estimates

This repository supports estimation of freshwater fishery catches and angler effort using roving-roving creel designs.

The main template script fw_creel.Rmd provide a workflow to fetch raw observed data, generate intermediate summaries, produce expanded point estimates and associated uncertainty measures, and output tables and figures. It builds on previous scripts associated with a rapid, established method ("PE") and a more recently developed Bayesian hierarchical state space method ("BSS").

These standard procedures are modified through user inputs in the uppermost params block of the script, defining the fishery of interest (fishery_name), the start and end dates of the period to be assessed, the focal species and encounter types of interest (est_catch_groups), and various other controls related to both the sampling design/protocol (e.g., what quantities were counted during index surveys; which days of the week are considered 'weekend' vs 'weekday') and the particular analysis (e.g., a minimum 'fishing time' duration threshold to filter interview).

Functions

This parameterized Rmarkdown workflow calls a sequence of R_functions/

• fetch_dwg brings into memory a set of electronically collected observations from the statewide freshwater creel database published at data.wa.gov); these consist of counts at index sites and along census survey sections as well as angler interviews and associated catch information
• prep_days associates time-strata information to the period of interest (e.g., week/month index, potential fishing hours, section-specific closures)

Next, the raw data are summarized as a list of objects shared by both the PE and BSS estimation methods

• prep_dwg_interview filters and reorganizes interview records, calculating times and joining catch values conditioned on user inputs in the params block
• prep_dwg_effort_census filters, reorganizes and aggregates survey counts, in particular associating the "count_sequence" index of the nearest-in-time index observations
• prep_dwg_effort_index similarly filters, reorganizes and aggregates counts at specific index sites

These objects are further processed to form the inputs for a "classic PE" method

• prep_inputs_pe_census_expan
• prep_inputs_pe_days_total
• prep_inputs_pe_ang_hrs_bank_boat
• prep_inputs_pe_ang_hrs_vhcl_trlr
• prep_inputs_pe_daily_cpue_catch_est
• prep_inputs_pe_df
• est_pe_effort
• est_pe_catch

The initial summary objects also form the basis for the state space method

• prep_inputs_bss translates the prepared tabular data into a list of the vector and matrix formats required by the stan model code and adds values for several priors; this function is iterated as a list of lists, with element of the outer list associated with each desired "catch group"

Each catch-group-specific list of state space inputs is then processed

• fit_bss wraps the rstan::stan() function, passing in the data list and various arguments controlling the MCMC process (e.g., number of chains, iterations, etc.)
• get_bss_overview
• get_bss_catch_daily
• get_bss_effort_daily

Finally, the calculated estimates are presented as tables, plots, and standalone workbooks...

PENDING FUNCTIONALIZED OUTPUTS