Application of ARIMA Models to Time Series of Recreational Catch and Effort

Autoregressive Integrated Moving Average (ARIMA) models are standard tools in time series analysis.  They are used extensively for modeling correlations in sequences, producing seasonally adjusted estimates, and forecasting in fields ranging from economics and finance to health and environmental sciences to computer sciences and electrical engineering.  In comparison, ARIMA model use in fisheries appears limited with relatively few reports in the literature.  This study explores the feasibility of modeling series of catch and effort estimates from recreational fishery surveys using ARIMA methodology.  Series of two-month (wave) estimates of catch and effort, from the Marine Recreational Fishery Statistics Survey (MRFSS), are modeled at various levels of spatial aggregation for states along the U.S. Atlantic and Gulf of Mexico coasts.  In addition to the standard series, log and logit transform series are modeled to account for non-normality.  Forecasts with simple retrospective analyses are provided.  Model performance improves in series with less variable correlation patterns: effort models fit better than do those for catch at the same level of aggregate; models tend to fit better in northern states with stronger seasonal fishing patterns; catch models fit better for species with relatively stable restrictive management regulations; shorter series (10-15 years) fit better than longer series (20+ years).  Forecasting results vary but appear useful in a number of cases.