Research Framework

Innovative Approach to
Aquaculture Monitoring

Comprehensive framework optimizing monitoring strategies, predictive modeling, and stakeholder-driven decision-making

Key Objectives & Methodology

Six integrated work packages combining sensor optimization, fish epidemiology, AI-driven predictive models, stakeholder co-creation, and knowledge transfer

01

Optimize Monitoring (WP1)

Will identify cost-effective sensor configurations through lab-scale benchmarking and comparative pen trials. Utilize Ensemble Kalman Filtering (EnKF) to enhance sensor robustness and cross-correct readings. Validate streamlined sensor combinations that eliminate redundancy while maintaining disease detection reliability across different aquaculture environments.

02

Epidemiology & Predictive AI (WP2 & WP3)

Will integrate historical disease records with real-time sensor data to develop epidemiological models linking environmental and biological factors to disease outbreaks. Create AI-driven predictive models using mixup augmentation and generative modeling to handle sparse data, achieving ≥80% accuracy in early warning systems. Apply privacy-preserving federated learning for secure multi-farm data sharing.

03

Stakeholder Co-Creation (WP4)

Will conduct structured field studies across 6-10 Philippine aquaculture sites to analyze how farmers interpret predictive model outputs and integrate them into decision-making. Organize co-creation workshops with farmers, industry experts, and policymakers to refine user interfaces, ensure practical applicability, and improve adoption rates through iterative feedback loops.

04

Knowledge Transfer (WP5)

Will conduct data-driven cross-validation using Norwegian aquaculture datasets from Manolin AS (accessing 1/3 of Norwegian data). Down-sample high-resolution data to simulate minimal-sensor setups, assessing whether AI-driven models can compensate for fewer sensors. Develop best-practice guidelines for adapting cost-effective Philippine solutions to Norway's high-precision aquaculture and vice versa.

05

Privacy-Preserving ML

Will develop federated learning and differential privacy frameworks for secure multi-farm data consolidation without compromising business confidentiality. Enable collaborative learning where farms contribute to improved predictive models while protecting proprietary operational data and competitive information.

06

Capacity Building (WP0)

Will train 3 PhD students (sensor optimization, fish epidemiology, AI-driven disease prediction) and 2 MSc students in aquaculture technology and data analytics. Publish 8+ peer-reviewed articles, organize 3 annual stakeholder workshops (Norway, Philippines, Indonesia), conduct industry seminars, develop policy briefs, and engage regulators to maximize project impact and ensure sustainable knowledge transfer.

How We Work

Combining lab-scale testing, field trials, AI-driven modeling, and cross-validation to develop practical, cost-effective solutions

Lab & Pen-Based Testing

Benchmark sensor configurations under controlled conditions before field deployment. Test sensitivity and stability across varying salinity, pH, temperature, and dissolved oxygen levels to validate minimal-sensor setups against high-end alternatives.

Philippine Field Trials

Two seasonal field trials (3-4 months each) across 6-10 sites in collaboration with BFAR, HOC PO Feed Corp, and local aquaculture operators. Real-world validation of low-cost monitoring solutions and stakeholder co-creation to ensure practical applicability.

Norwegian Back-Testing

Data-driven cross-validation using Manolin AS datasets (1/3 of Norwegian aquaculture data). Down-sample high-resolution data to simulate minimal-sensor setups, assessing whether AI-driven models can compensate for fewer sensors while maintaining prediction accuracy.

AI for Sparse Data

Mixup augmentation generates synthetic sensor readings to strengthen machine learning accuracy in data-sparse environments. Generative models simulate environmental stressors and disease outbreak scenarios to refine early-warning thresholds and improve robustness.

Privacy-Preserving Analytics

Federated learning and differential privacy techniques enable multi-farm data sharing for collaborative model training without exposing individual farm data. Protects business confidentiality while improving collective disease prediction capabilities.

Co-Creation with Farmers

Stakeholder workshops, focus groups, and iterative feedback sessions with farmers, industry experts, and policymakers ensure solutions align with real-world operations. User-centered design improves adoption rates and practical effectiveness.

Paradigm Shift in Aquaculture Technology

AQUAROM challenges the assumption that higher spending automatically leads to better outcomes

Challenges Status Quo

Will prove cost-effective solutions can rival expensive systems through optimized sensor configurations and intelligent data analysis.

Two-Way Knowledge Transfer

Philippines serves as testbed for innovation; Norway validates and adopts successful approaches in high-tech operations.

Industry-Driven Solutions

Solutions co-created with farmers and companies ensuring practical applicability and real-world effectiveness.

Globally Scalable

Benefits both high-tech and resource-limited aquaculture operations worldwide through adaptable frameworks.

Policy-Relevant

Will engage regulators in Norway and Philippines to ensure alignment with governance frameworks and industry standards.

Open Science

Committed to open-access data and algorithms following FAIR principles for maximum research impact.

Research Outputs & Impact

Technical Outputs

  • Cost-effective sensor configurations and optimization guidelines
  • Predictive disease models with ≥80% accuracy using limited data
  • Privacy-preserving ML framework for multi-farm data sharing
  • Epidemiological models linking environmental factors to disease outbreaks
  • Cross-validation study comparing low-cost vs. high-tech monitoring
  • Open-access datasets and algorithms (FAIR principles)

Capacity Building & Dissemination

  • 2 PhD students (Philippines/Indonesia)
  • 2 MSc students
  • 8+ peer-reviewed publications
  • 3 annual stakeholder workshops (Norway, Philippines, Indonesia)
  • Industry seminars and training programs
  • Policy briefs for regulators and decision-makers

Project Timeline (2026-2029)

2026
2027
2028
2029
Q1
Q2
Q3
Q4
Q1
Q2
Q3
Q4
Q1
Q2
Q3
Q4
Q1
Q2
Q3
Q4
WP0: Capacity Building
Continuous throughout project
WP1: Sensor Optimization
Lab & field testing
WP2: Fish Epidemiology
Disease modeling & analysis
WP3: AI Predictive Models
ML development & validation
WP4: Co-Creation
Workshops & field studies
WP5: Knowledge Transfer
Norwegian validation & dissemination

Key Milestones

2026 Q1: Project Kickoff

Launch, PhD recruitment, organizational establishment, and initial lab testing

2027 Q2: First Prototypes

Initial predictive models deployed, AI prototypes ready for testing

2028 Q2: First Field Trial

Philippine field deployment across 6-10 sites (3-4 months)

2029 Q2: Second Field Trial

Validation, iterative improvements, Norwegian cross-validation begins

2029 Q3: Cross-Validation

Norwegian back-testing completed, comparative analysis finalized

2029 Q4: Completion

Final reports, guidelines, policy briefs, and 8+ publications