Testing different methods of incorporating climate data into the assessment of US West Coast sablefish

Abstract

Schirripa, M. J., Goodyear, C. P., and Methot, R. M. 2009. Testing different methods of incorporating climate data into the assessment of US West Coast sablefish. – ICES Journal of Marine Science, 66: 1605–1613. The objective of this investigation was to evaluate different methods of including environmental variability directly into stock assessments and to demonstrate how this inclusion affects the estimation of recruitment parameters, stock status, and the conservation benchmarks used to manage a stock. Variations on two methods of incorporating environmental effects were tested. The first method (“model” method) utilizes a structural change in the stock–recruitment function to adjust the annual expected number of recruits by a value, either positive or negative, equal to that year's anomaly in the environmental variable. The second method (“data” method) allows for observation error in the environmental data and uses the time-series as an index to tune the vector of estimates of annual recruitment deviations. Simulation techniques were utilized to produce datasets of known quantities that were subsequently analysed with a widely used stock assessment platform. Under the circumstances simulated in this study, neither method could be said to have performed significantly better than the other in all situations. Because the two approaches handle years of missing data differently, the best approach is dictated by the available data, rather than a more appropriate method.

Shifts in fisheries management: adapting to regime shifts

“…Aside from these two operationalized examples, a number of studies have attempted to include some form of environmental forcing in stock assessments [25][26][27]. More typical is the identification of a stock-recruitment relationship with environmental forcing [7,[28][29][30].…”

Quantifying the population response of invasive water hyacinth, Eichhornia crassipes, to biological control and winter weather in Louisiana, USA

“…Thus, we assumed the trend in overwinter mortality was known and equal to the trend in WSI. Although it is theoretically possible (and preferable) to structure our model such that the trend in overwinter mortality was estimated by fitting our model directly to the WSI (Schirripa et al 2009), our data sources were not extensive or informative enough to statistically estimate annual deviations in both population growth and overwinter mortality separately. If such an approach had been feasible, we would likely have been able to reduce the level of under and overprediction of fall and spring water hyacinth abundance that resulted, especially during the initial years of the study period.…”

“…However, the dynamics of other invasive species may be driven by different anthropogenic or environmental factors that induce change over time in population rates. Assuming adequate data have been collected, models with time-varying parameters could be used to estimate population-level changes induced by such factors as disease outbreaks, introduction of competitors, and habitat or climate change (Schirripa et al 2009). In addition, changes in the methodology of long-term invasion monitoring programs that may have affected detection or catch rates can often be modeled using time-varying parameters, thus maintaining the utility of data collection programs that have undergone methodological changes (Linton and Bence 2011;Wilberg et al 2009).…”

“…Models with time-varying parameters offer a more flexible alternative. Time-varying parameterizations of logistic and age-structured population models are commonly used in fisheries science and related disciplines to capture strong temporal trends in population growth, mortality, capture/detection, and productivity rates across long time series (Chang et al 2015;Fu and Quinn II 2000;Jiao et al 2012;Linton and Bence 2011;Nesslage and Wilberg 2012;Schirripa et al 2009). Several approaches for incorporating time-varying parameters in population models exist, including (1) specification of a trend, (2) allowing the trend in an estimated model parameter to track an ancillary time series, (3) estimating the trend in a parameter using ancillary data as an index of the time-varying process of interest, or (4) modeling the time-varying parameter using a specific function (e.g.…”

“…Several approaches for incorporating time-varying parameters in population models exist, including (1) specification of a trend, (2) allowing the trend in an estimated model parameter to track an ancillary time series, (3) estimating the trend in a parameter using ancillary data as an index of the time-varying process of interest, or (4) modeling the time-varying parameter using a specific function (e.g. linear, random walk; Schirripa et al 2009;Wilberg and Bence 2006). The most appropriate parameterization for a given circumstance depends on the degree to which available data are informative with regard to the time-varying process of interest.…”