Improving landings forecasts using environmental covariates: A case study on the Indian oil sardine (Sardinella longiceps). (31st May 2021)

• Record Type: Journal Article
• Authors: Holmes, Elizabeth Eli; BR, Smitha; Nimit, Kumar; Maity, Sourav; Checkley, David M.; Wells, Mark L.; Trainer, Vera L.
• Is Part Of: Fisheries oceanography. Volume 30:Number 6(2021)
• Page Start: 623
• Page End: 642
• Publication Date: 2021-05-31
• Subjects: Atlantic Multidecadal Oscillation -- catch prediction -- climate -- dynamic linear modeling -- GAM modeling -- Indian oil sardine -- remote sensing -- sea surface temperature -- Southeast Arabian Sea; Fishery oceanography -- Periodicals; 639.2
• DOI: 10.1111/fog.12541
• Languages: English
• ISSNs: 1054-6006

Abstract

Abstract: Commercial landings of sardines are known to show strong year‐to‐year fluctuations. A key driver is thought to be environmental variability, to which small forage fish are especially sensitive. We examined the utility of including environmental covariates in forecasts for landings of the Indian oil sardine using a long‐term time series of quarterly catches. Potentially influential variables examined included precipitation, upwelling intensity, sea surface temperature (SST), and chlorophyll‐a concentration. All of these have been shown to be important for oil sardine growth and survival, spawning and/or movement into the nearshore fishing regions. However, improving out‐of‐sample landings forecasts using environmental covariates has often proven elusive. We tested the inclusion of environmental covariates in forecast models using generalized additive models, which allow for non‐linear responses, and dynamic linear models, which allow for time‐varying responses. Only two environmental covariates improved out‐of‐sample prediction: the 2.5‐year average regional SST and precipitation over land during June–July. The most significant improvement was with the SST covariate and post‐monsoon landings with a 19%–22% reduction in mean‐squared prediction error. Models with the second best covariate, monsoon precipitation over land, provided a 4%–8% reduction in prediction error. We also tested large‐scale ocean climate teleconnection indices. One, an index of the Atlantic Multidecadal Oscillation, also improved out‐of‐sample predictions similarly to the multiyear average regional SST. The earth's changing climate is associated with both rapid warming in the Western Indian Ocean and changes to monsoon rainfall patterns. Our work highlights these as key variables that can improve forecasting of oil sardine landings.