Cloudy with a chance of sardines: forecasting sardine distributions using regional climate models

Abstract

Despite the significant advances in making monthly or seasonal forecasts of weather, ocean hypoxia, harmful algal blooms and marine pathogens, few such forecasting efforts have extended to the ecology of upper trophic level marine species. Here, we test our ability to use short‐term (up to 9 months) predictions of ocean conditions to create a novel forecast of the spatial distribution of Pacific sardine, Sardinops sagax. Predictions of ocean conditions are derived using the output from the Climate Forecast System (CFS) model downscaled through the Regional Ocean Modeling System (ROMS). Using generalized additive models (GAMs), we estimated significant relationships between sardine presence in a test year (2009) and salinity and temperature. The model, fitted to 2009 data, had a moderate skill [area under the curve (AUC) = 0.67] in predicting 2009 sardine distributions, 5–8 months in advance. Preliminary tests indicate that the model also had the skill to predict sardine presence in August 2013 (AUC = 0.85) and August 2014 (AUC = 0.96), 4–5 months in advance. The approach could be used to provide fishery managers with an early warning of distributional shifts of this species, which migrates from the U.S.–Mexico border to as far north as British Columbia, Canada, in summers with warm water and other favorable ocean conditions. We expect seasonal and monthly forecasts of ocean conditions to be broadly useful for predicting spatial distributions of other pelagic and midwater species.

On the skill of seasonal sea surface temperature forecasts in the California Current System and its connection to ENSO variability

“…Furthermore, applications of seasonal SST forecasts often occur on scales smaller than the entire CCS (e.g., Kaplan et al 2016;Tommasi et al 2017). We therefore divide the CCS into northern, central, and southern sub-regions, with divisions at Cape Mendocino (~40.5°N) and Point Conception (~34.5°N), and evaluate SST forecast skill on these finer scales.…”

“…In Australian waters this approach has been extended to operational seasonal forecasts using dynamical climate forecast systems, with notable examples for coral reef stress on the Great Barrier Reef (Spillman et al 2013), bycatch reduction for Southern Bluefin Tuna off east Australia (Hobday et al 2011), and improved efficiency of the Southern Bluefin Tuna fishery in the Great Australian Bight (Eveson et al 2015). While similar efforts have not yet been operationalized in the CCS, Kaplan et al (2016) demonstrate one potential application using a sea surface temperature (SST) based habitat model in conjunction with downscaled seasonal forecasts to predict Pacific sardine distributions in the northern CCS. A second recent CCS climate forecast application aims to refine catch limits for climate-sensitive Pacific sardine by introducing short-term temperature forecasts (Tommasi et al 2017).…”

A Framework for Combining Seasonal Forecasts and Climate Projections to Aid Risk Management for Fisheries and Aquaculture

“…Fisheries managers can use forecast information to plan distribution of fishing effort, as has been proposed for the Californian sardine fishery based on forecasts with a downscaled regional ocean model (Kaplan et al, 2016). Aquaculture managers might be charged with managing production and harvest schedules, which can be informed by knowledge of likely (and unlikely) conditions over the coming months (i.e., seasonal forecasts).…”

“…Seasonal forecasting applications in Australia and elsewhere have been developed for a range of marine resource segments, including salmon and prawn aquaculture (Spillman and Hobday, 2014;Spillman et al, 2015), commercial tuna Eveson et al, 2015) and sardine fisheries (Kaplan et al, 2016), and recreational fisheries (Brodie et al, 2017). Depending on the application, these forecasting applications have delivered information on both environmental conditions, such as water temperature, rainfall, and air temperature, and habitat distribution, at lead times of up to 3 months , helping managers and fishers to plan activities based on predicted conditions (Eveson et al, 2015;Spillman et al, 2015).…”

Cooperative Research to Evaluate an Incidental Catch Distribution Forecast

“…Species distribution models provide the foundation for most proactive dynamic management systems, as species distributions are directly or indirectly related to environmental conditions (Mann, 1993). Environmental forecasts can be combined with species distribution models to predict over relatively short (i.e., days to months) time scales to inform fisheries (Hartog et al, 2011;Eveson et al, 2015;Kaplan et al, 2016), or over longer (i.e., decadal) time scales to predict shifts in distributions related to climate change (Hare et al, 2010;Lynch et al, 2015).…”