Abstract
Environmental change and fishing activity can produce directional trends in exploited fish populations with consequences for stock productivity. For herring in the Celtic Sea, size at age has been in steady decline since the mid 1980's. In the neighbouring herring stock off the Northwest coast of Ireland, reductions in size at age are noted after 1990. Here, gradient boosting regression trees were used to investigate trends in extended time series (1959–2012) of length-at-age across both populations and to identify important variables associated with the observed declines in size. The predominant signal detected was a non-linear negative relationship between adult size and mean Sea Surface Temperature during the first growing season. Herring length was negatively correlated with the Atlantic Multidecadal Oscillation. Weaker associations with indicators of food availability and population size were also detected. Across both populations a marked decline in length was observed at the upper end of the temperature range (~14∘C in the Celtic Sea and ~13∘C in the Northwest). Declines in length and associations with temperature were more pronounced in the Celtic Sea population which may be vulnerable to increasing sea temperatures due to its position at the southern limit of the species distribution.
Original language | English |
---|---|
Article number | 101154 |
Journal | Ecological Informatics |
Volume | 60 |
DOIs | |
Publication status | Published - Nov 2020 |
Keywords
- Clupea harengus
- Gradient boosting regression trees
- Growth
- Herring
- Interactions
- Multiple drivers
- Supervised machine learning
- Time-series