Quantifying linked rare events in fish and environmental Chesapeake Bay time series
Conference
Format: IPS Abstract
Keywords: chesapeake, climate change, extreme_event, fisheries, machine learning, rare
Session: Invited Session 10A - Size Estimation and Impact Assessment in Biological Populations
Thursday 5 December 1:30 p.m. - 3 p.m. (Australia/Adelaide)
Abstract
The nexus of climate change and its impact on rare weather events is increasingly recognized as a threat to ecological stability, including fish productivity and abundance. However, traditional assessments often rely on pre-defined biological anomaly thresholds, which may become obsolete with shifting baselines and animal adaptation. This study presents a data-driven approach to objectively identify linked rare events by leveraging a combination of statistical and machine learning techniques.
We analyzed long-term environmental and catch-per-unit-effort (CPUE) time series for the most important native and one invasive fish/shellfish species in the Chesapeake Bay. Rare events within environmental conditions and CPUE were identified using two complementary methods: (1) data-driven statistical identification of extremes based on outlier detection techniques, and (2) isolation random forest, a purely machine learning-based approach for anomaly detection. Subsequently, a correspondence analysis incorporating time series lags was employed to elucidate significant linkages between biological (CPUE) and environmental rare events.
This investigation unveils the potential of statistical and machine learning frameworks to objectively identify and link rare events in complex environmental systems. The findings offer valuable insights into climate-driven disruptions and their potential influence on Chesapeake Bay fish/shellfish populations. This approach can be readily adapted to other ecological systems facing similar challenges.