What This Analysis Does

This project identifies which brownfield sites in Greater Manchester pose the highest environmental risk and are best suited for nature-based restoration interventions like mycoforestry (using fungi to remediate contaminated soil).

Using satellite imagery, soil data, and terrain analysis, the model scores 1,585 registered brownfield sites based on their likelihood of spreading contamination to watercourses and groundwater.

Findings

Geographic Concentration

The highest-risk sites cluster in Salford M5, near the River Irwell. Three of the top 10 priority sites are located here, reflecting:

• Flat terrain (typical of former industrial warehouses and rail yards)
• Proximity to the River Irwell (contamination spread risk)
• Sandy soil (higher groundwater pollution potential)

What Makes a Site High-Risk?

The analysis combines three environmental factors:

1. Proximity to watercourses: Sites within 1-2km of rivers/streams can spread contamination through runoff
2. Soil permeability: Sandy soils allow pollutants to reach groundwater more easily than clay
3. Terrain flatness: Flat sites are more likely former industrial land (higher baseline contamination)

Machine Learning Insights

A predictive model was trained to identify which site characteristics best predict restoration suitability. The model found:

Terrain flatness contributes 13% — flat sites are more likely former industrial land, making them both riskier and more suitable for intervention.

Water and soil factors have minimal additional predictive power once size and terrain are accounted for.

How This Can Be Used

For Local Authorities

• Prioritise site inspections: Focus ecological assessments on the 746 high-risk sites rather than all 1,585
• Target funding applications: Use risk scores to justify restoration grants (e.g., Defra Environmental Land Management)
• Inform spatial planning: Consider contamination risk when approving developments on brownfield land

For Environmental NGOs

• Identify restoration opportunities: The Salford M5 cluster offers potential for grouped interventions (economies of scale)
• Pilot project selection: Choose mid-sized, high-risk sites for demonstration mycoforestry trials
• Community engagement: Use the interactive map to show residents which local sites pose risks

For Restoration Practitioners

• Site selection: Filter the dataset by size, risk score, and location to find suitable intervention sites
• Species matching: Flat, sandy sites near water may require different fungal species than hilly, clay sites
• Success tracking: Record outcomes from restored sites to improve future models

Important Limitations

The machine learning model uses synthetic data (rule-based assumptions about restoration suitability) rather than real restoration outcomes. With access to historical records tracking which sites were successfully restored, the model could be retrained to predict actual success rates.

Economic feasibility is not modelled. A high-risk site might be impractical to restore due to land ownership issues, access constraints, or prohibitive remediation costs.

Next Steps

To extend this work into operational decision-making tools, the following would be valuable:

• Soil contamination records: Historical testing data from councils to validate risk scores
• Restoration outcome tracking: Database of which sites were restored, which methods worked, and ecological recovery metrics
• Cost modelling: Integration of land acquisition costs, remediation expenses, and expected ecological benefits
• Stakeholder workshops: Input from planners, ecologists, and community groups to refine prioritisation criteria

Access the Data

The full dataset, interactive map, and analysis code are available:

• Interactive map: Explore all 1,585 sites with clickable risk scores
• GitHub repository: Full methodology and code
• Project details: Technical write-up with visualisations