World Aquaculture Singapore 2026

June 2-5, 2026

Singapore

AI-DRIVEN PRICE VOLATILITY MANAGEMENT IN AQUACULTURE: A RISK-AWARE FORECASTING AND HEDGING FRAMEWORK

Manuel Luna *, Olaya Pérez-Mon , Pelayo González, Pablo Rodríguez

Polytechnic School of Engineering

University of Oviedo

Asturias, Spain

lunamanuel@uniovi.es



Price volatility has become a central source of risk for  producers, directly affecting revenues, planning, and financial stability. However, as the technological transformation of production systems evolves, forecasting and quantifying risk remain challenging. In this context, this study propose a hybrid framework for forecasting and adaptively managing price volatility, tailored to the operational context of salmon production. This addresses both technical challenges , since effective machine learning methods often lack actionable measures of uncertainty ,  and practical ones, as the application of AI in aquaculture remains underexplored.

The framework consists of two sequential and interconnected phases:

(1) Short-term salmon prices are forecast using an XGBoost model enhanced with Conformal Prediction to obtain prediction intervals with guaranteed finite-sample coverage. Forecasts are generated in a rolling manner and retrained monthly, so that newly observed market information progressively improves accuracy and yields tighter uncertainty bounds as the selling date approaches (Fig 1).

(2) The resulting prediction intervals are embedded within aquaculture growth simulations to estimate the expected production value, together with its minimum and maximum bounds , and integrated into an adaptive hedging strategy based on Dynamic Portfolio Insurance.  These intervals are used to determine the value to be protected and to dynamically adjust hedge ratios over time using salmon futures contracts. This design  aims to mitigate downside risk while preserving upside potential, adapting continuously as new market information becomes available throughout the production cycle.

The proposed framework was empirically evaluated using real spot and futures data from the Atlantic salmon market. Results show that the conformal forecasts generate increasingly accurate inputs for risk-aware decision-making and that the adaptive hedging strategy stabilizes revenues under adverse price movements without imposing excessive costs in favorable conditions (Fig 2). This makes short-term price forecasts actionable and supports volatility-informed adaptive hedging, contributing to more resilient production strategies.