NEMAP: Predicting Marine Microplastic Density via Spatiotemporal Deep Learning
Overview
Microplastic toxicity in seafood is an emerging global concern, particularly in regions where both economies and diets heavily rely on aquaculture. To tackle this challenge, microplastic accumulation in the Northeastern United States was examined. We performed spatiotemporal modelling of marine microplastic concentration based on the data containing the following key parameters: sea surface temperature, pH, alkalinity, wind speed, salinity, phosphate/nitrate concentrations, coastal proximity, among others. After constructing a dataset through date-location matching, categorical prediction was implemented using classical machine learning algorithms (e.g., random forest, XGBoost) and deep learning architectures (LSTM and transformer). To address data imbalance, a novel contrastive learning-enhanced transformer framework was proposed. A comparative model analysis demonstrated that the transformer framework noticeably outperforms other models in prediction accuracy, validating its superiority in handling microplastic distribution heterogeneity. To better visualize outcomes, NEMAP, an interactive web application, was developed, integrating species distribution with transformer-predicted microplastic hotspots and quantifying contamination risk levels. This tool not only provides decision-making support to optimize seafood supply chains but also guides consumers with risk maps to avoid contaminated species, demonstrating significant potential in food safety and public health governance.
Competition history
- AJAS 2026
Related projects
ISEF · 2021
Developing a 3D Trajectory Modeling System to Predict Ocean Floor Microplastic Aggregation Using a Voxel-Based Neural Network Approach
ISEF · 2026
Autonomous Underwater Vehicle (AUV) for Real-Time 3D Microplastic Concentration and Toxicity Mapping via Deep Learning and Integrated Microscopy
ISEF · 2025
Global Microplastic Pollution-Year 5: A Novel Multi-Organism & Computational Approach for Investigating Microplastics' Size and Concentration-Dependent Effects for Translational Human Biomarker Discovery; and Deployment of Plant-Based, Economical, and Scalable Microplastic Remediation Systems
CSEF · 2026
No Nurdles: Novel Model for Predicting and Preventing Plastic Pellet/Nurdle Spills
ISEF · 2024
Global Microplastic Pollution - The Emerging Health Crisis: The Ecotoxicological Effects of Microplastics on Aquatic Organisms and Translating to Humans via A.I. Machine Learning & a Novel Low-Cost Water Filtration System
CWSF · 2026
In Situ Detection of Aquatic Microplastics using Laser-Based Holographic Imaging and Deep Learning
ISEF · 2026
In Situ Microplastic Detection Using Holographic Imaging and AI on an Autonomous Bionic Sea Turtle
AJAS · 2025
Predicting Microplastics Exposure via Neural Network Gene Expression Integration
Closest projects by meaning, across every fair and year in the corpus.
Browse more like this
Source: AAAS Annual Meeting (Confex) / American Junior Academy of Science