Physics-informed multi-task learning framework for modelling microplastic impacts on sea turtle nest temperature and...

Area of Science:

• Environmental Science
• Marine Biology
• Computational Ecology

Background:

• Coastal nesting beaches face microplastic (MP) pollution, potentially disrupting sea turtle nest thermal environments.
• Temperature-dependent sex determination (TSD) in sea turtles means nest temperature directly influences hatchling sex ratios.
• Understanding MP's impact on nest temperature and sex ratios is crucial for green turtle (Chelonia mydas) conservation.

Purpose of the Study:

• To investigate how microplastic abundance and characteristics influence green turtle nest temperature and sex ratio.
• To develop and validate a predictive model for microplastic-driven thermal and sex ratio shifts in nests.
• To identify key microplastic attributes affecting nest thermal dynamics.

Main Methods:

• Development of a physics-informed multi-task learning (PI-MTL) model to predict nest temperature and sex-ratio class.
• Detailed characterization of microplastics (color, shape, size) and their depth distribution.
• Incorporation of physics-informed consistency loss based on pivotal temperature and thermal bandwidths.
• Comparison of PI-MTL with single-task learning, kNN, and SVM models.
• Application of SHapley Additive exPlanations (SHAP) for feature importance analysis.

Main Results:

• The PI-MTL model achieved the lowest total absolute temperature error (4.037 °C) and 100% sex-ratio classification accuracy.
• PI-MTL provided more consistent and biologically plausible predictions compared to other machine learning methods.
• SHAP analysis identified black and blue microplastics as having the strongest influence on nest temperature.
• The model demonstrated robustness in predicting MP-driven shifts in nest temperature and sex ratio.

Conclusions:

• Integrating microplastic characterization with a PI-MTL framework offers a powerful approach for predicting environmental changes in sea turtle nests.
• The study highlights the significant impact of microplastic attributes, particularly color, on nest thermal profiles and hatchling sex ratios.
• This methodology provides a biologically grounded tool for informing conservation strategies in microplastic-affected nesting sites, especially in data-limited scenarios.